1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
2 @c 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009
3 @c Free Software Foundation, Inc.
4 @c This is part of the GCC manual.
5 @c For copying conditions, see the file gcc.texi.
12 @c man begin COPYRIGHT
13 Copyright @copyright{} 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
14 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009
15 Free Software Foundation, Inc.
17 Permission is granted to copy, distribute and/or modify this document
18 under the terms of the GNU Free Documentation License, Version 1.2 or
19 any later version published by the Free Software Foundation; with the
20 Invariant Sections being ``GNU General Public License'' and ``Funding
21 Free Software'', the Front-Cover texts being (a) (see below), and with
22 the Back-Cover Texts being (b) (see below). A copy of the license is
23 included in the gfdl(7) man page.
25 (a) The FSF's Front-Cover Text is:
29 (b) The FSF's Back-Cover Text is:
31 You have freedom to copy and modify this GNU Manual, like GNU
32 software. Copies published by the Free Software Foundation raise
33 funds for GNU development.
35 @c Set file name and title for the man page.
37 @settitle GNU project C and C++ compiler
39 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
40 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
41 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
42 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
43 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
44 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
45 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
47 Only the most useful options are listed here; see below for the
48 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
51 gpl(7), gfdl(7), fsf-funding(7),
52 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
53 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
54 @file{ld}, @file{binutils} and @file{gdb}.
57 For instructions on reporting bugs, see
61 See the Info entry for @command{gcc}, or
62 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
63 for contributors to GCC@.
68 @chapter GCC Command Options
69 @cindex GCC command options
70 @cindex command options
71 @cindex options, GCC command
73 @c man begin DESCRIPTION
74 When you invoke GCC, it normally does preprocessing, compilation,
75 assembly and linking. The ``overall options'' allow you to stop this
76 process at an intermediate stage. For example, the @option{-c} option
77 says not to run the linker. Then the output consists of object files
78 output by the assembler.
80 Other options are passed on to one stage of processing. Some options
81 control the preprocessor and others the compiler itself. Yet other
82 options control the assembler and linker; most of these are not
83 documented here, since you rarely need to use any of them.
85 @cindex C compilation options
86 Most of the command line options that you can use with GCC are useful
87 for C programs; when an option is only useful with another language
88 (usually C++), the explanation says so explicitly. If the description
89 for a particular option does not mention a source language, you can use
90 that option with all supported languages.
92 @cindex C++ compilation options
93 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
94 options for compiling C++ programs.
96 @cindex grouping options
97 @cindex options, grouping
98 The @command{gcc} program accepts options and file names as operands. Many
99 options have multi-letter names; therefore multiple single-letter options
100 may @emph{not} be grouped: @option{-dv} is very different from @w{@samp{-d
103 @cindex order of options
104 @cindex options, order
105 You can mix options and other arguments. For the most part, the order
106 you use doesn't matter. Order does matter when you use several
107 options of the same kind; for example, if you specify @option{-L} more
108 than once, the directories are searched in the order specified. Also,
109 the placement of the @option{-l} option is significant.
111 Many options have long names starting with @samp{-f} or with
112 @samp{-W}---for example,
113 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
114 these have both positive and negative forms; the negative form of
115 @option{-ffoo} would be @option{-fno-foo}. This manual documents
116 only one of these two forms, whichever one is not the default.
120 @xref{Option Index}, for an index to GCC's options.
123 * Option Summary:: Brief list of all options, without explanations.
124 * Overall Options:: Controlling the kind of output:
125 an executable, object files, assembler files,
126 or preprocessed source.
127 * Invoking G++:: Compiling C++ programs.
128 * C Dialect Options:: Controlling the variant of C language compiled.
129 * C++ Dialect Options:: Variations on C++.
130 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
132 * Language Independent Options:: Controlling how diagnostics should be
134 * Warning Options:: How picky should the compiler be?
135 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
136 * Optimize Options:: How much optimization?
137 * Preprocessor Options:: Controlling header files and macro definitions.
138 Also, getting dependency information for Make.
139 * Assembler Options:: Passing options to the assembler.
140 * Link Options:: Specifying libraries and so on.
141 * Directory Options:: Where to find header files and libraries.
142 Where to find the compiler executable files.
143 * Spec Files:: How to pass switches to sub-processes.
144 * Target Options:: Running a cross-compiler, or an old version of GCC.
145 * Submodel Options:: Specifying minor hardware or convention variations,
146 such as 68010 vs 68020.
147 * Code Gen Options:: Specifying conventions for function calls, data layout
149 * Environment Variables:: Env vars that affect GCC.
150 * Precompiled Headers:: Compiling a header once, and using it many times.
156 @section Option Summary
158 Here is a summary of all the options, grouped by type. Explanations are
159 in the following sections.
162 @item Overall Options
163 @xref{Overall Options,,Options Controlling the Kind of Output}.
164 @gccoptlist{-c -S -E -o @var{file} -combine -no-canonical-prefixes @gol
165 -pipe -pass-exit-codes @gol
166 -x @var{language} -v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help @gol
167 --version -wrapper@@@var{file} -fplugin=@var{file} -fplugin-arg-@var{name}=@var{arg}}
169 @item C Language Options
170 @xref{C Dialect Options,,Options Controlling C Dialect}.
171 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
172 -aux-info @var{filename} @gol
173 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
174 -fhosted -ffreestanding -fopenmp -fms-extensions @gol
175 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
176 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
177 -fsigned-bitfields -fsigned-char @gol
178 -funsigned-bitfields -funsigned-char}
180 @item C++ Language Options
181 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
182 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
183 -fconserve-space -ffriend-injection @gol
184 -fno-elide-constructors @gol
185 -fno-enforce-eh-specs @gol
186 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
187 -fno-implicit-templates @gol
188 -fno-implicit-inline-templates @gol
189 -fno-implement-inlines -fms-extensions @gol
190 -fno-nonansi-builtins -fno-operator-names @gol
191 -fno-optional-diags -fpermissive @gol
192 -fno-pretty-templates @gol
193 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
194 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
195 -fno-default-inline -fvisibility-inlines-hidden @gol
196 -fvisibility-ms-compat @gol
197 -Wabi -Wctor-dtor-privacy @gol
198 -Wnon-virtual-dtor -Wreorder @gol
199 -Weffc++ -Wstrict-null-sentinel @gol
200 -Wno-non-template-friend -Wold-style-cast @gol
201 -Woverloaded-virtual -Wno-pmf-conversions @gol
204 @item Objective-C and Objective-C++ Language Options
205 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
206 Objective-C and Objective-C++ Dialects}.
207 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
208 -fgnu-runtime -fnext-runtime @gol
209 -fno-nil-receivers @gol
210 -fobjc-call-cxx-cdtors @gol
211 -fobjc-direct-dispatch @gol
212 -fobjc-exceptions @gol
214 -freplace-objc-classes @gol
217 -Wassign-intercept @gol
218 -Wno-protocol -Wselector @gol
219 -Wstrict-selector-match @gol
220 -Wundeclared-selector}
222 @item Language Independent Options
223 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
224 @gccoptlist{-fmessage-length=@var{n} @gol
225 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
226 -fdiagnostics-show-option}
228 @item Warning Options
229 @xref{Warning Options,,Options to Request or Suppress Warnings}.
230 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
231 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
232 -Wno-attributes -Wno-builtin-macro-redefined @gol
233 -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual @gol
234 -Wchar-subscripts -Wclobbered -Wcomment @gol
235 -Wconversion -Wcoverage-mismatch -Wno-deprecated @gol
236 -Wno-deprecated-declarations -Wdisabled-optimization @gol
237 -Wno-div-by-zero -Wempty-body -Wenum-compare -Wno-endif-labels @gol
238 -Werror -Werror=* @gol
239 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
240 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
241 -Wformat-security -Wformat-y2k @gol
242 -Wframe-larger-than=@var{len} -Wjump-misses-init -Wignored-qualifiers @gol
243 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
244 -Winit-self -Winline @gol
245 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
246 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
247 -Wlogical-op -Wlong-long @gol
248 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
249 -Wmissing-format-attribute -Wmissing-include-dirs @gol
250 -Wmissing-noreturn -Wno-mudflap @gol
251 -Wno-multichar -Wnonnull -Wno-overflow @gol
252 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
253 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
254 -Wpointer-arith -Wno-pointer-to-int-cast @gol
255 -Wredundant-decls @gol
256 -Wreturn-type -Wsequence-point -Wshadow @gol
257 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
258 -Wstrict-aliasing -Wstrict-aliasing=n @gol
259 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
260 -Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand @gol
261 -Wsystem-headers -Wtrigraphs -Wtype-limits -Wundef -Wuninitialized @gol
262 -Wunknown-pragmas -Wno-pragmas -Wunreachable-code @gol
263 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
264 -Wunused-label -Wunused-parameter -Wno-unused-result -Wunused-value -Wunused-variable @gol
265 -Wvariadic-macros -Wvla @gol
266 -Wvolatile-register-var -Wwrite-strings}
268 @item C and Objective-C-only Warning Options
269 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
270 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
271 -Wold-style-declaration -Wold-style-definition @gol
272 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
273 -Wdeclaration-after-statement -Wpointer-sign}
275 @item Debugging Options
276 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
277 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
278 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
279 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
280 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
281 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
282 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
283 -fdump-statistics @gol
285 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
286 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
287 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
289 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
290 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
291 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
292 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
293 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
294 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
295 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
296 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
297 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
298 -fdump-tree-nrv -fdump-tree-vect @gol
299 -fdump-tree-sink @gol
300 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
301 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
302 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
303 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
304 -ftree-vectorizer-verbose=@var{n} @gol
305 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
306 -fdump-final-insns=@var{file} @gol
307 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
308 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
309 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
310 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
311 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
312 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
313 -ftest-coverage -ftime-report -fvar-tracking @gol
314 -fvar-tracking-assigments -fvar-tracking-assignments-toggle @gol
315 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
316 -ggdb -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
317 -gvms -gxcoff -gxcoff+ @gol
318 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
319 -fdebug-prefix-map=@var{old}=@var{new} @gol
320 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
321 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
322 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
323 -print-multi-directory -print-multi-lib @gol
324 -print-prog-name=@var{program} -print-search-dirs -Q @gol
325 -print-sysroot -print-sysroot-headers-suffix @gol
326 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
328 @item Optimization Options
329 @xref{Optimize Options,,Options that Control Optimization}.
331 -falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
332 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
333 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
334 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
335 -fcheck-data-deps -fconserve-stack -fcprop-registers -fcrossjumping @gol
336 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules -fcx-limited-range @gol
337 -fdata-sections -fdce -fdce @gol
338 -fdelayed-branch -fdelete-null-pointer-checks -fdse -fdse @gol
339 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffast-math @gol
340 -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
341 -fforward-propagate -ffunction-sections @gol
342 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm @gol
343 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
344 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
345 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg -fipa-pta @gol
346 -fipa-pure-const -fipa-reference -fipa-struct-reorg @gol
347 -fipa-type-escape -fira-algorithm=@var{algorithm} @gol
348 -fira-region=@var{region} -fira-coalesce -fno-ira-share-save-slots @gol
349 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
350 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
351 -floop-block -floop-interchange -floop-strip-mine -fgraphite-identity @gol
352 -floop-parallelize-all -flto -flto-compression-level -flto-report -fltrans @gol
353 -fltrans-output-list -fmerge-all-constants -fmerge-constants -fmodulo-sched @gol
354 -fmodulo-sched-allow-regmoves -fmove-loop-invariants -fmudflap @gol
355 -fmudflapir -fmudflapth -fno-branch-count-reg -fno-default-inline @gol
356 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
357 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
358 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
359 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
360 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
361 -fpeel-loops -fpredictive-commoning -fprefetch-loop-arrays @gol
362 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
363 -fprofile-generate=@var{path} @gol
364 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
365 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
366 -freorder-blocks-and-partition -freorder-functions @gol
367 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
368 -frounding-math -fsched2-use-superblocks @gol
369 -fsched2-use-traces -fsched-pressure @gol
370 -fsched-spec-load -fsched-spec-load-dangerous @gol
371 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
372 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
373 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
374 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
375 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
376 -fselective-scheduling -fselective-scheduling2 @gol
377 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
378 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
379 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
380 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer @gol
381 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
382 -ftree-copyrename -ftree-dce @gol
383 -ftree-dominator-opts -ftree-dse -ftree-forwprop -ftree-fre -ftree-loop-im @gol
384 -ftree-phiprop -ftree-loop-distribution @gol
385 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
386 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-pta -ftree-reassoc @gol
387 -ftree-sink -ftree-sra -ftree-switch-conversion @gol
388 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
389 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
390 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
391 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
392 -fwhole-program -fwhopr -fwpa -use-linker-plugin @gol
393 --param @var{name}=@var{value}
394 -O -O0 -O1 -O2 -O3 -Os}
396 @item Preprocessor Options
397 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
398 @gccoptlist{-A@var{question}=@var{answer} @gol
399 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
400 -C -dD -dI -dM -dN @gol
401 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
402 -idirafter @var{dir} @gol
403 -include @var{file} -imacros @var{file} @gol
404 -iprefix @var{file} -iwithprefix @var{dir} @gol
405 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
406 -imultilib @var{dir} -isysroot @var{dir} @gol
407 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
408 -P -fworking-directory -remap @gol
409 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
410 -Xpreprocessor @var{option}}
412 @item Assembler Option
413 @xref{Assembler Options,,Passing Options to the Assembler}.
414 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
417 @xref{Link Options,,Options for Linking}.
418 @gccoptlist{@var{object-file-name} -l@var{library} @gol
419 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
420 -s -static -static-libgcc -static-libstdc++ -shared @gol
421 -shared-libgcc -symbolic @gol
422 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
425 @item Directory Options
426 @xref{Directory Options,,Options for Directory Search}.
427 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir}
428 -specs=@var{file} -I- --sysroot=@var{dir}}
431 @c I wrote this xref this way to avoid overfull hbox. -- rms
432 @xref{Target Options}.
433 @gccoptlist{-V @var{version} -b @var{machine}}
435 @item Machine Dependent Options
436 @xref{Submodel Options,,Hardware Models and Configurations}.
437 @c This list is ordered alphanumerically by subsection name.
438 @c Try and put the significant identifier (CPU or system) first,
439 @c so users have a clue at guessing where the ones they want will be.
442 @gccoptlist{-EB -EL @gol
443 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
444 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
447 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
448 -mabi=@var{name} @gol
449 -mapcs-stack-check -mno-apcs-stack-check @gol
450 -mapcs-float -mno-apcs-float @gol
451 -mapcs-reentrant -mno-apcs-reentrant @gol
452 -msched-prolog -mno-sched-prolog @gol
453 -mlittle-endian -mbig-endian -mwords-little-endian @gol
454 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
455 -mfp16-format=@var{name}
456 -mthumb-interwork -mno-thumb-interwork @gol
457 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
458 -mstructure-size-boundary=@var{n} @gol
459 -mabort-on-noreturn @gol
460 -mlong-calls -mno-long-calls @gol
461 -msingle-pic-base -mno-single-pic-base @gol
462 -mpic-register=@var{reg} @gol
463 -mnop-fun-dllimport @gol
464 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
465 -mpoke-function-name @gol
467 -mtpcs-frame -mtpcs-leaf-frame @gol
468 -mcaller-super-interworking -mcallee-super-interworking @gol
470 -mword-relocations @gol
471 -mfix-cortex-m3-ldrd}
474 @gccoptlist{-mmcu=@var{mcu} -msize -mno-interrupts @gol
475 -mcall-prologues -mtiny-stack -mint8}
477 @emph{Blackfin Options}
478 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
479 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
480 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
481 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
482 -mno-id-shared-library -mshared-library-id=@var{n} @gol
483 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
484 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
485 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
489 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
490 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
491 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
492 -mstack-align -mdata-align -mconst-align @gol
493 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
494 -melf -maout -melinux -mlinux -sim -sim2 @gol
495 -mmul-bug-workaround -mno-mul-bug-workaround}
498 @gccoptlist{-mmac -mpush-args}
500 @emph{Darwin Options}
501 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
502 -arch_only -bind_at_load -bundle -bundle_loader @gol
503 -client_name -compatibility_version -current_version @gol
505 -dependency-file -dylib_file -dylinker_install_name @gol
506 -dynamic -dynamiclib -exported_symbols_list @gol
507 -filelist -flat_namespace -force_cpusubtype_ALL @gol
508 -force_flat_namespace -headerpad_max_install_names @gol
510 -image_base -init -install_name -keep_private_externs @gol
511 -multi_module -multiply_defined -multiply_defined_unused @gol
512 -noall_load -no_dead_strip_inits_and_terms @gol
513 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
514 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
515 -private_bundle -read_only_relocs -sectalign @gol
516 -sectobjectsymbols -whyload -seg1addr @gol
517 -sectcreate -sectobjectsymbols -sectorder @gol
518 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
519 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
520 -segprot -segs_read_only_addr -segs_read_write_addr @gol
521 -single_module -static -sub_library -sub_umbrella @gol
522 -twolevel_namespace -umbrella -undefined @gol
523 -unexported_symbols_list -weak_reference_mismatches @gol
524 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
525 -mkernel -mone-byte-bool}
527 @emph{DEC Alpha Options}
528 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
529 -mieee -mieee-with-inexact -mieee-conformant @gol
530 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
531 -mtrap-precision=@var{mode} -mbuild-constants @gol
532 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
533 -mbwx -mmax -mfix -mcix @gol
534 -mfloat-vax -mfloat-ieee @gol
535 -mexplicit-relocs -msmall-data -mlarge-data @gol
536 -msmall-text -mlarge-text @gol
537 -mmemory-latency=@var{time}}
539 @emph{DEC Alpha/VMS Options}
540 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
543 @gccoptlist{-msmall-model -mno-lsim}
546 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
547 -mhard-float -msoft-float @gol
548 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
549 -mdouble -mno-double @gol
550 -mmedia -mno-media -mmuladd -mno-muladd @gol
551 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
552 -mlinked-fp -mlong-calls -malign-labels @gol
553 -mlibrary-pic -macc-4 -macc-8 @gol
554 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
555 -moptimize-membar -mno-optimize-membar @gol
556 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
557 -mvliw-branch -mno-vliw-branch @gol
558 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
559 -mno-nested-cond-exec -mtomcat-stats @gol
563 @emph{GNU/Linux Options}
564 @gccoptlist{-muclibc}
566 @emph{H8/300 Options}
567 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
570 @gccoptlist{-march=@var{architecture-type} @gol
571 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
572 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
573 -mfixed-range=@var{register-range} @gol
574 -mjump-in-delay -mlinker-opt -mlong-calls @gol
575 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
576 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
577 -mno-jump-in-delay -mno-long-load-store @gol
578 -mno-portable-runtime -mno-soft-float @gol
579 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
580 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
581 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
582 -munix=@var{unix-std} -nolibdld -static -threads}
584 @emph{i386 and x86-64 Options}
585 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
586 -mfpmath=@var{unit} @gol
587 -masm=@var{dialect} -mno-fancy-math-387 @gol
588 -mno-fp-ret-in-387 -msoft-float @gol
589 -mno-wide-multiply -mrtd -malign-double @gol
590 -mpreferred-stack-boundary=@var{num}
591 -mincoming-stack-boundary=@var{num}
592 -mcld -mcx16 -msahf -mmovbe -mcrc32 -mrecip @gol
593 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
595 -msse4a -m3dnow -mpopcnt -mabm -mfma4 @gol
596 -mthreads -mno-align-stringops -minline-all-stringops @gol
597 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
598 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
599 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
600 -mveclibabi=@var{type} -mpc32 -mpc64 -mpc80 -mstackrealign @gol
601 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
602 -mcmodel=@var{code-model} -mabi=@var{name} @gol
603 -m32 -m64 -mlarge-data-threshold=@var{num} @gol
607 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
608 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
609 -mconstant-gp -mauto-pic -mfused-madd @gol
610 -minline-float-divide-min-latency @gol
611 -minline-float-divide-max-throughput @gol
612 -mno-inline-float-divide @gol
613 -minline-int-divide-min-latency @gol
614 -minline-int-divide-max-throughput @gol
615 -mno-inline-int-divide @gol
616 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
617 -mno-inline-sqrt @gol
618 -mdwarf2-asm -mearly-stop-bits @gol
619 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
620 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
621 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
622 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
623 -msched-spec-ldc -msched-spec-control-ldc @gol
624 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
625 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
626 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
627 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
629 @emph{IA-64/VMS Options}
630 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
632 @emph{M32R/D Options}
633 @gccoptlist{-m32r2 -m32rx -m32r @gol
635 -malign-loops -mno-align-loops @gol
636 -missue-rate=@var{number} @gol
637 -mbranch-cost=@var{number} @gol
638 -mmodel=@var{code-size-model-type} @gol
639 -msdata=@var{sdata-type} @gol
640 -mno-flush-func -mflush-func=@var{name} @gol
641 -mno-flush-trap -mflush-trap=@var{number} @gol
645 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
647 @emph{M680x0 Options}
648 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
649 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
650 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
651 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
652 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
653 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
654 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
655 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
658 @emph{M68hc1x Options}
659 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
660 -mauto-incdec -minmax -mlong-calls -mshort @gol
661 -msoft-reg-count=@var{count}}
664 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
665 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
666 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
667 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
668 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
671 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
672 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
673 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
674 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
678 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
679 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
680 -mips64 -mips64r2 @gol
681 -mips16 -mno-mips16 -mflip-mips16 @gol
682 -minterlink-mips16 -mno-interlink-mips16 @gol
683 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
684 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
685 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
686 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
687 -mfpu=@var{fpu-type} @gol
688 -msmartmips -mno-smartmips @gol
689 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
690 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
691 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
692 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
693 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
694 -membedded-data -mno-embedded-data @gol
695 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
696 -mcode-readable=@var{setting} @gol
697 -msplit-addresses -mno-split-addresses @gol
698 -mexplicit-relocs -mno-explicit-relocs @gol
699 -mcheck-zero-division -mno-check-zero-division @gol
700 -mdivide-traps -mdivide-breaks @gol
701 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
702 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
703 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
704 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
705 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
706 -mflush-func=@var{func} -mno-flush-func @gol
707 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
708 -mfp-exceptions -mno-fp-exceptions @gol
709 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
710 -mrelax-pic-calls -mno-relax-pic-calls}
713 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
714 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
715 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
716 -mno-base-addresses -msingle-exit -mno-single-exit}
718 @emph{MN10300 Options}
719 @gccoptlist{-mmult-bug -mno-mult-bug @gol
720 -mam33 -mno-am33 @gol
721 -mam33-2 -mno-am33-2 @gol
722 -mreturn-pointer-on-d0 @gol
725 @emph{PDP-11 Options}
726 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
727 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
728 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
729 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
730 -mbranch-expensive -mbranch-cheap @gol
731 -msplit -mno-split -munix-asm -mdec-asm}
733 @emph{picoChip Options}
734 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N}
735 -msymbol-as-address -mno-inefficient-warnings}
737 @emph{PowerPC Options}
738 See RS/6000 and PowerPC Options.
740 @emph{RS/6000 and PowerPC Options}
741 @gccoptlist{-mcpu=@var{cpu-type} @gol
742 -mtune=@var{cpu-type} @gol
743 -mpower -mno-power -mpower2 -mno-power2 @gol
744 -mpowerpc -mpowerpc64 -mno-powerpc @gol
745 -maltivec -mno-altivec @gol
746 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
747 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
748 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
749 -mfprnd -mno-fprnd @gol
750 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
751 -mnew-mnemonics -mold-mnemonics @gol
752 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
753 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
754 -malign-power -malign-natural @gol
755 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
756 -msingle-float -mdouble-float -msimple-fpu @gol
757 -mstring -mno-string -mupdate -mno-update @gol
758 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
759 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
760 -mstrict-align -mno-strict-align -mrelocatable @gol
761 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
762 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
763 -mdynamic-no-pic -maltivec -mswdiv @gol
764 -mprioritize-restricted-insns=@var{priority} @gol
765 -msched-costly-dep=@var{dependence_type} @gol
766 -minsert-sched-nops=@var{scheme} @gol
767 -mcall-sysv -mcall-netbsd @gol
768 -maix-struct-return -msvr4-struct-return @gol
769 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
770 -misel -mno-isel @gol
771 -misel=yes -misel=no @gol
773 -mspe=yes -mspe=no @gol
775 -mgen-cell-microcode -mwarn-cell-microcode @gol
776 -mvrsave -mno-vrsave @gol
777 -mmulhw -mno-mulhw @gol
778 -mdlmzb -mno-dlmzb @gol
779 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
780 -mprototype -mno-prototype @gol
781 -msim -mmvme -mads -myellowknife -memb -msdata @gol
782 -msdata=@var{opt} -mvxworks -G @var{num} -pthread}
784 @emph{S/390 and zSeries Options}
785 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
786 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
787 -mlong-double-64 -mlong-double-128 @gol
788 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
789 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
790 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
791 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
792 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
795 @gccoptlist{-meb -mel @gol
799 -mscore5 -mscore5u -mscore7 -mscore7d}
802 @gccoptlist{-m1 -m2 -m2e @gol
803 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
805 -m4-nofpu -m4-single-only -m4-single -m4 @gol
806 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
807 -m5-64media -m5-64media-nofpu @gol
808 -m5-32media -m5-32media-nofpu @gol
809 -m5-compact -m5-compact-nofpu @gol
810 -mb -ml -mdalign -mrelax @gol
811 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
812 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
813 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
814 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
815 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
819 @gccoptlist{-mcpu=@var{cpu-type} @gol
820 -mtune=@var{cpu-type} @gol
821 -mcmodel=@var{code-model} @gol
822 -m32 -m64 -mapp-regs -mno-app-regs @gol
823 -mfaster-structs -mno-faster-structs @gol
824 -mfpu -mno-fpu -mhard-float -msoft-float @gol
825 -mhard-quad-float -msoft-quad-float @gol
826 -mimpure-text -mno-impure-text -mlittle-endian @gol
827 -mstack-bias -mno-stack-bias @gol
828 -munaligned-doubles -mno-unaligned-doubles @gol
829 -mv8plus -mno-v8plus -mvis -mno-vis
830 -threads -pthreads -pthread}
833 @gccoptlist{-mwarn-reloc -merror-reloc @gol
834 -msafe-dma -munsafe-dma @gol
836 -msmall-mem -mlarge-mem -mstdmain @gol
837 -mfixed-range=@var{register-range}}
839 @emph{System V Options}
840 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
843 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
844 -mprolog-function -mno-prolog-function -mspace @gol
845 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
846 -mapp-regs -mno-app-regs @gol
847 -mdisable-callt -mno-disable-callt @gol
853 @gccoptlist{-mg -mgnu -munix}
855 @emph{VxWorks Options}
856 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
857 -Xbind-lazy -Xbind-now}
859 @emph{x86-64 Options}
860 See i386 and x86-64 Options.
862 @emph{i386 and x86-64 Windows Options}
863 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll
864 -mnop-fun-dllimport -mthread -municode -mwin32 -mwindows}
866 @emph{Xstormy16 Options}
869 @emph{Xtensa Options}
870 @gccoptlist{-mconst16 -mno-const16 @gol
871 -mfused-madd -mno-fused-madd @gol
872 -mserialize-volatile -mno-serialize-volatile @gol
873 -mtext-section-literals -mno-text-section-literals @gol
874 -mtarget-align -mno-target-align @gol
875 -mlongcalls -mno-longcalls}
877 @emph{zSeries Options}
878 See S/390 and zSeries Options.
880 @item Code Generation Options
881 @xref{Code Gen Options,,Options for Code Generation Conventions}.
882 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
883 -ffixed-@var{reg} -fexceptions @gol
884 -fnon-call-exceptions -funwind-tables @gol
885 -fasynchronous-unwind-tables @gol
886 -finhibit-size-directive -finstrument-functions @gol
887 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
888 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
889 -fno-common -fno-ident @gol
890 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
891 -fno-jump-tables @gol
892 -frecord-gcc-switches @gol
893 -freg-struct-return -fshort-enums @gol
894 -fshort-double -fshort-wchar @gol
895 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
896 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
897 -fno-stack-limit -fargument-alias -fargument-noalias @gol
898 -fargument-noalias-global -fargument-noalias-anything @gol
899 -fleading-underscore -ftls-model=@var{model} @gol
900 -ftrapv -fwrapv -fbounds-check @gol
905 * Overall Options:: Controlling the kind of output:
906 an executable, object files, assembler files,
907 or preprocessed source.
908 * C Dialect Options:: Controlling the variant of C language compiled.
909 * C++ Dialect Options:: Variations on C++.
910 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
912 * Language Independent Options:: Controlling how diagnostics should be
914 * Warning Options:: How picky should the compiler be?
915 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
916 * Optimize Options:: How much optimization?
917 * Preprocessor Options:: Controlling header files and macro definitions.
918 Also, getting dependency information for Make.
919 * Assembler Options:: Passing options to the assembler.
920 * Link Options:: Specifying libraries and so on.
921 * Directory Options:: Where to find header files and libraries.
922 Where to find the compiler executable files.
923 * Spec Files:: How to pass switches to sub-processes.
924 * Target Options:: Running a cross-compiler, or an old version of GCC.
927 @node Overall Options
928 @section Options Controlling the Kind of Output
930 Compilation can involve up to four stages: preprocessing, compilation
931 proper, assembly and linking, always in that order. GCC is capable of
932 preprocessing and compiling several files either into several
933 assembler input files, or into one assembler input file; then each
934 assembler input file produces an object file, and linking combines all
935 the object files (those newly compiled, and those specified as input)
936 into an executable file.
938 @cindex file name suffix
939 For any given input file, the file name suffix determines what kind of
944 C source code which must be preprocessed.
947 C source code which should not be preprocessed.
950 C++ source code which should not be preprocessed.
953 Objective-C source code. Note that you must link with the @file{libobjc}
954 library to make an Objective-C program work.
957 Objective-C source code which should not be preprocessed.
961 Objective-C++ source code. Note that you must link with the @file{libobjc}
962 library to make an Objective-C++ program work. Note that @samp{.M} refers
963 to a literal capital M@.
966 Objective-C++ source code which should not be preprocessed.
969 C, C++, Objective-C or Objective-C++ header file to be turned into a
974 @itemx @var{file}.cxx
975 @itemx @var{file}.cpp
976 @itemx @var{file}.CPP
977 @itemx @var{file}.c++
979 C++ source code which must be preprocessed. Note that in @samp{.cxx},
980 the last two letters must both be literally @samp{x}. Likewise,
981 @samp{.C} refers to a literal capital C@.
985 Objective-C++ source code which must be preprocessed.
988 Objective-C++ source code which should not be preprocessed.
993 @itemx @var{file}.hxx
994 @itemx @var{file}.hpp
995 @itemx @var{file}.HPP
996 @itemx @var{file}.h++
997 @itemx @var{file}.tcc
998 C++ header file to be turned into a precompiled header.
1001 @itemx @var{file}.for
1002 @itemx @var{file}.ftn
1003 Fixed form Fortran source code which should not be preprocessed.
1006 @itemx @var{file}.FOR
1007 @itemx @var{file}.fpp
1008 @itemx @var{file}.FPP
1009 @itemx @var{file}.FTN
1010 Fixed form Fortran source code which must be preprocessed (with the traditional
1013 @item @var{file}.f90
1014 @itemx @var{file}.f95
1015 @itemx @var{file}.f03
1016 @itemx @var{file}.f08
1017 Free form Fortran source code which should not be preprocessed.
1019 @item @var{file}.F90
1020 @itemx @var{file}.F95
1021 @itemx @var{file}.F03
1022 @itemx @var{file}.F08
1023 Free form Fortran source code which must be preprocessed (with the
1024 traditional preprocessor).
1026 @c FIXME: Descriptions of Java file types.
1032 @item @var{file}.ads
1033 Ada source code file which contains a library unit declaration (a
1034 declaration of a package, subprogram, or generic, or a generic
1035 instantiation), or a library unit renaming declaration (a package,
1036 generic, or subprogram renaming declaration). Such files are also
1039 @item @var{file}.adb
1040 Ada source code file containing a library unit body (a subprogram or
1041 package body). Such files are also called @dfn{bodies}.
1043 @c GCC also knows about some suffixes for languages not yet included:
1054 @itemx @var{file}.sx
1055 Assembler code which must be preprocessed.
1058 An object file to be fed straight into linking.
1059 Any file name with no recognized suffix is treated this way.
1063 You can specify the input language explicitly with the @option{-x} option:
1066 @item -x @var{language}
1067 Specify explicitly the @var{language} for the following input files
1068 (rather than letting the compiler choose a default based on the file
1069 name suffix). This option applies to all following input files until
1070 the next @option{-x} option. Possible values for @var{language} are:
1072 c c-header c-cpp-output
1073 c++ c++-header c++-cpp-output
1074 objective-c objective-c-header objective-c-cpp-output
1075 objective-c++ objective-c++-header objective-c++-cpp-output
1076 assembler assembler-with-cpp
1078 f77 f77-cpp-input f95 f95-cpp-input
1083 Turn off any specification of a language, so that subsequent files are
1084 handled according to their file name suffixes (as they are if @option{-x}
1085 has not been used at all).
1087 @item -pass-exit-codes
1088 @opindex pass-exit-codes
1089 Normally the @command{gcc} program will exit with the code of 1 if any
1090 phase of the compiler returns a non-success return code. If you specify
1091 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1092 numerically highest error produced by any phase that returned an error
1093 indication. The C, C++, and Fortran frontends return 4, if an internal
1094 compiler error is encountered.
1097 If you only want some of the stages of compilation, you can use
1098 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1099 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1100 @command{gcc} is to stop. Note that some combinations (for example,
1101 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1106 Compile or assemble the source files, but do not link. The linking
1107 stage simply is not done. The ultimate output is in the form of an
1108 object file for each source file.
1110 By default, the object file name for a source file is made by replacing
1111 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1113 Unrecognized input files, not requiring compilation or assembly, are
1118 Stop after the stage of compilation proper; do not assemble. The output
1119 is in the form of an assembler code file for each non-assembler input
1122 By default, the assembler file name for a source file is made by
1123 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1125 Input files that don't require compilation are ignored.
1129 Stop after the preprocessing stage; do not run the compiler proper. The
1130 output is in the form of preprocessed source code, which is sent to the
1133 Input files which don't require preprocessing are ignored.
1135 @cindex output file option
1138 Place output in file @var{file}. This applies regardless to whatever
1139 sort of output is being produced, whether it be an executable file,
1140 an object file, an assembler file or preprocessed C code.
1142 If @option{-o} is not specified, the default is to put an executable
1143 file in @file{a.out}, the object file for
1144 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1145 assembler file in @file{@var{source}.s}, a precompiled header file in
1146 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1151 Print (on standard error output) the commands executed to run the stages
1152 of compilation. Also print the version number of the compiler driver
1153 program and of the preprocessor and the compiler proper.
1157 Like @option{-v} except the commands are not executed and all command
1158 arguments are quoted. This is useful for shell scripts to capture the
1159 driver-generated command lines.
1163 Use pipes rather than temporary files for communication between the
1164 various stages of compilation. This fails to work on some systems where
1165 the assembler is unable to read from a pipe; but the GNU assembler has
1170 If you are compiling multiple source files, this option tells the driver
1171 to pass all the source files to the compiler at once (for those
1172 languages for which the compiler can handle this). This will allow
1173 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1174 language for which this is supported is C@. If you pass source files for
1175 multiple languages to the driver, using this option, the driver will invoke
1176 the compiler(s) that support IMA once each, passing each compiler all the
1177 source files appropriate for it. For those languages that do not support
1178 IMA this option will be ignored, and the compiler will be invoked once for
1179 each source file in that language. If you use this option in conjunction
1180 with @option{-save-temps}, the compiler will generate multiple
1182 (one for each source file), but only one (combined) @file{.o} or
1187 Print (on the standard output) a description of the command line options
1188 understood by @command{gcc}. If the @option{-v} option is also specified
1189 then @option{--help} will also be passed on to the various processes
1190 invoked by @command{gcc}, so that they can display the command line options
1191 they accept. If the @option{-Wextra} option has also been specified
1192 (prior to the @option{--help} option), then command line options which
1193 have no documentation associated with them will also be displayed.
1196 @opindex target-help
1197 Print (on the standard output) a description of target-specific command
1198 line options for each tool. For some targets extra target-specific
1199 information may also be printed.
1201 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1202 Print (on the standard output) a description of the command line
1203 options understood by the compiler that fit into all specified classes
1204 and qualifiers. These are the supported classes:
1207 @item @samp{optimizers}
1208 This will display all of the optimization options supported by the
1211 @item @samp{warnings}
1212 This will display all of the options controlling warning messages
1213 produced by the compiler.
1216 This will display target-specific options. Unlike the
1217 @option{--target-help} option however, target-specific options of the
1218 linker and assembler will not be displayed. This is because those
1219 tools do not currently support the extended @option{--help=} syntax.
1222 This will display the values recognized by the @option{--param}
1225 @item @var{language}
1226 This will display the options supported for @var{language}, where
1227 @var{language} is the name of one of the languages supported in this
1231 This will display the options that are common to all languages.
1234 These are the supported qualifiers:
1237 @item @samp{undocumented}
1238 Display only those options which are undocumented.
1241 Display options which take an argument that appears after an equal
1242 sign in the same continuous piece of text, such as:
1243 @samp{--help=target}.
1245 @item @samp{separate}
1246 Display options which take an argument that appears as a separate word
1247 following the original option, such as: @samp{-o output-file}.
1250 Thus for example to display all the undocumented target-specific
1251 switches supported by the compiler the following can be used:
1254 --help=target,undocumented
1257 The sense of a qualifier can be inverted by prefixing it with the
1258 @samp{^} character, so for example to display all binary warning
1259 options (i.e., ones that are either on or off and that do not take an
1260 argument), which have a description the following can be used:
1263 --help=warnings,^joined,^undocumented
1266 The argument to @option{--help=} should not consist solely of inverted
1269 Combining several classes is possible, although this usually
1270 restricts the output by so much that there is nothing to display. One
1271 case where it does work however is when one of the classes is
1272 @var{target}. So for example to display all the target-specific
1273 optimization options the following can be used:
1276 --help=target,optimizers
1279 The @option{--help=} option can be repeated on the command line. Each
1280 successive use will display its requested class of options, skipping
1281 those that have already been displayed.
1283 If the @option{-Q} option appears on the command line before the
1284 @option{--help=} option, then the descriptive text displayed by
1285 @option{--help=} is changed. Instead of describing the displayed
1286 options, an indication is given as to whether the option is enabled,
1287 disabled or set to a specific value (assuming that the compiler
1288 knows this at the point where the @option{--help=} option is used).
1290 Here is a truncated example from the ARM port of @command{gcc}:
1293 % gcc -Q -mabi=2 --help=target -c
1294 The following options are target specific:
1296 -mabort-on-noreturn [disabled]
1300 The output is sensitive to the effects of previous command line
1301 options, so for example it is possible to find out which optimizations
1302 are enabled at @option{-O2} by using:
1305 -Q -O2 --help=optimizers
1308 Alternatively you can discover which binary optimizations are enabled
1309 by @option{-O3} by using:
1312 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1313 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1314 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1317 @item -no-canonical-prefixes
1318 @opindex no-canonical-prefixes
1319 Do not expand any symbolic links, resolve references to @samp{/../}
1320 or @samp{/./}, or make the path absolute when generating a relative
1325 Display the version number and copyrights of the invoked GCC@.
1329 Invoke all subcommands under a wrapper program. It takes a single
1330 comma separated list as an argument, which will be used to invoke
1334 gcc -c t.c -wrapper gdb,--args
1337 This will invoke all subprograms of gcc under "gdb --args",
1338 thus cc1 invocation will be "gdb --args cc1 ...".
1340 @item -fplugin=@var{name}.so
1341 Load the plugin code in file @var{name}.so, assumed to be a
1342 shared object to be dlopen'd by the compiler. The base name of
1343 the shared object file is used to identify the plugin for the
1344 purposes of argument parsing (See
1345 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1346 Each plugin should define the callback functions specified in the
1349 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1350 Define an argument called @var{key} with a value of @var{value}
1351 for the plugin called @var{name}.
1353 @include @value{srcdir}/../libiberty/at-file.texi
1357 @section Compiling C++ Programs
1359 @cindex suffixes for C++ source
1360 @cindex C++ source file suffixes
1361 C++ source files conventionally use one of the suffixes @samp{.C},
1362 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1363 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1364 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1365 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1366 files with these names and compiles them as C++ programs even if you
1367 call the compiler the same way as for compiling C programs (usually
1368 with the name @command{gcc}).
1372 However, the use of @command{gcc} does not add the C++ library.
1373 @command{g++} is a program that calls GCC and treats @samp{.c},
1374 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1375 files unless @option{-x} is used, and automatically specifies linking
1376 against the C++ library. This program is also useful when
1377 precompiling a C header file with a @samp{.h} extension for use in C++
1378 compilations. On many systems, @command{g++} is also installed with
1379 the name @command{c++}.
1381 @cindex invoking @command{g++}
1382 When you compile C++ programs, you may specify many of the same
1383 command-line options that you use for compiling programs in any
1384 language; or command-line options meaningful for C and related
1385 languages; or options that are meaningful only for C++ programs.
1386 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1387 explanations of options for languages related to C@.
1388 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1389 explanations of options that are meaningful only for C++ programs.
1391 @node C Dialect Options
1392 @section Options Controlling C Dialect
1393 @cindex dialect options
1394 @cindex language dialect options
1395 @cindex options, dialect
1397 The following options control the dialect of C (or languages derived
1398 from C, such as C++, Objective-C and Objective-C++) that the compiler
1402 @cindex ANSI support
1406 In C mode, this is equivalent to @samp{-std=c89}. In C++ mode, it is
1407 equivalent to @samp{-std=c++98}.
1409 This turns off certain features of GCC that are incompatible with ISO
1410 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1411 such as the @code{asm} and @code{typeof} keywords, and
1412 predefined macros such as @code{unix} and @code{vax} that identify the
1413 type of system you are using. It also enables the undesirable and
1414 rarely used ISO trigraph feature. For the C compiler,
1415 it disables recognition of C++ style @samp{//} comments as well as
1416 the @code{inline} keyword.
1418 The alternate keywords @code{__asm__}, @code{__extension__},
1419 @code{__inline__} and @code{__typeof__} continue to work despite
1420 @option{-ansi}. You would not want to use them in an ISO C program, of
1421 course, but it is useful to put them in header files that might be included
1422 in compilations done with @option{-ansi}. Alternate predefined macros
1423 such as @code{__unix__} and @code{__vax__} are also available, with or
1424 without @option{-ansi}.
1426 The @option{-ansi} option does not cause non-ISO programs to be
1427 rejected gratuitously. For that, @option{-pedantic} is required in
1428 addition to @option{-ansi}. @xref{Warning Options}.
1430 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1431 option is used. Some header files may notice this macro and refrain
1432 from declaring certain functions or defining certain macros that the
1433 ISO standard doesn't call for; this is to avoid interfering with any
1434 programs that might use these names for other things.
1436 Functions that would normally be built in but do not have semantics
1437 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1438 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1439 built-in functions provided by GCC}, for details of the functions
1444 Determine the language standard. @xref{Standards,,Language Standards
1445 Supported by GCC}, for details of these standard versions. This option
1446 is currently only supported when compiling C or C++.
1448 The compiler can accept several base standards, such as @samp{c89} or
1449 @samp{c++98}, and GNU dialects of those standards, such as
1450 @samp{gnu89} or @samp{gnu++98}. By specifying a base standard, the
1451 compiler will accept all programs following that standard and those
1452 using GNU extensions that do not contradict it. For example,
1453 @samp{-std=c89} turns off certain features of GCC that are
1454 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1455 keywords, but not other GNU extensions that do not have a meaning in
1456 ISO C90, such as omitting the middle term of a @code{?:}
1457 expression. On the other hand, by specifying a GNU dialect of a
1458 standard, all features the compiler support are enabled, even when
1459 those features change the meaning of the base standard and some
1460 strict-conforming programs may be rejected. The particular standard
1461 is used by @option{-pedantic} to identify which features are GNU
1462 extensions given that version of the standard. For example
1463 @samp{-std=gnu89 -pedantic} would warn about C++ style @samp{//}
1464 comments, while @samp{-std=gnu99 -pedantic} would not.
1466 A value for this option must be provided; possible values are
1471 Support all ISO C90 programs (certain GNU extensions that conflict
1472 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1474 @item iso9899:199409
1475 ISO C90 as modified in amendment 1.
1481 ISO C99. Note that this standard is not yet fully supported; see
1482 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1483 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1486 GNU dialect of ISO C90 (including some C99 features). This
1487 is the default for C code.
1491 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1492 this will become the default. The name @samp{gnu9x} is deprecated.
1495 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1499 GNU dialect of @option{-std=c++98}. This is the default for
1503 The working draft of the upcoming ISO C++0x standard. This option
1504 enables experimental features that are likely to be included in
1505 C++0x. The working draft is constantly changing, and any feature that is
1506 enabled by this flag may be removed from future versions of GCC if it is
1507 not part of the C++0x standard.
1510 GNU dialect of @option{-std=c++0x}. This option enables
1511 experimental features that may be removed in future versions of GCC.
1514 @item -fgnu89-inline
1515 @opindex fgnu89-inline
1516 The option @option{-fgnu89-inline} tells GCC to use the traditional
1517 GNU semantics for @code{inline} functions when in C99 mode.
1518 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1519 is accepted and ignored by GCC versions 4.1.3 up to but not including
1520 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1521 C99 mode. Using this option is roughly equivalent to adding the
1522 @code{gnu_inline} function attribute to all inline functions
1523 (@pxref{Function Attributes}).
1525 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1526 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1527 specifies the default behavior). This option was first supported in
1528 GCC 4.3. This option is not supported in C89 or gnu89 mode.
1530 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1531 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1532 in effect for @code{inline} functions. @xref{Common Predefined
1533 Macros,,,cpp,The C Preprocessor}.
1535 @item -aux-info @var{filename}
1537 Output to the given filename prototyped declarations for all functions
1538 declared and/or defined in a translation unit, including those in header
1539 files. This option is silently ignored in any language other than C@.
1541 Besides declarations, the file indicates, in comments, the origin of
1542 each declaration (source file and line), whether the declaration was
1543 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1544 @samp{O} for old, respectively, in the first character after the line
1545 number and the colon), and whether it came from a declaration or a
1546 definition (@samp{C} or @samp{F}, respectively, in the following
1547 character). In the case of function definitions, a K&R-style list of
1548 arguments followed by their declarations is also provided, inside
1549 comments, after the declaration.
1553 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1554 keyword, so that code can use these words as identifiers. You can use
1555 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1556 instead. @option{-ansi} implies @option{-fno-asm}.
1558 In C++, this switch only affects the @code{typeof} keyword, since
1559 @code{asm} and @code{inline} are standard keywords. You may want to
1560 use the @option{-fno-gnu-keywords} flag instead, which has the same
1561 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1562 switch only affects the @code{asm} and @code{typeof} keywords, since
1563 @code{inline} is a standard keyword in ISO C99.
1566 @itemx -fno-builtin-@var{function}
1567 @opindex fno-builtin
1568 @cindex built-in functions
1569 Don't recognize built-in functions that do not begin with
1570 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1571 functions provided by GCC}, for details of the functions affected,
1572 including those which are not built-in functions when @option{-ansi} or
1573 @option{-std} options for strict ISO C conformance are used because they
1574 do not have an ISO standard meaning.
1576 GCC normally generates special code to handle certain built-in functions
1577 more efficiently; for instance, calls to @code{alloca} may become single
1578 instructions that adjust the stack directly, and calls to @code{memcpy}
1579 may become inline copy loops. The resulting code is often both smaller
1580 and faster, but since the function calls no longer appear as such, you
1581 cannot set a breakpoint on those calls, nor can you change the behavior
1582 of the functions by linking with a different library. In addition,
1583 when a function is recognized as a built-in function, GCC may use
1584 information about that function to warn about problems with calls to
1585 that function, or to generate more efficient code, even if the
1586 resulting code still contains calls to that function. For example,
1587 warnings are given with @option{-Wformat} for bad calls to
1588 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1589 known not to modify global memory.
1591 With the @option{-fno-builtin-@var{function}} option
1592 only the built-in function @var{function} is
1593 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1594 function is named that is not built-in in this version of GCC, this
1595 option is ignored. There is no corresponding
1596 @option{-fbuiltin-@var{function}} option; if you wish to enable
1597 built-in functions selectively when using @option{-fno-builtin} or
1598 @option{-ffreestanding}, you may define macros such as:
1601 #define abs(n) __builtin_abs ((n))
1602 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1607 @cindex hosted environment
1609 Assert that compilation takes place in a hosted environment. This implies
1610 @option{-fbuiltin}. A hosted environment is one in which the
1611 entire standard library is available, and in which @code{main} has a return
1612 type of @code{int}. Examples are nearly everything except a kernel.
1613 This is equivalent to @option{-fno-freestanding}.
1615 @item -ffreestanding
1616 @opindex ffreestanding
1617 @cindex hosted environment
1619 Assert that compilation takes place in a freestanding environment. This
1620 implies @option{-fno-builtin}. A freestanding environment
1621 is one in which the standard library may not exist, and program startup may
1622 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1623 This is equivalent to @option{-fno-hosted}.
1625 @xref{Standards,,Language Standards Supported by GCC}, for details of
1626 freestanding and hosted environments.
1630 @cindex openmp parallel
1631 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1632 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1633 compiler generates parallel code according to the OpenMP Application
1634 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1635 implies @option{-pthread}, and thus is only supported on targets that
1636 have support for @option{-pthread}.
1638 @item -fms-extensions
1639 @opindex fms-extensions
1640 Accept some non-standard constructs used in Microsoft header files.
1642 Some cases of unnamed fields in structures and unions are only
1643 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1644 fields within structs/unions}, for details.
1648 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1649 options for strict ISO C conformance) implies @option{-trigraphs}.
1651 @item -no-integrated-cpp
1652 @opindex no-integrated-cpp
1653 Performs a compilation in two passes: preprocessing and compiling. This
1654 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1655 @option{-B} option. The user supplied compilation step can then add in
1656 an additional preprocessing step after normal preprocessing but before
1657 compiling. The default is to use the integrated cpp (internal cpp)
1659 The semantics of this option will change if "cc1", "cc1plus", and
1660 "cc1obj" are merged.
1662 @cindex traditional C language
1663 @cindex C language, traditional
1665 @itemx -traditional-cpp
1666 @opindex traditional-cpp
1667 @opindex traditional
1668 Formerly, these options caused GCC to attempt to emulate a pre-standard
1669 C compiler. They are now only supported with the @option{-E} switch.
1670 The preprocessor continues to support a pre-standard mode. See the GNU
1671 CPP manual for details.
1673 @item -fcond-mismatch
1674 @opindex fcond-mismatch
1675 Allow conditional expressions with mismatched types in the second and
1676 third arguments. The value of such an expression is void. This option
1677 is not supported for C++.
1679 @item -flax-vector-conversions
1680 @opindex flax-vector-conversions
1681 Allow implicit conversions between vectors with differing numbers of
1682 elements and/or incompatible element types. This option should not be
1685 @item -funsigned-char
1686 @opindex funsigned-char
1687 Let the type @code{char} be unsigned, like @code{unsigned char}.
1689 Each kind of machine has a default for what @code{char} should
1690 be. It is either like @code{unsigned char} by default or like
1691 @code{signed char} by default.
1693 Ideally, a portable program should always use @code{signed char} or
1694 @code{unsigned char} when it depends on the signedness of an object.
1695 But many programs have been written to use plain @code{char} and
1696 expect it to be signed, or expect it to be unsigned, depending on the
1697 machines they were written for. This option, and its inverse, let you
1698 make such a program work with the opposite default.
1700 The type @code{char} is always a distinct type from each of
1701 @code{signed char} or @code{unsigned char}, even though its behavior
1702 is always just like one of those two.
1705 @opindex fsigned-char
1706 Let the type @code{char} be signed, like @code{signed char}.
1708 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1709 the negative form of @option{-funsigned-char}. Likewise, the option
1710 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1712 @item -fsigned-bitfields
1713 @itemx -funsigned-bitfields
1714 @itemx -fno-signed-bitfields
1715 @itemx -fno-unsigned-bitfields
1716 @opindex fsigned-bitfields
1717 @opindex funsigned-bitfields
1718 @opindex fno-signed-bitfields
1719 @opindex fno-unsigned-bitfields
1720 These options control whether a bit-field is signed or unsigned, when the
1721 declaration does not use either @code{signed} or @code{unsigned}. By
1722 default, such a bit-field is signed, because this is consistent: the
1723 basic integer types such as @code{int} are signed types.
1726 @node C++ Dialect Options
1727 @section Options Controlling C++ Dialect
1729 @cindex compiler options, C++
1730 @cindex C++ options, command line
1731 @cindex options, C++
1732 This section describes the command-line options that are only meaningful
1733 for C++ programs; but you can also use most of the GNU compiler options
1734 regardless of what language your program is in. For example, you
1735 might compile a file @code{firstClass.C} like this:
1738 g++ -g -frepo -O -c firstClass.C
1742 In this example, only @option{-frepo} is an option meant
1743 only for C++ programs; you can use the other options with any
1744 language supported by GCC@.
1746 Here is a list of options that are @emph{only} for compiling C++ programs:
1750 @item -fabi-version=@var{n}
1751 @opindex fabi-version
1752 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1753 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1754 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1755 the version that conforms most closely to the C++ ABI specification.
1756 Therefore, the ABI obtained using version 0 will change as ABI bugs
1759 The default is version 2.
1761 @item -fno-access-control
1762 @opindex fno-access-control
1763 Turn off all access checking. This switch is mainly useful for working
1764 around bugs in the access control code.
1768 Check that the pointer returned by @code{operator new} is non-null
1769 before attempting to modify the storage allocated. This check is
1770 normally unnecessary because the C++ standard specifies that
1771 @code{operator new} will only return @code{0} if it is declared
1772 @samp{throw()}, in which case the compiler will always check the
1773 return value even without this option. In all other cases, when
1774 @code{operator new} has a non-empty exception specification, memory
1775 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1776 @samp{new (nothrow)}.
1778 @item -fconserve-space
1779 @opindex fconserve-space
1780 Put uninitialized or runtime-initialized global variables into the
1781 common segment, as C does. This saves space in the executable at the
1782 cost of not diagnosing duplicate definitions. If you compile with this
1783 flag and your program mysteriously crashes after @code{main()} has
1784 completed, you may have an object that is being destroyed twice because
1785 two definitions were merged.
1787 This option is no longer useful on most targets, now that support has
1788 been added for putting variables into BSS without making them common.
1790 @item -fno-deduce-init-list
1791 @opindex fno-deduce-init-list
1792 Disable deduction of a template type parameter as
1793 std::initializer_list from a brace-enclosed initializer list, i.e.
1796 template <class T> auto forward(T t) -> decltype (realfn (t))
1803 forward(@{1,2@}); // call forward<std::initializer_list<int>>
1807 This option is present because this deduction is an extension to the
1808 current specification in the C++0x working draft, and there was
1809 some concern about potential overload resolution problems.
1811 @item -ffriend-injection
1812 @opindex ffriend-injection
1813 Inject friend functions into the enclosing namespace, so that they are
1814 visible outside the scope of the class in which they are declared.
1815 Friend functions were documented to work this way in the old Annotated
1816 C++ Reference Manual, and versions of G++ before 4.1 always worked
1817 that way. However, in ISO C++ a friend function which is not declared
1818 in an enclosing scope can only be found using argument dependent
1819 lookup. This option causes friends to be injected as they were in
1822 This option is for compatibility, and may be removed in a future
1825 @item -fno-elide-constructors
1826 @opindex fno-elide-constructors
1827 The C++ standard allows an implementation to omit creating a temporary
1828 which is only used to initialize another object of the same type.
1829 Specifying this option disables that optimization, and forces G++ to
1830 call the copy constructor in all cases.
1832 @item -fno-enforce-eh-specs
1833 @opindex fno-enforce-eh-specs
1834 Don't generate code to check for violation of exception specifications
1835 at runtime. This option violates the C++ standard, but may be useful
1836 for reducing code size in production builds, much like defining
1837 @samp{NDEBUG}. This does not give user code permission to throw
1838 exceptions in violation of the exception specifications; the compiler
1839 will still optimize based on the specifications, so throwing an
1840 unexpected exception will result in undefined behavior.
1843 @itemx -fno-for-scope
1845 @opindex fno-for-scope
1846 If @option{-ffor-scope} is specified, the scope of variables declared in
1847 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1848 as specified by the C++ standard.
1849 If @option{-fno-for-scope} is specified, the scope of variables declared in
1850 a @i{for-init-statement} extends to the end of the enclosing scope,
1851 as was the case in old versions of G++, and other (traditional)
1852 implementations of C++.
1854 The default if neither flag is given to follow the standard,
1855 but to allow and give a warning for old-style code that would
1856 otherwise be invalid, or have different behavior.
1858 @item -fno-gnu-keywords
1859 @opindex fno-gnu-keywords
1860 Do not recognize @code{typeof} as a keyword, so that code can use this
1861 word as an identifier. You can use the keyword @code{__typeof__} instead.
1862 @option{-ansi} implies @option{-fno-gnu-keywords}.
1864 @item -fno-implicit-templates
1865 @opindex fno-implicit-templates
1866 Never emit code for non-inline templates which are instantiated
1867 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1868 @xref{Template Instantiation}, for more information.
1870 @item -fno-implicit-inline-templates
1871 @opindex fno-implicit-inline-templates
1872 Don't emit code for implicit instantiations of inline templates, either.
1873 The default is to handle inlines differently so that compiles with and
1874 without optimization will need the same set of explicit instantiations.
1876 @item -fno-implement-inlines
1877 @opindex fno-implement-inlines
1878 To save space, do not emit out-of-line copies of inline functions
1879 controlled by @samp{#pragma implementation}. This will cause linker
1880 errors if these functions are not inlined everywhere they are called.
1882 @item -fms-extensions
1883 @opindex fms-extensions
1884 Disable pedantic warnings about constructs used in MFC, such as implicit
1885 int and getting a pointer to member function via non-standard syntax.
1887 @item -fno-nonansi-builtins
1888 @opindex fno-nonansi-builtins
1889 Disable built-in declarations of functions that are not mandated by
1890 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1891 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1893 @item -fno-operator-names
1894 @opindex fno-operator-names
1895 Do not treat the operator name keywords @code{and}, @code{bitand},
1896 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1897 synonyms as keywords.
1899 @item -fno-optional-diags
1900 @opindex fno-optional-diags
1901 Disable diagnostics that the standard says a compiler does not need to
1902 issue. Currently, the only such diagnostic issued by G++ is the one for
1903 a name having multiple meanings within a class.
1906 @opindex fpermissive
1907 Downgrade some diagnostics about nonconformant code from errors to
1908 warnings. Thus, using @option{-fpermissive} will allow some
1909 nonconforming code to compile.
1911 @item -fno-pretty-templates
1912 @opindex fno-pretty-templates
1913 When an error message refers to a specialization of a function
1914 template, the compiler will normally print the signature of the
1915 template followed by the template arguments and any typedefs or
1916 typenames in the signature (e.g. @code{void f(T) [with T = int]}
1917 rather than @code{void f(int)}) so that it's clear which template is
1918 involved. When an error message refers to a specialization of a class
1919 template, the compiler will omit any template arguments which match
1920 the default template arguments for that template. If either of these
1921 behaviors make it harder to understand the error message rather than
1922 easier, using @option{-fno-pretty-templates} will disable them.
1926 Enable automatic template instantiation at link time. This option also
1927 implies @option{-fno-implicit-templates}. @xref{Template
1928 Instantiation}, for more information.
1932 Disable generation of information about every class with virtual
1933 functions for use by the C++ runtime type identification features
1934 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1935 of the language, you can save some space by using this flag. Note that
1936 exception handling uses the same information, but it will generate it as
1937 needed. The @samp{dynamic_cast} operator can still be used for casts that
1938 do not require runtime type information, i.e.@: casts to @code{void *} or to
1939 unambiguous base classes.
1943 Emit statistics about front-end processing at the end of the compilation.
1944 This information is generally only useful to the G++ development team.
1946 @item -ftemplate-depth-@var{n}
1947 @opindex ftemplate-depth
1948 Set the maximum instantiation depth for template classes to @var{n}.
1949 A limit on the template instantiation depth is needed to detect
1950 endless recursions during template class instantiation. ANSI/ISO C++
1951 conforming programs must not rely on a maximum depth greater than 17
1952 (changed to 1024 in C++0x).
1954 @item -fno-threadsafe-statics
1955 @opindex fno-threadsafe-statics
1956 Do not emit the extra code to use the routines specified in the C++
1957 ABI for thread-safe initialization of local statics. You can use this
1958 option to reduce code size slightly in code that doesn't need to be
1961 @item -fuse-cxa-atexit
1962 @opindex fuse-cxa-atexit
1963 Register destructors for objects with static storage duration with the
1964 @code{__cxa_atexit} function rather than the @code{atexit} function.
1965 This option is required for fully standards-compliant handling of static
1966 destructors, but will only work if your C library supports
1967 @code{__cxa_atexit}.
1969 @item -fno-use-cxa-get-exception-ptr
1970 @opindex fno-use-cxa-get-exception-ptr
1971 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
1972 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
1973 if the runtime routine is not available.
1975 @item -fvisibility-inlines-hidden
1976 @opindex fvisibility-inlines-hidden
1977 This switch declares that the user does not attempt to compare
1978 pointers to inline methods where the addresses of the two functions
1979 were taken in different shared objects.
1981 The effect of this is that GCC may, effectively, mark inline methods with
1982 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1983 appear in the export table of a DSO and do not require a PLT indirection
1984 when used within the DSO@. Enabling this option can have a dramatic effect
1985 on load and link times of a DSO as it massively reduces the size of the
1986 dynamic export table when the library makes heavy use of templates.
1988 The behavior of this switch is not quite the same as marking the
1989 methods as hidden directly, because it does not affect static variables
1990 local to the function or cause the compiler to deduce that
1991 the function is defined in only one shared object.
1993 You may mark a method as having a visibility explicitly to negate the
1994 effect of the switch for that method. For example, if you do want to
1995 compare pointers to a particular inline method, you might mark it as
1996 having default visibility. Marking the enclosing class with explicit
1997 visibility will have no effect.
1999 Explicitly instantiated inline methods are unaffected by this option
2000 as their linkage might otherwise cross a shared library boundary.
2001 @xref{Template Instantiation}.
2003 @item -fvisibility-ms-compat
2004 @opindex fvisibility-ms-compat
2005 This flag attempts to use visibility settings to make GCC's C++
2006 linkage model compatible with that of Microsoft Visual Studio.
2008 The flag makes these changes to GCC's linkage model:
2012 It sets the default visibility to @code{hidden}, like
2013 @option{-fvisibility=hidden}.
2016 Types, but not their members, are not hidden by default.
2019 The One Definition Rule is relaxed for types without explicit
2020 visibility specifications which are defined in more than one different
2021 shared object: those declarations are permitted if they would have
2022 been permitted when this option was not used.
2025 In new code it is better to use @option{-fvisibility=hidden} and
2026 export those classes which are intended to be externally visible.
2027 Unfortunately it is possible for code to rely, perhaps accidentally,
2028 on the Visual Studio behavior.
2030 Among the consequences of these changes are that static data members
2031 of the same type with the same name but defined in different shared
2032 objects will be different, so changing one will not change the other;
2033 and that pointers to function members defined in different shared
2034 objects may not compare equal. When this flag is given, it is a
2035 violation of the ODR to define types with the same name differently.
2039 Do not use weak symbol support, even if it is provided by the linker.
2040 By default, G++ will use weak symbols if they are available. This
2041 option exists only for testing, and should not be used by end-users;
2042 it will result in inferior code and has no benefits. This option may
2043 be removed in a future release of G++.
2047 Do not search for header files in the standard directories specific to
2048 C++, but do still search the other standard directories. (This option
2049 is used when building the C++ library.)
2052 In addition, these optimization, warning, and code generation options
2053 have meanings only for C++ programs:
2056 @item -fno-default-inline
2057 @opindex fno-default-inline
2058 Do not assume @samp{inline} for functions defined inside a class scope.
2059 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2060 functions will have linkage like inline functions; they just won't be
2063 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2066 Warn when G++ generates code that is probably not compatible with the
2067 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2068 all such cases, there are probably some cases that are not warned about,
2069 even though G++ is generating incompatible code. There may also be
2070 cases where warnings are emitted even though the code that is generated
2073 You should rewrite your code to avoid these warnings if you are
2074 concerned about the fact that code generated by G++ may not be binary
2075 compatible with code generated by other compilers.
2077 The known incompatibilities at this point include:
2082 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2083 pack data into the same byte as a base class. For example:
2086 struct A @{ virtual void f(); int f1 : 1; @};
2087 struct B : public A @{ int f2 : 1; @};
2091 In this case, G++ will place @code{B::f2} into the same byte
2092 as@code{A::f1}; other compilers will not. You can avoid this problem
2093 by explicitly padding @code{A} so that its size is a multiple of the
2094 byte size on your platform; that will cause G++ and other compilers to
2095 layout @code{B} identically.
2098 Incorrect handling of tail-padding for virtual bases. G++ does not use
2099 tail padding when laying out virtual bases. For example:
2102 struct A @{ virtual void f(); char c1; @};
2103 struct B @{ B(); char c2; @};
2104 struct C : public A, public virtual B @{@};
2108 In this case, G++ will not place @code{B} into the tail-padding for
2109 @code{A}; other compilers will. You can avoid this problem by
2110 explicitly padding @code{A} so that its size is a multiple of its
2111 alignment (ignoring virtual base classes); that will cause G++ and other
2112 compilers to layout @code{C} identically.
2115 Incorrect handling of bit-fields with declared widths greater than that
2116 of their underlying types, when the bit-fields appear in a union. For
2120 union U @{ int i : 4096; @};
2124 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2125 union too small by the number of bits in an @code{int}.
2128 Empty classes can be placed at incorrect offsets. For example:
2138 struct C : public B, public A @{@};
2142 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2143 it should be placed at offset zero. G++ mistakenly believes that the
2144 @code{A} data member of @code{B} is already at offset zero.
2147 Names of template functions whose types involve @code{typename} or
2148 template template parameters can be mangled incorrectly.
2151 template <typename Q>
2152 void f(typename Q::X) @{@}
2154 template <template <typename> class Q>
2155 void f(typename Q<int>::X) @{@}
2159 Instantiations of these templates may be mangled incorrectly.
2163 It also warns psABI related changes. The known psABI changes at this
2169 For SYSV/x86-64, when passing union with long double, it is changed to
2170 pass in memory as specified in psABI. For example:
2180 @code{union U} will always be passed in memory.
2184 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2185 @opindex Wctor-dtor-privacy
2186 @opindex Wno-ctor-dtor-privacy
2187 Warn when a class seems unusable because all the constructors or
2188 destructors in that class are private, and it has neither friends nor
2189 public static member functions.
2191 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2192 @opindex Wnon-virtual-dtor
2193 @opindex Wno-non-virtual-dtor
2194 Warn when a class has virtual functions and accessible non-virtual
2195 destructor, in which case it would be possible but unsafe to delete
2196 an instance of a derived class through a pointer to the base class.
2197 This warning is also enabled if -Weffc++ is specified.
2199 @item -Wreorder @r{(C++ and Objective-C++ only)}
2201 @opindex Wno-reorder
2202 @cindex reordering, warning
2203 @cindex warning for reordering of member initializers
2204 Warn when the order of member initializers given in the code does not
2205 match the order in which they must be executed. For instance:
2211 A(): j (0), i (1) @{ @}
2215 The compiler will rearrange the member initializers for @samp{i}
2216 and @samp{j} to match the declaration order of the members, emitting
2217 a warning to that effect. This warning is enabled by @option{-Wall}.
2220 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2223 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2226 Warn about violations of the following style guidelines from Scott Meyers'
2227 @cite{Effective C++} book:
2231 Item 11: Define a copy constructor and an assignment operator for classes
2232 with dynamically allocated memory.
2235 Item 12: Prefer initialization to assignment in constructors.
2238 Item 14: Make destructors virtual in base classes.
2241 Item 15: Have @code{operator=} return a reference to @code{*this}.
2244 Item 23: Don't try to return a reference when you must return an object.
2248 Also warn about violations of the following style guidelines from
2249 Scott Meyers' @cite{More Effective C++} book:
2253 Item 6: Distinguish between prefix and postfix forms of increment and
2254 decrement operators.
2257 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2261 When selecting this option, be aware that the standard library
2262 headers do not obey all of these guidelines; use @samp{grep -v}
2263 to filter out those warnings.
2265 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2266 @opindex Wstrict-null-sentinel
2267 @opindex Wno-strict-null-sentinel
2268 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2269 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2270 to @code{__null}. Although it is a null pointer constant not a null pointer,
2271 it is guaranteed to be of the same size as a pointer. But this use is
2272 not portable across different compilers.
