1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
2 @c 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009
3 @c Free Software Foundation, Inc.
4 @c This is part of the GCC manual.
5 @c For copying conditions, see the file gcc.texi.
12 @c man begin COPYRIGHT
13 Copyright @copyright{} 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
14 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009
15 Free Software Foundation, Inc.
17 Permission is granted to copy, distribute and/or modify this document
18 under the terms of the GNU Free Documentation License, Version 1.2 or
19 any later version published by the Free Software Foundation; with the
20 Invariant Sections being ``GNU General Public License'' and ``Funding
21 Free Software'', the Front-Cover texts being (a) (see below), and with
22 the Back-Cover Texts being (b) (see below). A copy of the license is
23 included in the gfdl(7) man page.
25 (a) The FSF's Front-Cover Text is:
29 (b) The FSF's Back-Cover Text is:
31 You have freedom to copy and modify this GNU Manual, like GNU
32 software. Copies published by the Free Software Foundation raise
33 funds for GNU development.
35 @c Set file name and title for the man page.
37 @settitle GNU project C and C++ compiler
39 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
40 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
41 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
42 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
43 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
44 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
45 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
47 Only the most useful options are listed here; see below for the
48 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
51 gpl(7), gfdl(7), fsf-funding(7),
52 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
53 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
54 @file{ld}, @file{binutils} and @file{gdb}.
57 For instructions on reporting bugs, see
61 See the Info entry for @command{gcc}, or
62 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
63 for contributors to GCC@.
68 @chapter GCC Command Options
69 @cindex GCC command options
70 @cindex command options
71 @cindex options, GCC command
73 @c man begin DESCRIPTION
74 When you invoke GCC, it normally does preprocessing, compilation,
75 assembly and linking. The ``overall options'' allow you to stop this
76 process at an intermediate stage. For example, the @option{-c} option
77 says not to run the linker. Then the output consists of object files
78 output by the assembler.
80 Other options are passed on to one stage of processing. Some options
81 control the preprocessor and others the compiler itself. Yet other
82 options control the assembler and linker; most of these are not
83 documented here, since you rarely need to use any of them.
85 @cindex C compilation options
86 Most of the command line options that you can use with GCC are useful
87 for C programs; when an option is only useful with another language
88 (usually C++), the explanation says so explicitly. If the description
89 for a particular option does not mention a source language, you can use
90 that option with all supported languages.
92 @cindex C++ compilation options
93 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
94 options for compiling C++ programs.
96 @cindex grouping options
97 @cindex options, grouping
98 The @command{gcc} program accepts options and file names as operands. Many
99 options have multi-letter names; therefore multiple single-letter options
100 may @emph{not} be grouped: @option{-dv} is very different from @w{@samp{-d
103 @cindex order of options
104 @cindex options, order
105 You can mix options and other arguments. For the most part, the order
106 you use doesn't matter. Order does matter when you use several
107 options of the same kind; for example, if you specify @option{-L} more
108 than once, the directories are searched in the order specified. Also,
109 the placement of the @option{-l} option is significant.
111 Many options have long names starting with @samp{-f} or with
112 @samp{-W}---for example,
113 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
114 these have both positive and negative forms; the negative form of
115 @option{-ffoo} would be @option{-fno-foo}. This manual documents
116 only one of these two forms, whichever one is not the default.
120 @xref{Option Index}, for an index to GCC's options.
123 * Option Summary:: Brief list of all options, without explanations.
124 * Overall Options:: Controlling the kind of output:
125 an executable, object files, assembler files,
126 or preprocessed source.
127 * Invoking G++:: Compiling C++ programs.
128 * C Dialect Options:: Controlling the variant of C language compiled.
129 * C++ Dialect Options:: Variations on C++.
130 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
132 * Language Independent Options:: Controlling how diagnostics should be
134 * Warning Options:: How picky should the compiler be?
135 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
136 * Optimize Options:: How much optimization?
137 * Preprocessor Options:: Controlling header files and macro definitions.
138 Also, getting dependency information for Make.
139 * Assembler Options:: Passing options to the assembler.
140 * Link Options:: Specifying libraries and so on.
141 * Directory Options:: Where to find header files and libraries.
142 Where to find the compiler executable files.
143 * Spec Files:: How to pass switches to sub-processes.
144 * Target Options:: Running a cross-compiler, or an old version of GCC.
145 * Submodel Options:: Specifying minor hardware or convention variations,
146 such as 68010 vs 68020.
147 * Code Gen Options:: Specifying conventions for function calls, data layout
149 * Environment Variables:: Env vars that affect GCC.
150 * Precompiled Headers:: Compiling a header once, and using it many times.
156 @section Option Summary
158 Here is a summary of all the options, grouped by type. Explanations are
159 in the following sections.
162 @item Overall Options
163 @xref{Overall Options,,Options Controlling the Kind of Output}.
164 @gccoptlist{-c -S -E -o @var{file} -combine -no-canonical-prefixes @gol
165 -pipe -pass-exit-codes @gol
166 -x @var{language} -v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help @gol
167 --version -wrapper@@@var{file} -fplugin=@var{file} -fplugin-arg-@var{name}=@var{arg}}
169 @item C Language Options
170 @xref{C Dialect Options,,Options Controlling C Dialect}.
171 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
172 -aux-info @var{filename} @gol
173 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
174 -fhosted -ffreestanding -fopenmp -fms-extensions @gol
175 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
176 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
177 -fsigned-bitfields -fsigned-char @gol
178 -funsigned-bitfields -funsigned-char}
180 @item C++ Language Options
181 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
182 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
183 -fconserve-space -ffriend-injection @gol
184 -fno-elide-constructors @gol
185 -fno-enforce-eh-specs @gol
186 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
187 -fno-implicit-templates @gol
188 -fno-implicit-inline-templates @gol
189 -fno-implement-inlines -fms-extensions @gol
190 -fno-nonansi-builtins -fno-operator-names @gol
191 -fno-optional-diags -fpermissive @gol
192 -fno-pretty-templates @gol
193 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
194 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
195 -fno-default-inline -fvisibility-inlines-hidden @gol
196 -fvisibility-ms-compat @gol
197 -Wabi -Wctor-dtor-privacy @gol
198 -Wnon-virtual-dtor -Wreorder @gol
199 -Weffc++ -Wstrict-null-sentinel @gol
200 -Wno-non-template-friend -Wold-style-cast @gol
201 -Woverloaded-virtual -Wno-pmf-conversions @gol
204 @item Objective-C and Objective-C++ Language Options
205 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
206 Objective-C and Objective-C++ Dialects}.
207 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
208 -fgnu-runtime -fnext-runtime @gol
209 -fno-nil-receivers @gol
210 -fobjc-call-cxx-cdtors @gol
211 -fobjc-direct-dispatch @gol
212 -fobjc-exceptions @gol
214 -freplace-objc-classes @gol
217 -Wassign-intercept @gol
218 -Wno-protocol -Wselector @gol
219 -Wstrict-selector-match @gol
220 -Wundeclared-selector}
222 @item Language Independent Options
223 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
224 @gccoptlist{-fmessage-length=@var{n} @gol
225 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
226 -fdiagnostics-show-option}
228 @item Warning Options
229 @xref{Warning Options,,Options to Request or Suppress Warnings}.
230 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
231 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
232 -Wno-attributes -Wno-builtin-macro-redefined @gol
233 -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual @gol
234 -Wchar-subscripts -Wclobbered -Wcomment @gol
235 -Wconversion -Wcoverage-mismatch -Wno-deprecated @gol
236 -Wno-deprecated-declarations -Wdisabled-optimization @gol
237 -Wno-div-by-zero -Wempty-body -Wenum-compare -Wno-endif-labels @gol
238 -Werror -Werror=* @gol
239 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
240 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
241 -Wformat-security -Wformat-y2k @gol
242 -Wframe-larger-than=@var{len} -Wjump-misses-init -Wignored-qualifiers @gol
243 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
244 -Winit-self -Winline @gol
245 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
246 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
247 -Wlogical-op -Wlong-long @gol
248 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
249 -Wmissing-format-attribute -Wmissing-include-dirs @gol
250 -Wmissing-noreturn -Wno-mudflap @gol
251 -Wno-multichar -Wnonnull -Wno-overflow @gol
252 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
253 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
254 -Wpointer-arith -Wno-pointer-to-int-cast @gol
255 -Wredundant-decls @gol
256 -Wreturn-type -Wsequence-point -Wshadow @gol
257 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
258 -Wstrict-aliasing -Wstrict-aliasing=n @gol
259 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
260 -Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand @gol
261 -Wsystem-headers -Wtrigraphs -Wtype-limits -Wundef -Wuninitialized @gol
262 -Wunknown-pragmas -Wno-pragmas -Wunreachable-code @gol
263 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
264 -Wunused-label -Wunused-parameter -Wno-unused-result -Wunused-value -Wunused-variable @gol
265 -Wvariadic-macros -Wvla @gol
266 -Wvolatile-register-var -Wwrite-strings}
268 @item C and Objective-C-only Warning Options
269 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
270 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
271 -Wold-style-declaration -Wold-style-definition @gol
272 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
273 -Wdeclaration-after-statement -Wpointer-sign}
275 @item Debugging Options
276 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
277 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
278 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
279 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
280 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
281 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
282 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
283 -fdump-statistics @gol
285 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
286 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
287 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
289 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
290 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
291 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
292 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
293 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
294 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
295 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
296 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
297 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
298 -fdump-tree-nrv -fdump-tree-vect @gol
299 -fdump-tree-sink @gol
300 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
301 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
302 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
303 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
304 -ftree-vectorizer-verbose=@var{n} @gol
305 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
306 -fdump-final-insns=@var{file} @gol
307 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
308 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
309 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
310 -fenable-icf-debug @gol
311 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
312 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
313 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
314 -ftest-coverage -ftime-report -fvar-tracking @gol
315 -fvar-tracking-assigments -fvar-tracking-assignments-toggle @gol
316 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
317 -ggdb -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
318 -gvms -gxcoff -gxcoff+ @gol
319 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
320 -fdebug-prefix-map=@var{old}=@var{new} @gol
321 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
322 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
323 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
324 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
325 -print-prog-name=@var{program} -print-search-dirs -Q @gol
326 -print-sysroot -print-sysroot-headers-suffix @gol
327 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
329 @item Optimization Options
330 @xref{Optimize Options,,Options that Control Optimization}.
332 -falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
333 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
334 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
335 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
336 -fcheck-data-deps -fconserve-stack -fcprop-registers -fcrossjumping @gol
337 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules -fcx-limited-range @gol
338 -fdata-sections -fdce -fdce @gol
339 -fdelayed-branch -fdelete-null-pointer-checks -fdse -fdse @gol
340 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffast-math @gol
341 -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
342 -fforward-propagate -ffunction-sections @gol
343 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm @gol
344 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
345 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
346 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg -fipa-pta @gol
347 -fipa-pure-const -fipa-reference -fipa-struct-reorg @gol
348 -fipa-type-escape -fira-algorithm=@var{algorithm} @gol
349 -fira-region=@var{region} -fira-coalesce @gol
350 -fira-loop-pressure -fno-ira-share-save-slots @gol
351 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
352 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
353 -floop-block -floop-interchange -floop-strip-mine -fgraphite-identity @gol
354 -floop-parallelize-all -flto -flto-compression-level -flto-report -fltrans @gol
355 -fltrans-output-list -fmerge-all-constants -fmerge-constants -fmodulo-sched @gol
356 -fmodulo-sched-allow-regmoves -fmove-loop-invariants -fmudflap @gol
357 -fmudflapir -fmudflapth -fno-branch-count-reg -fno-default-inline @gol
358 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
359 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
360 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
361 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
362 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
363 -fpeel-loops -fpredictive-commoning -fprefetch-loop-arrays @gol
364 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
365 -fprofile-generate=@var{path} @gol
366 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
367 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
368 -freorder-blocks-and-partition -freorder-functions @gol
369 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
370 -frounding-math -fsched2-use-superblocks @gol
371 -fsched2-use-traces -fsched-pressure @gol
372 -fsched-spec-load -fsched-spec-load-dangerous @gol
373 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
374 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
375 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
376 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
377 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
378 -fselective-scheduling -fselective-scheduling2 @gol
379 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
380 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
381 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
382 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer @gol
383 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
384 -ftree-copyrename -ftree-dce @gol
385 -ftree-dominator-opts -ftree-dse -ftree-forwprop -ftree-fre -ftree-loop-im @gol
386 -ftree-phiprop -ftree-loop-distribution @gol
387 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
388 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-pta -ftree-reassoc @gol
389 -ftree-sink -ftree-sra -ftree-switch-conversion @gol
390 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
391 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
392 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
393 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
394 -fwhole-program -fwhopr -fwpa -use-linker-plugin @gol
395 --param @var{name}=@var{value}
396 -O -O0 -O1 -O2 -O3 -Os}
398 @item Preprocessor Options
399 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
400 @gccoptlist{-A@var{question}=@var{answer} @gol
401 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
402 -C -dD -dI -dM -dN @gol
403 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
404 -idirafter @var{dir} @gol
405 -include @var{file} -imacros @var{file} @gol
406 -iprefix @var{file} -iwithprefix @var{dir} @gol
407 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
408 -imultilib @var{dir} -isysroot @var{dir} @gol
409 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
410 -P -fworking-directory -remap @gol
411 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
412 -Xpreprocessor @var{option}}
414 @item Assembler Option
415 @xref{Assembler Options,,Passing Options to the Assembler}.
416 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
419 @xref{Link Options,,Options for Linking}.
420 @gccoptlist{@var{object-file-name} -l@var{library} @gol
421 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
422 -s -static -static-libgcc -static-libstdc++ -shared @gol
423 -shared-libgcc -symbolic @gol
424 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
427 @item Directory Options
428 @xref{Directory Options,,Options for Directory Search}.
429 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir}
430 -specs=@var{file} -I- --sysroot=@var{dir}}
433 @c I wrote this xref this way to avoid overfull hbox. -- rms
434 @xref{Target Options}.
435 @gccoptlist{-V @var{version} -b @var{machine}}
437 @item Machine Dependent Options
438 @xref{Submodel Options,,Hardware Models and Configurations}.
439 @c This list is ordered alphanumerically by subsection name.
440 @c Try and put the significant identifier (CPU or system) first,
441 @c so users have a clue at guessing where the ones they want will be.
444 @gccoptlist{-EB -EL @gol
445 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
446 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
449 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
450 -mabi=@var{name} @gol
451 -mapcs-stack-check -mno-apcs-stack-check @gol
452 -mapcs-float -mno-apcs-float @gol
453 -mapcs-reentrant -mno-apcs-reentrant @gol
454 -msched-prolog -mno-sched-prolog @gol
455 -mlittle-endian -mbig-endian -mwords-little-endian @gol
456 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
457 -mfp16-format=@var{name}
458 -mthumb-interwork -mno-thumb-interwork @gol
459 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
460 -mstructure-size-boundary=@var{n} @gol
461 -mabort-on-noreturn @gol
462 -mlong-calls -mno-long-calls @gol
463 -msingle-pic-base -mno-single-pic-base @gol
464 -mpic-register=@var{reg} @gol
465 -mnop-fun-dllimport @gol
466 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
467 -mpoke-function-name @gol
469 -mtpcs-frame -mtpcs-leaf-frame @gol
470 -mcaller-super-interworking -mcallee-super-interworking @gol
472 -mword-relocations @gol
473 -mfix-cortex-m3-ldrd}
476 @gccoptlist{-mmcu=@var{mcu} -mno-interrupts @gol
477 -mcall-prologues -mtiny-stack -mint8}
479 @emph{Blackfin Options}
480 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
481 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
482 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
483 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
484 -mno-id-shared-library -mshared-library-id=@var{n} @gol
485 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
486 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
487 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
491 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
492 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
493 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
494 -mstack-align -mdata-align -mconst-align @gol
495 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
496 -melf -maout -melinux -mlinux -sim -sim2 @gol
497 -mmul-bug-workaround -mno-mul-bug-workaround}
500 @gccoptlist{-mmac -mpush-args}
502 @emph{Darwin Options}
503 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
504 -arch_only -bind_at_load -bundle -bundle_loader @gol
505 -client_name -compatibility_version -current_version @gol
507 -dependency-file -dylib_file -dylinker_install_name @gol
508 -dynamic -dynamiclib -exported_symbols_list @gol
509 -filelist -flat_namespace -force_cpusubtype_ALL @gol
510 -force_flat_namespace -headerpad_max_install_names @gol
512 -image_base -init -install_name -keep_private_externs @gol
513 -multi_module -multiply_defined -multiply_defined_unused @gol
514 -noall_load -no_dead_strip_inits_and_terms @gol
515 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
516 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
517 -private_bundle -read_only_relocs -sectalign @gol
518 -sectobjectsymbols -whyload -seg1addr @gol
519 -sectcreate -sectobjectsymbols -sectorder @gol
520 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
521 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
522 -segprot -segs_read_only_addr -segs_read_write_addr @gol
523 -single_module -static -sub_library -sub_umbrella @gol
524 -twolevel_namespace -umbrella -undefined @gol
525 -unexported_symbols_list -weak_reference_mismatches @gol
526 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
527 -mkernel -mone-byte-bool}
529 @emph{DEC Alpha Options}
530 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
531 -mieee -mieee-with-inexact -mieee-conformant @gol
532 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
533 -mtrap-precision=@var{mode} -mbuild-constants @gol
534 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
535 -mbwx -mmax -mfix -mcix @gol
536 -mfloat-vax -mfloat-ieee @gol
537 -mexplicit-relocs -msmall-data -mlarge-data @gol
538 -msmall-text -mlarge-text @gol
539 -mmemory-latency=@var{time}}
541 @emph{DEC Alpha/VMS Options}
542 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
545 @gccoptlist{-msmall-model -mno-lsim}
548 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
549 -mhard-float -msoft-float @gol
550 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
551 -mdouble -mno-double @gol
552 -mmedia -mno-media -mmuladd -mno-muladd @gol
553 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
554 -mlinked-fp -mlong-calls -malign-labels @gol
555 -mlibrary-pic -macc-4 -macc-8 @gol
556 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
557 -moptimize-membar -mno-optimize-membar @gol
558 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
559 -mvliw-branch -mno-vliw-branch @gol
560 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
561 -mno-nested-cond-exec -mtomcat-stats @gol
565 @emph{GNU/Linux Options}
566 @gccoptlist{-muclibc}
568 @emph{H8/300 Options}
569 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
572 @gccoptlist{-march=@var{architecture-type} @gol
573 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
574 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
575 -mfixed-range=@var{register-range} @gol
576 -mjump-in-delay -mlinker-opt -mlong-calls @gol
577 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
578 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
579 -mno-jump-in-delay -mno-long-load-store @gol
580 -mno-portable-runtime -mno-soft-float @gol
581 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
582 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
583 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
584 -munix=@var{unix-std} -nolibdld -static -threads}
586 @emph{i386 and x86-64 Options}
587 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
588 -mfpmath=@var{unit} @gol
589 -masm=@var{dialect} -mno-fancy-math-387 @gol
590 -mno-fp-ret-in-387 -msoft-float @gol
591 -mno-wide-multiply -mrtd -malign-double @gol
592 -mpreferred-stack-boundary=@var{num}
593 -mincoming-stack-boundary=@var{num}
594 -mcld -mcx16 -msahf -mmovbe -mcrc32 -mrecip @gol
595 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
597 -msse4a -m3dnow -mpopcnt -mabm -mfma4 @gol
598 -mthreads -mno-align-stringops -minline-all-stringops @gol
599 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
600 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
601 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
602 -mveclibabi=@var{type} -mpc32 -mpc64 -mpc80 -mstackrealign @gol
603 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
604 -mcmodel=@var{code-model} -mabi=@var{name} @gol
605 -m32 -m64 -mlarge-data-threshold=@var{num} @gol
609 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
610 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
611 -mconstant-gp -mauto-pic -mfused-madd @gol
612 -minline-float-divide-min-latency @gol
613 -minline-float-divide-max-throughput @gol
614 -mno-inline-float-divide @gol
615 -minline-int-divide-min-latency @gol
616 -minline-int-divide-max-throughput @gol
617 -mno-inline-int-divide @gol
618 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
619 -mno-inline-sqrt @gol
620 -mdwarf2-asm -mearly-stop-bits @gol
621 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
622 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
623 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
624 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
625 -msched-spec-ldc -msched-spec-control-ldc @gol
626 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
627 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
628 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
629 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
631 @emph{IA-64/VMS Options}
632 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
634 @emph{M32R/D Options}
635 @gccoptlist{-m32r2 -m32rx -m32r @gol
637 -malign-loops -mno-align-loops @gol
638 -missue-rate=@var{number} @gol
639 -mbranch-cost=@var{number} @gol
640 -mmodel=@var{code-size-model-type} @gol
641 -msdata=@var{sdata-type} @gol
642 -mno-flush-func -mflush-func=@var{name} @gol
643 -mno-flush-trap -mflush-trap=@var{number} @gol
647 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
649 @emph{M680x0 Options}
650 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
651 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
652 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
653 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
654 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
655 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
656 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
657 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
660 @emph{M68hc1x Options}
661 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
662 -mauto-incdec -minmax -mlong-calls -mshort @gol
663 -msoft-reg-count=@var{count}}
666 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
667 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
668 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
669 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
670 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
673 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
674 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
675 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
676 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
680 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
681 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
682 -mips64 -mips64r2 @gol
683 -mips16 -mno-mips16 -mflip-mips16 @gol
684 -minterlink-mips16 -mno-interlink-mips16 @gol
685 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
686 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
687 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
688 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
689 -mfpu=@var{fpu-type} @gol
690 -msmartmips -mno-smartmips @gol
691 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
692 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
693 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
694 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
695 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
696 -membedded-data -mno-embedded-data @gol
697 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
698 -mcode-readable=@var{setting} @gol
699 -msplit-addresses -mno-split-addresses @gol
700 -mexplicit-relocs -mno-explicit-relocs @gol
701 -mcheck-zero-division -mno-check-zero-division @gol
702 -mdivide-traps -mdivide-breaks @gol
703 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
704 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
705 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
706 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
707 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
708 -mflush-func=@var{func} -mno-flush-func @gol
709 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
710 -mfp-exceptions -mno-fp-exceptions @gol
711 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
712 -mrelax-pic-calls -mno-relax-pic-calls}
715 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
716 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
717 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
718 -mno-base-addresses -msingle-exit -mno-single-exit}
720 @emph{MN10300 Options}
721 @gccoptlist{-mmult-bug -mno-mult-bug @gol
722 -mam33 -mno-am33 @gol
723 -mam33-2 -mno-am33-2 @gol
724 -mreturn-pointer-on-d0 @gol
727 @emph{PDP-11 Options}
728 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
729 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
730 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
731 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
732 -mbranch-expensive -mbranch-cheap @gol
733 -msplit -mno-split -munix-asm -mdec-asm}
735 @emph{picoChip Options}
736 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N}
737 -msymbol-as-address -mno-inefficient-warnings}
739 @emph{PowerPC Options}
740 See RS/6000 and PowerPC Options.
742 @emph{RS/6000 and PowerPC Options}
743 @gccoptlist{-mcpu=@var{cpu-type} @gol
744 -mtune=@var{cpu-type} @gol
745 -mpower -mno-power -mpower2 -mno-power2 @gol
746 -mpowerpc -mpowerpc64 -mno-powerpc @gol
747 -maltivec -mno-altivec @gol
748 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
749 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
750 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
751 -mfprnd -mno-fprnd @gol
752 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
753 -mnew-mnemonics -mold-mnemonics @gol
754 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
755 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
756 -malign-power -malign-natural @gol
757 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
758 -msingle-float -mdouble-float -msimple-fpu @gol
759 -mstring -mno-string -mupdate -mno-update @gol
760 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
761 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
762 -mstrict-align -mno-strict-align -mrelocatable @gol
763 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
764 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
765 -mdynamic-no-pic -maltivec -mswdiv @gol
766 -mprioritize-restricted-insns=@var{priority} @gol
767 -msched-costly-dep=@var{dependence_type} @gol
768 -minsert-sched-nops=@var{scheme} @gol
769 -mcall-sysv -mcall-netbsd @gol
770 -maix-struct-return -msvr4-struct-return @gol
771 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
772 -misel -mno-isel @gol
773 -misel=yes -misel=no @gol
775 -mspe=yes -mspe=no @gol
777 -mgen-cell-microcode -mwarn-cell-microcode @gol
778 -mvrsave -mno-vrsave @gol
779 -mmulhw -mno-mulhw @gol
780 -mdlmzb -mno-dlmzb @gol
781 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
782 -mprototype -mno-prototype @gol
783 -msim -mmvme -mads -myellowknife -memb -msdata @gol
784 -msdata=@var{opt} -mvxworks -G @var{num} -pthread}
786 @emph{S/390 and zSeries Options}
787 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
788 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
789 -mlong-double-64 -mlong-double-128 @gol
790 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
791 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
792 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
793 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
794 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
797 @gccoptlist{-meb -mel @gol
801 -mscore5 -mscore5u -mscore7 -mscore7d}
804 @gccoptlist{-m1 -m2 -m2e @gol
805 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
807 -m4-nofpu -m4-single-only -m4-single -m4 @gol
808 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
809 -m5-64media -m5-64media-nofpu @gol
810 -m5-32media -m5-32media-nofpu @gol
811 -m5-compact -m5-compact-nofpu @gol
812 -mb -ml -mdalign -mrelax @gol
813 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
814 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
815 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
816 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
817 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
821 @gccoptlist{-mcpu=@var{cpu-type} @gol
822 -mtune=@var{cpu-type} @gol
823 -mcmodel=@var{code-model} @gol
824 -m32 -m64 -mapp-regs -mno-app-regs @gol
825 -mfaster-structs -mno-faster-structs @gol
826 -mfpu -mno-fpu -mhard-float -msoft-float @gol
827 -mhard-quad-float -msoft-quad-float @gol
828 -mimpure-text -mno-impure-text -mlittle-endian @gol
829 -mstack-bias -mno-stack-bias @gol
830 -munaligned-doubles -mno-unaligned-doubles @gol
831 -mv8plus -mno-v8plus -mvis -mno-vis
832 -threads -pthreads -pthread}
835 @gccoptlist{-mwarn-reloc -merror-reloc @gol
836 -msafe-dma -munsafe-dma @gol
838 -msmall-mem -mlarge-mem -mstdmain @gol
839 -mfixed-range=@var{register-range}}
841 @emph{System V Options}
842 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
845 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
846 -mprolog-function -mno-prolog-function -mspace @gol
847 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
848 -mapp-regs -mno-app-regs @gol
849 -mdisable-callt -mno-disable-callt @gol
855 @gccoptlist{-mg -mgnu -munix}
857 @emph{VxWorks Options}
858 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
859 -Xbind-lazy -Xbind-now}
861 @emph{x86-64 Options}
862 See i386 and x86-64 Options.
864 @emph{i386 and x86-64 Windows Options}
865 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll
866 -mnop-fun-dllimport -mthread -municode -mwin32 -mwindows}
868 @emph{Xstormy16 Options}
871 @emph{Xtensa Options}
872 @gccoptlist{-mconst16 -mno-const16 @gol
873 -mfused-madd -mno-fused-madd @gol
874 -mserialize-volatile -mno-serialize-volatile @gol
875 -mtext-section-literals -mno-text-section-literals @gol
876 -mtarget-align -mno-target-align @gol
877 -mlongcalls -mno-longcalls}
879 @emph{zSeries Options}
880 See S/390 and zSeries Options.
882 @item Code Generation Options
883 @xref{Code Gen Options,,Options for Code Generation Conventions}.