2274 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2275 @opindex Wno-non-template-friend
2276 @opindex Wnon-template-friend
2277 Disable warnings when non-templatized friend functions are declared
2278 within a template. Since the advent of explicit template specification
2279 support in G++, if the name of the friend is an unqualified-id (i.e.,
2280 @samp{friend foo(int)}), the C++ language specification demands that the
2281 friend declare or define an ordinary, nontemplate function. (Section
2282 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2283 could be interpreted as a particular specialization of a templatized
2284 function. Because this non-conforming behavior is no longer the default
2285 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2286 check existing code for potential trouble spots and is on by default.
2287 This new compiler behavior can be turned off with
2288 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2289 but disables the helpful warning.
2291 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2292 @opindex Wold-style-cast
2293 @opindex Wno-old-style-cast
2294 Warn if an old-style (C-style) cast to a non-void type is used within
2295 a C++ program. The new-style casts (@samp{dynamic_cast},
2296 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2297 less vulnerable to unintended effects and much easier to search for.
2299 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2300 @opindex Woverloaded-virtual
2301 @opindex Wno-overloaded-virtual
2302 @cindex overloaded virtual fn, warning
2303 @cindex warning for overloaded virtual fn
2304 Warn when a function declaration hides virtual functions from a
2305 base class. For example, in:
2312 struct B: public A @{
2317 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2325 will fail to compile.
2327 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2328 @opindex Wno-pmf-conversions
2329 @opindex Wpmf-conversions
2330 Disable the diagnostic for converting a bound pointer to member function
2333 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2334 @opindex Wsign-promo
2335 @opindex Wno-sign-promo
2336 Warn when overload resolution chooses a promotion from unsigned or
2337 enumerated type to a signed type, over a conversion to an unsigned type of
2338 the same size. Previous versions of G++ would try to preserve
2339 unsignedness, but the standard mandates the current behavior.
2344 A& operator = (int);
2354 In this example, G++ will synthesize a default @samp{A& operator =
2355 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2358 @node Objective-C and Objective-C++ Dialect Options
2359 @section Options Controlling Objective-C and Objective-C++ Dialects
2361 @cindex compiler options, Objective-C and Objective-C++
2362 @cindex Objective-C and Objective-C++ options, command line
2363 @cindex options, Objective-C and Objective-C++
2364 (NOTE: This manual does not describe the Objective-C and Objective-C++
2365 languages themselves. See @xref{Standards,,Language Standards
2366 Supported by GCC}, for references.)
2368 This section describes the command-line options that are only meaningful
2369 for Objective-C and Objective-C++ programs, but you can also use most of
2370 the language-independent GNU compiler options.
2371 For example, you might compile a file @code{some_class.m} like this:
2374 gcc -g -fgnu-runtime -O -c some_class.m
2378 In this example, @option{-fgnu-runtime} is an option meant only for
2379 Objective-C and Objective-C++ programs; you can use the other options with
2380 any language supported by GCC@.
2382 Note that since Objective-C is an extension of the C language, Objective-C
2383 compilations may also use options specific to the C front-end (e.g.,
2384 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2385 C++-specific options (e.g., @option{-Wabi}).
2387 Here is a list of options that are @emph{only} for compiling Objective-C
2388 and Objective-C++ programs:
2391 @item -fconstant-string-class=@var{class-name}
2392 @opindex fconstant-string-class
2393 Use @var{class-name} as the name of the class to instantiate for each
2394 literal string specified with the syntax @code{@@"@dots{}"}. The default
2395 class name is @code{NXConstantString} if the GNU runtime is being used, and
2396 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2397 @option{-fconstant-cfstrings} option, if also present, will override the
2398 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2399 to be laid out as constant CoreFoundation strings.
2402 @opindex fgnu-runtime
2403 Generate object code compatible with the standard GNU Objective-C
2404 runtime. This is the default for most types of systems.
2406 @item -fnext-runtime
2407 @opindex fnext-runtime
2408 Generate output compatible with the NeXT runtime. This is the default
2409 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2410 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2413 @item -fno-nil-receivers
2414 @opindex fno-nil-receivers
2415 Assume that all Objective-C message dispatches (e.g.,
2416 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
2417 is not @code{nil}. This allows for more efficient entry points in the runtime
2418 to be used. Currently, this option is only available in conjunction with
2419 the NeXT runtime on Mac OS X 10.3 and later.
2421 @item -fobjc-call-cxx-cdtors
2422 @opindex fobjc-call-cxx-cdtors
2423 For each Objective-C class, check if any of its instance variables is a
2424 C++ object with a non-trivial default constructor. If so, synthesize a
2425 special @code{- (id) .cxx_construct} instance method that will run
2426 non-trivial default constructors on any such instance variables, in order,
2427 and then return @code{self}. Similarly, check if any instance variable
2428 is a C++ object with a non-trivial destructor, and if so, synthesize a
2429 special @code{- (void) .cxx_destruct} method that will run
2430 all such default destructors, in reverse order.
2432 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
2433 thusly generated will only operate on instance variables declared in the
2434 current Objective-C class, and not those inherited from superclasses. It
2435 is the responsibility of the Objective-C runtime to invoke all such methods
2436 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
2437 will be invoked by the runtime immediately after a new object
2438 instance is allocated; the @code{- (void) .cxx_destruct} methods will
2439 be invoked immediately before the runtime deallocates an object instance.
2441 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2442 support for invoking the @code{- (id) .cxx_construct} and
2443 @code{- (void) .cxx_destruct} methods.
2445 @item -fobjc-direct-dispatch
2446 @opindex fobjc-direct-dispatch
2447 Allow fast jumps to the message dispatcher. On Darwin this is
2448 accomplished via the comm page.
2450 @item -fobjc-exceptions
2451 @opindex fobjc-exceptions
2452 Enable syntactic support for structured exception handling in Objective-C,
2453 similar to what is offered by C++ and Java. This option is
2454 unavailable in conjunction with the NeXT runtime on Mac OS X 10.2 and
2463 @@catch (AnObjCClass *exc) @{
2470 @@catch (AnotherClass *exc) @{
2473 @@catch (id allOthers) @{
2483 The @code{@@throw} statement may appear anywhere in an Objective-C or
2484 Objective-C++ program; when used inside of a @code{@@catch} block, the
2485 @code{@@throw} may appear without an argument (as shown above), in which case
2486 the object caught by the @code{@@catch} will be rethrown.
2488 Note that only (pointers to) Objective-C objects may be thrown and
2489 caught using this scheme. When an object is thrown, it will be caught
2490 by the nearest @code{@@catch} clause capable of handling objects of that type,
2491 analogously to how @code{catch} blocks work in C++ and Java. A
2492 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2493 any and all Objective-C exceptions not caught by previous @code{@@catch}
2496 The @code{@@finally} clause, if present, will be executed upon exit from the
2497 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2498 regardless of whether any exceptions are thrown, caught or rethrown
2499 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2500 of the @code{finally} clause in Java.
2502 There are several caveats to using the new exception mechanism:
2506 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2507 idioms provided by the @code{NSException} class, the new
2508 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2509 systems, due to additional functionality needed in the (NeXT) Objective-C
2513 As mentioned above, the new exceptions do not support handling
2514 types other than Objective-C objects. Furthermore, when used from
2515 Objective-C++, the Objective-C exception model does not interoperate with C++
2516 exceptions at this time. This means you cannot @code{@@throw} an exception
2517 from Objective-C and @code{catch} it in C++, or vice versa
2518 (i.e., @code{throw @dots{} @@catch}).
2521 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2522 blocks for thread-safe execution:
2525 @@synchronized (ObjCClass *guard) @{
2530 Upon entering the @code{@@synchronized} block, a thread of execution shall
2531 first check whether a lock has been placed on the corresponding @code{guard}
2532 object by another thread. If it has, the current thread shall wait until
2533 the other thread relinquishes its lock. Once @code{guard} becomes available,
2534 the current thread will place its own lock on it, execute the code contained in
2535 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2536 making @code{guard} available to other threads).
2538 Unlike Java, Objective-C does not allow for entire methods to be marked
2539 @code{@@synchronized}. Note that throwing exceptions out of
2540 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2541 to be unlocked properly.
2545 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2547 @item -freplace-objc-classes
2548 @opindex freplace-objc-classes
2549 Emit a special marker instructing @command{ld(1)} not to statically link in
2550 the resulting object file, and allow @command{dyld(1)} to load it in at
2551 run time instead. This is used in conjunction with the Fix-and-Continue
2552 debugging mode, where the object file in question may be recompiled and
2553 dynamically reloaded in the course of program execution, without the need
2554 to restart the program itself. Currently, Fix-and-Continue functionality
2555 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2560 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2561 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2562 compile time) with static class references that get initialized at load time,
2563 which improves run-time performance. Specifying the @option{-fzero-link} flag
2564 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2565 to be retained. This is useful in Zero-Link debugging mode, since it allows
2566 for individual class implementations to be modified during program execution.
2570 Dump interface declarations for all classes seen in the source file to a
2571 file named @file{@var{sourcename}.decl}.
2573 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2574 @opindex Wassign-intercept
2575 @opindex Wno-assign-intercept
2576 Warn whenever an Objective-C assignment is being intercepted by the
2579 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2580 @opindex Wno-protocol
2582 If a class is declared to implement a protocol, a warning is issued for
2583 every method in the protocol that is not implemented by the class. The
2584 default behavior is to issue a warning for every method not explicitly
2585 implemented in the class, even if a method implementation is inherited
2586 from the superclass. If you use the @option{-Wno-protocol} option, then
2587 methods inherited from the superclass are considered to be implemented,
2588 and no warning is issued for them.
2590 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2592 @opindex Wno-selector
2593 Warn if multiple methods of different types for the same selector are
2594 found during compilation. The check is performed on the list of methods
2595 in the final stage of compilation. Additionally, a check is performed
2596 for each selector appearing in a @code{@@selector(@dots{})}
2597 expression, and a corresponding method for that selector has been found
2598 during compilation. Because these checks scan the method table only at
2599 the end of compilation, these warnings are not produced if the final
2600 stage of compilation is not reached, for example because an error is
2601 found during compilation, or because the @option{-fsyntax-only} option is
2604 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2605 @opindex Wstrict-selector-match
2606 @opindex Wno-strict-selector-match
2607 Warn if multiple methods with differing argument and/or return types are
2608 found for a given selector when attempting to send a message using this
2609 selector to a receiver of type @code{id} or @code{Class}. When this flag
2610 is off (which is the default behavior), the compiler will omit such warnings
2611 if any differences found are confined to types which share the same size
2614 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2615 @opindex Wundeclared-selector
2616 @opindex Wno-undeclared-selector
2617 Warn if a @code{@@selector(@dots{})} expression referring to an
2618 undeclared selector is found. A selector is considered undeclared if no
2619 method with that name has been declared before the
2620 @code{@@selector(@dots{})} expression, either explicitly in an
2621 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2622 an @code{@@implementation} section. This option always performs its
2623 checks as soon as a @code{@@selector(@dots{})} expression is found,
2624 while @option{-Wselector} only performs its checks in the final stage of
2625 compilation. This also enforces the coding style convention
2626 that methods and selectors must be declared before being used.
2628 @item -print-objc-runtime-info
2629 @opindex print-objc-runtime-info
2630 Generate C header describing the largest structure that is passed by
2635 @node Language Independent Options
2636 @section Options to Control Diagnostic Messages Formatting
2637 @cindex options to control diagnostics formatting
2638 @cindex diagnostic messages
2639 @cindex message formatting
2641 Traditionally, diagnostic messages have been formatted irrespective of
2642 the output device's aspect (e.g.@: its width, @dots{}). The options described
2643 below can be used to control the diagnostic messages formatting
2644 algorithm, e.g.@: how many characters per line, how often source location
2645 information should be reported. Right now, only the C++ front end can
2646 honor these options. However it is expected, in the near future, that
2647 the remaining front ends would be able to digest them correctly.
2650 @item -fmessage-length=@var{n}
2651 @opindex fmessage-length
2652 Try to format error messages so that they fit on lines of about @var{n}
2653 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2654 the front ends supported by GCC@. If @var{n} is zero, then no
2655 line-wrapping will be done; each error message will appear on a single
2658 @opindex fdiagnostics-show-location
2659 @item -fdiagnostics-show-location=once
2660 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2661 reporter to emit @emph{once} source location information; that is, in
2662 case the message is too long to fit on a single physical line and has to
2663 be wrapped, the source location won't be emitted (as prefix) again,
2664 over and over, in subsequent continuation lines. This is the default
2667 @item -fdiagnostics-show-location=every-line
2668 Only meaningful in line-wrapping mode. Instructs the diagnostic
2669 messages reporter to emit the same source location information (as
2670 prefix) for physical lines that result from the process of breaking
2671 a message which is too long to fit on a single line.
2673 @item -fdiagnostics-show-option
2674 @opindex fdiagnostics-show-option
2675 This option instructs the diagnostic machinery to add text to each
2676 diagnostic emitted, which indicates which command line option directly
2677 controls that diagnostic, when such an option is known to the
2678 diagnostic machinery.
2680 @item -Wcoverage-mismatch
2681 @opindex Wcoverage-mismatch
2682 Warn if feedback profiles do not match when using the
2683 @option{-fprofile-use} option.
2684 If a source file was changed between @option{-fprofile-gen} and
2685 @option{-fprofile-use}, the files with the profile feedback can fail
2686 to match the source file and GCC can not use the profile feedback
2687 information. By default, GCC emits an error message in this case.
2688 The option @option{-Wcoverage-mismatch} emits a warning instead of an
2689 error. GCC does not use appropriate feedback profiles, so using this
2690 option can result in poorly optimized code. This option is useful
2691 only in the case of very minor changes such as bug fixes to an
2696 @node Warning Options
2697 @section Options to Request or Suppress Warnings
2698 @cindex options to control warnings
2699 @cindex warning messages
2700 @cindex messages, warning
2701 @cindex suppressing warnings
2703 Warnings are diagnostic messages that report constructions which
2704 are not inherently erroneous but which are risky or suggest there
2705 may have been an error.
2707 The following language-independent options do not enable specific
2708 warnings but control the kinds of diagnostics produced by GCC.
2711 @cindex syntax checking
2713 @opindex fsyntax-only
2714 Check the code for syntax errors, but don't do anything beyond that.
2718 Inhibit all warning messages.
2723 Make all warnings into errors.
2728 Make the specified warning into an error. The specifier for a warning
2729 is appended, for example @option{-Werror=switch} turns the warnings
2730 controlled by @option{-Wswitch} into errors. This switch takes a
2731 negative form, to be used to negate @option{-Werror} for specific
2732 warnings, for example @option{-Wno-error=switch} makes
2733 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2734 is in effect. You can use the @option{-fdiagnostics-show-option}
2735 option to have each controllable warning amended with the option which
2736 controls it, to determine what to use with this option.
2738 Note that specifying @option{-Werror=}@var{foo} automatically implies
2739 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2742 @item -Wfatal-errors
2743 @opindex Wfatal-errors
2744 @opindex Wno-fatal-errors
2745 This option causes the compiler to abort compilation on the first error
2746 occurred rather than trying to keep going and printing further error
2751 You can request many specific warnings with options beginning
2752 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2753 implicit declarations. Each of these specific warning options also
2754 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2755 example, @option{-Wno-implicit}. This manual lists only one of the
2756 two forms, whichever is not the default. For further,
2757 language-specific options also refer to @ref{C++ Dialect Options} and
2758 @ref{Objective-C and Objective-C++ Dialect Options}.
2763 Issue all the warnings demanded by strict ISO C and ISO C++;
2764 reject all programs that use forbidden extensions, and some other
2765 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2766 version of the ISO C standard specified by any @option{-std} option used.
2768 Valid ISO C and ISO C++ programs should compile properly with or without
2769 this option (though a rare few will require @option{-ansi} or a
2770 @option{-std} option specifying the required version of ISO C)@. However,
2771 without this option, certain GNU extensions and traditional C and C++
2772 features are supported as well. With this option, they are rejected.
2774 @option{-pedantic} does not cause warning messages for use of the
2775 alternate keywords whose names begin and end with @samp{__}. Pedantic
2776 warnings are also disabled in the expression that follows
2777 @code{__extension__}. However, only system header files should use
2778 these escape routes; application programs should avoid them.
2779 @xref{Alternate Keywords}.
2781 Some users try to use @option{-pedantic} to check programs for strict ISO
2782 C conformance. They soon find that it does not do quite what they want:
2783 it finds some non-ISO practices, but not all---only those for which
2784 ISO C @emph{requires} a diagnostic, and some others for which
2785 diagnostics have been added.
2787 A feature to report any failure to conform to ISO C might be useful in
2788 some instances, but would require considerable additional work and would
2789 be quite different from @option{-pedantic}. We don't have plans to
2790 support such a feature in the near future.
2792 Where the standard specified with @option{-std} represents a GNU
2793 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2794 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2795 extended dialect is based. Warnings from @option{-pedantic} are given
2796 where they are required by the base standard. (It would not make sense
2797 for such warnings to be given only for features not in the specified GNU
2798 C dialect, since by definition the GNU dialects of C include all
2799 features the compiler supports with the given option, and there would be
2800 nothing to warn about.)
2802 @item -pedantic-errors
2803 @opindex pedantic-errors
2804 Like @option{-pedantic}, except that errors are produced rather than
2810 This enables all the warnings about constructions that some users
2811 consider questionable, and that are easy to avoid (or modify to
2812 prevent the warning), even in conjunction with macros. This also
2813 enables some language-specific warnings described in @ref{C++ Dialect
2814 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2816 @option{-Wall} turns on the following warning flags:
2818 @gccoptlist{-Waddress @gol
2819 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2821 -Wchar-subscripts @gol
2822 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
2824 -Wimplicit-function-declaration @gol
2827 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2828 -Wmissing-braces @gol
2834 -Wsequence-point @gol
2835 -Wsign-compare @r{(only in C++)} @gol
2836 -Wstrict-aliasing @gol
2837 -Wstrict-overflow=1 @gol
2840 -Wuninitialized @gol
2841 -Wunknown-pragmas @gol
2842 -Wunused-function @gol
2845 -Wunused-variable @gol
2846 -Wvolatile-register-var @gol
2849 Note that some warning flags are not implied by @option{-Wall}. Some of
2850 them warn about constructions that users generally do not consider
2851 questionable, but which occasionally you might wish to check for;
2852 others warn about constructions that are necessary or hard to avoid in
2853 some cases, and there is no simple way to modify the code to suppress
2854 the warning. Some of them are enabled by @option{-Wextra} but many of
2855 them must be enabled individually.
2861 This enables some extra warning flags that are not enabled by
2862 @option{-Wall}. (This option used to be called @option{-W}. The older
2863 name is still supported, but the newer name is more descriptive.)
2865 @gccoptlist{-Wclobbered @gol
2867 -Wignored-qualifiers @gol
2868 -Wmissing-field-initializers @gol
2869 -Wmissing-parameter-type @r{(C only)} @gol
2870 -Wold-style-declaration @r{(C only)} @gol
2871 -Woverride-init @gol
2874 -Wuninitialized @gol
2875 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2878 The option @option{-Wextra} also prints warning messages for the
2884 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2885 @samp{>}, or @samp{>=}.
2888 (C++ only) An enumerator and a non-enumerator both appear in a
2889 conditional expression.
2892 (C++ only) Ambiguous virtual bases.
2895 (C++ only) Subscripting an array which has been declared @samp{register}.
2898 (C++ only) Taking the address of a variable which has been declared
2902 (C++ only) A base class is not initialized in a derived class' copy
2907 @item -Wchar-subscripts
2908 @opindex Wchar-subscripts
2909 @opindex Wno-char-subscripts
2910 Warn if an array subscript has type @code{char}. This is a common cause
2911 of error, as programmers often forget that this type is signed on some
2913 This warning is enabled by @option{-Wall}.
2917 @opindex Wno-comment
2918 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2919 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2920 This warning is enabled by @option{-Wall}.
2925 @opindex ffreestanding
2926 @opindex fno-builtin
2927 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2928 the arguments supplied have types appropriate to the format string
2929 specified, and that the conversions specified in the format string make
2930 sense. This includes standard functions, and others specified by format
2931 attributes (@pxref{Function Attributes}), in the @code{printf},
2932 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2933 not in the C standard) families (or other target-specific families).
2934 Which functions are checked without format attributes having been
2935 specified depends on the standard version selected, and such checks of
2936 functions without the attribute specified are disabled by
2937 @option{-ffreestanding} or @option{-fno-builtin}.
2939 The formats are checked against the format features supported by GNU
2940 libc version 2.2. These include all ISO C90 and C99 features, as well
2941 as features from the Single Unix Specification and some BSD and GNU
2942 extensions. Other library implementations may not support all these
2943 features; GCC does not support warning about features that go beyond a
2944 particular library's limitations. However, if @option{-pedantic} is used
2945 with @option{-Wformat}, warnings will be given about format features not
2946 in the selected standard version (but not for @code{strfmon} formats,
2947 since those are not in any version of the C standard). @xref{C Dialect
2948 Options,,Options Controlling C Dialect}.
2950 Since @option{-Wformat} also checks for null format arguments for
2951 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2953 @option{-Wformat} is included in @option{-Wall}. For more control over some
2954 aspects of format checking, the options @option{-Wformat-y2k},
2955 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2956 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2957 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2960 @opindex Wformat-y2k
2961 @opindex Wno-format-y2k
2962 If @option{-Wformat} is specified, also warn about @code{strftime}
2963 formats which may yield only a two-digit year.
2965 @item -Wno-format-contains-nul
2966 @opindex Wno-format-contains-nul
2967 @opindex Wformat-contains-nul
2968 If @option{-Wformat} is specified, do not warn about format strings that
2971 @item -Wno-format-extra-args
2972 @opindex Wno-format-extra-args
2973 @opindex Wformat-extra-args
2974 If @option{-Wformat} is specified, do not warn about excess arguments to a
2975 @code{printf} or @code{scanf} format function. The C standard specifies
2976 that such arguments are ignored.
2978 Where the unused arguments lie between used arguments that are
2979 specified with @samp{$} operand number specifications, normally
2980 warnings are still given, since the implementation could not know what
2981 type to pass to @code{va_arg} to skip the unused arguments. However,
2982 in the case of @code{scanf} formats, this option will suppress the
2983 warning if the unused arguments are all pointers, since the Single
2984 Unix Specification says that such unused arguments are allowed.
2986 @item -Wno-format-zero-length @r{(C and Objective-C only)}
2987 @opindex Wno-format-zero-length
2988 @opindex Wformat-zero-length
2989 If @option{-Wformat} is specified, do not warn about zero-length formats.
2990 The C standard specifies that zero-length formats are allowed.
2992 @item -Wformat-nonliteral
2993 @opindex Wformat-nonliteral
2994 @opindex Wno-format-nonliteral
2995 If @option{-Wformat} is specified, also warn if the format string is not a
2996 string literal and so cannot be checked, unless the format function
2997 takes its format arguments as a @code{va_list}.
2999 @item -Wformat-security
3000 @opindex Wformat-security
3001 @opindex Wno-format-security
3002 If @option{-Wformat} is specified, also warn about uses of format
3003 functions that represent possible security problems. At present, this
3004 warns about calls to @code{printf} and @code{scanf} functions where the
3005 format string is not a string literal and there are no format arguments,
3006 as in @code{printf (foo);}. This may be a security hole if the format
3007 string came from untrusted input and contains @samp{%n}. (This is
3008 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3009 in future warnings may be added to @option{-Wformat-security} that are not
3010 included in @option{-Wformat-nonliteral}.)
3014 @opindex Wno-format=2
3015 Enable @option{-Wformat} plus format checks not included in
3016 @option{-Wformat}. Currently equivalent to @samp{-Wformat
3017 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
3019 @item -Wnonnull @r{(C and Objective-C only)}
3021 @opindex Wno-nonnull
3022 Warn about passing a null pointer for arguments marked as
3023 requiring a non-null value by the @code{nonnull} function attribute.
3025 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3026 can be disabled with the @option{-Wno-nonnull} option.
3028 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3030 @opindex Wno-init-self
3031 Warn about uninitialized variables which are initialized with themselves.
3032 Note this option can only be used with the @option{-Wuninitialized} option.
3034 For example, GCC will warn about @code{i} being uninitialized in the
3035 following snippet only when @option{-Winit-self} has been specified:
3046 @item -Wimplicit-int @r{(C and Objective-C only)}
3047 @opindex Wimplicit-int
3048 @opindex Wno-implicit-int
3049 Warn when a declaration does not specify a type.
3050 This warning is enabled by @option{-Wall}.
3052 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3053 @opindex Wimplicit-function-declaration
3054 @opindex Wno-implicit-function-declaration
3055 Give a warning whenever a function is used before being declared. In
3056 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3057 enabled by default and it is made into an error by
3058 @option{-pedantic-errors}. This warning is also enabled by
3063 @opindex Wno-implicit
3064 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3065 This warning is enabled by @option{-Wall}.
3067 @item -Wignored-qualifiers @r{(C and C++ only)}
3068 @opindex Wignored-qualifiers
3069 @opindex Wno-ignored-qualifiers
3070 Warn if the return type of a function has a type qualifier
3071 such as @code{const}. For ISO C such a type qualifier has no effect,
3072 since the value returned by a function is not an lvalue.
3073 For C++, the warning is only emitted for scalar types or @code{void}.
3074 ISO C prohibits qualified @code{void} return types on function
3075 definitions, so such return types always receive a warning
3076 even without this option.
3078 This warning is also enabled by @option{-Wextra}.
3083 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3084 a function with external linkage, returning int, taking either zero
3085 arguments, two, or three arguments of appropriate types. This warning
3086 is enabled by default in C++ and is enabled by either @option{-Wall}
3087 or @option{-pedantic}.
3089 @item -Wmissing-braces
3090 @opindex Wmissing-braces
3091 @opindex Wno-missing-braces
3092 Warn if an aggregate or union initializer is not fully bracketed. In
3093 the following example, the initializer for @samp{a} is not fully
3094 bracketed, but that for @samp{b} is fully bracketed.
3097 int a[2][2] = @{ 0, 1, 2, 3 @};
3098 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3101 This warning is enabled by @option{-Wall}.
3103 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3104 @opindex Wmissing-include-dirs
3105 @opindex Wno-missing-include-dirs
3106 Warn if a user-supplied include directory does not exist.
3109 @opindex Wparentheses
3110 @opindex Wno-parentheses
3111 Warn if parentheses are omitted in certain contexts, such
3112 as when there is an assignment in a context where a truth value
3113 is expected, or when operators are nested whose precedence people
3114 often get confused about.
3116 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3117 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3118 interpretation from that of ordinary mathematical notation.
3120 Also warn about constructions where there may be confusion to which
3121 @code{if} statement an @code{else} branch belongs. Here is an example of
3136 In C/C++, every @code{else} branch belongs to the innermost possible
3137 @code{if} statement, which in this example is @code{if (b)}. This is
3138 often not what the programmer expected, as illustrated in the above
3139 example by indentation the programmer chose. When there is the
3140 potential for this confusion, GCC will issue a warning when this flag
3141 is specified. To eliminate the warning, add explicit braces around
3142 the innermost @code{if} statement so there is no way the @code{else}
3143 could belong to the enclosing @code{if}. The resulting code would
3160 This warning is enabled by @option{-Wall}.
3162 @item -Wsequence-point
3163 @opindex Wsequence-point
3164 @opindex Wno-sequence-point
3165 Warn about code that may have undefined semantics because of violations
3166 of sequence point rules in the C and C++ standards.
3168 The C and C++ standards defines the order in which expressions in a C/C++
3169 program are evaluated in terms of @dfn{sequence points}, which represent
3170 a partial ordering between the execution of parts of the program: those
3171 executed before the sequence point, and those executed after it. These
3172 occur after the evaluation of a full expression (one which is not part
3173 of a larger expression), after the evaluation of the first operand of a
3174 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3175 function is called (but after the evaluation of its arguments and the
3176 expression denoting the called function), and in certain other places.
3177 Other than as expressed by the sequence point rules, the order of
3178 evaluation of subexpressions of an expression is not specified. All
3179 these rules describe only a partial order rather than a total order,
3180 since, for example, if two functions are called within one expression
3181 with no sequence point between them, the order in which the functions
3182 are called is not specified. However, the standards committee have
3183 ruled that function calls do not overlap.
3185 It is not specified when between sequence points modifications to the
3186 values of objects take effect. Programs whose behavior depends on this
3187 have undefined behavior; the C and C++ standards specify that ``Between
3188 the previous and next sequence point an object shall have its stored
3189 value modified at most once by the evaluation of an expression.
3190 Furthermore, the prior value shall be read only to determine the value
3191 to be stored.''. If a program breaks these rules, the results on any
3192 particular implementation are entirely unpredictable.
3194 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3195 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3196 diagnosed by this option, and it may give an occasional false positive
3197 result, but in general it has been found fairly effective at detecting
3198 this sort of problem in programs.
3200 The standard is worded confusingly, therefore there is some debate
3201 over the precise meaning of the sequence point rules in subtle cases.
3202 Links to discussions of the problem, including proposed formal
3203 definitions, may be found on the GCC readings page, at
3204 @w{@uref{http://gcc.gnu.org/readings.html}}.
3206 This warning is enabled by @option{-Wall} for C and C++.
3209 @opindex Wreturn-type
3210 @opindex Wno-return-type
3211 Warn whenever a function is defined with a return-type that defaults
3212 to @code{int}. Also warn about any @code{return} statement with no
3213 return-value in a function whose return-type is not @code{void}
3214 (falling off the end of the function body is considered returning
3215 without a value), and about a @code{return} statement with an
3216 expression in a function whose return-type is @code{void}.
3218 For C++, a function without return type always produces a diagnostic
3219 message, even when @option{-Wno-return-type} is specified. The only
3220 exceptions are @samp{main} and functions defined in system headers.
3222 This warning is enabled by @option{-Wall}.
3227 Warn whenever a @code{switch} statement has an index of enumerated type
3228 and lacks a @code{case} for one or more of the named codes of that
3229 enumeration. (The presence of a @code{default} label prevents this
3230 warning.) @code{case} labels outside the enumeration range also
3231 provoke warnings when this option is used (even if there is a
3232 @code{default} label).
3233 This warning is enabled by @option{-Wall}.
3235 @item -Wswitch-default
3236 @opindex Wswitch-default
3237 @opindex Wno-switch-default
3238 Warn whenever a @code{switch} statement does not have a @code{default}
3242 @opindex Wswitch-enum
3243 @opindex Wno-switch-enum
3244 Warn whenever a @code{switch} statement has an index of enumerated type
3245 and lacks a @code{case} for one or more of the named codes of that
3246 enumeration. @code{case} labels outside the enumeration range also
3247 provoke warnings when this option is used. The only difference
3248 between @option{-Wswitch} and this option is that this option gives a
3249 warning about an omitted enumeration code even if there is a
3250 @code{default} label.
3252 @item -Wsync-nand @r{(C and C++ only)}
3254 @opindex Wno-sync-nand
3255 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3256 built-in functions are used. These functions changed semantics in GCC 4.4.
3260 @opindex Wno-trigraphs
3261 Warn if any trigraphs are encountered that might change the meaning of
3262 the program (trigraphs within comments are not warned about).
3263 This warning is enabled by @option{-Wall}.
3265 @item -Wunused-function
3266 @opindex Wunused-function
3267 @opindex Wno-unused-function
3268 Warn whenever a static function is declared but not defined or a
3269 non-inline static function is unused.
3270 This warning is enabled by @option{-Wall}.
3272 @item -Wunused-label
3273 @opindex Wunused-label
3274 @opindex Wno-unused-label
3275 Warn whenever a label is declared but not used.
3276 This warning is enabled by @option{-Wall}.
3278 To suppress this warning use the @samp{unused} attribute
3279 (@pxref{Variable Attributes}).
3281 @item -Wunused-parameter
3282 @opindex Wunused-parameter
3283 @opindex Wno-unused-parameter
3284 Warn whenever a function parameter is unused aside from its declaration.
3286 To suppress this warning use the @samp{unused} attribute
3287 (@pxref{Variable Attributes}).
3289 @item -Wno-unused-result
3290 @opindex Wunused-result
3291 @opindex Wno-unused-result
3292 Do not warn if a caller of a function marked with attribute
3293 @code{warn_unused_result} (@pxref{Variable Attributes}) does not use
3294 its return value. The default is @option{-Wunused-result}.
3296 @item -Wunused-variable
3297 @opindex Wunused-variable
3298 @opindex Wno-unused-variable
3299 Warn whenever a local variable or non-constant static variable is unused
3300 aside from its declaration.
3301 This warning is enabled by @option{-Wall}.
3303 To suppress this warning use the @samp{unused} attribute
3304 (@pxref{Variable Attributes}).
3306 @item -Wunused-value
3307 @opindex Wunused-value
3308 @opindex Wno-unused-value
3309 Warn whenever a statement computes a result that is explicitly not
3310 used. To suppress this warning cast the unused expression to
3311 @samp{void}. This includes an expression-statement or the left-hand
3312 side of a comma expression that contains no side effects. For example,
3313 an expression such as @samp{x[i,j]} will cause a warning, while
3314 @samp{x[(void)i,j]} will not.
3316 This warning is enabled by @option{-Wall}.
3321 All the above @option{-Wunused} options combined.
3323 In order to get a warning about an unused function parameter, you must
3324 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3325 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3327 @item -Wuninitialized
3328 @opindex Wuninitialized
3329 @opindex Wno-uninitialized
3330 Warn if an automatic variable is used without first being initialized
3331 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3332 warn if a non-static reference or non-static @samp{const} member
3333 appears in a class without constructors.
3335 If you want to warn about code which uses the uninitialized value of the
3336 variable in its own initializer, use the @option{-Winit-self} option.
3338 These warnings occur for individual uninitialized or clobbered
3339 elements of structure, union or array variables as well as for
3340 variables which are uninitialized or clobbered as a whole. They do
3341 not occur for variables or elements declared @code{volatile}. Because
3342 these warnings depend on optimization, the exact variables or elements
3343 for which there are warnings will depend on the precise optimization
3344 options and version of GCC used.
3346 Note that there may be no warning about a variable that is used only
3347 to compute a value that itself is never used, because such
3348 computations may be deleted by data flow analysis before the warnings
3351 These warnings are made optional because GCC is not smart
3352 enough to see all the reasons why the code might be correct
3353 despite appearing to have an error. Here is one example of how
3374 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3375 always initialized, but GCC doesn't know this. Here is
3376 another common case:
3381 if (change_y) save_y = y, y = new_y;
3383 if (change_y) y = save_y;
3388 This has no bug because @code{save_y} is used only if it is set.
3390 @cindex @code{longjmp} warnings
3391 This option also warns when a non-volatile automatic variable might be
3392 changed by a call to @code{longjmp}. These warnings as well are possible
3393 only in optimizing compilation.
3395 The compiler sees only the calls to @code{setjmp}. It cannot know
3396 where @code{longjmp} will be called; in fact, a signal handler could
3397 call it at any point in the code. As a result, you may get a warning
3398 even when there is in fact no problem because @code{longjmp} cannot
3399 in fact be called at the place which would cause a problem.
3401 Some spurious warnings can be avoided if you declare all the functions
3402 you use that never return as @code{noreturn}. @xref{Function
3405 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3407 @item -Wunknown-pragmas
3408 @opindex Wunknown-pragmas
3409 @opindex Wno-unknown-pragmas
3410 @cindex warning for unknown pragmas
3411 @cindex unknown pragmas, warning
3412 @cindex pragmas, warning of unknown
3413 Warn when a #pragma directive is encountered which is not understood by
3414 GCC@. If this command line option is used, warnings will even be issued
3415 for unknown pragmas in system header files. This is not the case if
3416 the warnings were only enabled by the @option{-Wall} command line option.
3419 @opindex Wno-pragmas
3421 Do not warn about misuses of pragmas, such as incorrect parameters,
3422 invalid syntax, or conflicts between pragmas. See also
3423 @samp{-Wunknown-pragmas}.
3425 @item -Wstrict-aliasing
3426 @opindex Wstrict-aliasing
3427 @opindex Wno-strict-aliasing
3428 This option is only active when @option{-fstrict-aliasing} is active.
3429 It warns about code which might break the strict aliasing rules that the
3430 compiler is using for optimization. The warning does not catch all
3431 cases, but does attempt to catch the more common pitfalls. It is
3432 included in @option{-Wall}.
3433 It is equivalent to @option{-Wstrict-aliasing=3}
3435 @item -Wstrict-aliasing=n
3436 @opindex Wstrict-aliasing=n
3437 @opindex Wno-strict-aliasing=n
3438 This option is only active when @option{-fstrict-aliasing} is active.
3439 It warns about code which might break the strict aliasing rules that the
3440 compiler is using for optimization.
3441 Higher levels correspond to higher accuracy (fewer false positives).
3442 Higher levels also correspond to more effort, similar to the way -O works.
3443 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3446 Level 1: Most aggressive, quick, least accurate.
3447 Possibly useful when higher levels
3448 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3449 false negatives. However, it has many false positives.
3450 Warns for all pointer conversions between possibly incompatible types,
3451 even if never dereferenced. Runs in the frontend only.
3453 Level 2: Aggressive, quick, not too precise.
3454 May still have many false positives (not as many as level 1 though),
3455 and few false negatives (but possibly more than level 1).
3456 Unlike level 1, it only warns when an address is taken. Warns about
3457 incomplete types. Runs in the frontend only.
3459 Level 3 (default for @option{-Wstrict-aliasing}):
3460 Should have very few false positives and few false
3461 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3462 Takes care of the common punn+dereference pattern in the frontend:
3463 @code{*(int*)&some_float}.
3464 If optimization is enabled, it also runs in the backend, where it deals
3465 with multiple statement cases using flow-sensitive points-to information.
3466 Only warns when the converted pointer is dereferenced.
3467 Does not warn about incomplete types.
3469 @item -Wstrict-overflow
3470 @itemx -Wstrict-overflow=@var{n}
3471 @opindex Wstrict-overflow
3472 @opindex Wno-strict-overflow
3473 This option is only active when @option{-fstrict-overflow} is active.
3474 It warns about cases where the compiler optimizes based on the
3475 assumption that signed overflow does not occur. Note that it does not
3476 warn about all cases where the code might overflow: it only warns
3477 about cases where the compiler implements some optimization. Thus
3478 this warning depends on the optimization level.
3480 An optimization which assumes that signed overflow does not occur is
3481 perfectly safe if the values of the variables involved are such that
3482 overflow never does, in fact, occur. Therefore this warning can
3483 easily give a false positive: a warning about code which is not
3484 actually a problem. To help focus on important issues, several
3485 warning levels are defined. No warnings are issued for the use of
3486 undefined signed overflow when estimating how many iterations a loop
3487 will require, in particular when determining whether a loop will be
3491 @item -Wstrict-overflow=1
3492 Warn about cases which are both questionable and easy to avoid. For
3493 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3494 compiler will simplify this to @code{1}. This level of
3495 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3496 are not, and must be explicitly requested.
3498 @item -Wstrict-overflow=2
3499 Also warn about other cases where a comparison is simplified to a
3500 constant. For example: @code{abs (x) >= 0}. This can only be
3501 simplified when @option{-fstrict-overflow} is in effect, because
3502 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3503 zero. @option{-Wstrict-overflow} (with no level) is the same as
3504 @option{-Wstrict-overflow=2}.
3506 @item -Wstrict-overflow=3
3507 Also warn about other cases where a comparison is simplified. For
3508 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3510 @item -Wstrict-overflow=4
3511 Also warn about other simplifications not covered by the above cases.
3512 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3514 @item -Wstrict-overflow=5
3515 Also warn about cases where the compiler reduces the magnitude of a
3516 constant involved in a comparison. For example: @code{x + 2 > y} will
3517 be simplified to @code{x + 1 >= y}. This is reported only at the
3518 highest warning level because this simplification applies to many
3519 comparisons, so this warning level will give a very large number of
3523 @item -Warray-bounds
3524 @opindex Wno-array-bounds
3525 @opindex Warray-bounds
3526 This option is only active when @option{-ftree-vrp} is active
3527 (default for -O2 and above). It warns about subscripts to arrays
3528 that are always out of bounds. This warning is enabled by @option{-Wall}.
3530 @item -Wno-div-by-zero
3531 @opindex Wno-div-by-zero
3532 @opindex Wdiv-by-zero
3533 Do not warn about compile-time integer division by zero. Floating point
3534 division by zero is not warned about, as it can be a legitimate way of
3535 obtaining infinities and NaNs.
3537 @item -Wsystem-headers
3538 @opindex Wsystem-headers
3539 @opindex Wno-system-headers
3540 @cindex warnings from system headers
3541 @cindex system headers, warnings from
3542 Print warning messages for constructs found in system header files.
3543 Warnings from system headers are normally suppressed, on the assumption
3544 that they usually do not indicate real problems and would only make the
3545 compiler output harder to read. Using this command line option tells
3546 GCC to emit warnings from system headers as if they occurred in user
3547 code. However, note that using @option{-Wall} in conjunction with this
3548 option will @emph{not} warn about unknown pragmas in system
3549 headers---for that, @option{-Wunknown-pragmas} must also be used.
3552 @opindex Wfloat-equal
3553 @opindex Wno-float-equal
3554 Warn if floating point values are used in equality comparisons.
3556 The idea behind this is that sometimes it is convenient (for the
3557 programmer) to consider floating-point values as approximations to
3558 infinitely precise real numbers. If you are doing this, then you need
3559 to compute (by analyzing the code, or in some other way) the maximum or
3560 likely maximum error that the computation introduces, and allow for it
3561 when performing comparisons (and when producing output, but that's a
3562 different problem). In particular, instead of testing for equality, you
3563 would check to see whether the two values have ranges that overlap; and
3564 this is done with the relational operators, so equality comparisons are
3567 @item -Wtraditional @r{(C and Objective-C only)}
3568 @opindex Wtraditional
3569 @opindex Wno-traditional
3570 Warn about certain constructs that behave differently in traditional and
3571 ISO C@. Also warn about ISO C constructs that have no traditional C
3572 equivalent, and/or problematic constructs which should be avoided.
3576 Macro parameters that appear within string literals in the macro body.
3577 In traditional C macro replacement takes place within string literals,
3578 but does not in ISO C@.
3581 In traditional C, some preprocessor directives did not exist.
3582 Traditional preprocessors would only consider a line to be a directive
3583 if the @samp{#} appeared in column 1 on the line. Therefore
3584 @option{-Wtraditional} warns about directives that traditional C
3585 understands but would ignore because the @samp{#} does not appear as the
3586 first character on the line. It also suggests you hide directives like
3587 @samp{#pragma} not understood by traditional C by indenting them. Some
3588 traditional implementations would not recognize @samp{#elif}, so it
3589 suggests avoiding it altogether.
3592 A function-like macro that appears without arguments.
3595 The unary plus operator.
3598 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3599 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3600 constants.) Note, these suffixes appear in macros defined in the system
3601 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3602 Use of these macros in user code might normally lead to spurious
3603 warnings, however GCC's integrated preprocessor has enough context to
3604 avoid warning in these cases.
3607 A function declared external in one block and then used after the end of
3611 A @code{switch} statement has an operand of type @code{long}.
3614 A non-@code{static} function declaration follows a @code{static} one.
3615 This construct is not accepted by some traditional C compilers.
3618 The ISO type of an integer constant has a different width or
3619 signedness from its traditional type. This warning is only issued if
3620 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3621 typically represent bit patterns, are not warned about.
3624 Usage of ISO string concatenation is detected.
3627 Initialization of automatic aggregates.
3630 Identifier conflicts with labels. Traditional C lacks a separate
3631 namespace for labels.
3634 Initialization of unions. If the initializer is zero, the warning is
3635 omitted. This is done under the assumption that the zero initializer in
3636 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3637 initializer warnings and relies on default initialization to zero in the
3641 Conversions by prototypes between fixed/floating point values and vice
3642 versa. The absence of these prototypes when compiling with traditional
3643 C would cause serious problems. This is a subset of the possible
3644 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3647 Use of ISO C style function definitions. This warning intentionally is
3648 @emph{not} issued for prototype declarations or variadic functions
3649 because these ISO C features will appear in your code when using
3650 libiberty's traditional C compatibility macros, @code{PARAMS} and
3651 @code{VPARAMS}. This warning is also bypassed for nested functions
3652 because that feature is already a GCC extension and thus not relevant to
3653 traditional C compatibility.
3656 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3657 @opindex Wtraditional-conversion
3658 @opindex Wno-traditional-conversion
3659 Warn if a prototype causes a type conversion that is different from what
3660 would happen to the same argument in the absence of a prototype. This
3661 includes conversions of fixed point to floating and vice versa, and
3662 conversions changing the width or signedness of a fixed point argument
3663 except when the same as the default promotion.
3665 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3666 @opindex Wdeclaration-after-statement
3667 @opindex Wno-declaration-after-statement
3668 Warn when a declaration is found after a statement in a block. This
3669 construct, known from C++, was introduced with ISO C99 and is by default
3670 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3671 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3676 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3678 @item -Wno-endif-labels
3679 @opindex Wno-endif-labels
3680 @opindex Wendif-labels
3681 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3686 Warn whenever a local variable shadows another local variable, parameter or
3687 global variable or whenever a built-in function is shadowed.
3689 @item -Wlarger-than=@var{len}
3690 @opindex Wlarger-than=@var{len}
3691 @opindex Wlarger-than-@var{len}
3692 Warn whenever an object of larger than @var{len} bytes is defined.
3694 @item -Wframe-larger-than=@var{len}
3695 @opindex Wframe-larger-than
3696 Warn if the size of a function frame is larger than @var{len} bytes.
3697 The computation done to determine the stack frame size is approximate
3698 and not conservative.
3699 The actual requirements may be somewhat greater than @var{len}
3700 even if you do not get a warning. In addition, any space allocated
3701 via @code{alloca}, variable-length arrays, or related constructs
3702 is not included by the compiler when determining
3703 whether or not to issue a warning.
3705 @item -Wunsafe-loop-optimizations
3706 @opindex Wunsafe-loop-optimizations
3707 @opindex Wno-unsafe-loop-optimizations
3708 Warn if the loop cannot be optimized because the compiler could not
3709 assume anything on the bounds of the loop indices. With
3710 @option{-funsafe-loop-optimizations} warn if the compiler made
3713 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
3714 @opindex Wno-pedantic-ms-format
3715 @opindex Wpedantic-ms-format
3716 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
3717 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
3718 depending on the MS runtime, when you are using the options @option{-Wformat}
3719 and @option{-pedantic} without gnu-extensions.
3721 @item -Wpointer-arith
3722 @opindex Wpointer-arith
3723 @opindex Wno-pointer-arith
3724 Warn about anything that depends on the ``size of'' a function type or
3725 of @code{void}. GNU C assigns these types a size of 1, for
3726 convenience in calculations with @code{void *} pointers and pointers
3727 to functions. In C++, warn also when an arithmetic operation involves
3728 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3731 @opindex Wtype-limits
3732 @opindex Wno-type-limits
3733 Warn if a comparison is always true or always false due to the limited
3734 range of the data type, but do not warn for constant expressions. For
3735 example, warn if an unsigned variable is compared against zero with
3736 @samp{<} or @samp{>=}. This warning is also enabled by
3739 @item -Wbad-function-cast @r{(C and Objective-C only)}
3740 @opindex Wbad-function-cast
3741 @opindex Wno-bad-function-cast
3742 Warn whenever a function call is cast to a non-matching type.
3743 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3745 @item -Wc++-compat @r{(C and Objective-C only)}
3746 Warn about ISO C constructs that are outside of the common subset of
3747 ISO C and ISO C++, e.g.@: request for implicit conversion from
3748 @code{void *} to a pointer to non-@code{void} type.
3750 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3751 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3752 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3753 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3757 @opindex Wno-cast-qual
3758 Warn whenever a pointer is cast so as to remove a type qualifier from
3759 the target type. For example, warn if a @code{const char *} is cast
3760 to an ordinary @code{char *}.
3762 Also warn when making a cast which introduces a type qualifier in an
3763 unsafe way. For example, casting @code{char **} to @code{const char **}
3764 is unsafe, as in this example:
3767 /* p is char ** value. */
3768 const char **q = (const char **) p;
3769 /* Assignment of readonly string to const char * is OK. */
3771 /* Now char** pointer points to read-only memory. */
3776 @opindex Wcast-align
3777 @opindex Wno-cast-align
3778 Warn whenever a pointer is cast such that the required alignment of the
3779 target is increased. For example, warn if a @code{char *} is cast to
3780 an @code{int *} on machines where integers can only be accessed at
3781 two- or four-byte boundaries.
3783 @item -Wwrite-strings
3784 @opindex Wwrite-strings
3785 @opindex Wno-write-strings
3786 When compiling C, give string constants the type @code{const
3787 char[@var{length}]} so that copying the address of one into a
3788 non-@code{const} @code{char *} pointer will get a warning. These
3789 warnings will help you find at compile time code that can try to write
3790 into a string constant, but only if you have been very careful about
3791 using @code{const} in declarations and prototypes. Otherwise, it will
3792 just be a nuisance. This is why we did not make @option{-Wall} request
3795 When compiling C++, warn about the deprecated conversion from string
3796 literals to @code{char *}. This warning is enabled by default for C++
3801 @opindex Wno-clobbered
3802 Warn for variables that might be changed by @samp{longjmp} or
3803 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
3806 @opindex Wconversion
3807 @opindex Wno-conversion
3808 Warn for implicit conversions that may alter a value. This includes
3809 conversions between real and integer, like @code{abs (x)} when
3810 @code{x} is @code{double}; conversions between signed and unsigned,
3811 like @code{unsigned ui = -1}; and conversions to smaller types, like
3812 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
3813 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
3814 changed by the conversion like in @code{abs (2.0)}. Warnings about
3815 conversions between signed and unsigned integers can be disabled by
3816 using @option{-Wno-sign-conversion}.
3818 For C++, also warn for conversions between @code{NULL} and non-pointer
3819 types; confusing overload resolution for user-defined conversions; and
3820 conversions that will never use a type conversion operator:
3821 conversions to @code{void}, the same type, a base class or a reference
3822 to them. Warnings about conversions between signed and unsigned
3823 integers are disabled by default in C++ unless
3824 @option{-Wsign-conversion} is explicitly enabled.
3827 @opindex Wempty-body
3828 @opindex Wno-empty-body
3829 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
3830 while} statement. This warning is also enabled by @option{-Wextra}.
3832 @item -Wenum-compare
3833 @opindex Wenum-compare
3834 @opindex Wno-enum-compare
3835 Warn about a comparison between values of different enum types. In C++
3836 this warning is enabled by default. In C this warning is enabled by
3839 @item -Wjump-misses-init @r{(C, Objective-C only)}
3840 @opindex Wjump-misses-init
3841 @opindex Wno-jump-misses-init
3842 Warn if a @code{goto} statement or a @code{switch} statement jumps
3843 forward across the initialization of a variable, or jumps backward to a
3844 label after the variable has been initialized. This only warns about
3845 variables which are initialized when they are declared. This warning is
3846 only supported for C and Objective C; in C++ this sort of branch is an
3849 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
3850 can be disabled with the @option{-Wno-jump-misses-init} option.
3852 @item -Wsign-compare
3853 @opindex Wsign-compare
3854 @opindex Wno-sign-compare
3855 @cindex warning for comparison of signed and unsigned values
3856 @cindex comparison of signed and unsigned values, warning
3857 @cindex signed and unsigned values, comparison warning
3858 Warn when a comparison between signed and unsigned values could produce
3859 an incorrect result when the signed value is converted to unsigned.
3860 This warning is also enabled by @option{-Wextra}; to get the other warnings
3861 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3863 @item -Wsign-conversion
3864 @opindex Wsign-conversion
3865 @opindex Wno-sign-conversion
3866 Warn for implicit conversions that may change the sign of an integer
3867 value, like assigning a signed integer expression to an unsigned
3868 integer variable. An explicit cast silences the warning. In C, this
3869 option is enabled also by @option{-Wconversion}.
3873 @opindex Wno-address
3874 Warn about suspicious uses of memory addresses. These include using
3875 the address of a function in a conditional expression, such as
3876 @code{void func(void); if (func)}, and comparisons against the memory
3877 address of a string literal, such as @code{if (x == "abc")}. Such
3878 uses typically indicate a programmer error: the address of a function
3879 always evaluates to true, so their use in a conditional usually
3880 indicate that the programmer forgot the parentheses in a function
3881 call; and comparisons against string literals result in unspecified
3882 behavior and are not portable in C, so they usually indicate that the
3883 programmer intended to use @code{strcmp}. This warning is enabled by
3887 @opindex Wlogical-op
3888 @opindex Wno-logical-op
3889 Warn about suspicious uses of logical operators in expressions.
3890 This includes using logical operators in contexts where a
3891 bit-wise operator is likely to be expected.
3893 @item -Waggregate-return
3894 @opindex Waggregate-return
3895 @opindex Wno-aggregate-return
3896 Warn if any functions that return structures or unions are defined or
3897 called. (In languages where you can return an array, this also elicits
3900 @item -Wno-attributes
3901 @opindex Wno-attributes
3902 @opindex Wattributes
3903 Do not warn if an unexpected @code{__attribute__} is used, such as
3904 unrecognized attributes, function attributes applied to variables,
3905 etc. This will not stop errors for incorrect use of supported
3908 @item -Wno-builtin-macro-redefined
3909 @opindex Wno-builtin-macro-redefined
3910 @opindex Wbuiltin-macro-redefined
3911 Do not warn if certain built-in macros are redefined. This suppresses
3912 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
3913 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
3915 @item -Wstrict-prototypes @r{(C and Objective-C only)}
3916 @opindex Wstrict-prototypes
3917 @opindex Wno-strict-prototypes
3918 Warn if a function is declared or defined without specifying the
3919 argument types. (An old-style function definition is permitted without
3920 a warning if preceded by a declaration which specifies the argument
3923 @item -Wold-style-declaration @r{(C and Objective-C only)}
3924 @opindex Wold-style-declaration
3925 @opindex Wno-old-style-declaration
3926 Warn for obsolescent usages, according to the C Standard, in a
3927 declaration. For example, warn if storage-class specifiers like
3928 @code{static} are not the first things in a declaration. This warning
3929 is also enabled by @option{-Wextra}.
3931 @item -Wold-style-definition @r{(C and Objective-C only)}
3932 @opindex Wold-style-definition
3933 @opindex Wno-old-style-definition
3934 Warn if an old-style function definition is used. A warning is given
3935 even if there is a previous prototype.
3937 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
3938 @opindex Wmissing-parameter-type
3939 @opindex Wno-missing-parameter-type
3940 A function parameter is declared without a type specifier in K&R-style
3947 This warning is also enabled by @option{-Wextra}.
3949 @item -Wmissing-prototypes @r{(C and Objective-C only)}
3950 @opindex Wmissing-prototypes
3951 @opindex Wno-missing-prototypes
3952 Warn if a global function is defined without a previous prototype
3953 declaration. This warning is issued even if the definition itself
3954 provides a prototype. The aim is to detect global functions that fail
3955 to be declared in header files.
3957 @item -Wmissing-declarations
3958 @opindex Wmissing-declarations
3959 @opindex Wno-missing-declarations
3960 Warn if a global function is defined without a previous declaration.
3961 Do so even if the definition itself provides a prototype.
3962 Use this option to detect global functions that are not declared in
3963 header files. In C++, no warnings are issued for function templates,
3964 or for inline functions, or for functions in anonymous namespaces.
3966 @item -Wmissing-field-initializers
3967 @opindex Wmissing-field-initializers
3968 @opindex Wno-missing-field-initializers
3972 Warn if a structure's initializer has some fields missing. For
3973 example, the following code would cause such a warning, because
3974 @code{x.h} is implicitly zero:
3977 struct s @{ int f, g, h; @};
3978 struct s x = @{ 3, 4 @};
3981 This option does not warn about designated initializers, so the following
3982 modification would not trigger a warning:
3985 struct s @{ int f, g, h; @};
3986 struct s x = @{ .f = 3, .g = 4 @};
3989 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3990 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3992 @item -Wmissing-noreturn
3993 @opindex Wmissing-noreturn
3994 @opindex Wno-missing-noreturn
3995 Warn about functions which might be candidates for attribute @code{noreturn}.
3996 Note these are only possible candidates, not absolute ones. Care should
3997 be taken to manually verify functions actually do not ever return before
3998 adding the @code{noreturn} attribute, otherwise subtle code generation
3999 bugs could be introduced. You will not get a warning for @code{main} in
4000 hosted C environments.
4002 @item -Wmissing-format-attribute
4003 @opindex Wmissing-format-attribute
4004 @opindex Wno-missing-format-attribute
4007 Warn about function pointers which might be candidates for @code{format}
4008 attributes. Note these are only possible candidates, not absolute ones.
4009 GCC will guess that function pointers with @code{format} attributes that
4010 are used in assignment, initialization, parameter passing or return
4011 statements should have a corresponding @code{format} attribute in the
4012 resulting type. I.e.@: the left-hand side of the assignment or
4013 initialization, the type of the parameter variable, or the return type
4014 of the containing function respectively should also have a @code{format}
4015 attribute to avoid the warning.
4017 GCC will also warn about function definitions which might be
4018 candidates for @code{format} attributes. Again, these are only
4019 possible candidates. GCC will guess that @code{format} attributes
4020 might be appropriate for any function that calls a function like
4021 @code{vprintf} or @code{vscanf}, but this might not always be the
4022 case, and some functions for which @code{format} attributes are
4023 appropriate may not be detected.
4025 @item -Wno-multichar
4026 @opindex Wno-multichar
4028 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4029 Usually they indicate a typo in the user's code, as they have
4030 implementation-defined values, and should not be used in portable code.
4032 @item -Wnormalized=<none|id|nfc|nfkc>
4033 @opindex Wnormalized=
4036 @cindex character set, input normalization
4037 In ISO C and ISO C++, two identifiers are different if they are
4038 different sequences of characters. However, sometimes when characters
4039 outside the basic ASCII character set are used, you can have two
4040 different character sequences that look the same. To avoid confusion,
4041 the ISO 10646 standard sets out some @dfn{normalization rules} which
4042 when applied ensure that two sequences that look the same are turned into
4043 the same sequence. GCC can warn you if you are using identifiers which
4044 have not been normalized; this option controls that warning.
4046 There are four levels of warning that GCC supports. The default is
4047 @option{-Wnormalized=nfc}, which warns about any identifier which is
4048 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4049 recommended form for most uses.
4051 Unfortunately, there are some characters which ISO C and ISO C++ allow
4052 in identifiers that when turned into NFC aren't allowable as
4053 identifiers. That is, there's no way to use these symbols in portable
4054 ISO C or C++ and have all your identifiers in NFC@.
4055 @option{-Wnormalized=id} suppresses the warning for these characters.
4056 It is hoped that future versions of the standards involved will correct
4057 this, which is why this option is not the default.
4059 You can switch the warning off for all characters by writing
4060 @option{-Wnormalized=none}. You would only want to do this if you
4061 were using some other normalization scheme (like ``D''), because
4062 otherwise you can easily create bugs that are literally impossible to see.
4064 Some characters in ISO 10646 have distinct meanings but look identical
4065 in some fonts or display methodologies, especially once formatting has
4066 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4067 LETTER N'', will display just like a regular @code{n} which has been
4068 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4069 normalization scheme to convert all these into a standard form as
4070 well, and GCC will warn if your code is not in NFKC if you use
4071 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4072 about every identifier that contains the letter O because it might be
4073 confused with the digit 0, and so is not the default, but may be
4074 useful as a local coding convention if the programming environment is
4075 unable to be fixed to display these characters distinctly.
4077 @item -Wno-deprecated
4078 @opindex Wno-deprecated
4079 @opindex Wdeprecated
4080 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4082 @item -Wno-deprecated-declarations
4083 @opindex Wno-deprecated-declarations
4084 @opindex Wdeprecated-declarations
4085 Do not warn about uses of functions (@pxref{Function Attributes}),
4086 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4087 Attributes}) marked as deprecated by using the @code{deprecated}
4091 @opindex Wno-overflow
4093 Do not warn about compile-time overflow in constant expressions.
4095 @item -Woverride-init @r{(C and Objective-C only)}
4096 @opindex Woverride-init
4097 @opindex Wno-override-init
4101 Warn if an initialized field without side effects is overridden when
4102 using designated initializers (@pxref{Designated Inits, , Designated
4105 This warning is included in @option{-Wextra}. To get other
4106 @option{-Wextra} warnings without this one, use @samp{-Wextra
4107 -Wno-override-init}.
4112 Warn if a structure is given the packed attribute, but the packed
4113 attribute has no effect on the layout or size of the structure.
4114 Such structures may be mis-aligned for little benefit. For
4115 instance, in this code, the variable @code{f.x} in @code{struct bar}
4116 will be misaligned even though @code{struct bar} does not itself
4117 have the packed attribute:
4124 @} __attribute__((packed));
4132 @item -Wpacked-bitfield-compat
4133 @opindex Wpacked-bitfield-compat
4134 @opindex Wno-packed-bitfield-compat
4135 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4136 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4137 the change can lead to differences in the structure layout. GCC
4138 informs you when the offset of such a field has changed in GCC 4.4.
4139 For example there is no longer a 4-bit padding between field @code{a}
4140 and @code{b} in this structure:
4147 @} __attribute__ ((packed));
4150 This warning is enabled by default. Use
4151 @option{-Wno-packed-bitfield-compat} to disable this warning.
4156 Warn if padding is included in a structure, either to align an element
4157 of the structure or to align the whole structure. Sometimes when this
4158 happens it is possible to rearrange the fields of the structure to
4159 reduce the padding and so make the structure smaller.
4161 @item -Wredundant-decls
4162 @opindex Wredundant-decls
4163 @opindex Wno-redundant-decls
4164 Warn if anything is declared more than once in the same scope, even in
4165 cases where multiple declaration is valid and changes nothing.
4167 @item -Wnested-externs @r{(C and Objective-C only)}
4168 @opindex Wnested-externs
4169 @opindex Wno-nested-externs
4170 Warn if an @code{extern} declaration is encountered within a function.
4172 @item -Wunreachable-code
4173 @opindex Wunreachable-code
4174 @opindex Wno-unreachable-code
4175 Warn if the compiler detects that code will never be executed.
4177 This option is intended to warn when the compiler detects that at
4178 least a whole line of source code will never be executed, because
4179 some condition is never satisfied or because it is after a
4180 procedure that never returns.
4182 It is possible for this option to produce a warning even though there
4183 are circumstances under which part of the affected line can be executed,
4184 so care should be taken when removing apparently-unreachable code.
4186 For instance, when a function is inlined, a warning may mean that the
4187 line is unreachable in only one inlined copy of the function.
4189 This option is not made part of @option{-Wall} because in a debugging
4190 version of a program there is often substantial code which checks
4191 correct functioning of the program and is, hopefully, unreachable
4192 because the program does work. Another common use of unreachable
4193 code is to provide behavior which is selectable at compile-time.
4198 Warn if a function can not be inlined and it was declared as inline.
4199 Even with this option, the compiler will not warn about failures to
4200 inline functions declared in system headers.
4202 The compiler uses a variety of heuristics to determine whether or not
4203 to inline a function. For example, the compiler takes into account
4204 the size of the function being inlined and the amount of inlining
4205 that has already been done in the current function. Therefore,
4206 seemingly insignificant changes in the source program can cause the
4207 warnings produced by @option{-Winline} to appear or disappear.
4209 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4210 @opindex Wno-invalid-offsetof
4211 @opindex Winvalid-offsetof
4212 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4213 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4214 to a non-POD type is undefined. In existing C++ implementations,
4215 however, @samp{offsetof} typically gives meaningful results even when
4216 applied to certain kinds of non-POD types. (Such as a simple
4217 @samp{struct} that fails to be a POD type only by virtue of having a
4218 constructor.) This flag is for users who are aware that they are
4219 writing nonportable code and who have deliberately chosen to ignore the
4222 The restrictions on @samp{offsetof} may be relaxed in a future version
4223 of the C++ standard.
4225 @item -Wno-int-to-pointer-cast @r{(C and Objective-C only)}
4226 @opindex Wno-int-to-pointer-cast
4227 @opindex Wint-to-pointer-cast
4228 Suppress warnings from casts to pointer type of an integer of a
4231 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4232 @opindex Wno-pointer-to-int-cast
4233 @opindex Wpointer-to-int-cast
4234 Suppress warnings from casts from a pointer to an integer type of a
4238 @opindex Winvalid-pch
4239 @opindex Wno-invalid-pch
4240 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4241 the search path but can't be used.
4245 @opindex Wno-long-long
4246 Warn if @samp{long long} type is used. This is enabled by either
4247 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4248 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4250 @item -Wvariadic-macros
4251 @opindex Wvariadic-macros
4252 @opindex Wno-variadic-macros
4253 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4254 alternate syntax when in pedantic ISO C99 mode. This is default.
4255 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4260 Warn if variable length array is used in the code.
4261 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4262 the variable length array.
4264 @item -Wvolatile-register-var
4265 @opindex Wvolatile-register-var
4266 @opindex Wno-volatile-register-var
4267 Warn if a register variable is declared volatile. The volatile
4268 modifier does not inhibit all optimizations that may eliminate reads
4269 and/or writes to register variables. This warning is enabled by
4272 @item -Wdisabled-optimization
4273 @opindex Wdisabled-optimization
4274 @opindex Wno-disabled-optimization
4275 Warn if a requested optimization pass is disabled. This warning does
4276 not generally indicate that there is anything wrong with your code; it
4277 merely indicates that GCC's optimizers were unable to handle the code
4278 effectively. Often, the problem is that your code is too big or too
4279 complex; GCC will refuse to optimize programs when the optimization
4280 itself is likely to take inordinate amounts of time.
4282 @item -Wpointer-sign @r{(C and Objective-C only)}
4283 @opindex Wpointer-sign
4284 @opindex Wno-pointer-sign
4285 Warn for pointer argument passing or assignment with different signedness.
4286 This option is only supported for C and Objective-C@. It is implied by
4287 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4288 @option{-Wno-pointer-sign}.
4290 @item -Wstack-protector
4291 @opindex Wstack-protector
4292 @opindex Wno-stack-protector
4293 This option is only active when @option{-fstack-protector} is active. It
4294 warns about functions that will not be protected against stack smashing.
4297 @opindex Wno-mudflap
4298 Suppress warnings about constructs that cannot be instrumented by
4301 @item -Woverlength-strings
4302 @opindex Woverlength-strings
4303 @opindex Wno-overlength-strings
4304 Warn about string constants which are longer than the ``minimum
4305 maximum'' length specified in the C standard. Modern compilers
4306 generally allow string constants which are much longer than the
4307 standard's minimum limit, but very portable programs should avoid
4308 using longer strings.
4310 The limit applies @emph{after} string constant concatenation, and does
4311 not count the trailing NUL@. In C89, the limit was 509 characters; in
4312 C99, it was raised to 4095. C++98 does not specify a normative
4313 minimum maximum, so we do not diagnose overlength strings in C++@.
4315 This option is implied by @option{-pedantic}, and can be disabled with
4316 @option{-Wno-overlength-strings}.
4318 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
4319 @opindex Wunsuffixed-float-constants
4321 GCC will issue a warning for any floating constant that does not have
4322 a suffix. When used together with @option{-Wsystem-headers} it will
4323 warn about such constants in system header files. This can be useful
4324 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4325 from the decimal floating-point extension to C99.
4328 @node Debugging Options
4329 @section Options for Debugging Your Program or GCC
4330 @cindex options, debugging
4331 @cindex debugging information options
4333 GCC has various special options that are used for debugging
4334 either your program or GCC:
4339 Produce debugging information in the operating system's native format
4340 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4343 On most systems that use stabs format, @option{-g} enables use of extra
4344 debugging information that only GDB can use; this extra information
4345 makes debugging work better in GDB but will probably make other debuggers
4347 refuse to read the program. If you want to control for certain whether
4348 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4349 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4351 GCC allows you to use @option{-g} with
4352 @option{-O}. The shortcuts taken by optimized code may occasionally
4353 produce surprising results: some variables you declared may not exist
4354 at all; flow of control may briefly move where you did not expect it;
4355 some statements may not be executed because they compute constant
4356 results or their values were already at hand; some statements may
4357 execute in different places because they were moved out of loops.
4359 Nevertheless it proves possible to debug optimized output. This makes
4360 it reasonable to use the optimizer for programs that might have bugs.
4362 The following options are useful when GCC is generated with the
4363 capability for more than one debugging format.
4367 Produce debugging information for use by GDB@. This means to use the
4368 most expressive format available (DWARF 2, stabs, or the native format
4369 if neither of those are supported), including GDB extensions if at all
4374 Produce debugging information in stabs format (if that is supported),
4375 without GDB extensions. This is the format used by DBX on most BSD
4376 systems. On MIPS, Alpha and System V Release 4 systems this option
4377 produces stabs debugging output which is not understood by DBX or SDB@.
4378 On System V Release 4 systems this option requires the GNU assembler.
4380 @item -feliminate-unused-debug-symbols
4381 @opindex feliminate-unused-debug-symbols
4382 Produce debugging information in stabs format (if that is supported),
4383 for only symbols that are actually used.
4385 @item -femit-class-debug-always
4386 Instead of emitting debugging information for a C++ class in only one
4387 object file, emit it in all object files using the class. This option
4388 should be used only with debuggers that are unable to handle the way GCC
4389 normally emits debugging information for classes because using this
4390 option will increase the size of debugging information by as much as a
4395 Produce debugging information in stabs format (if that is supported),
4396 using GNU extensions understood only by the GNU debugger (GDB)@. The
4397 use of these extensions is likely to make other debuggers crash or
4398 refuse to read the program.
4402 Produce debugging information in COFF format (if that is supported).
4403 This is the format used by SDB on most System V systems prior to
4408 Produce debugging information in XCOFF format (if that is supported).
4409 This is the format used by the DBX debugger on IBM RS/6000 systems.
4413 Produce debugging information in XCOFF format (if that is supported),
4414 using GNU extensions understood only by the GNU debugger (GDB)@. The
4415 use of these extensions is likely to make other debuggers crash or
4416 refuse to read the program, and may cause assemblers other than the GNU
4417 assembler (GAS) to fail with an error.
4419 @item -gdwarf-@var{version}
4420 @opindex gdwarf-@var{version}
4421 Produce debugging information in DWARF format (if that is
4422 supported). This is the format used by DBX on IRIX 6. The value
4423 of @var{version} may be either 2, 3 or 4; the default version is 2.
4425 Note that with DWARF version 2 some ports require, and will always
4426 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4428 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
4429 for maximum benefit.
4431 @item -gstrict-dwarf
4432 @opindex gstrict-dwarf
4433 Disallow using extensions of later DWARF standard version than selected
4434 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
4435 DWARF extensions from later standard versions is allowed.
4437 @item -gno-strict-dwarf
4438 @opindex gno-strict-dwarf
4439 Allow using extensions of later DWARF standard version than selected with
4440 @option{-gdwarf-@var{version}}.
4444 Produce debugging information in VMS debug format (if that is
4445 supported). This is the format used by DEBUG on VMS systems.
4448 @itemx -ggdb@var{level}
4449 @itemx -gstabs@var{level}
4450 @itemx -gcoff@var{level}
4451 @itemx -gxcoff@var{level}
4452 @itemx -gvms@var{level}
4453 Request debugging information and also use @var{level} to specify how
4454 much information. The default level is 2.
4456 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4459 Level 1 produces minimal information, enough for making backtraces in
4460 parts of the program that you don't plan to debug. This includes
4461 descriptions of functions and external variables, but no information
4462 about local variables and no line numbers.
4464 Level 3 includes extra information, such as all the macro definitions
4465 present in the program. Some debuggers support macro expansion when
4466 you use @option{-g3}.
4468 @option{-gdwarf-2} does not accept a concatenated debug level, because
4469 GCC used to support an option @option{-gdwarf} that meant to generate
4470 debug information in version 1 of the DWARF format (which is very
4471 different from version 2), and it would have been too confusing. That
4472 debug format is long obsolete, but the option cannot be changed now.
4473 Instead use an additional @option{-g@var{level}} option to change the
4474 debug level for DWARF.
4478 Turn off generation of debug info, if leaving out this option would have
4479 generated it, or turn it on at level 2 otherwise. The position of this
4480 argument in the command line does not matter, it takes effect after all
4481 other options are processed, and it does so only once, no matter how
4482 many times it is given. This is mainly intended to be used with
4483 @option{-fcompare-debug}.
4485 @item -fdump-final-insns@r{[}=@var{file}@r{]}
4486 @opindex fdump-final-insns
4487 Dump the final internal representation (RTL) to @var{file}. If the
4488 optional argument is omitted (or if @var{file} is @code{.}), the name
4489 of the dump file will be determined by appending @code{.gkd} to the
4490 compilation output file name.
4492 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4493 @opindex fcompare-debug
4494 @opindex fno-compare-debug
4495 If no error occurs during compilation, run the compiler a second time,
4496 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4497 passed to the second compilation. Dump the final internal
4498 representation in both compilations, and print an error if they differ.
4500 If the equal sign is omitted, the default @option{-gtoggle} is used.
4502 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4503 and nonzero, implicitly enables @option{-fcompare-debug}. If
4504 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4505 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4508 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4509 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4510 of the final representation and the second compilation, preventing even
4511 @env{GCC_COMPARE_DEBUG} from taking effect.
4513 To verify full coverage during @option{-fcompare-debug} testing, set
4514 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4515 which GCC will reject as an invalid option in any actual compilation
4516 (rather than preprocessing, assembly or linking). To get just a
4517 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4518 not overridden} will do.
4520 @item -fcompare-debug-second
4521 @opindex fcompare-debug-second
4522 This option is implicitly passed to the compiler for the second
4523 compilation requested by @option{-fcompare-debug}, along with options to
4524 silence warnings, and omitting other options that would cause
4525 side-effect compiler outputs to files or to the standard output. Dump
4526 files and preserved temporary files are renamed so as to contain the
4527 @code{.gk} additional extension during the second compilation, to avoid
4528 overwriting those generated by the first.
4530 When this option is passed to the compiler driver, it causes the
4531 @emph{first} compilation to be skipped, which makes it useful for little
4532 other than debugging the compiler proper.
4534 @item -feliminate-dwarf2-dups
4535 @opindex feliminate-dwarf2-dups
4536 Compress DWARF2 debugging information by eliminating duplicated
4537 information about each symbol. This option only makes sense when
4538 generating DWARF2 debugging information with @option{-gdwarf-2}.
4540 @item -femit-struct-debug-baseonly
4541 Emit debug information for struct-like types
4542 only when the base name of the compilation source file
4543 matches the base name of file in which the struct was defined.
4545 This option substantially reduces the size of debugging information,
4546 but at significant potential loss in type information to the debugger.
4547 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4548 See @option{-femit-struct-debug-detailed} for more detailed control.
4550 This option works only with DWARF 2.
4552 @item -femit-struct-debug-reduced
4553 Emit debug information for struct-like types
4554 only when the base name of the compilation source file
4555 matches the base name of file in which the type was defined,
4556 unless the struct is a template or defined in a system header.
4558 This option significantly reduces the size of debugging information,
4559 with some potential loss in type information to the debugger.
4560 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4561 See @option{-femit-struct-debug-detailed} for more detailed control.
4563 This option works only with DWARF 2.
4565 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4566 Specify the struct-like types
4567 for which the compiler will generate debug information.
4568 The intent is to reduce duplicate struct debug information
4569 between different object files within the same program.
4571 This option is a detailed version of
4572 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4573 which will serve for most needs.
4575 A specification has the syntax
4576 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4578 The optional first word limits the specification to
4579 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4580 A struct type is used directly when it is the type of a variable, member.
4581 Indirect uses arise through pointers to structs.
4582 That is, when use of an incomplete struct would be legal, the use is indirect.
4584 @samp{struct one direct; struct two * indirect;}.
4586 The optional second word limits the specification to
4587 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4588 Generic structs are a bit complicated to explain.
4589 For C++, these are non-explicit specializations of template classes,
4590 or non-template classes within the above.
4591 Other programming languages have generics,
4592 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4594 The third word specifies the source files for those
4595 structs for which the compiler will emit debug information.
4596 The values @samp{none} and @samp{any} have the normal meaning.
4597 The value @samp{base} means that
4598 the base of name of the file in which the type declaration appears
4599 must match the base of the name of the main compilation file.
4600 In practice, this means that
4601 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4602 but types declared in other header will not.
4603 The value @samp{sys} means those types satisfying @samp{base}
4604 or declared in system or compiler headers.
4606 You may need to experiment to determine the best settings for your application.
4608 The default is @samp{-femit-struct-debug-detailed=all}.
4610 This option works only with DWARF 2.
4612 @item -fno-merge-debug-strings
4613 @opindex fmerge-debug-strings
4614 @opindex fno-merge-debug-strings
4615 Direct the linker to not merge together strings in the debugging
4616 information which are identical in different object files. Merging is
4617 not supported by all assemblers or linkers. Merging decreases the size
4618 of the debug information in the output file at the cost of increasing
4619 link processing time. Merging is enabled by default.
4621 @item -fdebug-prefix-map=@var{old}=@var{new}
4622 @opindex fdebug-prefix-map
4623 When compiling files in directory @file{@var{old}}, record debugging
4624 information describing them as in @file{@var{new}} instead.
4626 @item -fno-dwarf2-cfi-asm
4627 @opindex fdwarf2-cfi-asm
4628 @opindex fno-dwarf2-cfi-asm
4629 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4630 instead of using GAS @code{.cfi_*} directives.
4632 @cindex @command{prof}
4635 Generate extra code to write profile information suitable for the
4636 analysis program @command{prof}. You must use this option when compiling
4637 the source files you want data about, and you must also use it when
4640 @cindex @command{gprof}
4643 Generate extra code to write profile information suitable for the
4644 analysis program @command{gprof}. You must use this option when compiling
4645 the source files you want data about, and you must also use it when
4650 Makes the compiler print out each function name as it is compiled, and
4651 print some statistics about each pass when it finishes.
4654 @opindex ftime-report
4655 Makes the compiler print some statistics about the time consumed by each
4656 pass when it finishes.
4659 @opindex fmem-report
4660 Makes the compiler print some statistics about permanent memory
4661 allocation when it finishes.
4663 @item -fpre-ipa-mem-report
4664 @opindex fpre-ipa-mem-report
4665 @item -fpost-ipa-mem-report
4666 @opindex fpost-ipa-mem-report
4667 Makes the compiler print some statistics about permanent memory
4668 allocation before or after interprocedural optimization.
4670 @item -fprofile-arcs
4671 @opindex fprofile-arcs
4672 Add code so that program flow @dfn{arcs} are instrumented. During
4673 execution the program records how many times each branch and call is
4674 executed and how many times it is taken or returns. When the compiled
4675 program exits it saves this data to a file called
4676 @file{@var{auxname}.gcda} for each source file. The data may be used for
4677 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4678 test coverage analysis (@option{-ftest-coverage}). Each object file's
4679 @var{auxname} is generated from the name of the output file, if
4680 explicitly specified and it is not the final executable, otherwise it is
4681 the basename of the source file. In both cases any suffix is removed
4682 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4683 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4684 @xref{Cross-profiling}.
4686 @cindex @command{gcov}
4690 This option is used to compile and link code instrumented for coverage
4691 analysis. The option is a synonym for @option{-fprofile-arcs}
4692 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4693 linking). See the documentation for those options for more details.
4698 Compile the source files with @option{-fprofile-arcs} plus optimization
4699 and code generation options. For test coverage analysis, use the
4700 additional @option{-ftest-coverage} option. You do not need to profile
4701 every source file in a program.
4704 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4705 (the latter implies the former).
4708 Run the program on a representative workload to generate the arc profile
4709 information. This may be repeated any number of times. You can run
4710 concurrent instances of your program, and provided that the file system
4711 supports locking, the data files will be correctly updated. Also
4712 @code{fork} calls are detected and correctly handled (double counting
4716 For profile-directed optimizations, compile the source files again with
4717 the same optimization and code generation options plus
4718 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4719 Control Optimization}).
4722 For test coverage analysis, use @command{gcov} to produce human readable
4723 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4724 @command{gcov} documentation for further information.
4728 With @option{-fprofile-arcs}, for each function of your program GCC
4729 creates a program flow graph, then finds a spanning tree for the graph.
4730 Only arcs that are not on the spanning tree have to be instrumented: the
4731 compiler adds code to count the number of times that these arcs are
4732 executed. When an arc is the only exit or only entrance to a block, the
4733 instrumentation code can be added to the block; otherwise, a new basic
4734 block must be created to hold the instrumentation code.
4737 @item -ftest-coverage
4738 @opindex ftest-coverage
4739 Produce a notes file that the @command{gcov} code-coverage utility
4740 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4741 show program coverage. Each source file's note file is called
4742 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4743 above for a description of @var{auxname} and instructions on how to
4744 generate test coverage data. Coverage data will match the source files
4745 more closely, if you do not optimize.
4747 @item -fdbg-cnt-list
4748 @opindex fdbg-cnt-list
4749 Print the name and the counter upperbound for all debug counters.
4751 @item -fdbg-cnt=@var{counter-value-list}
4753 Set the internal debug counter upperbound. @var{counter-value-list}
4754 is a comma-separated list of @var{name}:@var{value} pairs
4755 which sets the upperbound of each debug counter @var{name} to @var{value}.
4756 All debug counters have the initial upperbound of @var{UINT_MAX},
4757 thus dbg_cnt() returns true always unless the upperbound is set by this option.
4758 e.g. With -fdbg-cnt=dce:10,tail_call:0
4759 dbg_cnt(dce) will return true only for first 10 invocations
4760 and dbg_cnt(tail_call) will return false always.
4762 @item -d@var{letters}
4763 @itemx -fdump-rtl-@var{pass}
4765 Says to make debugging dumps during compilation at times specified by
4766 @var{letters}. This is used for debugging the RTL-based passes of the
4767 compiler. The file names for most of the dumps are made by appending
4768 a pass number and a word to the @var{dumpname}, and the files are
4769 created in the directory of the output file. @var{dumpname} is
4770 generated from the name of the output file, if explicitly specified
4771 and it is not an executable, otherwise it is the basename of the
4772 source file. These switches may have different effects when
4773 @option{-E} is used for preprocessing.
4775 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
4776 @option{-d} option @var{letters}. Here are the possible
4777 letters for use in @var{pass} and @var{letters}, and their meanings:
4781 @item -fdump-rtl-alignments
4782 @opindex fdump-rtl-alignments
4783 Dump after branch alignments have been computed.
4785 @item -fdump-rtl-asmcons
4786 @opindex fdump-rtl-asmcons
4787 Dump after fixing rtl statements that have unsatisfied in/out constraints.
4789 @item -fdump-rtl-auto_inc_dec
4790 @opindex fdump-rtl-auto_inc_dec
4791 Dump after auto-inc-dec discovery. This pass is only run on
4792 architectures that have auto inc or auto dec instructions.
4794 @item -fdump-rtl-barriers
4795 @opindex fdump-rtl-barriers
4796 Dump after cleaning up the barrier instructions.
4798 @item -fdump-rtl-bbpart
4799 @opindex fdump-rtl-bbpart
4800 Dump after partitioning hot and cold basic blocks.
4802 @item -fdump-rtl-bbro
4803 @opindex fdump-rtl-bbro
4804 Dump after block reordering.
4806 @item -fdump-rtl-btl1
4807 @itemx -fdump-rtl-btl2
4808 @opindex fdump-rtl-btl2
4809 @opindex fdump-rtl-btl2
4810 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
4811 after the two branch
4812 target load optimization passes.
4814 @item -fdump-rtl-bypass
4815 @opindex fdump-rtl-bypass
4816 Dump after jump bypassing and control flow optimizations.
4818 @item -fdump-rtl-combine
4819 @opindex fdump-rtl-combine
4820 Dump after the RTL instruction combination pass.
4822 @item -fdump-rtl-compgotos
4823 @opindex fdump-rtl-compgotos
4824 Dump after duplicating the computed gotos.
4826 @item -fdump-rtl-ce1
4827 @itemx -fdump-rtl-ce2
4828 @itemx -fdump-rtl-ce3
4829 @opindex fdump-rtl-ce1
4830 @opindex fdump-rtl-ce2
4831 @opindex fdump-rtl-ce3
4832 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
4833 @option{-fdump-rtl-ce3} enable dumping after the three
4834 if conversion passes.
4836 @itemx -fdump-rtl-cprop_hardreg
4837 @opindex fdump-rtl-cprop_hardreg
4838 Dump after hard register copy propagation.
4840 @itemx -fdump-rtl-csa
4841 @opindex fdump-rtl-csa
4842 Dump after combining stack adjustments.
4844 @item -fdump-rtl-cse1
4845 @itemx -fdump-rtl-cse2
4846 @opindex fdump-rtl-cse1
4847 @opindex fdump-rtl-cse2
4848 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
4849 the two common sub-expression elimination passes.
4851 @itemx -fdump-rtl-dce
4852 @opindex fdump-rtl-dce
4853 Dump after the standalone dead code elimination passes.
4855 @itemx -fdump-rtl-dbr
4856 @opindex fdump-rtl-dbr
4857 Dump after delayed branch scheduling.
4859 @item -fdump-rtl-dce1
4860 @itemx -fdump-rtl-dce2
4861 @opindex fdump-rtl-dce1
4862 @opindex fdump-rtl-dce2
4863 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
4864 the two dead store elimination passes.
4867 @opindex fdump-rtl-eh
4868 Dump after finalization of EH handling code.
4870 @item -fdump-rtl-eh_ranges
4871 @opindex fdump-rtl-eh_ranges
4872 Dump after conversion of EH handling range regions.
4874 @item -fdump-rtl-expand
4875 @opindex fdump-rtl-expand
4876 Dump after RTL generation.
4878 @item -fdump-rtl-fwprop1
4879 @itemx -fdump-rtl-fwprop2
4880 @opindex fdump-rtl-fwprop1
4881 @opindex fdump-rtl-fwprop2
4882 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
4883 dumping after the two forward propagation passes.
4885 @item -fdump-rtl-gcse1
4886 @itemx -fdump-rtl-gcse2
4887 @opindex fdump-rtl-gcse1
4888 @opindex fdump-rtl-gcse2
4889 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
4890 after global common subexpression elimination.
4892 @item -fdump-rtl-init-regs
4893 @opindex fdump-rtl-init-regs
4894 Dump after the initialization of the registers.
4896 @item -fdump-rtl-initvals
4897 @opindex fdump-rtl-initvals
4898 Dump after the computation of the initial value sets.
4900 @itemx -fdump-rtl-into_cfglayout
4901 @opindex fdump-rtl-into_cfglayout
4902 Dump after converting to cfglayout mode.
4904 @item -fdump-rtl-ira
4905 @opindex fdump-rtl-ira
4906 Dump after iterated register allocation.
4908 @item -fdump-rtl-jump
4909 @opindex fdump-rtl-jump
4910 Dump after the second jump optimization.
4912 @item -fdump-rtl-loop2
4913 @opindex fdump-rtl-loop2
4914 @option{-fdump-rtl-loop2} enables dumping after the rtl
4915 loop optimization passes.
4917 @item -fdump-rtl-mach
4918 @opindex fdump-rtl-mach
4919 Dump after performing the machine dependent reorganization pass, if that
4922 @item -fdump-rtl-mode_sw
4923 @opindex fdump-rtl-mode_sw
4924 Dump after removing redundant mode switches.
4926 @item -fdump-rtl-rnreg
4927 @opindex fdump-rtl-rnreg
4928 Dump after register renumbering.
4930 @itemx -fdump-rtl-outof_cfglayout
4931 @opindex fdump-rtl-outof_cfglayout
4932 Dump after converting from cfglayout mode.
4934 @item -fdump-rtl-peephole2
4935 @opindex fdump-rtl-peephole2
4936 Dump after the peephole pass.
4938 @item -fdump-rtl-postreload
4939 @opindex fdump-rtl-postreload
4940 Dump after post-reload optimizations.
4942 @itemx -fdump-rtl-pro_and_epilogue
4943 @opindex fdump-rtl-pro_and_epilogue
4944 Dump after generating the function pro and epilogues.
4946 @item -fdump-rtl-regmove
4947 @opindex fdump-rtl-regmove
4948 Dump after the register move pass.
4950 @item -fdump-rtl-sched1
4951 @itemx -fdump-rtl-sched2
4952 @opindex fdump-rtl-sched1
4953 @opindex fdump-rtl-sched2
4954 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
4955 after the basic block scheduling passes.
4957 @item -fdump-rtl-see
4958 @opindex fdump-rtl-see
4959 Dump after sign extension elimination.
4961 @item -fdump-rtl-seqabstr
4962 @opindex fdump-rtl-seqabstr
4963 Dump after common sequence discovery.
4965 @item -fdump-rtl-shorten
4966 @opindex fdump-rtl-shorten
4967 Dump after shortening branches.
4969 @item -fdump-rtl-sibling
4970 @opindex fdump-rtl-sibling
4971 Dump after sibling call optimizations.
4973 @item -fdump-rtl-split1
4974 @itemx -fdump-rtl-split2
4975 @itemx -fdump-rtl-split3
4976 @itemx -fdump-rtl-split4
4977 @itemx -fdump-rtl-split5
4978 @opindex fdump-rtl-split1
4979 @opindex fdump-rtl-split2
4980 @opindex fdump-rtl-split3
4981 @opindex fdump-rtl-split4
4982 @opindex fdump-rtl-split5
4983 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
4984 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
4985 @option{-fdump-rtl-split5} enable dumping after five rounds of
4986 instruction splitting.
4988 @item -fdump-rtl-sms
4989 @opindex fdump-rtl-sms
4990 Dump after modulo scheduling. This pass is only run on some
4993 @item -fdump-rtl-stack
4994 @opindex fdump-rtl-stack
4995 Dump after conversion from GCC's "flat register file" registers to the
4996 x87's stack-like registers. This pass is only run on x86 variants.
4998 @item -fdump-rtl-subreg1
4999 @itemx -fdump-rtl-subreg2
5000 @opindex fdump-rtl-subreg1
5001 @opindex fdump-rtl-subreg2
5002 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
5003 the two subreg expansion passes.
5005 @item -fdump-rtl-unshare
5006 @opindex fdump-rtl-unshare
5007 Dump after all rtl has been unshared.
5009 @item -fdump-rtl-vartrack
5010 @opindex fdump-rtl-vartrack
5011 Dump after variable tracking.
5013 @item -fdump-rtl-vregs
5014 @opindex fdump-rtl-vregs
5015 Dump after converting virtual registers to hard registers.
5017 @item -fdump-rtl-web
5018 @opindex fdump-rtl-web
5019 Dump after live range splitting.
5021 @item -fdump-rtl-regclass
5022 @itemx -fdump-rtl-subregs_of_mode_init
5023 @itemx -fdump-rtl-subregs_of_mode_finish
5024 @itemx -fdump-rtl-dfinit
5025 @itemx -fdump-rtl-dfinish
5026 @opindex fdump-rtl-regclass
5027 @opindex fdump-rtl-subregs_of_mode_init
5028 @opindex fdump-rtl-subregs_of_mode_finish
5029 @opindex fdump-rtl-dfinit
5030 @opindex fdump-rtl-dfinish
5031 These dumps are defined but always produce empty files.
5033 @item -fdump-rtl-all
5034 @opindex fdump-rtl-all
5035 Produce all the dumps listed above.
5039 Annotate the assembler output with miscellaneous debugging information.
5043 Dump all macro definitions, at the end of preprocessing, in addition to
5048 Produce a core dump whenever an error occurs.
5052 Print statistics on memory usage, at the end of the run, to
5057 Annotate the assembler output with a comment indicating which
5058 pattern and alternative was used. The length of each instruction is
5063 Dump the RTL in the assembler output as a comment before each instruction.
5064 Also turns on @option{-dp} annotation.
5068 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
5069 dump a representation of the control flow graph suitable for viewing with VCG
5070 to @file{@var{file}.@var{pass}.vcg}.
5074 Just generate RTL for a function instead of compiling it. Usually used
5075 with @option{-fdump-rtl-expand}.
5079 Dump debugging information during parsing, to standard error.
5083 @opindex fdump-noaddr
5084 When doing debugging dumps, suppress address output. This makes it more
5085 feasible to use diff on debugging dumps for compiler invocations with
5086 different compiler binaries and/or different
5087 text / bss / data / heap / stack / dso start locations.
5089 @item -fdump-unnumbered
5090 @opindex fdump-unnumbered
5091 When doing debugging dumps, suppress instruction numbers and address output.
5092 This makes it more feasible to use diff on debugging dumps for compiler
5093 invocations with different options, in particular with and without
5096 @item -fdump-unnumbered-links
5097 @opindex fdump-unnumbered-links
5098 When doing debugging dumps (see @option{-d} option above), suppress
5099 instruction numbers for the links to the previous and next instructions
5102 @item -fdump-translation-unit @r{(C++ only)}
5103 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5104 @opindex fdump-translation-unit
5105 Dump a representation of the tree structure for the entire translation
5106 unit to a file. The file name is made by appending @file{.tu} to the
5107 source file name, and the file is created in the same directory as the
5108 output file. If the @samp{-@var{options}} form is used, @var{options}
5109 controls the details of the dump as described for the
5110 @option{-fdump-tree} options.
5112 @item -fdump-class-hierarchy @r{(C++ only)}
5113 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5114 @opindex fdump-class-hierarchy
5115 Dump a representation of each class's hierarchy and virtual function
5116 table layout to a file. The file name is made by appending
5117 @file{.class} to the source file name, and the file is created in the
5118 same directory as the output file. If the @samp{-@var{options}} form
5119 is used, @var{options} controls the details of the dump as described
5120 for the @option{-fdump-tree} options.
5122 @item -fdump-ipa-@var{switch}
5124 Control the dumping at various stages of inter-procedural analysis
5125 language tree to a file. The file name is generated by appending a
5126 switch specific suffix to the source file name, and the file is created
5127 in the same directory as the output file. The following dumps are
5132 Enables all inter-procedural analysis dumps.
5135 Dumps information about call-graph optimization, unused function removal,
5136 and inlining decisions.
5139 Dump after function inlining.
5143 @item -fdump-statistics-@var{option}
5144 @opindex fdump-statistics
5145 Enable and control dumping of pass statistics in a separate file. The
5146 file name is generated by appending a suffix ending in
5147 @samp{.statistics} to the source file name, and the file is created in
5148 the same directory as the output file. If the @samp{-@var{option}}
5149 form is used, @samp{-stats} will cause counters to be summed over the
5150 whole compilation unit while @samp{-details} will dump every event as
5151 the passes generate them. The default with no option is to sum
5152 counters for each function compiled.
5154 @item -fdump-tree-@var{switch}
5155 @itemx -fdump-tree-@var{switch}-@var{options}
5157 Control the dumping at various stages of processing the intermediate
5158 language tree to a file. The file name is generated by appending a
5159 switch specific suffix to the source file name, and the file is
5160 created in the same directory as the output file. If the
5161 @samp{-@var{options}} form is used, @var{options} is a list of
5162 @samp{-} separated options that control the details of the dump. Not
5163 all options are applicable to all dumps, those which are not
5164 meaningful will be ignored. The following options are available
5168 Print the address of each node. Usually this is not meaningful as it
5169 changes according to the environment and source file. Its primary use
5170 is for tying up a dump file with a debug environment.
5172 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
5173 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
5174 use working backward from mangled names in the assembly file.
5176 Inhibit dumping of members of a scope or body of a function merely
5177 because that scope has been reached. Only dump such items when they
5178 are directly reachable by some other path. When dumping pretty-printed
5179 trees, this option inhibits dumping the bodies of control structures.
5181 Print a raw representation of the tree. By default, trees are
5182 pretty-printed into a C-like representation.
5184 Enable more detailed dumps (not honored by every dump option).
5186 Enable dumping various statistics about the pass (not honored by every dump
5189 Enable showing basic block boundaries (disabled in raw dumps).
5191 Enable showing virtual operands for every statement.
5193 Enable showing line numbers for statements.
5195 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5197 Enable showing the tree dump for each statement.
5199 Enable showing the EH region number holding each statement.
5201 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5202 and @option{lineno}.
5205 The following tree dumps are possible:
5209 @opindex fdump-tree-original
5210 Dump before any tree based optimization, to @file{@var{file}.original}.
5213 @opindex fdump-tree-optimized
5214 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5217 @opindex fdump-tree-gimple
5218 Dump each function before and after the gimplification pass to a file. The
5219 file name is made by appending @file{.gimple} to the source file name.
5222 @opindex fdump-tree-cfg
5223 Dump the control flow graph of each function to a file. The file name is
5224 made by appending @file{.cfg} to the source file name.
5227 @opindex fdump-tree-vcg
5228 Dump the control flow graph of each function to a file in VCG format. The
5229 file name is made by appending @file{.vcg} to the source file name. Note
5230 that if the file contains more than one function, the generated file cannot
5231 be used directly by VCG@. You will need to cut and paste each function's
5232 graph into its own separate file first.
5235 @opindex fdump-tree-ch
5236 Dump each function after copying loop headers. The file name is made by
5237 appending @file{.ch} to the source file name.
5240 @opindex fdump-tree-ssa
5241 Dump SSA related information to a file. The file name is made by appending
5242 @file{.ssa} to the source file name.
5245 @opindex fdump-tree-alias
5246 Dump aliasing information for each function. The file name is made by
5247 appending @file{.alias} to the source file name.
5250 @opindex fdump-tree-ccp
5251 Dump each function after CCP@. The file name is made by appending
5252 @file{.ccp} to the source file name.
5255 @opindex fdump-tree-storeccp
5256 Dump each function after STORE-CCP@. The file name is made by appending
5257 @file{.storeccp} to the source file name.
5260 @opindex fdump-tree-pre
5261 Dump trees after partial redundancy elimination. The file name is made
5262 by appending @file{.pre} to the source file name.
5265 @opindex fdump-tree-fre
5266 Dump trees after full redundancy elimination. The file name is made
5267 by appending @file{.fre} to the source file name.
5270 @opindex fdump-tree-copyprop
5271 Dump trees after copy propagation. The file name is made
5272 by appending @file{.copyprop} to the source file name.
5274 @item store_copyprop
5275 @opindex fdump-tree-store_copyprop
5276 Dump trees after store copy-propagation. The file name is made
5277 by appending @file{.store_copyprop} to the source file name.
5280 @opindex fdump-tree-dce
5281 Dump each function after dead code elimination. The file name is made by
5282 appending @file{.dce} to the source file name.
5285 @opindex fdump-tree-mudflap
5286 Dump each function after adding mudflap instrumentation. The file name is
5287 made by appending @file{.mudflap} to the source file name.
5290 @opindex fdump-tree-sra
5291 Dump each function after performing scalar replacement of aggregates. The
5292 file name is made by appending @file{.sra} to the source file name.
5295 @opindex fdump-tree-sink
5296 Dump each function after performing code sinking. The file name is made
5297 by appending @file{.sink} to the source file name.
5300 @opindex fdump-tree-dom
5301 Dump each function after applying dominator tree optimizations. The file
5302 name is made by appending @file{.dom} to the source file name.
5305 @opindex fdump-tree-dse
5306 Dump each function after applying dead store elimination. The file
5307 name is made by appending @file{.dse} to the source file name.
5310 @opindex fdump-tree-phiopt
5311 Dump each function after optimizing PHI nodes into straightline code. The file
5312 name is made by appending @file{.phiopt} to the source file name.
5315 @opindex fdump-tree-forwprop
5316 Dump each function after forward propagating single use variables. The file
5317 name is made by appending @file{.forwprop} to the source file name.
5320 @opindex fdump-tree-copyrename
5321 Dump each function after applying the copy rename optimization. The file
5322 name is made by appending @file{.copyrename} to the source file name.
5325 @opindex fdump-tree-nrv
5326 Dump each function after applying the named return value optimization on
5327 generic trees. The file name is made by appending @file{.nrv} to the source
5331 @opindex fdump-tree-vect
5332 Dump each function after applying vectorization of loops. The file name is
5333 made by appending @file{.vect} to the source file name.
5336 @opindex fdump-tree-vrp
5337 Dump each function after Value Range Propagation (VRP). The file name
5338 is made by appending @file{.vrp} to the source file name.
5341 @opindex fdump-tree-all
5342 Enable all the available tree dumps with the flags provided in this option.
5345 @item -ftree-vectorizer-verbose=@var{n}
5346 @opindex ftree-vectorizer-verbose
5347 This option controls the amount of debugging output the vectorizer prints.
5348 This information is written to standard error, unless
5349 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5350 in which case it is output to the usual dump listing file, @file{.vect}.
5351 For @var{n}=0 no diagnostic information is reported.
5352 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5353 and the total number of loops that got vectorized.
5354 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5355 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5356 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5357 level that @option{-fdump-tree-vect-stats} uses.
5358 Higher verbosity levels mean either more information dumped for each
5359 reported loop, or same amount of information reported for more loops:
5360 If @var{n}=3, alignment related information is added to the reports.
5361 If @var{n}=4, data-references related information (e.g.@: memory dependences,
5362 memory access-patterns) is added to the reports.
5363 If @var{n}=5, the vectorizer reports also non-vectorized inner-most loops
5364 that did not pass the first analysis phase (i.e., may not be countable, or
5365 may have complicated control-flow).
5366 If @var{n}=6, the vectorizer reports also non-vectorized nested loops.
5367 For @var{n}=7, all the information the vectorizer generates during its
5368 analysis and transformation is reported. This is the same verbosity level
5369 that @option{-fdump-tree-vect-details} uses.
5371 @item -frandom-seed=@var{string}
5372 @opindex frandom-seed
5373 This option provides a seed that GCC uses when it would otherwise use
5374 random numbers. It is used to generate certain symbol names
5375 that have to be different in every compiled file. It is also used to
5376 place unique stamps in coverage data files and the object files that
5377 produce them. You can use the @option{-frandom-seed} option to produce
5378 reproducibly identical object files.
5380 The @var{string} should be different for every file you compile.
5382 @item -fsched-verbose=@var{n}
5383 @opindex fsched-verbose
5384 On targets that use instruction scheduling, this option controls the
5385 amount of debugging output the scheduler prints. This information is
5386 written to standard error, unless @option{-fdump-rtl-sched1} or
5387 @option{-fdump-rtl-sched2} is specified, in which case it is output
5388 to the usual dump listing file, @file{.sched} or @file{.sched2}
5389 respectively. However for @var{n} greater than nine, the output is
5390 always printed to standard error.
5392 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5393 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5394 For @var{n} greater than one, it also output basic block probabilities,
5395 detailed ready list information and unit/insn info. For @var{n} greater
5396 than two, it includes RTL at abort point, control-flow and regions info.
5397 And for @var{n} over four, @option{-fsched-verbose} also includes
5401 @itemx -save-temps=cwd
5403 Store the usual ``temporary'' intermediate files permanently; place them
5404 in the current directory and name them based on the source file. Thus,
5405 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5406 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5407 preprocessed @file{foo.i} output file even though the compiler now
5408 normally uses an integrated preprocessor.
5410 When used in combination with the @option{-x} command line option,
5411 @option{-save-temps} is sensible enough to avoid over writing an
5412 input source file with the same extension as an intermediate file.
5413 The corresponding intermediate file may be obtained by renaming the
5414 source file before using @option{-save-temps}.
5416 If you invoke GCC in parallel, compiling several different source
5417 files that share a common base name in different subdirectories or the
5418 same source file compiled for multiple output destinations, it is
5419 likely that the different parallel compilers will interfere with each
5420 other, and overwrite the temporary files. For instance:
5423 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5424 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5427 may result in @file{foo.i} and @file{foo.o} being written to
5428 simultaneously by both compilers.
5430 @item -save-temps=obj
5431 @opindex save-temps=obj
5432 Store the usual ``temporary'' intermediate files permanently. If the
5433 @option{-o} option is used, the temporary files are based on the
5434 object file. If the @option{-o} option is not used, the
5435 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5440 gcc -save-temps=obj -c foo.c
5441 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5442 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5445 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5446 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5447 @file{dir2/yfoobar.o}.
5449 @item -time@r{[}=@var{file}@r{]}
5451 Report the CPU time taken by each subprocess in the compilation
5452 sequence. For C source files, this is the compiler proper and assembler
5453 (plus the linker if linking is done).
5455 Without the specification of an output file, the output looks like this:
5462 The first number on each line is the ``user time'', that is time spent
5463 executing the program itself. The second number is ``system time'',
5464 time spent executing operating system routines on behalf of the program.
5465 Both numbers are in seconds.
5467 With the specification of an output file, the output is appended to the
5468 named file, and it looks like this:
5471 0.12 0.01 cc1 @var{options}
5472 0.00 0.01 as @var{options}
5475 The ``user time'' and the ``system time'' are moved before the program
5476 name, and the options passed to the program are displayed, so that one
5477 can later tell what file was being compiled, and with which options.
5479 @item -fvar-tracking
5480 @opindex fvar-tracking
5481 Run variable tracking pass. It computes where variables are stored at each
5482 position in code. Better debugging information is then generated
5483 (if the debugging information format supports this information).
5485 It is enabled by default when compiling with optimization (@option{-Os},
5486 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5487 the debug info format supports it.
5489 @item -fvar-tracking-assignments
5490 @opindex fvar-tracking-assignments
5491 @opindex fno-var-tracking-assignments
5492 Annotate assignments to user variables early in the compilation and
5493 attempt to carry the annotations over throughout the compilation all the
5494 way to the end, in an attempt to improve debug information while
5495 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
5497 It can be enabled even if var-tracking is disabled, in which case
5498 annotations will be created and maintained, but discarded at the end.
5500 @item -fvar-tracking-assignments-toggle
5501 @opindex fvar-tracking-assignments-toggle
5502 @opindex fno-var-tracking-assignments-toggle
5503 Toggle @option{-fvar-tracking-assignments}, in the same way that
5504 @option{-gtoggle} toggles @option{-g}.
5506 @item -print-file-name=@var{library}
5507 @opindex print-file-name
5508 Print the full absolute name of the library file @var{library} that
5509 would be used when linking---and don't do anything else. With this
5510 option, GCC does not compile or link anything; it just prints the
5513 @item -print-multi-directory
5514 @opindex print-multi-directory
5515 Print the directory name corresponding to the multilib selected by any
5516 other switches present in the command line. This directory is supposed
5517 to exist in @env{GCC_EXEC_PREFIX}.
5519 @item -print-multi-lib
5520 @opindex print-multi-lib
5521 Print the mapping from multilib directory names to compiler switches
5522 that enable them. The directory name is separated from the switches by
5523 @samp{;}, and each switch starts with an @samp{@@} instead of the
5524 @samp{-}, without spaces between multiple switches. This is supposed to
5525 ease shell-processing.
5527 @item -print-prog-name=@var{program}
5528 @opindex print-prog-name
5529 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5531 @item -print-libgcc-file-name
5532 @opindex print-libgcc-file-name
5533 Same as @option{-print-file-name=libgcc.a}.
5535 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5536 but you do want to link with @file{libgcc.a}. You can do
5539 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5542 @item -print-search-dirs
5543 @opindex print-search-dirs
5544 Print the name of the configured installation directory and a list of
5545 program and library directories @command{gcc} will search---and don't do anything else.
5547 This is useful when @command{gcc} prints the error message
5548 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5549 To resolve this you either need to put @file{cpp0} and the other compiler
5550 components where @command{gcc} expects to find them, or you can set the environment
5551 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5552 Don't forget the trailing @samp{/}.
5553 @xref{Environment Variables}.
5555 @item -print-sysroot
5556 @opindex print-sysroot
5557 Print the target sysroot directory that will be used during
5558 compilation. This is the target sysroot specified either at configure
5559 time or using the @option{--sysroot} option, possibly with an extra
5560 suffix that depends on compilation options. If no target sysroot is
5561 specified, the option prints nothing.
5563 @item -print-sysroot-headers-suffix
5564 @opindex print-sysroot-headers-suffix
5565 Print the suffix added to the target sysroot when searching for
5566 headers, or give an error if the compiler is not configured with such
5567 a suffix---and don't do anything else.
5570 @opindex dumpmachine
5571 Print the compiler's target machine (for example,
5572 @samp{i686-pc-linux-gnu})---and don't do anything else.
5575 @opindex dumpversion
5576 Print the compiler version (for example, @samp{3.0})---and don't do
5581 Print the compiler's built-in specs---and don't do anything else. (This
5582 is used when GCC itself is being built.) @xref{Spec Files}.
5584 @item -feliminate-unused-debug-types
5585 @opindex feliminate-unused-debug-types
5586 Normally, when producing DWARF2 output, GCC will emit debugging
5587 information for all types declared in a compilation
5588 unit, regardless of whether or not they are actually used
5589 in that compilation unit. Sometimes this is useful, such as
5590 if, in the debugger, you want to cast a value to a type that is
5591 not actually used in your program (but is declared). More often,
5592 however, this results in a significant amount of wasted space.
5593 With this option, GCC will avoid producing debug symbol output
5594 for types that are nowhere used in the source file being compiled.
5597 @node Optimize Options
5598 @section Options That Control Optimization
5599 @cindex optimize options
5600 @cindex options, optimization
5602 These options control various sorts of optimizations.
5604 Without any optimization option, the compiler's goal is to reduce the
5605 cost of compilation and to make debugging produce the expected
5606 results. Statements are independent: if you stop the program with a
5607 breakpoint between statements, you can then assign a new value to any
5608 variable or change the program counter to any other statement in the
5609 function and get exactly the results you would expect from the source
5612 Turning on optimization flags makes the compiler attempt to improve
5613 the performance and/or code size at the expense of compilation time
5614 and possibly the ability to debug the program.
5616 The compiler performs optimization based on the knowledge it has of the
5617 program. Compiling multiple files at once to a single output file mode allows
5618 the compiler to use information gained from all of the files when compiling
5621 Not all optimizations are controlled directly by a flag. Only
5622 optimizations that have a flag are listed in this section.
5624 Depending on the target and how GCC was configured, a slightly different
5625 set of optimizations may be enabled at each @option{-O} level than
5626 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
5627 to find out the exact set of optimizations that are enabled at each level.
5628 @xref{Overall Options}, for examples.
5635 Optimize. Optimizing compilation takes somewhat more time, and a lot
5636 more memory for a large function.
5638 With @option{-O}, the compiler tries to reduce code size and execution
5639 time, without performing any optimizations that take a great deal of
5642 @option{-O} turns on the following optimization flags:
5645 -fcprop-registers @gol
5648 -fdelayed-branch @gol
5650 -fguess-branch-probability @gol
5651 -fif-conversion2 @gol
5652 -fif-conversion @gol
5653 -fipa-pure-const @gol
5654 -fipa-reference @gol
5656 -fsplit-wide-types @gol
5657 -ftree-builtin-call-dce @gol
5660 -ftree-copyrename @gol
5662 -ftree-dominator-opts @gol
5664 -ftree-forwprop @gol
5672 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5673 where doing so does not interfere with debugging.
5677 Optimize even more. GCC performs nearly all supported optimizations
5678 that do not involve a space-speed tradeoff.
5679 As compared to @option{-O}, this option increases both compilation time
5680 and the performance of the generated code.
5682 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5683 also turns on the following optimization flags:
5684 @gccoptlist{-fthread-jumps @gol
5685 -falign-functions -falign-jumps @gol
5686 -falign-loops -falign-labels @gol
5689 -fcse-follow-jumps -fcse-skip-blocks @gol
5690 -fdelete-null-pointer-checks @gol
5691 -fexpensive-optimizations @gol
5692 -fgcse -fgcse-lm @gol
5693 -finline-small-functions @gol
5694 -findirect-inlining @gol
5696 -foptimize-sibling-calls @gol
5699 -freorder-blocks -freorder-functions @gol
5700 -frerun-cse-after-loop @gol
5701 -fsched-interblock -fsched-spec @gol
5702 -fschedule-insns -fschedule-insns2 @gol
5703 -fstrict-aliasing -fstrict-overflow @gol
5704 -ftree-switch-conversion @gol
5708 Please note the warning under @option{-fgcse} about
5709 invoking @option{-O2} on programs that use computed gotos.
5713 Optimize yet more. @option{-O3} turns on all optimizations specified
5714 by @option{-O2} and also turns on the @option{-finline-functions},
5715 @option{-funswitch-loops}, @option{-fpredictive-commoning},
5716 @option{-fgcse-after-reload} and @option{-ftree-vectorize} options.
5720 Reduce compilation time and make debugging produce the expected
5721 results. This is the default.
5725 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5726 do not typically increase code size. It also performs further
5727 optimizations designed to reduce code size.
5729 @option{-Os} disables the following optimization flags:
5730 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5731 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5732 -fprefetch-loop-arrays -ftree-vect-loop-version}
5734 If you use multiple @option{-O} options, with or without level numbers,
5735 the last such option is the one that is effective.
5738 Options of the form @option{-f@var{flag}} specify machine-independent
5739 flags. Most flags have both positive and negative forms; the negative
5740 form of @option{-ffoo} would be @option{-fno-foo}. In the table
5741 below, only one of the forms is listed---the one you typically will
5742 use. You can figure out the other form by either removing @samp{no-}
5745 The following options control specific optimizations. They are either
5746 activated by @option{-O} options or are related to ones that are. You
5747 can use the following flags in the rare cases when ``fine-tuning'' of
5748 optimizations to be performed is desired.
5751 @item -fno-default-inline
5752 @opindex fno-default-inline
5753 Do not make member functions inline by default merely because they are
5754 defined inside the class scope (C++ only). Otherwise, when you specify
5755 @w{@option{-O}}, member functions defined inside class scope are compiled
5756 inline by default; i.e., you don't need to add @samp{inline} in front of
5757 the member function name.
5759 @item -fno-defer-pop
5760 @opindex fno-defer-pop
5761 Always pop the arguments to each function call as soon as that function
5762 returns. For machines which must pop arguments after a function call,
5763 the compiler normally lets arguments accumulate on the stack for several
5764 function calls and pops them all at once.
5766 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5768 @item -fforward-propagate
5769 @opindex fforward-propagate
5770 Perform a forward propagation pass on RTL@. The pass tries to combine two
5771 instructions and checks if the result can be simplified. If loop unrolling
5772 is active, two passes are performed and the second is scheduled after
5775 This option is enabled by default at optimization levels @option{-O},
5776 @option{-O2}, @option{-O3}, @option{-Os}.
5778 @item -fomit-frame-pointer
5779 @opindex fomit-frame-pointer
5780 Don't keep the frame pointer in a register for functions that
5781 don't need one. This avoids the instructions to save, set up and
5782 restore frame pointers; it also makes an extra register available
5783 in many functions. @strong{It also makes debugging impossible on
5786 On some machines, such as the VAX, this flag has no effect, because
5787 the standard calling sequence automatically handles the frame pointer
5788 and nothing is saved by pretending it doesn't exist. The
5789 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
5790 whether a target machine supports this flag. @xref{Registers,,Register
5791 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
5793 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5795 @item -foptimize-sibling-calls
5796 @opindex foptimize-sibling-calls
5797 Optimize sibling and tail recursive calls.
5799 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5803 Don't pay attention to the @code{inline} keyword. Normally this option
5804 is used to keep the compiler from expanding any functions inline.
5805 Note that if you are not optimizing, no functions can be expanded inline.
5807 @item -finline-small-functions
5808 @opindex finline-small-functions
5809 Integrate functions into their callers when their body is smaller than expected
5810 function call code (so overall size of program gets smaller). The compiler
5811 heuristically decides which functions are simple enough to be worth integrating
5814 Enabled at level @option{-O2}.
5816 @item -findirect-inlining
5817 @opindex findirect-inlining
5818 Inline also indirect calls that are discovered to be known at compile
5819 time thanks to previous inlining. This option has any effect only
5820 when inlining itself is turned on by the @option{-finline-functions}
5821 or @option{-finline-small-functions} options.
5823 Enabled at level @option{-O2}.
5825 @item -finline-functions
5826 @opindex finline-functions
5827 Integrate all simple functions into their callers. The compiler
5828 heuristically decides which functions are simple enough to be worth
5829 integrating in this way.
5831 If all calls to a given function are integrated, and the function is
5832 declared @code{static}, then the function is normally not output as
5833 assembler code in its own right.
5835 Enabled at level @option{-O3}.
5837 @item -finline-functions-called-once
5838 @opindex finline-functions-called-once
5839 Consider all @code{static} functions called once for inlining into their
5840 caller even if they are not marked @code{inline}. If a call to a given
5841 function is integrated, then the function is not output as assembler code
5844 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
5846 @item -fearly-inlining
5847 @opindex fearly-inlining
5848 Inline functions marked by @code{always_inline} and functions whose body seems
5849 smaller than the function call overhead early before doing
5850 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
5851 makes profiling significantly cheaper and usually inlining faster on programs
5852 having large chains of nested wrapper functions.
5858 Perform interprocedural scalar replacement of aggregates, removal of
5859 unused parameters and replacement of parameters passed by reference
5860 by parameters passed by value.
5862 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
5864 @item -finline-limit=@var{n}
5865 @opindex finline-limit
5866 By default, GCC limits the size of functions that can be inlined. This flag
5867 allows coarse control of this limit. @var{n} is the size of functions that
5868 can be inlined in number of pseudo instructions.
5870 Inlining is actually controlled by a number of parameters, which may be
5871 specified individually by using @option{--param @var{name}=@var{value}}.
5872 The @option{-finline-limit=@var{n}} option sets some of these parameters
5876 @item max-inline-insns-single
5877 is set to @var{n}/2.
5878 @item max-inline-insns-auto
5879 is set to @var{n}/2.
5882 See below for a documentation of the individual
5883 parameters controlling inlining and for the defaults of these parameters.
5885 @emph{Note:} there may be no value to @option{-finline-limit} that results
5886 in default behavior.
5888 @emph{Note:} pseudo instruction represents, in this particular context, an
5889 abstract measurement of function's size. In no way does it represent a count
5890 of assembly instructions and as such its exact meaning might change from one
5891 release to an another.
5893 @item -fkeep-inline-functions
5894 @opindex fkeep-inline-functions
5895 In C, emit @code{static} functions that are declared @code{inline}
5896 into the object file, even if the function has been inlined into all
5897 of its callers. This switch does not affect functions using the
5898 @code{extern inline} extension in GNU C89@. In C++, emit any and all
5899 inline functions into the object file.
5901 @item -fkeep-static-consts
5902 @opindex fkeep-static-consts
5903 Emit variables declared @code{static const} when optimization isn't turned
5904 on, even if the variables aren't referenced.
5906 GCC enables this option by default. If you want to force the compiler to
5907 check if the variable was referenced, regardless of whether or not
5908 optimization is turned on, use the @option{-fno-keep-static-consts} option.
5910 @item -fmerge-constants
5911 @opindex fmerge-constants
5912 Attempt to merge identical constants (string constants and floating point
5913 constants) across compilation units.
5915 This option is the default for optimized compilation if the assembler and
5916 linker support it. Use @option{-fno-merge-constants} to inhibit this
5919 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5921 @item -fmerge-all-constants
5922 @opindex fmerge-all-constants
5923 Attempt to merge identical constants and identical variables.
5925 This option implies @option{-fmerge-constants}. In addition to
5926 @option{-fmerge-constants} this considers e.g.@: even constant initialized
5927 arrays or initialized constant variables with integral or floating point
5928 types. Languages like C or C++ require each variable, including multiple
5929 instances of the same variable in recursive calls, to have distinct locations,
5930 so using this option will result in non-conforming
5933 @item -fmodulo-sched
5934 @opindex fmodulo-sched
5935 Perform swing modulo scheduling immediately before the first scheduling
5936 pass. This pass looks at innermost loops and reorders their
5937 instructions by overlapping different iterations.
5939 @item -fmodulo-sched-allow-regmoves
5940 @opindex fmodulo-sched-allow-regmoves
5941 Perform more aggressive SMS based modulo scheduling with register moves
5942 allowed. By setting this flag certain anti-dependences edges will be
5943 deleted which will trigger the generation of reg-moves based on the
5944 life-range analysis. This option is effective only with
5945 @option{-fmodulo-sched} enabled.
5947 @item -fno-branch-count-reg
5948 @opindex fno-branch-count-reg
5949 Do not use ``decrement and branch'' instructions on a count register,
5950 but instead generate a sequence of instructions that decrement a
5951 register, compare it against zero, then branch based upon the result.
5952 This option is only meaningful on architectures that support such
5953 instructions, which include x86, PowerPC, IA-64 and S/390.
5955 The default is @option{-fbranch-count-reg}.
5957 @item -fno-function-cse
5958 @opindex fno-function-cse
5959 Do not put function addresses in registers; make each instruction that
5960 calls a constant function contain the function's address explicitly.
5962 This option results in less efficient code, but some strange hacks
5963 that alter the assembler output may be confused by the optimizations
5964 performed when this option is not used.
5966 The default is @option{-ffunction-cse}
5968 @item -fno-zero-initialized-in-bss
5969 @opindex fno-zero-initialized-in-bss
5970 If the target supports a BSS section, GCC by default puts variables that
5971 are initialized to zero into BSS@. This can save space in the resulting
5974 This option turns off this behavior because some programs explicitly
5975 rely on variables going to the data section. E.g., so that the
5976 resulting executable can find the beginning of that section and/or make
5977 assumptions based on that.
5979 The default is @option{-fzero-initialized-in-bss}.
5981 @item -fmudflap -fmudflapth -fmudflapir
5985 @cindex bounds checking
5987 For front-ends that support it (C and C++), instrument all risky
5988 pointer/array dereferencing operations, some standard library
5989 string/heap functions, and some other associated constructs with
5990 range/validity tests. Modules so instrumented should be immune to
5991 buffer overflows, invalid heap use, and some other classes of C/C++
5992 programming errors. The instrumentation relies on a separate runtime
5993 library (@file{libmudflap}), which will be linked into a program if
5994 @option{-fmudflap} is given at link time. Run-time behavior of the
5995 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
5996 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
5999 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
6000 link if your program is multi-threaded. Use @option{-fmudflapir}, in
6001 addition to @option{-fmudflap} or @option{-fmudflapth}, if
6002 instrumentation should ignore pointer reads. This produces less
6003 instrumentation (and therefore faster execution) and still provides
6004 some protection against outright memory corrupting writes, but allows
6005 erroneously read data to propagate within a program.
6007 @item -fthread-jumps
6008 @opindex fthread-jumps
6009 Perform optimizations where we check to see if a jump branches to a
6010 location where another comparison subsumed by the first is found. If
6011 so, the first branch is redirected to either the destination of the
6012 second branch or a point immediately following it, depending on whether
6013 the condition is known to be true or false.
6015 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6017 @item -fsplit-wide-types
6018 @opindex fsplit-wide-types
6019 When using a type that occupies multiple registers, such as @code{long
6020 long} on a 32-bit system, split the registers apart and allocate them
6021 independently. This normally generates better code for those types,
6022 but may make debugging more difficult.
6024 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
6027 @item -fcse-follow-jumps
6028 @opindex fcse-follow-jumps
6029 In common subexpression elimination (CSE), scan through jump instructions
6030 when the target of the jump is not reached by any other path. For
6031 example, when CSE encounters an @code{if} statement with an
6032 @code{else} clause, CSE will follow the jump when the condition
6035 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6037 @item -fcse-skip-blocks
6038 @opindex fcse-skip-blocks
6039 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
6040 follow jumps which conditionally skip over blocks. When CSE
6041 encounters a simple @code{if} statement with no else clause,
6042 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6043 body of the @code{if}.
6045 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6047 @item -frerun-cse-after-loop
6048 @opindex frerun-cse-after-loop
6049 Re-run common subexpression elimination after loop optimizations has been
6052 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6056 Perform a global common subexpression elimination pass.
6057 This pass also performs global constant and copy propagation.
6059 @emph{Note:} When compiling a program using computed gotos, a GCC
6060 extension, you may get better runtime performance if you disable
6061 the global common subexpression elimination pass by adding
6062 @option{-fno-gcse} to the command line.
6064 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6068 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
6069 attempt to move loads which are only killed by stores into themselves. This
6070 allows a loop containing a load/store sequence to be changed to a load outside
6071 the loop, and a copy/store within the loop.
6073 Enabled by default when gcse is enabled.
6077 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6078 global common subexpression elimination. This pass will attempt to move
6079 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6080 loops containing a load/store sequence can be changed to a load before
6081 the loop and a store after the loop.
6083 Not enabled at any optimization level.
6087 When @option{-fgcse-las} is enabled, the global common subexpression
6088 elimination pass eliminates redundant loads that come after stores to the
6089 same memory location (both partial and full redundancies).
6091 Not enabled at any optimization level.
6093 @item -fgcse-after-reload
6094 @opindex fgcse-after-reload
6095 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6096 pass is performed after reload. The purpose of this pass is to cleanup
6099 @item -funsafe-loop-optimizations
6100 @opindex funsafe-loop-optimizations
6101 If given, the loop optimizer will assume that loop indices do not
6102 overflow, and that the loops with nontrivial exit condition are not
6103 infinite. This enables a wider range of loop optimizations even if
6104 the loop optimizer itself cannot prove that these assumptions are valid.
6105 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
6106 if it finds this kind of loop.
6108 @item -fcrossjumping
6109 @opindex fcrossjumping
6110 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6111 resulting code may or may not perform better than without cross-jumping.
6113 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6115 @item -fauto-inc-dec
6116 @opindex fauto-inc-dec
6117 Combine increments or decrements of addresses with memory accesses.
6118 This pass is always skipped on architectures that do not have
6119 instructions to support this. Enabled by default at @option{-O} and
6120 higher on architectures that support this.
6124 Perform dead code elimination (DCE) on RTL@.
6125 Enabled by default at @option{-O} and higher.
6129 Perform dead store elimination (DSE) on RTL@.
6130 Enabled by default at @option{-O} and higher.
6132 @item -fif-conversion
6133 @opindex fif-conversion
6134 Attempt to transform conditional jumps into branch-less equivalents. This
6135 include use of conditional moves, min, max, set flags and abs instructions, and
6136 some tricks doable by standard arithmetics. The use of conditional execution
6137 on chips where it is available is controlled by @code{if-conversion2}.
6139 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6141 @item -fif-conversion2
6142 @opindex fif-conversion2
6143 Use conditional execution (where available) to transform conditional jumps into
6144 branch-less equivalents.
6146 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6148 @item -fdelete-null-pointer-checks
6149 @opindex fdelete-null-pointer-checks
6150 Assume that programs cannot safely dereference null pointers, and that
6151 no code or data element resides there. This enables simple constant
6152 folding optimizations at all optimization levels. In addition, other
6153 optimization passes in GCC use this flag to control global dataflow
6154 analyses that eliminate useless checks for null pointers; these assume
6155 that if a pointer is checked after it has already been dereferenced,
6158 Note however that in some environments this assumption is not true.
6159 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6160 for programs which depend on that behavior.
6162 Some targets, especially embedded ones, disable this option at all levels.
6163 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6164 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6165 are enabled independently at different optimization levels.
6167 @item -fexpensive-optimizations
6168 @opindex fexpensive-optimizations
6169 Perform a number of minor optimizations that are relatively expensive.
6171 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6173 @item -foptimize-register-move
6175 @opindex foptimize-register-move
6177 Attempt to reassign register numbers in move instructions and as
6178 operands of other simple instructions in order to maximize the amount of
6179 register tying. This is especially helpful on machines with two-operand
6182 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6185 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6187 @item -fira-algorithm=@var{algorithm}
6188 Use specified coloring algorithm for the integrated register
6189 allocator. The @var{algorithm} argument should be @code{priority} or
6190 @code{CB}. The first algorithm specifies Chow's priority coloring,
6191 the second one specifies Chaitin-Briggs coloring. The second
6192 algorithm can be unimplemented for some architectures. If it is
6193 implemented, it is the default because Chaitin-Briggs coloring as a
6194 rule generates a better code.
6196 @item -fira-region=@var{region}
6197 Use specified regions for the integrated register allocator. The
6198 @var{region} argument should be one of @code{all}, @code{mixed}, or
6199 @code{one}. The first value means using all loops as register
6200 allocation regions, the second value which is the default means using
6201 all loops except for loops with small register pressure as the
6202 regions, and third one means using all function as a single region.
6203 The first value can give best result for machines with small size and
6204 irregular register set, the third one results in faster and generates
6205 decent code and the smallest size code, and the default value usually
6206 give the best results in most cases and for most architectures.
6208 @item -fira-coalesce
6209 @opindex fira-coalesce
6210 Do optimistic register coalescing. This option might be profitable for
6211 architectures with big regular register files.
6213 @item -fno-ira-share-save-slots
6214 @opindex fno-ira-share-save-slots
6215 Switch off sharing stack slots used for saving call used hard
6216 registers living through a call. Each hard register will get a
6217 separate stack slot and as a result function stack frame will be
6220 @item -fno-ira-share-spill-slots
6221 @opindex fno-ira-share-spill-slots
6222 Switch off sharing stack slots allocated for pseudo-registers. Each
6223 pseudo-register which did not get a hard register will get a separate
6224 stack slot and as a result function stack frame will be bigger.
6226 @item -fira-verbose=@var{n}
6227 @opindex fira-verbose
6228 Set up how verbose dump file for the integrated register allocator
6229 will be. Default value is 5. If the value is greater or equal to 10,
6230 the dump file will be stderr as if the value were @var{n} minus 10.
6232 @item -fdelayed-branch
6233 @opindex fdelayed-branch
6234 If supported for the target machine, attempt to reorder instructions
6235 to exploit instruction slots available after delayed branch
6238 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6240 @item -fschedule-insns
6241 @opindex fschedule-insns
6242 If supported for the target machine, attempt to reorder instructions to
6243 eliminate execution stalls due to required data being unavailable. This
6244 helps machines that have slow floating point or memory load instructions
6245 by allowing other instructions to be issued until the result of the load
6246 or floating point instruction is required.
6248 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6250 @item -fschedule-insns2
6251 @opindex fschedule-insns2
6252 Similar to @option{-fschedule-insns}, but requests an additional pass of
6253 instruction scheduling after register allocation has been done. This is
6254 especially useful on machines with a relatively small number of
6255 registers and where memory load instructions take more than one cycle.
6257 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6259 @item -fno-sched-interblock
6260 @opindex fno-sched-interblock
6261 Don't schedule instructions across basic blocks. This is normally
6262 enabled by default when scheduling before register allocation, i.e.@:
6263 with @option{-fschedule-insns} or at @option{-O2} or higher.
6265 @item -fno-sched-spec
6266 @opindex fno-sched-spec
6267 Don't allow speculative motion of non-load instructions. 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 -fsched-pressure
6272 @opindex fsched-pressure
6273 Enable register pressure sensitive insn scheduling before the register
6274 allocation. This only makes sense when scheduling before register
6275 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
6276 @option{-O2} or higher. Usage of this option can improve the
6277 generated code and decrease its size by preventing register pressure
6278 increase above the number of available hard registers and as a
6279 consequence register spills in the register allocation.
6281 @item -fsched-spec-load
6282 @opindex fsched-spec-load
6283 Allow speculative motion of some load instructions. This only makes
6284 sense when scheduling before register allocation, i.e.@: with
6285 @option{-fschedule-insns} or at @option{-O2} or higher.
6287 @item -fsched-spec-load-dangerous
6288 @opindex fsched-spec-load-dangerous
6289 Allow speculative motion of more 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-stalled-insns
6294 @itemx -fsched-stalled-insns=@var{n}
6295 @opindex fsched-stalled-insns
6296 Define how many insns (if any) can be moved prematurely from the queue
6297 of stalled insns into the ready list, during the second scheduling pass.
6298 @option{-fno-sched-stalled-insns} means that no insns will be moved
6299 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6300 on how many queued insns can be moved prematurely.
6301 @option{-fsched-stalled-insns} without a value is equivalent to
6302 @option{-fsched-stalled-insns=1}.
6304 @item -fsched-stalled-insns-dep
6305 @itemx -fsched-stalled-insns-dep=@var{n}
6306 @opindex fsched-stalled-insns-dep
6307 Define how many insn groups (cycles) will be examined for a dependency
6308 on a stalled insn that is candidate for premature removal from the queue
6309 of stalled insns. This has an effect only during the second scheduling pass,
6310 and only if @option{-fsched-stalled-insns} is used.
6311 @option{-fno-sched-stalled-insns-dep} is equivalent to
6312 @option{-fsched-stalled-insns-dep=0}.
6313 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6314 @option{-fsched-stalled-insns-dep=1}.
6316 @item -fsched2-use-superblocks
6317 @opindex fsched2-use-superblocks
6318 When scheduling after register allocation, do use superblock scheduling
6319 algorithm. Superblock scheduling allows motion across basic block boundaries
6320 resulting on faster schedules. This option is experimental, as not all machine
6321 descriptions used by GCC model the CPU closely enough to avoid unreliable
6322 results from the algorithm.
6324 This only makes sense when scheduling after register allocation, i.e.@: with
6325 @option{-fschedule-insns2} or at @option{-O2} or higher.
6327 @item -fsched-group-heuristic
6328 @opindex fsched-group-heuristic
6329 Enable the group heuristic in the scheduler. This heuristic favors
6330 the instruction that belongs to a schedule group. This is enabled
6331 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6332 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6334 @item -fsched-critical-path-heuristic
6335 @opindex fsched-critical-path-heuristic
6336 Enable the critical-path heuristic in the scheduler. This heuristic favors
6337 instructions on the critical path. This is enabled by default when
6338 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6339 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6341 @item -fsched-spec-insn-heuristic
6342 @opindex fsched-spec-insn-heuristic
6343 Enable the speculative instruction heuristic in the scheduler. This
6344 heuristic favors speculative instructions with greater dependency weakness.
6345 This is enabled by default when scheduling is enabled, i.e.@:
6346 with @option{-fschedule-insns} or @option{-fschedule-insns2}
6347 or at @option{-O2} or higher.
6349 @item -fsched-rank-heuristic
6350 @opindex fsched-rank-heuristic
6351 Enable the rank heuristic in the scheduler. This heuristic favors
6352 the instruction belonging to a basic block with greater size or frequency.
6353 This is enabled by default when scheduling is enabled, i.e.@:
6354 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6355 at @option{-O2} or higher.
6357 @item -fsched-last-insn-heuristic
6358 @opindex fsched-last-insn-heuristic
6359 Enable the last-instruction heuristic in the scheduler. This heuristic
6360 favors the instruction that is less dependent on the last instruction
6361 scheduled. This is enabled by default when scheduling is enabled,
6362 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6363 at @option{-O2} or higher.
6365 @item -fsched-dep-count-heuristic
6366 @opindex fsched-dep-count-heuristic
6367 Enable the dependent-count heuristic in the scheduler. This heuristic
6368 favors the instruction that has more instructions depending on it.
6369 This is enabled by default when scheduling is enabled, i.e.@:
6370 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6371 at @option{-O2} or higher.
6373 @item -fsched2-use-traces
6374 @opindex fsched2-use-traces
6375 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
6376 allocation and additionally perform code duplication in order to increase the
6377 size of superblocks using tracer pass. See @option{-ftracer} for details on
6380 This mode should produce faster but significantly longer programs. Also
6381 without @option{-fbranch-probabilities} the traces constructed may not
6382 match the reality and hurt the performance. This only makes
6383 sense when scheduling after register allocation, i.e.@: with
6384 @option{-fschedule-insns2} or at @option{-O2} or higher.
6386 @item -freschedule-modulo-scheduled-loops
6387 @opindex freschedule-modulo-scheduled-loops
6388 The modulo scheduling comes before the traditional scheduling, if a loop
6389 was modulo scheduled we may want to prevent the later scheduling passes
6390 from changing its schedule, we use this option to control that.
6392 @item -fselective-scheduling
6393 @opindex fselective-scheduling
6394 Schedule instructions using selective scheduling algorithm. Selective
6395 scheduling runs instead of the first scheduler pass.
6397 @item -fselective-scheduling2
6398 @opindex fselective-scheduling2
6399 Schedule instructions using selective scheduling algorithm. Selective
6400 scheduling runs instead of the second scheduler pass.
6402 @item -fsel-sched-pipelining
6403 @opindex fsel-sched-pipelining
6404 Enable software pipelining of innermost loops during selective scheduling.
6405 This option has no effect until one of @option{-fselective-scheduling} or
6406 @option{-fselective-scheduling2} is turned on.
6408 @item -fsel-sched-pipelining-outer-loops
6409 @opindex fsel-sched-pipelining-outer-loops
6410 When pipelining loops during selective scheduling, also pipeline outer loops.
6411 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6413 @item -fcaller-saves
6414 @opindex fcaller-saves
6415 Enable values to be allocated in registers that will be clobbered by
6416 function calls, by emitting extra instructions to save and restore the
6417 registers around such calls. Such allocation is done only when it
6418 seems to result in better code than would otherwise be produced.
6420 This option is always enabled by default on certain machines, usually
6421 those which have no call-preserved registers to use instead.
6423 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6425 @item -fconserve-stack
6426 @opindex fconserve-stack
6427 Attempt to minimize stack usage. The compiler will attempt to use less
6428 stack space, even if that makes the program slower. This option
6429 implies setting the @option{large-stack-frame} parameter to 100
6430 and the @option{large-stack-frame-growth} parameter to 400.
6432 @item -ftree-reassoc
6433 @opindex ftree-reassoc
6434 Perform reassociation on trees. This flag is enabled by default
6435 at @option{-O} and higher.
6439 Perform partial redundancy elimination (PRE) on trees. This flag is
6440 enabled by default at @option{-O2} and @option{-O3}.
6442 @item -ftree-forwprop
6443 @opindex ftree-forwprop
6444 Perform forward propagation on trees. This flag is enabled by default
6445 at @option{-O} and higher.
6449 Perform full redundancy elimination (FRE) on trees. The difference
6450 between FRE and PRE is that FRE only considers expressions
6451 that are computed on all paths leading to the redundant computation.