884 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
885 -ffixed-@var{reg} -fexceptions @gol
886 -fnon-call-exceptions -funwind-tables @gol
887 -fasynchronous-unwind-tables @gol
888 -finhibit-size-directive -finstrument-functions @gol
889 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
890 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
891 -fno-common -fno-ident @gol
892 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
893 -fno-jump-tables @gol
894 -frecord-gcc-switches @gol
895 -freg-struct-return -fshort-enums @gol
896 -fshort-double -fshort-wchar @gol
897 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
898 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
899 -fno-stack-limit -fargument-alias -fargument-noalias @gol
900 -fargument-noalias-global -fargument-noalias-anything @gol
901 -fleading-underscore -ftls-model=@var{model} @gol
902 -ftrapv -fwrapv -fbounds-check @gol
907 * Overall Options:: Controlling the kind of output:
908 an executable, object files, assembler files,
909 or preprocessed source.
910 * C Dialect Options:: Controlling the variant of C language compiled.
911 * C++ Dialect Options:: Variations on C++.
912 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
914 * Language Independent Options:: Controlling how diagnostics should be
916 * Warning Options:: How picky should the compiler be?
917 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
918 * Optimize Options:: How much optimization?
919 * Preprocessor Options:: Controlling header files and macro definitions.
920 Also, getting dependency information for Make.
921 * Assembler Options:: Passing options to the assembler.
922 * Link Options:: Specifying libraries and so on.
923 * Directory Options:: Where to find header files and libraries.
924 Where to find the compiler executable files.
925 * Spec Files:: How to pass switches to sub-processes.
926 * Target Options:: Running a cross-compiler, or an old version of GCC.
929 @node Overall Options
930 @section Options Controlling the Kind of Output
932 Compilation can involve up to four stages: preprocessing, compilation
933 proper, assembly and linking, always in that order. GCC is capable of
934 preprocessing and compiling several files either into several
935 assembler input files, or into one assembler input file; then each
936 assembler input file produces an object file, and linking combines all
937 the object files (those newly compiled, and those specified as input)
938 into an executable file.
940 @cindex file name suffix
941 For any given input file, the file name suffix determines what kind of
946 C source code which must be preprocessed.
949 C source code which should not be preprocessed.
952 C++ source code which should not be preprocessed.
955 Objective-C source code. Note that you must link with the @file{libobjc}
956 library to make an Objective-C program work.
959 Objective-C source code which should not be preprocessed.
963 Objective-C++ source code. Note that you must link with the @file{libobjc}
964 library to make an Objective-C++ program work. Note that @samp{.M} refers
965 to a literal capital M@.
968 Objective-C++ source code which should not be preprocessed.
971 C, C++, Objective-C or Objective-C++ header file to be turned into a
976 @itemx @var{file}.cxx
977 @itemx @var{file}.cpp
978 @itemx @var{file}.CPP
979 @itemx @var{file}.c++
981 C++ source code which must be preprocessed. Note that in @samp{.cxx},
982 the last two letters must both be literally @samp{x}. Likewise,
983 @samp{.C} refers to a literal capital C@.
987 Objective-C++ source code which must be preprocessed.
990 Objective-C++ source code which should not be preprocessed.
995 @itemx @var{file}.hxx
996 @itemx @var{file}.hpp
997 @itemx @var{file}.HPP
998 @itemx @var{file}.h++
999 @itemx @var{file}.tcc
1000 C++ header file to be turned into a precompiled header.
1003 @itemx @var{file}.for
1004 @itemx @var{file}.ftn
1005 Fixed form Fortran source code which should not be preprocessed.
1008 @itemx @var{file}.FOR
1009 @itemx @var{file}.fpp
1010 @itemx @var{file}.FPP
1011 @itemx @var{file}.FTN
1012 Fixed form Fortran source code which must be preprocessed (with the traditional
1015 @item @var{file}.f90
1016 @itemx @var{file}.f95
1017 @itemx @var{file}.f03
1018 @itemx @var{file}.f08
1019 Free form Fortran source code which should not be preprocessed.
1021 @item @var{file}.F90
1022 @itemx @var{file}.F95
1023 @itemx @var{file}.F03
1024 @itemx @var{file}.F08
1025 Free form Fortran source code which must be preprocessed (with the
1026 traditional preprocessor).
1028 @c FIXME: Descriptions of Java file types.
1034 @item @var{file}.ads
1035 Ada source code file which contains a library unit declaration (a
1036 declaration of a package, subprogram, or generic, or a generic
1037 instantiation), or a library unit renaming declaration (a package,
1038 generic, or subprogram renaming declaration). Such files are also
1041 @item @var{file}.adb
1042 Ada source code file containing a library unit body (a subprogram or
1043 package body). Such files are also called @dfn{bodies}.
1045 @c GCC also knows about some suffixes for languages not yet included:
1056 @itemx @var{file}.sx
1057 Assembler code which must be preprocessed.
1060 An object file to be fed straight into linking.
1061 Any file name with no recognized suffix is treated this way.
1065 You can specify the input language explicitly with the @option{-x} option:
1068 @item -x @var{language}
1069 Specify explicitly the @var{language} for the following input files
1070 (rather than letting the compiler choose a default based on the file
1071 name suffix). This option applies to all following input files until
1072 the next @option{-x} option. Possible values for @var{language} are:
1074 c c-header c-cpp-output
1075 c++ c++-header c++-cpp-output
1076 objective-c objective-c-header objective-c-cpp-output
1077 objective-c++ objective-c++-header objective-c++-cpp-output
1078 assembler assembler-with-cpp
1080 f77 f77-cpp-input f95 f95-cpp-input
1085 Turn off any specification of a language, so that subsequent files are
1086 handled according to their file name suffixes (as they are if @option{-x}
1087 has not been used at all).
1089 @item -pass-exit-codes
1090 @opindex pass-exit-codes
1091 Normally the @command{gcc} program will exit with the code of 1 if any
1092 phase of the compiler returns a non-success return code. If you specify
1093 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1094 numerically highest error produced by any phase that returned an error
1095 indication. The C, C++, and Fortran frontends return 4, if an internal
1096 compiler error is encountered.
1099 If you only want some of the stages of compilation, you can use
1100 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1101 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1102 @command{gcc} is to stop. Note that some combinations (for example,
1103 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1108 Compile or assemble the source files, but do not link. The linking
1109 stage simply is not done. The ultimate output is in the form of an
1110 object file for each source file.
1112 By default, the object file name for a source file is made by replacing
1113 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1115 Unrecognized input files, not requiring compilation or assembly, are
1120 Stop after the stage of compilation proper; do not assemble. The output
1121 is in the form of an assembler code file for each non-assembler input
1124 By default, the assembler file name for a source file is made by
1125 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1127 Input files that don't require compilation are ignored.
1131 Stop after the preprocessing stage; do not run the compiler proper. The
1132 output is in the form of preprocessed source code, which is sent to the
1135 Input files which don't require preprocessing are ignored.
1137 @cindex output file option
1140 Place output in file @var{file}. This applies regardless to whatever
1141 sort of output is being produced, whether it be an executable file,
1142 an object file, an assembler file or preprocessed C code.
1144 If @option{-o} is not specified, the default is to put an executable
1145 file in @file{a.out}, the object file for
1146 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1147 assembler file in @file{@var{source}.s}, a precompiled header file in
1148 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1153 Print (on standard error output) the commands executed to run the stages
1154 of compilation. Also print the version number of the compiler driver
1155 program and of the preprocessor and the compiler proper.
1159 Like @option{-v} except the commands are not executed and all command
1160 arguments are quoted. This is useful for shell scripts to capture the
1161 driver-generated command lines.
1165 Use pipes rather than temporary files for communication between the
1166 various stages of compilation. This fails to work on some systems where
1167 the assembler is unable to read from a pipe; but the GNU assembler has
1172 If you are compiling multiple source files, this option tells the driver
1173 to pass all the source files to the compiler at once (for those
1174 languages for which the compiler can handle this). This will allow
1175 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1176 language for which this is supported is C@. If you pass source files for
1177 multiple languages to the driver, using this option, the driver will invoke
1178 the compiler(s) that support IMA once each, passing each compiler all the
1179 source files appropriate for it. For those languages that do not support
1180 IMA this option will be ignored, and the compiler will be invoked once for
1181 each source file in that language. If you use this option in conjunction
1182 with @option{-save-temps}, the compiler will generate multiple
1184 (one for each source file), but only one (combined) @file{.o} or
1189 Print (on the standard output) a description of the command line options
1190 understood by @command{gcc}. If the @option{-v} option is also specified
1191 then @option{--help} will also be passed on to the various processes
1192 invoked by @command{gcc}, so that they can display the command line options
1193 they accept. If the @option{-Wextra} option has also been specified
1194 (prior to the @option{--help} option), then command line options which
1195 have no documentation associated with them will also be displayed.
1198 @opindex target-help
1199 Print (on the standard output) a description of target-specific command
1200 line options for each tool. For some targets extra target-specific
1201 information may also be printed.
1203 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1204 Print (on the standard output) a description of the command line
1205 options understood by the compiler that fit into all specified classes
1206 and qualifiers. These are the supported classes:
1209 @item @samp{optimizers}
1210 This will display all of the optimization options supported by the
1213 @item @samp{warnings}
1214 This will display all of the options controlling warning messages
1215 produced by the compiler.
1218 This will display target-specific options. Unlike the
1219 @option{--target-help} option however, target-specific options of the
1220 linker and assembler will not be displayed. This is because those
1221 tools do not currently support the extended @option{--help=} syntax.
1224 This will display the values recognized by the @option{--param}
1227 @item @var{language}
1228 This will display the options supported for @var{language}, where
1229 @var{language} is the name of one of the languages supported in this
1233 This will display the options that are common to all languages.
1236 These are the supported qualifiers:
1239 @item @samp{undocumented}
1240 Display only those options which are undocumented.
1243 Display options which take an argument that appears after an equal
1244 sign in the same continuous piece of text, such as:
1245 @samp{--help=target}.
1247 @item @samp{separate}
1248 Display options which take an argument that appears as a separate word
1249 following the original option, such as: @samp{-o output-file}.
1252 Thus for example to display all the undocumented target-specific
1253 switches supported by the compiler the following can be used:
1256 --help=target,undocumented
1259 The sense of a qualifier can be inverted by prefixing it with the
1260 @samp{^} character, so for example to display all binary warning
1261 options (i.e., ones that are either on or off and that do not take an
1262 argument), which have a description the following can be used:
1265 --help=warnings,^joined,^undocumented
1268 The argument to @option{--help=} should not consist solely of inverted
1271 Combining several classes is possible, although this usually
1272 restricts the output by so much that there is nothing to display. One
1273 case where it does work however is when one of the classes is
1274 @var{target}. So for example to display all the target-specific
1275 optimization options the following can be used:
1278 --help=target,optimizers
1281 The @option{--help=} option can be repeated on the command line. Each
1282 successive use will display its requested class of options, skipping
1283 those that have already been displayed.
1285 If the @option{-Q} option appears on the command line before the
1286 @option{--help=} option, then the descriptive text displayed by
1287 @option{--help=} is changed. Instead of describing the displayed
1288 options, an indication is given as to whether the option is enabled,
1289 disabled or set to a specific value (assuming that the compiler
1290 knows this at the point where the @option{--help=} option is used).
1292 Here is a truncated example from the ARM port of @command{gcc}:
1295 % gcc -Q -mabi=2 --help=target -c
1296 The following options are target specific:
1298 -mabort-on-noreturn [disabled]
1302 The output is sensitive to the effects of previous command line
1303 options, so for example it is possible to find out which optimizations
1304 are enabled at @option{-O2} by using:
1307 -Q -O2 --help=optimizers
1310 Alternatively you can discover which binary optimizations are enabled
1311 by @option{-O3} by using:
1314 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1315 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1316 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1319 @item -no-canonical-prefixes
1320 @opindex no-canonical-prefixes
1321 Do not expand any symbolic links, resolve references to @samp{/../}
1322 or @samp{/./}, or make the path absolute when generating a relative
1327 Display the version number and copyrights of the invoked GCC@.
1331 Invoke all subcommands under a wrapper program. It takes a single
1332 comma separated list as an argument, which will be used to invoke
1336 gcc -c t.c -wrapper gdb,--args
1339 This will invoke all subprograms of gcc under "gdb --args",
1340 thus cc1 invocation will be "gdb --args cc1 ...".
1342 @item -fplugin=@var{name}.so
1343 Load the plugin code in file @var{name}.so, assumed to be a
1344 shared object to be dlopen'd by the compiler. The base name of
1345 the shared object file is used to identify the plugin for the
1346 purposes of argument parsing (See
1347 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1348 Each plugin should define the callback functions specified in the
1351 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1352 Define an argument called @var{key} with a value of @var{value}
1353 for the plugin called @var{name}.
1355 @include @value{srcdir}/../libiberty/at-file.texi
1359 @section Compiling C++ Programs
1361 @cindex suffixes for C++ source
1362 @cindex C++ source file suffixes
1363 C++ source files conventionally use one of the suffixes @samp{.C},
1364 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1365 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1366 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1367 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1368 files with these names and compiles them as C++ programs even if you
1369 call the compiler the same way as for compiling C programs (usually
1370 with the name @command{gcc}).
1374 However, the use of @command{gcc} does not add the C++ library.
1375 @command{g++} is a program that calls GCC and treats @samp{.c},
1376 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1377 files unless @option{-x} is used, and automatically specifies linking
1378 against the C++ library. This program is also useful when
1379 precompiling a C header file with a @samp{.h} extension for use in C++
1380 compilations. On many systems, @command{g++} is also installed with
1381 the name @command{c++}.
1383 @cindex invoking @command{g++}
1384 When you compile C++ programs, you may specify many of the same
1385 command-line options that you use for compiling programs in any
1386 language; or command-line options meaningful for C and related
1387 languages; or options that are meaningful only for C++ programs.
1388 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1389 explanations of options for languages related to C@.
1390 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1391 explanations of options that are meaningful only for C++ programs.
1393 @node C Dialect Options
1394 @section Options Controlling C Dialect
1395 @cindex dialect options
1396 @cindex language dialect options
1397 @cindex options, dialect
1399 The following options control the dialect of C (or languages derived
1400 from C, such as C++, Objective-C and Objective-C++) that the compiler
1404 @cindex ANSI support
1408 In C mode, this is equivalent to @samp{-std=c89}. In C++ mode, it is
1409 equivalent to @samp{-std=c++98}.
1411 This turns off certain features of GCC that are incompatible with ISO
1412 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1413 such as the @code{asm} and @code{typeof} keywords, and
1414 predefined macros such as @code{unix} and @code{vax} that identify the
1415 type of system you are using. It also enables the undesirable and
1416 rarely used ISO trigraph feature. For the C compiler,
1417 it disables recognition of C++ style @samp{//} comments as well as
1418 the @code{inline} keyword.
1420 The alternate keywords @code{__asm__}, @code{__extension__},
1421 @code{__inline__} and @code{__typeof__} continue to work despite
1422 @option{-ansi}. You would not want to use them in an ISO C program, of
1423 course, but it is useful to put them in header files that might be included
1424 in compilations done with @option{-ansi}. Alternate predefined macros
1425 such as @code{__unix__} and @code{__vax__} are also available, with or
1426 without @option{-ansi}.
1428 The @option{-ansi} option does not cause non-ISO programs to be
1429 rejected gratuitously. For that, @option{-pedantic} is required in
1430 addition to @option{-ansi}. @xref{Warning Options}.
1432 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1433 option is used. Some header files may notice this macro and refrain
1434 from declaring certain functions or defining certain macros that the
1435 ISO standard doesn't call for; this is to avoid interfering with any
1436 programs that might use these names for other things.
1438 Functions that would normally be built in but do not have semantics
1439 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1440 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1441 built-in functions provided by GCC}, for details of the functions
1446 Determine the language standard. @xref{Standards,,Language Standards
1447 Supported by GCC}, for details of these standard versions. This option
1448 is currently only supported when compiling C or C++.
1450 The compiler can accept several base standards, such as @samp{c89} or
1451 @samp{c++98}, and GNU dialects of those standards, such as
1452 @samp{gnu89} or @samp{gnu++98}. By specifying a base standard, the
1453 compiler will accept all programs following that standard and those
1454 using GNU extensions that do not contradict it. For example,
1455 @samp{-std=c89} turns off certain features of GCC that are
1456 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1457 keywords, but not other GNU extensions that do not have a meaning in
1458 ISO C90, such as omitting the middle term of a @code{?:}
1459 expression. On the other hand, by specifying a GNU dialect of a
1460 standard, all features the compiler support are enabled, even when
1461 those features change the meaning of the base standard and some
1462 strict-conforming programs may be rejected. The particular standard
1463 is used by @option{-pedantic} to identify which features are GNU
1464 extensions given that version of the standard. For example
1465 @samp{-std=gnu89 -pedantic} would warn about C++ style @samp{//}
1466 comments, while @samp{-std=gnu99 -pedantic} would not.
1468 A value for this option must be provided; possible values are
1473 Support all ISO C90 programs (certain GNU extensions that conflict
1474 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1476 @item iso9899:199409
1477 ISO C90 as modified in amendment 1.
1483 ISO C99. Note that this standard is not yet fully supported; see
1484 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1485 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1488 GNU dialect of ISO C90 (including some C99 features). This
1489 is the default for C code.
1493 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1494 this will become the default. The name @samp{gnu9x} is deprecated.
1497 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1501 GNU dialect of @option{-std=c++98}. This is the default for
1505 The working draft of the upcoming ISO C++0x standard. This option
1506 enables experimental features that are likely to be included in
1507 C++0x. The working draft is constantly changing, and any feature that is
1508 enabled by this flag may be removed from future versions of GCC if it is
1509 not part of the C++0x standard.
1512 GNU dialect of @option{-std=c++0x}. This option enables
1513 experimental features that may be removed in future versions of GCC.
1516 @item -fgnu89-inline
1517 @opindex fgnu89-inline
1518 The option @option{-fgnu89-inline} tells GCC to use the traditional
1519 GNU semantics for @code{inline} functions when in C99 mode.
1520 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1521 is accepted and ignored by GCC versions 4.1.3 up to but not including
1522 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1523 C99 mode. Using this option is roughly equivalent to adding the
1524 @code{gnu_inline} function attribute to all inline functions
1525 (@pxref{Function Attributes}).
1527 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1528 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1529 specifies the default behavior). This option was first supported in
1530 GCC 4.3. This option is not supported in C89 or gnu89 mode.
1532 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1533 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1534 in effect for @code{inline} functions. @xref{Common Predefined
1535 Macros,,,cpp,The C Preprocessor}.
1537 @item -aux-info @var{filename}
1539 Output to the given filename prototyped declarations for all functions
1540 declared and/or defined in a translation unit, including those in header
1541 files. This option is silently ignored in any language other than C@.
1543 Besides declarations, the file indicates, in comments, the origin of
1544 each declaration (source file and line), whether the declaration was
1545 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1546 @samp{O} for old, respectively, in the first character after the line
1547 number and the colon), and whether it came from a declaration or a
1548 definition (@samp{C} or @samp{F}, respectively, in the following
1549 character). In the case of function definitions, a K&R-style list of
1550 arguments followed by their declarations is also provided, inside
1551 comments, after the declaration.
1555 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1556 keyword, so that code can use these words as identifiers. You can use
1557 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1558 instead. @option{-ansi} implies @option{-fno-asm}.
1560 In C++, this switch only affects the @code{typeof} keyword, since
1561 @code{asm} and @code{inline} are standard keywords. You may want to
1562 use the @option{-fno-gnu-keywords} flag instead, which has the same
1563 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1564 switch only affects the @code{asm} and @code{typeof} keywords, since
1565 @code{inline} is a standard keyword in ISO C99.
1568 @itemx -fno-builtin-@var{function}
1569 @opindex fno-builtin
1570 @cindex built-in functions
1571 Don't recognize built-in functions that do not begin with
1572 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1573 functions provided by GCC}, for details of the functions affected,
1574 including those which are not built-in functions when @option{-ansi} or
1575 @option{-std} options for strict ISO C conformance are used because they
1576 do not have an ISO standard meaning.
1578 GCC normally generates special code to handle certain built-in functions
1579 more efficiently; for instance, calls to @code{alloca} may become single
1580 instructions that adjust the stack directly, and calls to @code{memcpy}
1581 may become inline copy loops. The resulting code is often both smaller
1582 and faster, but since the function calls no longer appear as such, you
1583 cannot set a breakpoint on those calls, nor can you change the behavior
1584 of the functions by linking with a different library. In addition,
1585 when a function is recognized as a built-in function, GCC may use
1586 information about that function to warn about problems with calls to
1587 that function, or to generate more efficient code, even if the
1588 resulting code still contains calls to that function. For example,
1589 warnings are given with @option{-Wformat} for bad calls to
1590 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1591 known not to modify global memory.