6452 This analysis is faster than PRE, though it exposes fewer redundancies.
6453 This flag is enabled by default at @option{-O} and higher.
6455 @item -ftree-phiprop
6456 @opindex ftree-phiprop
6457 Perform hoisting of loads from conditional pointers on trees. This
6458 pass is enabled by default at @option{-O} and higher.
6460 @item -ftree-copy-prop
6461 @opindex ftree-copy-prop
6462 Perform copy propagation on trees. This pass eliminates unnecessary
6463 copy operations. This flag is enabled by default at @option{-O} and
6466 @item -fipa-pure-const
6467 @opindex fipa-pure-const
6468 Discover which functions are pure or constant.
6469 Enabled by default at @option{-O} and higher.
6471 @item -fipa-reference
6472 @opindex fipa-reference
6473 Discover which static variables do not escape cannot escape the
6475 Enabled by default at @option{-O} and higher.
6477 @item -fipa-struct-reorg
6478 @opindex fipa-struct-reorg
6479 Perform structure reorganization optimization, that change C-like structures
6480 layout in order to better utilize spatial locality. This transformation is
6481 affective for programs containing arrays of structures. Available in two
6482 compilation modes: profile-based (enabled with @option{-fprofile-generate})
6483 or static (which uses built-in heuristics). Require @option{-fipa-type-escape}
6484 to provide the safety of this transformation. It works only in whole program
6485 mode, so it requires @option{-fwhole-program} and @option{-combine} to be
6486 enabled. Structures considered @samp{cold} by this transformation are not
6487 affected (see @option{--param struct-reorg-cold-struct-ratio=@var{value}}).
6489 With this flag, the program debug info reflects a new structure layout.
6493 Perform interprocedural pointer analysis. This option is experimental
6494 and does not affect generated code.
6498 Perform interprocedural constant propagation.
6499 This optimization analyzes the program to determine when values passed
6500 to functions are constants and then optimizes accordingly.
6501 This optimization can substantially increase performance
6502 if the application has constants passed to functions.
6503 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
6505 @item -fipa-cp-clone
6506 @opindex fipa-cp-clone
6507 Perform function cloning to make interprocedural constant propagation stronger.
6508 When enabled, interprocedural constant propagation will perform function cloning
6509 when externally visible function can be called with constant arguments.
6510 Because this optimization can create multiple copies of functions,
6511 it may significantly increase code size
6512 (see @option{--param ipcp-unit-growth=@var{value}}).
6513 This flag is enabled by default at @option{-O3}.
6515 @item -fipa-matrix-reorg
6516 @opindex fipa-matrix-reorg
6517 Perform matrix flattening and transposing.
6518 Matrix flattening tries to replace an @math{m}-dimensional matrix
6519 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
6520 This reduces the level of indirection needed for accessing the elements
6521 of the matrix. The second optimization is matrix transposing that
6522 attempts to change the order of the matrix's dimensions in order to
6523 improve cache locality.
6524 Both optimizations need the @option{-fwhole-program} flag.
6525 Transposing is enabled only if profiling information is available.
6529 Perform forward store motion on trees. This flag is
6530 enabled by default at @option{-O} and higher.
6534 Perform sparse conditional constant propagation (CCP) on trees. This
6535 pass only operates on local scalar variables and is enabled by default
6536 at @option{-O} and higher.
6538 @item -ftree-switch-conversion
6539 Perform conversion of simple initializations in a switch to
6540 initializations from a scalar array. This flag is enabled by default
6541 at @option{-O2} and higher.
6545 Perform dead code elimination (DCE) on trees. This flag is enabled by
6546 default at @option{-O} and higher.
6548 @item -ftree-builtin-call-dce
6549 @opindex ftree-builtin-call-dce
6550 Perform conditional dead code elimination (DCE) for calls to builtin functions
6551 that may set @code{errno} but are otherwise side-effect free. This flag is
6552 enabled by default at @option{-O2} and higher if @option{-Os} is not also
6555 @item -ftree-dominator-opts
6556 @opindex ftree-dominator-opts
6557 Perform a variety of simple scalar cleanups (constant/copy
6558 propagation, redundancy elimination, range propagation and expression
6559 simplification) based on a dominator tree traversal. This also
6560 performs jump threading (to reduce jumps to jumps). This flag is
6561 enabled by default at @option{-O} and higher.
6565 Perform dead store elimination (DSE) on trees. A dead store is a store into
6566 a memory location which will later be overwritten by another store without
6567 any intervening loads. In this case the earlier store can be deleted. This
6568 flag is enabled by default at @option{-O} and higher.
6572 Perform loop header copying on trees. This is beneficial since it increases
6573 effectiveness of code motion optimizations. It also saves one jump. This flag
6574 is enabled by default at @option{-O} and higher. It is not enabled
6575 for @option{-Os}, since it usually increases code size.
6577 @item -ftree-loop-optimize
6578 @opindex ftree-loop-optimize
6579 Perform loop optimizations on trees. This flag is enabled by default
6580 at @option{-O} and higher.
6582 @item -ftree-loop-linear
6583 @opindex ftree-loop-linear
6584 Perform linear loop transformations on tree. This flag can improve cache
6585 performance and allow further loop optimizations to take place.
6587 @item -floop-interchange
6588 Perform loop interchange transformations on loops. Interchanging two
6589 nested loops switches the inner and outer loops. For example, given a
6594 A(J, I) = A(J, I) * C
6598 loop interchange will transform the loop as if the user had written:
6602 A(J, I) = A(J, I) * C
6606 which can be beneficial when @code{N} is larger than the caches,
6607 because in Fortran, the elements of an array are stored in memory
6608 contiguously by column, and the original loop iterates over rows,
6609 potentially creating at each access a cache miss. This optimization
6610 applies to all the languages supported by GCC and is not limited to
6611 Fortran. To use this code transformation, GCC has to be configured
6612 with @option{--with-ppl} and @option{--with-cloog} to enable the
6613 Graphite loop transformation infrastructure.
6615 @item -floop-strip-mine
6616 Perform loop strip mining transformations on loops. Strip mining
6617 splits a loop into two nested loops. The outer loop has strides
6618 equal to the strip size and the inner loop has strides of the
6619 original loop within a strip. For example, given a loop like:
6625 loop strip mining will transform the loop as if the user had written:
6628 DO I = II, min (II + 3, N)
6633 This optimization applies to all the languages supported by GCC and is
6634 not limited to Fortran. To use this code transformation, GCC has to
6635 be configured with @option{--with-ppl} and @option{--with-cloog} to
6636 enable the Graphite loop transformation infrastructure.
6639 Perform loop blocking transformations on loops. Blocking strip mines
6640 each loop in the loop nest such that the memory accesses of the
6641 element loops fit inside caches. For example, given a loop like:
6645 A(J, I) = B(I) + C(J)
6649 loop blocking will transform the loop as if the user had written:
6653 DO I = II, min (II + 63, N)
6654 DO J = JJ, min (JJ + 63, M)
6655 A(J, I) = B(I) + C(J)
6661 which can be beneficial when @code{M} is larger than the caches,
6662 because the innermost loop will iterate over a smaller amount of data
6663 that can be kept in the caches. This optimization applies to all the
6664 languages supported by GCC and is not limited to Fortran. To use this
6665 code transformation, GCC has to be configured with @option{--with-ppl}
6666 and @option{--with-cloog} to enable the Graphite loop transformation
6669 @item -fgraphite-identity
6670 @opindex fgraphite-identity
6671 Enable the identity transformation for graphite. For every SCoP we generate
6672 the polyhedral representation and transform it back to gimple. Using
6673 @option{-fgraphite-identity} we can check the costs or benefits of the
6674 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
6675 are also performed by the code generator CLooG, like index splitting and
6676 dead code elimination in loops.
6678 @item -floop-parallelize-all
6679 Use the Graphite data dependence analysis to identify loops that can
6680 be parallelized. Parallelize all the loops that can be analyzed to
6681 not contain loop carried dependences without checking that it is
6682 profitable to parallelize the loops.
6684 @item -fcheck-data-deps
6685 @opindex fcheck-data-deps
6686 Compare the results of several data dependence analyzers. This option
6687 is used for debugging the data dependence analyzers.
6689 @item -ftree-loop-distribution
6690 Perform loop distribution. This flag can improve cache performance on
6691 big loop bodies and allow further loop optimizations, like
6692 parallelization or vectorization, to take place. For example, the loop
6709 @item -ftree-loop-im
6710 @opindex ftree-loop-im
6711 Perform loop invariant motion on trees. This pass moves only invariants that
6712 would be hard to handle at RTL level (function calls, operations that expand to
6713 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
6714 operands of conditions that are invariant out of the loop, so that we can use
6715 just trivial invariantness analysis in loop unswitching. The pass also includes
6718 @item -ftree-loop-ivcanon
6719 @opindex ftree-loop-ivcanon
6720 Create a canonical counter for number of iterations in the loop for that
6721 determining number of iterations requires complicated analysis. Later
6722 optimizations then may determine the number easily. Useful especially
6723 in connection with unrolling.
6727 Perform induction variable optimizations (strength reduction, induction
6728 variable merging and induction variable elimination) on trees.
6730 @item -ftree-parallelize-loops=n
6731 @opindex ftree-parallelize-loops
6732 Parallelize loops, i.e., split their iteration space to run in n threads.
6733 This is only possible for loops whose iterations are independent
6734 and can be arbitrarily reordered. The optimization is only
6735 profitable on multiprocessor machines, for loops that are CPU-intensive,
6736 rather than constrained e.g.@: by memory bandwidth. This option
6737 implies @option{-pthread}, and thus is only supported on targets
6738 that have support for @option{-pthread}.
6742 Perform function-local points-to analysis on trees. This flag is
6743 enabled by default at @option{-O} and higher.
6747 Perform scalar replacement of aggregates. This pass replaces structure
6748 references with scalars to prevent committing structures to memory too
6749 early. This flag is enabled by default at @option{-O} and higher.
6751 @item -ftree-copyrename
6752 @opindex ftree-copyrename
6753 Perform copy renaming on trees. This pass attempts to rename compiler
6754 temporaries to other variables at copy locations, usually resulting in
6755 variable names which more closely resemble the original variables. This flag
6756 is enabled by default at @option{-O} and higher.
6760 Perform temporary expression replacement during the SSA->normal phase. Single
6761 use/single def temporaries are replaced at their use location with their
6762 defining expression. This results in non-GIMPLE code, but gives the expanders
6763 much more complex trees to work on resulting in better RTL generation. This is
6764 enabled by default at @option{-O} and higher.
6766 @item -ftree-vectorize
6767 @opindex ftree-vectorize
6768 Perform loop vectorization on trees. This flag is enabled by default at
6771 @item -ftree-vect-loop-version
6772 @opindex ftree-vect-loop-version
6773 Perform loop versioning when doing loop vectorization on trees. When a loop
6774 appears to be vectorizable except that data alignment or data dependence cannot
6775 be determined at compile time then vectorized and non-vectorized versions of
6776 the loop are generated along with runtime checks for alignment or dependence
6777 to control which version is executed. This option is enabled by default
6778 except at level @option{-Os} where it is disabled.
6780 @item -fvect-cost-model
6781 @opindex fvect-cost-model
6782 Enable cost model for vectorization.
6786 Perform Value Range Propagation on trees. This is similar to the
6787 constant propagation pass, but instead of values, ranges of values are
6788 propagated. This allows the optimizers to remove unnecessary range
6789 checks like array bound checks and null pointer checks. This is
6790 enabled by default at @option{-O2} and higher. Null pointer check
6791 elimination is only done if @option{-fdelete-null-pointer-checks} is
6796 Perform tail duplication to enlarge superblock size. This transformation
6797 simplifies the control flow of the function allowing other optimizations to do
6800 @item -funroll-loops
6801 @opindex funroll-loops
6802 Unroll loops whose number of iterations can be determined at compile
6803 time or upon entry to the loop. @option{-funroll-loops} implies
6804 @option{-frerun-cse-after-loop}. This option makes code larger,
6805 and may or may not make it run faster.
6807 @item -funroll-all-loops
6808 @opindex funroll-all-loops
6809 Unroll all loops, even if their number of iterations is uncertain when
6810 the loop is entered. This usually makes programs run more slowly.
6811 @option{-funroll-all-loops} implies the same options as
6812 @option{-funroll-loops},
6814 @item -fsplit-ivs-in-unroller
6815 @opindex fsplit-ivs-in-unroller
6816 Enables expressing of values of induction variables in later iterations
6817 of the unrolled loop using the value in the first iteration. This breaks
6818 long dependency chains, thus improving efficiency of the scheduling passes.
6820 Combination of @option{-fweb} and CSE is often sufficient to obtain the
6821 same effect. However in cases the loop body is more complicated than
6822 a single basic block, this is not reliable. It also does not work at all
6823 on some of the architectures due to restrictions in the CSE pass.
6825 This optimization is enabled by default.
6827 @item -fvariable-expansion-in-unroller
6828 @opindex fvariable-expansion-in-unroller
6829 With this option, the compiler will create multiple copies of some
6830 local variables when unrolling a loop which can result in superior code.
6832 @item -fpredictive-commoning
6833 @opindex fpredictive-commoning
6834 Perform predictive commoning optimization, i.e., reusing computations
6835 (especially memory loads and stores) performed in previous
6836 iterations of loops.
6838 This option is enabled at level @option{-O3}.
6840 @item -fprefetch-loop-arrays
6841 @opindex fprefetch-loop-arrays
6842 If supported by the target machine, generate instructions to prefetch
6843 memory to improve the performance of loops that access large arrays.
6845 This option may generate better or worse code; results are highly
6846 dependent on the structure of loops within the source code.
6848 Disabled at level @option{-Os}.
6851 @itemx -fno-peephole2
6852 @opindex fno-peephole
6853 @opindex fno-peephole2
6854 Disable any machine-specific peephole optimizations. The difference
6855 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
6856 are implemented in the compiler; some targets use one, some use the
6857 other, a few use both.
6859 @option{-fpeephole} is enabled by default.
6860 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6862 @item -fno-guess-branch-probability
6863 @opindex fno-guess-branch-probability
6864 Do not guess branch probabilities using heuristics.
6866 GCC will use heuristics to guess branch probabilities if they are
6867 not provided by profiling feedback (@option{-fprofile-arcs}). These
6868 heuristics are based on the control flow graph. If some branch probabilities
6869 are specified by @samp{__builtin_expect}, then the heuristics will be
6870 used to guess branch probabilities for the rest of the control flow graph,
6871 taking the @samp{__builtin_expect} info into account. The interactions
6872 between the heuristics and @samp{__builtin_expect} can be complex, and in
6873 some cases, it may be useful to disable the heuristics so that the effects
6874 of @samp{__builtin_expect} are easier to understand.
6876 The default is @option{-fguess-branch-probability} at levels
6877 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6879 @item -freorder-blocks
6880 @opindex freorder-blocks
6881 Reorder basic blocks in the compiled function in order to reduce number of
6882 taken branches and improve code locality.
6884 Enabled at levels @option{-O2}, @option{-O3}.
6886 @item -freorder-blocks-and-partition
6887 @opindex freorder-blocks-and-partition
6888 In addition to reordering basic blocks in the compiled function, in order
6889 to reduce number of taken branches, partitions hot and cold basic blocks
6890 into separate sections of the assembly and .o files, to improve
6891 paging and cache locality performance.
6893 This optimization is automatically turned off in the presence of
6894 exception handling, for linkonce sections, for functions with a user-defined
6895 section attribute and on any architecture that does not support named
6898 @item -freorder-functions
6899 @opindex freorder-functions
6900 Reorder functions in the object file in order to
6901 improve code locality. This is implemented by using special
6902 subsections @code{.text.hot} for most frequently executed functions and
6903 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
6904 the linker so object file format must support named sections and linker must
6905 place them in a reasonable way.
6907 Also profile feedback must be available in to make this option effective. See
6908 @option{-fprofile-arcs} for details.
6910 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6912 @item -fstrict-aliasing
6913 @opindex fstrict-aliasing
6914 Allow the compiler to assume the strictest aliasing rules applicable to
6915 the language being compiled. For C (and C++), this activates
6916 optimizations based on the type of expressions. In particular, an
6917 object of one type is assumed never to reside at the same address as an
6918 object of a different type, unless the types are almost the same. For
6919 example, an @code{unsigned int} can alias an @code{int}, but not a
6920 @code{void*} or a @code{double}. A character type may alias any other
6923 @anchor{Type-punning}Pay special attention to code like this:
6936 The practice of reading from a different union member than the one most
6937 recently written to (called ``type-punning'') is common. Even with
6938 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
6939 is accessed through the union type. So, the code above will work as
6940 expected. @xref{Structures unions enumerations and bit-fields
6941 implementation}. However, this code might not:
6952 Similarly, access by taking the address, casting the resulting pointer
6953 and dereferencing the result has undefined behavior, even if the cast
6954 uses a union type, e.g.:
6958 return ((union a_union *) &d)->i;
6962 The @option{-fstrict-aliasing} option is enabled at levels
6963 @option{-O2}, @option{-O3}, @option{-Os}.
6965 @item -fstrict-overflow
6966 @opindex fstrict-overflow
6967 Allow the compiler to assume strict signed overflow rules, depending
6968 on the language being compiled. For C (and C++) this means that
6969 overflow when doing arithmetic with signed numbers is undefined, which
6970 means that the compiler may assume that it will not happen. This
6971 permits various optimizations. For example, the compiler will assume
6972 that an expression like @code{i + 10 > i} will always be true for
6973 signed @code{i}. This assumption is only valid if signed overflow is
6974 undefined, as the expression is false if @code{i + 10} overflows when
6975 using twos complement arithmetic. When this option is in effect any
6976 attempt to determine whether an operation on signed numbers will
6977 overflow must be written carefully to not actually involve overflow.
6979 This option also allows the compiler to assume strict pointer
6980 semantics: given a pointer to an object, if adding an offset to that
6981 pointer does not produce a pointer to the same object, the addition is
6982 undefined. This permits the compiler to conclude that @code{p + u >
6983 p} is always true for a pointer @code{p} and unsigned integer
6984 @code{u}. This assumption is only valid because pointer wraparound is
6985 undefined, as the expression is false if @code{p + u} overflows using
6986 twos complement arithmetic.
6988 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
6989 that integer signed overflow is fully defined: it wraps. When
6990 @option{-fwrapv} is used, there is no difference between
6991 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
6992 integers. With @option{-fwrapv} certain types of overflow are
6993 permitted. For example, if the compiler gets an overflow when doing
6994 arithmetic on constants, the overflowed value can still be used with
6995 @option{-fwrapv}, but not otherwise.
6997 The @option{-fstrict-overflow} option is enabled at levels
6998 @option{-O2}, @option{-O3}, @option{-Os}.
7000 @item -falign-functions
7001 @itemx -falign-functions=@var{n}
7002 @opindex falign-functions
7003 Align the start of functions to the next power-of-two greater than
7004 @var{n}, skipping up to @var{n} bytes. For instance,
7005 @option{-falign-functions=32} aligns functions to the next 32-byte
7006 boundary, but @option{-falign-functions=24} would align to the next
7007 32-byte boundary only if this can be done by skipping 23 bytes or less.
7009 @option{-fno-align-functions} and @option{-falign-functions=1} are
7010 equivalent and mean that functions will not be aligned.
7012 Some assemblers only support this flag when @var{n} is a power of two;
7013 in that case, it is rounded up.
7015 If @var{n} is not specified or is zero, use a machine-dependent default.
7017 Enabled at levels @option{-O2}, @option{-O3}.
7019 @item -falign-labels
7020 @itemx -falign-labels=@var{n}
7021 @opindex falign-labels
7022 Align all branch targets to a power-of-two boundary, skipping up to
7023 @var{n} bytes like @option{-falign-functions}. This option can easily
7024 make code slower, because it must insert dummy operations for when the
7025 branch target is reached in the usual flow of the code.
7027 @option{-fno-align-labels} and @option{-falign-labels=1} are
7028 equivalent and mean that labels will not be aligned.
7030 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
7031 are greater than this value, then their values are used instead.
7033 If @var{n} is not specified or is zero, use a machine-dependent default
7034 which is very likely to be @samp{1}, meaning no alignment.
7036 Enabled at levels @option{-O2}, @option{-O3}.
7039 @itemx -falign-loops=@var{n}
7040 @opindex falign-loops
7041 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7042 like @option{-falign-functions}. The hope is that the loop will be
7043 executed many times, which will make up for any execution of the dummy
7046 @option{-fno-align-loops} and @option{-falign-loops=1} are
7047 equivalent and mean that loops will not be aligned.
7049 If @var{n} is not specified or is zero, use a machine-dependent default.
7051 Enabled at levels @option{-O2}, @option{-O3}.
7054 @itemx -falign-jumps=@var{n}
7055 @opindex falign-jumps
7056 Align branch targets to a power-of-two boundary, for branch targets
7057 where the targets can only be reached by jumping, skipping up to @var{n}
7058 bytes like @option{-falign-functions}. In this case, no dummy operations
7061 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7062 equivalent and mean that loops will not be aligned.
7064 If @var{n} is not specified or is zero, use a machine-dependent default.
7066 Enabled at levels @option{-O2}, @option{-O3}.
7068 @item -funit-at-a-time
7069 @opindex funit-at-a-time
7070 This option is left for compatibility reasons. @option{-funit-at-a-time}
7071 has no effect, while @option{-fno-unit-at-a-time} implies
7072 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7076 @item -fno-toplevel-reorder
7077 @opindex fno-toplevel-reorder
7078 Do not reorder top-level functions, variables, and @code{asm}
7079 statements. Output them in the same order that they appear in the
7080 input file. When this option is used, unreferenced static variables
7081 will not be removed. This option is intended to support existing code
7082 which relies on a particular ordering. For new code, it is better to
7085 Enabled at level @option{-O0}. When disabled explicitly, it also imply
7086 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
7091 Constructs webs as commonly used for register allocation purposes and assign
7092 each web individual pseudo register. This allows the register allocation pass
7093 to operate on pseudos directly, but also strengthens several other optimization
7094 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7095 however, make debugging impossible, since variables will no longer stay in a
7098 Enabled by default with @option{-funroll-loops}.
7100 @item -fwhole-program
7101 @opindex fwhole-program
7102 Assume that the current compilation unit represents the whole program being
7103 compiled. All public functions and variables with the exception of @code{main}
7104 and those merged by attribute @code{externally_visible} become static functions
7105 and in effect are optimized more aggressively by interprocedural optimizers.
7106 While this option is equivalent to proper use of the @code{static} keyword for
7107 programs consisting of a single file, in combination with option
7108 @option{-combine}, @option{-flto} or @option{-fwhopr} this flag can be used to
7109 compile many smaller scale programs since the functions and variables become
7110 local for the whole combined compilation unit, not for the single source file
7113 This option implies @option{-fwhole-file} for Fortran programs.
7117 This option runs the standard link-time optimizer. When invoked
7118 with source code, it generates GIMPLE (one of GCC's internal
7119 representations) and writes it to special ELF sections in the object
7120 file. When the object files are linked together, all the function
7121 bodies are read from these ELF sections and instantiated as if they
7122 had been part of the same translation unit.
7124 To use the link-timer optimizer, @option{-flto} needs to be specified at
7125 compile time and during the final link. For example,
7128 gcc -c -O2 -flto foo.c
7129 gcc -c -O2 -flto bar.c
7130 gcc -o myprog -flto -O2 foo.o bar.o
7133 The first two invocations to GCC will save a bytecode representation
7134 of GIMPLE into special ELF sections inside @file{foo.o} and
7135 @file{bar.o}. The final invocation will read the GIMPLE bytecode from
7136 @file{foo.o} and @file{bar.o}, merge the two files into a single
7137 internal image, and compile the result as usual. Since both
7138 @file{foo.o} and @file{bar.o} are merged into a single image, this
7139 causes all the inter-procedural analyses and optimizations in GCC to
7140 work across the two files as if they were a single one. This means,
7141 for example, that the inliner will be able to inline functions in
7142 @file{bar.o} into functions in @file{foo.o} and vice-versa.
7144 Another (simpler) way to enable link-time optimization is,
7147 gcc -o myprog -flto -O2 foo.c bar.c
7150 The above will generate bytecode for @file{foo.c} and @file{bar.c},
7151 merge them together into a single GIMPLE representation and optimize
7152 them as usual to produce @file{myprog}.
7154 The only important thing to keep in mind is that to enable link-time
7155 optimizations the @option{-flto} flag needs to be passed to both the
7156 compile and the link commands.
7158 Note that when a file is compiled with @option{-flto}, the generated
7159 object file will be larger than a regular object file because it will
7160 contain GIMPLE bytecodes and the usual final code. This means that
7161 object files with LTO information can be linked as a normal object
7162 file. So, in the previous example, if the final link is done with
7165 gcc -o myprog foo.o bar.o
7168 The only difference will be that no inter-procedural optimizations
7169 will be applied to produce @file{myprog}. The two object files
7170 @file{foo.o} and @file{bar.o} will be simply sent to the regular
7173 Additionally, the optimization flags used to compile individual files
7174 are not necessarily related to those used at link-time. For instance,
7177 gcc -c -O0 -flto foo.c
7178 gcc -c -O0 -flto bar.c
7179 gcc -o myprog -flto -O3 foo.o bar.o
7182 This will produce individual object files with unoptimized assembler
7183 code, but the resulting binary @file{myprog} will be optimized at
7184 @option{-O3}. Now, if the final binary is generated without
7185 @option{-flto}, then @file{myprog} will not be optimized.
7187 When producing the final binary with @option{-flto}, GCC will only
7188 apply link-time optimizations to those files that contain bytecode.
7189 Therefore, you can mix and match object files and libraries with
7190 GIMPLE bytecodes and final object code. GCC will automatically select
7191 which files to optimize in LTO mode and which files to link without
7194 There are some code generation flags that GCC will preserve when
7195 generating bytecodes, as they need to be used during the final link
7196 stage. Currently, the following options are saved into the GIMPLE
7197 bytecode files: @option{-fPIC}, @option{-fcommon} and all the
7198 @option{-m} target flags.
7200 At link time, these options are read-in and reapplied. Note that the
7201 current implementation makes no attempt at recognizing conflicting
7202 values for these options. If two or more files have a conflicting
7203 value (e.g., one file is compiled with @option{-fPIC} and another
7204 isn't), the compiler will simply use the last value read from the
7205 bytecode files. It is recommended, then, that all the files
7206 participating in the same link be compiled with the same options.
7208 Another feature of LTO is that it is possible to apply interprocedural
7209 optimizations on files written in different languages. This requires
7210 some support in the language front end. Currently, the C, C++ and
7211 Fortran front ends are capable of emitting GIMPLE bytecodes, so
7212 something like this should work
7217 gfortran -c -flto baz.f90
7218 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
7221 Notice that the final link is done with @command{g++} to get the C++
7222 runtime libraries and @option{-lgfortran} is added to get the Fortran
7223 runtime libraries. In general, when mixing languages in LTO mode, you
7224 should use the same link command used when mixing languages in a
7225 regular (non-LTO) compilation. This means that if your build process
7226 was mixing languages before, all you need to add is @option{-flto} to
7227 all the compile and link commands.
7229 If object files containing GIMPLE bytecode are stored in a library
7230 archive, say @file{libfoo.a}, it is possible to extract and use them
7231 in an LTO link if you are using @command{gold} as the linker (which,
7232 in turn requires GCC to be configured with @option{--enable-gold}).
7233 To enable this feature, use the flag @option{-use-linker-plugin} at
7237 gcc -o myprog -O2 -flto -use-linker-plugin a.o b.o -lfoo
7240 With the linker plugin enabled, @command{gold} will extract the needed
7241 GIMPLE files from @file{libfoo.a} and pass them on to the running GCC
7242 to make them part of the aggregated GIMPLE image to be optimized.
7244 If you are not using @command{gold} and/or do not specify
7245 @option{-use-linker-plugin} then the objects inside @file{libfoo.a}
7246 will be extracted and linked as usual, but they will not participate
7247 in the LTO optimization process.
7249 Link time optimizations do not require the presence of the whole
7250 program to operate. If the program does not require any symbols to
7251 be exported, it is possible to combine @option{-flto} and
7252 @option{-fwhopr} with @option{-fwhole-program} to allow the
7253 interprocedural optimizers to use more aggressive assumptions which
7254 may lead to improved optimization opportunities.
7256 Regarding portability: the current implementation of LTO makes no
7257 attempt at generating bytecode that can be ported between different
7258 types of hosts. The bytecode files are versioned and there is a
7259 strict version check, so bytecode files generated in one version of
7260 GCC will not work with an older/newer version of GCC.
7262 This option is disabled by default.
7266 This option is identical in functionality to @option{-flto} but it
7267 differs in how the final link stage is executed. Instead of loading
7268 all the function bodies in memory, the callgraph is analyzed and
7269 optimization decisions are made (whole program analysis or WPA). Once
7270 optimization decisions are made, the callgraph is partitioned and the
7271 different sections are compiled separately (local transformations or
7272 LTRANS)@. This process allows optimizations on very large programs
7273 that otherwise would not fit in memory. This option enables
7274 @option{-fwpa} and @option{-fltrans} automatically.
7276 Disabled by default.
7280 This is an internal option used by GCC when compiling with
7281 @option{-fwhopr}. You should never need to use it.
7283 This option runs the link-time optimizer in the whole-program-analysis
7284 (WPA) mode, which reads in summary information from all inputs and
7285 performs a whole-program analysis based on summary information only.
7286 It generates object files for subsequent runs of the link-time
7287 optimizer where individual object files are optimized using both
7288 summary information from the WPA mode and the actual function bodies.
7289 It then drives the LTRANS phase.
7291 Disabled by default.
7295 This is an internal option used by GCC when compiling with
7296 @option{-fwhopr}. You should never need to use it.
7298 This option runs the link-time optimizer in the local-transformation (LTRANS)
7299 mode, which reads in output from a previous run of the LTO in WPA mode.
7300 In the LTRANS mode, LTO optimizes an object and produces the final assembly.
7302 Disabled by default.
7304 @item -fltrans-output-list=@var{file}
7305 @opindex fltrans-output-list
7306 This is an internal option used by GCC when compiling with
7307 @option{-fwhopr}. You should never need to use it.
7309 This option specifies a file to which the names of LTRANS output files are
7310 written. This option is only meaningful in conjunction with @option{-fwpa}.
7312 Disabled by default.
7314 @item -flto-compression-level=@var{n}
7315 This option specifies the level of compression used for intermediate
7316 language written to LTO object files, and is only meaningful in
7317 conjunction with LTO mode (@option{-fwhopr}, @option{-flto}). Valid
7318 values are 0 (no compression) to 9 (maximum compression). Values
7319 outside this range are clamped to either 0 or 9. If the option is not
7320 given, a default balanced compression setting is used.
7323 Prints a report with internal details on the workings of the link-time
7324 optimizer. The contents of this report vary from version to version,
7325 it is meant to be useful to GCC developers when processing object
7326 files in LTO mode (via @option{-fwhopr} or @option{-flto}).
7328 Disabled by default.
7330 @item -use-linker-plugin
7331 Enables the extraction of objects with GIMPLE bytecode information
7332 from library archives. This option relies on features available only
7333 in @command{gold}, so to use this you must configure GCC with
7334 @option{--enable-gold}. See @option{-flto} for a description on the
7335 effect of this flag and how to use it.
7337 Disabled by default.
7339 @item -fcprop-registers
7340 @opindex fcprop-registers
7341 After register allocation and post-register allocation instruction splitting,
7342 we perform a copy-propagation pass to try to reduce scheduling dependencies
7343 and occasionally eliminate the copy.
7345 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7347 @item -fprofile-correction
7348 @opindex fprofile-correction
7349 Profiles collected using an instrumented binary for multi-threaded programs may
7350 be inconsistent due to missed counter updates. When this option is specified,
7351 GCC will use heuristics to correct or smooth out such inconsistencies. By
7352 default, GCC will emit an error message when an inconsistent profile is detected.
7354 @item -fprofile-dir=@var{path}
7355 @opindex fprofile-dir
7357 Set the directory to search the profile data files in to @var{path}.
7358 This option affects only the profile data generated by
7359 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
7360 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
7361 and its related options.
7362 By default, GCC will use the current directory as @var{path}
7363 thus the profile data file will appear in the same directory as the object file.
7365 @item -fprofile-generate
7366 @itemx -fprofile-generate=@var{path}
7367 @opindex fprofile-generate
7369 Enable options usually used for instrumenting application to produce
7370 profile useful for later recompilation with profile feedback based
7371 optimization. You must use @option{-fprofile-generate} both when
7372 compiling and when linking your program.
7374 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
7376 If @var{path} is specified, GCC will look at the @var{path} to find
7377 the profile feedback data files. See @option{-fprofile-dir}.
7380 @itemx -fprofile-use=@var{path}
7381 @opindex fprofile-use
7382 Enable profile feedback directed optimizations, and optimizations
7383 generally profitable only with profile feedback available.
7385 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
7386 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
7388 By default, GCC emits an error message if the feedback profiles do not
7389 match the source code. This error can be turned into a warning by using
7390 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
7393 If @var{path} is specified, GCC will look at the @var{path} to find
7394 the profile feedback data files. See @option{-fprofile-dir}.
7397 The following options control compiler behavior regarding floating
7398 point arithmetic. These options trade off between speed and
7399 correctness. All must be specifically enabled.
7403 @opindex ffloat-store
7404 Do not store floating point variables in registers, and inhibit other
7405 options that might change whether a floating point value is taken from a
7408 @cindex floating point precision
7409 This option prevents undesirable excess precision on machines such as
7410 the 68000 where the floating registers (of the 68881) keep more
7411 precision than a @code{double} is supposed to have. Similarly for the
7412 x86 architecture. For most programs, the excess precision does only
7413 good, but a few programs rely on the precise definition of IEEE floating
7414 point. Use @option{-ffloat-store} for such programs, after modifying
7415 them to store all pertinent intermediate computations into variables.
7417 @item -fexcess-precision=@var{style}
7418 @opindex fexcess-precision
7419 This option allows further control over excess precision on machines
7420 where floating-point registers have more precision than the IEEE
7421 @code{float} and @code{double} types and the processor does not
7422 support operations rounding to those types. By default,
7423 @option{-fexcess-precision=fast} is in effect; this means that
7424 operations are carried out in the precision of the registers and that
7425 it is unpredictable when rounding to the types specified in the source
7426 code takes place. When compiling C, if
7427 @option{-fexcess-precision=standard} is specified then excess
7428 precision will follow the rules specified in ISO C99; in particular,
7429 both casts and assignments cause values to be rounded to their
7430 semantic types (whereas @option{-ffloat-store} only affects
7431 assignments). This option is enabled by default for C if a strict
7432 conformance option such as @option{-std=c99} is used.
7435 @option{-fexcess-precision=standard} is not implemented for languages
7436 other than C, and has no effect if
7437 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
7438 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
7439 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
7440 semantics apply without excess precision, and in the latter, rounding
7445 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
7446 @option{-ffinite-math-only}, @option{-fno-rounding-math},
7447 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
7449 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
7451 This option is not turned on by any @option{-O} option since
7452 it can result in incorrect output for programs which depend on
7453 an exact implementation of IEEE or ISO rules/specifications for
7454 math functions. It may, however, yield faster code for programs
7455 that do not require the guarantees of these specifications.
7457 @item -fno-math-errno
7458 @opindex fno-math-errno
7459 Do not set ERRNO after calling math functions that are executed
7460 with a single instruction, e.g., sqrt. A program that relies on
7461 IEEE exceptions for math error handling may want to use this flag
7462 for speed while maintaining IEEE arithmetic compatibility.
7464 This option is not turned on by any @option{-O} option since
7465 it can result in incorrect output for programs which depend on
7466 an exact implementation of IEEE or ISO rules/specifications for
7467 math functions. It may, however, yield faster code for programs
7468 that do not require the guarantees of these specifications.
7470 The default is @option{-fmath-errno}.
7472 On Darwin systems, the math library never sets @code{errno}. There is
7473 therefore no reason for the compiler to consider the possibility that
7474 it might, and @option{-fno-math-errno} is the default.
7476 @item -funsafe-math-optimizations
7477 @opindex funsafe-math-optimizations
7479 Allow optimizations for floating-point arithmetic that (a) assume
7480 that arguments and results are valid and (b) may violate IEEE or
7481 ANSI standards. When used at link-time, it may include libraries
7482 or startup files that change the default FPU control word or other
7483 similar optimizations.
7485 This option is not turned on by any @option{-O} option since
7486 it can result in incorrect output for programs which depend on
7487 an exact implementation of IEEE or ISO rules/specifications for
7488 math functions. It may, however, yield faster code for programs
7489 that do not require the guarantees of these specifications.
7490 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
7491 @option{-fassociative-math} and @option{-freciprocal-math}.
7493 The default is @option{-fno-unsafe-math-optimizations}.
7495 @item -fassociative-math
7496 @opindex fassociative-math
7498 Allow re-association of operands in series of floating-point operations.
7499 This violates the ISO C and C++ language standard by possibly changing
7500 computation result. NOTE: re-ordering may change the sign of zero as
7501 well as ignore NaNs and inhibit or create underflow or overflow (and
7502 thus cannot be used on a code which relies on rounding behavior like
7503 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
7504 and thus may not be used when ordered comparisons are required.
7505 This option requires that both @option{-fno-signed-zeros} and
7506 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
7507 much sense with @option{-frounding-math}.
7509 The default is @option{-fno-associative-math}.
7511 @item -freciprocal-math
7512 @opindex freciprocal-math
7514 Allow the reciprocal of a value to be used instead of dividing by
7515 the value if this enables optimizations. For example @code{x / y}
7516 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
7517 is subject to common subexpression elimination. Note that this loses
7518 precision and increases the number of flops operating on the value.
7520 The default is @option{-fno-reciprocal-math}.
7522 @item -ffinite-math-only
7523 @opindex ffinite-math-only
7524 Allow optimizations for floating-point arithmetic that assume
7525 that arguments and results are not NaNs or +-Infs.
7527 This option is not turned on by any @option{-O} option since
7528 it can result in incorrect output for programs which depend on
7529 an exact implementation of IEEE or ISO rules/specifications for
7530 math functions. It may, however, yield faster code for programs
7531 that do not require the guarantees of these specifications.
7533 The default is @option{-fno-finite-math-only}.
7535 @item -fno-signed-zeros
7536 @opindex fno-signed-zeros
7537 Allow optimizations for floating point arithmetic that ignore the
7538 signedness of zero. IEEE arithmetic specifies the behavior of
7539 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
7540 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
7541 This option implies that the sign of a zero result isn't significant.
7543 The default is @option{-fsigned-zeros}.
7545 @item -fno-trapping-math
7546 @opindex fno-trapping-math
7547 Compile code assuming that floating-point operations cannot generate
7548 user-visible traps. These traps include division by zero, overflow,
7549 underflow, inexact result and invalid operation. This option requires
7550 that @option{-fno-signaling-nans} be in effect. Setting this option may
7551 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
7553 This option should never be turned on by any @option{-O} option since
7554 it can result in incorrect output for programs which depend on
7555 an exact implementation of IEEE or ISO rules/specifications for
7558 The default is @option{-ftrapping-math}.
7560 @item -frounding-math
7561 @opindex frounding-math
7562 Disable transformations and optimizations that assume default floating
7563 point rounding behavior. This is round-to-zero for all floating point
7564 to integer conversions, and round-to-nearest for all other arithmetic
7565 truncations. This option should be specified for programs that change
7566 the FP rounding mode dynamically, or that may be executed with a
7567 non-default rounding mode. This option disables constant folding of
7568 floating point expressions at compile-time (which may be affected by
7569 rounding mode) and arithmetic transformations that are unsafe in the
7570 presence of sign-dependent rounding modes.
7572 The default is @option{-fno-rounding-math}.
7574 This option is experimental and does not currently guarantee to
7575 disable all GCC optimizations that are affected by rounding mode.
7576 Future versions of GCC may provide finer control of this setting
7577 using C99's @code{FENV_ACCESS} pragma. This command line option
7578 will be used to specify the default state for @code{FENV_ACCESS}.
7580 @item -fsignaling-nans
7581 @opindex fsignaling-nans
7582 Compile code assuming that IEEE signaling NaNs may generate user-visible
7583 traps during floating-point operations. Setting this option disables
7584 optimizations that may change the number of exceptions visible with
7585 signaling NaNs. This option implies @option{-ftrapping-math}.
7587 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
7590 The default is @option{-fno-signaling-nans}.
7592 This option is experimental and does not currently guarantee to
7593 disable all GCC optimizations that affect signaling NaN behavior.
7595 @item -fsingle-precision-constant
7596 @opindex fsingle-precision-constant
7597 Treat floating point constant as single precision constant instead of
7598 implicitly converting it to double precision constant.
7600 @item -fcx-limited-range
7601 @opindex fcx-limited-range
7602 When enabled, this option states that a range reduction step is not
7603 needed when performing complex division. Also, there is no checking
7604 whether the result of a complex multiplication or division is @code{NaN
7605 + I*NaN}, with an attempt to rescue the situation in that case. The
7606 default is @option{-fno-cx-limited-range}, but is enabled by
7607 @option{-ffast-math}.
7609 This option controls the default setting of the ISO C99
7610 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
7613 @item -fcx-fortran-rules
7614 @opindex fcx-fortran-rules
7615 Complex multiplication and division follow Fortran rules. Range
7616 reduction is done as part of complex division, but there is no checking
7617 whether the result of a complex multiplication or division is @code{NaN
7618 + I*NaN}, with an attempt to rescue the situation in that case.
7620 The default is @option{-fno-cx-fortran-rules}.
7624 The following options control optimizations that may improve
7625 performance, but are not enabled by any @option{-O} options. This
7626 section includes experimental options that may produce broken code.
7629 @item -fbranch-probabilities
7630 @opindex fbranch-probabilities
7631 After running a program compiled with @option{-fprofile-arcs}
7632 (@pxref{Debugging Options,, Options for Debugging Your Program or
7633 @command{gcc}}), you can compile it a second time using
7634 @option{-fbranch-probabilities}, to improve optimizations based on
7635 the number of times each branch was taken. When the program
7636 compiled with @option{-fprofile-arcs} exits it saves arc execution
7637 counts to a file called @file{@var{sourcename}.gcda} for each source
7638 file. The information in this data file is very dependent on the
7639 structure of the generated code, so you must use the same source code
7640 and the same optimization options for both compilations.
7642 With @option{-fbranch-probabilities}, GCC puts a
7643 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
7644 These can be used to improve optimization. Currently, they are only
7645 used in one place: in @file{reorg.c}, instead of guessing which path a
7646 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
7647 exactly determine which path is taken more often.
7649 @item -fprofile-values
7650 @opindex fprofile-values
7651 If combined with @option{-fprofile-arcs}, it adds code so that some
7652 data about values of expressions in the program is gathered.
7654 With @option{-fbranch-probabilities}, it reads back the data gathered
7655 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
7656 notes to instructions for their later usage in optimizations.
7658 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
7662 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
7663 a code to gather information about values of expressions.
7665 With @option{-fbranch-probabilities}, it reads back the data gathered
7666 and actually performs the optimizations based on them.
7667 Currently the optimizations include specialization of division operation
7668 using the knowledge about the value of the denominator.
7670 @item -frename-registers
7671 @opindex frename-registers
7672 Attempt to avoid false dependencies in scheduled code by making use
7673 of registers left over after register allocation. This optimization
7674 will most benefit processors with lots of registers. Depending on the
7675 debug information format adopted by the target, however, it can
7676 make debugging impossible, since variables will no longer stay in
7677 a ``home register''.
7679 Enabled by default with @option{-funroll-loops}.
7683 Perform tail duplication to enlarge superblock size. This transformation
7684 simplifies the control flow of the function allowing other optimizations to do
7687 Enabled with @option{-fprofile-use}.
7689 @item -funroll-loops
7690 @opindex funroll-loops
7691 Unroll loops whose number of iterations can be determined at compile time or
7692 upon entry to the loop. @option{-funroll-loops} implies
7693 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
7694 It also turns on complete loop peeling (i.e.@: complete removal of loops with
7695 small constant number of iterations). This option makes code larger, and may
7696 or may not make it run faster.
7698 Enabled with @option{-fprofile-use}.
7700 @item -funroll-all-loops
7701 @opindex funroll-all-loops
7702 Unroll all loops, even if their number of iterations is uncertain when
7703 the loop is entered. This usually makes programs run more slowly.
7704 @option{-funroll-all-loops} implies the same options as
7705 @option{-funroll-loops}.
7708 @opindex fpeel-loops
7709 Peels the loops for that there is enough information that they do not
7710 roll much (from profile feedback). It also turns on complete loop peeling
7711 (i.e.@: complete removal of loops with small constant number of iterations).
7713 Enabled with @option{-fprofile-use}.
7715 @item -fmove-loop-invariants
7716 @opindex fmove-loop-invariants
7717 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
7718 at level @option{-O1}
7720 @item -funswitch-loops
7721 @opindex funswitch-loops
7722 Move branches with loop invariant conditions out of the loop, with duplicates
7723 of the loop on both branches (modified according to result of the condition).
7725 @item -ffunction-sections
7726 @itemx -fdata-sections
7727 @opindex ffunction-sections
7728 @opindex fdata-sections
7729 Place each function or data item into its own section in the output
7730 file if the target supports arbitrary sections. The name of the
7731 function or the name of the data item determines the section's name
7734 Use these options on systems where the linker can perform optimizations
7735 to improve locality of reference in the instruction space. Most systems
7736 using the ELF object format and SPARC processors running Solaris 2 have
7737 linkers with such optimizations. AIX may have these optimizations in
7740 Only use these options when there are significant benefits from doing
7741 so. When you specify these options, the assembler and linker will
7742 create larger object and executable files and will also be slower.
7743 You will not be able to use @code{gprof} on all systems if you
7744 specify this option and you may have problems with debugging if
7745 you specify both this option and @option{-g}.
7747 @item -fbranch-target-load-optimize
7748 @opindex fbranch-target-load-optimize
7749 Perform branch target register load optimization before prologue / epilogue
7751 The use of target registers can typically be exposed only during reload,
7752 thus hoisting loads out of loops and doing inter-block scheduling needs
7753 a separate optimization pass.
7755 @item -fbranch-target-load-optimize2
7756 @opindex fbranch-target-load-optimize2
7757 Perform branch target register load optimization after prologue / epilogue
7760 @item -fbtr-bb-exclusive
7761 @opindex fbtr-bb-exclusive
7762 When performing branch target register load optimization, don't reuse
7763 branch target registers in within any basic block.
7765 @item -fstack-protector
7766 @opindex fstack-protector
7767 Emit extra code to check for buffer overflows, such as stack smashing
7768 attacks. This is done by adding a guard variable to functions with
7769 vulnerable objects. This includes functions that call alloca, and
7770 functions with buffers larger than 8 bytes. The guards are initialized
7771 when a function is entered and then checked when the function exits.
7772 If a guard check fails, an error message is printed and the program exits.
7774 @item -fstack-protector-all
7775 @opindex fstack-protector-all
7776 Like @option{-fstack-protector} except that all functions are protected.
7778 @item -fsection-anchors
7779 @opindex fsection-anchors
7780 Try to reduce the number of symbolic address calculations by using
7781 shared ``anchor'' symbols to address nearby objects. This transformation
7782 can help to reduce the number of GOT entries and GOT accesses on some
7785 For example, the implementation of the following function @code{foo}:
7789 int foo (void) @{ return a + b + c; @}
7792 would usually calculate the addresses of all three variables, but if you
7793 compile it with @option{-fsection-anchors}, it will access the variables
7794 from a common anchor point instead. The effect is similar to the
7795 following pseudocode (which isn't valid C):
7800 register int *xr = &x;
7801 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
7805 Not all targets support this option.
7807 @item --param @var{name}=@var{value}
7809 In some places, GCC uses various constants to control the amount of
7810 optimization that is done. For example, GCC will not inline functions
7811 that contain more that a certain number of instructions. You can
7812 control some of these constants on the command-line using the
7813 @option{--param} option.
7815 The names of specific parameters, and the meaning of the values, are
7816 tied to the internals of the compiler, and are subject to change
7817 without notice in future releases.
7819 In each case, the @var{value} is an integer. The allowable choices for
7820 @var{name} are given in the following table:
7823 @item struct-reorg-cold-struct-ratio
7824 The threshold ratio (as a percentage) between a structure frequency
7825 and the frequency of the hottest structure in the program. This parameter
7826 is used by struct-reorg optimization enabled by @option{-fipa-struct-reorg}.
7827 We say that if the ratio of a structure frequency, calculated by profiling,
7828 to the hottest structure frequency in the program is less than this
7829 parameter, then structure reorganization is not applied to this structure.
7832 @item predictable-branch-cost-outcome
7833 When branch is predicted to be taken with probability lower than this threshold
7834 (in percent), then it is considered well predictable. The default is 10.
7836 @item max-crossjump-edges
7837 The maximum number of incoming edges to consider for crossjumping.
7838 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
7839 the number of edges incoming to each block. Increasing values mean
7840 more aggressive optimization, making the compile time increase with
7841 probably small improvement in executable size.
7843 @item min-crossjump-insns
7844 The minimum number of instructions which must be matched at the end
7845 of two blocks before crossjumping will be performed on them. This
7846 value is ignored in the case where all instructions in the block being
7847 crossjumped from are matched. The default value is 5.
7849 @item max-grow-copy-bb-insns
7850 The maximum code size expansion factor when copying basic blocks
7851 instead of jumping. The expansion is relative to a jump instruction.
7852 The default value is 8.
7854 @item max-goto-duplication-insns
7855 The maximum number of instructions to duplicate to a block that jumps
7856 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
7857 passes, GCC factors computed gotos early in the compilation process,
7858 and unfactors them as late as possible. Only computed jumps at the
7859 end of a basic blocks with no more than max-goto-duplication-insns are
7860 unfactored. The default value is 8.
7862 @item max-delay-slot-insn-search
7863 The maximum number of instructions to consider when looking for an
7864 instruction to fill a delay slot. If more than this arbitrary number of
7865 instructions is searched, the time savings from filling the delay slot
7866 will be minimal so stop searching. Increasing values mean more
7867 aggressive optimization, making the compile time increase with probably
7868 small improvement in executable run time.
7870 @item max-delay-slot-live-search
7871 When trying to fill delay slots, the maximum number of instructions to
7872 consider when searching for a block with valid live register
7873 information. Increasing this arbitrarily chosen value means more
7874 aggressive optimization, increasing the compile time. This parameter
7875 should be removed when the delay slot code is rewritten to maintain the
7878 @item max-gcse-memory
7879 The approximate maximum amount of memory that will be allocated in
7880 order to perform the global common subexpression elimination
7881 optimization. If more memory than specified is required, the
7882 optimization will not be done.
7884 @item max-pending-list-length
7885 The maximum number of pending dependencies scheduling will allow
7886 before flushing the current state and starting over. Large functions
7887 with few branches or calls can create excessively large lists which
7888 needlessly consume memory and resources.
7890 @item max-inline-insns-single
7891 Several parameters control the tree inliner used in gcc.
7892 This number sets the maximum number of instructions (counted in GCC's
7893 internal representation) in a single function that the tree inliner
7894 will consider for inlining. This only affects functions declared
7895 inline and methods implemented in a class declaration (C++).
7896 The default value is 300.
7898 @item max-inline-insns-auto
7899 When you use @option{-finline-functions} (included in @option{-O3}),
7900 a lot of functions that would otherwise not be considered for inlining
7901 by the compiler will be investigated. To those functions, a different
7902 (more restrictive) limit compared to functions declared inline can
7904 The default value is 50.
7906 @item large-function-insns
7907 The limit specifying really large functions. For functions larger than this
7908 limit after inlining, inlining is constrained by
7909 @option{--param large-function-growth}. This parameter is useful primarily
7910 to avoid extreme compilation time caused by non-linear algorithms used by the
7912 The default value is 2700.
7914 @item large-function-growth
7915 Specifies maximal growth of large function caused by inlining in percents.
7916 The default value is 100 which limits large function growth to 2.0 times
7919 @item large-unit-insns
7920 The limit specifying large translation unit. Growth caused by inlining of
7921 units larger than this limit is limited by @option{--param inline-unit-growth}.
7922 For small units this might be too tight (consider unit consisting of function A
7923 that is inline and B that just calls A three time. If B is small relative to
7924 A, the growth of unit is 300\% and yet such inlining is very sane. For very
7925 large units consisting of small inlineable functions however the overall unit
7926 growth limit is needed to avoid exponential explosion of code size. Thus for
7927 smaller units, the size is increased to @option{--param large-unit-insns}
7928 before applying @option{--param inline-unit-growth}. The default is 10000
7930 @item inline-unit-growth
7931 Specifies maximal overall growth of the compilation unit caused by inlining.
7932 The default value is 30 which limits unit growth to 1.3 times the original
7935 @item ipcp-unit-growth
7936 Specifies maximal overall growth of the compilation unit caused by
7937 interprocedural constant propagation. The default value is 10 which limits
7938 unit growth to 1.1 times the original size.
7940 @item large-stack-frame
7941 The limit specifying large stack frames. While inlining the algorithm is trying
7942 to not grow past this limit too much. Default value is 256 bytes.
7944 @item large-stack-frame-growth
7945 Specifies maximal growth of large stack frames caused by inlining in percents.
7946 The default value is 1000 which limits large stack frame growth to 11 times
7949 @item max-inline-insns-recursive
7950 @itemx max-inline-insns-recursive-auto
7951 Specifies maximum number of instructions out-of-line copy of self recursive inline
7952 function can grow into by performing recursive inlining.
7954 For functions declared inline @option{--param max-inline-insns-recursive} is
7955 taken into account. For function not declared inline, recursive inlining
7956 happens only when @option{-finline-functions} (included in @option{-O3}) is
7957 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
7958 default value is 450.
7960 @item max-inline-recursive-depth
7961 @itemx max-inline-recursive-depth-auto
7962 Specifies maximum recursion depth used by the recursive inlining.
7964 For functions declared inline @option{--param max-inline-recursive-depth} is
7965 taken into account. For function not declared inline, recursive inlining
7966 happens only when @option{-finline-functions} (included in @option{-O3}) is
7967 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
7970 @item min-inline-recursive-probability
7971 Recursive inlining is profitable only for function having deep recursion
7972 in average and can hurt for function having little recursion depth by
7973 increasing the prologue size or complexity of function body to other
7976 When profile feedback is available (see @option{-fprofile-generate}) the actual
7977 recursion depth can be guessed from probability that function will recurse via
7978 given call expression. This parameter limits inlining only to call expression
7979 whose probability exceeds given threshold (in percents). The default value is
7982 @item early-inlining-insns
7983 Specify growth that early inliner can make. In effect it increases amount of
7984 inlining for code having large abstraction penalty. The default value is 8.
7986 @item max-early-inliner-iterations
7987 @itemx max-early-inliner-iterations
7988 Limit of iterations of early inliner. This basically bounds number of nested
7989 indirect calls early inliner can resolve. Deeper chains are still handled by
7992 @item min-vect-loop-bound
7993 The minimum number of iterations under which a loop will not get vectorized
7994 when @option{-ftree-vectorize} is used. The number of iterations after
7995 vectorization needs to be greater than the value specified by this option
7996 to allow vectorization. The default value is 0.
7998 @item max-unrolled-insns
7999 The maximum number of instructions that a loop should have if that loop
8000 is unrolled, and if the loop is unrolled, it determines how many times
8001 the loop code is unrolled.
8003 @item max-average-unrolled-insns
8004 The maximum number of instructions biased by probabilities of their execution
8005 that a loop should have if that loop is unrolled, and if the loop is unrolled,
8006 it determines how many times the loop code is unrolled.
8008 @item max-unroll-times
8009 The maximum number of unrollings of a single loop.
8011 @item max-peeled-insns
8012 The maximum number of instructions that a loop should have if that loop
8013 is peeled, and if the loop is peeled, it determines how many times
8014 the loop code is peeled.
8016 @item max-peel-times
8017 The maximum number of peelings of a single loop.
8019 @item max-completely-peeled-insns
8020 The maximum number of insns of a completely peeled loop.
8022 @item max-completely-peel-times
8023 The maximum number of iterations of a loop to be suitable for complete peeling.
8025 @item max-unswitch-insns
8026 The maximum number of insns of an unswitched loop.
8028 @item max-unswitch-level
8029 The maximum number of branches unswitched in a single loop.
8032 The minimum cost of an expensive expression in the loop invariant motion.
8034 @item iv-consider-all-candidates-bound
8035 Bound on number of candidates for induction variables below that
8036 all candidates are considered for each use in induction variable
8037 optimizations. Only the most relevant candidates are considered
8038 if there are more candidates, to avoid quadratic time complexity.
8040 @item iv-max-considered-uses
8041 The induction variable optimizations give up on loops that contain more
8042 induction variable uses.
8044 @item iv-always-prune-cand-set-bound
8045 If number of candidates in the set is smaller than this value,
8046 we always try to remove unnecessary ivs from the set during its
8047 optimization when a new iv is added to the set.
8049 @item scev-max-expr-size
8050 Bound on size of expressions used in the scalar evolutions analyzer.
8051 Large expressions slow the analyzer.
8053 @item omega-max-vars
8054 The maximum number of variables in an Omega constraint system.
8055 The default value is 128.
8057 @item omega-max-geqs
8058 The maximum number of inequalities in an Omega constraint system.
8059 The default value is 256.
8062 The maximum number of equalities in an Omega constraint system.
8063 The default value is 128.
8065 @item omega-max-wild-cards
8066 The maximum number of wildcard variables that the Omega solver will
8067 be able to insert. The default value is 18.
8069 @item omega-hash-table-size
8070 The size of the hash table in the Omega solver. The default value is
8073 @item omega-max-keys
8074 The maximal number of keys used by the Omega solver. The default
8077 @item omega-eliminate-redundant-constraints
8078 When set to 1, use expensive methods to eliminate all redundant
8079 constraints. The default value is 0.
8081 @item vect-max-version-for-alignment-checks
8082 The maximum number of runtime checks that can be performed when
8083 doing loop versioning for alignment in the vectorizer. See option
8084 ftree-vect-loop-version for more information.
8086 @item vect-max-version-for-alias-checks
8087 The maximum number of runtime checks that can be performed when
8088 doing loop versioning for alias in the vectorizer. See option
8089 ftree-vect-loop-version for more information.
8091 @item max-iterations-to-track
8093 The maximum number of iterations of a loop the brute force algorithm
8094 for analysis of # of iterations of the loop tries to evaluate.
8096 @item hot-bb-count-fraction
8097 Select fraction of the maximal count of repetitions of basic block in program
8098 given basic block needs to have to be considered hot.
8100 @item hot-bb-frequency-fraction
8101 Select fraction of the maximal frequency of executions of basic block in
8102 function given basic block needs to have to be considered hot
8104 @item max-predicted-iterations
8105 The maximum number of loop iterations we predict statically. This is useful
8106 in cases where function contain single loop with known bound and other loop
8107 with unknown. We predict the known number of iterations correctly, while
8108 the unknown number of iterations average to roughly 10. This means that the
8109 loop without bounds would appear artificially cold relative to the other one.
8111 @item align-threshold
8113 Select fraction of the maximal frequency of executions of basic block in
8114 function given basic block will get aligned.
8116 @item align-loop-iterations
8118 A loop expected to iterate at lest the selected number of iterations will get
8121 @item tracer-dynamic-coverage
8122 @itemx tracer-dynamic-coverage-feedback
8124 This value is used to limit superblock formation once the given percentage of
8125 executed instructions is covered. This limits unnecessary code size
8128 The @option{tracer-dynamic-coverage-feedback} is used only when profile
8129 feedback is available. The real profiles (as opposed to statically estimated
8130 ones) are much less balanced allowing the threshold to be larger value.
8132 @item tracer-max-code-growth
8133 Stop tail duplication once code growth has reached given percentage. This is
8134 rather hokey argument, as most of the duplicates will be eliminated later in
8135 cross jumping, so it may be set to much higher values than is the desired code
8138 @item tracer-min-branch-ratio
8140 Stop reverse growth when the reverse probability of best edge is less than this
8141 threshold (in percent).
8143 @item tracer-min-branch-ratio
8144 @itemx tracer-min-branch-ratio-feedback
8146 Stop forward growth if the best edge do have probability lower than this
8149 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
8150 compilation for profile feedback and one for compilation without. The value
8151 for compilation with profile feedback needs to be more conservative (higher) in
8152 order to make tracer effective.
8154 @item max-cse-path-length
8156 Maximum number of basic blocks on path that cse considers. The default is 10.
8159 The maximum instructions CSE process before flushing. The default is 1000.
8161 @item ggc-min-expand
8163 GCC uses a garbage collector to manage its own memory allocation. This
8164 parameter specifies the minimum percentage by which the garbage
8165 collector's heap should be allowed to expand between collections.
8166 Tuning this may improve compilation speed; it has no effect on code
8169 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
8170 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
8171 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
8172 GCC is not able to calculate RAM on a particular platform, the lower
8173 bound of 30% is used. Setting this parameter and
8174 @option{ggc-min-heapsize} to zero causes a full collection to occur at
8175 every opportunity. This is extremely slow, but can be useful for
8178 @item ggc-min-heapsize
8180 Minimum size of the garbage collector's heap before it begins bothering
8181 to collect garbage. The first collection occurs after the heap expands
8182 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
8183 tuning this may improve compilation speed, and has no effect on code
8186 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
8187 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
8188 with a lower bound of 4096 (four megabytes) and an upper bound of
8189 131072 (128 megabytes). If GCC is not able to calculate RAM on a
8190 particular platform, the lower bound is used. Setting this parameter
8191 very large effectively disables garbage collection. Setting this
8192 parameter and @option{ggc-min-expand} to zero causes a full collection
8193 to occur at every opportunity.
8195 @item max-reload-search-insns
8196 The maximum number of instruction reload should look backward for equivalent
8197 register. Increasing values mean more aggressive optimization, making the
8198 compile time increase with probably slightly better performance. The default
8201 @item max-cselib-memory-locations
8202 The maximum number of memory locations cselib should take into account.
8203 Increasing values mean more aggressive optimization, making the compile time
8204 increase with probably slightly better performance. The default value is 500.
8206 @item reorder-blocks-duplicate
8207 @itemx reorder-blocks-duplicate-feedback
8209 Used by basic block reordering pass to decide whether to use unconditional
8210 branch or duplicate the code on its destination. Code is duplicated when its
8211 estimated size is smaller than this value multiplied by the estimated size of
8212 unconditional jump in the hot spots of the program.
8214 The @option{reorder-block-duplicate-feedback} is used only when profile
8215 feedback is available and may be set to higher values than
8216 @option{reorder-block-duplicate} since information about the hot spots is more
8219 @item max-sched-ready-insns
8220 The maximum number of instructions ready to be issued the scheduler should
8221 consider at any given time during the first scheduling pass. Increasing
8222 values mean more thorough searches, making the compilation time increase
8223 with probably little benefit. The default value is 100.
8225 @item max-sched-region-blocks
8226 The maximum number of blocks in a region to be considered for
8227 interblock scheduling. The default value is 10.
8229 @item max-pipeline-region-blocks
8230 The maximum number of blocks in a region to be considered for
8231 pipelining in the selective scheduler. The default value is 15.
8233 @item max-sched-region-insns
8234 The maximum number of insns in a region to be considered for
8235 interblock scheduling. The default value is 100.
8237 @item max-pipeline-region-insns
8238 The maximum number of insns in a region to be considered for
8239 pipelining in the selective scheduler. The default value is 200.
8242 The minimum probability (in percents) of reaching a source block
8243 for interblock speculative scheduling. The default value is 40.
8245 @item max-sched-extend-regions-iters
8246 The maximum number of iterations through CFG to extend regions.
8247 0 - disable region extension,
8248 N - do at most N iterations.
8249 The default value is 0.
8251 @item max-sched-insn-conflict-delay
8252 The maximum conflict delay for an insn to be considered for speculative motion.
8253 The default value is 3.
8255 @item sched-spec-prob-cutoff
8256 The minimal probability of speculation success (in percents), so that
8257 speculative insn will be scheduled.
8258 The default value is 40.
8260 @item sched-mem-true-dep-cost
8261 Minimal distance (in CPU cycles) between store and load targeting same
8262 memory locations. The default value is 1.
8264 @item selsched-max-lookahead
8265 The maximum size of the lookahead window of selective scheduling. It is a
8266 depth of search for available instructions.
8267 The default value is 50.
8269 @item selsched-max-sched-times
8270 The maximum number of times that an instruction will be scheduled during
8271 selective scheduling. This is the limit on the number of iterations
8272 through which the instruction may be pipelined. The default value is 2.
8274 @item selsched-max-insns-to-rename
8275 The maximum number of best instructions in the ready list that are considered
8276 for renaming in the selective scheduler. The default value is 2.
8278 @item max-last-value-rtl
8279 The maximum size measured as number of RTLs that can be recorded in an expression
8280 in combiner for a pseudo register as last known value of that register. The default
8283 @item integer-share-limit
8284 Small integer constants can use a shared data structure, reducing the
8285 compiler's memory usage and increasing its speed. This sets the maximum
8286 value of a shared integer constant. The default value is 256.
8288 @item min-virtual-mappings
8289 Specifies the minimum number of virtual mappings in the incremental
8290 SSA updater that should be registered to trigger the virtual mappings
8291 heuristic defined by virtual-mappings-ratio. The default value is
8294 @item virtual-mappings-ratio
8295 If the number of virtual mappings is virtual-mappings-ratio bigger
8296 than the number of virtual symbols to be updated, then the incremental
8297 SSA updater switches to a full update for those symbols. The default
8300 @item ssp-buffer-size
8301 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
8302 protection when @option{-fstack-protection} is used.
8304 @item max-jump-thread-duplication-stmts
8305 Maximum number of statements allowed in a block that needs to be
8306 duplicated when threading jumps.
8308 @item max-fields-for-field-sensitive
8309 Maximum number of fields in a structure we will treat in
8310 a field sensitive manner during pointer analysis. The default is zero
8311 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
8313 @item prefetch-latency
8314 Estimate on average number of instructions that are executed before
8315 prefetch finishes. The distance we prefetch ahead is proportional
8316 to this constant. Increasing this number may also lead to less
8317 streams being prefetched (see @option{simultaneous-prefetches}).
8319 @item simultaneous-prefetches
8320 Maximum number of prefetches that can run at the same time.
8322 @item l1-cache-line-size
8323 The size of cache line in L1 cache, in bytes.
8326 The size of L1 cache, in kilobytes.
8329 The size of L2 cache, in kilobytes.
8331 @item min-insn-to-prefetch-ratio
8332 The minimum ratio between the number of instructions and the
8333 number of prefetches to enable prefetching in a loop with an
8336 @item prefetch-min-insn-to-mem-ratio
8337 The minimum ratio between the number of instructions and the
8338 number of memory references to enable prefetching in a loop.
8340 @item use-canonical-types
8341 Whether the compiler should use the ``canonical'' type system. By
8342 default, this should always be 1, which uses a more efficient internal
8343 mechanism for comparing types in C++ and Objective-C++. However, if
8344 bugs in the canonical type system are causing compilation failures,
8345 set this value to 0 to disable canonical types.
8347 @item switch-conversion-max-branch-ratio
8348 Switch initialization conversion will refuse to create arrays that are
8349 bigger than @option{switch-conversion-max-branch-ratio} times the number of
8350 branches in the switch.
8352 @item max-partial-antic-length
8353 Maximum length of the partial antic set computed during the tree
8354 partial redundancy elimination optimization (@option{-ftree-pre}) when
8355 optimizing at @option{-O3} and above. For some sorts of source code
8356 the enhanced partial redundancy elimination optimization can run away,
8357 consuming all of the memory available on the host machine. This
8358 parameter sets a limit on the length of the sets that are computed,
8359 which prevents the runaway behavior. Setting a value of 0 for
8360 this parameter will allow an unlimited set length.
8362 @item sccvn-max-scc-size
8363 Maximum size of a strongly connected component (SCC) during SCCVN
8364 processing. If this limit is hit, SCCVN processing for the whole
8365 function will not be done and optimizations depending on it will
8366 be disabled. The default maximum SCC size is 10000.
8368 @item ira-max-loops-num
8369 IRA uses a regional register allocation by default. If a function
8370 contains loops more than number given by the parameter, only at most
8371 given number of the most frequently executed loops will form regions
8372 for the regional register allocation. The default value of the
8375 @item ira-max-conflict-table-size
8376 Although IRA uses a sophisticated algorithm of compression conflict
8377 table, the table can be still big for huge functions. If the conflict
8378 table for a function could be more than size in MB given by the
8379 parameter, the conflict table is not built and faster, simpler, and
8380 lower quality register allocation algorithm will be used. The
8381 algorithm do not use pseudo-register conflicts. The default value of
8382 the parameter is 2000.
8384 @item loop-invariant-max-bbs-in-loop
8385 Loop invariant motion can be very expensive, both in compile time and
8386 in amount of needed compile time memory, with very large loops. Loops
8387 with more basic blocks than this parameter won't have loop invariant
8388 motion optimization performed on them. The default value of the
8389 parameter is 1000 for -O1 and 10000 for -O2 and above.
8391 @item min-nondebug-insn-uid
8392 Use uids starting at this parameter for nondebug insns. The range below
8393 the parameter is reserved exclusively for debug insns created by
8394 @option{-fvar-tracking-assignments}, but debug insns may get
8395 (non-overlapping) uids above it if the reserved range is exhausted.
8397 @item ipa-sra-ptr-growth-factor
8398 IPA-SRA will replace a pointer to an aggregate with one or more new
8399 parameters only when their cumulative size is less or equal to
8400 @option{ipa-sra-ptr-growth-factor} times the size of the original
8406 @node Preprocessor Options
8407 @section Options Controlling the Preprocessor
8408 @cindex preprocessor options
8409 @cindex options, preprocessor
8411 These options control the C preprocessor, which is run on each C source
8412 file before actual compilation.
8414 If you use the @option{-E} option, nothing is done except preprocessing.
8415 Some of these options make sense only together with @option{-E} because
8416 they cause the preprocessor output to be unsuitable for actual
8420 @item -Wp,@var{option}
8422 You can use @option{-Wp,@var{option}} to bypass the compiler driver
8423 and pass @var{option} directly through to the preprocessor. If
8424 @var{option} contains commas, it is split into multiple options at the
8425 commas. However, many options are modified, translated or interpreted
8426 by the compiler driver before being passed to the preprocessor, and
8427 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
8428 interface is undocumented and subject to change, so whenever possible
8429 you should avoid using @option{-Wp} and let the driver handle the
8432 @item -Xpreprocessor @var{option}
8433 @opindex Xpreprocessor
8434 Pass @var{option} as an option to the preprocessor. You can use this to
8435 supply system-specific preprocessor options which GCC does not know how to
8438 If you want to pass an option that takes an argument, you must use
8439 @option{-Xpreprocessor} twice, once for the option and once for the argument.
8442 @include cppopts.texi
8444 @node Assembler Options
8445 @section Passing Options to the Assembler
8447 @c prevent bad page break with this line
8448 You can pass options to the assembler.
8451 @item -Wa,@var{option}
8453 Pass @var{option} as an option to the assembler. If @var{option}
8454 contains commas, it is split into multiple options at the commas.
8456 @item -Xassembler @var{option}
8458 Pass @var{option} as an option to the assembler. You can use this to
8459 supply system-specific assembler options which GCC does not know how to
8462 If you want to pass an option that takes an argument, you must use
8463 @option{-Xassembler} twice, once for the option and once for the argument.
8468 @section Options for Linking
8469 @cindex link options
8470 @cindex options, linking
8472 These options come into play when the compiler links object files into
8473 an executable output file. They are meaningless if the compiler is
8474 not doing a link step.
8478 @item @var{object-file-name}
8479 A file name that does not end in a special recognized suffix is
8480 considered to name an object file or library. (Object files are
8481 distinguished from libraries by the linker according to the file
8482 contents.) If linking is done, these object files are used as input
8491 If any of these options is used, then the linker is not run, and
8492 object file names should not be used as arguments. @xref{Overall
8496 @item -l@var{library}
8497 @itemx -l @var{library}
8499 Search the library named @var{library} when linking. (The second
8500 alternative with the library as a separate argument is only for
8501 POSIX compliance and is not recommended.)
8503 It makes a difference where in the command you write this option; the
8504 linker searches and processes libraries and object files in the order they
8505 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
8506 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
8507 to functions in @samp{z}, those functions may not be loaded.
8509 The linker searches a standard list of directories for the library,
8510 which is actually a file named @file{lib@var{library}.a}. The linker
8511 then uses this file as if it had been specified precisely by name.
8513 The directories searched include several standard system directories
8514 plus any that you specify with @option{-L}.
8516 Normally the files found this way are library files---archive files
8517 whose members are object files. The linker handles an archive file by
8518 scanning through it for members which define symbols that have so far
8519 been referenced but not defined. But if the file that is found is an
8520 ordinary object file, it is linked in the usual fashion. The only
8521 difference between using an @option{-l} option and specifying a file name
8522 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
8523 and searches several directories.
8527 You need this special case of the @option{-l} option in order to
8528 link an Objective-C or Objective-C++ program.
8531 @opindex nostartfiles
8532 Do not use the standard system startup files when linking.
8533 The standard system libraries are used normally, unless @option{-nostdlib}
8534 or @option{-nodefaultlibs} is used.
8536 @item -nodefaultlibs
8537 @opindex nodefaultlibs
8538 Do not use the standard system libraries when linking.
8539 Only the libraries you specify will be passed to the linker, options
8540 specifying linkage of the system libraries, such as @code{-static-libgcc}
8541 or @code{-shared-libgcc}, will be ignored.
8542 The standard startup files are used normally, unless @option{-nostartfiles}
8543 is used. The compiler may generate calls to @code{memcmp},
8544 @code{memset}, @code{memcpy} and @code{memmove}.
8545 These entries are usually resolved by entries in
8546 libc. These entry points should be supplied through some other
8547 mechanism when this option is specified.
8551 Do not use the standard system startup files or libraries when linking.
8552 No startup files and only the libraries you specify will be passed to
8553 the linker, options specifying linkage of the system libraries, such as
8554 @code{-static-libgcc} or @code{-shared-libgcc}, will be ignored.
8555 The compiler may generate calls to @code{memcmp}, @code{memset},
8556 @code{memcpy} and @code{memmove}.
8557 These entries are usually resolved by entries in
8558 libc. These entry points should be supplied through some other
8559 mechanism when this option is specified.
8561 @cindex @option{-lgcc}, use with @option{-nostdlib}
8562 @cindex @option{-nostdlib} and unresolved references
8563 @cindex unresolved references and @option{-nostdlib}
8564 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
8565 @cindex @option{-nodefaultlibs} and unresolved references
8566 @cindex unresolved references and @option{-nodefaultlibs}
8567 One of the standard libraries bypassed by @option{-nostdlib} and
8568 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
8569 that GCC uses to overcome shortcomings of particular machines, or special
8570 needs for some languages.
8571 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
8572 Collection (GCC) Internals},
8573 for more discussion of @file{libgcc.a}.)
8574 In most cases, you need @file{libgcc.a} even when you want to avoid
8575 other standard libraries. In other words, when you specify @option{-nostdlib}
8576 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
8577 This ensures that you have no unresolved references to internal GCC
8578 library subroutines. (For example, @samp{__main}, used to ensure C++
8579 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
8580 GNU Compiler Collection (GCC) Internals}.)
8584 Produce a position independent executable on targets which support it.
8585 For predictable results, you must also specify the same set of options
8586 that were used to generate code (@option{-fpie}, @option{-fPIE},
8587 or model suboptions) when you specify this option.
8591 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
8592 that support it. This instructs the linker to add all symbols, not
8593 only used ones, to the dynamic symbol table. This option is needed
8594 for some uses of @code{dlopen} or to allow obtaining backtraces
8595 from within a program.
8599 Remove all symbol table and relocation information from the executable.
8603 On systems that support dynamic linking, this prevents linking with the shared
8604 libraries. On other systems, this option has no effect.
8608 Produce a shared object which can then be linked with other objects to
8609 form an executable. Not all systems support this option. For predictable
8610 results, you must also specify the same set of options that were used to
8611 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
8612 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
8613 needs to build supplementary stub code for constructors to work. On
8614 multi-libbed systems, @samp{gcc -shared} must select the correct support
8615 libraries to link against. Failing to supply the correct flags may lead
8616 to subtle defects. Supplying them in cases where they are not necessary
8619 @item -shared-libgcc
8620 @itemx -static-libgcc
8621 @opindex shared-libgcc
8622 @opindex static-libgcc
8623 On systems that provide @file{libgcc} as a shared library, these options
8624 force the use of either the shared or static version respectively.
8625 If no shared version of @file{libgcc} was built when the compiler was
8626 configured, these options have no effect.
8628 There are several situations in which an application should use the
8629 shared @file{libgcc} instead of the static version. The most common
8630 of these is when the application wishes to throw and catch exceptions
8631 across different shared libraries. In that case, each of the libraries
8632 as well as the application itself should use the shared @file{libgcc}.
8634 Therefore, the G++ and GCJ drivers automatically add
8635 @option{-shared-libgcc} whenever you build a shared library or a main
8636 executable, because C++ and Java programs typically use exceptions, so
8637 this is the right thing to do.
8639 If, instead, you use the GCC driver to create shared libraries, you may
8640 find that they will not always be linked with the shared @file{libgcc}.
8641 If GCC finds, at its configuration time, that you have a non-GNU linker
8642 or a GNU linker that does not support option @option{--eh-frame-hdr},
8643 it will link the shared version of @file{libgcc} into shared libraries
8644 by default. Otherwise, it will take advantage of the linker and optimize
8645 away the linking with the shared version of @file{libgcc}, linking with
8646 the static version of libgcc by default. This allows exceptions to
8647 propagate through such shared libraries, without incurring relocation
8648 costs at library load time.
8650 However, if a library or main executable is supposed to throw or catch
8651 exceptions, you must link it using the G++ or GCJ driver, as appropriate
8652 for the languages used in the program, or using the option
8653 @option{-shared-libgcc}, such that it is linked with the shared
8656 @item -static-libstdc++
8657 When the @command{g++} program is used to link a C++ program, it will
8658 normally automatically link against @option{libstdc++}. If
8659 @file{libstdc++} is available as a shared library, and the
8660 @option{-static} option is not used, then this will link against the
8661 shared version of @file{libstdc++}. That is normally fine. However, it
8662 is sometimes useful to freeze the version of @file{libstdc++} used by
8663 the program without going all the way to a fully static link. The
8664 @option{-static-libstdc++} option directs the @command{g++} driver to
8665 link @file{libstdc++} statically, without necessarily linking other
8666 libraries statically.
8670 Bind references to global symbols when building a shared object. Warn
8671 about any unresolved references (unless overridden by the link editor
8672 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
8675 @item -T @var{script}
8677 @cindex linker script
8678 Use @var{script} as the linker script. This option is supported by most
8679 systems using the GNU linker. On some targets, such as bare-board
8680 targets without an operating system, the @option{-T} option may be required
8681 when linking to avoid references to undefined symbols.
8683 @item -Xlinker @var{option}
8685 Pass @var{option} as an option to the linker. You can use this to
8686 supply system-specific linker options which GCC does not know how to
8689 If you want to pass an option that takes a separate argument, you must use
8690 @option{-Xlinker} twice, once for the option and once for the argument.
8691 For example, to pass @option{-assert definitions}, you must write
8692 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
8693 @option{-Xlinker "-assert definitions"}, because this passes the entire
8694 string as a single argument, which is not what the linker expects.
8696 When using the GNU linker, it is usually more convenient to pass
8697 arguments to linker options using the @option{@var{option}=@var{value}}
8698 syntax than as separate arguments. For example, you can specify
8699 @samp{-Xlinker -Map=output.map} rather than
8700 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
8701 this syntax for command-line options.
8703 @item -Wl,@var{option}
8705 Pass @var{option} as an option to the linker. If @var{option} contains
8706 commas, it is split into multiple options at the commas. You can use this
8707 syntax to pass an argument to the option.
8708 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
8709 linker. When using the GNU linker, you can also get the same effect with
8710 @samp{-Wl,-Map=output.map}.
8712 @item -u @var{symbol}
8714 Pretend the symbol @var{symbol} is undefined, to force linking of
8715 library modules to define it. You can use @option{-u} multiple times with
8716 different symbols to force loading of additional library modules.
8719 @node Directory Options
8720 @section Options for Directory Search
8721 @cindex directory options
8722 @cindex options, directory search
8725 These options specify directories to search for header files, for
8726 libraries and for parts of the compiler:
8731 Add the directory @var{dir} to the head of the list of directories to be
8732 searched for header files. This can be used to override a system header
8733 file, substituting your own version, since these directories are
8734 searched before the system header file directories. However, you should
8735 not use this option to add directories that contain vendor-supplied
8736 system header files (use @option{-isystem} for that). If you use more than
8737 one @option{-I} option, the directories are scanned in left-to-right
8738 order; the standard system directories come after.
8740 If a standard system include directory, or a directory specified with
8741 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
8742 option will be ignored. The directory will still be searched but as a
8743 system directory at its normal position in the system include chain.
8744 This is to ensure that GCC's procedure to fix buggy system headers and
8745 the ordering for the include_next directive are not inadvertently changed.
8746 If you really need to change the search order for system directories,
8747 use the @option{-nostdinc} and/or @option{-isystem} options.
8749 @item -iquote@var{dir}
8751 Add the directory @var{dir} to the head of the list of directories to
8752 be searched for header files only for the case of @samp{#include
8753 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
8754 otherwise just like @option{-I}.
8758 Add directory @var{dir} to the list of directories to be searched
8761 @item -B@var{prefix}
8763 This option specifies where to find the executables, libraries,
8764 include files, and data files of the compiler itself.
8766 The compiler driver program runs one or more of the subprograms
8767 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
8768 @var{prefix} as a prefix for each program it tries to run, both with and
8769 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
8771 For each subprogram to be run, the compiler driver first tries the
8772 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
8773 was not specified, the driver tries two standard prefixes, which are
8774 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
8775 those results in a file name that is found, the unmodified program
8776 name is searched for using the directories specified in your
8777 @env{PATH} environment variable.
8779 The compiler will check to see if the path provided by the @option{-B}
8780 refers to a directory, and if necessary it will add a directory
8781 separator character at the end of the path.
8783 @option{-B} prefixes that effectively specify directory names also apply
8784 to libraries in the linker, because the compiler translates these
8785 options into @option{-L} options for the linker. They also apply to
8786 includes files in the preprocessor, because the compiler translates these
8787 options into @option{-isystem} options for the preprocessor. In this case,
8788 the compiler appends @samp{include} to the prefix.
8790 The run-time support file @file{libgcc.a} can also be searched for using
8791 the @option{-B} prefix, if needed. If it is not found there, the two
8792 standard prefixes above are tried, and that is all. The file is left
8793 out of the link if it is not found by those means.
8795 Another way to specify a prefix much like the @option{-B} prefix is to use
8796 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
8799 As a special kludge, if the path provided by @option{-B} is
8800 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
8801 9, then it will be replaced by @file{[dir/]include}. This is to help
8802 with boot-strapping the compiler.
8804 @item -specs=@var{file}
8806 Process @var{file} after the compiler reads in the standard @file{specs}
8807 file, in order to override the defaults that the @file{gcc} driver
8808 program uses when determining what switches to pass to @file{cc1},
8809 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
8810 @option{-specs=@var{file}} can be specified on the command line, and they
8811 are processed in order, from left to right.
8813 @item --sysroot=@var{dir}
8815 Use @var{dir} as the logical root directory for headers and libraries.
8816 For example, if the compiler would normally search for headers in
8817 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
8818 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
8820 If you use both this option and the @option{-isysroot} option, then
8821 the @option{--sysroot} option will apply to libraries, but the
8822 @option{-isysroot} option will apply to header files.
8824 The GNU linker (beginning with version 2.16) has the necessary support
8825 for this option. If your linker does not support this option, the
8826 header file aspect of @option{--sysroot} will still work, but the
8827 library aspect will not.
8831 This option has been deprecated. Please use @option{-iquote} instead for
8832 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
8833 Any directories you specify with @option{-I} options before the @option{-I-}
8834 option are searched only for the case of @samp{#include "@var{file}"};
8835 they are not searched for @samp{#include <@var{file}>}.
8837 If additional directories are specified with @option{-I} options after
8838 the @option{-I-}, these directories are searched for all @samp{#include}
8839 directives. (Ordinarily @emph{all} @option{-I} directories are used
8842 In addition, the @option{-I-} option inhibits the use of the current
8843 directory (where the current input file came from) as the first search
8844 directory for @samp{#include "@var{file}"}. There is no way to
8845 override this effect of @option{-I-}. With @option{-I.} you can specify
8846 searching the directory which was current when the compiler was
8847 invoked. That is not exactly the same as what the preprocessor does
8848 by default, but it is often satisfactory.
8850 @option{-I-} does not inhibit the use of the standard system directories
8851 for header files. Thus, @option{-I-} and @option{-nostdinc} are
8858 @section Specifying subprocesses and the switches to pass to them
8861 @command{gcc} is a driver program. It performs its job by invoking a
8862 sequence of other programs to do the work of compiling, assembling and
8863 linking. GCC interprets its command-line parameters and uses these to
8864 deduce which programs it should invoke, and which command-line options
8865 it ought to place on their command lines. This behavior is controlled
8866 by @dfn{spec strings}. In most cases there is one spec string for each
8867 program that GCC can invoke, but a few programs have multiple spec
8868 strings to control their behavior. The spec strings built into GCC can
8869 be overridden by using the @option{-specs=} command-line switch to specify
8872 @dfn{Spec files} are plaintext files that are used to construct spec
8873 strings. They consist of a sequence of directives separated by blank
8874 lines. The type of directive is determined by the first non-whitespace
8875 character on the line and it can be one of the following:
8878 @item %@var{command}
8879 Issues a @var{command} to the spec file processor. The commands that can
8883 @item %include <@var{file}>
8885 Search for @var{file} and insert its text at the current point in the
8888 @item %include_noerr <@var{file}>
8889 @cindex %include_noerr
8890 Just like @samp{%include}, but do not generate an error message if the include
8891 file cannot be found.
8893 @item %rename @var{old_name} @var{new_name}
8895 Rename the spec string @var{old_name} to @var{new_name}.
8899 @item *[@var{spec_name}]:
8900 This tells the compiler to create, override or delete the named spec
8901 string. All lines after this directive up to the next directive or
8902 blank line are considered to be the text for the spec string. If this
8903 results in an empty string then the spec will be deleted. (Or, if the
8904 spec did not exist, then nothing will happened.) Otherwise, if the spec
8905 does not currently exist a new spec will be created. If the spec does
8906 exist then its contents will be overridden by the text of this
8907 directive, unless the first character of that text is the @samp{+}
8908 character, in which case the text will be appended to the spec.
8910 @item [@var{suffix}]:
8911 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
8912 and up to the next directive or blank line are considered to make up the
8913 spec string for the indicated suffix. When the compiler encounters an
8914 input file with the named suffix, it will processes the spec string in
8915 order to work out how to compile that file. For example:
8922 This says that any input file whose name ends in @samp{.ZZ} should be
8923 passed to the program @samp{z-compile}, which should be invoked with the
8924 command-line switch @option{-input} and with the result of performing the
8925 @samp{%i} substitution. (See below.)
8927 As an alternative to providing a spec string, the text that follows a
8928 suffix directive can be one of the following:
8931 @item @@@var{language}
8932 This says that the suffix is an alias for a known @var{language}. This is
8933 similar to using the @option{-x} command-line switch to GCC to specify a
8934 language explicitly. For example:
8941 Says that .ZZ files are, in fact, C++ source files.
8944 This causes an error messages saying:
8947 @var{name} compiler not installed on this system.
8951 GCC already has an extensive list of suffixes built into it.
8952 This directive will add an entry to the end of the list of suffixes, but
8953 since the list is searched from the end backwards, it is effectively
8954 possible to override earlier entries using this technique.
8958 GCC has the following spec strings built into it. Spec files can
8959 override these strings or create their own. Note that individual
8960 targets can also add their own spec strings to this list.
8963 asm Options to pass to the assembler
8964 asm_final Options to pass to the assembler post-processor
8965 cpp Options to pass to the C preprocessor
8966 cc1 Options to pass to the C compiler
8967 cc1plus Options to pass to the C++ compiler
8968 endfile Object files to include at the end of the link
8969 link Options to pass to the linker
8970 lib Libraries to include on the command line to the linker
8971 libgcc Decides which GCC support library to pass to the linker
8972 linker Sets the name of the linker
8973 predefines Defines to be passed to the C preprocessor
8974 signed_char Defines to pass to CPP to say whether @code{char} is signed
8976 startfile Object files to include at the start of the link
8979 Here is a small example of a spec file:
8985 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
8988 This example renames the spec called @samp{lib} to @samp{old_lib} and
8989 then overrides the previous definition of @samp{lib} with a new one.
8990 The new definition adds in some extra command-line options before
8991 including the text of the old definition.
8993 @dfn{Spec strings} are a list of command-line options to be passed to their
8994 corresponding program. In addition, the spec strings can contain
8995 @samp{%}-prefixed sequences to substitute variable text or to
8996 conditionally insert text into the command line. Using these constructs
8997 it is possible to generate quite complex command lines.
8999 Here is a table of all defined @samp{%}-sequences for spec
9000 strings. Note that spaces are not generated automatically around the
9001 results of expanding these sequences. Therefore you can concatenate them
9002 together or combine them with constant text in a single argument.
9006 Substitute one @samp{%} into the program name or argument.
9009 Substitute the name of the input file being processed.
9012 Substitute the basename of the input file being processed.
9013 This is the substring up to (and not including) the last period
9014 and not including the directory.
9017 This is the same as @samp{%b}, but include the file suffix (text after
9021 Marks the argument containing or following the @samp{%d} as a
9022 temporary file name, so that that file will be deleted if GCC exits
9023 successfully. Unlike @samp{%g}, this contributes no text to the
9026 @item %g@var{suffix}
9027 Substitute a file name that has suffix @var{suffix} and is chosen
9028 once per compilation, and mark the argument in the same way as
9029 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
9030 name is now chosen in a way that is hard to predict even when previously
9031 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
9032 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
9033 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
9034 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
9035 was simply substituted with a file name chosen once per compilation,
9036 without regard to any appended suffix (which was therefore treated
9037 just like ordinary text), making such attacks more likely to succeed.
9039 @item %u@var{suffix}
9040 Like @samp{%g}, but generates a new temporary file name even if
9041 @samp{%u@var{suffix}} was already seen.
9043 @item %U@var{suffix}
9044 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
9045 new one if there is no such last file name. In the absence of any
9046 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
9047 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
9048 would involve the generation of two distinct file names, one
9049 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
9050 simply substituted with a file name chosen for the previous @samp{%u},
9051 without regard to any appended suffix.
9053 @item %j@var{suffix}
9054 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
9055 writable, and if save-temps is off; otherwise, substitute the name
9056 of a temporary file, just like @samp{%u}. This temporary file is not
9057 meant for communication between processes, but rather as a junk
9060 @item %|@var{suffix}
9061 @itemx %m@var{suffix}
9062 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
9063 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
9064 all. These are the two most common ways to instruct a program that it
9065 should read from standard input or write to standard output. If you
9066 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
9067 construct: see for example @file{f/lang-specs.h}.
9069 @item %.@var{SUFFIX}
9070 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
9071 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
9072 terminated by the next space or %.
9075 Marks the argument containing or following the @samp{%w} as the
9076 designated output file of this compilation. This puts the argument
9077 into the sequence of arguments that @samp{%o} will substitute later.
9080 Substitutes the names of all the output files, with spaces
9081 automatically placed around them. You should write spaces
9082 around the @samp{%o} as well or the results are undefined.
9083 @samp{%o} is for use in the specs for running the linker.
9084 Input files whose names have no recognized suffix are not compiled
9085 at all, but they are included among the output files, so they will
9089 Substitutes the suffix for object files. Note that this is
9090 handled specially when it immediately follows @samp{%g, %u, or %U},
9091 because of the need for those to form complete file names. The
9092 handling is such that @samp{%O} is treated exactly as if it had already
9093 been substituted, except that @samp{%g, %u, and %U} do not currently
9094 support additional @var{suffix} characters following @samp{%O} as they would
9095 following, for example, @samp{.o}.
9098 Substitutes the standard macro predefinitions for the
9099 current target machine. Use this when running @code{cpp}.
9102 Like @samp{%p}, but puts @samp{__} before and after the name of each
9103 predefined macro, except for macros that start with @samp{__} or with
9104 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
9108 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
9109 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
9110 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
9111 and @option{-imultilib} as necessary.
9114 Current argument is the name of a library or startup file of some sort.
9115 Search for that file in a standard list of directories and substitute
9116 the full name found. The current working directory is included in the
9117 list of directories scanned.
9120 Current argument is the name of a linker script. Search for that file
9121 in the current list of directories to scan for libraries. If the file
9122 is located insert a @option{--script} option into the command line
9123 followed by the full path name found. If the file is not found then
9124 generate an error message. Note: the current working directory is not
9128 Print @var{str} as an error message. @var{str} is terminated by a newline.
9129 Use this when inconsistent options are detected.
9132 Substitute the contents of spec string @var{name} at this point.
9135 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
9137 @item %x@{@var{option}@}
9138 Accumulate an option for @samp{%X}.
9141 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
9145 Output the accumulated assembler options specified by @option{-Wa}.
9148 Output the accumulated preprocessor options specified by @option{-Wp}.
9151 Process the @code{asm} spec. This is used to compute the
9152 switches to be passed to the assembler.
9155 Process the @code{asm_final} spec. This is a spec string for
9156 passing switches to an assembler post-processor, if such a program is
9160 Process the @code{link} spec. This is the spec for computing the
9161 command line passed to the linker. Typically it will make use of the
9162 @samp{%L %G %S %D and %E} sequences.
9165 Dump out a @option{-L} option for each directory that GCC believes might
9166 contain startup files. If the target supports multilibs then the
9167 current multilib directory will be prepended to each of these paths.
9170 Process the @code{lib} spec. This is a spec string for deciding which
9171 libraries should be included on the command line to the linker.
9174 Process the @code{libgcc} spec. This is a spec string for deciding
9175 which GCC support library should be included on the command line to the linker.
9178 Process the @code{startfile} spec. This is a spec for deciding which
9179 object files should be the first ones passed to the linker. Typically
9180 this might be a file named @file{crt0.o}.
9183 Process the @code{endfile} spec. This is a spec string that specifies
9184 the last object files that will be passed to the linker.
9187 Process the @code{cpp} spec. This is used to construct the arguments
9188 to be passed to the C preprocessor.
9191 Process the @code{cc1} spec. This is used to construct the options to be
9192 passed to the actual C compiler (@samp{cc1}).
9195 Process the @code{cc1plus} spec. This is used to construct the options to be
9196 passed to the actual C++ compiler (@samp{cc1plus}).
9199 Substitute the variable part of a matched option. See below.
9200 Note that each comma in the substituted string is replaced by
9204 Remove all occurrences of @code{-S} from the command line. Note---this
9205 command is position dependent. @samp{%} commands in the spec string
9206 before this one will see @code{-S}, @samp{%} commands in the spec string
9207 after this one will not.
9209 @item %:@var{function}(@var{args})
9210 Call the named function @var{function}, passing it @var{args}.
9211 @var{args} is first processed as a nested spec string, then split
9212 into an argument vector in the usual fashion. The function returns
9213 a string which is processed as if it had appeared literally as part
9214 of the current spec.
9216 The following built-in spec functions are provided:
9220 The @code{getenv} spec function takes two arguments: an environment
9221 variable name and a string. If the environment variable is not
9222 defined, a fatal error is issued. Otherwise, the return value is the
9223 value of the environment variable concatenated with the string. For
9224 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
9227 %:getenv(TOPDIR /include)
9230 expands to @file{/path/to/top/include}.
9232 @item @code{if-exists}
9233 The @code{if-exists} spec function takes one argument, an absolute
9234 pathname to a file. If the file exists, @code{if-exists} returns the
9235 pathname. Here is a small example of its usage:
9239 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
9242 @item @code{if-exists-else}
9243 The @code{if-exists-else} spec function is similar to the @code{if-exists}
9244 spec function, except that it takes two arguments. The first argument is
9245 an absolute pathname to a file. If the file exists, @code{if-exists-else}
9246 returns the pathname. If it does not exist, it returns the second argument.
9247 This way, @code{if-exists-else} can be used to select one file or another,
9248 based on the existence of the first. Here is a small example of its usage:
9252 crt0%O%s %:if-exists(crti%O%s) \
9253 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
9256 @item @code{replace-outfile}
9257 The @code{replace-outfile} spec function takes two arguments. It looks for the
9258 first argument in the outfiles array and replaces it with the second argument. Here
9259 is a small example of its usage:
9262 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
9265 @item @code{print-asm-header}
9266 The @code{print-asm-header} function takes no arguments and simply
9267 prints a banner like:
9273 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
9276 It is used to separate compiler options from assembler options
9277 in the @option{--target-help} output.
9281 Substitutes the @code{-S} switch, if that switch was given to GCC@.
9282 If that switch was not specified, this substitutes nothing. Note that
9283 the leading dash is omitted when specifying this option, and it is
9284 automatically inserted if the substitution is performed. Thus the spec
9285 string @samp{%@{foo@}} would match the command-line option @option{-foo}
9286 and would output the command line option @option{-foo}.
9288 @item %W@{@code{S}@}
9289 Like %@{@code{S}@} but mark last argument supplied within as a file to be
9292 @item %@{@code{S}*@}
9293 Substitutes all the switches specified to GCC whose names start
9294 with @code{-S}, but which also take an argument. This is used for
9295 switches like @option{-o}, @option{-D}, @option{-I}, etc.
9296 GCC considers @option{-o foo} as being
9297 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
9298 text, including the space. Thus two arguments would be generated.
9300 @item %@{@code{S}*&@code{T}*@}
9301 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
9302 (the order of @code{S} and @code{T} in the spec is not significant).
9303 There can be any number of ampersand-separated variables; for each the
9304 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
9306 @item %@{@code{S}:@code{X}@}
9307 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
9309 @item %@{!@code{S}:@code{X}@}
9310 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
9312 @item %@{@code{S}*:@code{X}@}
9313 Substitutes @code{X} if one or more switches whose names start with
9314 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
9315 once, no matter how many such switches appeared. However, if @code{%*}
9316 appears somewhere in @code{X}, then @code{X} will be substituted once
9317 for each matching switch, with the @code{%*} replaced by the part of
9318 that switch that matched the @code{*}.
9320 @item %@{.@code{S}:@code{X}@}
9321 Substitutes @code{X}, if processing a file with suffix @code{S}.
9323 @item %@{!.@code{S}:@code{X}@}
9324 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
9326 @item %@{,@code{S}:@code{X}@}
9327 Substitutes @code{X}, if processing a file for language @code{S}.
9329 @item %@{!,@code{S}:@code{X}@}
9330 Substitutes @code{X}, if not processing a file for language @code{S}.
9332 @item %@{@code{S}|@code{P}:@code{X}@}
9333 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
9334 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
9335 @code{*} sequences as well, although they have a stronger binding than
9336 the @samp{|}. If @code{%*} appears in @code{X}, all of the
9337 alternatives must be starred, and only the first matching alternative
9340 For example, a spec string like this:
9343 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
9346 will output the following command-line options from the following input
9347 command-line options:
9352 -d fred.c -foo -baz -boggle
9353 -d jim.d -bar -baz -boggle
9356 @item %@{S:X; T:Y; :D@}
9358 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
9359 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
9360 be as many clauses as you need. This may be combined with @code{.},
9361 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
9366 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
9367 construct may contain other nested @samp{%} constructs or spaces, or
9368 even newlines. They are processed as usual, as described above.
9369 Trailing white space in @code{X} is ignored. White space may also
9370 appear anywhere on the left side of the colon in these constructs,
9371 except between @code{.} or @code{*} and the corresponding word.
9373 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
9374 handled specifically in these constructs. If another value of
9375 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
9376 @option{-W} switch is found later in the command line, the earlier
9377 switch value is ignored, except with @{@code{S}*@} where @code{S} is
9378 just one letter, which passes all matching options.
9380 The character @samp{|} at the beginning of the predicate text is used to
9381 indicate that a command should be piped to the following command, but
9382 only if @option{-pipe} is specified.
9384 It is built into GCC which switches take arguments and which do not.
9385 (You might think it would be useful to generalize this to allow each
9386 compiler's spec to say which switches take arguments. But this cannot
9387 be done in a consistent fashion. GCC cannot even decide which input
9388 files have been specified without knowing which switches take arguments,
9389 and it must know which input files to compile in order to tell which
9392 GCC also knows implicitly that arguments starting in @option{-l} are to be
9393 treated as compiler output files, and passed to the linker in their
9394 proper position among the other output files.
9396 @c man begin OPTIONS
9398 @node Target Options
9399 @section Specifying Target Machine and Compiler Version
9400 @cindex target options
9401 @cindex cross compiling
9402 @cindex specifying machine version
9403 @cindex specifying compiler version and target machine
9404 @cindex compiler version, specifying
9405 @cindex target machine, specifying
9407 The usual way to run GCC is to run the executable called @file{gcc}, or
9408 @file{<machine>-gcc} when cross-compiling, or
9409 @file{<machine>-gcc-<version>} to run a version other than the one that
9410 was installed last. Sometimes this is inconvenient, so GCC provides
9411 options that will switch to another cross-compiler or version.
9414 @item -b @var{machine}
9416 The argument @var{machine} specifies the target machine for compilation.
9418 The value to use for @var{machine} is the same as was specified as the
9419 machine type when configuring GCC as a cross-compiler. For
9420 example, if a cross-compiler was configured with @samp{configure
9421 arm-elf}, meaning to compile for an arm processor with elf binaries,
9422 then you would specify @option{-b arm-elf} to run that cross compiler.
9423 Because there are other options beginning with @option{-b}, the
9424 configuration must contain a hyphen, or @option{-b} alone should be one
9425 argument followed by the configuration in the next argument.
9427 @item -V @var{version}
9429 The argument @var{version} specifies which version of GCC to run.
9430 This is useful when multiple versions are installed. For example,
9431 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
9434 The @option{-V} and @option{-b} options work by running the
9435 @file{<machine>-gcc-<version>} executable, so there's no real reason to
9436 use them if you can just run that directly.
9438 @node Submodel Options
9439 @section Hardware Models and Configurations
9440 @cindex submodel options
9441 @cindex specifying hardware config
9442 @cindex hardware models and configurations, specifying
9443 @cindex machine dependent options
9445 Earlier we discussed the standard option @option{-b} which chooses among
9446 different installed compilers for completely different target
9447 machines, such as VAX vs.@: 68000 vs.@: 80386.
9449 In addition, each of these target machine types can have its own
9450 special options, starting with @samp{-m}, to choose among various
9451 hardware models or configurations---for example, 68010 vs 68020,
9452 floating coprocessor or none. A single installed version of the
9453 compiler can compile for any model or configuration, according to the
9456 Some configurations of the compiler also support additional special
9457 options, usually for compatibility with other compilers on the same
9460 @c This list is ordered alphanumerically by subsection name.
9461 @c It should be the same order and spelling as these options are listed
9462 @c in Machine Dependent Options
9468 * Blackfin Options::
9472 * DEC Alpha Options::
9473 * DEC Alpha/VMS Options::
9476 * GNU/Linux Options::
9479 * i386 and x86-64 Options::
9480 * i386 and x86-64 Windows Options::
9482 * IA-64/VMS Options::
9493 * picoChip Options::
9495 * RS/6000 and PowerPC Options::
9496 * S/390 and zSeries Options::
9501 * System V Options::
9506 * Xstormy16 Options::
9512 @subsection ARC Options
9515 These options are defined for ARC implementations:
9520 Compile code for little endian mode. This is the default.
9524 Compile code for big endian mode.
9527 @opindex mmangle-cpu
9528 Prepend the name of the cpu to all public symbol names.
9529 In multiple-processor systems, there are many ARC variants with different
9530 instruction and register set characteristics. This flag prevents code
9531 compiled for one cpu to be linked with code compiled for another.
9532 No facility exists for handling variants that are ``almost identical''.
9533 This is an all or nothing option.
9535 @item -mcpu=@var{cpu}
9537 Compile code for ARC variant @var{cpu}.
9538 Which variants are supported depend on the configuration.
9539 All variants support @option{-mcpu=base}, this is the default.
9541 @item -mtext=@var{text-section}
9542 @itemx -mdata=@var{data-section}
9543 @itemx -mrodata=@var{readonly-data-section}
9547 Put functions, data, and readonly data in @var{text-section},
9548 @var{data-section}, and @var{readonly-data-section} respectively
9549 by default. This can be overridden with the @code{section} attribute.
9550 @xref{Variable Attributes}.
9552 @item -mfix-cortex-m3-ldrd
9553 @opindex mfix-cortex-m3-ldrd
9554 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
9555 with overlapping destination and base registers are used. This option avoids
9556 generating these instructions. This option is enabled by default when
9557 @option{-mcpu=cortex-m3} is specified.
9562 @subsection ARM Options
9565 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
9569 @item -mabi=@var{name}
9571 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
9572 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
9575 @opindex mapcs-frame
9576 Generate a stack frame that is compliant with the ARM Procedure Call
9577 Standard for all functions, even if this is not strictly necessary for
9578 correct execution of the code. Specifying @option{-fomit-frame-pointer}
9579 with this option will cause the stack frames not to be generated for
9580 leaf functions. The default is @option{-mno-apcs-frame}.
9584 This is a synonym for @option{-mapcs-frame}.
9587 @c not currently implemented
9588 @item -mapcs-stack-check
9589 @opindex mapcs-stack-check
9590 Generate code to check the amount of stack space available upon entry to
9591 every function (that actually uses some stack space). If there is
9592 insufficient space available then either the function
9593 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
9594 called, depending upon the amount of stack space required. The run time
9595 system is required to provide these functions. The default is
9596 @option{-mno-apcs-stack-check}, since this produces smaller code.
9598 @c not currently implemented
9600 @opindex mapcs-float
9601 Pass floating point arguments using the float point registers. This is
9602 one of the variants of the APCS@. This option is recommended if the
9603 target hardware has a floating point unit or if a lot of floating point
9604 arithmetic is going to be performed by the code. The default is
9605 @option{-mno-apcs-float}, since integer only code is slightly increased in
9606 size if @option{-mapcs-float} is used.
9608 @c not currently implemented
9609 @item -mapcs-reentrant
9610 @opindex mapcs-reentrant
9611 Generate reentrant, position independent code. The default is
9612 @option{-mno-apcs-reentrant}.
9615 @item -mthumb-interwork
9616 @opindex mthumb-interwork
9617 Generate code which supports calling between the ARM and Thumb
9618 instruction sets. Without this option the two instruction sets cannot
9619 be reliably used inside one program. The default is
9620 @option{-mno-thumb-interwork}, since slightly larger code is generated
9621 when @option{-mthumb-interwork} is specified.
9623 @item -mno-sched-prolog
9624 @opindex mno-sched-prolog
9625 Prevent the reordering of instructions in the function prolog, or the
9626 merging of those instruction with the instructions in the function's
9627 body. This means that all functions will start with a recognizable set
9628 of instructions (or in fact one of a choice from a small set of
9629 different function prologues), and this information can be used to
9630 locate the start if functions inside an executable piece of code. The
9631 default is @option{-msched-prolog}.
9633 @item -mfloat-abi=@var{name}
9635 Specifies which floating-point ABI to use. Permissible values
9636 are: @samp{soft}, @samp{softfp} and @samp{hard}.
9638 Specifying @samp{soft} causes GCC to generate output containing
9639 library calls for floating-point operations.
9640 @samp{softfp} allows the generation of code using hardware floating-point
9641 instructions, but still uses the soft-float calling conventions.
9642 @samp{hard} allows generation of floating-point instructions
9643 and uses FPU-specific calling conventions.
9645 The default depends on the specific target configuration. Note that
9646 the hard-float and soft-float ABIs are not link-compatible; you must
9647 compile your entire program with the same ABI, and link with a
9648 compatible set of libraries.
9651 @opindex mhard-float
9652 Equivalent to @option{-mfloat-abi=hard}.
9655 @opindex msoft-float
9656 Equivalent to @option{-mfloat-abi=soft}.
9658 @item -mlittle-endian
9659 @opindex mlittle-endian
9660 Generate code for a processor running in little-endian mode. This is
9661 the default for all standard configurations.
9664 @opindex mbig-endian
9665 Generate code for a processor running in big-endian mode; the default is
9666 to compile code for a little-endian processor.
9668 @item -mwords-little-endian
9669 @opindex mwords-little-endian
9670 This option only applies when generating code for big-endian processors.
9671 Generate code for a little-endian word order but a big-endian byte
9672 order. That is, a byte order of the form @samp{32107654}. Note: this
9673 option should only be used if you require compatibility with code for
9674 big-endian ARM processors generated by versions of the compiler prior to
9677 @item -mcpu=@var{name}
9679 This specifies the name of the target ARM processor. GCC uses this name
9680 to determine what kind of instructions it can emit when generating
9681 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
9682 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
9683 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
9684 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
9685 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
9687 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
9688 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
9689 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
9690 @samp{strongarm1110},
9691 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
9692 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
9693 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
9694 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
9695 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
9696 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
9697 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
9698 @samp{cortex-a8}, @samp{cortex-a9},
9699 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-m3},
9702 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9704 @item -mtune=@var{name}
9706 This option is very similar to the @option{-mcpu=} option, except that
9707 instead of specifying the actual target processor type, and hence
9708 restricting which instructions can be used, it specifies that GCC should
9709 tune the performance of the code as if the target were of the type
9710 specified in this option, but still choosing the instructions that it
9711 will generate based on the cpu specified by a @option{-mcpu=} option.
9712 For some ARM implementations better performance can be obtained by using
9715 @item -march=@var{name}
9717 This specifies the name of the target ARM architecture. GCC uses this
9718 name to determine what kind of instructions it can emit when generating
9719 assembly code. This option can be used in conjunction with or instead
9720 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
9721 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
9722 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
9723 @samp{armv6}, @samp{armv6j},
9724 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
9725 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
9726 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9728 @item -mfpu=@var{name}
9729 @itemx -mfpe=@var{number}
9730 @itemx -mfp=@var{number}
9734 This specifies what floating point hardware (or hardware emulation) is
9735 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
9736 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-d16},
9737 @samp{neon}, and @samp{neon-fp16}. @option{-mfp} and @option{-mfpe}
9738 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
9739 with older versions of GCC@.
9741 If @option{-msoft-float} is specified this specifies the format of
9742 floating point values.
9744 @item -mfp16-format=@var{name}
9745 @opindex mfp16-format
9746 Specify the format of the @code{__fp16} half-precision floating-point type.
9747 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
9748 the default is @samp{none}, in which case the @code{__fp16} type is not
9749 defined. @xref{Half-Precision}, for more information.
9751 @item -mstructure-size-boundary=@var{n}
9752 @opindex mstructure-size-boundary
9753 The size of all structures and unions will be rounded up to a multiple
9754 of the number of bits set by this option. Permissible values are 8, 32
9755 and 64. The default value varies for different toolchains. For the COFF
9756 targeted toolchain the default value is 8. A value of 64 is only allowed
9757 if the underlying ABI supports it.
9759 Specifying the larger number can produce faster, more efficient code, but
9760 can also increase the size of the program. Different values are potentially
9761 incompatible. Code compiled with one value cannot necessarily expect to
9762 work with code or libraries compiled with another value, if they exchange
9763 information using structures or unions.
9765 @item -mabort-on-noreturn
9766 @opindex mabort-on-noreturn
9767 Generate a call to the function @code{abort} at the end of a
9768 @code{noreturn} function. It will be executed if the function tries to
9772 @itemx -mno-long-calls
9773 @opindex mlong-calls
9774 @opindex mno-long-calls
9775 Tells the compiler to perform function calls by first loading the
9776 address of the function into a register and then performing a subroutine
9777 call on this register. This switch is needed if the target function
9778 will lie outside of the 64 megabyte addressing range of the offset based
9779 version of subroutine call instruction.
9781 Even if this switch is enabled, not all function calls will be turned
9782 into long calls. The heuristic is that static functions, functions
9783 which have the @samp{short-call} attribute, functions that are inside
9784 the scope of a @samp{#pragma no_long_calls} directive and functions whose
9785 definitions have already been compiled within the current compilation
9786 unit, will not be turned into long calls. The exception to this rule is
9787 that weak function definitions, functions with the @samp{long-call}
9788 attribute or the @samp{section} attribute, and functions that are within
9789 the scope of a @samp{#pragma long_calls} directive, will always be
9790 turned into long calls.
9792 This feature is not enabled by default. Specifying
9793 @option{-mno-long-calls} will restore the default behavior, as will
9794 placing the function calls within the scope of a @samp{#pragma
9795 long_calls_off} directive. Note these switches have no effect on how
9796 the compiler generates code to handle function calls via function
9799 @item -msingle-pic-base
9800 @opindex msingle-pic-base
9801 Treat the register used for PIC addressing as read-only, rather than
9802 loading it in the prologue for each function. The run-time system is
9803 responsible for initializing this register with an appropriate value
9804 before execution begins.
9806 @item -mpic-register=@var{reg}
9807 @opindex mpic-register
9808 Specify the register to be used for PIC addressing. The default is R10
9809 unless stack-checking is enabled, when R9 is used.
9811 @item -mcirrus-fix-invalid-insns
9812 @opindex mcirrus-fix-invalid-insns
9813 @opindex mno-cirrus-fix-invalid-insns
9814 Insert NOPs into the instruction stream to in order to work around
9815 problems with invalid Maverick instruction combinations. This option
9816 is only valid if the @option{-mcpu=ep9312} option has been used to
9817 enable generation of instructions for the Cirrus Maverick floating
9818 point co-processor. This option is not enabled by default, since the
9819 problem is only present in older Maverick implementations. The default
9820 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
9823 @item -mpoke-function-name
9824 @opindex mpoke-function-name
9825 Write the name of each function into the text section, directly
9826 preceding the function prologue. The generated code is similar to this:
9830 .ascii "arm_poke_function_name", 0
9833 .word 0xff000000 + (t1 - t0)
9834 arm_poke_function_name
9836 stmfd sp!, @{fp, ip, lr, pc@}
9840 When performing a stack backtrace, code can inspect the value of
9841 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
9842 location @code{pc - 12} and the top 8 bits are set, then we know that
9843 there is a function name embedded immediately preceding this location
9844 and has length @code{((pc[-3]) & 0xff000000)}.
9848 Generate code for the Thumb instruction set. The default is to
9849 use the 32-bit ARM instruction set.
9850 This option automatically enables either 16-bit Thumb-1 or
9851 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
9852 and @option{-march=@var{name}} options. This option is not passed to the
9853 assembler. If you want to force assembler files to be interpreted as Thumb code,
9854 either add a @samp{.thumb} directive to the source or pass the @option{-mthumb}
9855 option directly to the assembler by prefixing it with @option{-Wa}.
9858 @opindex mtpcs-frame
9859 Generate a stack frame that is compliant with the Thumb Procedure Call
9860 Standard for all non-leaf functions. (A leaf function is one that does
9861 not call any other functions.) The default is @option{-mno-tpcs-frame}.
9863 @item -mtpcs-leaf-frame
9864 @opindex mtpcs-leaf-frame
9865 Generate a stack frame that is compliant with the Thumb Procedure Call
9866 Standard for all leaf functions. (A leaf function is one that does
9867 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
9869 @item -mcallee-super-interworking
9870 @opindex mcallee-super-interworking
9871 Gives all externally visible functions in the file being compiled an ARM
9872 instruction set header which switches to Thumb mode before executing the
9873 rest of the function. This allows these functions to be called from
9874 non-interworking code. This option is not valid in AAPCS configurations
9875 because interworking is enabled by default.
9877 @item -mcaller-super-interworking
9878 @opindex mcaller-super-interworking
9879 Allows calls via function pointers (including virtual functions) to
9880 execute correctly regardless of whether the target code has been
9881 compiled for interworking or not. There is a small overhead in the cost
9882 of executing a function pointer if this option is enabled. This option
9883 is not valid in AAPCS configurations because interworking is enabled
9886 @item -mtp=@var{name}
9888 Specify the access model for the thread local storage pointer. The valid
9889 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
9890 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
9891 (supported in the arm6k architecture), and @option{auto}, which uses the
9892 best available method for the selected processor. The default setting is
9895 @item -mword-relocations
9896 @opindex mword-relocations
9897 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
9898 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
9899 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
9905 @subsection AVR Options
9908 These options are defined for AVR implementations:
9911 @item -mmcu=@var{mcu}
9913 Specify ATMEL AVR instruction set or MCU type.
9915 Instruction set avr1 is for the minimal AVR core, not supported by the C
9916 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
9917 attiny11, attiny12, attiny15, attiny28).
9919 Instruction set avr2 (default) is for the classic AVR core with up to
9920 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
9921 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
9922 at90c8534, at90s8535).
9924 Instruction set avr3 is for the classic AVR core with up to 128K program
9925 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
9927 Instruction set avr4 is for the enhanced AVR core with up to 8K program
9928 memory space (MCU types: atmega8, atmega83, atmega85).
9930 Instruction set avr5 is for the enhanced AVR core with up to 128K program
9931 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
9932 atmega64, atmega128, at43usb355, at94k).
9936 Output instruction sizes to the asm file.
9938 @item -mno-interrupts
9939 @opindex mno-interrupts
9940 Generated code is not compatible with hardware interrupts.
9941 Code size will be smaller.
9943 @item -mcall-prologues
9944 @opindex mcall-prologues
9945 Functions prologues/epilogues expanded as call to appropriate
9946 subroutines. Code size will be smaller.
9949 @opindex mtiny-stack
9950 Change only the low 8 bits of the stack pointer.
9954 Assume int to be 8 bit integer. This affects the sizes of all types: A
9955 char will be 1 byte, an int will be 1 byte, a long will be 2 bytes
9956 and long long will be 4 bytes. Please note that this option does not
9957 comply to the C standards, but it will provide you with smaller code
9961 @node Blackfin Options
9962 @subsection Blackfin Options
9963 @cindex Blackfin Options
9966 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
9968 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
9969 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
9970 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
9971 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
9972 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
9973 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
9974 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
9976 The optional @var{sirevision} specifies the silicon revision of the target
9977 Blackfin processor. Any workarounds available for the targeted silicon revision
9978 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
9979 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
9980 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
9981 hexadecimal digits representing the major and minor numbers in the silicon
9982 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
9983 is not defined. If @var{sirevision} is @samp{any}, the
9984 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
9985 If this optional @var{sirevision} is not used, GCC assumes the latest known
9986 silicon revision of the targeted Blackfin processor.
9988 Support for @samp{bf561} is incomplete. For @samp{bf561},
9989 Only the processor macro is defined.
9990 Without this option, @samp{bf532} is used as the processor by default.
9991 The corresponding predefined processor macros for @var{cpu} is to
9992 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
9993 provided by libgloss to be linked in if @option{-msim} is not given.
9997 Specifies that the program will be run on the simulator. This causes
9998 the simulator BSP provided by libgloss to be linked in. This option
9999 has effect only for @samp{bfin-elf} toolchain.
10000 Certain other options, such as @option{-mid-shared-library} and
10001 @option{-mfdpic}, imply @option{-msim}.
10003 @item -momit-leaf-frame-pointer
10004 @opindex momit-leaf-frame-pointer
10005 Don't keep the frame pointer in a register for leaf functions. This
10006 avoids the instructions to save, set up and restore frame pointers and
10007 makes an extra register available in leaf functions. The option
10008 @option{-fomit-frame-pointer} removes the frame pointer for all functions
10009 which might make debugging harder.
10011 @item -mspecld-anomaly
10012 @opindex mspecld-anomaly
10013 When enabled, the compiler will ensure that the generated code does not
10014 contain speculative loads after jump instructions. If this option is used,
10015 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
10017 @item -mno-specld-anomaly
10018 @opindex mno-specld-anomaly
10019 Don't generate extra code to prevent speculative loads from occurring.
10021 @item -mcsync-anomaly
10022 @opindex mcsync-anomaly
10023 When enabled, the compiler will ensure that the generated code does not
10024 contain CSYNC or SSYNC instructions too soon after conditional branches.
10025 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
10027 @item -mno-csync-anomaly
10028 @opindex mno-csync-anomaly
10029 Don't generate extra code to prevent CSYNC or SSYNC instructions from
10030 occurring too soon after a conditional branch.
10034 When enabled, the compiler is free to take advantage of the knowledge that
10035 the entire program fits into the low 64k of memory.
10038 @opindex mno-low-64k
10039 Assume that the program is arbitrarily large. This is the default.
10041 @item -mstack-check-l1
10042 @opindex mstack-check-l1
10043 Do stack checking using information placed into L1 scratchpad memory by the
10046 @item -mid-shared-library
10047 @opindex mid-shared-library
10048 Generate code that supports shared libraries via the library ID method.
10049 This allows for execute in place and shared libraries in an environment
10050 without virtual memory management. This option implies @option{-fPIC}.
10051 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10053 @item -mno-id-shared-library
10054 @opindex mno-id-shared-library
10055 Generate code that doesn't assume ID based shared libraries are being used.
10056 This is the default.
10058 @item -mleaf-id-shared-library
10059 @opindex mleaf-id-shared-library
10060 Generate code that supports shared libraries via the library ID method,
10061 but assumes that this library or executable won't link against any other
10062 ID shared libraries. That allows the compiler to use faster code for jumps
10065 @item -mno-leaf-id-shared-library
10066 @opindex mno-leaf-id-shared-library
10067 Do not assume that the code being compiled won't link against any ID shared
10068 libraries. Slower code will be generated for jump and call insns.
10070 @item -mshared-library-id=n
10071 @opindex mshared-library-id
10072 Specified the identification number of the ID based shared library being
10073 compiled. Specifying a value of 0 will generate more compact code, specifying
10074 other values will force the allocation of that number to the current
10075 library but is no more space or time efficient than omitting this option.
10079 Generate code that allows the data segment to be located in a different
10080 area of memory from the text segment. This allows for execute in place in
10081 an environment without virtual memory management by eliminating relocations
10082 against the text section.
10084 @item -mno-sep-data
10085 @opindex mno-sep-data
10086 Generate code that assumes that the data segment follows the text segment.
10087 This is the default.
10090 @itemx -mno-long-calls
10091 @opindex mlong-calls
10092 @opindex mno-long-calls
10093 Tells the compiler to perform function calls by first loading the
10094 address of the function into a register and then performing a subroutine
10095 call on this register. This switch is needed if the target function
10096 will lie outside of the 24 bit addressing range of the offset based
10097 version of subroutine call instruction.
10099 This feature is not enabled by default. Specifying
10100 @option{-mno-long-calls} will restore the default behavior. Note these
10101 switches have no effect on how the compiler generates code to handle
10102 function calls via function pointers.
10106 Link with the fast floating-point library. This library relaxes some of
10107 the IEEE floating-point standard's rules for checking inputs against
10108 Not-a-Number (NAN), in the interest of performance.
10111 @opindex minline-plt
10112 Enable inlining of PLT entries in function calls to functions that are
10113 not known to bind locally. It has no effect without @option{-mfdpic}.
10116 @opindex mmulticore
10117 Build standalone application for multicore Blackfin processor. Proper
10118 start files and link scripts will be used to support multicore.
10119 This option defines @code{__BFIN_MULTICORE}. It can only be used with
10120 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
10121 @option{-mcorea} or @option{-mcoreb}. If it's used without
10122 @option{-mcorea} or @option{-mcoreb}, single application/dual core
10123 programming model is used. In this model, the main function of Core B
10124 should be named as coreb_main. If it's used with @option{-mcorea} or
10125 @option{-mcoreb}, one application per core programming model is used.
10126 If this option is not used, single core application programming
10131 Build standalone application for Core A of BF561 when using
10132 one application per core programming model. Proper start files
10133 and link scripts will be used to support Core A. This option
10134 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
10138 Build standalone application for Core B of BF561 when using
10139 one application per core programming model. Proper start files
10140 and link scripts will be used to support Core B. This option
10141 defines @code{__BFIN_COREB}. When this option is used, coreb_main
10142 should be used instead of main. It must be used with
10143 @option{-mmulticore}.
10147 Build standalone application for SDRAM. Proper start files and
10148 link scripts will be used to put the application into SDRAM.
10149 Loader should initialize SDRAM before loading the application
10150 into SDRAM. This option defines @code{__BFIN_SDRAM}.
10154 Assume that ICPLBs are enabled at runtime. This has an effect on certain
10155 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
10156 are enabled; for standalone applications the default is off.
10160 @subsection CRIS Options
10161 @cindex CRIS Options
10163 These options are defined specifically for the CRIS ports.
10166 @item -march=@var{architecture-type}
10167 @itemx -mcpu=@var{architecture-type}
10170 Generate code for the specified architecture. The choices for
10171 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
10172 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
10173 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
10176 @item -mtune=@var{architecture-type}
10178 Tune to @var{architecture-type} everything applicable about the generated
10179 code, except for the ABI and the set of available instructions. The
10180 choices for @var{architecture-type} are the same as for
10181 @option{-march=@var{architecture-type}}.
10183 @item -mmax-stack-frame=@var{n}
10184 @opindex mmax-stack-frame
10185 Warn when the stack frame of a function exceeds @var{n} bytes.
10191 The options @option{-metrax4} and @option{-metrax100} are synonyms for
10192 @option{-march=v3} and @option{-march=v8} respectively.
10194 @item -mmul-bug-workaround
10195 @itemx -mno-mul-bug-workaround
10196 @opindex mmul-bug-workaround
10197 @opindex mno-mul-bug-workaround
10198 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
10199 models where it applies. This option is active by default.
10203 Enable CRIS-specific verbose debug-related information in the assembly
10204 code. This option also has the effect to turn off the @samp{#NO_APP}
10205 formatted-code indicator to the assembler at the beginning of the
10210 Do not use condition-code results from previous instruction; always emit
10211 compare and test instructions before use of condition codes.
10213 @item -mno-side-effects
10214 @opindex mno-side-effects
10215 Do not emit instructions with side-effects in addressing modes other than
10218 @item -mstack-align
10219 @itemx -mno-stack-align
10220 @itemx -mdata-align
10221 @itemx -mno-data-align
10222 @itemx -mconst-align
10223 @itemx -mno-const-align
10224 @opindex mstack-align
10225 @opindex mno-stack-align
10226 @opindex mdata-align
10227 @opindex mno-data-align
10228 @opindex mconst-align
10229 @opindex mno-const-align
10230 These options (no-options) arranges (eliminate arrangements) for the
10231 stack-frame, individual data and constants to be aligned for the maximum
10232 single data access size for the chosen CPU model. The default is to
10233 arrange for 32-bit alignment. ABI details such as structure layout are
10234 not affected by these options.
10242 Similar to the stack- data- and const-align options above, these options
10243 arrange for stack-frame, writable data and constants to all be 32-bit,
10244 16-bit or 8-bit aligned. The default is 32-bit alignment.
10246 @item -mno-prologue-epilogue
10247 @itemx -mprologue-epilogue
10248 @opindex mno-prologue-epilogue
10249 @opindex mprologue-epilogue
10250 With @option{-mno-prologue-epilogue}, the normal function prologue and
10251 epilogue that sets up the stack-frame are omitted and no return
10252 instructions or return sequences are generated in the code. Use this
10253 option only together with visual inspection of the compiled code: no
10254 warnings or errors are generated when call-saved registers must be saved,
10255 or storage for local variable needs to be allocated.
10259 @opindex mno-gotplt
10261 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
10262 instruction sequences that load addresses for functions from the PLT part
10263 of the GOT rather than (traditional on other architectures) calls to the
10264 PLT@. The default is @option{-mgotplt}.
10268 Legacy no-op option only recognized with the cris-axis-elf and
10269 cris-axis-linux-gnu targets.
10273 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
10277 This option, recognized for the cris-axis-elf arranges
10278 to link with input-output functions from a simulator library. Code,
10279 initialized data and zero-initialized data are allocated consecutively.
10283 Like @option{-sim}, but pass linker options to locate initialized data at
10284 0x40000000 and zero-initialized data at 0x80000000.
10288 @subsection CRX Options
10289 @cindex CRX Options
10291 These options are defined specifically for the CRX ports.
10297 Enable the use of multiply-accumulate instructions. Disabled by default.
10300 @opindex mpush-args
10301 Push instructions will be used to pass outgoing arguments when functions
10302 are called. Enabled by default.
10305 @node Darwin Options
10306 @subsection Darwin Options
10307 @cindex Darwin options
10309 These options are defined for all architectures running the Darwin operating
10312 FSF GCC on Darwin does not create ``fat'' object files; it will create
10313 an object file for the single architecture that it was built to
10314 target. Apple's GCC on Darwin does create ``fat'' files if multiple
10315 @option{-arch} options are used; it does so by running the compiler or
10316 linker multiple times and joining the results together with
10319 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
10320 @samp{i686}) is determined by the flags that specify the ISA
10321 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
10322 @option{-force_cpusubtype_ALL} option can be used to override this.
10324 The Darwin tools vary in their behavior when presented with an ISA
10325 mismatch. The assembler, @file{as}, will only permit instructions to
10326 be used that are valid for the subtype of the file it is generating,
10327 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
10328 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
10329 and print an error if asked to create a shared library with a less
10330 restrictive subtype than its input files (for instance, trying to put
10331 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
10332 for executables, @file{ld}, will quietly give the executable the most
10333 restrictive subtype of any of its input files.
10338 Add the framework directory @var{dir} to the head of the list of
10339 directories to be searched for header files. These directories are
10340 interleaved with those specified by @option{-I} options and are
10341 scanned in a left-to-right order.
10343 A framework directory is a directory with frameworks in it. A
10344 framework is a directory with a @samp{"Headers"} and/or
10345 @samp{"PrivateHeaders"} directory contained directly in it that ends
10346 in @samp{".framework"}. The name of a framework is the name of this
10347 directory excluding the @samp{".framework"}. Headers associated with
10348 the framework are found in one of those two directories, with
10349 @samp{"Headers"} being searched first. A subframework is a framework
10350 directory that is in a framework's @samp{"Frameworks"} directory.
10351 Includes of subframework headers can only appear in a header of a
10352 framework that contains the subframework, or in a sibling subframework
10353 header. Two subframeworks are siblings if they occur in the same
10354 framework. A subframework should not have the same name as a
10355 framework, a warning will be issued if this is violated. Currently a
10356 subframework cannot have subframeworks, in the future, the mechanism
10357 may be extended to support this. The standard frameworks can be found
10358 in @samp{"/System/Library/Frameworks"} and
10359 @samp{"/Library/Frameworks"}. An example include looks like
10360 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
10361 the name of the framework and header.h is found in the
10362 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
10364 @item -iframework@var{dir}
10365 @opindex iframework
10366 Like @option{-F} except the directory is a treated as a system
10367 directory. The main difference between this @option{-iframework} and
10368 @option{-F} is that with @option{-iframework} the compiler does not
10369 warn about constructs contained within header files found via
10370 @var{dir}. This option is valid only for the C family of languages.
10374 Emit debugging information for symbols that are used. For STABS
10375 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
10376 This is by default ON@.
10380 Emit debugging information for all symbols and types.
10382 @item -mmacosx-version-min=@var{version}
10383 The earliest version of MacOS X that this executable will run on
10384 is @var{version}. Typical values of @var{version} include @code{10.1},
10385 @code{10.2}, and @code{10.3.9}.
10387 If the compiler was built to use the system's headers by default,
10388 then the default for this option is the system version on which the
10389 compiler is running, otherwise the default is to make choices which
10390 are compatible with as many systems and code bases as possible.
10394 Enable kernel development mode. The @option{-mkernel} option sets
10395 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
10396 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
10397 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
10398 applicable. This mode also sets @option{-mno-altivec},
10399 @option{-msoft-float}, @option{-fno-builtin} and
10400 @option{-mlong-branch} for PowerPC targets.
10402 @item -mone-byte-bool
10403 @opindex mone-byte-bool
10404 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
10405 By default @samp{sizeof(bool)} is @samp{4} when compiling for
10406 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
10407 option has no effect on x86.
10409 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
10410 to generate code that is not binary compatible with code generated
10411 without that switch. Using this switch may require recompiling all
10412 other modules in a program, including system libraries. Use this
10413 switch to conform to a non-default data model.
10415 @item -mfix-and-continue
10416 @itemx -ffix-and-continue
10417 @itemx -findirect-data
10418 @opindex mfix-and-continue
10419 @opindex ffix-and-continue
10420 @opindex findirect-data
10421 Generate code suitable for fast turn around development. Needed to
10422 enable gdb to dynamically load @code{.o} files into already running
10423 programs. @option{-findirect-data} and @option{-ffix-and-continue}
10424 are provided for backwards compatibility.
10428 Loads all members of static archive libraries.
10429 See man ld(1) for more information.
10431 @item -arch_errors_fatal
10432 @opindex arch_errors_fatal
10433 Cause the errors having to do with files that have the wrong architecture
10436 @item -bind_at_load
10437 @opindex bind_at_load
10438 Causes the output file to be marked such that the dynamic linker will
10439 bind all undefined references when the file is loaded or launched.
10443 Produce a Mach-o bundle format file.
10444 See man ld(1) for more information.
10446 @item -bundle_loader @var{executable}
10447 @opindex bundle_loader
10448 This option specifies the @var{executable} that will be loading the build
10449 output file being linked. See man ld(1) for more information.
10452 @opindex dynamiclib
10453 When passed this option, GCC will produce a dynamic library instead of
10454 an executable when linking, using the Darwin @file{libtool} command.
10456 @item -force_cpusubtype_ALL
10457 @opindex force_cpusubtype_ALL
10458 This causes GCC's output file to have the @var{ALL} subtype, instead of
10459 one controlled by the @option{-mcpu} or @option{-march} option.
10461 @item -allowable_client @var{client_name}
10462 @itemx -client_name
10463 @itemx -compatibility_version
10464 @itemx -current_version
10466 @itemx -dependency-file
10468 @itemx -dylinker_install_name
10470 @itemx -exported_symbols_list
10472 @itemx -flat_namespace
10473 @itemx -force_flat_namespace
10474 @itemx -headerpad_max_install_names
10477 @itemx -install_name
10478 @itemx -keep_private_externs
10479 @itemx -multi_module
10480 @itemx -multiply_defined
10481 @itemx -multiply_defined_unused
10483 @itemx -no_dead_strip_inits_and_terms
10484 @itemx -nofixprebinding
10485 @itemx -nomultidefs
10487 @itemx -noseglinkedit
10488 @itemx -pagezero_size
10490 @itemx -prebind_all_twolevel_modules
10491 @itemx -private_bundle
10492 @itemx -read_only_relocs
10494 @itemx -sectobjectsymbols
10498 @itemx -sectobjectsymbols
10501 @itemx -segs_read_only_addr
10502 @itemx -segs_read_write_addr
10503 @itemx -seg_addr_table
10504 @itemx -seg_addr_table_filename
10505 @itemx -seglinkedit
10507 @itemx -segs_read_only_addr
10508 @itemx -segs_read_write_addr
10509 @itemx -single_module
10511 @itemx -sub_library
10512 @itemx -sub_umbrella
10513 @itemx -twolevel_namespace
10516 @itemx -unexported_symbols_list
10517 @itemx -weak_reference_mismatches
10518 @itemx -whatsloaded
10519 @opindex allowable_client
10520 @opindex client_name
10521 @opindex compatibility_version
10522 @opindex current_version
10523 @opindex dead_strip
10524 @opindex dependency-file
10525 @opindex dylib_file
10526 @opindex dylinker_install_name
10528 @opindex exported_symbols_list
10530 @opindex flat_namespace
10531 @opindex force_flat_namespace
10532 @opindex headerpad_max_install_names
10533 @opindex image_base
10535 @opindex install_name
10536 @opindex keep_private_externs
10537 @opindex multi_module
10538 @opindex multiply_defined
10539 @opindex multiply_defined_unused
10540 @opindex noall_load
10541 @opindex no_dead_strip_inits_and_terms
10542 @opindex nofixprebinding
10543 @opindex nomultidefs
10545 @opindex noseglinkedit
10546 @opindex pagezero_size
10548 @opindex prebind_all_twolevel_modules
10549 @opindex private_bundle
10550 @opindex read_only_relocs
10552 @opindex sectobjectsymbols
10555 @opindex sectcreate
10556 @opindex sectobjectsymbols
10559 @opindex segs_read_only_addr
10560 @opindex segs_read_write_addr
10561 @opindex seg_addr_table
10562 @opindex seg_addr_table_filename
10563 @opindex seglinkedit
10565 @opindex segs_read_only_addr
10566 @opindex segs_read_write_addr
10567 @opindex single_module
10569 @opindex sub_library
10570 @opindex sub_umbrella
10571 @opindex twolevel_namespace
10574 @opindex unexported_symbols_list
10575 @opindex weak_reference_mismatches
10576 @opindex whatsloaded
10577 These options are passed to the Darwin linker. The Darwin linker man page
10578 describes them in detail.
10581 @node DEC Alpha Options
10582 @subsection DEC Alpha Options
10584 These @samp{-m} options are defined for the DEC Alpha implementations:
10587 @item -mno-soft-float
10588 @itemx -msoft-float
10589 @opindex mno-soft-float
10590 @opindex msoft-float
10591 Use (do not use) the hardware floating-point instructions for
10592 floating-point operations. When @option{-msoft-float} is specified,
10593 functions in @file{libgcc.a} will be used to perform floating-point
10594 operations. Unless they are replaced by routines that emulate the
10595 floating-point operations, or compiled in such a way as to call such
10596 emulations routines, these routines will issue floating-point
10597 operations. If you are compiling for an Alpha without floating-point
10598 operations, you must ensure that the library is built so as not to call
10601 Note that Alpha implementations without floating-point operations are
10602 required to have floating-point registers.