1593 With the @option{-fno-builtin-@var{function}} option
1594 only the built-in function @var{function} is
1595 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1596 function is named that is not built-in in this version of GCC, this
1597 option is ignored. There is no corresponding
1598 @option{-fbuiltin-@var{function}} option; if you wish to enable
1599 built-in functions selectively when using @option{-fno-builtin} or
1600 @option{-ffreestanding}, you may define macros such as:
1603 #define abs(n) __builtin_abs ((n))
1604 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1609 @cindex hosted environment
1611 Assert that compilation takes place in a hosted environment. This implies
1612 @option{-fbuiltin}. A hosted environment is one in which the
1613 entire standard library is available, and in which @code{main} has a return
1614 type of @code{int}. Examples are nearly everything except a kernel.
1615 This is equivalent to @option{-fno-freestanding}.
1617 @item -ffreestanding
1618 @opindex ffreestanding
1619 @cindex hosted environment
1621 Assert that compilation takes place in a freestanding environment. This
1622 implies @option{-fno-builtin}. A freestanding environment
1623 is one in which the standard library may not exist, and program startup may
1624 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1625 This is equivalent to @option{-fno-hosted}.
1627 @xref{Standards,,Language Standards Supported by GCC}, for details of
1628 freestanding and hosted environments.
1632 @cindex openmp parallel
1633 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1634 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1635 compiler generates parallel code according to the OpenMP Application
1636 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1637 implies @option{-pthread}, and thus is only supported on targets that
1638 have support for @option{-pthread}.
1640 @item -fms-extensions
1641 @opindex fms-extensions
1642 Accept some non-standard constructs used in Microsoft header files.
1644 Some cases of unnamed fields in structures and unions are only
1645 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1646 fields within structs/unions}, for details.
1650 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1651 options for strict ISO C conformance) implies @option{-trigraphs}.
1653 @item -no-integrated-cpp
1654 @opindex no-integrated-cpp
1655 Performs a compilation in two passes: preprocessing and compiling. This
1656 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1657 @option{-B} option. The user supplied compilation step can then add in
1658 an additional preprocessing step after normal preprocessing but before
1659 compiling. The default is to use the integrated cpp (internal cpp)
1661 The semantics of this option will change if "cc1", "cc1plus", and
1662 "cc1obj" are merged.
1664 @cindex traditional C language
1665 @cindex C language, traditional
1667 @itemx -traditional-cpp
1668 @opindex traditional-cpp
1669 @opindex traditional
1670 Formerly, these options caused GCC to attempt to emulate a pre-standard
1671 C compiler. They are now only supported with the @option{-E} switch.
1672 The preprocessor continues to support a pre-standard mode. See the GNU
1673 CPP manual for details.
1675 @item -fcond-mismatch
1676 @opindex fcond-mismatch
1677 Allow conditional expressions with mismatched types in the second and
1678 third arguments. The value of such an expression is void. This option
1679 is not supported for C++.
1681 @item -flax-vector-conversions
1682 @opindex flax-vector-conversions
1683 Allow implicit conversions between vectors with differing numbers of
1684 elements and/or incompatible element types. This option should not be
1687 @item -funsigned-char
1688 @opindex funsigned-char
1689 Let the type @code{char} be unsigned, like @code{unsigned char}.
1691 Each kind of machine has a default for what @code{char} should
1692 be. It is either like @code{unsigned char} by default or like
1693 @code{signed char} by default.
1695 Ideally, a portable program should always use @code{signed char} or
1696 @code{unsigned char} when it depends on the signedness of an object.
1697 But many programs have been written to use plain @code{char} and
1698 expect it to be signed, or expect it to be unsigned, depending on the
1699 machines they were written for. This option, and its inverse, let you
1700 make such a program work with the opposite default.
1702 The type @code{char} is always a distinct type from each of
1703 @code{signed char} or @code{unsigned char}, even though its behavior
1704 is always just like one of those two.
1707 @opindex fsigned-char
1708 Let the type @code{char} be signed, like @code{signed char}.
1710 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1711 the negative form of @option{-funsigned-char}. Likewise, the option
1712 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1714 @item -fsigned-bitfields
1715 @itemx -funsigned-bitfields
1716 @itemx -fno-signed-bitfields
1717 @itemx -fno-unsigned-bitfields
1718 @opindex fsigned-bitfields
1719 @opindex funsigned-bitfields
1720 @opindex fno-signed-bitfields
1721 @opindex fno-unsigned-bitfields
1722 These options control whether a bit-field is signed or unsigned, when the
1723 declaration does not use either @code{signed} or @code{unsigned}. By
1724 default, such a bit-field is signed, because this is consistent: the
1725 basic integer types such as @code{int} are signed types.
1728 @node C++ Dialect Options
1729 @section Options Controlling C++ Dialect
1731 @cindex compiler options, C++
1732 @cindex C++ options, command line
1733 @cindex options, C++
1734 This section describes the command-line options that are only meaningful
1735 for C++ programs; but you can also use most of the GNU compiler options
1736 regardless of what language your program is in. For example, you
1737 might compile a file @code{firstClass.C} like this:
1740 g++ -g -frepo -O -c firstClass.C
1744 In this example, only @option{-frepo} is an option meant
1745 only for C++ programs; you can use the other options with any
1746 language supported by GCC@.
1748 Here is a list of options that are @emph{only} for compiling C++ programs:
1752 @item -fabi-version=@var{n}
1753 @opindex fabi-version
1754 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1755 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1756 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1757 the version that conforms most closely to the C++ ABI specification.
1758 Therefore, the ABI obtained using version 0 will change as ABI bugs
1761 The default is version 2.
1763 @item -fno-access-control
1764 @opindex fno-access-control
1765 Turn off all access checking. This switch is mainly useful for working
1766 around bugs in the access control code.
1770 Check that the pointer returned by @code{operator new} is non-null
1771 before attempting to modify the storage allocated. This check is
1772 normally unnecessary because the C++ standard specifies that
1773 @code{operator new} will only return @code{0} if it is declared
1774 @samp{throw()}, in which case the compiler will always check the
1775 return value even without this option. In all other cases, when
1776 @code{operator new} has a non-empty exception specification, memory
1777 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1778 @samp{new (nothrow)}.
1780 @item -fconserve-space
1781 @opindex fconserve-space
1782 Put uninitialized or runtime-initialized global variables into the
1783 common segment, as C does. This saves space in the executable at the
1784 cost of not diagnosing duplicate definitions. If you compile with this
1785 flag and your program mysteriously crashes after @code{main()} has
1786 completed, you may have an object that is being destroyed twice because
1787 two definitions were merged.
1789 This option is no longer useful on most targets, now that support has
1790 been added for putting variables into BSS without making them common.
1792 @item -fno-deduce-init-list
1793 @opindex fno-deduce-init-list
1794 Disable deduction of a template type parameter as
1795 std::initializer_list from a brace-enclosed initializer list, i.e.
1798 template <class T> auto forward(T t) -> decltype (realfn (t))
1805 forward(@{1,2@}); // call forward<std::initializer_list<int>>
1809 This option is present because this deduction is an extension to the
1810 current specification in the C++0x working draft, and there was
1811 some concern about potential overload resolution problems.
1813 @item -ffriend-injection
1814 @opindex ffriend-injection
1815 Inject friend functions into the enclosing namespace, so that they are
1816 visible outside the scope of the class in which they are declared.
1817 Friend functions were documented to work this way in the old Annotated
1818 C++ Reference Manual, and versions of G++ before 4.1 always worked
1819 that way. However, in ISO C++ a friend function which is not declared
1820 in an enclosing scope can only be found using argument dependent
1821 lookup. This option causes friends to be injected as they were in
1824 This option is for compatibility, and may be removed in a future
1827 @item -fno-elide-constructors
1828 @opindex fno-elide-constructors
1829 The C++ standard allows an implementation to omit creating a temporary
1830 which is only used to initialize another object of the same type.
1831 Specifying this option disables that optimization, and forces G++ to
1832 call the copy constructor in all cases.
1834 @item -fno-enforce-eh-specs
1835 @opindex fno-enforce-eh-specs
1836 Don't generate code to check for violation of exception specifications
1837 at runtime. This option violates the C++ standard, but may be useful
1838 for reducing code size in production builds, much like defining
1839 @samp{NDEBUG}. This does not give user code permission to throw
1840 exceptions in violation of the exception specifications; the compiler
1841 will still optimize based on the specifications, so throwing an
1842 unexpected exception will result in undefined behavior.
1845 @itemx -fno-for-scope
1847 @opindex fno-for-scope
1848 If @option{-ffor-scope} is specified, the scope of variables declared in
1849 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1850 as specified by the C++ standard.
1851 If @option{-fno-for-scope} is specified, the scope of variables declared in
1852 a @i{for-init-statement} extends to the end of the enclosing scope,
1853 as was the case in old versions of G++, and other (traditional)
1854 implementations of C++.
1856 The default if neither flag is given to follow the standard,
1857 but to allow and give a warning for old-style code that would
1858 otherwise be invalid, or have different behavior.
1860 @item -fno-gnu-keywords
1861 @opindex fno-gnu-keywords
1862 Do not recognize @code{typeof} as a keyword, so that code can use this
1863 word as an identifier. You can use the keyword @code{__typeof__} instead.
1864 @option{-ansi} implies @option{-fno-gnu-keywords}.
1866 @item -fno-implicit-templates
1867 @opindex fno-implicit-templates
1868 Never emit code for non-inline templates which are instantiated
1869 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1870 @xref{Template Instantiation}, for more information.
1872 @item -fno-implicit-inline-templates
1873 @opindex fno-implicit-inline-templates
1874 Don't emit code for implicit instantiations of inline templates, either.
1875 The default is to handle inlines differently so that compiles with and
1876 without optimization will need the same set of explicit instantiations.
1878 @item -fno-implement-inlines
1879 @opindex fno-implement-inlines
1880 To save space, do not emit out-of-line copies of inline functions
1881 controlled by @samp{#pragma implementation}. This will cause linker
1882 errors if these functions are not inlined everywhere they are called.
1884 @item -fms-extensions
1885 @opindex fms-extensions
1886 Disable pedantic warnings about constructs used in MFC, such as implicit
1887 int and getting a pointer to member function via non-standard syntax.
1889 @item -fno-nonansi-builtins
1890 @opindex fno-nonansi-builtins
1891 Disable built-in declarations of functions that are not mandated by
1892 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1893 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1895 @item -fno-operator-names
1896 @opindex fno-operator-names
1897 Do not treat the operator name keywords @code{and}, @code{bitand},
1898 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1899 synonyms as keywords.
1901 @item -fno-optional-diags
1902 @opindex fno-optional-diags
1903 Disable diagnostics that the standard says a compiler does not need to
1904 issue. Currently, the only such diagnostic issued by G++ is the one for
1905 a name having multiple meanings within a class.
1908 @opindex fpermissive
1909 Downgrade some diagnostics about nonconformant code from errors to
1910 warnings. Thus, using @option{-fpermissive} will allow some
1911 nonconforming code to compile.
1913 @item -fno-pretty-templates
1914 @opindex fno-pretty-templates
1915 When an error message refers to a specialization of a function
1916 template, the compiler will normally print the signature of the
1917 template followed by the template arguments and any typedefs or
1918 typenames in the signature (e.g. @code{void f(T) [with T = int]}
1919 rather than @code{void f(int)}) so that it's clear which template is
1920 involved. When an error message refers to a specialization of a class
1921 template, the compiler will omit any template arguments which match
1922 the default template arguments for that template. If either of these
1923 behaviors make it harder to understand the error message rather than
1924 easier, using @option{-fno-pretty-templates} will disable them.
1928 Enable automatic template instantiation at link time. This option also
1929 implies @option{-fno-implicit-templates}. @xref{Template
1930 Instantiation}, for more information.
1934 Disable generation of information about every class with virtual
1935 functions for use by the C++ runtime type identification features
1936 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1937 of the language, you can save some space by using this flag. Note that
1938 exception handling uses the same information, but it will generate it as
1939 needed. The @samp{dynamic_cast} operator can still be used for casts that
1940 do not require runtime type information, i.e.@: casts to @code{void *} or to
1941 unambiguous base classes.
1945 Emit statistics about front-end processing at the end of the compilation.
1946 This information is generally only useful to the G++ development team.
1948 @item -ftemplate-depth-@var{n}
1949 @opindex ftemplate-depth
1950 Set the maximum instantiation depth for template classes to @var{n}.
1951 A limit on the template instantiation depth is needed to detect
1952 endless recursions during template class instantiation. ANSI/ISO C++
1953 conforming programs must not rely on a maximum depth greater than 17
1954 (changed to 1024 in C++0x).
1956 @item -fno-threadsafe-statics
1957 @opindex fno-threadsafe-statics
1958 Do not emit the extra code to use the routines specified in the C++
1959 ABI for thread-safe initialization of local statics. You can use this
1960 option to reduce code size slightly in code that doesn't need to be
1963 @item -fuse-cxa-atexit
1964 @opindex fuse-cxa-atexit
1965 Register destructors for objects with static storage duration with the
1966 @code{__cxa_atexit} function rather than the @code{atexit} function.
1967 This option is required for fully standards-compliant handling of static
1968 destructors, but will only work if your C library supports
1969 @code{__cxa_atexit}.
1971 @item -fno-use-cxa-get-exception-ptr
1972 @opindex fno-use-cxa-get-exception-ptr
1973 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
1974 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
1975 if the runtime routine is not available.
1977 @item -fvisibility-inlines-hidden
1978 @opindex fvisibility-inlines-hidden
1979 This switch declares that the user does not attempt to compare
1980 pointers to inline methods where the addresses of the two functions
1981 were taken in different shared objects.
1983 The effect of this is that GCC may, effectively, mark inline methods with
1984 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1985 appear in the export table of a DSO and do not require a PLT indirection
1986 when used within the DSO@. Enabling this option can have a dramatic effect
1987 on load and link times of a DSO as it massively reduces the size of the
1988 dynamic export table when the library makes heavy use of templates.
1990 The behavior of this switch is not quite the same as marking the
1991 methods as hidden directly, because it does not affect static variables
1992 local to the function or cause the compiler to deduce that
1993 the function is defined in only one shared object.
1995 You may mark a method as having a visibility explicitly to negate the
1996 effect of the switch for that method. For example, if you do want to
1997 compare pointers to a particular inline method, you might mark it as
1998 having default visibility. Marking the enclosing class with explicit
1999 visibility will have no effect.
2001 Explicitly instantiated inline methods are unaffected by this option
2002 as their linkage might otherwise cross a shared library boundary.
2003 @xref{Template Instantiation}.
2005 @item -fvisibility-ms-compat
2006 @opindex fvisibility-ms-compat
2007 This flag attempts to use visibility settings to make GCC's C++
2008 linkage model compatible with that of Microsoft Visual Studio.
2010 The flag makes these changes to GCC's linkage model:
2014 It sets the default visibility to @code{hidden}, like
2015 @option{-fvisibility=hidden}.
2018 Types, but not their members, are not hidden by default.
2021 The One Definition Rule is relaxed for types without explicit
2022 visibility specifications which are defined in more than one different
2023 shared object: those declarations are permitted if they would have
2024 been permitted when this option was not used.
2027 In new code it is better to use @option{-fvisibility=hidden} and
2028 export those classes which are intended to be externally visible.
2029 Unfortunately it is possible for code to rely, perhaps accidentally,
2030 on the Visual Studio behavior.
2032 Among the consequences of these changes are that static data members
2033 of the same type with the same name but defined in different shared
2034 objects will be different, so changing one will not change the other;
2035 and that pointers to function members defined in different shared
2036 objects may not compare equal. When this flag is given, it is a
2037 violation of the ODR to define types with the same name differently.
2041 Do not use weak symbol support, even if it is provided by the linker.
2042 By default, G++ will use weak symbols if they are available. This
2043 option exists only for testing, and should not be used by end-users;
2044 it will result in inferior code and has no benefits. This option may
2045 be removed in a future release of G++.
2049 Do not search for header files in the standard directories specific to
2050 C++, but do still search the other standard directories. (This option
2051 is used when building the C++ library.)
2054 In addition, these optimization, warning, and code generation options
2055 have meanings only for C++ programs:
2058 @item -fno-default-inline
2059 @opindex fno-default-inline
2060 Do not assume @samp{inline} for functions defined inside a class scope.
2061 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2062 functions will have linkage like inline functions; they just won't be
2065 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2068 Warn when G++ generates code that is probably not compatible with the
2069 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2070 all such cases, there are probably some cases that are not warned about,
2071 even though G++ is generating incompatible code. There may also be
2072 cases where warnings are emitted even though the code that is generated
2075 You should rewrite your code to avoid these warnings if you are
2076 concerned about the fact that code generated by G++ may not be binary
2077 compatible with code generated by other compilers.
2079 The known incompatibilities at this point include:
2084 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2085 pack data into the same byte as a base class. For example:
2088 struct A @{ virtual void f(); int f1 : 1; @};
2089 struct B : public A @{ int f2 : 1; @};
2093 In this case, G++ will place @code{B::f2} into the same byte
2094 as@code{A::f1}; other compilers will not. You can avoid this problem
2095 by explicitly padding @code{A} so that its size is a multiple of the
2096 byte size on your platform; that will cause G++ and other compilers to
2097 layout @code{B} identically.
2100 Incorrect handling of tail-padding for virtual bases. G++ does not use
2101 tail padding when laying out virtual bases. For example:
2104 struct A @{ virtual void f(); char c1; @};
2105 struct B @{ B(); char c2; @};
2106 struct C : public A, public virtual B @{@};
2110 In this case, G++ will not place @code{B} into the tail-padding for
2111 @code{A}; other compilers will. You can avoid this problem by
2112 explicitly padding @code{A} so that its size is a multiple of its
2113 alignment (ignoring virtual base classes); that will cause G++ and other
2114 compilers to layout @code{C} identically.
2117 Incorrect handling of bit-fields with declared widths greater than that
2118 of their underlying types, when the bit-fields appear in a union. For
2122 union U @{ int i : 4096; @};
2126 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2127 union too small by the number of bits in an @code{int}.
2130 Empty classes can be placed at incorrect offsets. For example:
2140 struct C : public B, public A @{@};
2144 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2145 it should be placed at offset zero. G++ mistakenly believes that the
2146 @code{A} data member of @code{B} is already at offset zero.
2149 Names of template functions whose types involve @code{typename} or
2150 template template parameters can be mangled incorrectly.
2153 template <typename Q>
2154 void f(typename Q::X) @{@}
2156 template <template <typename> class Q>
2157 void f(typename Q<int>::X) @{@}
2161 Instantiations of these templates may be mangled incorrectly.
2165 It also warns psABI related changes. The known psABI changes at this
2171 For SYSV/x86-64, when passing union with long double, it is changed to
2172 pass in memory as specified in psABI. For example:
2182 @code{union U} will always be passed in memory.
2186 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2187 @opindex Wctor-dtor-privacy
2188 @opindex Wno-ctor-dtor-privacy
2189 Warn when a class seems unusable because all the constructors or
2190 destructors in that class are private, and it has neither friends nor
2191 public static member functions.
2193 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2194 @opindex Wnon-virtual-dtor
2195 @opindex Wno-non-virtual-dtor
2196 Warn when a class has virtual functions and accessible non-virtual
2197 destructor, in which case it would be possible but unsafe to delete
2198 an instance of a derived class through a pointer to the base class.
2199 This warning is also enabled if -Weffc++ is specified.
2201 @item -Wreorder @r{(C++ and Objective-C++ only)}
2203 @opindex Wno-reorder
2204 @cindex reordering, warning
2205 @cindex warning for reordering of member initializers
2206 Warn when the order of member initializers given in the code does not
2207 match the order in which they must be executed. For instance:
2213 A(): j (0), i (1) @{ @}
2217 The compiler will rearrange the member initializers for @samp{i}
2218 and @samp{j} to match the declaration order of the members, emitting
2219 a warning to that effect. This warning is enabled by @option{-Wall}.
2222 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2225 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2228 Warn about violations of the following style guidelines from Scott Meyers'
2229 @cite{Effective C++} book:
2233 Item 11: Define a copy constructor and an assignment operator for classes
2234 with dynamically allocated memory.
2237 Item 12: Prefer initialization to assignment in constructors.
2240 Item 14: Make destructors virtual in base classes.
2243 Item 15: Have @code{operator=} return a reference to @code{*this}.
2246 Item 23: Don't try to return a reference when you must return an object.
2250 Also warn about violations of the following style guidelines from
2251 Scott Meyers' @cite{More Effective C++} book:
2255 Item 6: Distinguish between prefix and postfix forms of increment and
2256 decrement operators.
2259 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2263 When selecting this option, be aware that the standard library
2264 headers do not obey all of these guidelines; use @samp{grep -v}
2265 to filter out those warnings.
2267 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2268 @opindex Wstrict-null-sentinel
2269 @opindex Wno-strict-null-sentinel
2270 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2271 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2272 to @code{__null}. Although it is a null pointer constant not a null pointer,
2273 it is guaranteed to be of the same size as a pointer. But this use is
2274 not portable across different compilers.
2276 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2277 @opindex Wno-non-template-friend
2278 @opindex Wnon-template-friend
2279 Disable warnings when non-templatized friend functions are declared
2280 within a template. Since the advent of explicit template specification
2281 support in G++, if the name of the friend is an unqualified-id (i.e.,
2282 @samp{friend foo(int)}), the C++ language specification demands that the
2283 friend declare or define an ordinary, nontemplate function. (Section
2284 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2285 could be interpreted as a particular specialization of a templatized
2286 function. Because this non-conforming behavior is no longer the default
2287 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2288 check existing code for potential trouble spots and is on by default.
2289 This new compiler behavior can be turned off with
2290 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2291 but disables the helpful warning.
2293 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2294 @opindex Wold-style-cast
2295 @opindex Wno-old-style-cast
2296 Warn if an old-style (C-style) cast to a non-void type is used within
2297 a C++ program. The new-style casts (@samp{dynamic_cast},
2298 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2299 less vulnerable to unintended effects and much easier to search for.
2301 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2302 @opindex Woverloaded-virtual
2303 @opindex Wno-overloaded-virtual
2304 @cindex overloaded virtual fn, warning
2305 @cindex warning for overloaded virtual fn
2306 Warn when a function declaration hides virtual functions from a
2307 base class. For example, in:
2314 struct B: public A @{
2319 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2327 will fail to compile.
2329 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2330 @opindex Wno-pmf-conversions
2331 @opindex Wpmf-conversions
2332 Disable the diagnostic for converting a bound pointer to member function
2335 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2336 @opindex Wsign-promo
2337 @opindex Wno-sign-promo
2338 Warn when overload resolution chooses a promotion from unsigned or
2339 enumerated type to a signed type, over a conversion to an unsigned type of
2340 the same size. Previous versions of G++ would try to preserve
2341 unsignedness, but the standard mandates the current behavior.
2346 A& operator = (int);
2356 In this example, G++ will synthesize a default @samp{A& operator =
2357 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2360 @node Objective-C and Objective-C++ Dialect Options
2361 @section Options Controlling Objective-C and Objective-C++ Dialects
2363 @cindex compiler options, Objective-C and Objective-C++
2364 @cindex Objective-C and Objective-C++ options, command line
2365 @cindex options, Objective-C and Objective-C++
2366 (NOTE: This manual does not describe the Objective-C and Objective-C++
2367 languages themselves. See @xref{Standards,,Language Standards
2368 Supported by GCC}, for references.)
2370 This section describes the command-line options that are only meaningful
2371 for Objective-C and Objective-C++ programs, but you can also use most of
2372 the language-independent GNU compiler options.
2373 For example, you might compile a file @code{some_class.m} like this:
2376 gcc -g -fgnu-runtime -O -c some_class.m
2380 In this example, @option{-fgnu-runtime} is an option meant only for
2381 Objective-C and Objective-C++ programs; you can use the other options with
2382 any language supported by GCC@.