10605 @itemx -mno-fp-regs
10607 @opindex mno-fp-regs
10608 Generate code that uses (does not use) the floating-point register set.
10609 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
10610 register set is not used, floating point operands are passed in integer
10611 registers as if they were integers and floating-point results are passed
10612 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
10613 so any function with a floating-point argument or return value called by code
10614 compiled with @option{-mno-fp-regs} must also be compiled with that
10617 A typical use of this option is building a kernel that does not use,
10618 and hence need not save and restore, any floating-point registers.
10622 The Alpha architecture implements floating-point hardware optimized for
10623 maximum performance. It is mostly compliant with the IEEE floating
10624 point standard. However, for full compliance, software assistance is
10625 required. This option generates code fully IEEE compliant code
10626 @emph{except} that the @var{inexact-flag} is not maintained (see below).
10627 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
10628 defined during compilation. The resulting code is less efficient but is
10629 able to correctly support denormalized numbers and exceptional IEEE
10630 values such as not-a-number and plus/minus infinity. Other Alpha
10631 compilers call this option @option{-ieee_with_no_inexact}.
10633 @item -mieee-with-inexact
10634 @opindex mieee-with-inexact
10635 This is like @option{-mieee} except the generated code also maintains
10636 the IEEE @var{inexact-flag}. Turning on this option causes the
10637 generated code to implement fully-compliant IEEE math. In addition to
10638 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
10639 macro. On some Alpha implementations the resulting code may execute
10640 significantly slower than the code generated by default. Since there is
10641 very little code that depends on the @var{inexact-flag}, you should
10642 normally not specify this option. Other Alpha compilers call this
10643 option @option{-ieee_with_inexact}.
10645 @item -mfp-trap-mode=@var{trap-mode}
10646 @opindex mfp-trap-mode
10647 This option controls what floating-point related traps are enabled.
10648 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
10649 The trap mode can be set to one of four values:
10653 This is the default (normal) setting. The only traps that are enabled
10654 are the ones that cannot be disabled in software (e.g., division by zero
10658 In addition to the traps enabled by @samp{n}, underflow traps are enabled
10662 Like @samp{u}, but the instructions are marked to be safe for software
10663 completion (see Alpha architecture manual for details).
10666 Like @samp{su}, but inexact traps are enabled as well.
10669 @item -mfp-rounding-mode=@var{rounding-mode}
10670 @opindex mfp-rounding-mode
10671 Selects the IEEE rounding mode. Other Alpha compilers call this option
10672 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
10677 Normal IEEE rounding mode. Floating point numbers are rounded towards
10678 the nearest machine number or towards the even machine number in case
10682 Round towards minus infinity.
10685 Chopped rounding mode. Floating point numbers are rounded towards zero.
10688 Dynamic rounding mode. A field in the floating point control register
10689 (@var{fpcr}, see Alpha architecture reference manual) controls the
10690 rounding mode in effect. The C library initializes this register for
10691 rounding towards plus infinity. Thus, unless your program modifies the
10692 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
10695 @item -mtrap-precision=@var{trap-precision}
10696 @opindex mtrap-precision
10697 In the Alpha architecture, floating point traps are imprecise. This
10698 means without software assistance it is impossible to recover from a
10699 floating trap and program execution normally needs to be terminated.
10700 GCC can generate code that can assist operating system trap handlers
10701 in determining the exact location that caused a floating point trap.
10702 Depending on the requirements of an application, different levels of
10703 precisions can be selected:
10707 Program precision. This option is the default and means a trap handler
10708 can only identify which program caused a floating point exception.
10711 Function precision. The trap handler can determine the function that
10712 caused a floating point exception.
10715 Instruction precision. The trap handler can determine the exact
10716 instruction that caused a floating point exception.
10719 Other Alpha compilers provide the equivalent options called
10720 @option{-scope_safe} and @option{-resumption_safe}.
10722 @item -mieee-conformant
10723 @opindex mieee-conformant
10724 This option marks the generated code as IEEE conformant. You must not
10725 use this option unless you also specify @option{-mtrap-precision=i} and either
10726 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
10727 is to emit the line @samp{.eflag 48} in the function prologue of the
10728 generated assembly file. Under DEC Unix, this has the effect that
10729 IEEE-conformant math library routines will be linked in.
10731 @item -mbuild-constants
10732 @opindex mbuild-constants
10733 Normally GCC examines a 32- or 64-bit integer constant to
10734 see if it can construct it from smaller constants in two or three
10735 instructions. If it cannot, it will output the constant as a literal and
10736 generate code to load it from the data segment at runtime.
10738 Use this option to require GCC to construct @emph{all} integer constants
10739 using code, even if it takes more instructions (the maximum is six).
10741 You would typically use this option to build a shared library dynamic
10742 loader. Itself a shared library, it must relocate itself in memory
10743 before it can find the variables and constants in its own data segment.
10749 Select whether to generate code to be assembled by the vendor-supplied
10750 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
10768 Indicate whether GCC should generate code to use the optional BWX,
10769 CIX, FIX and MAX instruction sets. The default is to use the instruction
10770 sets supported by the CPU type specified via @option{-mcpu=} option or that
10771 of the CPU on which GCC was built if none was specified.
10774 @itemx -mfloat-ieee
10775 @opindex mfloat-vax
10776 @opindex mfloat-ieee
10777 Generate code that uses (does not use) VAX F and G floating point
10778 arithmetic instead of IEEE single and double precision.
10780 @item -mexplicit-relocs
10781 @itemx -mno-explicit-relocs
10782 @opindex mexplicit-relocs
10783 @opindex mno-explicit-relocs
10784 Older Alpha assemblers provided no way to generate symbol relocations
10785 except via assembler macros. Use of these macros does not allow
10786 optimal instruction scheduling. GNU binutils as of version 2.12
10787 supports a new syntax that allows the compiler to explicitly mark
10788 which relocations should apply to which instructions. This option
10789 is mostly useful for debugging, as GCC detects the capabilities of
10790 the assembler when it is built and sets the default accordingly.
10793 @itemx -mlarge-data
10794 @opindex msmall-data
10795 @opindex mlarge-data
10796 When @option{-mexplicit-relocs} is in effect, static data is
10797 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
10798 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
10799 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
10800 16-bit relocations off of the @code{$gp} register. This limits the
10801 size of the small data area to 64KB, but allows the variables to be
10802 directly accessed via a single instruction.
10804 The default is @option{-mlarge-data}. With this option the data area
10805 is limited to just below 2GB@. Programs that require more than 2GB of
10806 data must use @code{malloc} or @code{mmap} to allocate the data in the
10807 heap instead of in the program's data segment.
10809 When generating code for shared libraries, @option{-fpic} implies
10810 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
10813 @itemx -mlarge-text
10814 @opindex msmall-text
10815 @opindex mlarge-text
10816 When @option{-msmall-text} is used, the compiler assumes that the
10817 code of the entire program (or shared library) fits in 4MB, and is
10818 thus reachable with a branch instruction. When @option{-msmall-data}
10819 is used, the compiler can assume that all local symbols share the
10820 same @code{$gp} value, and thus reduce the number of instructions
10821 required for a function call from 4 to 1.
10823 The default is @option{-mlarge-text}.
10825 @item -mcpu=@var{cpu_type}
10827 Set the instruction set and instruction scheduling parameters for
10828 machine type @var{cpu_type}. You can specify either the @samp{EV}
10829 style name or the corresponding chip number. GCC supports scheduling
10830 parameters for the EV4, EV5 and EV6 family of processors and will
10831 choose the default values for the instruction set from the processor
10832 you specify. If you do not specify a processor type, GCC will default
10833 to the processor on which the compiler was built.
10835 Supported values for @var{cpu_type} are
10841 Schedules as an EV4 and has no instruction set extensions.
10845 Schedules as an EV5 and has no instruction set extensions.
10849 Schedules as an EV5 and supports the BWX extension.
10854 Schedules as an EV5 and supports the BWX and MAX extensions.
10858 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
10862 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
10865 Native Linux/GNU toolchains also support the value @samp{native},
10866 which selects the best architecture option for the host processor.
10867 @option{-mcpu=native} has no effect if GCC does not recognize
10870 @item -mtune=@var{cpu_type}
10872 Set only the instruction scheduling parameters for machine type
10873 @var{cpu_type}. The instruction set is not changed.
10875 Native Linux/GNU toolchains also support the value @samp{native},
10876 which selects the best architecture option for the host processor.
10877 @option{-mtune=native} has no effect if GCC does not recognize
10880 @item -mmemory-latency=@var{time}
10881 @opindex mmemory-latency
10882 Sets the latency the scheduler should assume for typical memory
10883 references as seen by the application. This number is highly
10884 dependent on the memory access patterns used by the application
10885 and the size of the external cache on the machine.
10887 Valid options for @var{time} are
10891 A decimal number representing clock cycles.
10897 The compiler contains estimates of the number of clock cycles for
10898 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
10899 (also called Dcache, Scache, and Bcache), as well as to main memory.
10900 Note that L3 is only valid for EV5.
10905 @node DEC Alpha/VMS Options
10906 @subsection DEC Alpha/VMS Options
10908 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
10911 @item -mvms-return-codes
10912 @opindex mvms-return-codes
10913 Return VMS condition codes from main. The default is to return POSIX
10914 style condition (e.g.@: error) codes.
10916 @item -mdebug-main=@var{prefix}
10917 @opindex mdebug-main=@var{prefix}
10918 Flag the first routine whose name starts with @var{prefix} as the main
10919 routine for the debugger.
10923 Default to 64bit memory allocation routines.
10927 @subsection FR30 Options
10928 @cindex FR30 Options
10930 These options are defined specifically for the FR30 port.
10934 @item -msmall-model
10935 @opindex msmall-model
10936 Use the small address space model. This can produce smaller code, but
10937 it does assume that all symbolic values and addresses will fit into a
10942 Assume that run-time support has been provided and so there is no need
10943 to include the simulator library (@file{libsim.a}) on the linker
10949 @subsection FRV Options
10950 @cindex FRV Options
10956 Only use the first 32 general purpose registers.
10961 Use all 64 general purpose registers.
10966 Use only the first 32 floating point registers.
10971 Use all 64 floating point registers
10974 @opindex mhard-float
10976 Use hardware instructions for floating point operations.
10979 @opindex msoft-float
10981 Use library routines for floating point operations.
10986 Dynamically allocate condition code registers.
10991 Do not try to dynamically allocate condition code registers, only
10992 use @code{icc0} and @code{fcc0}.
10997 Change ABI to use double word insns.
11002 Do not use double word instructions.
11007 Use floating point double instructions.
11010 @opindex mno-double
11012 Do not use floating point double instructions.
11017 Use media instructions.
11022 Do not use media instructions.
11027 Use multiply and add/subtract instructions.
11030 @opindex mno-muladd
11032 Do not use multiply and add/subtract instructions.
11037 Select the FDPIC ABI, that uses function descriptors to represent
11038 pointers to functions. Without any PIC/PIE-related options, it
11039 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
11040 assumes GOT entries and small data are within a 12-bit range from the
11041 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
11042 are computed with 32 bits.
11043 With a @samp{bfin-elf} target, this option implies @option{-msim}.
11046 @opindex minline-plt
11048 Enable inlining of PLT entries in function calls to functions that are
11049 not known to bind locally. It has no effect without @option{-mfdpic}.
11050 It's enabled by default if optimizing for speed and compiling for
11051 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
11052 optimization option such as @option{-O3} or above is present in the
11058 Assume a large TLS segment when generating thread-local code.
11063 Do not assume a large TLS segment when generating thread-local code.
11068 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
11069 that is known to be in read-only sections. It's enabled by default,
11070 except for @option{-fpic} or @option{-fpie}: even though it may help
11071 make the global offset table smaller, it trades 1 instruction for 4.
11072 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
11073 one of which may be shared by multiple symbols, and it avoids the need
11074 for a GOT entry for the referenced symbol, so it's more likely to be a
11075 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
11077 @item -multilib-library-pic
11078 @opindex multilib-library-pic
11080 Link with the (library, not FD) pic libraries. It's implied by
11081 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
11082 @option{-fpic} without @option{-mfdpic}. You should never have to use
11086 @opindex mlinked-fp
11088 Follow the EABI requirement of always creating a frame pointer whenever
11089 a stack frame is allocated. This option is enabled by default and can
11090 be disabled with @option{-mno-linked-fp}.
11093 @opindex mlong-calls
11095 Use indirect addressing to call functions outside the current
11096 compilation unit. This allows the functions to be placed anywhere
11097 within the 32-bit address space.
11099 @item -malign-labels
11100 @opindex malign-labels
11102 Try to align labels to an 8-byte boundary by inserting nops into the
11103 previous packet. This option only has an effect when VLIW packing
11104 is enabled. It doesn't create new packets; it merely adds nops to
11107 @item -mlibrary-pic
11108 @opindex mlibrary-pic
11110 Generate position-independent EABI code.
11115 Use only the first four media accumulator registers.
11120 Use all eight media accumulator registers.
11125 Pack VLIW instructions.
11130 Do not pack VLIW instructions.
11133 @opindex mno-eflags
11135 Do not mark ABI switches in e_flags.
11138 @opindex mcond-move
11140 Enable the use of conditional-move instructions (default).
11142 This switch is mainly for debugging the compiler and will likely be removed
11143 in a future version.
11145 @item -mno-cond-move
11146 @opindex mno-cond-move
11148 Disable the use of conditional-move instructions.
11150 This switch is mainly for debugging the compiler and will likely be removed
11151 in a future version.
11156 Enable the use of conditional set instructions (default).
11158 This switch is mainly for debugging the compiler and will likely be removed
11159 in a future version.
11164 Disable the use of conditional set instructions.
11166 This switch is mainly for debugging the compiler and will likely be removed
11167 in a future version.
11170 @opindex mcond-exec
11172 Enable the use of conditional execution (default).
11174 This switch is mainly for debugging the compiler and will likely be removed
11175 in a future version.
11177 @item -mno-cond-exec
11178 @opindex mno-cond-exec
11180 Disable the use of conditional execution.
11182 This switch is mainly for debugging the compiler and will likely be removed
11183 in a future version.
11185 @item -mvliw-branch
11186 @opindex mvliw-branch
11188 Run a pass to pack branches into VLIW instructions (default).
11190 This switch is mainly for debugging the compiler and will likely be removed
11191 in a future version.
11193 @item -mno-vliw-branch
11194 @opindex mno-vliw-branch
11196 Do not run a pass to pack branches into VLIW instructions.
11198 This switch is mainly for debugging the compiler and will likely be removed
11199 in a future version.
11201 @item -mmulti-cond-exec
11202 @opindex mmulti-cond-exec
11204 Enable optimization of @code{&&} and @code{||} in conditional execution
11207 This switch is mainly for debugging the compiler and will likely be removed
11208 in a future version.
11210 @item -mno-multi-cond-exec
11211 @opindex mno-multi-cond-exec
11213 Disable optimization of @code{&&} and @code{||} in conditional execution.
11215 This switch is mainly for debugging the compiler and will likely be removed
11216 in a future version.
11218 @item -mnested-cond-exec
11219 @opindex mnested-cond-exec
11221 Enable nested conditional execution optimizations (default).
11223 This switch is mainly for debugging the compiler and will likely be removed
11224 in a future version.
11226 @item -mno-nested-cond-exec
11227 @opindex mno-nested-cond-exec
11229 Disable nested conditional execution optimizations.
11231 This switch is mainly for debugging the compiler and will likely be removed
11232 in a future version.
11234 @item -moptimize-membar
11235 @opindex moptimize-membar
11237 This switch removes redundant @code{membar} instructions from the
11238 compiler generated code. It is enabled by default.
11240 @item -mno-optimize-membar
11241 @opindex mno-optimize-membar
11243 This switch disables the automatic removal of redundant @code{membar}
11244 instructions from the generated code.
11246 @item -mtomcat-stats
11247 @opindex mtomcat-stats
11249 Cause gas to print out tomcat statistics.
11251 @item -mcpu=@var{cpu}
11254 Select the processor type for which to generate code. Possible values are
11255 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
11256 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
11260 @node GNU/Linux Options
11261 @subsection GNU/Linux Options
11263 These @samp{-m} options are defined for GNU/Linux targets:
11268 Use the GNU C library instead of uClibc. This is the default except
11269 on @samp{*-*-linux-*uclibc*} targets.
11273 Use uClibc instead of the GNU C library. This is the default on
11274 @samp{*-*-linux-*uclibc*} targets.
11277 @node H8/300 Options
11278 @subsection H8/300 Options
11280 These @samp{-m} options are defined for the H8/300 implementations:
11285 Shorten some address references at link time, when possible; uses the
11286 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
11287 ld, Using ld}, for a fuller description.
11291 Generate code for the H8/300H@.
11295 Generate code for the H8S@.
11299 Generate code for the H8S and H8/300H in the normal mode. This switch
11300 must be used either with @option{-mh} or @option{-ms}.
11304 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
11308 Make @code{int} data 32 bits by default.
11311 @opindex malign-300
11312 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
11313 The default for the H8/300H and H8S is to align longs and floats on 4
11315 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
11316 This option has no effect on the H8/300.
11320 @subsection HPPA Options
11321 @cindex HPPA Options
11323 These @samp{-m} options are defined for the HPPA family of computers:
11326 @item -march=@var{architecture-type}
11328 Generate code for the specified architecture. The choices for
11329 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
11330 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
11331 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
11332 architecture option for your machine. Code compiled for lower numbered
11333 architectures will run on higher numbered architectures, but not the
11336 @item -mpa-risc-1-0
11337 @itemx -mpa-risc-1-1
11338 @itemx -mpa-risc-2-0
11339 @opindex mpa-risc-1-0
11340 @opindex mpa-risc-1-1
11341 @opindex mpa-risc-2-0
11342 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
11345 @opindex mbig-switch
11346 Generate code suitable for big switch tables. Use this option only if
11347 the assembler/linker complain about out of range branches within a switch
11350 @item -mjump-in-delay
11351 @opindex mjump-in-delay
11352 Fill delay slots of function calls with unconditional jump instructions
11353 by modifying the return pointer for the function call to be the target
11354 of the conditional jump.
11356 @item -mdisable-fpregs
11357 @opindex mdisable-fpregs
11358 Prevent floating point registers from being used in any manner. This is
11359 necessary for compiling kernels which perform lazy context switching of
11360 floating point registers. If you use this option and attempt to perform
11361 floating point operations, the compiler will abort.
11363 @item -mdisable-indexing
11364 @opindex mdisable-indexing
11365 Prevent the compiler from using indexing address modes. This avoids some
11366 rather obscure problems when compiling MIG generated code under MACH@.
11368 @item -mno-space-regs
11369 @opindex mno-space-regs
11370 Generate code that assumes the target has no space registers. This allows
11371 GCC to generate faster indirect calls and use unscaled index address modes.
11373 Such code is suitable for level 0 PA systems and kernels.
11375 @item -mfast-indirect-calls
11376 @opindex mfast-indirect-calls
11377 Generate code that assumes calls never cross space boundaries. This
11378 allows GCC to emit code which performs faster indirect calls.
11380 This option will not work in the presence of shared libraries or nested
11383 @item -mfixed-range=@var{register-range}
11384 @opindex mfixed-range
11385 Generate code treating the given register range as fixed registers.
11386 A fixed register is one that the register allocator can not use. This is
11387 useful when compiling kernel code. A register range is specified as
11388 two registers separated by a dash. Multiple register ranges can be
11389 specified separated by a comma.
11391 @item -mlong-load-store
11392 @opindex mlong-load-store
11393 Generate 3-instruction load and store sequences as sometimes required by
11394 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
11397 @item -mportable-runtime
11398 @opindex mportable-runtime
11399 Use the portable calling conventions proposed by HP for ELF systems.
11403 Enable the use of assembler directives only GAS understands.
11405 @item -mschedule=@var{cpu-type}
11407 Schedule code according to the constraints for the machine type
11408 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
11409 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
11410 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
11411 proper scheduling option for your machine. The default scheduling is
11415 @opindex mlinker-opt
11416 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
11417 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
11418 linkers in which they give bogus error messages when linking some programs.
11421 @opindex msoft-float
11422 Generate output containing library calls for floating point.
11423 @strong{Warning:} the requisite libraries are not available for all HPPA
11424 targets. Normally the facilities of the machine's usual C compiler are
11425 used, but this cannot be done directly in cross-compilation. You must make
11426 your own arrangements to provide suitable library functions for
11429 @option{-msoft-float} changes the calling convention in the output file;
11430 therefore, it is only useful if you compile @emph{all} of a program with
11431 this option. In particular, you need to compile @file{libgcc.a}, the
11432 library that comes with GCC, with @option{-msoft-float} in order for
11437 Generate the predefine, @code{_SIO}, for server IO@. The default is
11438 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
11439 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
11440 options are available under HP-UX and HI-UX@.
11444 Use GNU ld specific options. This passes @option{-shared} to ld when
11445 building a shared library. It is the default when GCC is configured,
11446 explicitly or implicitly, with the GNU linker. This option does not
11447 have any affect on which ld is called, it only changes what parameters
11448 are passed to that ld. The ld that is called is determined by the
11449 @option{--with-ld} configure option, GCC's program search path, and
11450 finally by the user's @env{PATH}. The linker used by GCC can be printed
11451 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
11452 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11456 Use HP ld specific options. This passes @option{-b} to ld when building
11457 a shared library and passes @option{+Accept TypeMismatch} to ld on all
11458 links. It is the default when GCC is configured, explicitly or
11459 implicitly, with the HP linker. This option does not have any affect on
11460 which ld is called, it only changes what parameters are passed to that
11461 ld. The ld that is called is determined by the @option{--with-ld}
11462 configure option, GCC's program search path, and finally by the user's
11463 @env{PATH}. The linker used by GCC can be printed using @samp{which
11464 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
11465 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11468 @opindex mno-long-calls
11469 Generate code that uses long call sequences. This ensures that a call
11470 is always able to reach linker generated stubs. The default is to generate
11471 long calls only when the distance from the call site to the beginning
11472 of the function or translation unit, as the case may be, exceeds a
11473 predefined limit set by the branch type being used. The limits for
11474 normal calls are 7,600,000 and 240,000 bytes, respectively for the
11475 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
11478 Distances are measured from the beginning of functions when using the
11479 @option{-ffunction-sections} option, or when using the @option{-mgas}
11480 and @option{-mno-portable-runtime} options together under HP-UX with
11483 It is normally not desirable to use this option as it will degrade
11484 performance. However, it may be useful in large applications,
11485 particularly when partial linking is used to build the application.
11487 The types of long calls used depends on the capabilities of the
11488 assembler and linker, and the type of code being generated. The
11489 impact on systems that support long absolute calls, and long pic
11490 symbol-difference or pc-relative calls should be relatively small.
11491 However, an indirect call is used on 32-bit ELF systems in pic code
11492 and it is quite long.
11494 @item -munix=@var{unix-std}
11496 Generate compiler predefines and select a startfile for the specified
11497 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
11498 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
11499 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
11500 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
11501 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
11504 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
11505 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
11506 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
11507 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
11508 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
11509 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
11511 It is @emph{important} to note that this option changes the interfaces
11512 for various library routines. It also affects the operational behavior
11513 of the C library. Thus, @emph{extreme} care is needed in using this
11516 Library code that is intended to operate with more than one UNIX
11517 standard must test, set and restore the variable @var{__xpg4_extended_mask}
11518 as appropriate. Most GNU software doesn't provide this capability.
11522 Suppress the generation of link options to search libdld.sl when the
11523 @option{-static} option is specified on HP-UX 10 and later.
11527 The HP-UX implementation of setlocale in libc has a dependency on
11528 libdld.sl. There isn't an archive version of libdld.sl. Thus,
11529 when the @option{-static} option is specified, special link options
11530 are needed to resolve this dependency.
11532 On HP-UX 10 and later, the GCC driver adds the necessary options to
11533 link with libdld.sl when the @option{-static} option is specified.
11534 This causes the resulting binary to be dynamic. On the 64-bit port,
11535 the linkers generate dynamic binaries by default in any case. The
11536 @option{-nolibdld} option can be used to prevent the GCC driver from
11537 adding these link options.
11541 Add support for multithreading with the @dfn{dce thread} library
11542 under HP-UX@. This option sets flags for both the preprocessor and
11546 @node i386 and x86-64 Options
11547 @subsection Intel 386 and AMD x86-64 Options
11548 @cindex i386 Options
11549 @cindex x86-64 Options
11550 @cindex Intel 386 Options
11551 @cindex AMD x86-64 Options
11553 These @samp{-m} options are defined for the i386 and x86-64 family of
11557 @item -mtune=@var{cpu-type}
11559 Tune to @var{cpu-type} everything applicable about the generated code, except
11560 for the ABI and the set of available instructions. The choices for
11561 @var{cpu-type} are:
11564 Produce code optimized for the most common IA32/AMD64/EM64T processors.
11565 If you know the CPU on which your code will run, then you should use
11566 the corresponding @option{-mtune} option instead of
11567 @option{-mtune=generic}. But, if you do not know exactly what CPU users
11568 of your application will have, then you should use this option.
11570 As new processors are deployed in the marketplace, the behavior of this
11571 option will change. Therefore, if you upgrade to a newer version of
11572 GCC, the code generated option will change to reflect the processors
11573 that were most common when that version of GCC was released.
11575 There is no @option{-march=generic} option because @option{-march}
11576 indicates the instruction set the compiler can use, and there is no
11577 generic instruction set applicable to all processors. In contrast,
11578 @option{-mtune} indicates the processor (or, in this case, collection of
11579 processors) for which the code is optimized.
11581 This selects the CPU to tune for at compilation time by determining
11582 the processor type of the compiling machine. Using @option{-mtune=native}
11583 will produce code optimized for the local machine under the constraints
11584 of the selected instruction set. Using @option{-march=native} will
11585 enable all instruction subsets supported by the local machine (hence
11586 the result might not run on different machines).
11588 Original Intel's i386 CPU@.
11590 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
11591 @item i586, pentium
11592 Intel Pentium CPU with no MMX support.
11594 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
11596 Intel PentiumPro CPU@.
11598 Same as @code{generic}, but when used as @code{march} option, PentiumPro
11599 instruction set will be used, so the code will run on all i686 family chips.
11601 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
11602 @item pentium3, pentium3m
11603 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
11606 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
11607 support. Used by Centrino notebooks.
11608 @item pentium4, pentium4m
11609 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
11611 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
11614 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
11615 SSE2 and SSE3 instruction set support.
11617 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
11618 instruction set support.
11620 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
11621 instruction set support.
11623 AMD K6 CPU with MMX instruction set support.
11625 Improved versions of AMD K6 CPU with MMX and 3dNOW!@: instruction set support.
11626 @item athlon, athlon-tbird
11627 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and SSE prefetch instructions
11629 @item athlon-4, athlon-xp, athlon-mp
11630 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and full SSE
11631 instruction set support.
11632 @item k8, opteron, athlon64, athlon-fx
11633 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
11634 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW!@: and 64-bit instruction set extensions.)
11635 @item k8-sse3, opteron-sse3, athlon64-sse3
11636 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
11637 @item amdfam10, barcelona
11638 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
11639 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3dNOW!, enhanced 3dNOW!, ABM and 64-bit
11640 instruction set extensions.)
11642 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
11645 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!@:
11646 instruction set support.
11648 Via C3 CPU with MMX and 3dNOW!@: instruction set support. (No scheduling is
11649 implemented for this chip.)
11651 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
11652 implemented for this chip.)
11654 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
11657 While picking a specific @var{cpu-type} will schedule things appropriately
11658 for that particular chip, the compiler will not generate any code that
11659 does not run on the i386 without the @option{-march=@var{cpu-type}} option
11662 @item -march=@var{cpu-type}
11664 Generate instructions for the machine type @var{cpu-type}. The choices
11665 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
11666 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
11668 @item -mcpu=@var{cpu-type}
11670 A deprecated synonym for @option{-mtune}.
11672 @item -mfpmath=@var{unit}
11674 Generate floating point arithmetics for selected unit @var{unit}. The choices
11675 for @var{unit} are:
11679 Use the standard 387 floating point coprocessor present majority of chips and
11680 emulated otherwise. Code compiled with this option will run almost everywhere.
11681 The temporary results are computed in 80bit precision instead of precision
11682 specified by the type resulting in slightly different results compared to most
11683 of other chips. See @option{-ffloat-store} for more detailed description.
11685 This is the default choice for i386 compiler.
11688 Use scalar floating point instructions present in the SSE instruction set.
11689 This instruction set is supported by Pentium3 and newer chips, in the AMD line
11690 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
11691 instruction set supports only single precision arithmetics, thus the double and
11692 extended precision arithmetics is still done using 387. Later version, present
11693 only in Pentium4 and the future AMD x86-64 chips supports double precision
11696 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
11697 or @option{-msse2} switches to enable SSE extensions and make this option
11698 effective. For the x86-64 compiler, these extensions are enabled by default.
11700 The resulting code should be considerably faster in the majority of cases and avoid
11701 the numerical instability problems of 387 code, but may break some existing
11702 code that expects temporaries to be 80bit.
11704 This is the default choice for the x86-64 compiler.
11709 Attempt to utilize both instruction sets at once. This effectively double the
11710 amount of available registers and on chips with separate execution units for
11711 387 and SSE the execution resources too. Use this option with care, as it is
11712 still experimental, because the GCC register allocator does not model separate
11713 functional units well resulting in instable performance.
11716 @item -masm=@var{dialect}
11717 @opindex masm=@var{dialect}
11718 Output asm instructions using selected @var{dialect}. Supported
11719 choices are @samp{intel} or @samp{att} (the default one). Darwin does
11720 not support @samp{intel}.
11723 @itemx -mno-ieee-fp
11725 @opindex mno-ieee-fp
11726 Control whether or not the compiler uses IEEE floating point
11727 comparisons. These handle correctly the case where the result of a
11728 comparison is unordered.
11731 @opindex msoft-float
11732 Generate output containing library calls for floating point.
11733 @strong{Warning:} the requisite libraries are not part of GCC@.
11734 Normally the facilities of the machine's usual C compiler are used, but
11735 this can't be done directly in cross-compilation. You must make your
11736 own arrangements to provide suitable library functions for
11739 On machines where a function returns floating point results in the 80387
11740 register stack, some floating point opcodes may be emitted even if
11741 @option{-msoft-float} is used.
11743 @item -mno-fp-ret-in-387
11744 @opindex mno-fp-ret-in-387
11745 Do not use the FPU registers for return values of functions.
11747 The usual calling convention has functions return values of types
11748 @code{float} and @code{double} in an FPU register, even if there
11749 is no FPU@. The idea is that the operating system should emulate
11752 The option @option{-mno-fp-ret-in-387} causes such values to be returned
11753 in ordinary CPU registers instead.
11755 @item -mno-fancy-math-387
11756 @opindex mno-fancy-math-387
11757 Some 387 emulators do not support the @code{sin}, @code{cos} and
11758 @code{sqrt} instructions for the 387. Specify this option to avoid
11759 generating those instructions. This option is the default on FreeBSD,
11760 OpenBSD and NetBSD@. This option is overridden when @option{-march}
11761 indicates that the target cpu will always have an FPU and so the
11762 instruction will not need emulation. As of revision 2.6.1, these
11763 instructions are not generated unless you also use the
11764 @option{-funsafe-math-optimizations} switch.
11766 @item -malign-double
11767 @itemx -mno-align-double
11768 @opindex malign-double
11769 @opindex mno-align-double
11770 Control whether GCC aligns @code{double}, @code{long double}, and
11771 @code{long long} variables on a two word boundary or a one word
11772 boundary. Aligning @code{double} variables on a two word boundary will
11773 produce code that runs somewhat faster on a @samp{Pentium} at the
11774 expense of more memory.
11776 On x86-64, @option{-malign-double} is enabled by default.
11778 @strong{Warning:} if you use the @option{-malign-double} switch,
11779 structures containing the above types will be aligned differently than
11780 the published application binary interface specifications for the 386
11781 and will not be binary compatible with structures in code compiled
11782 without that switch.
11784 @item -m96bit-long-double
11785 @itemx -m128bit-long-double
11786 @opindex m96bit-long-double
11787 @opindex m128bit-long-double
11788 These switches control the size of @code{long double} type. The i386
11789 application binary interface specifies the size to be 96 bits,
11790 so @option{-m96bit-long-double} is the default in 32 bit mode.
11792 Modern architectures (Pentium and newer) would prefer @code{long double}
11793 to be aligned to an 8 or 16 byte boundary. In arrays or structures
11794 conforming to the ABI, this would not be possible. So specifying a
11795 @option{-m128bit-long-double} will align @code{long double}
11796 to a 16 byte boundary by padding the @code{long double} with an additional
11799 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
11800 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
11802 Notice that neither of these options enable any extra precision over the x87
11803 standard of 80 bits for a @code{long double}.
11805 @strong{Warning:} if you override the default value for your target ABI, the
11806 structures and arrays containing @code{long double} variables will change
11807 their size as well as function calling convention for function taking
11808 @code{long double} will be modified. Hence they will not be binary
11809 compatible with arrays or structures in code compiled without that switch.
11811 @item -mlarge-data-threshold=@var{number}
11812 @opindex mlarge-data-threshold=@var{number}
11813 When @option{-mcmodel=medium} is specified, the data greater than
11814 @var{threshold} are placed in large data section. This value must be the
11815 same across all object linked into the binary and defaults to 65535.
11819 Use a different function-calling convention, in which functions that
11820 take a fixed number of arguments return with the @code{ret} @var{num}
11821 instruction, which pops their arguments while returning. This saves one
11822 instruction in the caller since there is no need to pop the arguments
11825 You can specify that an individual function is called with this calling
11826 sequence with the function attribute @samp{stdcall}. You can also
11827 override the @option{-mrtd} option by using the function attribute
11828 @samp{cdecl}. @xref{Function Attributes}.
11830 @strong{Warning:} this calling convention is incompatible with the one
11831 normally used on Unix, so you cannot use it if you need to call
11832 libraries compiled with the Unix compiler.
11834 Also, you must provide function prototypes for all functions that
11835 take variable numbers of arguments (including @code{printf});
11836 otherwise incorrect code will be generated for calls to those
11839 In addition, seriously incorrect code will result if you call a
11840 function with too many arguments. (Normally, extra arguments are
11841 harmlessly ignored.)
11843 @item -mregparm=@var{num}
11845 Control how many registers are used to pass integer arguments. By
11846 default, no registers are used to pass arguments, and at most 3
11847 registers can be used. You can control this behavior for a specific
11848 function by using the function attribute @samp{regparm}.
11849 @xref{Function Attributes}.
11851 @strong{Warning:} if you use this switch, and
11852 @var{num} is nonzero, then you must build all modules with the same
11853 value, including any libraries. This includes the system libraries and
11857 @opindex msseregparm
11858 Use SSE register passing conventions for float and double arguments
11859 and return values. You can control this behavior for a specific
11860 function by using the function attribute @samp{sseregparm}.
11861 @xref{Function Attributes}.
11863 @strong{Warning:} if you use this switch then you must build all
11864 modules with the same value, including any libraries. This includes
11865 the system libraries and startup modules.
11874 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
11875 is specified, the significands of results of floating-point operations are
11876 rounded to 24 bits (single precision); @option{-mpc64} rounds the
11877 significands of results of floating-point operations to 53 bits (double
11878 precision) and @option{-mpc80} rounds the significands of results of
11879 floating-point operations to 64 bits (extended double precision), which is
11880 the default. When this option is used, floating-point operations in higher
11881 precisions are not available to the programmer without setting the FPU
11882 control word explicitly.
11884 Setting the rounding of floating-point operations to less than the default
11885 80 bits can speed some programs by 2% or more. Note that some mathematical
11886 libraries assume that extended precision (80 bit) floating-point operations
11887 are enabled by default; routines in such libraries could suffer significant
11888 loss of accuracy, typically through so-called "catastrophic cancellation",
11889 when this option is used to set the precision to less than extended precision.
11891 @item -mstackrealign
11892 @opindex mstackrealign
11893 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
11894 option will generate an alternate prologue and epilogue that realigns the
11895 runtime stack if necessary. This supports mixing legacy codes that keep
11896 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
11897 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
11898 applicable to individual functions.
11900 @item -mpreferred-stack-boundary=@var{num}
11901 @opindex mpreferred-stack-boundary
11902 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
11903 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
11904 the default is 4 (16 bytes or 128 bits).
11906 @item -mincoming-stack-boundary=@var{num}
11907 @opindex mincoming-stack-boundary
11908 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
11909 boundary. If @option{-mincoming-stack-boundary} is not specified,
11910 the one specified by @option{-mpreferred-stack-boundary} will be used.
11912 On Pentium and PentiumPro, @code{double} and @code{long double} values
11913 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
11914 suffer significant run time performance penalties. On Pentium III, the
11915 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
11916 properly if it is not 16 byte aligned.
11918 To ensure proper alignment of this values on the stack, the stack boundary
11919 must be as aligned as that required by any value stored on the stack.
11920 Further, every function must be generated such that it keeps the stack
11921 aligned. Thus calling a function compiled with a higher preferred
11922 stack boundary from a function compiled with a lower preferred stack
11923 boundary will most likely misalign the stack. It is recommended that
11924 libraries that use callbacks always use the default setting.
11926 This extra alignment does consume extra stack space, and generally
11927 increases code size. Code that is sensitive to stack space usage, such
11928 as embedded systems and operating system kernels, may want to reduce the
11929 preferred alignment to @option{-mpreferred-stack-boundary=2}.
11969 These switches enable or disable the use of instructions in the MMX,
11970 SSE, SSE2, SSE3, SSSE3, SSE4.1, AVX, AES, PCLMUL, SSE4A, FMA4, ABM or
11971 3DNow!@: extended instruction sets.
11972 These extensions are also available as built-in functions: see
11973 @ref{X86 Built-in Functions}, for details of the functions enabled and
11974 disabled by these switches.
11976 To have SSE/SSE2 instructions generated automatically from floating-point
11977 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
11979 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
11980 generates new AVX instructions or AVX equivalence for all SSEx instructions
11983 These options will enable GCC to use these extended instructions in
11984 generated code, even without @option{-mfpmath=sse}. Applications which
11985 perform runtime CPU detection must compile separate files for each
11986 supported architecture, using the appropriate flags. In particular,
11987 the file containing the CPU detection code should be compiled without
11992 This option instructs GCC to emit a @code{cld} instruction in the prologue
11993 of functions that use string instructions. String instructions depend on
11994 the DF flag to select between autoincrement or autodecrement mode. While the
11995 ABI specifies the DF flag to be cleared on function entry, some operating
11996 systems violate this specification by not clearing the DF flag in their
11997 exception dispatchers. The exception handler can be invoked with the DF flag
11998 set which leads to wrong direction mode, when string instructions are used.
11999 This option can be enabled by default on 32-bit x86 targets by configuring
12000 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
12001 instructions can be suppressed with the @option{-mno-cld} compiler option
12006 This option will enable GCC to use CMPXCHG16B instruction in generated code.
12007 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
12008 data types. This is useful for high resolution counters that could be updated
12009 by multiple processors (or cores). This instruction is generated as part of
12010 atomic built-in functions: see @ref{Atomic Builtins} for details.
12014 This option will enable GCC to use SAHF instruction in generated 64-bit code.
12015 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
12016 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
12017 SAHF are load and store instructions, respectively, for certain status flags.
12018 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
12019 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
12023 This option will enable GCC to use movbe instruction to implement
12024 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
12028 This option will enable built-in functions, @code{__builtin_ia32_crc32qi},
12029 @code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and
12030 @code{__builtin_ia32_crc32di} to generate the crc32 machine instruction.
12034 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
12035 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
12036 to increase precision instead of DIVSS and SQRTSS (and their vectorized
12037 variants) for single precision floating point arguments. These instructions
12038 are generated only when @option{-funsafe-math-optimizations} is enabled
12039 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
12040 Note that while the throughput of the sequence is higher than the throughput
12041 of the non-reciprocal instruction, the precision of the sequence can be
12042 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
12044 @item -mveclibabi=@var{type}
12045 @opindex mveclibabi
12046 Specifies the ABI type to use for vectorizing intrinsics using an
12047 external library. Supported types are @code{svml} for the Intel short
12048 vector math library and @code{acml} for the AMD math core library style
12049 of interfacing. GCC will currently emit calls to @code{vmldExp2},
12050 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
12051 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
12052 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
12053 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
12054 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
12055 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
12056 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
12057 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
12058 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
12059 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
12060 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
12061 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
12062 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
12063 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
12064 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
12065 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
12066 compatible library will have to be specified at link time.
12068 @item -mabi=@var{name}
12070 Generate code for the specified calling convention. Permissible values
12071 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
12072 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
12073 ABI when targeting Windows. On all other systems, the default is the
12074 SYSV ABI. You can control this behavior for a specific function by
12075 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
12076 @xref{Function Attributes}.
12079 @itemx -mno-push-args
12080 @opindex mpush-args
12081 @opindex mno-push-args
12082 Use PUSH operations to store outgoing parameters. This method is shorter
12083 and usually equally fast as method using SUB/MOV operations and is enabled
12084 by default. In some cases disabling it may improve performance because of
12085 improved scheduling and reduced dependencies.
12087 @item -maccumulate-outgoing-args
12088 @opindex maccumulate-outgoing-args
12089 If enabled, the maximum amount of space required for outgoing arguments will be
12090 computed in the function prologue. This is faster on most modern CPUs
12091 because of reduced dependencies, improved scheduling and reduced stack usage
12092 when preferred stack boundary is not equal to 2. The drawback is a notable
12093 increase in code size. This switch implies @option{-mno-push-args}.
12097 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
12098 on thread-safe exception handling must compile and link all code with the
12099 @option{-mthreads} option. When compiling, @option{-mthreads} defines
12100 @option{-D_MT}; when linking, it links in a special thread helper library
12101 @option{-lmingwthrd} which cleans up per thread exception handling data.
12103 @item -mno-align-stringops
12104 @opindex mno-align-stringops
12105 Do not align destination of inlined string operations. This switch reduces
12106 code size and improves performance in case the destination is already aligned,
12107 but GCC doesn't know about it.
12109 @item -minline-all-stringops
12110 @opindex minline-all-stringops
12111 By default GCC inlines string operations only when destination is known to be
12112 aligned at least to 4 byte boundary. This enables more inlining, increase code
12113 size, but may improve performance of code that depends on fast memcpy, strlen
12114 and memset for short lengths.
12116 @item -minline-stringops-dynamically
12117 @opindex minline-stringops-dynamically
12118 For string operation of unknown size, inline runtime checks so for small
12119 blocks inline code is used, while for large blocks library call is used.
12121 @item -mstringop-strategy=@var{alg}
12122 @opindex mstringop-strategy=@var{alg}
12123 Overwrite internal decision heuristic about particular algorithm to inline
12124 string operation with. The allowed values are @code{rep_byte},
12125 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
12126 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
12127 expanding inline loop, @code{libcall} for always expanding library call.
12129 @item -momit-leaf-frame-pointer
12130 @opindex momit-leaf-frame-pointer
12131 Don't keep the frame pointer in a register for leaf functions. This
12132 avoids the instructions to save, set up and restore frame pointers and
12133 makes an extra register available in leaf functions. The option
12134 @option{-fomit-frame-pointer} removes the frame pointer for all functions
12135 which might make debugging harder.
12137 @item -mtls-direct-seg-refs
12138 @itemx -mno-tls-direct-seg-refs
12139 @opindex mtls-direct-seg-refs
12140 Controls whether TLS variables may be accessed with offsets from the
12141 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
12142 or whether the thread base pointer must be added. Whether or not this
12143 is legal depends on the operating system, and whether it maps the
12144 segment to cover the entire TLS area.
12146 For systems that use GNU libc, the default is on.
12149 @itemx -mno-sse2avx
12151 Specify that the assembler should encode SSE instructions with VEX
12152 prefix. The option @option{-mavx} turns this on by default.
12155 These @samp{-m} switches are supported in addition to the above
12156 on AMD x86-64 processors in 64-bit environments.
12163 Generate code for a 32-bit or 64-bit environment.
12164 The 32-bit environment sets int, long and pointer to 32 bits and
12165 generates code that runs on any i386 system.
12166 The 64-bit environment sets int to 32 bits and long and pointer
12167 to 64 bits and generates code for AMD's x86-64 architecture. For
12168 darwin only the -m64 option turns off the @option{-fno-pic} and
12169 @option{-mdynamic-no-pic} options.
12171 @item -mno-red-zone
12172 @opindex mno-red-zone
12173 Do not use a so called red zone for x86-64 code. The red zone is mandated
12174 by the x86-64 ABI, it is a 128-byte area beyond the location of the
12175 stack pointer that will not be modified by signal or interrupt handlers
12176 and therefore can be used for temporary data without adjusting the stack
12177 pointer. The flag @option{-mno-red-zone} disables this red zone.
12179 @item -mcmodel=small
12180 @opindex mcmodel=small
12181 Generate code for the small code model: the program and its symbols must
12182 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
12183 Programs can be statically or dynamically linked. This is the default
12186 @item -mcmodel=kernel
12187 @opindex mcmodel=kernel
12188 Generate code for the kernel code model. The kernel runs in the
12189 negative 2 GB of the address space.
12190 This model has to be used for Linux kernel code.
12192 @item -mcmodel=medium
12193 @opindex mcmodel=medium
12194 Generate code for the medium model: The program is linked in the lower 2
12195 GB of the address space. Small symbols are also placed there. Symbols
12196 with sizes larger than @option{-mlarge-data-threshold} are put into
12197 large data or bss sections and can be located above 2GB. Programs can
12198 be statically or dynamically linked.
12200 @item -mcmodel=large
12201 @opindex mcmodel=large
12202 Generate code for the large model: This model makes no assumptions
12203 about addresses and sizes of sections.
12206 @node IA-64 Options
12207 @subsection IA-64 Options
12208 @cindex IA-64 Options
12210 These are the @samp{-m} options defined for the Intel IA-64 architecture.
12214 @opindex mbig-endian
12215 Generate code for a big endian target. This is the default for HP-UX@.
12217 @item -mlittle-endian
12218 @opindex mlittle-endian
12219 Generate code for a little endian target. This is the default for AIX5
12225 @opindex mno-gnu-as
12226 Generate (or don't) code for the GNU assembler. This is the default.
12227 @c Also, this is the default if the configure option @option{--with-gnu-as}
12233 @opindex mno-gnu-ld
12234 Generate (or don't) code for the GNU linker. This is the default.
12235 @c Also, this is the default if the configure option @option{--with-gnu-ld}
12240 Generate code that does not use a global pointer register. The result
12241 is not position independent code, and violates the IA-64 ABI@.
12243 @item -mvolatile-asm-stop
12244 @itemx -mno-volatile-asm-stop
12245 @opindex mvolatile-asm-stop
12246 @opindex mno-volatile-asm-stop
12247 Generate (or don't) a stop bit immediately before and after volatile asm
12250 @item -mregister-names
12251 @itemx -mno-register-names
12252 @opindex mregister-names
12253 @opindex mno-register-names
12254 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
12255 the stacked registers. This may make assembler output more readable.
12261 Disable (or enable) optimizations that use the small data section. This may
12262 be useful for working around optimizer bugs.
12264 @item -mconstant-gp
12265 @opindex mconstant-gp
12266 Generate code that uses a single constant global pointer value. This is
12267 useful when compiling kernel code.
12271 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
12272 This is useful when compiling firmware code.
12274 @item -minline-float-divide-min-latency
12275 @opindex minline-float-divide-min-latency
12276 Generate code for inline divides of floating point values
12277 using the minimum latency algorithm.
12279 @item -minline-float-divide-max-throughput
12280 @opindex minline-float-divide-max-throughput
12281 Generate code for inline divides of floating point values
12282 using the maximum throughput algorithm.
12284 @item -mno-inline-float-divide
12285 @opindex mno-inline-float-divide
12286 Do not generate inline code for divides of floating point values.
12288 @item -minline-int-divide-min-latency
12289 @opindex minline-int-divide-min-latency
12290 Generate code for inline divides of integer values
12291 using the minimum latency algorithm.
12293 @item -minline-int-divide-max-throughput
12294 @opindex minline-int-divide-max-throughput
12295 Generate code for inline divides of integer values
12296 using the maximum throughput algorithm.
12298 @item -mno-inline-int-divide
12299 @opindex mno-inline-int-divide
12300 Do not generate inline code for divides of integer values.
12302 @item -minline-sqrt-min-latency
12303 @opindex minline-sqrt-min-latency
12304 Generate code for inline square roots
12305 using the minimum latency algorithm.
12307 @item -minline-sqrt-max-throughput
12308 @opindex minline-sqrt-max-throughput
12309 Generate code for inline square roots
12310 using the maximum throughput algorithm.
12312 @item -mno-inline-sqrt
12313 @opindex mno-inline-sqrt
12314 Do not generate inline code for sqrt.
12317 @itemx -mno-fused-madd
12318 @opindex mfused-madd
12319 @opindex mno-fused-madd
12320 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
12321 instructions. The default is to use these instructions.
12323 @item -mno-dwarf2-asm
12324 @itemx -mdwarf2-asm
12325 @opindex mno-dwarf2-asm
12326 @opindex mdwarf2-asm
12327 Don't (or do) generate assembler code for the DWARF2 line number debugging
12328 info. This may be useful when not using the GNU assembler.
12330 @item -mearly-stop-bits
12331 @itemx -mno-early-stop-bits
12332 @opindex mearly-stop-bits
12333 @opindex mno-early-stop-bits
12334 Allow stop bits to be placed earlier than immediately preceding the
12335 instruction that triggered the stop bit. This can improve instruction
12336 scheduling, but does not always do so.
12338 @item -mfixed-range=@var{register-range}
12339 @opindex mfixed-range
12340 Generate code treating the given register range as fixed registers.
12341 A fixed register is one that the register allocator can not use. This is
12342 useful when compiling kernel code. A register range is specified as
12343 two registers separated by a dash. Multiple register ranges can be
12344 specified separated by a comma.
12346 @item -mtls-size=@var{tls-size}
12348 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
12351 @item -mtune=@var{cpu-type}
12353 Tune the instruction scheduling for a particular CPU, Valid values are
12354 itanium, itanium1, merced, itanium2, and mckinley.
12360 Generate code for a 32-bit or 64-bit environment.
12361 The 32-bit environment sets int, long and pointer to 32 bits.
12362 The 64-bit environment sets int to 32 bits and long and pointer
12363 to 64 bits. These are HP-UX specific flags.
12365 @item -mno-sched-br-data-spec
12366 @itemx -msched-br-data-spec
12367 @opindex mno-sched-br-data-spec
12368 @opindex msched-br-data-spec
12369 (Dis/En)able data speculative scheduling before reload.
12370 This will result in generation of the ld.a instructions and
12371 the corresponding check instructions (ld.c / chk.a).
12372 The default is 'disable'.
12374 @item -msched-ar-data-spec
12375 @itemx -mno-sched-ar-data-spec
12376 @opindex msched-ar-data-spec
12377 @opindex mno-sched-ar-data-spec
12378 (En/Dis)able data speculative scheduling after reload.
12379 This will result in generation of the ld.a instructions and
12380 the corresponding check instructions (ld.c / chk.a).
12381 The default is 'enable'.
12383 @item -mno-sched-control-spec
12384 @itemx -msched-control-spec
12385 @opindex mno-sched-control-spec
12386 @opindex msched-control-spec
12387 (Dis/En)able control speculative scheduling. This feature is
12388 available only during region scheduling (i.e.@: before reload).
12389 This will result in generation of the ld.s instructions and
12390 the corresponding check instructions chk.s .
12391 The default is 'disable'.
12393 @item -msched-br-in-data-spec
12394 @itemx -mno-sched-br-in-data-spec
12395 @opindex msched-br-in-data-spec
12396 @opindex mno-sched-br-in-data-spec
12397 (En/Dis)able speculative scheduling of the instructions that
12398 are dependent on the data speculative loads before reload.
12399 This is effective only with @option{-msched-br-data-spec} enabled.
12400 The default is 'enable'.
12402 @item -msched-ar-in-data-spec
12403 @itemx -mno-sched-ar-in-data-spec
12404 @opindex msched-ar-in-data-spec
12405 @opindex mno-sched-ar-in-data-spec
12406 (En/Dis)able speculative scheduling of the instructions that
12407 are dependent on the data speculative loads after reload.
12408 This is effective only with @option{-msched-ar-data-spec} enabled.
12409 The default is 'enable'.
12411 @item -msched-in-control-spec
12412 @itemx -mno-sched-in-control-spec
12413 @opindex msched-in-control-spec
12414 @opindex mno-sched-in-control-spec
12415 (En/Dis)able speculative scheduling of the instructions that
12416 are dependent on the control speculative loads.
12417 This is effective only with @option{-msched-control-spec} enabled.
12418 The default is 'enable'.
12420 @item -mno-sched-prefer-non-data-spec-insns
12421 @itemx -msched-prefer-non-data-spec-insns
12422 @opindex mno-sched-prefer-non-data-spec-insns
12423 @opindex msched-prefer-non-data-spec-insns
12424 If enabled, data speculative instructions will be chosen for schedule
12425 only if there are no other choices at the moment. This will make
12426 the use of the data speculation much more conservative.
12427 The default is 'disable'.
12429 @item -mno-sched-prefer-non-control-spec-insns
12430 @itemx -msched-prefer-non-control-spec-insns
12431 @opindex mno-sched-prefer-non-control-spec-insns
12432 @opindex msched-prefer-non-control-spec-insns
12433 If enabled, control speculative instructions will be chosen for schedule
12434 only if there are no other choices at the moment. This will make
12435 the use of the control speculation much more conservative.
12436 The default is 'disable'.
12438 @item -mno-sched-count-spec-in-critical-path
12439 @itemx -msched-count-spec-in-critical-path
12440 @opindex mno-sched-count-spec-in-critical-path
12441 @opindex msched-count-spec-in-critical-path
12442 If enabled, speculative dependencies will be considered during
12443 computation of the instructions priorities. This will make the use of the
12444 speculation a bit more conservative.
12445 The default is 'disable'.
12447 @item -msched-spec-ldc
12448 @opindex msched-spec-ldc
12449 Use a simple data speculation check. This option is on by default.
12451 @item -msched-control-spec-ldc
12452 @opindex msched-spec-ldc
12453 Use a simple check for control speculation. This option is on by default.
12455 @item -msched-stop-bits-after-every-cycle
12456 @opindex msched-stop-bits-after-every-cycle
12457 Place a stop bit after every cycle when scheduling. This option is on
12460 @item -msched-fp-mem-deps-zero-cost
12461 @opindex msched-fp-mem-deps-zero-cost
12462 Assume that floating-point stores and loads are not likely to cause a conflict
12463 when placed into the same instruction group. This option is disabled by
12466 @item -msel-sched-dont-check-control-spec
12467 @opindex msel-sched-dont-check-control-spec
12468 Generate checks for control speculation in selective scheduling.
12469 This flag is disabled by default.
12471 @item -msched-max-memory-insns=@var{max-insns}
12472 @opindex msched-max-memory-insns
12473 Limit on the number of memory insns per instruction group, giving lower
12474 priority to subsequent memory insns attempting to schedule in the same
12475 instruction group. Frequently useful to prevent cache bank conflicts.
12476 The default value is 1.
12478 @item -msched-max-memory-insns-hard-limit
12479 @opindex msched-max-memory-insns-hard-limit
12480 Disallow more than `msched-max-memory-insns' in instruction group.
12481 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
12482 when limit is reached but may still schedule memory operations.
12486 @node IA-64/VMS Options
12487 @subsection IA-64/VMS Options
12489 These @samp{-m} options are defined for the IA-64/VMS implementations:
12492 @item -mvms-return-codes
12493 @opindex mvms-return-codes
12494 Return VMS condition codes from main. The default is to return POSIX
12495 style condition (e.g.@ error) codes.
12497 @item -mdebug-main=@var{prefix}
12498 @opindex mdebug-main=@var{prefix}
12499 Flag the first routine whose name starts with @var{prefix} as the main
12500 routine for the debugger.
12504 Default to 64bit memory allocation routines.
12508 @subsection M32C Options
12509 @cindex M32C options
12512 @item -mcpu=@var{name}
12514 Select the CPU for which code is generated. @var{name} may be one of
12515 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
12516 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
12517 the M32C/80 series.
12521 Specifies that the program will be run on the simulator. This causes
12522 an alternate runtime library to be linked in which supports, for
12523 example, file I/O@. You must not use this option when generating
12524 programs that will run on real hardware; you must provide your own
12525 runtime library for whatever I/O functions are needed.
12527 @item -memregs=@var{number}
12529 Specifies the number of memory-based pseudo-registers GCC will use
12530 during code generation. These pseudo-registers will be used like real
12531 registers, so there is a tradeoff between GCC's ability to fit the
12532 code into available registers, and the performance penalty of using
12533 memory instead of registers. Note that all modules in a program must
12534 be compiled with the same value for this option. Because of that, you
12535 must not use this option with the default runtime libraries gcc
12540 @node M32R/D Options
12541 @subsection M32R/D Options
12542 @cindex M32R/D options
12544 These @option{-m} options are defined for Renesas M32R/D architectures:
12549 Generate code for the M32R/2@.
12553 Generate code for the M32R/X@.
12557 Generate code for the M32R@. This is the default.
12559 @item -mmodel=small
12560 @opindex mmodel=small
12561 Assume all objects live in the lower 16MB of memory (so that their addresses
12562 can be loaded with the @code{ld24} instruction), and assume all subroutines
12563 are reachable with the @code{bl} instruction.
12564 This is the default.
12566 The addressability of a particular object can be set with the
12567 @code{model} attribute.
12569 @item -mmodel=medium
12570 @opindex mmodel=medium
12571 Assume objects may be anywhere in the 32-bit address space (the compiler
12572 will generate @code{seth/add3} instructions to load their addresses), and
12573 assume all subroutines are reachable with the @code{bl} instruction.
12575 @item -mmodel=large
12576 @opindex mmodel=large
12577 Assume objects may be anywhere in the 32-bit address space (the compiler
12578 will generate @code{seth/add3} instructions to load their addresses), and
12579 assume subroutines may not be reachable with the @code{bl} instruction
12580 (the compiler will generate the much slower @code{seth/add3/jl}
12581 instruction sequence).
12584 @opindex msdata=none
12585 Disable use of the small data area. Variables will be put into
12586 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
12587 @code{section} attribute has been specified).
12588 This is the default.
12590 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
12591 Objects may be explicitly put in the small data area with the
12592 @code{section} attribute using one of these sections.
12594 @item -msdata=sdata
12595 @opindex msdata=sdata
12596 Put small global and static data in the small data area, but do not
12597 generate special code to reference them.
12600 @opindex msdata=use
12601 Put small global and static data in the small data area, and generate
12602 special instructions to reference them.
12606 @cindex smaller data references
12607 Put global and static objects less than or equal to @var{num} bytes
12608 into the small data or bss sections instead of the normal data or bss
12609 sections. The default value of @var{num} is 8.
12610 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
12611 for this option to have any effect.
12613 All modules should be compiled with the same @option{-G @var{num}} value.
12614 Compiling with different values of @var{num} may or may not work; if it
12615 doesn't the linker will give an error message---incorrect code will not be
12620 Makes the M32R specific code in the compiler display some statistics
12621 that might help in debugging programs.
12623 @item -malign-loops
12624 @opindex malign-loops
12625 Align all loops to a 32-byte boundary.
12627 @item -mno-align-loops
12628 @opindex mno-align-loops
12629 Do not enforce a 32-byte alignment for loops. This is the default.
12631 @item -missue-rate=@var{number}
12632 @opindex missue-rate=@var{number}
12633 Issue @var{number} instructions per cycle. @var{number} can only be 1
12636 @item -mbranch-cost=@var{number}
12637 @opindex mbranch-cost=@var{number}
12638 @var{number} can only be 1 or 2. If it is 1 then branches will be
12639 preferred over conditional code, if it is 2, then the opposite will
12642 @item -mflush-trap=@var{number}
12643 @opindex mflush-trap=@var{number}
12644 Specifies the trap number to use to flush the cache. The default is
12645 12. Valid numbers are between 0 and 15 inclusive.
12647 @item -mno-flush-trap
12648 @opindex mno-flush-trap
12649 Specifies that the cache cannot be flushed by using a trap.
12651 @item -mflush-func=@var{name}
12652 @opindex mflush-func=@var{name}
12653 Specifies the name of the operating system function to call to flush
12654 the cache. The default is @emph{_flush_cache}, but a function call
12655 will only be used if a trap is not available.
12657 @item -mno-flush-func
12658 @opindex mno-flush-func
12659 Indicates that there is no OS function for flushing the cache.
12663 @node M680x0 Options
12664 @subsection M680x0 Options
12665 @cindex M680x0 options
12667 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
12668 The default settings depend on which architecture was selected when
12669 the compiler was configured; the defaults for the most common choices
12673 @item -march=@var{arch}
12675 Generate code for a specific M680x0 or ColdFire instruction set
12676 architecture. Permissible values of @var{arch} for M680x0
12677 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
12678 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
12679 architectures are selected according to Freescale's ISA classification
12680 and the permissible values are: @samp{isaa}, @samp{isaaplus},
12681 @samp{isab} and @samp{isac}.
12683 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
12684 code for a ColdFire target. The @var{arch} in this macro is one of the
12685 @option{-march} arguments given above.
12687 When used together, @option{-march} and @option{-mtune} select code
12688 that runs on a family of similar processors but that is optimized
12689 for a particular microarchitecture.
12691 @item -mcpu=@var{cpu}
12693 Generate code for a specific M680x0 or ColdFire processor.
12694 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
12695 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
12696 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
12697 below, which also classifies the CPUs into families:
12699 @multitable @columnfractions 0.20 0.80
12700 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
12701 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
12702 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
12703 @item @samp{5206e} @tab @samp{5206e}
12704 @item @samp{5208} @tab @samp{5207} @samp{5208}
12705 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
12706 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
12707 @item @samp{5216} @tab @samp{5214} @samp{5216}
12708 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
12709 @item @samp{5225} @tab @samp{5224} @samp{5225}
12710 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
12711 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
12712 @item @samp{5249} @tab @samp{5249}
12713 @item @samp{5250} @tab @samp{5250}
12714 @item @samp{5271} @tab @samp{5270} @samp{5271}
12715 @item @samp{5272} @tab @samp{5272}
12716 @item @samp{5275} @tab @samp{5274} @samp{5275}
12717 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
12718 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
12719 @item @samp{5307} @tab @samp{5307}
12720 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
12721 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
12722 @item @samp{5407} @tab @samp{5407}
12723 @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}
12726 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
12727 @var{arch} is compatible with @var{cpu}. Other combinations of
12728 @option{-mcpu} and @option{-march} are rejected.
12730 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
12731 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
12732 where the value of @var{family} is given by the table above.
12734 @item -mtune=@var{tune}
12736 Tune the code for a particular microarchitecture, within the
12737 constraints set by @option{-march} and @option{-mcpu}.
12738 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
12739 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
12740 and @samp{cpu32}. The ColdFire microarchitectures
12741 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
12743 You can also use @option{-mtune=68020-40} for code that needs
12744 to run relatively well on 68020, 68030 and 68040 targets.
12745 @option{-mtune=68020-60} is similar but includes 68060 targets
12746 as well. These two options select the same tuning decisions as
12747 @option{-m68020-40} and @option{-m68020-60} respectively.
12749 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
12750 when tuning for 680x0 architecture @var{arch}. It also defines
12751 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
12752 option is used. If gcc is tuning for a range of architectures,
12753 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
12754 it defines the macros for every architecture in the range.
12756 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
12757 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
12758 of the arguments given above.
12764 Generate output for a 68000. This is the default
12765 when the compiler is configured for 68000-based systems.
12766 It is equivalent to @option{-march=68000}.
12768 Use this option for microcontrollers with a 68000 or EC000 core,
12769 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
12773 Generate output for a 68010. This is the default
12774 when the compiler is configured for 68010-based systems.
12775 It is equivalent to @option{-march=68010}.
12781 Generate output for a 68020. This is the default
12782 when the compiler is configured for 68020-based systems.
12783 It is equivalent to @option{-march=68020}.
12787 Generate output for a 68030. This is the default when the compiler is
12788 configured for 68030-based systems. It is equivalent to
12789 @option{-march=68030}.
12793 Generate output for a 68040. This is the default when the compiler is
12794 configured for 68040-based systems. It is equivalent to
12795 @option{-march=68040}.
12797 This option inhibits the use of 68881/68882 instructions that have to be
12798 emulated by software on the 68040. Use this option if your 68040 does not
12799 have code to emulate those instructions.
12803 Generate output for a 68060. This is the default when the compiler is
12804 configured for 68060-based systems. It is equivalent to
12805 @option{-march=68060}.
12807 This option inhibits the use of 68020 and 68881/68882 instructions that
12808 have to be emulated by software on the 68060. Use this option if your 68060
12809 does not have code to emulate those instructions.
12813 Generate output for a CPU32. This is the default
12814 when the compiler is configured for CPU32-based systems.
12815 It is equivalent to @option{-march=cpu32}.
12817 Use this option for microcontrollers with a
12818 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
12819 68336, 68340, 68341, 68349 and 68360.
12823 Generate output for a 520X ColdFire CPU@. This is the default
12824 when the compiler is configured for 520X-based systems.
12825 It is equivalent to @option{-mcpu=5206}, and is now deprecated
12826 in favor of that option.
12828 Use this option for microcontroller with a 5200 core, including
12829 the MCF5202, MCF5203, MCF5204 and MCF5206.
12833 Generate output for a 5206e ColdFire CPU@. The option is now
12834 deprecated in favor of the equivalent @option{-mcpu=5206e}.
12838 Generate output for a member of the ColdFire 528X family.
12839 The option is now deprecated in favor of the equivalent
12840 @option{-mcpu=528x}.
12844 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
12845 in favor of the equivalent @option{-mcpu=5307}.
12849 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
12850 in favor of the equivalent @option{-mcpu=5407}.
12854 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
12855 This includes use of hardware floating point instructions.
12856 The option is equivalent to @option{-mcpu=547x}, and is now
12857 deprecated in favor of that option.
12861 Generate output for a 68040, without using any of the new instructions.
12862 This results in code which can run relatively efficiently on either a
12863 68020/68881 or a 68030 or a 68040. The generated code does use the
12864 68881 instructions that are emulated on the 68040.
12866 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
12870 Generate output for a 68060, without using any of the new instructions.
12871 This results in code which can run relatively efficiently on either a
12872 68020/68881 or a 68030 or a 68040. The generated code does use the
12873 68881 instructions that are emulated on the 68060.
12875 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
12879 @opindex mhard-float
12881 Generate floating-point instructions. This is the default for 68020
12882 and above, and for ColdFire devices that have an FPU@. It defines the
12883 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
12884 on ColdFire targets.
12887 @opindex msoft-float
12888 Do not generate floating-point instructions; use library calls instead.
12889 This is the default for 68000, 68010, and 68832 targets. It is also
12890 the default for ColdFire devices that have no FPU.
12896 Generate (do not generate) ColdFire hardware divide and remainder
12897 instructions. If @option{-march} is used without @option{-mcpu},
12898 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
12899 architectures. Otherwise, the default is taken from the target CPU
12900 (either the default CPU, or the one specified by @option{-mcpu}). For
12901 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
12902 @option{-mcpu=5206e}.
12904 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
12908 Consider type @code{int} to be 16 bits wide, like @code{short int}.
12909 Additionally, parameters passed on the stack are also aligned to a
12910 16-bit boundary even on targets whose API mandates promotion to 32-bit.
12914 Do not consider type @code{int} to be 16 bits wide. This is the default.
12917 @itemx -mno-bitfield
12918 @opindex mnobitfield
12919 @opindex mno-bitfield
12920 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
12921 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
12925 Do use the bit-field instructions. The @option{-m68020} option implies
12926 @option{-mbitfield}. This is the default if you use a configuration
12927 designed for a 68020.
12931 Use a different function-calling convention, in which functions
12932 that take a fixed number of arguments return with the @code{rtd}
12933 instruction, which pops their arguments while returning. This
12934 saves one instruction in the caller since there is no need to pop
12935 the arguments there.
12937 This calling convention is incompatible with the one normally
12938 used on Unix, so you cannot use it if you need to call libraries
12939 compiled with the Unix compiler.
12941 Also, you must provide function prototypes for all functions that
12942 take variable numbers of arguments (including @code{printf});
12943 otherwise incorrect code will be generated for calls to those
12946 In addition, seriously incorrect code will result if you call a
12947 function with too many arguments. (Normally, extra arguments are
12948 harmlessly ignored.)
12950 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
12951 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
12955 Do not use the calling conventions selected by @option{-mrtd}.
12956 This is the default.
12959 @itemx -mno-align-int
12960 @opindex malign-int
12961 @opindex mno-align-int
12962 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
12963 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
12964 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
12965 Aligning variables on 32-bit boundaries produces code that runs somewhat
12966 faster on processors with 32-bit busses at the expense of more memory.
12968 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
12969 align structures containing the above types differently than
12970 most published application binary interface specifications for the m68k.