2384 Note that since Objective-C is an extension of the C language, Objective-C
2385 compilations may also use options specific to the C front-end (e.g.,
2386 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2387 C++-specific options (e.g., @option{-Wabi}).
2389 Here is a list of options that are @emph{only} for compiling Objective-C
2390 and Objective-C++ programs:
2393 @item -fconstant-string-class=@var{class-name}
2394 @opindex fconstant-string-class
2395 Use @var{class-name} as the name of the class to instantiate for each
2396 literal string specified with the syntax @code{@@"@dots{}"}. The default
2397 class name is @code{NXConstantString} if the GNU runtime is being used, and
2398 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2399 @option{-fconstant-cfstrings} option, if also present, will override the
2400 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2401 to be laid out as constant CoreFoundation strings.
2404 @opindex fgnu-runtime
2405 Generate object code compatible with the standard GNU Objective-C
2406 runtime. This is the default for most types of systems.
2408 @item -fnext-runtime
2409 @opindex fnext-runtime
2410 Generate output compatible with the NeXT runtime. This is the default
2411 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2412 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2415 @item -fno-nil-receivers
2416 @opindex fno-nil-receivers
2417 Assume that all Objective-C message dispatches (e.g.,
2418 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
2419 is not @code{nil}. This allows for more efficient entry points in the runtime
2420 to be used. Currently, this option is only available in conjunction with
2421 the NeXT runtime on Mac OS X 10.3 and later.
2423 @item -fobjc-call-cxx-cdtors
2424 @opindex fobjc-call-cxx-cdtors
2425 For each Objective-C class, check if any of its instance variables is a
2426 C++ object with a non-trivial default constructor. If so, synthesize a
2427 special @code{- (id) .cxx_construct} instance method that will run
2428 non-trivial default constructors on any such instance variables, in order,
2429 and then return @code{self}. Similarly, check if any instance variable
2430 is a C++ object with a non-trivial destructor, and if so, synthesize a
2431 special @code{- (void) .cxx_destruct} method that will run
2432 all such default destructors, in reverse order.
2434 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
2435 thusly generated will only operate on instance variables declared in the
2436 current Objective-C class, and not those inherited from superclasses. It
2437 is the responsibility of the Objective-C runtime to invoke all such methods
2438 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
2439 will be invoked by the runtime immediately after a new object
2440 instance is allocated; the @code{- (void) .cxx_destruct} methods will
2441 be invoked immediately before the runtime deallocates an object instance.
2443 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2444 support for invoking the @code{- (id) .cxx_construct} and
2445 @code{- (void) .cxx_destruct} methods.
2447 @item -fobjc-direct-dispatch
2448 @opindex fobjc-direct-dispatch
2449 Allow fast jumps to the message dispatcher. On Darwin this is
2450 accomplished via the comm page.
2452 @item -fobjc-exceptions
2453 @opindex fobjc-exceptions
2454 Enable syntactic support for structured exception handling in Objective-C,
2455 similar to what is offered by C++ and Java. This option is
2456 unavailable in conjunction with the NeXT runtime on Mac OS X 10.2 and
2465 @@catch (AnObjCClass *exc) @{
2472 @@catch (AnotherClass *exc) @{
2475 @@catch (id allOthers) @{
2485 The @code{@@throw} statement may appear anywhere in an Objective-C or
2486 Objective-C++ program; when used inside of a @code{@@catch} block, the
2487 @code{@@throw} may appear without an argument (as shown above), in which case
2488 the object caught by the @code{@@catch} will be rethrown.
2490 Note that only (pointers to) Objective-C objects may be thrown and
2491 caught using this scheme. When an object is thrown, it will be caught
2492 by the nearest @code{@@catch} clause capable of handling objects of that type,
2493 analogously to how @code{catch} blocks work in C++ and Java. A
2494 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2495 any and all Objective-C exceptions not caught by previous @code{@@catch}
2498 The @code{@@finally} clause, if present, will be executed upon exit from the
2499 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2500 regardless of whether any exceptions are thrown, caught or rethrown
2501 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2502 of the @code{finally} clause in Java.
2504 There are several caveats to using the new exception mechanism:
2508 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2509 idioms provided by the @code{NSException} class, the new
2510 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2511 systems, due to additional functionality needed in the (NeXT) Objective-C
2515 As mentioned above, the new exceptions do not support handling
2516 types other than Objective-C objects. Furthermore, when used from
2517 Objective-C++, the Objective-C exception model does not interoperate with C++
2518 exceptions at this time. This means you cannot @code{@@throw} an exception
2519 from Objective-C and @code{catch} it in C++, or vice versa
2520 (i.e., @code{throw @dots{} @@catch}).
2523 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2524 blocks for thread-safe execution:
2527 @@synchronized (ObjCClass *guard) @{
2532 Upon entering the @code{@@synchronized} block, a thread of execution shall
2533 first check whether a lock has been placed on the corresponding @code{guard}
2534 object by another thread. If it has, the current thread shall wait until
2535 the other thread relinquishes its lock. Once @code{guard} becomes available,
2536 the current thread will place its own lock on it, execute the code contained in
2537 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2538 making @code{guard} available to other threads).
2540 Unlike Java, Objective-C does not allow for entire methods to be marked
2541 @code{@@synchronized}. Note that throwing exceptions out of
2542 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2543 to be unlocked properly.
2547 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2549 @item -freplace-objc-classes
2550 @opindex freplace-objc-classes
2551 Emit a special marker instructing @command{ld(1)} not to statically link in
2552 the resulting object file, and allow @command{dyld(1)} to load it in at
2553 run time instead. This is used in conjunction with the Fix-and-Continue
2554 debugging mode, where the object file in question may be recompiled and
2555 dynamically reloaded in the course of program execution, without the need
2556 to restart the program itself. Currently, Fix-and-Continue functionality
2557 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2562 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2563 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2564 compile time) with static class references that get initialized at load time,
2565 which improves run-time performance. Specifying the @option{-fzero-link} flag
2566 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2567 to be retained. This is useful in Zero-Link debugging mode, since it allows
2568 for individual class implementations to be modified during program execution.
2572 Dump interface declarations for all classes seen in the source file to a
2573 file named @file{@var{sourcename}.decl}.
2575 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2576 @opindex Wassign-intercept
2577 @opindex Wno-assign-intercept
2578 Warn whenever an Objective-C assignment is being intercepted by the
2581 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2582 @opindex Wno-protocol
2584 If a class is declared to implement a protocol, a warning is issued for
2585 every method in the protocol that is not implemented by the class. The
2586 default behavior is to issue a warning for every method not explicitly
2587 implemented in the class, even if a method implementation is inherited
2588 from the superclass. If you use the @option{-Wno-protocol} option, then
2589 methods inherited from the superclass are considered to be implemented,
2590 and no warning is issued for them.
2592 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2594 @opindex Wno-selector
2595 Warn if multiple methods of different types for the same selector are
2596 found during compilation. The check is performed on the list of methods
2597 in the final stage of compilation. Additionally, a check is performed
2598 for each selector appearing in a @code{@@selector(@dots{})}
2599 expression, and a corresponding method for that selector has been found
2600 during compilation. Because these checks scan the method table only at
2601 the end of compilation, these warnings are not produced if the final
2602 stage of compilation is not reached, for example because an error is
2603 found during compilation, or because the @option{-fsyntax-only} option is
2606 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2607 @opindex Wstrict-selector-match
2608 @opindex Wno-strict-selector-match
2609 Warn if multiple methods with differing argument and/or return types are
2610 found for a given selector when attempting to send a message using this
2611 selector to a receiver of type @code{id} or @code{Class}. When this flag
2612 is off (which is the default behavior), the compiler will omit such warnings
2613 if any differences found are confined to types which share the same size
2616 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2617 @opindex Wundeclared-selector
2618 @opindex Wno-undeclared-selector
2619 Warn if a @code{@@selector(@dots{})} expression referring to an
2620 undeclared selector is found. A selector is considered undeclared if no
2621 method with that name has been declared before the
2622 @code{@@selector(@dots{})} expression, either explicitly in an
2623 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2624 an @code{@@implementation} section. This option always performs its
2625 checks as soon as a @code{@@selector(@dots{})} expression is found,
2626 while @option{-Wselector} only performs its checks in the final stage of
2627 compilation. This also enforces the coding style convention
2628 that methods and selectors must be declared before being used.
2630 @item -print-objc-runtime-info
2631 @opindex print-objc-runtime-info
2632 Generate C header describing the largest structure that is passed by
2637 @node Language Independent Options
2638 @section Options to Control Diagnostic Messages Formatting
2639 @cindex options to control diagnostics formatting
2640 @cindex diagnostic messages
2641 @cindex message formatting
2643 Traditionally, diagnostic messages have been formatted irrespective of
2644 the output device's aspect (e.g.@: its width, @dots{}). The options described
2645 below can be used to control the diagnostic messages formatting
2646 algorithm, e.g.@: how many characters per line, how often source location
2647 information should be reported. Right now, only the C++ front end can
2648 honor these options. However it is expected, in the near future, that
2649 the remaining front ends would be able to digest them correctly.
2652 @item -fmessage-length=@var{n}
2653 @opindex fmessage-length
2654 Try to format error messages so that they fit on lines of about @var{n}
2655 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2656 the front ends supported by GCC@. If @var{n} is zero, then no
2657 line-wrapping will be done; each error message will appear on a single
2660 @opindex fdiagnostics-show-location
2661 @item -fdiagnostics-show-location=once
2662 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2663 reporter to emit @emph{once} source location information; that is, in
2664 case the message is too long to fit on a single physical line and has to
2665 be wrapped, the source location won't be emitted (as prefix) again,
2666 over and over, in subsequent continuation lines. This is the default
2669 @item -fdiagnostics-show-location=every-line
2670 Only meaningful in line-wrapping mode. Instructs the diagnostic
2671 messages reporter to emit the same source location information (as
2672 prefix) for physical lines that result from the process of breaking
2673 a message which is too long to fit on a single line.
2675 @item -fdiagnostics-show-option
2676 @opindex fdiagnostics-show-option
2677 This option instructs the diagnostic machinery to add text to each
2678 diagnostic emitted, which indicates which command line option directly
2679 controls that diagnostic, when such an option is known to the
2680 diagnostic machinery.
2682 @item -Wcoverage-mismatch
2683 @opindex Wcoverage-mismatch
2684 Warn if feedback profiles do not match when using the
2685 @option{-fprofile-use} option.
2686 If a source file was changed between @option{-fprofile-gen} and
2687 @option{-fprofile-use}, the files with the profile feedback can fail
2688 to match the source file and GCC can not use the profile feedback
2689 information. By default, GCC emits an error message in this case.
2690 The option @option{-Wcoverage-mismatch} emits a warning instead of an
2691 error. GCC does not use appropriate feedback profiles, so using this
2692 option can result in poorly optimized code. This option is useful
2693 only in the case of very minor changes such as bug fixes to an
2698 @node Warning Options
2699 @section Options to Request or Suppress Warnings
2700 @cindex options to control warnings
2701 @cindex warning messages
2702 @cindex messages, warning
2703 @cindex suppressing warnings
2705 Warnings are diagnostic messages that report constructions which
2706 are not inherently erroneous but which are risky or suggest there
2707 may have been an error.
2709 The following language-independent options do not enable specific
2710 warnings but control the kinds of diagnostics produced by GCC.
2713 @cindex syntax checking
2715 @opindex fsyntax-only
2716 Check the code for syntax errors, but don't do anything beyond that.
2720 Inhibit all warning messages.
2725 Make all warnings into errors.
2730 Make the specified warning into an error. The specifier for a warning
2731 is appended, for example @option{-Werror=switch} turns the warnings
2732 controlled by @option{-Wswitch} into errors. This switch takes a
2733 negative form, to be used to negate @option{-Werror} for specific
2734 warnings, for example @option{-Wno-error=switch} makes
2735 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2736 is in effect. You can use the @option{-fdiagnostics-show-option}
2737 option to have each controllable warning amended with the option which
2738 controls it, to determine what to use with this option.
2740 Note that specifying @option{-Werror=}@var{foo} automatically implies
2741 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2744 @item -Wfatal-errors
2745 @opindex Wfatal-errors
2746 @opindex Wno-fatal-errors
2747 This option causes the compiler to abort compilation on the first error
2748 occurred rather than trying to keep going and printing further error
2753 You can request many specific warnings with options beginning
2754 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2755 implicit declarations. Each of these specific warning options also
2756 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2757 example, @option{-Wno-implicit}. This manual lists only one of the
2758 two forms, whichever is not the default. For further,
2759 language-specific options also refer to @ref{C++ Dialect Options} and
2760 @ref{Objective-C and Objective-C++ Dialect Options}.
2765 Issue all the warnings demanded by strict ISO C and ISO C++;
2766 reject all programs that use forbidden extensions, and some other
2767 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2768 version of the ISO C standard specified by any @option{-std} option used.
2770 Valid ISO C and ISO C++ programs should compile properly with or without
2771 this option (though a rare few will require @option{-ansi} or a
2772 @option{-std} option specifying the required version of ISO C)@. However,
2773 without this option, certain GNU extensions and traditional C and C++
2774 features are supported as well. With this option, they are rejected.
2776 @option{-pedantic} does not cause warning messages for use of the
2777 alternate keywords whose names begin and end with @samp{__}. Pedantic
2778 warnings are also disabled in the expression that follows
2779 @code{__extension__}. However, only system header files should use
2780 these escape routes; application programs should avoid them.
2781 @xref{Alternate Keywords}.
2783 Some users try to use @option{-pedantic} to check programs for strict ISO
2784 C conformance. They soon find that it does not do quite what they want:
2785 it finds some non-ISO practices, but not all---only those for which
2786 ISO C @emph{requires} a diagnostic, and some others for which
2787 diagnostics have been added.
2789 A feature to report any failure to conform to ISO C might be useful in
2790 some instances, but would require considerable additional work and would
2791 be quite different from @option{-pedantic}. We don't have plans to
2792 support such a feature in the near future.
2794 Where the standard specified with @option{-std} represents a GNU
2795 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2796 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2797 extended dialect is based. Warnings from @option{-pedantic} are given
2798 where they are required by the base standard. (It would not make sense
2799 for such warnings to be given only for features not in the specified GNU
2800 C dialect, since by definition the GNU dialects of C include all
2801 features the compiler supports with the given option, and there would be
2802 nothing to warn about.)
2804 @item -pedantic-errors
2805 @opindex pedantic-errors
2806 Like @option{-pedantic}, except that errors are produced rather than
2812 This enables all the warnings about constructions that some users
2813 consider questionable, and that are easy to avoid (or modify to
2814 prevent the warning), even in conjunction with macros. This also
2815 enables some language-specific warnings described in @ref{C++ Dialect
2816 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2818 @option{-Wall} turns on the following warning flags:
2820 @gccoptlist{-Waddress @gol
2821 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2823 -Wchar-subscripts @gol
2824 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
2826 -Wimplicit-function-declaration @gol
2829 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2830 -Wmissing-braces @gol
2836 -Wsequence-point @gol
2837 -Wsign-compare @r{(only in C++)} @gol
2838 -Wstrict-aliasing @gol
2839 -Wstrict-overflow=1 @gol
2842 -Wuninitialized @gol
2843 -Wunknown-pragmas @gol
2844 -Wunused-function @gol
2847 -Wunused-variable @gol
2848 -Wvolatile-register-var @gol
2851 Note that some warning flags are not implied by @option{-Wall}. Some of
2852 them warn about constructions that users generally do not consider
2853 questionable, but which occasionally you might wish to check for;
2854 others warn about constructions that are necessary or hard to avoid in
2855 some cases, and there is no simple way to modify the code to suppress
2856 the warning. Some of them are enabled by @option{-Wextra} but many of
2857 them must be enabled individually.
2863 This enables some extra warning flags that are not enabled by
2864 @option{-Wall}. (This option used to be called @option{-W}. The older
2865 name is still supported, but the newer name is more descriptive.)
2867 @gccoptlist{-Wclobbered @gol
2869 -Wignored-qualifiers @gol
2870 -Wmissing-field-initializers @gol
2871 -Wmissing-parameter-type @r{(C only)} @gol
2872 -Wold-style-declaration @r{(C only)} @gol
2873 -Woverride-init @gol
2876 -Wuninitialized @gol
2877 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2880 The option @option{-Wextra} also prints warning messages for the
2886 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2887 @samp{>}, or @samp{>=}.
2890 (C++ only) An enumerator and a non-enumerator both appear in a
2891 conditional expression.
2894 (C++ only) Ambiguous virtual bases.
2897 (C++ only) Subscripting an array which has been declared @samp{register}.
2900 (C++ only) Taking the address of a variable which has been declared
2904 (C++ only) A base class is not initialized in a derived class' copy
2909 @item -Wchar-subscripts
2910 @opindex Wchar-subscripts
2911 @opindex Wno-char-subscripts
2912 Warn if an array subscript has type @code{char}. This is a common cause
2913 of error, as programmers often forget that this type is signed on some
2915 This warning is enabled by @option{-Wall}.
2919 @opindex Wno-comment
2920 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2921 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2922 This warning is enabled by @option{-Wall}.
2927 @opindex ffreestanding
2928 @opindex fno-builtin
2929 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2930 the arguments supplied have types appropriate to the format string
2931 specified, and that the conversions specified in the format string make
2932 sense. This includes standard functions, and others specified by format
2933 attributes (@pxref{Function Attributes}), in the @code{printf},
2934 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2935 not in the C standard) families (or other target-specific families).
2936 Which functions are checked without format attributes having been
2937 specified depends on the standard version selected, and such checks of
2938 functions without the attribute specified are disabled by
2939 @option{-ffreestanding} or @option{-fno-builtin}.
2941 The formats are checked against the format features supported by GNU
2942 libc version 2.2. These include all ISO C90 and C99 features, as well
2943 as features from the Single Unix Specification and some BSD and GNU
2944 extensions. Other library implementations may not support all these
2945 features; GCC does not support warning about features that go beyond a
2946 particular library's limitations. However, if @option{-pedantic} is used
2947 with @option{-Wformat}, warnings will be given about format features not
2948 in the selected standard version (but not for @code{strfmon} formats,
2949 since those are not in any version of the C standard). @xref{C Dialect
2950 Options,,Options Controlling C Dialect}.
2952 Since @option{-Wformat} also checks for null format arguments for
2953 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2955 @option{-Wformat} is included in @option{-Wall}. For more control over some
2956 aspects of format checking, the options @option{-Wformat-y2k},
2957 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2958 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2959 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2962 @opindex Wformat-y2k
2963 @opindex Wno-format-y2k
2964 If @option{-Wformat} is specified, also warn about @code{strftime}
2965 formats which may yield only a two-digit year.
2967 @item -Wno-format-contains-nul
2968 @opindex Wno-format-contains-nul
2969 @opindex Wformat-contains-nul
2970 If @option{-Wformat} is specified, do not warn about format strings that
2973 @item -Wno-format-extra-args
2974 @opindex Wno-format-extra-args
2975 @opindex Wformat-extra-args
2976 If @option{-Wformat} is specified, do not warn about excess arguments to a
2977 @code{printf} or @code{scanf} format function. The C standard specifies
2978 that such arguments are ignored.
2980 Where the unused arguments lie between used arguments that are
2981 specified with @samp{$} operand number specifications, normally
2982 warnings are still given, since the implementation could not know what
2983 type to pass to @code{va_arg} to skip the unused arguments. However,
2984 in the case of @code{scanf} formats, this option will suppress the
2985 warning if the unused arguments are all pointers, since the Single
2986 Unix Specification says that such unused arguments are allowed.
2988 @item -Wno-format-zero-length @r{(C and Objective-C only)}
2989 @opindex Wno-format-zero-length
2990 @opindex Wformat-zero-length
2991 If @option{-Wformat} is specified, do not warn about zero-length formats.
2992 The C standard specifies that zero-length formats are allowed.
2994 @item -Wformat-nonliteral
2995 @opindex Wformat-nonliteral
2996 @opindex Wno-format-nonliteral
2997 If @option{-Wformat} is specified, also warn if the format string is not a
2998 string literal and so cannot be checked, unless the format function
2999 takes its format arguments as a @code{va_list}.
3001 @item -Wformat-security
3002 @opindex Wformat-security
3003 @opindex Wno-format-security
3004 If @option{-Wformat} is specified, also warn about uses of format
3005 functions that represent possible security problems. At present, this
3006 warns about calls to @code{printf} and @code{scanf} functions where the
3007 format string is not a string literal and there are no format arguments,
3008 as in @code{printf (foo);}. This may be a security hole if the format
3009 string came from untrusted input and contains @samp{%n}. (This is
3010 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3011 in future warnings may be added to @option{-Wformat-security} that are not
3012 included in @option{-Wformat-nonliteral}.)
3016 @opindex Wno-format=2
3017 Enable @option{-Wformat} plus format checks not included in
3018 @option{-Wformat}. Currently equivalent to @samp{-Wformat
3019 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
3021 @item -Wnonnull @r{(C and Objective-C only)}
3023 @opindex Wno-nonnull
3024 Warn about passing a null pointer for arguments marked as
3025 requiring a non-null value by the @code{nonnull} function attribute.
3027 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3028 can be disabled with the @option{-Wno-nonnull} option.
3030 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3032 @opindex Wno-init-self
3033 Warn about uninitialized variables which are initialized with themselves.
3034 Note this option can only be used with the @option{-Wuninitialized} option.
3036 For example, GCC will warn about @code{i} being uninitialized in the
3037 following snippet only when @option{-Winit-self} has been specified:
3048 @item -Wimplicit-int @r{(C and Objective-C only)}
3049 @opindex Wimplicit-int
3050 @opindex Wno-implicit-int
3051 Warn when a declaration does not specify a type.
3052 This warning is enabled by @option{-Wall}.
3054 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3055 @opindex Wimplicit-function-declaration
3056 @opindex Wno-implicit-function-declaration
3057 Give a warning whenever a function is used before being declared. In
3058 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3059 enabled by default and it is made into an error by
3060 @option{-pedantic-errors}. This warning is also enabled by
3065 @opindex Wno-implicit
3066 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3067 This warning is enabled by @option{-Wall}.
3069 @item -Wignored-qualifiers @r{(C and C++ only)}
3070 @opindex Wignored-qualifiers
3071 @opindex Wno-ignored-qualifiers
3072 Warn if the return type of a function has a type qualifier
3073 such as @code{const}. For ISO C such a type qualifier has no effect,
3074 since the value returned by a function is not an lvalue.
3075 For C++, the warning is only emitted for scalar types or @code{void}.
3076 ISO C prohibits qualified @code{void} return types on function
3077 definitions, so such return types always receive a warning
3078 even without this option.
3080 This warning is also enabled by @option{-Wextra}.
3085 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3086 a function with external linkage, returning int, taking either zero
3087 arguments, two, or three arguments of appropriate types. This warning
3088 is enabled by default in C++ and is enabled by either @option{-Wall}
3089 or @option{-pedantic}.
3091 @item -Wmissing-braces
3092 @opindex Wmissing-braces
3093 @opindex Wno-missing-braces
3094 Warn if an aggregate or union initializer is not fully bracketed. In
3095 the following example, the initializer for @samp{a} is not fully
3096 bracketed, but that for @samp{b} is fully bracketed.
3099 int a[2][2] = @{ 0, 1, 2, 3 @};
3100 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3103 This warning is enabled by @option{-Wall}.
3105 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3106 @opindex Wmissing-include-dirs
3107 @opindex Wno-missing-include-dirs
3108 Warn if a user-supplied include directory does not exist.
3111 @opindex Wparentheses
3112 @opindex Wno-parentheses
3113 Warn if parentheses are omitted in certain contexts, such
3114 as when there is an assignment in a context where a truth value
3115 is expected, or when operators are nested whose precedence people
3116 often get confused about.
3118 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3119 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3120 interpretation from that of ordinary mathematical notation.
3122 Also warn about constructions where there may be confusion to which
3123 @code{if} statement an @code{else} branch belongs. Here is an example of
3138 In C/C++, every @code{else} branch belongs to the innermost possible
3139 @code{if} statement, which in this example is @code{if (b)}. This is
3140 often not what the programmer expected, as illustrated in the above
3141 example by indentation the programmer chose. When there is the
3142 potential for this confusion, GCC will issue a warning when this flag
3143 is specified. To eliminate the warning, add explicit braces around
3144 the innermost @code{if} statement so there is no way the @code{else}
3145 could belong to the enclosing @code{if}. The resulting code would
3162 This warning is enabled by @option{-Wall}.
3164 @item -Wsequence-point
3165 @opindex Wsequence-point
3166 @opindex Wno-sequence-point
3167 Warn about code that may have undefined semantics because of violations
3168 of sequence point rules in the C and C++ standards.
3170 The C and C++ standards defines the order in which expressions in a C/C++
3171 program are evaluated in terms of @dfn{sequence points}, which represent
3172 a partial ordering between the execution of parts of the program: those
3173 executed before the sequence point, and those executed after it. These
3174 occur after the evaluation of a full expression (one which is not part
3175 of a larger expression), after the evaluation of the first operand of a
3176 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3177 function is called (but after the evaluation of its arguments and the
3178 expression denoting the called function), and in certain other places.
3179 Other than as expressed by the sequence point rules, the order of
3180 evaluation of subexpressions of an expression is not specified. All
3181 these rules describe only a partial order rather than a total order,
3182 since, for example, if two functions are called within one expression
3183 with no sequence point between them, the order in which the functions
3184 are called is not specified. However, the standards committee have
3185 ruled that function calls do not overlap.
3187 It is not specified when between sequence points modifications to the
3188 values of objects take effect. Programs whose behavior depends on this
3189 have undefined behavior; the C and C++ standards specify that ``Between
3190 the previous and next sequence point an object shall have its stored
3191 value modified at most once by the evaluation of an expression.
3192 Furthermore, the prior value shall be read only to determine the value
3193 to be stored.''. If a program breaks these rules, the results on any
3194 particular implementation are entirely unpredictable.
3196 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3197 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3198 diagnosed by this option, and it may give an occasional false positive
3199 result, but in general it has been found fairly effective at detecting
3200 this sort of problem in programs.
3202 The standard is worded confusingly, therefore there is some debate
3203 over the precise meaning of the sequence point rules in subtle cases.
3204 Links to discussions of the problem, including proposed formal
3205 definitions, may be found on the GCC readings page, at
3206 @w{@uref{http://gcc.gnu.org/readings.html}}.
3208 This warning is enabled by @option{-Wall} for C and C++.
3211 @opindex Wreturn-type
3212 @opindex Wno-return-type
3213 Warn whenever a function is defined with a return-type that defaults
3214 to @code{int}. Also warn about any @code{return} statement with no
3215 return-value in a function whose return-type is not @code{void}
3216 (falling off the end of the function body is considered returning
3217 without a value), and about a @code{return} statement with an
3218 expression in a function whose return-type is @code{void}.
3220 For C++, a function without return type always produces a diagnostic
3221 message, even when @option{-Wno-return-type} is specified. The only
3222 exceptions are @samp{main} and functions defined in system headers.
3224 This warning is enabled by @option{-Wall}.
3229 Warn whenever a @code{switch} statement has an index of enumerated type
3230 and lacks a @code{case} for one or more of the named codes of that
3231 enumeration. (The presence of a @code{default} label prevents this
3232 warning.) @code{case} labels outside the enumeration range also
3233 provoke warnings when this option is used (even if there is a
3234 @code{default} label).
3235 This warning is enabled by @option{-Wall}.
3237 @item -Wswitch-default
3238 @opindex Wswitch-default
3239 @opindex Wno-switch-default
3240 Warn whenever a @code{switch} statement does not have a @code{default}
3244 @opindex Wswitch-enum
3245 @opindex Wno-switch-enum
3246 Warn whenever a @code{switch} statement has an index of enumerated type
3247 and lacks a @code{case} for one or more of the named codes of that
3248 enumeration. @code{case} labels outside the enumeration range also
3249 provoke warnings when this option is used. The only difference
3250 between @option{-Wswitch} and this option is that this option gives a
3251 warning about an omitted enumeration code even if there is a
3252 @code{default} label.
3254 @item -Wsync-nand @r{(C and C++ only)}
3256 @opindex Wno-sync-nand
3257 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3258 built-in functions are used. These functions changed semantics in GCC 4.4.
3262 @opindex Wno-trigraphs
3263 Warn if any trigraphs are encountered that might change the meaning of
3264 the program (trigraphs within comments are not warned about).
3265 This warning is enabled by @option{-Wall}.
3267 @item -Wunused-function
3268 @opindex Wunused-function
3269 @opindex Wno-unused-function
3270 Warn whenever a static function is declared but not defined or a
3271 non-inline static function is unused.
3272 This warning is enabled by @option{-Wall}.
3274 @item -Wunused-label
3275 @opindex Wunused-label
3276 @opindex Wno-unused-label
3277 Warn whenever a label is declared but not used.
3278 This warning is enabled by @option{-Wall}.
3280 To suppress this warning use the @samp{unused} attribute
3281 (@pxref{Variable Attributes}).
3283 @item -Wunused-parameter
3284 @opindex Wunused-parameter
3285 @opindex Wno-unused-parameter
3286 Warn whenever a function parameter is unused aside from its declaration.
3288 To suppress this warning use the @samp{unused} attribute
3289 (@pxref{Variable Attributes}).
3291 @item -Wno-unused-result
3292 @opindex Wunused-result
3293 @opindex Wno-unused-result
3294 Do not warn if a caller of a function marked with attribute
3295 @code{warn_unused_result} (@pxref{Variable Attributes}) does not use
3296 its return value. The default is @option{-Wunused-result}.
3298 @item -Wunused-variable
3299 @opindex Wunused-variable
3300 @opindex Wno-unused-variable
3301 Warn whenever a local variable or non-constant static variable is unused
3302 aside from its declaration.
3303 This warning is enabled by @option{-Wall}.
3305 To suppress this warning use the @samp{unused} attribute
3306 (@pxref{Variable Attributes}).
3308 @item -Wunused-value
3309 @opindex Wunused-value
3310 @opindex Wno-unused-value
3311 Warn whenever a statement computes a result that is explicitly not
3312 used. To suppress this warning cast the unused expression to
3313 @samp{void}. This includes an expression-statement or the left-hand
3314 side of a comma expression that contains no side effects. For example,
3315 an expression such as @samp{x[i,j]} will cause a warning, while
3316 @samp{x[(void)i,j]} will not.
3318 This warning is enabled by @option{-Wall}.
3323 All the above @option{-Wunused} options combined.
3325 In order to get a warning about an unused function parameter, you must
3326 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3327 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3329 @item -Wuninitialized
3330 @opindex Wuninitialized
3331 @opindex Wno-uninitialized
3332 Warn if an automatic variable is used without first being initialized
3333 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3334 warn if a non-static reference or non-static @samp{const} member
3335 appears in a class without constructors.
3337 If you want to warn about code which uses the uninitialized value of the
3338 variable in its own initializer, use the @option{-Winit-self} option.
3340 These warnings occur for individual uninitialized or clobbered
3341 elements of structure, union or array variables as well as for
3342 variables which are uninitialized or clobbered as a whole. They do
3343 not occur for variables or elements declared @code{volatile}. Because
3344 these warnings depend on optimization, the exact variables or elements
3345 for which there are warnings will depend on the precise optimization
3346 options and version of GCC used.
3348 Note that there may be no warning about a variable that is used only
3349 to compute a value that itself is never used, because such
3350 computations may be deleted by data flow analysis before the warnings
3353 These warnings are made optional because GCC is not smart
3354 enough to see all the reasons why the code might be correct
3355 despite appearing to have an error. Here is one example of how
3376 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3377 always initialized, but GCC doesn't know this. Here is
3378 another common case:
3383 if (change_y) save_y = y, y = new_y;
3385 if (change_y) y = save_y;
3390 This has no bug because @code{save_y} is used only if it is set.
3392 @cindex @code{longjmp} warnings
3393 This option also warns when a non-volatile automatic variable might be
3394 changed by a call to @code{longjmp}. These warnings as well are possible
3395 only in optimizing compilation.
3397 The compiler sees only the calls to @code{setjmp}. It cannot know
3398 where @code{longjmp} will be called; in fact, a signal handler could
3399 call it at any point in the code. As a result, you may get a warning
3400 even when there is in fact no problem because @code{longjmp} cannot
3401 in fact be called at the place which would cause a problem.
3403 Some spurious warnings can be avoided if you declare all the functions
3404 you use that never return as @code{noreturn}. @xref{Function
3407 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3409 @item -Wunknown-pragmas
3410 @opindex Wunknown-pragmas
3411 @opindex Wno-unknown-pragmas
3412 @cindex warning for unknown pragmas
3413 @cindex unknown pragmas, warning
3414 @cindex pragmas, warning of unknown
3415 Warn when a #pragma directive is encountered which is not understood by
3416 GCC@. If this command line option is used, warnings will even be issued
3417 for unknown pragmas in system header files. This is not the case if
3418 the warnings were only enabled by the @option{-Wall} command line option.
3421 @opindex Wno-pragmas
3423 Do not warn about misuses of pragmas, such as incorrect parameters,
3424 invalid syntax, or conflicts between pragmas. See also
3425 @samp{-Wunknown-pragmas}.
3427 @item -Wstrict-aliasing
3428 @opindex Wstrict-aliasing
3429 @opindex Wno-strict-aliasing
3430 This option is only active when @option{-fstrict-aliasing} is active.
3431 It warns about code which might break the strict aliasing rules that the
3432 compiler is using for optimization. The warning does not catch all
3433 cases, but does attempt to catch the more common pitfalls. It is
3434 included in @option{-Wall}.
3435 It is equivalent to @option{-Wstrict-aliasing=3}
3437 @item -Wstrict-aliasing=n
3438 @opindex Wstrict-aliasing=n
3439 @opindex Wno-strict-aliasing=n
3440 This option is only active when @option{-fstrict-aliasing} is active.
3441 It warns about code which might break the strict aliasing rules that the
3442 compiler is using for optimization.
3443 Higher levels correspond to higher accuracy (fewer false positives).
3444 Higher levels also correspond to more effort, similar to the way -O works.
3445 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3448 Level 1: Most aggressive, quick, least accurate.
3449 Possibly useful when higher levels
3450 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3451 false negatives. However, it has many false positives.
3452 Warns for all pointer conversions between possibly incompatible types,
3453 even if never dereferenced. Runs in the frontend only.
3455 Level 2: Aggressive, quick, not too precise.
3456 May still have many false positives (not as many as level 1 though),
3457 and few false negatives (but possibly more than level 1).
3458 Unlike level 1, it only warns when an address is taken. Warns about
3459 incomplete types. Runs in the frontend only.
3461 Level 3 (default for @option{-Wstrict-aliasing}):
3462 Should have very few false positives and few false
3463 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3464 Takes care of the common punn+dereference pattern in the frontend:
3465 @code{*(int*)&some_float}.
3466 If optimization is enabled, it also runs in the backend, where it deals
3467 with multiple statement cases using flow-sensitive points-to information.
3468 Only warns when the converted pointer is dereferenced.
3469 Does not warn about incomplete types.
3471 @item -Wstrict-overflow
3472 @itemx -Wstrict-overflow=@var{n}
3473 @opindex Wstrict-overflow
3474 @opindex Wno-strict-overflow
3475 This option is only active when @option{-fstrict-overflow} is active.
3476 It warns about cases where the compiler optimizes based on the
3477 assumption that signed overflow does not occur. Note that it does not
3478 warn about all cases where the code might overflow: it only warns
3479 about cases where the compiler implements some optimization. Thus
3480 this warning depends on the optimization level.
3482 An optimization which assumes that signed overflow does not occur is
3483 perfectly safe if the values of the variables involved are such that
3484 overflow never does, in fact, occur. Therefore this warning can
3485 easily give a false positive: a warning about code which is not
3486 actually a problem. To help focus on important issues, several
3487 warning levels are defined. No warnings are issued for the use of
3488 undefined signed overflow when estimating how many iterations a loop
3489 will require, in particular when determining whether a loop will be
3493 @item -Wstrict-overflow=1
3494 Warn about cases which are both questionable and easy to avoid. For
3495 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3496 compiler will simplify this to @code{1}. This level of
3497 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3498 are not, and must be explicitly requested.
3500 @item -Wstrict-overflow=2
3501 Also warn about other cases where a comparison is simplified to a
3502 constant. For example: @code{abs (x) >= 0}. This can only be
3503 simplified when @option{-fstrict-overflow} is in effect, because
3504 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3505 zero. @option{-Wstrict-overflow} (with no level) is the same as
3506 @option{-Wstrict-overflow=2}.
3508 @item -Wstrict-overflow=3
3509 Also warn about other cases where a comparison is simplified. For
3510 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3512 @item -Wstrict-overflow=4
3513 Also warn about other simplifications not covered by the above cases.
3514 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3516 @item -Wstrict-overflow=5
3517 Also warn about cases where the compiler reduces the magnitude of a
3518 constant involved in a comparison. For example: @code{x + 2 > y} will
3519 be simplified to @code{x + 1 >= y}. This is reported only at the
3520 highest warning level because this simplification applies to many
3521 comparisons, so this warning level will give a very large number of
3525 @item -Warray-bounds
3526 @opindex Wno-array-bounds
3527 @opindex Warray-bounds
3528 This option is only active when @option{-ftree-vrp} is active
3529 (default for -O2 and above). It warns about subscripts to arrays
3530 that are always out of bounds. This warning is enabled by @option{-Wall}.
3532 @item -Wno-div-by-zero
3533 @opindex Wno-div-by-zero
3534 @opindex Wdiv-by-zero
3535 Do not warn about compile-time integer division by zero. Floating point
3536 division by zero is not warned about, as it can be a legitimate way of
3537 obtaining infinities and NaNs.
3539 @item -Wsystem-headers
3540 @opindex Wsystem-headers
3541 @opindex Wno-system-headers
3542 @cindex warnings from system headers
3543 @cindex system headers, warnings from
3544 Print warning messages for constructs found in system header files.
3545 Warnings from system headers are normally suppressed, on the assumption
3546 that they usually do not indicate real problems and would only make the
3547 compiler output harder to read. Using this command line option tells
3548 GCC to emit warnings from system headers as if they occurred in user
3549 code. However, note that using @option{-Wall} in conjunction with this
3550 option will @emph{not} warn about unknown pragmas in system
3551 headers---for that, @option{-Wunknown-pragmas} must also be used.
3554 @opindex Wfloat-equal
3555 @opindex Wno-float-equal
3556 Warn if floating point values are used in equality comparisons.
3558 The idea behind this is that sometimes it is convenient (for the
3559 programmer) to consider floating-point values as approximations to
3560 infinitely precise real numbers. If you are doing this, then you need
3561 to compute (by analyzing the code, or in some other way) the maximum or
3562 likely maximum error that the computation introduces, and allow for it
3563 when performing comparisons (and when producing output, but that's a
3564 different problem). In particular, instead of testing for equality, you
3565 would check to see whether the two values have ranges that overlap; and
3566 this is done with the relational operators, so equality comparisons are
3569 @item -Wtraditional @r{(C and Objective-C only)}
3570 @opindex Wtraditional
3571 @opindex Wno-traditional
3572 Warn about certain constructs that behave differently in traditional and
3573 ISO C@. Also warn about ISO C constructs that have no traditional C
3574 equivalent, and/or problematic constructs which should be avoided.
3578 Macro parameters that appear within string literals in the macro body.
3579 In traditional C macro replacement takes place within string literals,
3580 but does not in ISO C@.
3583 In traditional C, some preprocessor directives did not exist.
3584 Traditional preprocessors would only consider a line to be a directive
3585 if the @samp{#} appeared in column 1 on the line. Therefore
3586 @option{-Wtraditional} warns about directives that traditional C
3587 understands but would ignore because the @samp{#} does not appear as the
3588 first character on the line. It also suggests you hide directives like
3589 @samp{#pragma} not understood by traditional C by indenting them. Some
3590 traditional implementations would not recognize @samp{#elif}, so it
3591 suggests avoiding it altogether.
3594 A function-like macro that appears without arguments.
3597 The unary plus operator.
3600 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3601 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3602 constants.) Note, these suffixes appear in macros defined in the system
3603 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3604 Use of these macros in user code might normally lead to spurious
3605 warnings, however GCC's integrated preprocessor has enough context to
3606 avoid warning in these cases.
3609 A function declared external in one block and then used after the end of
3613 A @code{switch} statement has an operand of type @code{long}.
3616 A non-@code{static} function declaration follows a @code{static} one.
3617 This construct is not accepted by some traditional C compilers.
3620 The ISO type of an integer constant has a different width or
3621 signedness from its traditional type. This warning is only issued if
3622 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3623 typically represent bit patterns, are not warned about.
3626 Usage of ISO string concatenation is detected.
3629 Initialization of automatic aggregates.
3632 Identifier conflicts with labels. Traditional C lacks a separate
3633 namespace for labels.
3636 Initialization of unions. If the initializer is zero, the warning is
3637 omitted. This is done under the assumption that the zero initializer in
3638 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3639 initializer warnings and relies on default initialization to zero in the
3643 Conversions by prototypes between fixed/floating point values and vice
3644 versa. The absence of these prototypes when compiling with traditional
3645 C would cause serious problems. This is a subset of the possible
3646 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3649 Use of ISO C style function definitions. This warning intentionally is
3650 @emph{not} issued for prototype declarations or variadic functions
3651 because these ISO C features will appear in your code when using
3652 libiberty's traditional C compatibility macros, @code{PARAMS} and
3653 @code{VPARAMS}. This warning is also bypassed for nested functions
3654 because that feature is already a GCC extension and thus not relevant to
3655 traditional C compatibility.
3658 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3659 @opindex Wtraditional-conversion
3660 @opindex Wno-traditional-conversion
3661 Warn if a prototype causes a type conversion that is different from what
3662 would happen to the same argument in the absence of a prototype. This
3663 includes conversions of fixed point to floating and vice versa, and
3664 conversions changing the width or signedness of a fixed point argument
3665 except when the same as the default promotion.
3667 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3668 @opindex Wdeclaration-after-statement
3669 @opindex Wno-declaration-after-statement
3670 Warn when a declaration is found after a statement in a block. This
3671 construct, known from C++, was introduced with ISO C99 and is by default
3672 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3673 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3678 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3680 @item -Wno-endif-labels
3681 @opindex Wno-endif-labels
3682 @opindex Wendif-labels
3683 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3688 Warn whenever a local variable shadows another local variable, parameter or
3689 global variable or whenever a built-in function is shadowed.
3691 @item -Wlarger-than=@var{len}
3692 @opindex Wlarger-than=@var{len}
3693 @opindex Wlarger-than-@var{len}
3694 Warn whenever an object of larger than @var{len} bytes is defined.
3696 @item -Wframe-larger-than=@var{len}
3697 @opindex Wframe-larger-than
3698 Warn if the size of a function frame is larger than @var{len} bytes.
3699 The computation done to determine the stack frame size is approximate
3700 and not conservative.
3701 The actual requirements may be somewhat greater than @var{len}
3702 even if you do not get a warning. In addition, any space allocated
3703 via @code{alloca}, variable-length arrays, or related constructs
3704 is not included by the compiler when determining
3705 whether or not to issue a warning.
3707 @item -Wunsafe-loop-optimizations
3708 @opindex Wunsafe-loop-optimizations
3709 @opindex Wno-unsafe-loop-optimizations
3710 Warn if the loop cannot be optimized because the compiler could not
3711 assume anything on the bounds of the loop indices. With
3712 @option{-funsafe-loop-optimizations} warn if the compiler made
3715 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
3716 @opindex Wno-pedantic-ms-format
3717 @opindex Wpedantic-ms-format
3718 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
3719 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
3720 depending on the MS runtime, when you are using the options @option{-Wformat}
3721 and @option{-pedantic} without gnu-extensions.
3723 @item -Wpointer-arith
3724 @opindex Wpointer-arith
3725 @opindex Wno-pointer-arith
3726 Warn about anything that depends on the ``size of'' a function type or
3727 of @code{void}. GNU C assigns these types a size of 1, for
3728 convenience in calculations with @code{void *} pointers and pointers
3729 to functions. In C++, warn also when an arithmetic operation involves
3730 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3733 @opindex Wtype-limits
3734 @opindex Wno-type-limits
3735 Warn if a comparison is always true or always false due to the limited
3736 range of the data type, but do not warn for constant expressions. For
3737 example, warn if an unsigned variable is compared against zero with
3738 @samp{<} or @samp{>=}. This warning is also enabled by
3741 @item -Wbad-function-cast @r{(C and Objective-C only)}
3742 @opindex Wbad-function-cast
3743 @opindex Wno-bad-function-cast
3744 Warn whenever a function call is cast to a non-matching type.
3745 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3747 @item -Wc++-compat @r{(C and Objective-C only)}
3748 Warn about ISO C constructs that are outside of the common subset of
3749 ISO C and ISO C++, e.g.@: request for implicit conversion from
3750 @code{void *} to a pointer to non-@code{void} type.
3752 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3753 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3754 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3755 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3759 @opindex Wno-cast-qual
3760 Warn whenever a pointer is cast so as to remove a type qualifier from
3761 the target type. For example, warn if a @code{const char *} is cast
3762 to an ordinary @code{char *}.
3764 Also warn when making a cast which introduces a type qualifier in an
3765 unsafe way. For example, casting @code{char **} to @code{const char **}
3766 is unsafe, as in this example:
3769 /* p is char ** value. */
3770 const char **q = (const char **) p;
3771 /* Assignment of readonly string to const char * is OK. */
3773 /* Now char** pointer points to read-only memory. */
3778 @opindex Wcast-align
3779 @opindex Wno-cast-align
3780 Warn whenever a pointer is cast such that the required alignment of the
3781 target is increased. For example, warn if a @code{char *} is cast to
3782 an @code{int *} on machines where integers can only be accessed at
3783 two- or four-byte boundaries.
3785 @item -Wwrite-strings
3786 @opindex Wwrite-strings
3787 @opindex Wno-write-strings
3788 When compiling C, give string constants the type @code{const
3789 char[@var{length}]} so that copying the address of one into a
3790 non-@code{const} @code{char *} pointer will get a warning. These
3791 warnings will help you find at compile time code that can try to write
3792 into a string constant, but only if you have been very careful about
3793 using @code{const} in declarations and prototypes. Otherwise, it will
3794 just be a nuisance. This is why we did not make @option{-Wall} request
3797 When compiling C++, warn about the deprecated conversion from string
3798 literals to @code{char *}. This warning is enabled by default for C++
3803 @opindex Wno-clobbered
3804 Warn for variables that might be changed by @samp{longjmp} or
3805 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
3808 @opindex Wconversion
3809 @opindex Wno-conversion
3810 Warn for implicit conversions that may alter a value. This includes
3811 conversions between real and integer, like @code{abs (x)} when
3812 @code{x} is @code{double}; conversions between signed and unsigned,
3813 like @code{unsigned ui = -1}; and conversions to smaller types, like
3814 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
3815 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
3816 changed by the conversion like in @code{abs (2.0)}. Warnings about
3817 conversions between signed and unsigned integers can be disabled by
3818 using @option{-Wno-sign-conversion}.
3820 For C++, also warn for conversions between @code{NULL} and non-pointer
3821 types; confusing overload resolution for user-defined conversions; and
3822 conversions that will never use a type conversion operator:
3823 conversions to @code{void}, the same type, a base class or a reference
3824 to them. Warnings about conversions between signed and unsigned
3825 integers are disabled by default in C++ unless
3826 @option{-Wsign-conversion} is explicitly enabled.
3829 @opindex Wempty-body
3830 @opindex Wno-empty-body
3831 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
3832 while} statement. This warning is also enabled by @option{-Wextra}.