12974 Use the pc-relative addressing mode of the 68000 directly, instead of
12975 using a global offset table. At present, this option implies @option{-fpic},
12976 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
12977 not presently supported with @option{-mpcrel}, though this could be supported for
12978 68020 and higher processors.
12980 @item -mno-strict-align
12981 @itemx -mstrict-align
12982 @opindex mno-strict-align
12983 @opindex mstrict-align
12984 Do not (do) assume that unaligned memory references will be handled by
12988 Generate code that allows the data segment to be located in a different
12989 area of memory from the text segment. This allows for execute in place in
12990 an environment without virtual memory management. This option implies
12993 @item -mno-sep-data
12994 Generate code that assumes that the data segment follows the text segment.
12995 This is the default.
12997 @item -mid-shared-library
12998 Generate code that supports shared libraries via the library ID method.
12999 This allows for execute in place and shared libraries in an environment
13000 without virtual memory management. This option implies @option{-fPIC}.
13002 @item -mno-id-shared-library
13003 Generate code that doesn't assume ID based shared libraries are being used.
13004 This is the default.
13006 @item -mshared-library-id=n
13007 Specified the identification number of the ID based shared library being
13008 compiled. Specifying a value of 0 will generate more compact code, specifying
13009 other values will force the allocation of that number to the current
13010 library but is no more space or time efficient than omitting this option.
13016 When generating position-independent code for ColdFire, generate code
13017 that works if the GOT has more than 8192 entries. This code is
13018 larger and slower than code generated without this option. On M680x0
13019 processors, this option is not needed; @option{-fPIC} suffices.
13021 GCC normally uses a single instruction to load values from the GOT@.
13022 While this is relatively efficient, it only works if the GOT
13023 is smaller than about 64k. Anything larger causes the linker
13024 to report an error such as:
13026 @cindex relocation truncated to fit (ColdFire)
13028 relocation truncated to fit: R_68K_GOT16O foobar
13031 If this happens, you should recompile your code with @option{-mxgot}.
13032 It should then work with very large GOTs. However, code generated with
13033 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
13034 the value of a global symbol.
13036 Note that some linkers, including newer versions of the GNU linker,
13037 can create multiple GOTs and sort GOT entries. If you have such a linker,
13038 you should only need to use @option{-mxgot} when compiling a single
13039 object file that accesses more than 8192 GOT entries. Very few do.
13041 These options have no effect unless GCC is generating
13042 position-independent code.
13046 @node M68hc1x Options
13047 @subsection M68hc1x Options
13048 @cindex M68hc1x options
13050 These are the @samp{-m} options defined for the 68hc11 and 68hc12
13051 microcontrollers. The default values for these options depends on
13052 which style of microcontroller was selected when the compiler was configured;
13053 the defaults for the most common choices are given below.
13060 Generate output for a 68HC11. This is the default
13061 when the compiler is configured for 68HC11-based systems.
13067 Generate output for a 68HC12. This is the default
13068 when the compiler is configured for 68HC12-based systems.
13074 Generate output for a 68HCS12.
13076 @item -mauto-incdec
13077 @opindex mauto-incdec
13078 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
13085 Enable the use of 68HC12 min and max instructions.
13088 @itemx -mno-long-calls
13089 @opindex mlong-calls
13090 @opindex mno-long-calls
13091 Treat all calls as being far away (near). If calls are assumed to be
13092 far away, the compiler will use the @code{call} instruction to
13093 call a function and the @code{rtc} instruction for returning.
13097 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13099 @item -msoft-reg-count=@var{count}
13100 @opindex msoft-reg-count
13101 Specify the number of pseudo-soft registers which are used for the
13102 code generation. The maximum number is 32. Using more pseudo-soft
13103 register may or may not result in better code depending on the program.
13104 The default is 4 for 68HC11 and 2 for 68HC12.
13108 @node MCore Options
13109 @subsection MCore Options
13110 @cindex MCore options
13112 These are the @samp{-m} options defined for the Motorola M*Core
13118 @itemx -mno-hardlit
13120 @opindex mno-hardlit
13121 Inline constants into the code stream if it can be done in two
13122 instructions or less.
13128 Use the divide instruction. (Enabled by default).
13130 @item -mrelax-immediate
13131 @itemx -mno-relax-immediate
13132 @opindex mrelax-immediate
13133 @opindex mno-relax-immediate
13134 Allow arbitrary sized immediates in bit operations.
13136 @item -mwide-bitfields
13137 @itemx -mno-wide-bitfields
13138 @opindex mwide-bitfields
13139 @opindex mno-wide-bitfields
13140 Always treat bit-fields as int-sized.
13142 @item -m4byte-functions
13143 @itemx -mno-4byte-functions
13144 @opindex m4byte-functions
13145 @opindex mno-4byte-functions
13146 Force all functions to be aligned to a four byte boundary.
13148 @item -mcallgraph-data
13149 @itemx -mno-callgraph-data
13150 @opindex mcallgraph-data
13151 @opindex mno-callgraph-data
13152 Emit callgraph information.
13155 @itemx -mno-slow-bytes
13156 @opindex mslow-bytes
13157 @opindex mno-slow-bytes
13158 Prefer word access when reading byte quantities.
13160 @item -mlittle-endian
13161 @itemx -mbig-endian
13162 @opindex mlittle-endian
13163 @opindex mbig-endian
13164 Generate code for a little endian target.
13170 Generate code for the 210 processor.
13174 Assume that run-time support has been provided and so omit the
13175 simulator library (@file{libsim.a)} from the linker command line.
13177 @item -mstack-increment=@var{size}
13178 @opindex mstack-increment
13179 Set the maximum amount for a single stack increment operation. Large
13180 values can increase the speed of programs which contain functions
13181 that need a large amount of stack space, but they can also trigger a
13182 segmentation fault if the stack is extended too much. The default
13188 @subsection MeP Options
13189 @cindex MeP options
13195 Enables the @code{abs} instruction, which is the absolute difference
13196 between two registers.
13200 Enables all the optional instructions - average, multiply, divide, bit
13201 operations, leading zero, absolute difference, min/max, clip, and
13207 Enables the @code{ave} instruction, which computes the average of two
13210 @item -mbased=@var{n}
13212 Variables of size @var{n} bytes or smaller will be placed in the
13213 @code{.based} section by default. Based variables use the @code{$tp}
13214 register as a base register, and there is a 128 byte limit to the
13215 @code{.based} section.
13219 Enables the bit operation instructions - bit test (@code{btstm}), set
13220 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
13221 test-and-set (@code{tas}).
13223 @item -mc=@var{name}
13225 Selects which section constant data will be placed in. @var{name} may
13226 be @code{tiny}, @code{near}, or @code{far}.
13230 Enables the @code{clip} instruction. Note that @code{-mclip} is not
13231 useful unless you also provide @code{-mminmax}.
13233 @item -mconfig=@var{name}
13235 Selects one of the build-in core configurations. Each MeP chip has
13236 one or more modules in it; each module has a core CPU and a variety of
13237 coprocessors, optional instructions, and peripherals. The
13238 @code{MeP-Integrator} tool, not part of GCC, provides these
13239 configurations through this option; using this option is the same as
13240 using all the corresponding command line options. The default
13241 configuration is @code{default}.
13245 Enables the coprocessor instructions. By default, this is a 32-bit
13246 coprocessor. Note that the coprocessor is normally enabled via the
13247 @code{-mconfig=} option.
13251 Enables the 32-bit coprocessor's instructions.
13255 Enables the 64-bit coprocessor's instructions.
13259 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
13263 Causes constant variables to be placed in the @code{.near} section.
13267 Enables the @code{div} and @code{divu} instructions.
13271 Generate big-endian code.
13275 Generate little-endian code.
13277 @item -mio-volatile
13278 @opindex mio-volatile
13279 Tells the compiler that any variable marked with the @code{io}
13280 attribute is to be considered volatile.
13284 Causes variables to be assigned to the @code{.far} section by default.
13288 Enables the @code{leadz} (leading zero) instruction.
13292 Causes variables to be assigned to the @code{.near} section by default.
13296 Enables the @code{min} and @code{max} instructions.
13300 Enables the multiplication and multiply-accumulate instructions.
13304 Disables all the optional instructions enabled by @code{-mall-opts}.
13308 Enables the @code{repeat} and @code{erepeat} instructions, used for
13309 low-overhead looping.
13313 Causes all variables to default to the @code{.tiny} section. Note
13314 that there is a 65536 byte limit to this section. Accesses to these
13315 variables use the @code{%gp} base register.
13319 Enables the saturation instructions. Note that the compiler does not
13320 currently generate these itself, but this option is included for
13321 compatibility with other tools, like @code{as}.
13325 Link the SDRAM-based runtime instead of the default ROM-based runtime.
13329 Link the simulator runtime libraries.
13333 Link the simulator runtime libraries, excluding built-in support
13334 for reset and exception vectors and tables.
13338 Causes all functions to default to the @code{.far} section. Without
13339 this option, functions default to the @code{.near} section.
13341 @item -mtiny=@var{n}
13343 Variables that are @var{n} bytes or smaller will be allocated to the
13344 @code{.tiny} section. These variables use the @code{$gp} base
13345 register. The default for this option is 4, but note that there's a
13346 65536 byte limit to the @code{.tiny} section.
13351 @subsection MIPS Options
13352 @cindex MIPS options
13358 Generate big-endian code.
13362 Generate little-endian code. This is the default for @samp{mips*el-*-*}
13365 @item -march=@var{arch}
13367 Generate code that will run on @var{arch}, which can be the name of a
13368 generic MIPS ISA, or the name of a particular processor.
13370 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
13371 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
13372 The processor names are:
13373 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
13374 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
13375 @samp{5kc}, @samp{5kf},
13377 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
13378 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
13379 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
13380 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
13381 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
13382 @samp{loongson2e}, @samp{loongson2f},
13386 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
13387 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
13388 @samp{rm7000}, @samp{rm9000},
13389 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
13392 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
13393 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
13395 The special value @samp{from-abi} selects the
13396 most compatible architecture for the selected ABI (that is,
13397 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
13399 Native Linux/GNU toolchains also support the value @samp{native},
13400 which selects the best architecture option for the host processor.
13401 @option{-march=native} has no effect if GCC does not recognize
13404 In processor names, a final @samp{000} can be abbreviated as @samp{k}
13405 (for example, @samp{-march=r2k}). Prefixes are optional, and
13406 @samp{vr} may be written @samp{r}.
13408 Names of the form @samp{@var{n}f2_1} refer to processors with
13409 FPUs clocked at half the rate of the core, names of the form
13410 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
13411 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
13412 processors with FPUs clocked a ratio of 3:2 with respect to the core.
13413 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
13414 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
13415 accepted as synonyms for @samp{@var{n}f1_1}.
13417 GCC defines two macros based on the value of this option. The first
13418 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
13419 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
13420 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
13421 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
13422 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
13424 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
13425 above. In other words, it will have the full prefix and will not
13426 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
13427 the macro names the resolved architecture (either @samp{"mips1"} or
13428 @samp{"mips3"}). It names the default architecture when no
13429 @option{-march} option is given.
13431 @item -mtune=@var{arch}
13433 Optimize for @var{arch}. Among other things, this option controls
13434 the way instructions are scheduled, and the perceived cost of arithmetic
13435 operations. The list of @var{arch} values is the same as for
13438 When this option is not used, GCC will optimize for the processor
13439 specified by @option{-march}. By using @option{-march} and
13440 @option{-mtune} together, it is possible to generate code that will
13441 run on a family of processors, but optimize the code for one
13442 particular member of that family.
13444 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
13445 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
13446 @samp{-march} ones described above.
13450 Equivalent to @samp{-march=mips1}.
13454 Equivalent to @samp{-march=mips2}.
13458 Equivalent to @samp{-march=mips3}.
13462 Equivalent to @samp{-march=mips4}.
13466 Equivalent to @samp{-march=mips32}.
13470 Equivalent to @samp{-march=mips32r2}.
13474 Equivalent to @samp{-march=mips64}.
13478 Equivalent to @samp{-march=mips64r2}.
13483 @opindex mno-mips16
13484 Generate (do not generate) MIPS16 code. If GCC is targetting a
13485 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
13487 MIPS16 code generation can also be controlled on a per-function basis
13488 by means of @code{mips16} and @code{nomips16} attributes.
13489 @xref{Function Attributes}, for more information.
13491 @item -mflip-mips16
13492 @opindex mflip-mips16
13493 Generate MIPS16 code on alternating functions. This option is provided
13494 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
13495 not intended for ordinary use in compiling user code.
13497 @item -minterlink-mips16
13498 @itemx -mno-interlink-mips16
13499 @opindex minterlink-mips16
13500 @opindex mno-interlink-mips16
13501 Require (do not require) that non-MIPS16 code be link-compatible with
13504 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
13505 it must either use a call or an indirect jump. @option{-minterlink-mips16}
13506 therefore disables direct jumps unless GCC knows that the target of the
13507 jump is not MIPS16.
13519 Generate code for the given ABI@.
13521 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
13522 generates 64-bit code when you select a 64-bit architecture, but you
13523 can use @option{-mgp32} to get 32-bit code instead.
13525 For information about the O64 ABI, see
13526 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
13528 GCC supports a variant of the o32 ABI in which floating-point registers
13529 are 64 rather than 32 bits wide. You can select this combination with
13530 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
13531 and @samp{mfhc1} instructions and is therefore only supported for
13532 MIPS32R2 processors.
13534 The register assignments for arguments and return values remain the
13535 same, but each scalar value is passed in a single 64-bit register
13536 rather than a pair of 32-bit registers. For example, scalar
13537 floating-point values are returned in @samp{$f0} only, not a
13538 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
13539 remains the same, but all 64 bits are saved.
13542 @itemx -mno-abicalls
13544 @opindex mno-abicalls
13545 Generate (do not generate) code that is suitable for SVR4-style
13546 dynamic objects. @option{-mabicalls} is the default for SVR4-based
13551 Generate (do not generate) code that is fully position-independent,
13552 and that can therefore be linked into shared libraries. This option
13553 only affects @option{-mabicalls}.
13555 All @option{-mabicalls} code has traditionally been position-independent,
13556 regardless of options like @option{-fPIC} and @option{-fpic}. However,
13557 as an extension, the GNU toolchain allows executables to use absolute
13558 accesses for locally-binding symbols. It can also use shorter GP
13559 initialization sequences and generate direct calls to locally-defined
13560 functions. This mode is selected by @option{-mno-shared}.
13562 @option{-mno-shared} depends on binutils 2.16 or higher and generates
13563 objects that can only be linked by the GNU linker. However, the option
13564 does not affect the ABI of the final executable; it only affects the ABI
13565 of relocatable objects. Using @option{-mno-shared} will generally make
13566 executables both smaller and quicker.
13568 @option{-mshared} is the default.
13574 Assume (do not assume) that the static and dynamic linkers
13575 support PLTs and copy relocations. This option only affects
13576 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
13577 has no effect without @samp{-msym32}.
13579 You can make @option{-mplt} the default by configuring
13580 GCC with @option{--with-mips-plt}. The default is
13581 @option{-mno-plt} otherwise.
13587 Lift (do not lift) the usual restrictions on the size of the global
13590 GCC normally uses a single instruction to load values from the GOT@.
13591 While this is relatively efficient, it will only work if the GOT
13592 is smaller than about 64k. Anything larger will cause the linker
13593 to report an error such as:
13595 @cindex relocation truncated to fit (MIPS)
13597 relocation truncated to fit: R_MIPS_GOT16 foobar
13600 If this happens, you should recompile your code with @option{-mxgot}.
13601 It should then work with very large GOTs, although it will also be
13602 less efficient, since it will take three instructions to fetch the
13603 value of a global symbol.
13605 Note that some linkers can create multiple GOTs. If you have such a
13606 linker, you should only need to use @option{-mxgot} when a single object
13607 file accesses more than 64k's worth of GOT entries. Very few do.
13609 These options have no effect unless GCC is generating position
13614 Assume that general-purpose registers are 32 bits wide.
13618 Assume that general-purpose registers are 64 bits wide.
13622 Assume that floating-point registers are 32 bits wide.
13626 Assume that floating-point registers are 64 bits wide.
13629 @opindex mhard-float
13630 Use floating-point coprocessor instructions.
13633 @opindex msoft-float
13634 Do not use floating-point coprocessor instructions. Implement
13635 floating-point calculations using library calls instead.
13637 @item -msingle-float
13638 @opindex msingle-float
13639 Assume that the floating-point coprocessor only supports single-precision
13642 @item -mdouble-float
13643 @opindex mdouble-float
13644 Assume that the floating-point coprocessor supports double-precision
13645 operations. This is the default.
13651 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
13652 implement atomic memory built-in functions. When neither option is
13653 specified, GCC will use the instructions if the target architecture
13656 @option{-mllsc} is useful if the runtime environment can emulate the
13657 instructions and @option{-mno-llsc} can be useful when compiling for
13658 nonstandard ISAs. You can make either option the default by
13659 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
13660 respectively. @option{--with-llsc} is the default for some
13661 configurations; see the installation documentation for details.
13667 Use (do not use) revision 1 of the MIPS DSP ASE@.
13668 @xref{MIPS DSP Built-in Functions}. This option defines the
13669 preprocessor macro @samp{__mips_dsp}. It also defines
13670 @samp{__mips_dsp_rev} to 1.
13676 Use (do not use) revision 2 of the MIPS DSP ASE@.
13677 @xref{MIPS DSP Built-in Functions}. This option defines the
13678 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
13679 It also defines @samp{__mips_dsp_rev} to 2.
13682 @itemx -mno-smartmips
13683 @opindex msmartmips
13684 @opindex mno-smartmips
13685 Use (do not use) the MIPS SmartMIPS ASE.
13687 @item -mpaired-single
13688 @itemx -mno-paired-single
13689 @opindex mpaired-single
13690 @opindex mno-paired-single
13691 Use (do not use) paired-single floating-point instructions.
13692 @xref{MIPS Paired-Single Support}. This option requires
13693 hardware floating-point support to be enabled.
13699 Use (do not use) MIPS Digital Media Extension instructions.
13700 This option can only be used when generating 64-bit code and requires
13701 hardware floating-point support to be enabled.
13706 @opindex mno-mips3d
13707 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
13708 The option @option{-mips3d} implies @option{-mpaired-single}.
13714 Use (do not use) MT Multithreading instructions.
13718 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
13719 an explanation of the default and the way that the pointer size is
13724 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
13726 The default size of @code{int}s, @code{long}s and pointers depends on
13727 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
13728 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
13729 32-bit @code{long}s. Pointers are the same size as @code{long}s,
13730 or the same size as integer registers, whichever is smaller.
13736 Assume (do not assume) that all symbols have 32-bit values, regardless
13737 of the selected ABI@. This option is useful in combination with
13738 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
13739 to generate shorter and faster references to symbolic addresses.
13743 Put definitions of externally-visible data in a small data section
13744 if that data is no bigger than @var{num} bytes. GCC can then access
13745 the data more efficiently; see @option{-mgpopt} for details.
13747 The default @option{-G} option depends on the configuration.
13749 @item -mlocal-sdata
13750 @itemx -mno-local-sdata
13751 @opindex mlocal-sdata
13752 @opindex mno-local-sdata
13753 Extend (do not extend) the @option{-G} behavior to local data too,
13754 such as to static variables in C@. @option{-mlocal-sdata} is the
13755 default for all configurations.
13757 If the linker complains that an application is using too much small data,
13758 you might want to try rebuilding the less performance-critical parts with
13759 @option{-mno-local-sdata}. You might also want to build large
13760 libraries with @option{-mno-local-sdata}, so that the libraries leave
13761 more room for the main program.
13763 @item -mextern-sdata
13764 @itemx -mno-extern-sdata
13765 @opindex mextern-sdata
13766 @opindex mno-extern-sdata
13767 Assume (do not assume) that externally-defined data will be in
13768 a small data section if that data is within the @option{-G} limit.
13769 @option{-mextern-sdata} is the default for all configurations.
13771 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
13772 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
13773 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
13774 is placed in a small data section. If @var{Var} is defined by another
13775 module, you must either compile that module with a high-enough
13776 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
13777 definition. If @var{Var} is common, you must link the application
13778 with a high-enough @option{-G} setting.
13780 The easiest way of satisfying these restrictions is to compile
13781 and link every module with the same @option{-G} option. However,
13782 you may wish to build a library that supports several different
13783 small data limits. You can do this by compiling the library with
13784 the highest supported @option{-G} setting and additionally using
13785 @option{-mno-extern-sdata} to stop the library from making assumptions
13786 about externally-defined data.
13792 Use (do not use) GP-relative accesses for symbols that are known to be
13793 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
13794 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
13797 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
13798 might not hold the value of @code{_gp}. For example, if the code is
13799 part of a library that might be used in a boot monitor, programs that
13800 call boot monitor routines will pass an unknown value in @code{$gp}.
13801 (In such situations, the boot monitor itself would usually be compiled
13802 with @option{-G0}.)
13804 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
13805 @option{-mno-extern-sdata}.
13807 @item -membedded-data
13808 @itemx -mno-embedded-data
13809 @opindex membedded-data
13810 @opindex mno-embedded-data
13811 Allocate variables to the read-only data section first if possible, then
13812 next in the small data section if possible, otherwise in data. This gives
13813 slightly slower code than the default, but reduces the amount of RAM required
13814 when executing, and thus may be preferred for some embedded systems.
13816 @item -muninit-const-in-rodata
13817 @itemx -mno-uninit-const-in-rodata
13818 @opindex muninit-const-in-rodata
13819 @opindex mno-uninit-const-in-rodata
13820 Put uninitialized @code{const} variables in the read-only data section.
13821 This option is only meaningful in conjunction with @option{-membedded-data}.
13823 @item -mcode-readable=@var{setting}
13824 @opindex mcode-readable
13825 Specify whether GCC may generate code that reads from executable sections.
13826 There are three possible settings:
13829 @item -mcode-readable=yes
13830 Instructions may freely access executable sections. This is the
13833 @item -mcode-readable=pcrel
13834 MIPS16 PC-relative load instructions can access executable sections,
13835 but other instructions must not do so. This option is useful on 4KSc
13836 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
13837 It is also useful on processors that can be configured to have a dual
13838 instruction/data SRAM interface and that, like the M4K, automatically
13839 redirect PC-relative loads to the instruction RAM.
13841 @item -mcode-readable=no
13842 Instructions must not access executable sections. This option can be
13843 useful on targets that are configured to have a dual instruction/data
13844 SRAM interface but that (unlike the M4K) do not automatically redirect
13845 PC-relative loads to the instruction RAM.
13848 @item -msplit-addresses
13849 @itemx -mno-split-addresses
13850 @opindex msplit-addresses
13851 @opindex mno-split-addresses
13852 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
13853 relocation operators. This option has been superseded by
13854 @option{-mexplicit-relocs} but is retained for backwards compatibility.
13856 @item -mexplicit-relocs
13857 @itemx -mno-explicit-relocs
13858 @opindex mexplicit-relocs
13859 @opindex mno-explicit-relocs
13860 Use (do not use) assembler relocation operators when dealing with symbolic
13861 addresses. The alternative, selected by @option{-mno-explicit-relocs},
13862 is to use assembler macros instead.
13864 @option{-mexplicit-relocs} is the default if GCC was configured
13865 to use an assembler that supports relocation operators.
13867 @item -mcheck-zero-division
13868 @itemx -mno-check-zero-division
13869 @opindex mcheck-zero-division
13870 @opindex mno-check-zero-division
13871 Trap (do not trap) on integer division by zero.
13873 The default is @option{-mcheck-zero-division}.
13875 @item -mdivide-traps
13876 @itemx -mdivide-breaks
13877 @opindex mdivide-traps
13878 @opindex mdivide-breaks
13879 MIPS systems check for division by zero by generating either a
13880 conditional trap or a break instruction. Using traps results in
13881 smaller code, but is only supported on MIPS II and later. Also, some
13882 versions of the Linux kernel have a bug that prevents trap from
13883 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
13884 allow conditional traps on architectures that support them and
13885 @option{-mdivide-breaks} to force the use of breaks.
13887 The default is usually @option{-mdivide-traps}, but this can be
13888 overridden at configure time using @option{--with-divide=breaks}.
13889 Divide-by-zero checks can be completely disabled using
13890 @option{-mno-check-zero-division}.
13895 @opindex mno-memcpy
13896 Force (do not force) the use of @code{memcpy()} for non-trivial block
13897 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
13898 most constant-sized copies.
13901 @itemx -mno-long-calls
13902 @opindex mlong-calls
13903 @opindex mno-long-calls
13904 Disable (do not disable) use of the @code{jal} instruction. Calling
13905 functions using @code{jal} is more efficient but requires the caller
13906 and callee to be in the same 256 megabyte segment.
13908 This option has no effect on abicalls code. The default is
13909 @option{-mno-long-calls}.
13915 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
13916 instructions, as provided by the R4650 ISA@.
13919 @itemx -mno-fused-madd
13920 @opindex mfused-madd
13921 @opindex mno-fused-madd
13922 Enable (disable) use of the floating point multiply-accumulate
13923 instructions, when they are available. The default is
13924 @option{-mfused-madd}.
13926 When multiply-accumulate instructions are used, the intermediate
13927 product is calculated to infinite precision and is not subject to
13928 the FCSR Flush to Zero bit. This may be undesirable in some
13933 Tell the MIPS assembler to not run its preprocessor over user
13934 assembler files (with a @samp{.s} suffix) when assembling them.
13937 @itemx -mno-fix-r4000
13938 @opindex mfix-r4000
13939 @opindex mno-fix-r4000
13940 Work around certain R4000 CPU errata:
13943 A double-word or a variable shift may give an incorrect result if executed
13944 immediately after starting an integer division.
13946 A double-word or a variable shift may give an incorrect result if executed
13947 while an integer multiplication is in progress.
13949 An integer division may give an incorrect result if started in a delay slot
13950 of a taken branch or a jump.
13954 @itemx -mno-fix-r4400
13955 @opindex mfix-r4400
13956 @opindex mno-fix-r4400
13957 Work around certain R4400 CPU errata:
13960 A double-word or a variable shift may give an incorrect result if executed
13961 immediately after starting an integer division.
13965 @itemx -mno-fix-r10000
13966 @opindex mfix-r10000
13967 @opindex mno-fix-r10000
13968 Work around certain R10000 errata:
13971 @code{ll}/@code{sc} sequences may not behave atomically on revisions
13972 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
13975 This option can only be used if the target architecture supports
13976 branch-likely instructions. @option{-mfix-r10000} is the default when
13977 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
13981 @itemx -mno-fix-vr4120
13982 @opindex mfix-vr4120
13983 Work around certain VR4120 errata:
13986 @code{dmultu} does not always produce the correct result.
13988 @code{div} and @code{ddiv} do not always produce the correct result if one
13989 of the operands is negative.
13991 The workarounds for the division errata rely on special functions in
13992 @file{libgcc.a}. At present, these functions are only provided by
13993 the @code{mips64vr*-elf} configurations.
13995 Other VR4120 errata require a nop to be inserted between certain pairs of
13996 instructions. These errata are handled by the assembler, not by GCC itself.
13999 @opindex mfix-vr4130
14000 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
14001 workarounds are implemented by the assembler rather than by GCC,
14002 although GCC will avoid using @code{mflo} and @code{mfhi} if the
14003 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
14004 instructions are available instead.
14007 @itemx -mno-fix-sb1
14009 Work around certain SB-1 CPU core errata.
14010 (This flag currently works around the SB-1 revision 2
14011 ``F1'' and ``F2'' floating point errata.)
14013 @item -mr10k-cache-barrier=@var{setting}
14014 @opindex mr10k-cache-barrier
14015 Specify whether GCC should insert cache barriers to avoid the
14016 side-effects of speculation on R10K processors.
14018 In common with many processors, the R10K tries to predict the outcome
14019 of a conditional branch and speculatively executes instructions from
14020 the ``taken'' branch. It later aborts these instructions if the
14021 predicted outcome was wrong. However, on the R10K, even aborted
14022 instructions can have side effects.
14024 This problem only affects kernel stores and, depending on the system,
14025 kernel loads. As an example, a speculatively-executed store may load
14026 the target memory into cache and mark the cache line as dirty, even if
14027 the store itself is later aborted. If a DMA operation writes to the
14028 same area of memory before the ``dirty'' line is flushed, the cached
14029 data will overwrite the DMA-ed data. See the R10K processor manual
14030 for a full description, including other potential problems.
14032 One workaround is to insert cache barrier instructions before every memory
14033 access that might be speculatively executed and that might have side
14034 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
14035 controls GCC's implementation of this workaround. It assumes that
14036 aborted accesses to any byte in the following regions will not have
14041 the memory occupied by the current function's stack frame;
14044 the memory occupied by an incoming stack argument;
14047 the memory occupied by an object with a link-time-constant address.
14050 It is the kernel's responsibility to ensure that speculative
14051 accesses to these regions are indeed safe.
14053 If the input program contains a function declaration such as:
14059 then the implementation of @code{foo} must allow @code{j foo} and
14060 @code{jal foo} to be executed speculatively. GCC honors this
14061 restriction for functions it compiles itself. It expects non-GCC
14062 functions (such as hand-written assembly code) to do the same.
14064 The option has three forms:
14067 @item -mr10k-cache-barrier=load-store
14068 Insert a cache barrier before a load or store that might be
14069 speculatively executed and that might have side effects even
14072 @item -mr10k-cache-barrier=store
14073 Insert a cache barrier before a store that might be speculatively
14074 executed and that might have side effects even if aborted.
14076 @item -mr10k-cache-barrier=none
14077 Disable the insertion of cache barriers. This is the default setting.
14080 @item -mflush-func=@var{func}
14081 @itemx -mno-flush-func
14082 @opindex mflush-func
14083 Specifies the function to call to flush the I and D caches, or to not
14084 call any such function. If called, the function must take the same
14085 arguments as the common @code{_flush_func()}, that is, the address of the
14086 memory range for which the cache is being flushed, the size of the
14087 memory range, and the number 3 (to flush both caches). The default
14088 depends on the target GCC was configured for, but commonly is either
14089 @samp{_flush_func} or @samp{__cpu_flush}.
14091 @item mbranch-cost=@var{num}
14092 @opindex mbranch-cost
14093 Set the cost of branches to roughly @var{num} ``simple'' instructions.
14094 This cost is only a heuristic and is not guaranteed to produce
14095 consistent results across releases. A zero cost redundantly selects
14096 the default, which is based on the @option{-mtune} setting.
14098 @item -mbranch-likely
14099 @itemx -mno-branch-likely
14100 @opindex mbranch-likely
14101 @opindex mno-branch-likely
14102 Enable or disable use of Branch Likely instructions, regardless of the
14103 default for the selected architecture. By default, Branch Likely
14104 instructions may be generated if they are supported by the selected
14105 architecture. An exception is for the MIPS32 and MIPS64 architectures
14106 and processors which implement those architectures; for those, Branch
14107 Likely instructions will not be generated by default because the MIPS32
14108 and MIPS64 architectures specifically deprecate their use.
14110 @item -mfp-exceptions
14111 @itemx -mno-fp-exceptions
14112 @opindex mfp-exceptions
14113 Specifies whether FP exceptions are enabled. This affects how we schedule
14114 FP instructions for some processors. The default is that FP exceptions are
14117 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
14118 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
14121 @item -mvr4130-align
14122 @itemx -mno-vr4130-align
14123 @opindex mvr4130-align
14124 The VR4130 pipeline is two-way superscalar, but can only issue two
14125 instructions together if the first one is 8-byte aligned. When this
14126 option is enabled, GCC will align pairs of instructions that it
14127 thinks should execute in parallel.
14129 This option only has an effect when optimizing for the VR4130.
14130 It normally makes code faster, but at the expense of making it bigger.
14131 It is enabled by default at optimization level @option{-O3}.
14136 Enable (disable) generation of @code{synci} instructions on
14137 architectures that support it. The @code{synci} instructions (if
14138 enabled) will be generated when @code{__builtin___clear_cache()} is
14141 This option defaults to @code{-mno-synci}, but the default can be
14142 overridden by configuring with @code{--with-synci}.
14144 When compiling code for single processor systems, it is generally safe
14145 to use @code{synci}. However, on many multi-core (SMP) systems, it
14146 will not invalidate the instruction caches on all cores and may lead
14147 to undefined behavior.
14149 @item -mrelax-pic-calls
14150 @itemx -mno-relax-pic-calls
14151 @opindex mrelax-pic-calls
14152 Try to turn PIC calls that are normally dispatched via register
14153 @code{$25} into direct calls. This is only possible if the linker can
14154 resolve the destination at link-time and if the destination is within
14155 range for a direct call.
14157 @option{-mrelax-pic-calls} is the default if GCC was configured to use
14158 an assembler and a linker that supports the @code{.reloc} assembly
14159 directive and @code{-mexplicit-relocs} is in effect. With
14160 @code{-mno-explicit-relocs}, this optimization can be performed by the
14161 assembler and the linker alone without help from the compiler.
14165 @subsection MMIX Options
14166 @cindex MMIX Options
14168 These options are defined for the MMIX:
14172 @itemx -mno-libfuncs
14174 @opindex mno-libfuncs
14175 Specify that intrinsic library functions are being compiled, passing all
14176 values in registers, no matter the size.
14179 @itemx -mno-epsilon
14181 @opindex mno-epsilon
14182 Generate floating-point comparison instructions that compare with respect
14183 to the @code{rE} epsilon register.
14185 @item -mabi=mmixware
14187 @opindex mabi=mmixware
14189 Generate code that passes function parameters and return values that (in
14190 the called function) are seen as registers @code{$0} and up, as opposed to
14191 the GNU ABI which uses global registers @code{$231} and up.
14193 @item -mzero-extend
14194 @itemx -mno-zero-extend
14195 @opindex mzero-extend
14196 @opindex mno-zero-extend
14197 When reading data from memory in sizes shorter than 64 bits, use (do not
14198 use) zero-extending load instructions by default, rather than
14199 sign-extending ones.
14202 @itemx -mno-knuthdiv
14204 @opindex mno-knuthdiv
14205 Make the result of a division yielding a remainder have the same sign as
14206 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
14207 remainder follows the sign of the dividend. Both methods are
14208 arithmetically valid, the latter being almost exclusively used.
14210 @item -mtoplevel-symbols
14211 @itemx -mno-toplevel-symbols
14212 @opindex mtoplevel-symbols
14213 @opindex mno-toplevel-symbols
14214 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
14215 code can be used with the @code{PREFIX} assembly directive.
14219 Generate an executable in the ELF format, rather than the default
14220 @samp{mmo} format used by the @command{mmix} simulator.
14222 @item -mbranch-predict
14223 @itemx -mno-branch-predict
14224 @opindex mbranch-predict
14225 @opindex mno-branch-predict
14226 Use (do not use) the probable-branch instructions, when static branch
14227 prediction indicates a probable branch.
14229 @item -mbase-addresses
14230 @itemx -mno-base-addresses
14231 @opindex mbase-addresses
14232 @opindex mno-base-addresses
14233 Generate (do not generate) code that uses @emph{base addresses}. Using a
14234 base address automatically generates a request (handled by the assembler
14235 and the linker) for a constant to be set up in a global register. The
14236 register is used for one or more base address requests within the range 0
14237 to 255 from the value held in the register. The generally leads to short
14238 and fast code, but the number of different data items that can be
14239 addressed is limited. This means that a program that uses lots of static
14240 data may require @option{-mno-base-addresses}.
14242 @item -msingle-exit
14243 @itemx -mno-single-exit
14244 @opindex msingle-exit
14245 @opindex mno-single-exit
14246 Force (do not force) generated code to have a single exit point in each
14250 @node MN10300 Options
14251 @subsection MN10300 Options
14252 @cindex MN10300 options
14254 These @option{-m} options are defined for Matsushita MN10300 architectures:
14259 Generate code to avoid bugs in the multiply instructions for the MN10300
14260 processors. This is the default.
14262 @item -mno-mult-bug
14263 @opindex mno-mult-bug
14264 Do not generate code to avoid bugs in the multiply instructions for the
14265 MN10300 processors.
14269 Generate code which uses features specific to the AM33 processor.
14273 Do not generate code which uses features specific to the AM33 processor. This
14276 @item -mreturn-pointer-on-d0
14277 @opindex mreturn-pointer-on-d0
14278 When generating a function which returns a pointer, return the pointer
14279 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
14280 only in a0, and attempts to call such functions without a prototype
14281 would result in errors. Note that this option is on by default; use
14282 @option{-mno-return-pointer-on-d0} to disable it.
14286 Do not link in the C run-time initialization object file.
14290 Indicate to the linker that it should perform a relaxation optimization pass
14291 to shorten branches, calls and absolute memory addresses. This option only
14292 has an effect when used on the command line for the final link step.
14294 This option makes symbolic debugging impossible.
14297 @node PDP-11 Options
14298 @subsection PDP-11 Options
14299 @cindex PDP-11 Options
14301 These options are defined for the PDP-11:
14306 Use hardware FPP floating point. This is the default. (FIS floating
14307 point on the PDP-11/40 is not supported.)
14310 @opindex msoft-float
14311 Do not use hardware floating point.
14315 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
14319 Return floating-point results in memory. This is the default.
14323 Generate code for a PDP-11/40.
14327 Generate code for a PDP-11/45. This is the default.
14331 Generate code for a PDP-11/10.
14333 @item -mbcopy-builtin
14334 @opindex mbcopy-builtin
14335 Use inline @code{movmemhi} patterns for copying memory. This is the
14340 Do not use inline @code{movmemhi} patterns for copying memory.
14346 Use 16-bit @code{int}. This is the default.
14352 Use 32-bit @code{int}.
14355 @itemx -mno-float32
14357 @opindex mno-float32
14358 Use 64-bit @code{float}. This is the default.
14361 @itemx -mno-float64
14363 @opindex mno-float64
14364 Use 32-bit @code{float}.
14368 Use @code{abshi2} pattern. This is the default.
14372 Do not use @code{abshi2} pattern.
14374 @item -mbranch-expensive
14375 @opindex mbranch-expensive
14376 Pretend that branches are expensive. This is for experimenting with
14377 code generation only.
14379 @item -mbranch-cheap
14380 @opindex mbranch-cheap
14381 Do not pretend that branches are expensive. This is the default.
14385 Generate code for a system with split I&D@.
14389 Generate code for a system without split I&D@. This is the default.
14393 Use Unix assembler syntax. This is the default when configured for
14394 @samp{pdp11-*-bsd}.
14398 Use DEC assembler syntax. This is the default when configured for any
14399 PDP-11 target other than @samp{pdp11-*-bsd}.
14402 @node picoChip Options
14403 @subsection picoChip Options
14404 @cindex picoChip options
14406 These @samp{-m} options are defined for picoChip implementations:
14410 @item -mae=@var{ae_type}
14412 Set the instruction set, register set, and instruction scheduling
14413 parameters for array element type @var{ae_type}. Supported values
14414 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
14416 @option{-mae=ANY} selects a completely generic AE type. Code
14417 generated with this option will run on any of the other AE types. The
14418 code will not be as efficient as it would be if compiled for a specific
14419 AE type, and some types of operation (e.g., multiplication) will not
14420 work properly on all types of AE.
14422 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
14423 for compiled code, and is the default.
14425 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
14426 option may suffer from poor performance of byte (char) manipulation,
14427 since the DSP AE does not provide hardware support for byte load/stores.
14429 @item -msymbol-as-address
14430 Enable the compiler to directly use a symbol name as an address in a
14431 load/store instruction, without first loading it into a
14432 register. Typically, the use of this option will generate larger
14433 programs, which run faster than when the option isn't used. However, the
14434 results vary from program to program, so it is left as a user option,
14435 rather than being permanently enabled.
14437 @item -mno-inefficient-warnings
14438 Disables warnings about the generation of inefficient code. These
14439 warnings can be generated, for example, when compiling code which
14440 performs byte-level memory operations on the MAC AE type. The MAC AE has
14441 no hardware support for byte-level memory operations, so all byte
14442 load/stores must be synthesized from word load/store operations. This is
14443 inefficient and a warning will be generated indicating to the programmer
14444 that they should rewrite the code to avoid byte operations, or to target
14445 an AE type which has the necessary hardware support. This option enables
14446 the warning to be turned off.
14450 @node PowerPC Options
14451 @subsection PowerPC Options
14452 @cindex PowerPC options
14454 These are listed under @xref{RS/6000 and PowerPC Options}.
14456 @node RS/6000 and PowerPC Options
14457 @subsection IBM RS/6000 and PowerPC Options
14458 @cindex RS/6000 and PowerPC Options
14459 @cindex IBM RS/6000 and PowerPC Options
14461 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
14468 @itemx -mno-powerpc
14469 @itemx -mpowerpc-gpopt
14470 @itemx -mno-powerpc-gpopt
14471 @itemx -mpowerpc-gfxopt
14472 @itemx -mno-powerpc-gfxopt
14474 @itemx -mno-powerpc64
14478 @itemx -mno-popcntb
14480 @itemx -mno-popcntd
14488 @itemx -mno-hard-dfp
14492 @opindex mno-power2
14494 @opindex mno-powerpc
14495 @opindex mpowerpc-gpopt
14496 @opindex mno-powerpc-gpopt
14497 @opindex mpowerpc-gfxopt
14498 @opindex mno-powerpc-gfxopt
14499 @opindex mpowerpc64
14500 @opindex mno-powerpc64
14504 @opindex mno-popcntb
14506 @opindex mno-popcntd
14512 @opindex mno-mfpgpr
14514 @opindex mno-hard-dfp
14515 GCC supports two related instruction set architectures for the
14516 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
14517 instructions supported by the @samp{rios} chip set used in the original
14518 RS/6000 systems and the @dfn{PowerPC} instruction set is the
14519 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
14520 the IBM 4xx, 6xx, and follow-on microprocessors.
14522 Neither architecture is a subset of the other. However there is a
14523 large common subset of instructions supported by both. An MQ
14524 register is included in processors supporting the POWER architecture.
14526 You use these options to specify which instructions are available on the
14527 processor you are using. The default value of these options is
14528 determined when configuring GCC@. Specifying the
14529 @option{-mcpu=@var{cpu_type}} overrides the specification of these
14530 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
14531 rather than the options listed above.
14533 The @option{-mpower} option allows GCC to generate instructions that
14534 are found only in the POWER architecture and to use the MQ register.
14535 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
14536 to generate instructions that are present in the POWER2 architecture but
14537 not the original POWER architecture.
14539 The @option{-mpowerpc} option allows GCC to generate instructions that
14540 are found only in the 32-bit subset of the PowerPC architecture.
14541 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
14542 GCC to use the optional PowerPC architecture instructions in the
14543 General Purpose group, including floating-point square root. Specifying
14544 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
14545 use the optional PowerPC architecture instructions in the Graphics
14546 group, including floating-point select.
14548 The @option{-mmfcrf} option allows GCC to generate the move from
14549 condition register field instruction implemented on the POWER4
14550 processor and other processors that support the PowerPC V2.01
14552 The @option{-mpopcntb} option allows GCC to generate the popcount and
14553 double precision FP reciprocal estimate instruction implemented on the
14554 POWER5 processor and other processors that support the PowerPC V2.02
14556 The @option{-mpopcntd} option allows GCC to generate the popcount
14557 instruction implemented on the POWER7 processor and other processors
14558 that support the PowerPC V2.06 architecture.
14559 The @option{-mfprnd} option allows GCC to generate the FP round to
14560 integer instructions implemented on the POWER5+ processor and other
14561 processors that support the PowerPC V2.03 architecture.
14562 The @option{-mcmpb} option allows GCC to generate the compare bytes
14563 instruction implemented on the POWER6 processor and other processors
14564 that support the PowerPC V2.05 architecture.
14565 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
14566 general purpose register instructions implemented on the POWER6X
14567 processor and other processors that support the extended PowerPC V2.05
14569 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
14570 point instructions implemented on some POWER processors.
14572 The @option{-mpowerpc64} option allows GCC to generate the additional
14573 64-bit instructions that are found in the full PowerPC64 architecture
14574 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
14575 @option{-mno-powerpc64}.
14577 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
14578 will use only the instructions in the common subset of both
14579 architectures plus some special AIX common-mode calls, and will not use
14580 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
14581 permits GCC to use any instruction from either architecture and to
14582 allow use of the MQ register; specify this for the Motorola MPC601.
14584 @item -mnew-mnemonics
14585 @itemx -mold-mnemonics
14586 @opindex mnew-mnemonics
14587 @opindex mold-mnemonics
14588 Select which mnemonics to use in the generated assembler code. With
14589 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
14590 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
14591 assembler mnemonics defined for the POWER architecture. Instructions
14592 defined in only one architecture have only one mnemonic; GCC uses that
14593 mnemonic irrespective of which of these options is specified.
14595 GCC defaults to the mnemonics appropriate for the architecture in
14596 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
14597 value of these option. Unless you are building a cross-compiler, you
14598 should normally not specify either @option{-mnew-mnemonics} or
14599 @option{-mold-mnemonics}, but should instead accept the default.
14601 @item -mcpu=@var{cpu_type}
14603 Set architecture type, register usage, choice of mnemonics, and
14604 instruction scheduling parameters for machine type @var{cpu_type}.
14605 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
14606 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
14607 @samp{505}, @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
14608 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
14609 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
14610 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
14611 @samp{e300c3}, @samp{e500mc}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
14612 @samp{power}, @samp{power2}, @samp{power3}, @samp{power4},
14613 @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x}, @samp{power7},
14614 @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
14615 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
14617 @option{-mcpu=common} selects a completely generic processor. Code
14618 generated under this option will run on any POWER or PowerPC processor.
14619 GCC will use only the instructions in the common subset of both
14620 architectures, and will not use the MQ register. GCC assumes a generic
14621 processor model for scheduling purposes.
14623 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
14624 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
14625 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
14626 types, with an appropriate, generic processor model assumed for
14627 scheduling purposes.
14629 The other options specify a specific processor. Code generated under
14630 those options will run best on that processor, and may not run at all on
14633 The @option{-mcpu} options automatically enable or disable the
14636 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
14637 -mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol
14638 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
14639 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx}
14641 The particular options set for any particular CPU will vary between
14642 compiler versions, depending on what setting seems to produce optimal
14643 code for that CPU; it doesn't necessarily reflect the actual hardware's
14644 capabilities. If you wish to set an individual option to a particular
14645 value, you may specify it after the @option{-mcpu} option, like
14646 @samp{-mcpu=970 -mno-altivec}.
14648 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
14649 not enabled or disabled by the @option{-mcpu} option at present because
14650 AIX does not have full support for these options. You may still
14651 enable or disable them individually if you're sure it'll work in your
14654 @item -mtune=@var{cpu_type}
14656 Set the instruction scheduling parameters for machine type
14657 @var{cpu_type}, but do not set the architecture type, register usage, or
14658 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
14659 values for @var{cpu_type} are used for @option{-mtune} as for
14660 @option{-mcpu}. If both are specified, the code generated will use the
14661 architecture, registers, and mnemonics set by @option{-mcpu}, but the
14662 scheduling parameters set by @option{-mtune}.
14668 Generate code to compute division as reciprocal estimate and iterative
14669 refinement, creating opportunities for increased throughput. This
14670 feature requires: optional PowerPC Graphics instruction set for single
14671 precision and FRE instruction for double precision, assuming divides
14672 cannot generate user-visible traps, and the domain values not include
14673 Infinities, denormals or zero denominator.
14676 @itemx -mno-altivec
14678 @opindex mno-altivec
14679 Generate code that uses (does not use) AltiVec instructions, and also
14680 enable the use of built-in functions that allow more direct access to
14681 the AltiVec instruction set. You may also need to set
14682 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
14688 @opindex mno-vrsave
14689 Generate VRSAVE instructions when generating AltiVec code.
14691 @item -mgen-cell-microcode
14692 @opindex mgen-cell-microcode
14693 Generate Cell microcode instructions
14695 @item -mwarn-cell-microcode
14696 @opindex mwarn-cell-microcode
14697 Warning when a Cell microcode instruction is going to emitted. An example
14698 of a Cell microcode instruction is a variable shift.
14701 @opindex msecure-plt
14702 Generate code that allows ld and ld.so to build executables and shared
14703 libraries with non-exec .plt and .got sections. This is a PowerPC
14704 32-bit SYSV ABI option.
14708 Generate code that uses a BSS .plt section that ld.so fills in, and
14709 requires .plt and .got sections that are both writable and executable.
14710 This is a PowerPC 32-bit SYSV ABI option.
14716 This switch enables or disables the generation of ISEL instructions.
14718 @item -misel=@var{yes/no}
14719 This switch has been deprecated. Use @option{-misel} and
14720 @option{-mno-isel} instead.
14726 This switch enables or disables the generation of SPE simd
14732 @opindex mno-paired
14733 This switch enables or disables the generation of PAIRED simd
14736 @item -mspe=@var{yes/no}
14737 This option has been deprecated. Use @option{-mspe} and
14738 @option{-mno-spe} instead.
14744 Generate code that uses (does not use) vector/scalar (VSX)
14745 instructions, and also enable the use of built-in functions that allow
14746 more direct access to the VSX instruction set.
14748 @item -mfloat-gprs=@var{yes/single/double/no}
14749 @itemx -mfloat-gprs
14750 @opindex mfloat-gprs
14751 This switch enables or disables the generation of floating point
14752 operations on the general purpose registers for architectures that
14755 The argument @var{yes} or @var{single} enables the use of
14756 single-precision floating point operations.
14758 The argument @var{double} enables the use of single and
14759 double-precision floating point operations.
14761 The argument @var{no} disables floating point operations on the
14762 general purpose registers.
14764 This option is currently only available on the MPC854x.
14770 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
14771 targets (including GNU/Linux). The 32-bit environment sets int, long
14772 and pointer to 32 bits and generates code that runs on any PowerPC
14773 variant. The 64-bit environment sets int to 32 bits and long and
14774 pointer to 64 bits, and generates code for PowerPC64, as for
14775 @option{-mpowerpc64}.
14778 @itemx -mno-fp-in-toc
14779 @itemx -mno-sum-in-toc
14780 @itemx -mminimal-toc
14782 @opindex mno-fp-in-toc
14783 @opindex mno-sum-in-toc
14784 @opindex mminimal-toc
14785 Modify generation of the TOC (Table Of Contents), which is created for
14786 every executable file. The @option{-mfull-toc} option is selected by
14787 default. In that case, GCC will allocate at least one TOC entry for
14788 each unique non-automatic variable reference in your program. GCC
14789 will also place floating-point constants in the TOC@. However, only
14790 16,384 entries are available in the TOC@.
14792 If you receive a linker error message that saying you have overflowed
14793 the available TOC space, you can reduce the amount of TOC space used
14794 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
14795 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
14796 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
14797 generate code to calculate the sum of an address and a constant at
14798 run-time instead of putting that sum into the TOC@. You may specify one
14799 or both of these options. Each causes GCC to produce very slightly
14800 slower and larger code at the expense of conserving TOC space.
14802 If you still run out of space in the TOC even when you specify both of
14803 these options, specify @option{-mminimal-toc} instead. This option causes
14804 GCC to make only one TOC entry for every file. When you specify this
14805 option, GCC will produce code that is slower and larger but which
14806 uses extremely little TOC space. You may wish to use this option
14807 only on files that contain less frequently executed code.
14813 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
14814 @code{long} type, and the infrastructure needed to support them.
14815 Specifying @option{-maix64} implies @option{-mpowerpc64} and
14816 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
14817 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
14820 @itemx -mno-xl-compat
14821 @opindex mxl-compat
14822 @opindex mno-xl-compat
14823 Produce code that conforms more closely to IBM XL compiler semantics
14824 when using AIX-compatible ABI@. Pass floating-point arguments to
14825 prototyped functions beyond the register save area (RSA) on the stack
14826 in addition to argument FPRs. Do not assume that most significant
14827 double in 128-bit long double value is properly rounded when comparing
14828 values and converting to double. Use XL symbol names for long double
14831 The AIX calling convention was extended but not initially documented to
14832 handle an obscure K&R C case of calling a function that takes the
14833 address of its arguments with fewer arguments than declared. IBM XL
14834 compilers access floating point arguments which do not fit in the
14835 RSA from the stack when a subroutine is compiled without
14836 optimization. Because always storing floating-point arguments on the
14837 stack is inefficient and rarely needed, this option is not enabled by
14838 default and only is necessary when calling subroutines compiled by IBM
14839 XL compilers without optimization.
14843 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
14844 application written to use message passing with special startup code to
14845 enable the application to run. The system must have PE installed in the
14846 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
14847 must be overridden with the @option{-specs=} option to specify the
14848 appropriate directory location. The Parallel Environment does not
14849 support threads, so the @option{-mpe} option and the @option{-pthread}
14850 option are incompatible.
14852 @item -malign-natural
14853 @itemx -malign-power
14854 @opindex malign-natural
14855 @opindex malign-power
14856 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
14857 @option{-malign-natural} overrides the ABI-defined alignment of larger
14858 types, such as floating-point doubles, on their natural size-based boundary.
14859 The option @option{-malign-power} instructs GCC to follow the ABI-specified
14860 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
14862 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
14866 @itemx -mhard-float
14867 @opindex msoft-float
14868 @opindex mhard-float
14869 Generate code that does not use (uses) the floating-point register set.
14870 Software floating point emulation is provided if you use the
14871 @option{-msoft-float} option, and pass the option to GCC when linking.
14873 @item -msingle-float
14874 @itemx -mdouble-float
14875 @opindex msingle-float
14876 @opindex mdouble-float
14877 Generate code for single or double-precision floating point operations.
14878 @option{-mdouble-float} implies @option{-msingle-float}.
14881 @opindex msimple-fpu
14882 Do not generate sqrt and div instructions for hardware floating point unit.
14886 Specify type of floating point unit. Valid values are @var{sp_lite}
14887 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
14888 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
14889 and @var{dp_full} (equivalent to -mdouble-float).
14892 @opindex mxilinx-fpu
14893 Perform optimizations for floating point unit on Xilinx PPC 405/440.
14896 @itemx -mno-multiple
14898 @opindex mno-multiple
14899 Generate code that uses (does not use) the load multiple word
14900 instructions and the store multiple word instructions. These
14901 instructions are generated by default on POWER systems, and not
14902 generated on PowerPC systems. Do not use @option{-mmultiple} on little
14903 endian PowerPC systems, since those instructions do not work when the
14904 processor is in little endian mode. The exceptions are PPC740 and
14905 PPC750 which permit the instructions usage in little endian mode.
14910 @opindex mno-string
14911 Generate code that uses (does not use) the load string instructions
14912 and the store string word instructions to save multiple registers and
14913 do small block moves. These instructions are generated by default on
14914 POWER systems, and not generated on PowerPC systems. Do not use
14915 @option{-mstring} on little endian PowerPC systems, since those
14916 instructions do not work when the processor is in little endian mode.
14917 The exceptions are PPC740 and PPC750 which permit the instructions
14918 usage in little endian mode.
14923 @opindex mno-update
14924 Generate code that uses (does not use) the load or store instructions
14925 that update the base register to the address of the calculated memory
14926 location. These instructions are generated by default. If you use
14927 @option{-mno-update}, there is a small window between the time that the
14928 stack pointer is updated and the address of the previous frame is
14929 stored, which means code that walks the stack frame across interrupts or
14930 signals may get corrupted data.
14932 @item -mavoid-indexed-addresses
14933 @item -mno-avoid-indexed-addresses
14934 @opindex mavoid-indexed-addresses
14935 @opindex mno-avoid-indexed-addresses
14936 Generate code that tries to avoid (not avoid) the use of indexed load
14937 or store instructions. These instructions can incur a performance
14938 penalty on Power6 processors in certain situations, such as when
14939 stepping through large arrays that cross a 16M boundary. This option
14940 is enabled by default when targetting Power6 and disabled otherwise.
14943 @itemx -mno-fused-madd
14944 @opindex mfused-madd
14945 @opindex mno-fused-madd
14946 Generate code that uses (does not use) the floating point multiply and
14947 accumulate instructions. These instructions are generated by default if
14948 hardware floating is used.
14954 Generate code that uses (does not use) the half-word multiply and
14955 multiply-accumulate instructions on the IBM 405, 440 and 464 processors.
14956 These instructions are generated by default when targetting those
14963 Generate code that uses (does not use) the string-search @samp{dlmzb}
14964 instruction on the IBM 405, 440 and 464 processors. This instruction is
14965 generated by default when targetting those processors.
14967 @item -mno-bit-align
14969 @opindex mno-bit-align
14970 @opindex mbit-align
14971 On System V.4 and embedded PowerPC systems do not (do) force structures
14972 and unions that contain bit-fields to be aligned to the base type of the
14975 For example, by default a structure containing nothing but 8
14976 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
14977 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
14978 the structure would be aligned to a 1 byte boundary and be one byte in
14981 @item -mno-strict-align
14982 @itemx -mstrict-align
14983 @opindex mno-strict-align
14984 @opindex mstrict-align
14985 On System V.4 and embedded PowerPC systems do not (do) assume that
14986 unaligned memory references will be handled by the system.
14988 @item -mrelocatable
14989 @itemx -mno-relocatable
14990 @opindex mrelocatable
14991 @opindex mno-relocatable
14992 On embedded PowerPC systems generate code that allows (does not allow)
14993 the program to be relocated to a different address at runtime. If you
14994 use @option{-mrelocatable} on any module, all objects linked together must
14995 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
14997 @item -mrelocatable-lib
14998 @itemx -mno-relocatable-lib
14999 @opindex mrelocatable-lib
15000 @opindex mno-relocatable-lib
15001 On embedded PowerPC systems generate code that allows (does not allow)
15002 the program to be relocated to a different address at runtime. Modules
15003 compiled with @option{-mrelocatable-lib} can be linked with either modules
15004 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
15005 with modules compiled with the @option{-mrelocatable} options.
15011 On System V.4 and embedded PowerPC systems do not (do) assume that
15012 register 2 contains a pointer to a global area pointing to the addresses
15013 used in the program.
15016 @itemx -mlittle-endian
15018 @opindex mlittle-endian
15019 On System V.4 and embedded PowerPC systems compile code for the
15020 processor in little endian mode. The @option{-mlittle-endian} option is
15021 the same as @option{-mlittle}.
15024 @itemx -mbig-endian
15026 @opindex mbig-endian
15027 On System V.4 and embedded PowerPC systems compile code for the
15028 processor in big endian mode. The @option{-mbig-endian} option is
15029 the same as @option{-mbig}.
15031 @item -mdynamic-no-pic
15032 @opindex mdynamic-no-pic
15033 On Darwin and Mac OS X systems, compile code so that it is not
15034 relocatable, but that its external references are relocatable. The
15035 resulting code is suitable for applications, but not shared
15038 @item -mprioritize-restricted-insns=@var{priority}
15039 @opindex mprioritize-restricted-insns
15040 This option controls the priority that is assigned to
15041 dispatch-slot restricted instructions during the second scheduling
15042 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
15043 @var{no/highest/second-highest} priority to dispatch slot restricted
15046 @item -msched-costly-dep=@var{dependence_type}
15047 @opindex msched-costly-dep
15048 This option controls which dependences are considered costly
15049 by the target during instruction scheduling. The argument
15050 @var{dependence_type} takes one of the following values:
15051 @var{no}: no dependence is costly,
15052 @var{all}: all dependences are costly,
15053 @var{true_store_to_load}: a true dependence from store to load is costly,
15054 @var{store_to_load}: any dependence from store to load is costly,
15055 @var{number}: any dependence which latency >= @var{number} is costly.
15057 @item -minsert-sched-nops=@var{scheme}
15058 @opindex minsert-sched-nops
15059 This option controls which nop insertion scheme will be used during
15060 the second scheduling pass. The argument @var{scheme} takes one of the
15062 @var{no}: Don't insert nops.
15063 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
15064 according to the scheduler's grouping.
15065 @var{regroup_exact}: Insert nops to force costly dependent insns into
15066 separate groups. Insert exactly as many nops as needed to force an insn
15067 to a new group, according to the estimated processor grouping.
15068 @var{number}: Insert nops to force costly dependent insns into
15069 separate groups. Insert @var{number} nops to force an insn to a new group.
15072 @opindex mcall-sysv
15073 On System V.4 and embedded PowerPC systems compile code using calling
15074 conventions that adheres to the March 1995 draft of the System V
15075 Application Binary Interface, PowerPC processor supplement. This is the
15076 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
15078 @item -mcall-sysv-eabi
15080 @opindex mcall-sysv-eabi
15081 @opindex mcall-eabi
15082 Specify both @option{-mcall-sysv} and @option{-meabi} options.
15084 @item -mcall-sysv-noeabi
15085 @opindex mcall-sysv-noeabi
15086 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
15088 @item -mcall-aixdesc
15090 On System V.4 and embedded PowerPC systems compile code for the AIX
15094 @opindex mcall-linux
15095 On System V.4 and embedded PowerPC systems compile code for the
15096 Linux-based GNU system.
15100 On System V.4 and embedded PowerPC systems compile code for the
15101 Hurd-based GNU system.
15103 @item -mcall-freebsd
15104 @opindex mcall-freebsd
15105 On System V.4 and embedded PowerPC systems compile code for the
15106 FreeBSD operating system.
15108 @item -mcall-netbsd
15109 @opindex mcall-netbsd
15110 On System V.4 and embedded PowerPC systems compile code for the
15111 NetBSD operating system.
15113 @item -mcall-openbsd
15114 @opindex mcall-netbsd
15115 On System V.4 and embedded PowerPC systems compile code for the
15116 OpenBSD operating system.
15118 @item -maix-struct-return
15119 @opindex maix-struct-return
15120 Return all structures in memory (as specified by the AIX ABI)@.
15122 @item -msvr4-struct-return
15123 @opindex msvr4-struct-return
15124 Return structures smaller than 8 bytes in registers (as specified by the
15127 @item -mabi=@var{abi-type}
15129 Extend the current ABI with a particular extension, or remove such extension.
15130 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
15131 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
15135 Extend the current ABI with SPE ABI extensions. This does not change
15136 the default ABI, instead it adds the SPE ABI extensions to the current
15140 @opindex mabi=no-spe
15141 Disable Booke SPE ABI extensions for the current ABI@.
15143 @item -mabi=ibmlongdouble
15144 @opindex mabi=ibmlongdouble
15145 Change the current ABI to use IBM extended precision long double.
15146 This is a PowerPC 32-bit SYSV ABI option.
15148 @item -mabi=ieeelongdouble
15149 @opindex mabi=ieeelongdouble
15150 Change the current ABI to use IEEE extended precision long double.
15151 This is a PowerPC 32-bit Linux ABI option.
15154 @itemx -mno-prototype
15155 @opindex mprototype
15156 @opindex mno-prototype
15157 On System V.4 and embedded PowerPC systems assume that all calls to
15158 variable argument functions are properly prototyped. Otherwise, the
15159 compiler must insert an instruction before every non prototyped call to
15160 set or clear bit 6 of the condition code register (@var{CR}) to
15161 indicate whether floating point values were passed in the floating point
15162 registers in case the function takes a variable arguments. With
15163 @option{-mprototype}, only calls to prototyped variable argument functions
15164 will set or clear the bit.
15168 On embedded PowerPC systems, assume that the startup module is called
15169 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
15170 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
15175 On embedded PowerPC systems, assume that the startup module is called
15176 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
15181 On embedded PowerPC systems, assume that the startup module is called
15182 @file{crt0.o} and the standard C libraries are @file{libads.a} and
15185 @item -myellowknife
15186 @opindex myellowknife
15187 On embedded PowerPC systems, assume that the startup module is called
15188 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
15193 On System V.4 and embedded PowerPC systems, specify that you are
15194 compiling for a VxWorks system.
15198 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
15199 header to indicate that @samp{eabi} extended relocations are used.
15205 On System V.4 and embedded PowerPC systems do (do not) adhere to the
15206 Embedded Applications Binary Interface (eabi) which is a set of
15207 modifications to the System V.4 specifications. Selecting @option{-meabi}
15208 means that the stack is aligned to an 8 byte boundary, a function
15209 @code{__eabi} is called to from @code{main} to set up the eabi
15210 environment, and the @option{-msdata} option can use both @code{r2} and
15211 @code{r13} to point to two separate small data areas. Selecting
15212 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
15213 do not call an initialization function from @code{main}, and the
15214 @option{-msdata} option will only use @code{r13} to point to a single
15215 small data area. The @option{-meabi} option is on by default if you
15216 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
15219 @opindex msdata=eabi
15220 On System V.4 and embedded PowerPC systems, put small initialized
15221 @code{const} global and static data in the @samp{.sdata2} section, which
15222 is pointed to by register @code{r2}. Put small initialized
15223 non-@code{const} global and static data in the @samp{.sdata} section,
15224 which is pointed to by register @code{r13}. Put small uninitialized
15225 global and static data in the @samp{.sbss} section, which is adjacent to
15226 the @samp{.sdata} section. The @option{-msdata=eabi} option is
15227 incompatible with the @option{-mrelocatable} option. The
15228 @option{-msdata=eabi} option also sets the @option{-memb} option.
15231 @opindex msdata=sysv
15232 On System V.4 and embedded PowerPC systems, put small global and static
15233 data in the @samp{.sdata} section, which is pointed to by register
15234 @code{r13}. Put small uninitialized global and static data in the
15235 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
15236 The @option{-msdata=sysv} option is incompatible with the
15237 @option{-mrelocatable} option.
15239 @item -msdata=default
15241 @opindex msdata=default
15243 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
15244 compile code the same as @option{-msdata=eabi}, otherwise compile code the
15245 same as @option{-msdata=sysv}.
15248 @opindex msdata=data
15249 On System V.4 and embedded PowerPC systems, put small global
15250 data in the @samp{.sdata} section. Put small uninitialized global
15251 data in the @samp{.sbss} section. Do not use register @code{r13}
15252 to address small data however. This is the default behavior unless
15253 other @option{-msdata} options are used.
15257 @opindex msdata=none
15259 On embedded PowerPC systems, put all initialized global and static data
15260 in the @samp{.data} section, and all uninitialized data in the
15261 @samp{.bss} section.
15265 @cindex smaller data references (PowerPC)
15266 @cindex .sdata/.sdata2 references (PowerPC)
15267 On embedded PowerPC systems, put global and static items less than or
15268 equal to @var{num} bytes into the small data or bss sections instead of
15269 the normal data or bss section. By default, @var{num} is 8. The
15270 @option{-G @var{num}} switch is also passed to the linker.
15271 All modules should be compiled with the same @option{-G @var{num}} value.
15274 @itemx -mno-regnames
15276 @opindex mno-regnames
15277 On System V.4 and embedded PowerPC systems do (do not) emit register
15278 names in the assembly language output using symbolic forms.
15281 @itemx -mno-longcall
15283 @opindex mno-longcall
15284 By default assume that all calls are far away so that a longer more
15285 expensive calling sequence is required. This is required for calls
15286 further than 32 megabytes (33,554,432 bytes) from the current location.
15287 A short call will be generated if the compiler knows
15288 the call cannot be that far away. This setting can be overridden by
15289 the @code{shortcall} function attribute, or by @code{#pragma
15292 Some linkers are capable of detecting out-of-range calls and generating
15293 glue code on the fly. On these systems, long calls are unnecessary and
15294 generate slower code. As of this writing, the AIX linker can do this,
15295 as can the GNU linker for PowerPC/64. It is planned to add this feature
15296 to the GNU linker for 32-bit PowerPC systems as well.
15298 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
15299 callee, L42'', plus a ``branch island'' (glue code). The two target
15300 addresses represent the callee and the ``branch island''. The
15301 Darwin/PPC linker will prefer the first address and generate a ``bl
15302 callee'' if the PPC ``bl'' instruction will reach the callee directly;
15303 otherwise, the linker will generate ``bl L42'' to call the ``branch
15304 island''. The ``branch island'' is appended to the body of the
15305 calling function; it computes the full 32-bit address of the callee
15308 On Mach-O (Darwin) systems, this option directs the compiler emit to
15309 the glue for every direct call, and the Darwin linker decides whether
15310 to use or discard it.
15312 In the future, we may cause GCC to ignore all longcall specifications
15313 when the linker is known to generate glue.
15315 @item -mtls-markers
15316 @itemx -mno-tls-markers
15317 @opindex mtls-markers
15318 @opindex mno-tls-markers
15319 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
15320 specifying the function argument. The relocation allows ld to
15321 reliably associate function call with argument setup instructions for
15322 TLS optimization, which in turn allows gcc to better schedule the
15327 Adds support for multithreading with the @dfn{pthreads} library.
15328 This option sets flags for both the preprocessor and linker.
15332 @node S/390 and zSeries Options
15333 @subsection S/390 and zSeries Options
15334 @cindex S/390 and zSeries Options
15336 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
15340 @itemx -msoft-float
15341 @opindex mhard-float
15342 @opindex msoft-float
15343 Use (do not use) the hardware floating-point instructions and registers
15344 for floating-point operations. When @option{-msoft-float} is specified,
15345 functions in @file{libgcc.a} will be used to perform floating-point
15346 operations. When @option{-mhard-float} is specified, the compiler
15347 generates IEEE floating-point instructions. This is the default.
15350 @itemx -mno-hard-dfp
15352 @opindex mno-hard-dfp
15353 Use (do not use) the hardware decimal-floating-point instructions for
15354 decimal-floating-point operations. When @option{-mno-hard-dfp} is
15355 specified, functions in @file{libgcc.a} will be used to perform
15356 decimal-floating-point operations. When @option{-mhard-dfp} is
15357 specified, the compiler generates decimal-floating-point hardware
15358 instructions. This is the default for @option{-march=z9-ec} or higher.
15360 @item -mlong-double-64
15361 @itemx -mlong-double-128
15362 @opindex mlong-double-64
15363 @opindex mlong-double-128
15364 These switches control the size of @code{long double} type. A size
15365 of 64bit makes the @code{long double} type equivalent to the @code{double}
15366 type. This is the default.
15369 @itemx -mno-backchain
15370 @opindex mbackchain
15371 @opindex mno-backchain
15372 Store (do not store) the address of the caller's frame as backchain pointer
15373 into the callee's stack frame.
15374 A backchain may be needed to allow debugging using tools that do not understand
15375 DWARF-2 call frame information.
15376 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
15377 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
15378 the backchain is placed into the topmost word of the 96/160 byte register
15381 In general, code compiled with @option{-mbackchain} is call-compatible with
15382 code compiled with @option{-mmo-backchain}; however, use of the backchain
15383 for debugging purposes usually requires that the whole binary is built with
15384 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
15385 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
15386 to build a linux kernel use @option{-msoft-float}.
15388 The default is to not maintain the backchain.
15390 @item -mpacked-stack
15391 @itemx -mno-packed-stack
15392 @opindex mpacked-stack
15393 @opindex mno-packed-stack
15394 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
15395 specified, the compiler uses the all fields of the 96/160 byte register save
15396 area only for their default purpose; unused fields still take up stack space.
15397 When @option{-mpacked-stack} is specified, register save slots are densely
15398 packed at the top of the register save area; unused space is reused for other
15399 purposes, allowing for more efficient use of the available stack space.
15400 However, when @option{-mbackchain} is also in effect, the topmost word of
15401 the save area is always used to store the backchain, and the return address
15402 register is always saved two words below the backchain.
15404 As long as the stack frame backchain is not used, code generated with
15405 @option{-mpacked-stack} is call-compatible with code generated with
15406 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
15407 S/390 or zSeries generated code that uses the stack frame backchain at run
15408 time, not just for debugging purposes. Such code is not call-compatible
15409 with code compiled with @option{-mpacked-stack}. Also, note that the
15410 combination of @option{-mbackchain},
15411 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
15412 to build a linux kernel use @option{-msoft-float}.
15414 The default is to not use the packed stack layout.
15417 @itemx -mno-small-exec
15418 @opindex msmall-exec
15419 @opindex mno-small-exec
15420 Generate (or do not generate) code using the @code{bras} instruction
15421 to do subroutine calls.
15422 This only works reliably if the total executable size does not
15423 exceed 64k. The default is to use the @code{basr} instruction instead,
15424 which does not have this limitation.
15430 When @option{-m31} is specified, generate code compliant to the
15431 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
15432 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
15433 particular to generate 64-bit instructions. For the @samp{s390}
15434 targets, the default is @option{-m31}, while the @samp{s390x}
15435 targets default to @option{-m64}.
15441 When @option{-mzarch} is specified, generate code using the
15442 instructions available on z/Architecture.
15443 When @option{-mesa} is specified, generate code using the
15444 instructions available on ESA/390. Note that @option{-mesa} is
15445 not possible with @option{-m64}.
15446 When generating code compliant to the GNU/Linux for S/390 ABI,
15447 the default is @option{-mesa}. When generating code compliant
15448 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
15454 Generate (or do not generate) code using the @code{mvcle} instruction
15455 to perform block moves. When @option{-mno-mvcle} is specified,
15456 use a @code{mvc} loop instead. This is the default unless optimizing for
15463 Print (or do not print) additional debug information when compiling.
15464 The default is to not print debug information.
15466 @item -march=@var{cpu-type}
15468 Generate code that will run on @var{cpu-type}, which is the name of a system
15469 representing a certain processor type. Possible values for
15470 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
15471 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
15472 When generating code using the instructions available on z/Architecture,
15473 the default is @option{-march=z900}. Otherwise, the default is
15474 @option{-march=g5}.
15476 @item -mtune=@var{cpu-type}
15478 Tune to @var{cpu-type} everything applicable about the generated code,
15479 except for the ABI and the set of available instructions.
15480 The list of @var{cpu-type} values is the same as for @option{-march}.
15481 The default is the value used for @option{-march}.
15484 @itemx -mno-tpf-trace
15485 @opindex mtpf-trace
15486 @opindex mno-tpf-trace
15487 Generate code that adds (does not add) in TPF OS specific branches to trace
15488 routines in the operating system. This option is off by default, even
15489 when compiling for the TPF OS@.
15492 @itemx -mno-fused-madd
15493 @opindex mfused-madd
15494 @opindex mno-fused-madd
15495 Generate code that uses (does not use) the floating point multiply and
15496 accumulate instructions. These instructions are generated by default if
15497 hardware floating point is used.
15499 @item -mwarn-framesize=@var{framesize}
15500 @opindex mwarn-framesize
15501 Emit a warning if the current function exceeds the given frame size. Because
15502 this is a compile time check it doesn't need to be a real problem when the program
15503 runs. It is intended to identify functions which most probably cause
15504 a stack overflow. It is useful to be used in an environment with limited stack
15505 size e.g.@: the linux kernel.
15507 @item -mwarn-dynamicstack
15508 @opindex mwarn-dynamicstack
15509 Emit a warning if the function calls alloca or uses dynamically
15510 sized arrays. This is generally a bad idea with a limited stack size.
15512 @item -mstack-guard=@var{stack-guard}
15513 @itemx -mstack-size=@var{stack-size}
15514 @opindex mstack-guard
15515 @opindex mstack-size
15516 If these options are provided the s390 back end emits additional instructions in
15517 the function prologue which trigger a trap if the stack size is @var{stack-guard}
15518 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
15519 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
15520 the frame size of the compiled function is chosen.
15521 These options are intended to be used to help debugging stack overflow problems.
15522 The additionally emitted code causes only little overhead and hence can also be
15523 used in production like systems without greater performance degradation. The given
15524 values have to be exact powers of 2 and @var{stack-size} has to be greater than
15525 @var{stack-guard} without exceeding 64k.
15526 In order to be efficient the extra code makes the assumption that the stack starts
15527 at an address aligned to the value given by @var{stack-size}.
15528 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
15531 @node Score Options
15532 @subsection Score Options
15533 @cindex Score Options
15535 These options are defined for Score implementations:
15540 Compile code for big endian mode. This is the default.
15544 Compile code for little endian mode.
15548 Disable generate bcnz instruction.
15552 Enable generate unaligned load and store instruction.
15556 Enable the use of multiply-accumulate instructions. Disabled by default.
15560 Specify the SCORE5 as the target architecture.
15564 Specify the SCORE5U of the target architecture.
15568 Specify the SCORE7 as the target architecture. This is the default.
15572 Specify the SCORE7D as the target architecture.
15576 @subsection SH Options
15578 These @samp{-m} options are defined for the SH implementations:
15583 Generate code for the SH1.
15587 Generate code for the SH2.
15590 Generate code for the SH2e.
15594 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
15595 that the floating-point unit is not used.
15597 @item -m2a-single-only
15598 @opindex m2a-single-only
15599 Generate code for the SH2a-FPU, in such a way that no double-precision
15600 floating point operations are used.
15603 @opindex m2a-single
15604 Generate code for the SH2a-FPU assuming the floating-point unit is in
15605 single-precision mode by default.
15609 Generate code for the SH2a-FPU assuming the floating-point unit is in
15610 double-precision mode by default.
15614 Generate code for the SH3.
15618 Generate code for the SH3e.
15622 Generate code for the SH4 without a floating-point unit.
15624 @item -m4-single-only
15625 @opindex m4-single-only
15626 Generate code for the SH4 with a floating-point unit that only
15627 supports single-precision arithmetic.
15631 Generate code for the SH4 assuming the floating-point unit is in
15632 single-precision mode by default.
15636 Generate code for the SH4.
15640 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
15641 floating-point unit is not used.
15643 @item -m4a-single-only
15644 @opindex m4a-single-only
15645 Generate code for the SH4a, in such a way that no double-precision
15646 floating point operations are used.
15649 @opindex m4a-single
15650 Generate code for the SH4a assuming the floating-point unit is in
15651 single-precision mode by default.
15655 Generate code for the SH4a.
15659 Same as @option{-m4a-nofpu}, except that it implicitly passes
15660 @option{-dsp} to the assembler. GCC doesn't generate any DSP
15661 instructions at the moment.
15665 Compile code for the processor in big endian mode.
15669 Compile code for the processor in little endian mode.
15673 Align doubles at 64-bit boundaries. Note that this changes the calling
15674 conventions, and thus some functions from the standard C library will
15675 not work unless you recompile it first with @option{-mdalign}.
15679 Shorten some address references at link time, when possible; uses the
15680 linker option @option{-relax}.
15684 Use 32-bit offsets in @code{switch} tables. The default is to use
15689 Enable the use of bit manipulation instructions on SH2A.
15693 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
15694 alignment constraints.
15698 Comply with the calling conventions defined by Renesas.
15702 Comply with the calling conventions defined by Renesas.
15706 Comply with the calling conventions defined for GCC before the Renesas
15707 conventions were available. This option is the default for all
15708 targets of the SH toolchain except for @samp{sh-symbianelf}.
15711 @opindex mnomacsave
15712 Mark the @code{MAC} register as call-clobbered, even if
15713 @option{-mhitachi} is given.
15717 Increase IEEE-compliance of floating-point code.
15718 At the moment, this is equivalent to @option{-fno-finite-math-only}.
15719 When generating 16 bit SH opcodes, getting IEEE-conforming results for
15720 comparisons of NANs / infinities incurs extra overhead in every
15721 floating point comparison, therefore the default is set to
15722 @option{-ffinite-math-only}.
15724 @item -minline-ic_invalidate
15725 @opindex minline-ic_invalidate
15726 Inline code to invalidate instruction cache entries after setting up
15727 nested function trampolines.
15728 This option has no effect if -musermode is in effect and the selected
15729 code generation option (e.g. -m4) does not allow the use of the icbi
15731 If the selected code generation option does not allow the use of the icbi
15732 instruction, and -musermode is not in effect, the inlined code will
15733 manipulate the instruction cache address array directly with an associative
15734 write. This not only requires privileged mode, but it will also
15735 fail if the cache line had been mapped via the TLB and has become unmapped.
15739 Dump instruction size and location in the assembly code.
15742 @opindex mpadstruct
15743 This option is deprecated. It pads structures to multiple of 4 bytes,
15744 which is incompatible with the SH ABI@.
15748 Optimize for space instead of speed. Implied by @option{-Os}.
15751 @opindex mprefergot
15752 When generating position-independent code, emit function calls using
15753 the Global Offset Table instead of the Procedure Linkage Table.
15757 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
15758 if the inlined code would not work in user mode.
15759 This is the default when the target is @code{sh-*-linux*}.
15761 @item -multcost=@var{number}
15762 @opindex multcost=@var{number}
15763 Set the cost to assume for a multiply insn.
15765 @item -mdiv=@var{strategy}
15766 @opindex mdiv=@var{strategy}
15767 Set the division strategy to use for SHmedia code. @var{strategy} must be
15768 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
15769 inv:call2, inv:fp .
15770 "fp" performs the operation in floating point. This has a very high latency,
15771 but needs only a few instructions, so it might be a good choice if
15772 your code has enough easily exploitable ILP to allow the compiler to
15773 schedule the floating point instructions together with other instructions.
15774 Division by zero causes a floating point exception.
15775 "inv" uses integer operations to calculate the inverse of the divisor,
15776 and then multiplies the dividend with the inverse. This strategy allows
15777 cse and hoisting of the inverse calculation. Division by zero calculates
15778 an unspecified result, but does not trap.
15779 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
15780 have been found, or if the entire operation has been hoisted to the same
15781 place, the last stages of the inverse calculation are intertwined with the
15782 final multiply to reduce the overall latency, at the expense of using a few
15783 more instructions, and thus offering fewer scheduling opportunities with
15785 "call" calls a library function that usually implements the inv:minlat
15787 This gives high code density for m5-*media-nofpu compilations.
15788 "call2" uses a different entry point of the same library function, where it
15789 assumes that a pointer to a lookup table has already been set up, which
15790 exposes the pointer load to cse / code hoisting optimizations.
15791 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
15792 code generation, but if the code stays unoptimized, revert to the "call",
15793 "call2", or "fp" strategies, respectively. Note that the
15794 potentially-trapping side effect of division by zero is carried by a
15795 separate instruction, so it is possible that all the integer instructions
15796 are hoisted out, but the marker for the side effect stays where it is.
15797 A recombination to fp operations or a call is not possible in that case.
15798 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
15799 that the inverse calculation was nor separated from the multiply, they speed
15800 up division where the dividend fits into 20 bits (plus sign where applicable),
15801 by inserting a test to skip a number of operations in this case; this test
15802 slows down the case of larger dividends. inv20u assumes the case of a such
15803 a small dividend to be unlikely, and inv20l assumes it to be likely.
15805 @item -mdivsi3_libfunc=@var{name}
15806 @opindex mdivsi3_libfunc=@var{name}
15807 Set the name of the library function used for 32 bit signed division to
15808 @var{name}. This only affect the name used in the call and inv:call
15809 division strategies, and the compiler will still expect the same
15810 sets of input/output/clobbered registers as if this option was not present.
15812 @item -mfixed-range=@var{register-range}
15813 @opindex mfixed-range
15814 Generate code treating the given register range as fixed registers.
15815 A fixed register is one that the register allocator can not use. This is
15816 useful when compiling kernel code. A register range is specified as
15817 two registers separated by a dash. Multiple register ranges can be
15818 specified separated by a comma.
15820 @item -madjust-unroll
15821 @opindex madjust-unroll
15822 Throttle unrolling to avoid thrashing target registers.
15823 This option only has an effect if the gcc code base supports the
15824 TARGET_ADJUST_UNROLL_MAX target hook.
15826 @item -mindexed-addressing
15827 @opindex mindexed-addressing
15828 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
15829 This is only safe if the hardware and/or OS implement 32 bit wrap-around
15830 semantics for the indexed addressing mode. The architecture allows the
15831 implementation of processors with 64 bit MMU, which the OS could use to
15832 get 32 bit addressing, but since no current hardware implementation supports
15833 this or any other way to make the indexed addressing mode safe to use in
15834 the 32 bit ABI, the default is -mno-indexed-addressing.
15836 @item -mgettrcost=@var{number}
15837 @opindex mgettrcost=@var{number}
15838 Set the cost assumed for the gettr instruction to @var{number}.
15839 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
15843 Assume pt* instructions won't trap. This will generally generate better
15844 scheduled code, but is unsafe on current hardware. The current architecture
15845 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
15846 This has the unintentional effect of making it unsafe to schedule ptabs /
15847 ptrel before a branch, or hoist it out of a loop. For example,
15848 __do_global_ctors, a part of libgcc that runs constructors at program
15849 startup, calls functions in a list which is delimited by @minus{}1. With the
15850 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
15851 That means that all the constructors will be run a bit quicker, but when
15852 the loop comes to the end of the list, the program crashes because ptabs
15853 loads @minus{}1 into a target register. Since this option is unsafe for any
15854 hardware implementing the current architecture specification, the default
15855 is -mno-pt-fixed. Unless the user specifies a specific cost with
15856 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
15857 this deters register allocation using target registers for storing
15860 @item -minvalid-symbols
15861 @opindex minvalid-symbols
15862 Assume symbols might be invalid. Ordinary function symbols generated by
15863 the compiler will always be valid to load with movi/shori/ptabs or
15864 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
15865 to generate symbols that will cause ptabs / ptrel to trap.
15866 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
15867 It will then prevent cross-basic-block cse, hoisting and most scheduling
15868 of symbol loads. The default is @option{-mno-invalid-symbols}.
15871 @node SPARC Options
15872 @subsection SPARC Options
15873 @cindex SPARC options
15875 These @samp{-m} options are supported on the SPARC:
15878 @item -mno-app-regs
15880 @opindex mno-app-regs
15882 Specify @option{-mapp-regs} to generate output using the global registers
15883 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
15886 To be fully SVR4 ABI compliant at the cost of some performance loss,
15887 specify @option{-mno-app-regs}. You should compile libraries and system
15888 software with this option.
15891 @itemx -mhard-float
15893 @opindex mhard-float
15894 Generate output containing floating point instructions. This is the
15898 @itemx -msoft-float
15900 @opindex msoft-float
15901 Generate output containing library calls for floating point.
15902 @strong{Warning:} the requisite libraries are not available for all SPARC
15903 targets. Normally the facilities of the machine's usual C compiler are
15904 used, but this cannot be done directly in cross-compilation. You must make
15905 your own arrangements to provide suitable library functions for
15906 cross-compilation. The embedded targets @samp{sparc-*-aout} and
15907 @samp{sparclite-*-*} do provide software floating point support.
15909 @option{-msoft-float} changes the calling convention in the output file;
15910 therefore, it is only useful if you compile @emph{all} of a program with
15911 this option. In particular, you need to compile @file{libgcc.a}, the
15912 library that comes with GCC, with @option{-msoft-float} in order for
15915 @item -mhard-quad-float
15916 @opindex mhard-quad-float
15917 Generate output containing quad-word (long double) floating point
15920 @item -msoft-quad-float
15921 @opindex msoft-quad-float
15922 Generate output containing library calls for quad-word (long double)
15923 floating point instructions. The functions called are those specified
15924 in the SPARC ABI@. This is the default.
15926 As of this writing, there are no SPARC implementations that have hardware
15927 support for the quad-word floating point instructions. They all invoke
15928 a trap handler for one of these instructions, and then the trap handler
15929 emulates the effect of the instruction. Because of the trap handler overhead,
15930 this is much slower than calling the ABI library routines. Thus the
15931 @option{-msoft-quad-float} option is the default.
15933 @item -mno-unaligned-doubles
15934 @itemx -munaligned-doubles
15935 @opindex mno-unaligned-doubles
15936 @opindex munaligned-doubles
15937 Assume that doubles have 8 byte alignment. This is the default.
15939 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
15940 alignment only if they are contained in another type, or if they have an
15941 absolute address. Otherwise, it assumes they have 4 byte alignment.
15942 Specifying this option avoids some rare compatibility problems with code
15943 generated by other compilers. It is not the default because it results
15944 in a performance loss, especially for floating point code.
15946 @item -mno-faster-structs
15947 @itemx -mfaster-structs
15948 @opindex mno-faster-structs
15949 @opindex mfaster-structs
15950 With @option{-mfaster-structs}, the compiler assumes that structures
15951 should have 8 byte alignment. This enables the use of pairs of
15952 @code{ldd} and @code{std} instructions for copies in structure
15953 assignment, in place of twice as many @code{ld} and @code{st} pairs.
15954 However, the use of this changed alignment directly violates the SPARC
15955 ABI@. Thus, it's intended only for use on targets where the developer
15956 acknowledges that their resulting code will not be directly in line with
15957 the rules of the ABI@.
15959 @item -mimpure-text
15960 @opindex mimpure-text
15961 @option{-mimpure-text}, used in addition to @option{-shared}, tells
15962 the compiler to not pass @option{-z text} to the linker when linking a
15963 shared object. Using this option, you can link position-dependent
15964 code into a shared object.
15966 @option{-mimpure-text} suppresses the ``relocations remain against
15967 allocatable but non-writable sections'' linker error message.
15968 However, the necessary relocations will trigger copy-on-write, and the
15969 shared object is not actually shared across processes. Instead of
15970 using @option{-mimpure-text}, you should compile all source code with
15971 @option{-fpic} or @option{-fPIC}.
15973 This option is only available on SunOS and Solaris.
15975 @item -mcpu=@var{cpu_type}
15977 Set the instruction set, register set, and instruction scheduling parameters
15978 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
15979 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
15980 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
15981 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
15982 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
15984 Default instruction scheduling parameters are used for values that select
15985 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
15986 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
15988 Here is a list of each supported architecture and their supported
15993 v8: supersparc, hypersparc
15994 sparclite: f930, f934, sparclite86x
15996 v9: ultrasparc, ultrasparc3, niagara, niagara2
15999 By default (unless configured otherwise), GCC generates code for the V7
16000 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
16001 additionally optimizes it for the Cypress CY7C602 chip, as used in the
16002 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
16003 SPARCStation 1, 2, IPX etc.
16005 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
16006 architecture. The only difference from V7 code is that the compiler emits
16007 the integer multiply and integer divide instructions which exist in SPARC-V8
16008 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
16009 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
16012 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
16013 the SPARC architecture. This adds the integer multiply, integer divide step
16014 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
16015 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
16016 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
16017 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
16018 MB86934 chip, which is the more recent SPARClite with FPU@.
16020 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
16021 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
16022 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
16023 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
16024 optimizes it for the TEMIC SPARClet chip.
16026 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
16027 architecture. This adds 64-bit integer and floating-point move instructions,
16028 3 additional floating-point condition code registers and conditional move
16029 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
16030 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
16031 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
16032 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
16033 @option{-mcpu=niagara}, the compiler additionally optimizes it for
16034 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
16035 additionally optimizes it for Sun UltraSPARC T2 chips.
16037 @item -mtune=@var{cpu_type}
16039 Set the instruction scheduling parameters for machine type
16040 @var{cpu_type}, but do not set the instruction set or register set that the
16041 option @option{-mcpu=@var{cpu_type}} would.
16043 The same values for @option{-mcpu=@var{cpu_type}} can be used for
16044 @option{-mtune=@var{cpu_type}}, but the only useful values are those
16045 that select a particular cpu implementation. Those are @samp{cypress},
16046 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
16047 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
16048 @samp{ultrasparc3}, @samp{niagara}, and @samp{niagara2}.
16053 @opindex mno-v8plus
16054 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
16055 difference from the V8 ABI is that the global and out registers are
16056 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
16057 mode for all SPARC-V9 processors.
16063 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
16064 Visual Instruction Set extensions. The default is @option{-mno-vis}.
16067 These @samp{-m} options are supported in addition to the above
16068 on SPARC-V9 processors in 64-bit environments:
16071 @item -mlittle-endian
16072 @opindex mlittle-endian
16073 Generate code for a processor running in little-endian mode. It is only
16074 available for a few configurations and most notably not on Solaris and Linux.
16080 Generate code for a 32-bit or 64-bit environment.
16081 The 32-bit environment sets int, long and pointer to 32 bits.
16082 The 64-bit environment sets int to 32 bits and long and pointer
16085 @item -mcmodel=medlow
16086 @opindex mcmodel=medlow
16087 Generate code for the Medium/Low code model: 64-bit addresses, programs
16088 must be linked in the low 32 bits of memory. Programs can be statically
16089 or dynamically linked.
16091 @item -mcmodel=medmid
16092 @opindex mcmodel=medmid
16093 Generate code for the Medium/Middle code model: 64-bit addresses, programs
16094 must be linked in the low 44 bits of memory, the text and data segments must
16095 be less than 2GB in size and the data segment must be located within 2GB of
16098 @item -mcmodel=medany
16099 @opindex mcmodel=medany
16100 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
16101 may be linked anywhere in memory, the text and data segments must be less
16102 than 2GB in size and the data segment must be located within 2GB of the
16105 @item -mcmodel=embmedany
16106 @opindex mcmodel=embmedany
16107 Generate code for the Medium/Anywhere code model for embedded systems:
16108 64-bit addresses, the text and data segments must be less than 2GB in
16109 size, both starting anywhere in memory (determined at link time). The
16110 global register %g4 points to the base of the data segment. Programs
16111 are statically linked and PIC is not supported.
16114 @itemx -mno-stack-bias
16115 @opindex mstack-bias
16116 @opindex mno-stack-bias
16117 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
16118 frame pointer if present, are offset by @minus{}2047 which must be added back
16119 when making stack frame references. This is the default in 64-bit mode.
16120 Otherwise, assume no such offset is present.
16123 These switches are supported in addition to the above on Solaris:
16128 Add support for multithreading using the Solaris threads library. This
16129 option sets flags for both the preprocessor and linker. This option does
16130 not affect the thread safety of object code produced by the compiler or
16131 that of libraries supplied with it.
16135 Add support for multithreading using the POSIX threads library. This
16136 option sets flags for both the preprocessor and linker. This option does
16137 not affect the thread safety of object code produced by the compiler or
16138 that of libraries supplied with it.
16142 This is a synonym for @option{-pthreads}.
16146 @subsection SPU Options
16147 @cindex SPU options
16149 These @samp{-m} options are supported on the SPU:
16153 @itemx -merror-reloc
16154 @opindex mwarn-reloc
16155 @opindex merror-reloc
16157 The loader for SPU does not handle dynamic relocations. By default, GCC
16158 will give an error when it generates code that requires a dynamic
16159 relocation. @option{-mno-error-reloc} disables the error,
16160 @option{-mwarn-reloc} will generate a warning instead.
16163 @itemx -munsafe-dma
16165 @opindex munsafe-dma
16167 Instructions which initiate or test completion of DMA must not be
16168 reordered with respect to loads and stores of the memory which is being
16169 accessed. Users typically address this problem using the volatile
16170 keyword, but that can lead to inefficient code in places where the
16171 memory is known to not change. Rather than mark the memory as volatile
16172 we treat the DMA instructions as potentially effecting all memory. With
16173 @option{-munsafe-dma} users must use the volatile keyword to protect
16176 @item -mbranch-hints
16177 @opindex mbranch-hints
16179 By default, GCC will generate a branch hint instruction to avoid
16180 pipeline stalls for always taken or probably taken branches. A hint
16181 will not be generated closer than 8 instructions away from its branch.
16182 There is little reason to disable them, except for debugging purposes,
16183 or to make an object a little bit smaller.
16187 @opindex msmall-mem
16188 @opindex mlarge-mem
16190 By default, GCC generates code assuming that addresses are never larger
16191 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
16192 a full 32 bit address.
16197 By default, GCC links against startup code that assumes the SPU-style
16198 main function interface (which has an unconventional parameter list).
16199 With @option{-mstdmain}, GCC will link your program against startup
16200 code that assumes a C99-style interface to @code{main}, including a
16201 local copy of @code{argv} strings.
16203 @item -mfixed-range=@var{register-range}
16204 @opindex mfixed-range
16205 Generate code treating the given register range as fixed registers.
16206 A fixed register is one that the register allocator can not use. This is
16207 useful when compiling kernel code. A register range is specified as
16208 two registers separated by a dash. Multiple register ranges can be
16209 specified separated by a comma.
16212 @itemx -mdual-nops=@var{n}
16213 @opindex mdual-nops
16214 By default, GCC will insert nops to increase dual issue when it expects
16215 it to increase performance. @var{n} can be a value from 0 to 10. A
16216 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
16217 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
16219 @item -mhint-max-nops=@var{n}
16220 @opindex mhint-max-nops
16221 Maximum number of nops to insert for a branch hint. A branch hint must
16222 be at least 8 instructions away from the branch it is effecting. GCC
16223 will insert up to @var{n} nops to enforce this, otherwise it will not
16224 generate the branch hint.
16226 @item -mhint-max-distance=@var{n}
16227 @opindex mhint-max-distance
16228 The encoding of the branch hint instruction limits the hint to be within
16229 256 instructions of the branch it is effecting. By default, GCC makes
16230 sure it is within 125.
16233 @opindex msafe-hints
16234 Work around a hardware bug which causes the SPU to stall indefinitely.
16235 By default, GCC will insert the @code{hbrp} instruction to make sure
16236 this stall won't happen.
16240 @node System V Options
16241 @subsection Options for System V
16243 These additional options are available on System V Release 4 for
16244 compatibility with other compilers on those systems:
16249 Create a shared object.
16250 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
16254 Identify the versions of each tool used by the compiler, in a
16255 @code{.ident} assembler directive in the output.
16259 Refrain from adding @code{.ident} directives to the output file (this is
16262 @item -YP,@var{dirs}
16264 Search the directories @var{dirs}, and no others, for libraries
16265 specified with @option{-l}.
16267 @item -Ym,@var{dir}
16269 Look in the directory @var{dir} to find the M4 preprocessor.
16270 The assembler uses this option.
16271 @c This is supposed to go with a -Yd for predefined M4 macro files, but
16272 @c the generic assembler that comes with Solaris takes just -Ym.
16276 @subsection V850 Options
16277 @cindex V850 Options
16279 These @samp{-m} options are defined for V850 implementations:
16283 @itemx -mno-long-calls
16284 @opindex mlong-calls
16285 @opindex mno-long-calls
16286 Treat all calls as being far away (near). If calls are assumed to be
16287 far away, the compiler will always load the functions address up into a
16288 register, and call indirect through the pointer.
16294 Do not optimize (do optimize) basic blocks that use the same index
16295 pointer 4 or more times to copy pointer into the @code{ep} register, and
16296 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
16297 option is on by default if you optimize.
16299 @item -mno-prolog-function
16300 @itemx -mprolog-function
16301 @opindex mno-prolog-function
16302 @opindex mprolog-function
16303 Do not use (do use) external functions to save and restore registers
16304 at the prologue and epilogue of a function. The external functions
16305 are slower, but use less code space if more than one function saves
16306 the same number of registers. The @option{-mprolog-function} option
16307 is on by default if you optimize.
16311 Try to make the code as small as possible. At present, this just turns
16312 on the @option{-mep} and @option{-mprolog-function} options.
16314 @item -mtda=@var{n}
16316 Put static or global variables whose size is @var{n} bytes or less into
16317 the tiny data area that register @code{ep} points to. The tiny data
16318 area can hold up to 256 bytes in total (128 bytes for byte references).
16320 @item -msda=@var{n}
16322 Put static or global variables whose size is @var{n} bytes or less into
16323 the small data area that register @code{gp} points to. The small data
16324 area can hold up to 64 kilobytes.
16326 @item -mzda=@var{n}
16328 Put static or global variables whose size is @var{n} bytes or less into
16329 the first 32 kilobytes of memory.
16333 Specify that the target processor is the V850.
16336 @opindex mbig-switch
16337 Generate code suitable for big switch tables. Use this option only if
16338 the assembler/linker complain about out of range branches within a switch
16343 This option will cause r2 and r5 to be used in the code generated by
16344 the compiler. This setting is the default.
16346 @item -mno-app-regs
16347 @opindex mno-app-regs
16348 This option will cause r2 and r5 to be treated as fixed registers.
16352 Specify that the target processor is the V850E1. The preprocessor
16353 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
16354 this option is used.
16358 Specify that the target processor is the V850E@. The preprocessor
16359 constant @samp{__v850e__} will be defined if this option is used.
16361 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
16362 are defined then a default target processor will be chosen and the
16363 relevant @samp{__v850*__} preprocessor constant will be defined.
16365 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
16366 defined, regardless of which processor variant is the target.
16368 @item -mdisable-callt
16369 @opindex mdisable-callt
16370 This option will suppress generation of the CALLT instruction for the
16371 v850e and v850e1 flavors of the v850 architecture. The default is
16372 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
16377 @subsection VAX Options
16378 @cindex VAX options
16380 These @samp{-m} options are defined for the VAX:
16385 Do not output certain jump instructions (@code{aobleq} and so on)
16386 that the Unix assembler for the VAX cannot handle across long
16391 Do output those jump instructions, on the assumption that you
16392 will assemble with the GNU assembler.
16396 Output code for g-format floating point numbers instead of d-format.
16399 @node VxWorks Options
16400 @subsection VxWorks Options
16401 @cindex VxWorks Options
16403 The options in this section are defined for all VxWorks targets.
16404 Options specific to the target hardware are listed with the other
16405 options for that target.
16410 GCC can generate code for both VxWorks kernels and real time processes
16411 (RTPs). This option switches from the former to the latter. It also
16412 defines the preprocessor macro @code{__RTP__}.
16415 @opindex non-static
16416 Link an RTP executable against shared libraries rather than static
16417 libraries. The options @option{-static} and @option{-shared} can
16418 also be used for RTPs (@pxref{Link Options}); @option{-static}
16425 These options are passed down to the linker. They are defined for
16426 compatibility with Diab.
16429 @opindex Xbind-lazy
16430 Enable lazy binding of function calls. This option is equivalent to
16431 @option{-Wl,-z,now} and is defined for compatibility with Diab.
16435 Disable lazy binding of function calls. This option is the default and
16436 is defined for compatibility with Diab.
16439 @node x86-64 Options
16440 @subsection x86-64 Options
16441 @cindex x86-64 options
16443 These are listed under @xref{i386 and x86-64 Options}.
16445 @node i386 and x86-64 Windows Options
16446 @subsection i386 and x86-64 Windows Options
16447 @cindex i386 and x86-64 Windows Options
16449 These additional options are available for Windows targets:
16454 This option is available for Cygwin and MinGW targets. It
16455 specifies that a console application is to be generated, by
16456 instructing the linker to set the PE header subsystem type
16457 required for console applications.
16458 This is the default behavior for Cygwin and MinGW targets.
16462 This option is available for Cygwin targets. It specifies that
16463 the Cygwin internal interface is to be used for predefined
16464 preprocessor macros, C runtime libraries and related linker
16465 paths and options. For Cygwin targets this is the default behavior.
16466 This option is deprecated and will be removed in a future release.
16469 @opindex mno-cygwin
16470 This option is available for Cygwin targets. It specifies that
16471 the MinGW internal interface is to be used instead of Cygwin's, by
16472 setting MinGW-related predefined macros and linker paths and default
16474 This option is deprecated and will be removed in a future release.
16478 This option is available for Cygwin and MinGW targets. It
16479 specifies that a DLL - a dynamic link library - is to be
16480 generated, enabling the selection of the required runtime
16481 startup object and entry point.
16483 @item -mnop-fun-dllimport
16484 @opindex mnop-fun-dllimport
16485 This option is available for Cygwin and MinGW targets. It
16486 specifies that the dllimport attribute should be ignored.
16490 This option is available for MinGW targets. It specifies
16491 that MinGW-specific thread support is to be used.
16495 This option is available for mingw-w64 targets. It specifies
16496 that the UNICODE macro is getting pre-defined and that the
16497 unicode capable runtime startup code is choosen.
16501 This option is available for Cygwin and MinGW targets. It
16502 specifies that the typical Windows pre-defined macros are to
16503 be set in the pre-processor, but does not influence the choice
16504 of runtime library/startup code.
16508 This option is available for Cygwin and MinGW targets. It
16509 specifies that a GUI application is to be generated by
16510 instructing the linker to set the PE header subsystem type
16513 @item -mpe-aligned-commons
16514 @opindex mpe-aligned-commons
16515 This option is available for Cygwin and MinGW targets. It
16516 specifies that the GNU extension to the PE file format that
16517 permits the correct alignment of COMMON variables should be
16518 used when generating code. It will be enabled by default if
16519 GCC detects that the target assembler found during configuration
16520 supports the feature.
16523 See also under @ref{i386 and x86-64 Options} for standard options.
16525 @node Xstormy16 Options
16526 @subsection Xstormy16 Options
16527 @cindex Xstormy16 Options
16529 These options are defined for Xstormy16:
16534 Choose startup files and linker script suitable for the simulator.
16537 @node Xtensa Options
16538 @subsection Xtensa Options
16539 @cindex Xtensa Options
16541 These options are supported for Xtensa targets:
16545 @itemx -mno-const16
16547 @opindex mno-const16
16548 Enable or disable use of @code{CONST16} instructions for loading
16549 constant values. The @code{CONST16} instruction is currently not a
16550 standard option from Tensilica. When enabled, @code{CONST16}
16551 instructions are always used in place of the standard @code{L32R}
16552 instructions. The use of @code{CONST16} is enabled by default only if
16553 the @code{L32R} instruction is not available.
16556 @itemx -mno-fused-madd
16557 @opindex mfused-madd
16558 @opindex mno-fused-madd
16559 Enable or disable use of fused multiply/add and multiply/subtract
16560 instructions in the floating-point option. This has no effect if the
16561 floating-point option is not also enabled. Disabling fused multiply/add
16562 and multiply/subtract instructions forces the compiler to use separate
16563 instructions for the multiply and add/subtract operations. This may be
16564 desirable in some cases where strict IEEE 754-compliant results are
16565 required: the fused multiply add/subtract instructions do not round the
16566 intermediate result, thereby producing results with @emph{more} bits of
16567 precision than specified by the IEEE standard. Disabling fused multiply
16568 add/subtract instructions also ensures that the program output is not
16569 sensitive to the compiler's ability to combine multiply and add/subtract
16572 @item -mserialize-volatile
16573 @itemx -mno-serialize-volatile
16574 @opindex mserialize-volatile
16575 @opindex mno-serialize-volatile
16576 When this option is enabled, GCC inserts @code{MEMW} instructions before
16577 @code{volatile} memory references to guarantee sequential consistency.
16578 The default is @option{-mserialize-volatile}. Use
16579 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
16581 @item -mtext-section-literals
16582 @itemx -mno-text-section-literals
16583 @opindex mtext-section-literals
16584 @opindex mno-text-section-literals
16585 Control the treatment of literal pools. The default is
16586 @option{-mno-text-section-literals}, which places literals in a separate
16587 section in the output file. This allows the literal pool to be placed
16588 in a data RAM/ROM, and it also allows the linker to combine literal
16589 pools from separate object files to remove redundant literals and
16590 improve code size. With @option{-mtext-section-literals}, the literals
16591 are interspersed in the text section in order to keep them as close as
16592 possible to their references. This may be necessary for large assembly
16595 @item -mtarget-align
16596 @itemx -mno-target-align
16597 @opindex mtarget-align
16598 @opindex mno-target-align
16599 When this option is enabled, GCC instructs the assembler to
16600 automatically align instructions to reduce branch penalties at the
16601 expense of some code density. The assembler attempts to widen density
16602 instructions to align branch targets and the instructions following call
16603 instructions. If there are not enough preceding safe density
16604 instructions to align a target, no widening will be performed. The
16605 default is @option{-mtarget-align}. These options do not affect the
16606 treatment of auto-aligned instructions like @code{LOOP}, which the
16607 assembler will always align, either by widening density instructions or
16608 by inserting no-op instructions.
16611 @itemx -mno-longcalls
16612 @opindex mlongcalls
16613 @opindex mno-longcalls
16614 When this option is enabled, GCC instructs the assembler to translate
16615 direct calls to indirect calls unless it can determine that the target
16616 of a direct call is in the range allowed by the call instruction. This
16617 translation typically occurs for calls to functions in other source
16618 files. Specifically, the assembler translates a direct @code{CALL}
16619 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
16620 The default is @option{-mno-longcalls}. This option should be used in
16621 programs where the call target can potentially be out of range. This
16622 option is implemented in the assembler, not the compiler, so the
16623 assembly code generated by GCC will still show direct call
16624 instructions---look at the disassembled object code to see the actual
16625 instructions. Note that the assembler will use an indirect call for
16626 every cross-file call, not just those that really will be out of range.
16629 @node zSeries Options
16630 @subsection zSeries Options
16631 @cindex zSeries options
16633 These are listed under @xref{S/390 and zSeries Options}.
16635 @node Code Gen Options
16636 @section Options for Code Generation Conventions
16637 @cindex code generation conventions
16638 @cindex options, code generation
16639 @cindex run-time options
16641 These machine-independent options control the interface conventions
16642 used in code generation.
16644 Most of them have both positive and negative forms; the negative form
16645 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
16646 one of the forms is listed---the one which is not the default. You
16647 can figure out the other form by either removing @samp{no-} or adding
16651 @item -fbounds-check
16652 @opindex fbounds-check
16653 For front-ends that support it, generate additional code to check that
16654 indices used to access arrays are within the declared range. This is
16655 currently only supported by the Java and Fortran front-ends, where
16656 this option defaults to true and false respectively.
16660 This option generates traps for signed overflow on addition, subtraction,
16661 multiplication operations.
16665 This option instructs the compiler to assume that signed arithmetic
16666 overflow of addition, subtraction and multiplication wraps around
16667 using twos-complement representation. This flag enables some optimizations
16668 and disables others. This option is enabled by default for the Java
16669 front-end, as required by the Java language specification.
16672 @opindex fexceptions
16673 Enable exception handling. Generates extra code needed to propagate
16674 exceptions. For some targets, this implies GCC will generate frame
16675 unwind information for all functions, which can produce significant data
16676 size overhead, although it does not affect execution. If you do not
16677 specify this option, GCC will enable it by default for languages like
16678 C++ which normally require exception handling, and disable it for
16679 languages like C that do not normally require it. However, you may need
16680 to enable this option when compiling C code that needs to interoperate
16681 properly with exception handlers written in C++. You may also wish to
16682 disable this option if you are compiling older C++ programs that don't
16683 use exception handling.
16685 @item -fnon-call-exceptions
16686 @opindex fnon-call-exceptions
16687 Generate code that allows trapping instructions to throw exceptions.
16688 Note that this requires platform-specific runtime support that does
16689 not exist everywhere. Moreover, it only allows @emph{trapping}
16690 instructions to throw exceptions, i.e.@: memory references or floating
16691 point instructions. It does not allow exceptions to be thrown from
16692 arbitrary signal handlers such as @code{SIGALRM}.
16694 @item -funwind-tables
16695 @opindex funwind-tables
16696 Similar to @option{-fexceptions}, except that it will just generate any needed
16697 static data, but will not affect the generated code in any other way.
16698 You will normally not enable this option; instead, a language processor
16699 that needs this handling would enable it on your behalf.
16701 @item -fasynchronous-unwind-tables
16702 @opindex fasynchronous-unwind-tables
16703 Generate unwind table in dwarf2 format, if supported by target machine. The
16704 table is exact at each instruction boundary, so it can be used for stack
16705 unwinding from asynchronous events (such as debugger or garbage collector).
16707 @item -fpcc-struct-return
16708 @opindex fpcc-struct-return
16709 Return ``short'' @code{struct} and @code{union} values in memory like
16710 longer ones, rather than in registers. This convention is less
16711 efficient, but it has the advantage of allowing intercallability between
16712 GCC-compiled files and files compiled with other compilers, particularly
16713 the Portable C Compiler (pcc).
16715 The precise convention for returning structures in memory depends
16716 on the target configuration macros.
16718 Short structures and unions are those whose size and alignment match
16719 that of some integer type.
16721 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
16722 switch is not binary compatible with code compiled with the
16723 @option{-freg-struct-return} switch.
16724 Use it to conform to a non-default application binary interface.
16726 @item -freg-struct-return
16727 @opindex freg-struct-return
16728 Return @code{struct} and @code{union} values in registers when possible.
16729 This is more efficient for small structures than
16730 @option{-fpcc-struct-return}.
16732 If you specify neither @option{-fpcc-struct-return} nor
16733 @option{-freg-struct-return}, GCC defaults to whichever convention is
16734 standard for the target. If there is no standard convention, GCC
16735 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
16736 the principal compiler. In those cases, we can choose the standard, and
16737 we chose the more efficient register return alternative.
16739 @strong{Warning:} code compiled with the @option{-freg-struct-return}
16740 switch is not binary compatible with code compiled with the
16741 @option{-fpcc-struct-return} switch.
16742 Use it to conform to a non-default application binary interface.
16744 @item -fshort-enums
16745 @opindex fshort-enums
16746 Allocate to an @code{enum} type only as many bytes as it needs for the
16747 declared range of possible values. Specifically, the @code{enum} type
16748 will be equivalent to the smallest integer type which has enough room.
16750 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
16751 code that is not binary compatible with code generated without that switch.
16752 Use it to conform to a non-default application binary interface.
16754 @item -fshort-double
16755 @opindex fshort-double
16756 Use the same size for @code{double} as for @code{float}.
16758 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
16759 code that is not binary compatible with code generated without that switch.
16760 Use it to conform to a non-default application binary interface.
16762 @item -fshort-wchar
16763 @opindex fshort-wchar
16764 Override the underlying type for @samp{wchar_t} to be @samp{short
16765 unsigned int} instead of the default for the target. This option is
16766 useful for building programs to run under WINE@.
16768 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
16769 code that is not binary compatible with code generated without that switch.
16770 Use it to conform to a non-default application binary interface.
16773 @opindex fno-common
16774 In C code, controls the placement of uninitialized global variables.
16775 Unix C compilers have traditionally permitted multiple definitions of
16776 such variables in different compilation units by placing the variables
16778 This is the behavior specified by @option{-fcommon}, and is the default
16779 for GCC on most targets.
16780 On the other hand, this behavior is not required by ISO C, and on some
16781 targets may carry a speed or code size penalty on variable references.
16782 The @option{-fno-common} option specifies that the compiler should place
16783 uninitialized global variables in the data section of the object file,
16784 rather than generating them as common blocks.
16785 This has the effect that if the same variable is declared
16786 (without @code{extern}) in two different compilations,
16787 you will get a multiple-definition error when you link them.
16788 In this case, you must compile with @option{-fcommon} instead.
16789 Compiling with @option{-fno-common} is useful on targets for which
16790 it provides better performance, or if you wish to verify that the
16791 program will work on other systems which always treat uninitialized
16792 variable declarations this way.
16796 Ignore the @samp{#ident} directive.
16798 @item -finhibit-size-directive
16799 @opindex finhibit-size-directive
16800 Don't output a @code{.size} assembler directive, or anything else that
16801 would cause trouble if the function is split in the middle, and the
16802 two halves are placed at locations far apart in memory. This option is
16803 used when compiling @file{crtstuff.c}; you should not need to use it
16806 @item -fverbose-asm
16807 @opindex fverbose-asm
16808 Put extra commentary information in the generated assembly code to
16809 make it more readable. This option is generally only of use to those
16810 who actually need to read the generated assembly code (perhaps while
16811 debugging the compiler itself).
16813 @option{-fno-verbose-asm}, the default, causes the
16814 extra information to be omitted and is useful when comparing two assembler
16817 @item -frecord-gcc-switches
16818 @opindex frecord-gcc-switches
16819 This switch causes the command line that was used to invoke the
16820 compiler to be recorded into the object file that is being created.
16821 This switch is only implemented on some targets and the exact format
16822 of the recording is target and binary file format dependent, but it
16823 usually takes the form of a section containing ASCII text. This
16824 switch is related to the @option{-fverbose-asm} switch, but that
16825 switch only records information in the assembler output file as
16826 comments, so it never reaches the object file.
16830 @cindex global offset table
16832 Generate position-independent code (PIC) suitable for use in a shared
16833 library, if supported for the target machine. Such code accesses all
16834 constant addresses through a global offset table (GOT)@. The dynamic
16835 loader resolves the GOT entries when the program starts (the dynamic
16836 loader is not part of GCC; it is part of the operating system). If
16837 the GOT size for the linked executable exceeds a machine-specific
16838 maximum size, you get an error message from the linker indicating that
16839 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
16840 instead. (These maximums are 8k on the SPARC and 32k
16841 on the m68k and RS/6000. The 386 has no such limit.)
16843 Position-independent code requires special support, and therefore works
16844 only on certain machines. For the 386, GCC supports PIC for System V
16845 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
16846 position-independent.
16848 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
16853 If supported for the target machine, emit position-independent code,
16854 suitable for dynamic linking and avoiding any limit on the size of the
16855 global offset table. This option makes a difference on the m68k,
16856 PowerPC and SPARC@.
16858 Position-independent code requires special support, and therefore works
16859 only on certain machines.
16861 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
16868 These options are similar to @option{-fpic} and @option{-fPIC}, but
16869 generated position independent code can be only linked into executables.
16870 Usually these options are used when @option{-pie} GCC option will be
16871 used during linking.
16873 @option{-fpie} and @option{-fPIE} both define the macros
16874 @code{__pie__} and @code{__PIE__}. The macros have the value 1
16875 for @option{-fpie} and 2 for @option{-fPIE}.
16877 @item -fno-jump-tables
16878 @opindex fno-jump-tables
16879 Do not use jump tables for switch statements even where it would be
16880 more efficient than other code generation strategies. This option is
16881 of use in conjunction with @option{-fpic} or @option{-fPIC} for
16882 building code which forms part of a dynamic linker and cannot
16883 reference the address of a jump table. On some targets, jump tables
16884 do not require a GOT and this option is not needed.
16886 @item -ffixed-@var{reg}
16888 Treat the register named @var{reg} as a fixed register; generated code
16889 should never refer to it (except perhaps as a stack pointer, frame
16890 pointer or in some other fixed role).
16892 @var{reg} must be the name of a register. The register names accepted
16893 are machine-specific and are defined in the @code{REGISTER_NAMES}
16894 macro in the machine description macro file.
16896 This flag does not have a negative form, because it specifies a
16899 @item -fcall-used-@var{reg}
16900 @opindex fcall-used
16901 Treat the register named @var{reg} as an allocable register that is
16902 clobbered by function calls. It may be allocated for temporaries or
16903 variables that do not live across a call. Functions compiled this way
16904 will not save and restore the register @var{reg}.
16906 It is an error to used this flag with the frame pointer or stack pointer.
16907 Use of this flag for other registers that have fixed pervasive roles in
16908 the machine's execution model will produce disastrous results.
16910 This flag does not have a negative form, because it specifies a
16913 @item -fcall-saved-@var{reg}
16914 @opindex fcall-saved
16915 Treat the register named @var{reg} as an allocable register saved by
16916 functions. It may be allocated even for temporaries or variables that
16917 live across a call. Functions compiled this way will save and restore
16918 the register @var{reg} if they use it.
16920 It is an error to used this flag with the frame pointer or stack pointer.
16921 Use of this flag for other registers that have fixed pervasive roles in
16922 the machine's execution model will produce disastrous results.
16924 A different sort of disaster will result from the use of this flag for
16925 a register in which function values may be returned.
16927 This flag does not have a negative form, because it specifies a
16930 @item -fpack-struct[=@var{n}]
16931 @opindex fpack-struct
16932 Without a value specified, pack all structure members together without
16933 holes. When a value is specified (which must be a small power of two), pack
16934 structure members according to this value, representing the maximum
16935 alignment (that is, objects with default alignment requirements larger than
16936 this will be output potentially unaligned at the next fitting location.
16938 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
16939 code that is not binary compatible with code generated without that switch.
16940 Additionally, it makes the code suboptimal.
16941 Use it to conform to a non-default application binary interface.
16943 @item -finstrument-functions
16944 @opindex finstrument-functions
16945 Generate instrumentation calls for entry and exit to functions. Just
16946 after function entry and just before function exit, the following
16947 profiling functions will be called with the address of the current
16948 function and its call site. (On some platforms,
16949 @code{__builtin_return_address} does not work beyond the current
16950 function, so the call site information may not be available to the
16951 profiling functions otherwise.)
16954 void __cyg_profile_func_enter (void *this_fn,
16956 void __cyg_profile_func_exit (void *this_fn,
16960 The first argument is the address of the start of the current function,
16961 which may be looked up exactly in the symbol table.
16963 This instrumentation is also done for functions expanded inline in other
16964 functions. The profiling calls will indicate where, conceptually, the
16965 inline function is entered and exited. This means that addressable
16966 versions of such functions must be available. If all your uses of a
16967 function are expanded inline, this may mean an additional expansion of
16968 code size. If you use @samp{extern inline} in your C code, an
16969 addressable version of such functions must be provided. (This is
16970 normally the case anyways, but if you get lucky and the optimizer always
16971 expands the functions inline, you might have gotten away without
16972 providing static copies.)
16974 A function may be given the attribute @code{no_instrument_function}, in
16975 which case this instrumentation will not be done. This can be used, for
16976 example, for the profiling functions listed above, high-priority
16977 interrupt routines, and any functions from which the profiling functions
16978 cannot safely be called (perhaps signal handlers, if the profiling
16979 routines generate output or allocate memory).
16981 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
16982 @opindex finstrument-functions-exclude-file-list
16984 Set the list of functions that are excluded from instrumentation (see
16985 the description of @code{-finstrument-functions}). If the file that
16986 contains a function definition matches with one of @var{file}, then
16987 that function is not instrumented. The match is done on substrings:
16988 if the @var{file} parameter is a substring of the file name, it is
16989 considered to be a match.
16992 @code{-finstrument-functions-exclude-file-list=/bits/stl,include/sys}
16993 will exclude any inline function defined in files whose pathnames
16994 contain @code{/bits/stl} or @code{include/sys}.
16996 If, for some reason, you want to include letter @code{','} in one of
16997 @var{sym}, write @code{'\,'}. For example,
16998 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
16999 (note the single quote surrounding the option).
17001 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
17002 @opindex finstrument-functions-exclude-function-list
17004 This is similar to @code{-finstrument-functions-exclude-file-list},
17005 but this option sets the list of function names to be excluded from
17006 instrumentation. The function name to be matched is its user-visible
17007 name, such as @code{vector<int> blah(const vector<int> &)}, not the
17008 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
17009 match is done on substrings: if the @var{sym} parameter is a substring
17010 of the function name, it is considered to be a match. For C99 and C++
17011 extended identifiers, the function name must be given in UTF-8, not
17012 using universal character names.
17014 @item -fstack-check
17015 @opindex fstack-check
17016 Generate code to verify that you do not go beyond the boundary of the
17017 stack. You should specify this flag if you are running in an
17018 environment with multiple threads, but only rarely need to specify it in
17019 a single-threaded environment since stack overflow is automatically
17020 detected on nearly all systems if there is only one stack.
17022 Note that this switch does not actually cause checking to be done; the
17023 operating system or the language runtime must do that. The switch causes
17024 generation of code to ensure that they see the stack being extended.
17026 You can additionally specify a string parameter: @code{no} means no
17027 checking, @code{generic} means force the use of old-style checking,
17028 @code{specific} means use the best checking method and is equivalent
17029 to bare @option{-fstack-check}.
17031 Old-style checking is a generic mechanism that requires no specific
17032 target support in the compiler but comes with the following drawbacks:
17036 Modified allocation strategy for large objects: they will always be
17037 allocated dynamically if their size exceeds a fixed threshold.
17040 Fixed limit on the size of the static frame of functions: when it is
17041 topped by a particular function, stack checking is not reliable and
17042 a warning is issued by the compiler.
17045 Inefficiency: because of both the modified allocation strategy and the
17046 generic implementation, the performances of the code are hampered.
17049 Note that old-style stack checking is also the fallback method for
17050 @code{specific} if no target support has been added in the compiler.
17052 @item -fstack-limit-register=@var{reg}
17053 @itemx -fstack-limit-symbol=@var{sym}
17054 @itemx -fno-stack-limit
17055 @opindex fstack-limit-register
17056 @opindex fstack-limit-symbol
17057 @opindex fno-stack-limit
17058 Generate code to ensure that the stack does not grow beyond a certain value,
17059 either the value of a register or the address of a symbol. If the stack
17060 would grow beyond the value, a signal is raised. For most targets,
17061 the signal is raised before the stack overruns the boundary, so
17062 it is possible to catch the signal without taking special precautions.
17064 For instance, if the stack starts at absolute address @samp{0x80000000}
17065 and grows downwards, you can use the flags
17066 @option{-fstack-limit-symbol=__stack_limit} and
17067 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
17068 of 128KB@. Note that this may only work with the GNU linker.
17070 @cindex aliasing of parameters
17071 @cindex parameters, aliased
17072 @item -fargument-alias
17073 @itemx -fargument-noalias
17074 @itemx -fargument-noalias-global
17075 @itemx -fargument-noalias-anything
17076 @opindex fargument-alias
17077 @opindex fargument-noalias
17078 @opindex fargument-noalias-global
17079 @opindex fargument-noalias-anything
17080 Specify the possible relationships among parameters and between
17081 parameters and global data.
17083 @option{-fargument-alias} specifies that arguments (parameters) may
17084 alias each other and may alias global storage.@*
17085 @option{-fargument-noalias} specifies that arguments do not alias
17086 each other, but may alias global storage.@*
17087 @option{-fargument-noalias-global} specifies that arguments do not
17088 alias each other and do not alias global storage.
17089 @option{-fargument-noalias-anything} specifies that arguments do not
17090 alias any other storage.
17092 Each language will automatically use whatever option is required by
17093 the language standard. You should not need to use these options yourself.
17095 @item -fleading-underscore
17096 @opindex fleading-underscore
17097 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
17098 change the way C symbols are represented in the object file. One use
17099 is to help link with legacy assembly code.
17101 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
17102 generate code that is not binary compatible with code generated without that
17103 switch. Use it to conform to a non-default application binary interface.
17104 Not all targets provide complete support for this switch.
17106 @item -ftls-model=@var{model}
17107 @opindex ftls-model
17108 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
17109 The @var{model} argument should be one of @code{global-dynamic},
17110 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
17112 The default without @option{-fpic} is @code{initial-exec}; with
17113 @option{-fpic} the default is @code{global-dynamic}.
17115 @item -fvisibility=@var{default|internal|hidden|protected}
17116 @opindex fvisibility
17117 Set the default ELF image symbol visibility to the specified option---all
17118 symbols will be marked with this unless overridden within the code.
17119 Using this feature can very substantially improve linking and
17120 load times of shared object libraries, produce more optimized
17121 code, provide near-perfect API export and prevent symbol clashes.
17122 It is @strong{strongly} recommended that you use this in any shared objects
17125 Despite the nomenclature, @code{default} always means public ie;
17126 available to be linked against from outside the shared object.
17127 @code{protected} and @code{internal} are pretty useless in real-world
17128 usage so the only other commonly used option will be @code{hidden}.
17129 The default if @option{-fvisibility} isn't specified is
17130 @code{default}, i.e., make every
17131 symbol public---this causes the same behavior as previous versions of
17134 A good explanation of the benefits offered by ensuring ELF
17135 symbols have the correct visibility is given by ``How To Write
17136 Shared Libraries'' by Ulrich Drepper (which can be found at
17137 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
17138 solution made possible by this option to marking things hidden when
17139 the default is public is to make the default hidden and mark things
17140 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
17141 and @code{__attribute__ ((visibility("default")))} instead of
17142 @code{__declspec(dllexport)} you get almost identical semantics with
17143 identical syntax. This is a great boon to those working with
17144 cross-platform projects.
17146 For those adding visibility support to existing code, you may find
17147 @samp{#pragma GCC visibility} of use. This works by you enclosing
17148 the declarations you wish to set visibility for with (for example)
17149 @samp{#pragma GCC visibility push(hidden)} and
17150 @samp{#pragma GCC visibility pop}.
17151 Bear in mind that symbol visibility should be viewed @strong{as
17152 part of the API interface contract} and thus all new code should
17153 always specify visibility when it is not the default ie; declarations
17154 only for use within the local DSO should @strong{always} be marked explicitly
17155 as hidden as so to avoid PLT indirection overheads---making this
17156 abundantly clear also aids readability and self-documentation of the code.
17157 Note that due to ISO C++ specification requirements, operator new and
17158 operator delete must always be of default visibility.
17160 Be aware that headers from outside your project, in particular system
17161 headers and headers from any other library you use, may not be
17162 expecting to be compiled with visibility other than the default. You
17163 may need to explicitly say @samp{#pragma GCC visibility push(default)}
17164 before including any such headers.
17166 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
17167 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
17168 no modifications. However, this means that calls to @samp{extern}
17169 functions with no explicit visibility will use the PLT, so it is more
17170 effective to use @samp{__attribute ((visibility))} and/or
17171 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
17172 declarations should be treated as hidden.
17174 Note that @samp{-fvisibility} does affect C++ vague linkage
17175 entities. This means that, for instance, an exception class that will
17176 be thrown between DSOs must be explicitly marked with default
17177 visibility so that the @samp{type_info} nodes will be unified between
17180 An overview of these techniques, their benefits and how to use them
17181 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
17187 @node Environment Variables
17188 @section Environment Variables Affecting GCC
17189 @cindex environment variables
17191 @c man begin ENVIRONMENT
17192 This section describes several environment variables that affect how GCC
17193 operates. Some of them work by specifying directories or prefixes to use
17194 when searching for various kinds of files. Some are used to specify other
17195 aspects of the compilation environment.
17197 Note that you can also specify places to search using options such as
17198 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
17199 take precedence over places specified using environment variables, which
17200 in turn take precedence over those specified by the configuration of GCC@.
17201 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
17202 GNU Compiler Collection (GCC) Internals}.
17207 @c @itemx LC_COLLATE
17209 @c @itemx LC_MONETARY
17210 @c @itemx LC_NUMERIC
17215 @c @findex LC_COLLATE
17216 @findex LC_MESSAGES
17217 @c @findex LC_MONETARY
17218 @c @findex LC_NUMERIC
17222 These environment variables control the way that GCC uses
17223 localization information that allow GCC to work with different
17224 national conventions. GCC inspects the locale categories
17225 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
17226 so. These locale categories can be set to any value supported by your
17227 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
17228 Kingdom encoded in UTF-8.
17230 The @env{LC_CTYPE} environment variable specifies character
17231 classification. GCC uses it to determine the character boundaries in
17232 a string; this is needed for some multibyte encodings that contain quote
17233 and escape characters that would otherwise be interpreted as a string
17236 The @env{LC_MESSAGES} environment variable specifies the language to
17237 use in diagnostic messages.
17239 If the @env{LC_ALL} environment variable is set, it overrides the value
17240 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
17241 and @env{LC_MESSAGES} default to the value of the @env{LANG}
17242 environment variable. If none of these variables are set, GCC
17243 defaults to traditional C English behavior.
17247 If @env{TMPDIR} is set, it specifies the directory to use for temporary
17248 files. GCC uses temporary files to hold the output of one stage of
17249 compilation which is to be used as input to the next stage: for example,
17250 the output of the preprocessor, which is the input to the compiler
17253 @item GCC_EXEC_PREFIX
17254 @findex GCC_EXEC_PREFIX
17255 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
17256 names of the subprograms executed by the compiler. No slash is added
17257 when this prefix is combined with the name of a subprogram, but you can
17258 specify a prefix that ends with a slash if you wish.
17260 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
17261 an appropriate prefix to use based on the pathname it was invoked with.
17263 If GCC cannot find the subprogram using the specified prefix, it
17264 tries looking in the usual places for the subprogram.
17266 The default value of @env{GCC_EXEC_PREFIX} is
17267 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
17268 the installed compiler. In many cases @var{prefix} is the value
17269 of @code{prefix} when you ran the @file{configure} script.
17271 Other prefixes specified with @option{-B} take precedence over this prefix.
17273 This prefix is also used for finding files such as @file{crt0.o} that are
17276 In addition, the prefix is used in an unusual way in finding the
17277 directories to search for header files. For each of the standard
17278 directories whose name normally begins with @samp{/usr/local/lib/gcc}
17279 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
17280 replacing that beginning with the specified prefix to produce an
17281 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
17282 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
17283 These alternate directories are searched first; the standard directories
17284 come next. If a standard directory begins with the configured
17285 @var{prefix} then the value of @var{prefix} is replaced by
17286 @env{GCC_EXEC_PREFIX} when looking for header files.
17288 @item COMPILER_PATH
17289 @findex COMPILER_PATH
17290 The value of @env{COMPILER_PATH} is a colon-separated list of
17291 directories, much like @env{PATH}. GCC tries the directories thus
17292 specified when searching for subprograms, if it can't find the
17293 subprograms using @env{GCC_EXEC_PREFIX}.
17296 @findex LIBRARY_PATH
17297 The value of @env{LIBRARY_PATH} is a colon-separated list of
17298 directories, much like @env{PATH}. When configured as a native compiler,
17299 GCC tries the directories thus specified when searching for special
17300 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
17301 using GCC also uses these directories when searching for ordinary
17302 libraries for the @option{-l} option (but directories specified with
17303 @option{-L} come first).
17307 @cindex locale definition
17308 This variable is used to pass locale information to the compiler. One way in
17309 which this information is used is to determine the character set to be used
17310 when character literals, string literals and comments are parsed in C and C++.
17311 When the compiler is configured to allow multibyte characters,
17312 the following values for @env{LANG} are recognized:
17316 Recognize JIS characters.
17318 Recognize SJIS characters.
17320 Recognize EUCJP characters.
17323 If @env{LANG} is not defined, or if it has some other value, then the
17324 compiler will use mblen and mbtowc as defined by the default locale to
17325 recognize and translate multibyte characters.
17329 Some additional environments variables affect the behavior of the
17332 @include cppenv.texi
17336 @node Precompiled Headers
17337 @section Using Precompiled Headers
17338 @cindex precompiled headers
17339 @cindex speed of compilation
17341 Often large projects have many header files that are included in every
17342 source file. The time the compiler takes to process these header files
17343 over and over again can account for nearly all of the time required to
17344 build the project. To make builds faster, GCC allows users to
17345 `precompile' a header file; then, if builds can use the precompiled
17346 header file they will be much faster.
17348 To create a precompiled header file, simply compile it as you would any
17349 other file, if necessary using the @option{-x} option to make the driver
17350 treat it as a C or C++ header file. You will probably want to use a
17351 tool like @command{make} to keep the precompiled header up-to-date when
17352 the headers it contains change.
17354 A precompiled header file will be searched for when @code{#include} is
17355 seen in the compilation. As it searches for the included file
17356 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
17357 compiler looks for a precompiled header in each directory just before it
17358 looks for the include file in that directory. The name searched for is
17359 the name specified in the @code{#include} with @samp{.gch} appended. If
17360 the precompiled header file can't be used, it is ignored.
17362 For instance, if you have @code{#include "all.h"}, and you have
17363 @file{all.h.gch} in the same directory as @file{all.h}, then the
17364 precompiled header file will be used if possible, and the original
17365 header will be used otherwise.
17367 Alternatively, you might decide to put the precompiled header file in a
17368 directory and use @option{-I} to ensure that directory is searched
17369 before (or instead of) the directory containing the original header.
17370 Then, if you want to check that the precompiled header file is always
17371 used, you can put a file of the same name as the original header in this
17372 directory containing an @code{#error} command.
17374 This also works with @option{-include}. So yet another way to use
17375 precompiled headers, good for projects not designed with precompiled
17376 header files in mind, is to simply take most of the header files used by
17377 a project, include them from another header file, precompile that header
17378 file, and @option{-include} the precompiled header. If the header files
17379 have guards against multiple inclusion, they will be skipped because
17380 they've already been included (in the precompiled header).
17382 If you need to precompile the same header file for different
17383 languages, targets, or compiler options, you can instead make a
17384 @emph{directory} named like @file{all.h.gch}, and put each precompiled
17385 header in the directory, perhaps using @option{-o}. It doesn't matter
17386 what you call the files in the directory, every precompiled header in
17387 the directory will be considered. The first precompiled header
17388 encountered in the directory that is valid for this compilation will
17389 be used; they're searched in no particular order.
17391 There are many other possibilities, limited only by your imagination,
17392 good sense, and the constraints of your build system.
17394 A precompiled header file can be used only when these conditions apply:
17398 Only one precompiled header can be used in a particular compilation.
17401 A precompiled header can't be used once the first C token is seen. You
17402 can have preprocessor directives before a precompiled header; you can
17403 even include a precompiled header from inside another header, so long as
17404 there are no C tokens before the @code{#include}.
17407 The precompiled header file must be produced for the same language as
17408 the current compilation. You can't use a C precompiled header for a C++
17412 The precompiled header file must have been produced by the same compiler
17413 binary as the current compilation is using.
17416 Any macros defined before the precompiled header is included must
17417 either be defined in the same way as when the precompiled header was
17418 generated, or must not affect the precompiled header, which usually
17419 means that they don't appear in the precompiled header at all.
17421 The @option{-D} option is one way to define a macro before a
17422 precompiled header is included; using a @code{#define} can also do it.
17423 There are also some options that define macros implicitly, like
17424 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
17427 @item If debugging information is output when using the precompiled
17428 header, using @option{-g} or similar, the same kind of debugging information
17429 must have been output when building the precompiled header. However,
17430 a precompiled header built using @option{-g} can be used in a compilation
17431 when no debugging information is being output.
17433 @item The same @option{-m} options must generally be used when building
17434 and using the precompiled header. @xref{Submodel Options},
17435 for any cases where this rule is relaxed.
17437 @item Each of the following options must be the same when building and using
17438 the precompiled header:
17440 @gccoptlist{-fexceptions}
17443 Some other command-line options starting with @option{-f},
17444 @option{-p}, or @option{-O} must be defined in the same way as when
17445 the precompiled header was generated. At present, it's not clear
17446 which options are safe to change and which are not; the safest choice
17447 is to use exactly the same options when generating and using the
17448 precompiled header. The following are known to be safe:
17450 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
17451 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
17452 -fsched-verbose=<number> -fschedule-insns -fvisibility= @gol
17457 For all of these except the last, the compiler will automatically
17458 ignore the precompiled header if the conditions aren't met. If you
17459 find an option combination that doesn't work and doesn't cause the
17460 precompiled header to be ignored, please consider filing a bug report,
17463 If you do use differing options when generating and using the
17464 precompiled header, the actual behavior will be a mixture of the
17465 behavior for the options. For instance, if you use @option{-g} to
17466 generate the precompiled header but not when using it, you may or may
17467 not get debugging information for routines in the precompiled header.