3834 @item -Wenum-compare
3835 @opindex Wenum-compare
3836 @opindex Wno-enum-compare
3837 Warn about a comparison between values of different enum types. In C++
3838 this warning is enabled by default. In C this warning is enabled by
3841 @item -Wjump-misses-init @r{(C, Objective-C only)}
3842 @opindex Wjump-misses-init
3843 @opindex Wno-jump-misses-init
3844 Warn if a @code{goto} statement or a @code{switch} statement jumps
3845 forward across the initialization of a variable, or jumps backward to a
3846 label after the variable has been initialized. This only warns about
3847 variables which are initialized when they are declared. This warning is
3848 only supported for C and Objective C; in C++ this sort of branch is an
3851 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
3852 can be disabled with the @option{-Wno-jump-misses-init} option.
3854 @item -Wsign-compare
3855 @opindex Wsign-compare
3856 @opindex Wno-sign-compare
3857 @cindex warning for comparison of signed and unsigned values
3858 @cindex comparison of signed and unsigned values, warning
3859 @cindex signed and unsigned values, comparison warning
3860 Warn when a comparison between signed and unsigned values could produce
3861 an incorrect result when the signed value is converted to unsigned.
3862 This warning is also enabled by @option{-Wextra}; to get the other warnings
3863 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3865 @item -Wsign-conversion
3866 @opindex Wsign-conversion
3867 @opindex Wno-sign-conversion
3868 Warn for implicit conversions that may change the sign of an integer
3869 value, like assigning a signed integer expression to an unsigned
3870 integer variable. An explicit cast silences the warning. In C, this
3871 option is enabled also by @option{-Wconversion}.
3875 @opindex Wno-address
3876 Warn about suspicious uses of memory addresses. These include using
3877 the address of a function in a conditional expression, such as
3878 @code{void func(void); if (func)}, and comparisons against the memory
3879 address of a string literal, such as @code{if (x == "abc")}. Such
3880 uses typically indicate a programmer error: the address of a function
3881 always evaluates to true, so their use in a conditional usually
3882 indicate that the programmer forgot the parentheses in a function
3883 call; and comparisons against string literals result in unspecified
3884 behavior and are not portable in C, so they usually indicate that the
3885 programmer intended to use @code{strcmp}. This warning is enabled by
3889 @opindex Wlogical-op
3890 @opindex Wno-logical-op
3891 Warn about suspicious uses of logical operators in expressions.
3892 This includes using logical operators in contexts where a
3893 bit-wise operator is likely to be expected.
3895 @item -Waggregate-return
3896 @opindex Waggregate-return
3897 @opindex Wno-aggregate-return
3898 Warn if any functions that return structures or unions are defined or
3899 called. (In languages where you can return an array, this also elicits
3902 @item -Wno-attributes
3903 @opindex Wno-attributes
3904 @opindex Wattributes
3905 Do not warn if an unexpected @code{__attribute__} is used, such as
3906 unrecognized attributes, function attributes applied to variables,
3907 etc. This will not stop errors for incorrect use of supported
3910 @item -Wno-builtin-macro-redefined
3911 @opindex Wno-builtin-macro-redefined
3912 @opindex Wbuiltin-macro-redefined
3913 Do not warn if certain built-in macros are redefined. This suppresses
3914 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
3915 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
3917 @item -Wstrict-prototypes @r{(C and Objective-C only)}
3918 @opindex Wstrict-prototypes
3919 @opindex Wno-strict-prototypes
3920 Warn if a function is declared or defined without specifying the
3921 argument types. (An old-style function definition is permitted without
3922 a warning if preceded by a declaration which specifies the argument
3925 @item -Wold-style-declaration @r{(C and Objective-C only)}
3926 @opindex Wold-style-declaration
3927 @opindex Wno-old-style-declaration
3928 Warn for obsolescent usages, according to the C Standard, in a
3929 declaration. For example, warn if storage-class specifiers like
3930 @code{static} are not the first things in a declaration. This warning
3931 is also enabled by @option{-Wextra}.
3933 @item -Wold-style-definition @r{(C and Objective-C only)}
3934 @opindex Wold-style-definition
3935 @opindex Wno-old-style-definition
3936 Warn if an old-style function definition is used. A warning is given
3937 even if there is a previous prototype.
3939 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
3940 @opindex Wmissing-parameter-type
3941 @opindex Wno-missing-parameter-type
3942 A function parameter is declared without a type specifier in K&R-style
3949 This warning is also enabled by @option{-Wextra}.
3951 @item -Wmissing-prototypes @r{(C and Objective-C only)}
3952 @opindex Wmissing-prototypes
3953 @opindex Wno-missing-prototypes
3954 Warn if a global function is defined without a previous prototype
3955 declaration. This warning is issued even if the definition itself
3956 provides a prototype. The aim is to detect global functions that fail
3957 to be declared in header files.
3959 @item -Wmissing-declarations
3960 @opindex Wmissing-declarations
3961 @opindex Wno-missing-declarations
3962 Warn if a global function is defined without a previous declaration.
3963 Do so even if the definition itself provides a prototype.
3964 Use this option to detect global functions that are not declared in
3965 header files. In C++, no warnings are issued for function templates,
3966 or for inline functions, or for functions in anonymous namespaces.
3968 @item -Wmissing-field-initializers
3969 @opindex Wmissing-field-initializers
3970 @opindex Wno-missing-field-initializers
3974 Warn if a structure's initializer has some fields missing. For
3975 example, the following code would cause such a warning, because
3976 @code{x.h} is implicitly zero:
3979 struct s @{ int f, g, h; @};
3980 struct s x = @{ 3, 4 @};
3983 This option does not warn about designated initializers, so the following
3984 modification would not trigger a warning:
3987 struct s @{ int f, g, h; @};
3988 struct s x = @{ .f = 3, .g = 4 @};
3991 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3992 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3994 @item -Wmissing-noreturn
3995 @opindex Wmissing-noreturn
3996 @opindex Wno-missing-noreturn
3997 Warn about functions which might be candidates for attribute @code{noreturn}.
3998 Note these are only possible candidates, not absolute ones. Care should
3999 be taken to manually verify functions actually do not ever return before
4000 adding the @code{noreturn} attribute, otherwise subtle code generation
4001 bugs could be introduced. You will not get a warning for @code{main} in
4002 hosted C environments.
4004 @item -Wmissing-format-attribute
4005 @opindex Wmissing-format-attribute
4006 @opindex Wno-missing-format-attribute
4009 Warn about function pointers which might be candidates for @code{format}
4010 attributes. Note these are only possible candidates, not absolute ones.
4011 GCC will guess that function pointers with @code{format} attributes that
4012 are used in assignment, initialization, parameter passing or return
4013 statements should have a corresponding @code{format} attribute in the
4014 resulting type. I.e.@: the left-hand side of the assignment or
4015 initialization, the type of the parameter variable, or the return type
4016 of the containing function respectively should also have a @code{format}
4017 attribute to avoid the warning.
4019 GCC will also warn about function definitions which might be
4020 candidates for @code{format} attributes. Again, these are only
4021 possible candidates. GCC will guess that @code{format} attributes
4022 might be appropriate for any function that calls a function like
4023 @code{vprintf} or @code{vscanf}, but this might not always be the
4024 case, and some functions for which @code{format} attributes are
4025 appropriate may not be detected.
4027 @item -Wno-multichar
4028 @opindex Wno-multichar
4030 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4031 Usually they indicate a typo in the user's code, as they have
4032 implementation-defined values, and should not be used in portable code.
4034 @item -Wnormalized=<none|id|nfc|nfkc>
4035 @opindex Wnormalized=
4038 @cindex character set, input normalization
4039 In ISO C and ISO C++, two identifiers are different if they are
4040 different sequences of characters. However, sometimes when characters
4041 outside the basic ASCII character set are used, you can have two
4042 different character sequences that look the same. To avoid confusion,
4043 the ISO 10646 standard sets out some @dfn{normalization rules} which
4044 when applied ensure that two sequences that look the same are turned into
4045 the same sequence. GCC can warn you if you are using identifiers which
4046 have not been normalized; this option controls that warning.
4048 There are four levels of warning that GCC supports. The default is
4049 @option{-Wnormalized=nfc}, which warns about any identifier which is
4050 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4051 recommended form for most uses.
4053 Unfortunately, there are some characters which ISO C and ISO C++ allow
4054 in identifiers that when turned into NFC aren't allowable as
4055 identifiers. That is, there's no way to use these symbols in portable
4056 ISO C or C++ and have all your identifiers in NFC@.
4057 @option{-Wnormalized=id} suppresses the warning for these characters.
4058 It is hoped that future versions of the standards involved will correct
4059 this, which is why this option is not the default.
4061 You can switch the warning off for all characters by writing
4062 @option{-Wnormalized=none}. You would only want to do this if you
4063 were using some other normalization scheme (like ``D''), because
4064 otherwise you can easily create bugs that are literally impossible to see.
4066 Some characters in ISO 10646 have distinct meanings but look identical
4067 in some fonts or display methodologies, especially once formatting has
4068 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4069 LETTER N'', will display just like a regular @code{n} which has been
4070 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4071 normalization scheme to convert all these into a standard form as
4072 well, and GCC will warn if your code is not in NFKC if you use
4073 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4074 about every identifier that contains the letter O because it might be
4075 confused with the digit 0, and so is not the default, but may be
4076 useful as a local coding convention if the programming environment is
4077 unable to be fixed to display these characters distinctly.
4079 @item -Wno-deprecated
4080 @opindex Wno-deprecated
4081 @opindex Wdeprecated
4082 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4084 @item -Wno-deprecated-declarations
4085 @opindex Wno-deprecated-declarations
4086 @opindex Wdeprecated-declarations
4087 Do not warn about uses of functions (@pxref{Function Attributes}),
4088 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4089 Attributes}) marked as deprecated by using the @code{deprecated}
4093 @opindex Wno-overflow
4095 Do not warn about compile-time overflow in constant expressions.
4097 @item -Woverride-init @r{(C and Objective-C only)}
4098 @opindex Woverride-init
4099 @opindex Wno-override-init
4103 Warn if an initialized field without side effects is overridden when
4104 using designated initializers (@pxref{Designated Inits, , Designated
4107 This warning is included in @option{-Wextra}. To get other
4108 @option{-Wextra} warnings without this one, use @samp{-Wextra
4109 -Wno-override-init}.
4114 Warn if a structure is given the packed attribute, but the packed
4115 attribute has no effect on the layout or size of the structure.
4116 Such structures may be mis-aligned for little benefit. For
4117 instance, in this code, the variable @code{f.x} in @code{struct bar}
4118 will be misaligned even though @code{struct bar} does not itself
4119 have the packed attribute:
4126 @} __attribute__((packed));
4134 @item -Wpacked-bitfield-compat
4135 @opindex Wpacked-bitfield-compat
4136 @opindex Wno-packed-bitfield-compat
4137 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4138 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4139 the change can lead to differences in the structure layout. GCC
4140 informs you when the offset of such a field has changed in GCC 4.4.
4141 For example there is no longer a 4-bit padding between field @code{a}
4142 and @code{b} in this structure:
4149 @} __attribute__ ((packed));
4152 This warning is enabled by default. Use
4153 @option{-Wno-packed-bitfield-compat} to disable this warning.
4158 Warn if padding is included in a structure, either to align an element
4159 of the structure or to align the whole structure. Sometimes when this
4160 happens it is possible to rearrange the fields of the structure to
4161 reduce the padding and so make the structure smaller.
4163 @item -Wredundant-decls
4164 @opindex Wredundant-decls
4165 @opindex Wno-redundant-decls
4166 Warn if anything is declared more than once in the same scope, even in
4167 cases where multiple declaration is valid and changes nothing.
4169 @item -Wnested-externs @r{(C and Objective-C only)}
4170 @opindex Wnested-externs
4171 @opindex Wno-nested-externs
4172 Warn if an @code{extern} declaration is encountered within a function.
4174 @item -Wunreachable-code
4175 @opindex Wunreachable-code
4176 @opindex Wno-unreachable-code
4177 Warn if the compiler detects that code will never be executed.
4179 This option is intended to warn when the compiler detects that at
4180 least a whole line of source code will never be executed, because
4181 some condition is never satisfied or because it is after a
4182 procedure that never returns.
4184 It is possible for this option to produce a warning even though there
4185 are circumstances under which part of the affected line can be executed,
4186 so care should be taken when removing apparently-unreachable code.
4188 For instance, when a function is inlined, a warning may mean that the
4189 line is unreachable in only one inlined copy of the function.
4191 This option is not made part of @option{-Wall} because in a debugging
4192 version of a program there is often substantial code which checks
4193 correct functioning of the program and is, hopefully, unreachable
4194 because the program does work. Another common use of unreachable
4195 code is to provide behavior which is selectable at compile-time.
4200 Warn if a function can not be inlined and it was declared as inline.
4201 Even with this option, the compiler will not warn about failures to
4202 inline functions declared in system headers.
4204 The compiler uses a variety of heuristics to determine whether or not
4205 to inline a function. For example, the compiler takes into account
4206 the size of the function being inlined and the amount of inlining
4207 that has already been done in the current function. Therefore,
4208 seemingly insignificant changes in the source program can cause the
4209 warnings produced by @option{-Winline} to appear or disappear.
4211 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4212 @opindex Wno-invalid-offsetof
4213 @opindex Winvalid-offsetof
4214 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4215 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4216 to a non-POD type is undefined. In existing C++ implementations,
4217 however, @samp{offsetof} typically gives meaningful results even when
4218 applied to certain kinds of non-POD types. (Such as a simple
4219 @samp{struct} that fails to be a POD type only by virtue of having a
4220 constructor.) This flag is for users who are aware that they are
4221 writing nonportable code and who have deliberately chosen to ignore the
4224 The restrictions on @samp{offsetof} may be relaxed in a future version
4225 of the C++ standard.
4227 @item -Wno-int-to-pointer-cast @r{(C and Objective-C only)}
4228 @opindex Wno-int-to-pointer-cast
4229 @opindex Wint-to-pointer-cast
4230 Suppress warnings from casts to pointer type of an integer of a
4233 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4234 @opindex Wno-pointer-to-int-cast
4235 @opindex Wpointer-to-int-cast
4236 Suppress warnings from casts from a pointer to an integer type of a
4240 @opindex Winvalid-pch
4241 @opindex Wno-invalid-pch
4242 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4243 the search path but can't be used.
4247 @opindex Wno-long-long
4248 Warn if @samp{long long} type is used. This is enabled by either
4249 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4250 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4252 @item -Wvariadic-macros
4253 @opindex Wvariadic-macros
4254 @opindex Wno-variadic-macros
4255 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4256 alternate syntax when in pedantic ISO C99 mode. This is default.
4257 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4262 Warn if variable length array is used in the code.
4263 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4264 the variable length array.
4266 @item -Wvolatile-register-var
4267 @opindex Wvolatile-register-var
4268 @opindex Wno-volatile-register-var
4269 Warn if a register variable is declared volatile. The volatile
4270 modifier does not inhibit all optimizations that may eliminate reads
4271 and/or writes to register variables. This warning is enabled by
4274 @item -Wdisabled-optimization
4275 @opindex Wdisabled-optimization
4276 @opindex Wno-disabled-optimization
4277 Warn if a requested optimization pass is disabled. This warning does
4278 not generally indicate that there is anything wrong with your code; it
4279 merely indicates that GCC's optimizers were unable to handle the code
4280 effectively. Often, the problem is that your code is too big or too
4281 complex; GCC will refuse to optimize programs when the optimization
4282 itself is likely to take inordinate amounts of time.
4284 @item -Wpointer-sign @r{(C and Objective-C only)}
4285 @opindex Wpointer-sign
4286 @opindex Wno-pointer-sign
4287 Warn for pointer argument passing or assignment with different signedness.
4288 This option is only supported for C and Objective-C@. It is implied by
4289 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4290 @option{-Wno-pointer-sign}.
4292 @item -Wstack-protector
4293 @opindex Wstack-protector
4294 @opindex Wno-stack-protector
4295 This option is only active when @option{-fstack-protector} is active. It
4296 warns about functions that will not be protected against stack smashing.
4299 @opindex Wno-mudflap
4300 Suppress warnings about constructs that cannot be instrumented by
4303 @item -Woverlength-strings
4304 @opindex Woverlength-strings
4305 @opindex Wno-overlength-strings
4306 Warn about string constants which are longer than the ``minimum
4307 maximum'' length specified in the C standard. Modern compilers
4308 generally allow string constants which are much longer than the
4309 standard's minimum limit, but very portable programs should avoid
4310 using longer strings.
4312 The limit applies @emph{after} string constant concatenation, and does
4313 not count the trailing NUL@. In C89, the limit was 509 characters; in
4314 C99, it was raised to 4095. C++98 does not specify a normative
4315 minimum maximum, so we do not diagnose overlength strings in C++@.
4317 This option is implied by @option{-pedantic}, and can be disabled with
4318 @option{-Wno-overlength-strings}.
4320 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
4321 @opindex Wunsuffixed-float-constants
4323 GCC will issue a warning for any floating constant that does not have
4324 a suffix. When used together with @option{-Wsystem-headers} it will
4325 warn about such constants in system header files. This can be useful
4326 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4327 from the decimal floating-point extension to C99.
4330 @node Debugging Options
4331 @section Options for Debugging Your Program or GCC
4332 @cindex options, debugging
4333 @cindex debugging information options
4335 GCC has various special options that are used for debugging
4336 either your program or GCC:
4341 Produce debugging information in the operating system's native format
4342 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4345 On most systems that use stabs format, @option{-g} enables use of extra
4346 debugging information that only GDB can use; this extra information
4347 makes debugging work better in GDB but will probably make other debuggers
4349 refuse to read the program. If you want to control for certain whether
4350 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4351 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4353 GCC allows you to use @option{-g} with
4354 @option{-O}. The shortcuts taken by optimized code may occasionally
4355 produce surprising results: some variables you declared may not exist
4356 at all; flow of control may briefly move where you did not expect it;
4357 some statements may not be executed because they compute constant
4358 results or their values were already at hand; some statements may
4359 execute in different places because they were moved out of loops.
4361 Nevertheless it proves possible to debug optimized output. This makes
4362 it reasonable to use the optimizer for programs that might have bugs.
4364 The following options are useful when GCC is generated with the
4365 capability for more than one debugging format.
4369 Produce debugging information for use by GDB@. This means to use the
4370 most expressive format available (DWARF 2, stabs, or the native format
4371 if neither of those are supported), including GDB extensions if at all
4376 Produce debugging information in stabs format (if that is supported),
4377 without GDB extensions. This is the format used by DBX on most BSD
4378 systems. On MIPS, Alpha and System V Release 4 systems this option
4379 produces stabs debugging output which is not understood by DBX or SDB@.
4380 On System V Release 4 systems this option requires the GNU assembler.
4382 @item -feliminate-unused-debug-symbols
4383 @opindex feliminate-unused-debug-symbols
4384 Produce debugging information in stabs format (if that is supported),
4385 for only symbols that are actually used.
4387 @item -femit-class-debug-always
4388 Instead of emitting debugging information for a C++ class in only one
4389 object file, emit it in all object files using the class. This option
4390 should be used only with debuggers that are unable to handle the way GCC
4391 normally emits debugging information for classes because using this
4392 option will increase the size of debugging information by as much as a
4397 Produce debugging information in stabs format (if that is supported),
4398 using GNU extensions understood only by the GNU debugger (GDB)@. The
4399 use of these extensions is likely to make other debuggers crash or
4400 refuse to read the program.
4404 Produce debugging information in COFF format (if that is supported).
4405 This is the format used by SDB on most System V systems prior to
4410 Produce debugging information in XCOFF format (if that is supported).
4411 This is the format used by the DBX debugger on IBM RS/6000 systems.
4415 Produce debugging information in XCOFF format (if that is supported),
4416 using GNU extensions understood only by the GNU debugger (GDB)@. The
4417 use of these extensions is likely to make other debuggers crash or
4418 refuse to read the program, and may cause assemblers other than the GNU
4419 assembler (GAS) to fail with an error.
4421 @item -gdwarf-@var{version}
4422 @opindex gdwarf-@var{version}
4423 Produce debugging information in DWARF format (if that is
4424 supported). This is the format used by DBX on IRIX 6. The value
4425 of @var{version} may be either 2, 3 or 4; the default version is 2.
4427 Note that with DWARF version 2 some ports require, and will always
4428 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4430 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
4431 for maximum benefit.
4433 @item -gstrict-dwarf
4434 @opindex gstrict-dwarf
4435 Disallow using extensions of later DWARF standard version than selected
4436 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
4437 DWARF extensions from later standard versions is allowed.
4439 @item -gno-strict-dwarf
4440 @opindex gno-strict-dwarf
4441 Allow using extensions of later DWARF standard version than selected with
4442 @option{-gdwarf-@var{version}}.
4446 Produce debugging information in VMS debug format (if that is
4447 supported). This is the format used by DEBUG on VMS systems.
4450 @itemx -ggdb@var{level}
4451 @itemx -gstabs@var{level}
4452 @itemx -gcoff@var{level}
4453 @itemx -gxcoff@var{level}
4454 @itemx -gvms@var{level}
4455 Request debugging information and also use @var{level} to specify how
4456 much information. The default level is 2.
4458 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4461 Level 1 produces minimal information, enough for making backtraces in
4462 parts of the program that you don't plan to debug. This includes
4463 descriptions of functions and external variables, but no information
4464 about local variables and no line numbers.
4466 Level 3 includes extra information, such as all the macro definitions
4467 present in the program. Some debuggers support macro expansion when
4468 you use @option{-g3}.
4470 @option{-gdwarf-2} does not accept a concatenated debug level, because
4471 GCC used to support an option @option{-gdwarf} that meant to generate
4472 debug information in version 1 of the DWARF format (which is very
4473 different from version 2), and it would have been too confusing. That
4474 debug format is long obsolete, but the option cannot be changed now.
4475 Instead use an additional @option{-g@var{level}} option to change the
4476 debug level for DWARF.
4480 Turn off generation of debug info, if leaving out this option would have
4481 generated it, or turn it on at level 2 otherwise. The position of this
4482 argument in the command line does not matter, it takes effect after all
4483 other options are processed, and it does so only once, no matter how
4484 many times it is given. This is mainly intended to be used with
4485 @option{-fcompare-debug}.
4487 @item -fdump-final-insns@r{[}=@var{file}@r{]}
4488 @opindex fdump-final-insns
4489 Dump the final internal representation (RTL) to @var{file}. If the
4490 optional argument is omitted (or if @var{file} is @code{.}), the name
4491 of the dump file will be determined by appending @code{.gkd} to the
4492 compilation output file name.
4494 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4495 @opindex fcompare-debug
4496 @opindex fno-compare-debug
4497 If no error occurs during compilation, run the compiler a second time,
4498 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4499 passed to the second compilation. Dump the final internal
4500 representation in both compilations, and print an error if they differ.
4502 If the equal sign is omitted, the default @option{-gtoggle} is used.
4504 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4505 and nonzero, implicitly enables @option{-fcompare-debug}. If
4506 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4507 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4510 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4511 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4512 of the final representation and the second compilation, preventing even
4513 @env{GCC_COMPARE_DEBUG} from taking effect.
4515 To verify full coverage during @option{-fcompare-debug} testing, set
4516 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4517 which GCC will reject as an invalid option in any actual compilation
4518 (rather than preprocessing, assembly or linking). To get just a
4519 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4520 not overridden} will do.
4522 @item -fcompare-debug-second
4523 @opindex fcompare-debug-second
4524 This option is implicitly passed to the compiler for the second
4525 compilation requested by @option{-fcompare-debug}, along with options to
4526 silence warnings, and omitting other options that would cause
4527 side-effect compiler outputs to files or to the standard output. Dump
4528 files and preserved temporary files are renamed so as to contain the
4529 @code{.gk} additional extension during the second compilation, to avoid
4530 overwriting those generated by the first.
4532 When this option is passed to the compiler driver, it causes the
4533 @emph{first} compilation to be skipped, which makes it useful for little
4534 other than debugging the compiler proper.
4536 @item -feliminate-dwarf2-dups
4537 @opindex feliminate-dwarf2-dups
4538 Compress DWARF2 debugging information by eliminating duplicated
4539 information about each symbol. This option only makes sense when
4540 generating DWARF2 debugging information with @option{-gdwarf-2}.
4542 @item -femit-struct-debug-baseonly
4543 Emit debug information for struct-like types
4544 only when the base name of the compilation source file
4545 matches the base name of file in which the struct was defined.
4547 This option substantially reduces the size of debugging information,
4548 but at significant potential loss in type information to the debugger.
4549 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4550 See @option{-femit-struct-debug-detailed} for more detailed control.
4552 This option works only with DWARF 2.
4554 @item -femit-struct-debug-reduced
4555 Emit debug information for struct-like types
4556 only when the base name of the compilation source file
4557 matches the base name of file in which the type was defined,
4558 unless the struct is a template or defined in a system header.
4560 This option significantly reduces the size of debugging information,
4561 with some potential loss in type information to the debugger.
4562 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4563 See @option{-femit-struct-debug-detailed} for more detailed control.
4565 This option works only with DWARF 2.
4567 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4568 Specify the struct-like types
4569 for which the compiler will generate debug information.
4570 The intent is to reduce duplicate struct debug information
4571 between different object files within the same program.
4573 This option is a detailed version of
4574 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4575 which will serve for most needs.
4577 A specification has the syntax
4578 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4580 The optional first word limits the specification to
4581 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4582 A struct type is used directly when it is the type of a variable, member.
4583 Indirect uses arise through pointers to structs.
4584 That is, when use of an incomplete struct would be legal, the use is indirect.
4586 @samp{struct one direct; struct two * indirect;}.
4588 The optional second word limits the specification to
4589 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4590 Generic structs are a bit complicated to explain.
4591 For C++, these are non-explicit specializations of template classes,
4592 or non-template classes within the above.
4593 Other programming languages have generics,
4594 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4596 The third word specifies the source files for those
4597 structs for which the compiler will emit debug information.
4598 The values @samp{none} and @samp{any} have the normal meaning.
4599 The value @samp{base} means that
4600 the base of name of the file in which the type declaration appears
4601 must match the base of the name of the main compilation file.
4602 In practice, this means that
4603 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4604 but types declared in other header will not.
4605 The value @samp{sys} means those types satisfying @samp{base}
4606 or declared in system or compiler headers.
4608 You may need to experiment to determine the best settings for your application.
4610 The default is @samp{-femit-struct-debug-detailed=all}.
4612 This option works only with DWARF 2.
4614 @item -fenable-icf-debug
4615 @opindex fenable-icf-debug
4616 Generate additional debug information to support identical code folding (ICF).
4617 This option only works with DWARF version 2 or higher.
4619 @item -fno-merge-debug-strings
4620 @opindex fmerge-debug-strings
4621 @opindex fno-merge-debug-strings
4622 Direct the linker to not merge together strings in the debugging
4623 information which are identical in different object files. Merging is
4624 not supported by all assemblers or linkers. Merging decreases the size
4625 of the debug information in the output file at the cost of increasing
4626 link processing time. Merging is enabled by default.
4628 @item -fdebug-prefix-map=@var{old}=@var{new}
4629 @opindex fdebug-prefix-map
4630 When compiling files in directory @file{@var{old}}, record debugging
4631 information describing them as in @file{@var{new}} instead.
4633 @item -fno-dwarf2-cfi-asm
4634 @opindex fdwarf2-cfi-asm
4635 @opindex fno-dwarf2-cfi-asm
4636 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4637 instead of using GAS @code{.cfi_*} directives.
4639 @cindex @command{prof}
4642 Generate extra code to write profile information suitable for the
4643 analysis program @command{prof}. You must use this option when compiling
4644 the source files you want data about, and you must also use it when
4647 @cindex @command{gprof}
4650 Generate extra code to write profile information suitable for the
4651 analysis program @command{gprof}. You must use this option when compiling
4652 the source files you want data about, and you must also use it when
4657 Makes the compiler print out each function name as it is compiled, and
4658 print some statistics about each pass when it finishes.
4661 @opindex ftime-report
4662 Makes the compiler print some statistics about the time consumed by each
4663 pass when it finishes.
4666 @opindex fmem-report
4667 Makes the compiler print some statistics about permanent memory
4668 allocation when it finishes.
4670 @item -fpre-ipa-mem-report
4671 @opindex fpre-ipa-mem-report
4672 @item -fpost-ipa-mem-report
4673 @opindex fpost-ipa-mem-report
4674 Makes the compiler print some statistics about permanent memory
4675 allocation before or after interprocedural optimization.
4677 @item -fprofile-arcs
4678 @opindex fprofile-arcs
4679 Add code so that program flow @dfn{arcs} are instrumented. During
4680 execution the program records how many times each branch and call is
4681 executed and how many times it is taken or returns. When the compiled
4682 program exits it saves this data to a file called
4683 @file{@var{auxname}.gcda} for each source file. The data may be used for
4684 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4685 test coverage analysis (@option{-ftest-coverage}). Each object file's
4686 @var{auxname} is generated from the name of the output file, if
4687 explicitly specified and it is not the final executable, otherwise it is
4688 the basename of the source file. In both cases any suffix is removed
4689 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4690 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4691 @xref{Cross-profiling}.
4693 @cindex @command{gcov}
4697 This option is used to compile and link code instrumented for coverage
4698 analysis. The option is a synonym for @option{-fprofile-arcs}
4699 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4700 linking). See the documentation for those options for more details.
4705 Compile the source files with @option{-fprofile-arcs} plus optimization
4706 and code generation options. For test coverage analysis, use the
4707 additional @option{-ftest-coverage} option. You do not need to profile
4708 every source file in a program.
4711 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4712 (the latter implies the former).
4715 Run the program on a representative workload to generate the arc profile
4716 information. This may be repeated any number of times. You can run
4717 concurrent instances of your program, and provided that the file system
4718 supports locking, the data files will be correctly updated. Also
4719 @code{fork} calls are detected and correctly handled (double counting
4723 For profile-directed optimizations, compile the source files again with
4724 the same optimization and code generation options plus
4725 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4726 Control Optimization}).
4729 For test coverage analysis, use @command{gcov} to produce human readable
4730 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4731 @command{gcov} documentation for further information.
4735 With @option{-fprofile-arcs}, for each function of your program GCC
4736 creates a program flow graph, then finds a spanning tree for the graph.
4737 Only arcs that are not on the spanning tree have to be instrumented: the
4738 compiler adds code to count the number of times that these arcs are
4739 executed. When an arc is the only exit or only entrance to a block, the
4740 instrumentation code can be added to the block; otherwise, a new basic
4741 block must be created to hold the instrumentation code.
4744 @item -ftest-coverage
4745 @opindex ftest-coverage
4746 Produce a notes file that the @command{gcov} code-coverage utility
4747 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4748 show program coverage. Each source file's note file is called
4749 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4750 above for a description of @var{auxname} and instructions on how to
4751 generate test coverage data. Coverage data will match the source files
4752 more closely, if you do not optimize.
4754 @item -fdbg-cnt-list
4755 @opindex fdbg-cnt-list
4756 Print the name and the counter upperbound for all debug counters.
4758 @item -fdbg-cnt=@var{counter-value-list}
4760 Set the internal debug counter upperbound. @var{counter-value-list}
4761 is a comma-separated list of @var{name}:@var{value} pairs
4762 which sets the upperbound of each debug counter @var{name} to @var{value}.
4763 All debug counters have the initial upperbound of @var{UINT_MAX},
4764 thus dbg_cnt() returns true always unless the upperbound is set by this option.
4765 e.g. With -fdbg-cnt=dce:10,tail_call:0
4766 dbg_cnt(dce) will return true only for first 10 invocations
4767 and dbg_cnt(tail_call) will return false always.
4769 @item -d@var{letters}
4770 @itemx -fdump-rtl-@var{pass}
4772 Says to make debugging dumps during compilation at times specified by
4773 @var{letters}. This is used for debugging the RTL-based passes of the
4774 compiler. The file names for most of the dumps are made by appending
4775 a pass number and a word to the @var{dumpname}, and the files are
4776 created in the directory of the output file. @var{dumpname} is
4777 generated from the name of the output file, if explicitly specified
4778 and it is not an executable, otherwise it is the basename of the
4779 source file. These switches may have different effects when
4780 @option{-E} is used for preprocessing.
4782 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
4783 @option{-d} option @var{letters}. Here are the possible
4784 letters for use in @var{pass} and @var{letters}, and their meanings:
4788 @item -fdump-rtl-alignments
4789 @opindex fdump-rtl-alignments
4790 Dump after branch alignments have been computed.
4792 @item -fdump-rtl-asmcons
4793 @opindex fdump-rtl-asmcons
4794 Dump after fixing rtl statements that have unsatisfied in/out constraints.
4796 @item -fdump-rtl-auto_inc_dec
4797 @opindex fdump-rtl-auto_inc_dec
4798 Dump after auto-inc-dec discovery. This pass is only run on
4799 architectures that have auto inc or auto dec instructions.
4801 @item -fdump-rtl-barriers
4802 @opindex fdump-rtl-barriers
4803 Dump after cleaning up the barrier instructions.
4805 @item -fdump-rtl-bbpart
4806 @opindex fdump-rtl-bbpart
4807 Dump after partitioning hot and cold basic blocks.
4809 @item -fdump-rtl-bbro
4810 @opindex fdump-rtl-bbro
4811 Dump after block reordering.
4813 @item -fdump-rtl-btl1
4814 @itemx -fdump-rtl-btl2
4815 @opindex fdump-rtl-btl2
4816 @opindex fdump-rtl-btl2
4817 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
4818 after the two branch
4819 target load optimization passes.
4821 @item -fdump-rtl-bypass
4822 @opindex fdump-rtl-bypass
4823 Dump after jump bypassing and control flow optimizations.
4825 @item -fdump-rtl-combine
4826 @opindex fdump-rtl-combine
4827 Dump after the RTL instruction combination pass.
4829 @item -fdump-rtl-compgotos
4830 @opindex fdump-rtl-compgotos
4831 Dump after duplicating the computed gotos.
4833 @item -fdump-rtl-ce1
4834 @itemx -fdump-rtl-ce2
4835 @itemx -fdump-rtl-ce3
4836 @opindex fdump-rtl-ce1
4837 @opindex fdump-rtl-ce2
4838 @opindex fdump-rtl-ce3
4839 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
4840 @option{-fdump-rtl-ce3} enable dumping after the three
4841 if conversion passes.
4843 @itemx -fdump-rtl-cprop_hardreg
4844 @opindex fdump-rtl-cprop_hardreg
4845 Dump after hard register copy propagation.
4847 @itemx -fdump-rtl-csa
4848 @opindex fdump-rtl-csa
4849 Dump after combining stack adjustments.
4851 @item -fdump-rtl-cse1
4852 @itemx -fdump-rtl-cse2
4853 @opindex fdump-rtl-cse1
4854 @opindex fdump-rtl-cse2
4855 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
4856 the two common sub-expression elimination passes.
4858 @itemx -fdump-rtl-dce
4859 @opindex fdump-rtl-dce
4860 Dump after the standalone dead code elimination passes.
4862 @itemx -fdump-rtl-dbr
4863 @opindex fdump-rtl-dbr
4864 Dump after delayed branch scheduling.
4866 @item -fdump-rtl-dce1
4867 @itemx -fdump-rtl-dce2
4868 @opindex fdump-rtl-dce1
4869 @opindex fdump-rtl-dce2
4870 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
4871 the two dead store elimination passes.
4874 @opindex fdump-rtl-eh
4875 Dump after finalization of EH handling code.
4877 @item -fdump-rtl-eh_ranges
4878 @opindex fdump-rtl-eh_ranges
4879 Dump after conversion of EH handling range regions.
4881 @item -fdump-rtl-expand
4882 @opindex fdump-rtl-expand
4883 Dump after RTL generation.
4885 @item -fdump-rtl-fwprop1
4886 @itemx -fdump-rtl-fwprop2
4887 @opindex fdump-rtl-fwprop1
4888 @opindex fdump-rtl-fwprop2
4889 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
4890 dumping after the two forward propagation passes.
4892 @item -fdump-rtl-gcse1
4893 @itemx -fdump-rtl-gcse2
4894 @opindex fdump-rtl-gcse1
4895 @opindex fdump-rtl-gcse2
4896 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
4897 after global common subexpression elimination.
4899 @item -fdump-rtl-init-regs
4900 @opindex fdump-rtl-init-regs
4901 Dump after the initialization of the registers.
4903 @item -fdump-rtl-initvals
4904 @opindex fdump-rtl-initvals
4905 Dump after the computation of the initial value sets.
4907 @itemx -fdump-rtl-into_cfglayout
4908 @opindex fdump-rtl-into_cfglayout
4909 Dump after converting to cfglayout mode.
4911 @item -fdump-rtl-ira
4912 @opindex fdump-rtl-ira
4913 Dump after iterated register allocation.
4915 @item -fdump-rtl-jump
4916 @opindex fdump-rtl-jump
4917 Dump after the second jump optimization.
4919 @item -fdump-rtl-loop2
4920 @opindex fdump-rtl-loop2
4921 @option{-fdump-rtl-loop2} enables dumping after the rtl
4922 loop optimization passes.
4924 @item -fdump-rtl-mach
4925 @opindex fdump-rtl-mach
4926 Dump after performing the machine dependent reorganization pass, if that
4929 @item -fdump-rtl-mode_sw
4930 @opindex fdump-rtl-mode_sw
4931 Dump after removing redundant mode switches.
4933 @item -fdump-rtl-rnreg
4934 @opindex fdump-rtl-rnreg
4935 Dump after register renumbering.
4937 @itemx -fdump-rtl-outof_cfglayout
4938 @opindex fdump-rtl-outof_cfglayout
4939 Dump after converting from cfglayout mode.
4941 @item -fdump-rtl-peephole2
4942 @opindex fdump-rtl-peephole2
4943 Dump after the peephole pass.
4945 @item -fdump-rtl-postreload
4946 @opindex fdump-rtl-postreload
4947 Dump after post-reload optimizations.
4949 @itemx -fdump-rtl-pro_and_epilogue
4950 @opindex fdump-rtl-pro_and_epilogue
4951 Dump after generating the function pro and epilogues.
4953 @item -fdump-rtl-regmove
4954 @opindex fdump-rtl-regmove
4955 Dump after the register move pass.
4957 @item -fdump-rtl-sched1
4958 @itemx -fdump-rtl-sched2
4959 @opindex fdump-rtl-sched1
4960 @opindex fdump-rtl-sched2
4961 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
4962 after the basic block scheduling passes.
4964 @item -fdump-rtl-see
4965 @opindex fdump-rtl-see
4966 Dump after sign extension elimination.
4968 @item -fdump-rtl-seqabstr
4969 @opindex fdump-rtl-seqabstr
4970 Dump after common sequence discovery.
4972 @item -fdump-rtl-shorten
4973 @opindex fdump-rtl-shorten
4974 Dump after shortening branches.
4976 @item -fdump-rtl-sibling
4977 @opindex fdump-rtl-sibling
4978 Dump after sibling call optimizations.
4980 @item -fdump-rtl-split1
4981 @itemx -fdump-rtl-split2
4982 @itemx -fdump-rtl-split3
4983 @itemx -fdump-rtl-split4
4984 @itemx -fdump-rtl-split5
4985 @opindex fdump-rtl-split1
4986 @opindex fdump-rtl-split2
4987 @opindex fdump-rtl-split3
4988 @opindex fdump-rtl-split4
4989 @opindex fdump-rtl-split5
4990 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
4991 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
4992 @option{-fdump-rtl-split5} enable dumping after five rounds of
4993 instruction splitting.
4995 @item -fdump-rtl-sms
4996 @opindex fdump-rtl-sms
4997 Dump after modulo scheduling. This pass is only run on some
5000 @item -fdump-rtl-stack
5001 @opindex fdump-rtl-stack
5002 Dump after conversion from GCC's "flat register file" registers to the
5003 x87's stack-like registers. This pass is only run on x86 variants.
5005 @item -fdump-rtl-subreg1
5006 @itemx -fdump-rtl-subreg2
5007 @opindex fdump-rtl-subreg1
5008 @opindex fdump-rtl-subreg2
5009 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
5010 the two subreg expansion passes.
5012 @item -fdump-rtl-unshare
5013 @opindex fdump-rtl-unshare
5014 Dump after all rtl has been unshared.
5016 @item -fdump-rtl-vartrack
5017 @opindex fdump-rtl-vartrack
5018 Dump after variable tracking.
5020 @item -fdump-rtl-vregs
5021 @opindex fdump-rtl-vregs
5022 Dump after converting virtual registers to hard registers.
5024 @item -fdump-rtl-web
5025 @opindex fdump-rtl-web
5026 Dump after live range splitting.
5028 @item -fdump-rtl-regclass
5029 @itemx -fdump-rtl-subregs_of_mode_init
5030 @itemx -fdump-rtl-subregs_of_mode_finish
5031 @itemx -fdump-rtl-dfinit
5032 @itemx -fdump-rtl-dfinish
5033 @opindex fdump-rtl-regclass
5034 @opindex fdump-rtl-subregs_of_mode_init
5035 @opindex fdump-rtl-subregs_of_mode_finish
5036 @opindex fdump-rtl-dfinit
5037 @opindex fdump-rtl-dfinish
5038 These dumps are defined but always produce empty files.
5040 @item -fdump-rtl-all
5041 @opindex fdump-rtl-all
5042 Produce all the dumps listed above.
5046 Annotate the assembler output with miscellaneous debugging information.
5050 Dump all macro definitions, at the end of preprocessing, in addition to
5055 Produce a core dump whenever an error occurs.
5059 Print statistics on memory usage, at the end of the run, to
5064 Annotate the assembler output with a comment indicating which
5065 pattern and alternative was used. The length of each instruction is
5070 Dump the RTL in the assembler output as a comment before each instruction.
5071 Also turns on @option{-dp} annotation.
5075 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
5076 dump a representation of the control flow graph suitable for viewing with VCG
5077 to @file{@var{file}.@var{pass}.vcg}.
5081 Just generate RTL for a function instead of compiling it. Usually used
5082 with @option{-fdump-rtl-expand}.
5086 Dump debugging information during parsing, to standard error.
5090 @opindex fdump-noaddr
5091 When doing debugging dumps, suppress address output. This makes it more
5092 feasible to use diff on debugging dumps for compiler invocations with
5093 different compiler binaries and/or different
5094 text / bss / data / heap / stack / dso start locations.
5096 @item -fdump-unnumbered
5097 @opindex fdump-unnumbered
5098 When doing debugging dumps, suppress instruction numbers and address output.
5099 This makes it more feasible to use diff on debugging dumps for compiler
5100 invocations with different options, in particular with and without
5103 @item -fdump-unnumbered-links
5104 @opindex fdump-unnumbered-links
5105 When doing debugging dumps (see @option{-d} option above), suppress
5106 instruction numbers for the links to the previous and next instructions
5109 @item -fdump-translation-unit @r{(C++ only)}
5110 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5111 @opindex fdump-translation-unit
5112 Dump a representation of the tree structure for the entire translation
5113 unit to a file. The file name is made by appending @file{.tu} to the
5114 source file name, and the file is created in the same directory as the
5115 output file. If the @samp{-@var{options}} form is used, @var{options}
5116 controls the details of the dump as described for the
5117 @option{-fdump-tree} options.
5119 @item -fdump-class-hierarchy @r{(C++ only)}
5120 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5121 @opindex fdump-class-hierarchy
5122 Dump a representation of each class's hierarchy and virtual function
5123 table layout to a file. The file name is made by appending
5124 @file{.class} to the source file name, and the file is created in the
5125 same directory as the output file. If the @samp{-@var{options}} form
5126 is used, @var{options} controls the details of the dump as described
5127 for the @option{-fdump-tree} options.
5129 @item -fdump-ipa-@var{switch}
5131 Control the dumping at various stages of inter-procedural analysis
5132 language tree to a file. The file name is generated by appending a
5133 switch specific suffix to the source file name, and the file is created
5134 in the same directory as the output file. The following dumps are
5139 Enables all inter-procedural analysis dumps.
5142 Dumps information about call-graph optimization, unused function removal,
5143 and inlining decisions.
5146 Dump after function inlining.
5150 @item -fdump-statistics-@var{option}
5151 @opindex fdump-statistics
5152 Enable and control dumping of pass statistics in a separate file. The
5153 file name is generated by appending a suffix ending in
5154 @samp{.statistics} to the source file name, and the file is created in
5155 the same directory as the output file. If the @samp{-@var{option}}
5156 form is used, @samp{-stats} will cause counters to be summed over the
5157 whole compilation unit while @samp{-details} will dump every event as
5158 the passes generate them. The default with no option is to sum
5159 counters for each function compiled.
5161 @item -fdump-tree-@var{switch}
5162 @itemx -fdump-tree-@var{switch}-@var{options}
5164 Control the dumping at various stages of processing the intermediate
5165 language tree to a file. The file name is generated by appending a
5166 switch specific suffix to the source file name, and the file is
5167 created in the same directory as the output file. If the
5168 @samp{-@var{options}} form is used, @var{options} is a list of
5169 @samp{-} separated options that control the details of the dump. Not
5170 all options are applicable to all dumps, those which are not
5171 meaningful will be ignored. The following options are available
5175 Print the address of each node. Usually this is not meaningful as it
5176 changes according to the environment and source file. Its primary use
5177 is for tying up a dump file with a debug environment.
5179 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
5180 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
5181 use working backward from mangled names in the assembly file.
5183 Inhibit dumping of members of a scope or body of a function merely
5184 because that scope has been reached. Only dump such items when they
5185 are directly reachable by some other path. When dumping pretty-printed
5186 trees, this option inhibits dumping the bodies of control structures.
5188 Print a raw representation of the tree. By default, trees are
5189 pretty-printed into a C-like representation.
5191 Enable more detailed dumps (not honored by every dump option).
5193 Enable dumping various statistics about the pass (not honored by every dump
5196 Enable showing basic block boundaries (disabled in raw dumps).
5198 Enable showing virtual operands for every statement.
5200 Enable showing line numbers for statements.
5202 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5204 Enable showing the tree dump for each statement.
5206 Enable showing the EH region number holding each statement.
5208 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5209 and @option{lineno}.
5212 The following tree dumps are possible:
5216 @opindex fdump-tree-original
5217 Dump before any tree based optimization, to @file{@var{file}.original}.
5220 @opindex fdump-tree-optimized
5221 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5224 @opindex fdump-tree-gimple
5225 Dump each function before and after the gimplification pass to a file. The
5226 file name is made by appending @file{.gimple} to the source file name.
5229 @opindex fdump-tree-cfg
5230 Dump the control flow graph of each function to a file. The file name is
5231 made by appending @file{.cfg} to the source file name.
5234 @opindex fdump-tree-vcg
5235 Dump the control flow graph of each function to a file in VCG format. The
5236 file name is made by appending @file{.vcg} to the source file name. Note
5237 that if the file contains more than one function, the generated file cannot
5238 be used directly by VCG@. You will need to cut and paste each function's
5239 graph into its own separate file first.
5242 @opindex fdump-tree-ch
5243 Dump each function after copying loop headers. The file name is made by
5244 appending @file{.ch} to the source file name.
5247 @opindex fdump-tree-ssa
5248 Dump SSA related information to a file. The file name is made by appending
5249 @file{.ssa} to the source file name.
5252 @opindex fdump-tree-alias
5253 Dump aliasing information for each function. The file name is made by
5254 appending @file{.alias} to the source file name.
5257 @opindex fdump-tree-ccp
5258 Dump each function after CCP@. The file name is made by appending
5259 @file{.ccp} to the source file name.
5262 @opindex fdump-tree-storeccp
5263 Dump each function after STORE-CCP@. The file name is made by appending
5264 @file{.storeccp} to the source file name.
5267 @opindex fdump-tree-pre
5268 Dump trees after partial redundancy elimination. The file name is made
5269 by appending @file{.pre} to the source file name.
5272 @opindex fdump-tree-fre
5273 Dump trees after full redundancy elimination. The file name is made
5274 by appending @file{.fre} to the source file name.
5277 @opindex fdump-tree-copyprop
5278 Dump trees after copy propagation. The file name is made
5279 by appending @file{.copyprop} to the source file name.
5281 @item store_copyprop
5282 @opindex fdump-tree-store_copyprop
5283 Dump trees after store copy-propagation. The file name is made
5284 by appending @file{.store_copyprop} to the source file name.
5287 @opindex fdump-tree-dce
5288 Dump each function after dead code elimination. The file name is made by
5289 appending @file{.dce} to the source file name.
5292 @opindex fdump-tree-mudflap
5293 Dump each function after adding mudflap instrumentation. The file name is
5294 made by appending @file{.mudflap} to the source file name.
5297 @opindex fdump-tree-sra
5298 Dump each function after performing scalar replacement of aggregates. The
5299 file name is made by appending @file{.sra} to the source file name.
5302 @opindex fdump-tree-sink
5303 Dump each function after performing code sinking. The file name is made
5304 by appending @file{.sink} to the source file name.
5307 @opindex fdump-tree-dom
5308 Dump each function after applying dominator tree optimizations. The file
5309 name is made by appending @file{.dom} to the source file name.
5312 @opindex fdump-tree-dse
5313 Dump each function after applying dead store elimination. The file
5314 name is made by appending @file{.dse} to the source file name.
5317 @opindex fdump-tree-phiopt
5318 Dump each function after optimizing PHI nodes into straightline code. The file
5319 name is made by appending @file{.phiopt} to the source file name.
5322 @opindex fdump-tree-forwprop
5323 Dump each function after forward propagating single use variables. The file
5324 name is made by appending @file{.forwprop} to the source file name.
5327 @opindex fdump-tree-copyrename
5328 Dump each function after applying the copy rename optimization. The file
5329 name is made by appending @file{.copyrename} to the source file name.
5332 @opindex fdump-tree-nrv
5333 Dump each function after applying the named return value optimization on
5334 generic trees. The file name is made by appending @file{.nrv} to the source
5338 @opindex fdump-tree-vect
5339 Dump each function after applying vectorization of loops. The file name is
5340 made by appending @file{.vect} to the source file name.
5343 @opindex fdump-tree-vrp
5344 Dump each function after Value Range Propagation (VRP). The file name
5345 is made by appending @file{.vrp} to the source file name.
5348 @opindex fdump-tree-all
5349 Enable all the available tree dumps with the flags provided in this option.
5352 @item -ftree-vectorizer-verbose=@var{n}
5353 @opindex ftree-vectorizer-verbose
5354 This option controls the amount of debugging output the vectorizer prints.
5355 This information is written to standard error, unless
5356 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5357 in which case it is output to the usual dump listing file, @file{.vect}.
5358 For @var{n}=0 no diagnostic information is reported.
5359 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5360 and the total number of loops that got vectorized.
5361 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5362 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5363 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5364 level that @option{-fdump-tree-vect-stats} uses.
5365 Higher verbosity levels mean either more information dumped for each
5366 reported loop, or same amount of information reported for more loops:
5367 If @var{n}=3, alignment related information is added to the reports.
5368 If @var{n}=4, data-references related information (e.g.@: memory dependences,
5369 memory access-patterns) is added to the reports.
5370 If @var{n}=5, the vectorizer reports also non-vectorized inner-most loops
5371 that did not pass the first analysis phase (i.e., may not be countable, or
5372 may have complicated control-flow).
5373 If @var{n}=6, the vectorizer reports also non-vectorized nested loops.
5374 For @var{n}=7, all the information the vectorizer generates during its
5375 analysis and transformation is reported. This is the same verbosity level
5376 that @option{-fdump-tree-vect-details} uses.
5378 @item -frandom-seed=@var{string}
5379 @opindex frandom-seed
5380 This option provides a seed that GCC uses when it would otherwise use
5381 random numbers. It is used to generate certain symbol names
5382 that have to be different in every compiled file. It is also used to
5383 place unique stamps in coverage data files and the object files that
5384 produce them. You can use the @option{-frandom-seed} option to produce
5385 reproducibly identical object files.
5387 The @var{string} should be different for every file you compile.
5389 @item -fsched-verbose=@var{n}
5390 @opindex fsched-verbose
5391 On targets that use instruction scheduling, this option controls the
5392 amount of debugging output the scheduler prints. This information is
5393 written to standard error, unless @option{-fdump-rtl-sched1} or
5394 @option{-fdump-rtl-sched2} is specified, in which case it is output
5395 to the usual dump listing file, @file{.sched} or @file{.sched2}
5396 respectively. However for @var{n} greater than nine, the output is
5397 always printed to standard error.
5399 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5400 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5401 For @var{n} greater than one, it also output basic block probabilities,
5402 detailed ready list information and unit/insn info. For @var{n} greater
5403 than two, it includes RTL at abort point, control-flow and regions info.
5404 And for @var{n} over four, @option{-fsched-verbose} also includes
5408 @itemx -save-temps=cwd
5410 Store the usual ``temporary'' intermediate files permanently; place them
5411 in the current directory and name them based on the source file. Thus,
5412 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5413 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5414 preprocessed @file{foo.i} output file even though the compiler now
5415 normally uses an integrated preprocessor.
5417 When used in combination with the @option{-x} command line option,
5418 @option{-save-temps} is sensible enough to avoid over writing an
5419 input source file with the same extension as an intermediate file.
5420 The corresponding intermediate file may be obtained by renaming the
5421 source file before using @option{-save-temps}.
5423 If you invoke GCC in parallel, compiling several different source
5424 files that share a common base name in different subdirectories or the
5425 same source file compiled for multiple output destinations, it is
5426 likely that the different parallel compilers will interfere with each
5427 other, and overwrite the temporary files. For instance:
5430 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5431 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5434 may result in @file{foo.i} and @file{foo.o} being written to
5435 simultaneously by both compilers.
5437 @item -save-temps=obj
5438 @opindex save-temps=obj
5439 Store the usual ``temporary'' intermediate files permanently. If the
5440 @option{-o} option is used, the temporary files are based on the
5441 object file. If the @option{-o} option is not used, the
5442 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5447 gcc -save-temps=obj -c foo.c
5448 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5449 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5452 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5453 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5454 @file{dir2/yfoobar.o}.
5456 @item -time@r{[}=@var{file}@r{]}
5458 Report the CPU time taken by each subprocess in the compilation
5459 sequence. For C source files, this is the compiler proper and assembler
5460 (plus the linker if linking is done).
5462 Without the specification of an output file, the output looks like this:
5469 The first number on each line is the ``user time'', that is time spent
5470 executing the program itself. The second number is ``system time'',
5471 time spent executing operating system routines on behalf of the program.
5472 Both numbers are in seconds.
5474 With the specification of an output file, the output is appended to the
5475 named file, and it looks like this:
5478 0.12 0.01 cc1 @var{options}
5479 0.00 0.01 as @var{options}
5482 The ``user time'' and the ``system time'' are moved before the program
5483 name, and the options passed to the program are displayed, so that one
5484 can later tell what file was being compiled, and with which options.
5486 @item -fvar-tracking
5487 @opindex fvar-tracking
5488 Run variable tracking pass. It computes where variables are stored at each
5489 position in code. Better debugging information is then generated
5490 (if the debugging information format supports this information).
5492 It is enabled by default when compiling with optimization (@option{-Os},
5493 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5494 the debug info format supports it.
5496 @item -fvar-tracking-assignments
5497 @opindex fvar-tracking-assignments
5498 @opindex fno-var-tracking-assignments
5499 Annotate assignments to user variables early in the compilation and
5500 attempt to carry the annotations over throughout the compilation all the
5501 way to the end, in an attempt to improve debug information while
5502 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
5504 It can be enabled even if var-tracking is disabled, in which case
5505 annotations will be created and maintained, but discarded at the end.
5507 @item -fvar-tracking-assignments-toggle
5508 @opindex fvar-tracking-assignments-toggle
5509 @opindex fno-var-tracking-assignments-toggle
5510 Toggle @option{-fvar-tracking-assignments}, in the same way that
5511 @option{-gtoggle} toggles @option{-g}.
5513 @item -print-file-name=@var{library}
5514 @opindex print-file-name
5515 Print the full absolute name of the library file @var{library} that
5516 would be used when linking---and don't do anything else. With this
5517 option, GCC does not compile or link anything; it just prints the
5520 @item -print-multi-directory
5521 @opindex print-multi-directory
5522 Print the directory name corresponding to the multilib selected by any
5523 other switches present in the command line. This directory is supposed
5524 to exist in @env{GCC_EXEC_PREFIX}.
5526 @item -print-multi-lib
5527 @opindex print-multi-lib
5528 Print the mapping from multilib directory names to compiler switches
5529 that enable them. The directory name is separated from the switches by
5530 @samp{;}, and each switch starts with an @samp{@@} instead of the
5531 @samp{-}, without spaces between multiple switches. This is supposed to
5532 ease shell-processing.
5534 @item -print-multi-os-directory
5535 @opindex print-multi-os-directory
5536 Print the path to OS libraries for the selected
5537 multilib, relative to some @file{lib} subdirectory. If OS libraries are
5538 present in the @file{lib} subdirectory and no multilibs are used, this is
5539 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
5540 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
5541 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
5542 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
5544 @item -print-prog-name=@var{program}
5545 @opindex print-prog-name
5546 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5548 @item -print-libgcc-file-name
5549 @opindex print-libgcc-file-name
5550 Same as @option{-print-file-name=libgcc.a}.
5552 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5553 but you do want to link with @file{libgcc.a}. You can do
5556 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5559 @item -print-search-dirs
5560 @opindex print-search-dirs
5561 Print the name of the configured installation directory and a list of
5562 program and library directories @command{gcc} will search---and don't do anything else.
5564 This is useful when @command{gcc} prints the error message
5565 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5566 To resolve this you either need to put @file{cpp0} and the other compiler
5567 components where @command{gcc} expects to find them, or you can set the environment
5568 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5569 Don't forget the trailing @samp{/}.
5570 @xref{Environment Variables}.
5572 @item -print-sysroot
5573 @opindex print-sysroot
5574 Print the target sysroot directory that will be used during
5575 compilation. This is the target sysroot specified either at configure
5576 time or using the @option{--sysroot} option, possibly with an extra
5577 suffix that depends on compilation options. If no target sysroot is
5578 specified, the option prints nothing.
5580 @item -print-sysroot-headers-suffix
5581 @opindex print-sysroot-headers-suffix
5582 Print the suffix added to the target sysroot when searching for
5583 headers, or give an error if the compiler is not configured with such
5584 a suffix---and don't do anything else.
5587 @opindex dumpmachine
5588 Print the compiler's target machine (for example,
5589 @samp{i686-pc-linux-gnu})---and don't do anything else.
5592 @opindex dumpversion
5593 Print the compiler version (for example, @samp{3.0})---and don't do
5598 Print the compiler's built-in specs---and don't do anything else. (This
5599 is used when GCC itself is being built.) @xref{Spec Files}.
5601 @item -feliminate-unused-debug-types
5602 @opindex feliminate-unused-debug-types
5603 Normally, when producing DWARF2 output, GCC will emit debugging
5604 information for all types declared in a compilation
5605 unit, regardless of whether or not they are actually used
5606 in that compilation unit. Sometimes this is useful, such as
5607 if, in the debugger, you want to cast a value to a type that is
5608 not actually used in your program (but is declared). More often,
5609 however, this results in a significant amount of wasted space.
5610 With this option, GCC will avoid producing debug symbol output
5611 for types that are nowhere used in the source file being compiled.
5614 @node Optimize Options
5615 @section Options That Control Optimization
5616 @cindex optimize options
5617 @cindex options, optimization
5619 These options control various sorts of optimizations.
5621 Without any optimization option, the compiler's goal is to reduce the
5622 cost of compilation and to make debugging produce the expected
5623 results. Statements are independent: if you stop the program with a
5624 breakpoint between statements, you can then assign a new value to any
5625 variable or change the program counter to any other statement in the
5626 function and get exactly the results you would expect from the source
5629 Turning on optimization flags makes the compiler attempt to improve
5630 the performance and/or code size at the expense of compilation time
5631 and possibly the ability to debug the program.
5633 The compiler performs optimization based on the knowledge it has of the
5634 program. Compiling multiple files at once to a single output file mode allows
5635 the compiler to use information gained from all of the files when compiling
5638 Not all optimizations are controlled directly by a flag. Only
5639 optimizations that have a flag are listed in this section.
5641 Most of the optimizations are not enabled if a @option{-O} level is not set on
5642 the command line, even if individual optimization flags are specified.
5644 Depending on the target and how GCC was configured, a slightly different
5645 set of optimizations may be enabled at each @option{-O} level than
5646 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
5647 to find out the exact set of optimizations that are enabled at each level.
5648 @xref{Overall Options}, for examples.
5655 Optimize. Optimizing compilation takes somewhat more time, and a lot
5656 more memory for a large function.
5658 With @option{-O}, the compiler tries to reduce code size and execution
5659 time, without performing any optimizations that take a great deal of
5662 @option{-O} turns on the following optimization flags:
5665 -fcprop-registers @gol
5668 -fdelayed-branch @gol
5670 -fguess-branch-probability @gol
5671 -fif-conversion2 @gol
5672 -fif-conversion @gol
5673 -fipa-pure-const @gol
5674 -fipa-reference @gol
5676 -fsplit-wide-types @gol
5677 -ftree-builtin-call-dce @gol
5680 -ftree-copyrename @gol
5682 -ftree-dominator-opts @gol
5684 -ftree-forwprop @gol
5692 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5693 where doing so does not interfere with debugging.
5697 Optimize even more. GCC performs nearly all supported optimizations
5698 that do not involve a space-speed tradeoff.
5699 As compared to @option{-O}, this option increases both compilation time
5700 and the performance of the generated code.
5702 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5703 also turns on the following optimization flags:
5704 @gccoptlist{-fthread-jumps @gol
5705 -falign-functions -falign-jumps @gol
5706 -falign-loops -falign-labels @gol
5709 -fcse-follow-jumps -fcse-skip-blocks @gol
5710 -fdelete-null-pointer-checks @gol
5711 -fexpensive-optimizations @gol
5712 -fgcse -fgcse-lm @gol
5713 -finline-small-functions @gol
5714 -findirect-inlining @gol
5716 -foptimize-sibling-calls @gol
5719 -freorder-blocks -freorder-functions @gol
5720 -frerun-cse-after-loop @gol
5721 -fsched-interblock -fsched-spec @gol
5722 -fschedule-insns -fschedule-insns2 @gol
5723 -fstrict-aliasing -fstrict-overflow @gol
5724 -ftree-switch-conversion @gol
5728 Please note the warning under @option{-fgcse} about
5729 invoking @option{-O2} on programs that use computed gotos.
5733 Optimize yet more. @option{-O3} turns on all optimizations specified
5734 by @option{-O2} and also turns on the @option{-finline-functions},
5735 @option{-funswitch-loops}, @option{-fpredictive-commoning},
5736 @option{-fgcse-after-reload} and @option{-ftree-vectorize} options.
5740 Reduce compilation time and make debugging produce the expected
5741 results. This is the default.
5745 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5746 do not typically increase code size. It also performs further
5747 optimizations designed to reduce code size.
5749 @option{-Os} disables the following optimization flags:
5750 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5751 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5752 -fprefetch-loop-arrays -ftree-vect-loop-version}
5754 If you use multiple @option{-O} options, with or without level numbers,
5755 the last such option is the one that is effective.
5758 Options of the form @option{-f@var{flag}} specify machine-independent
5759 flags. Most flags have both positive and negative forms; the negative
5760 form of @option{-ffoo} would be @option{-fno-foo}. In the table
5761 below, only one of the forms is listed---the one you typically will
5762 use. You can figure out the other form by either removing @samp{no-}
5765 The following options control specific optimizations. They are either
5766 activated by @option{-O} options or are related to ones that are. You
5767 can use the following flags in the rare cases when ``fine-tuning'' of
5768 optimizations to be performed is desired.
5771 @item -fno-default-inline
5772 @opindex fno-default-inline
5773 Do not make member functions inline by default merely because they are
5774 defined inside the class scope (C++ only). Otherwise, when you specify
5775 @w{@option{-O}}, member functions defined inside class scope are compiled
5776 inline by default; i.e., you don't need to add @samp{inline} in front of
5777 the member function name.
5779 @item -fno-defer-pop
5780 @opindex fno-defer-pop
5781 Always pop the arguments to each function call as soon as that function
5782 returns. For machines which must pop arguments after a function call,
5783 the compiler normally lets arguments accumulate on the stack for several
5784 function calls and pops them all at once.
5786 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5788 @item -fforward-propagate
5789 @opindex fforward-propagate
5790 Perform a forward propagation pass on RTL@. The pass tries to combine two
5791 instructions and checks if the result can be simplified. If loop unrolling
5792 is active, two passes are performed and the second is scheduled after
5795 This option is enabled by default at optimization levels @option{-O},
5796 @option{-O2}, @option{-O3}, @option{-Os}.
5798 @item -fomit-frame-pointer
5799 @opindex fomit-frame-pointer
5800 Don't keep the frame pointer in a register for functions that
5801 don't need one. This avoids the instructions to save, set up and
5802 restore frame pointers; it also makes an extra register available
5803 in many functions. @strong{It also makes debugging impossible on
5806 On some machines, such as the VAX, this flag has no effect, because
5807 the standard calling sequence automatically handles the frame pointer
5808 and nothing is saved by pretending it doesn't exist. The
5809 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
5810 whether a target machine supports this flag. @xref{Registers,,Register
5811 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
5813 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5815 @item -foptimize-sibling-calls
5816 @opindex foptimize-sibling-calls
5817 Optimize sibling and tail recursive calls.
5819 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5823 Don't pay attention to the @code{inline} keyword. Normally this option
5824 is used to keep the compiler from expanding any functions inline.
5825 Note that if you are not optimizing, no functions can be expanded inline.
5827 @item -finline-small-functions
5828 @opindex finline-small-functions
5829 Integrate functions into their callers when their body is smaller than expected
5830 function call code (so overall size of program gets smaller). The compiler
5831 heuristically decides which functions are simple enough to be worth integrating
5834 Enabled at level @option{-O2}.
5836 @item -findirect-inlining
5837 @opindex findirect-inlining
5838 Inline also indirect calls that are discovered to be known at compile
5839 time thanks to previous inlining. This option has any effect only
5840 when inlining itself is turned on by the @option{-finline-functions}
5841 or @option{-finline-small-functions} options.
5843 Enabled at level @option{-O2}.
5845 @item -finline-functions
5846 @opindex finline-functions
5847 Integrate all simple functions into their callers. The compiler
5848 heuristically decides which functions are simple enough to be worth
5849 integrating in this way.
5851 If all calls to a given function are integrated, and the function is
5852 declared @code{static}, then the function is normally not output as
5853 assembler code in its own right.
5855 Enabled at level @option{-O3}.
5857 @item -finline-functions-called-once
5858 @opindex finline-functions-called-once
5859 Consider all @code{static} functions called once for inlining into their
5860 caller even if they are not marked @code{inline}. If a call to a given
5861 function is integrated, then the function is not output as assembler code
5864 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
5866 @item -fearly-inlining
5867 @opindex fearly-inlining
5868 Inline functions marked by @code{always_inline} and functions whose body seems
5869 smaller than the function call overhead early before doing
5870 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
5871 makes profiling significantly cheaper and usually inlining faster on programs
5872 having large chains of nested wrapper functions.
5878 Perform interprocedural scalar replacement of aggregates, removal of
5879 unused parameters and replacement of parameters passed by reference
5880 by parameters passed by value.
5882 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
5884 @item -finline-limit=@var{n}
5885 @opindex finline-limit
5886 By default, GCC limits the size of functions that can be inlined. This flag
5887 allows coarse control of this limit. @var{n} is the size of functions that
5888 can be inlined in number of pseudo instructions.
5890 Inlining is actually controlled by a number of parameters, which may be
5891 specified individually by using @option{--param @var{name}=@var{value}}.
5892 The @option{-finline-limit=@var{n}} option sets some of these parameters
5896 @item max-inline-insns-single
5897 is set to @var{n}/2.
5898 @item max-inline-insns-auto
5899 is set to @var{n}/2.
5902 See below for a documentation of the individual
5903 parameters controlling inlining and for the defaults of these parameters.
5905 @emph{Note:} there may be no value to @option{-finline-limit} that results
5906 in default behavior.
5908 @emph{Note:} pseudo instruction represents, in this particular context, an
5909 abstract measurement of function's size. In no way does it represent a count
5910 of assembly instructions and as such its exact meaning might change from one
5911 release to an another.
5913 @item -fkeep-inline-functions
5914 @opindex fkeep-inline-functions
5915 In C, emit @code{static} functions that are declared @code{inline}
5916 into the object file, even if the function has been inlined into all
5917 of its callers. This switch does not affect functions using the
5918 @code{extern inline} extension in GNU C89@. In C++, emit any and all
5919 inline functions into the object file.
5921 @item -fkeep-static-consts
5922 @opindex fkeep-static-consts
5923 Emit variables declared @code{static const} when optimization isn't turned
5924 on, even if the variables aren't referenced.
5926 GCC enables this option by default. If you want to force the compiler to
5927 check if the variable was referenced, regardless of whether or not
5928 optimization is turned on, use the @option{-fno-keep-static-consts} option.
5930 @item -fmerge-constants
5931 @opindex fmerge-constants
5932 Attempt to merge identical constants (string constants and floating point
5933 constants) across compilation units.
5935 This option is the default for optimized compilation if the assembler and
5936 linker support it. Use @option{-fno-merge-constants} to inhibit this
5939 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5941 @item -fmerge-all-constants
5942 @opindex fmerge-all-constants
5943 Attempt to merge identical constants and identical variables.
5945 This option implies @option{-fmerge-constants}. In addition to
5946 @option{-fmerge-constants} this considers e.g.@: even constant initialized
5947 arrays or initialized constant variables with integral or floating point
5948 types. Languages like C or C++ require each variable, including multiple
5949 instances of the same variable in recursive calls, to have distinct locations,
5950 so using this option will result in non-conforming
5953 @item -fmodulo-sched
5954 @opindex fmodulo-sched
5955 Perform swing modulo scheduling immediately before the first scheduling
5956 pass. This pass looks at innermost loops and reorders their
5957 instructions by overlapping different iterations.
5959 @item -fmodulo-sched-allow-regmoves
5960 @opindex fmodulo-sched-allow-regmoves
5961 Perform more aggressive SMS based modulo scheduling with register moves
5962 allowed. By setting this flag certain anti-dependences edges will be
5963 deleted which will trigger the generation of reg-moves based on the
5964 life-range analysis. This option is effective only with
5965 @option{-fmodulo-sched} enabled.
5967 @item -fno-branch-count-reg
5968 @opindex fno-branch-count-reg
5969 Do not use ``decrement and branch'' instructions on a count register,
5970 but instead generate a sequence of instructions that decrement a
5971 register, compare it against zero, then branch based upon the result.
5972 This option is only meaningful on architectures that support such
5973 instructions, which include x86, PowerPC, IA-64 and S/390.
5975 The default is @option{-fbranch-count-reg}.
5977 @item -fno-function-cse
5978 @opindex fno-function-cse
5979 Do not put function addresses in registers; make each instruction that
5980 calls a constant function contain the function's address explicitly.
5982 This option results in less efficient code, but some strange hacks
5983 that alter the assembler output may be confused by the optimizations
5984 performed when this option is not used.
5986 The default is @option{-ffunction-cse}
5988 @item -fno-zero-initialized-in-bss
5989 @opindex fno-zero-initialized-in-bss
5990 If the target supports a BSS section, GCC by default puts variables that
5991 are initialized to zero into BSS@. This can save space in the resulting
5994 This option turns off this behavior because some programs explicitly
5995 rely on variables going to the data section. E.g., so that the
5996 resulting executable can find the beginning of that section and/or make
5997 assumptions based on that.
5999 The default is @option{-fzero-initialized-in-bss}.
6001 @item -fmudflap -fmudflapth -fmudflapir
6005 @cindex bounds checking
6007 For front-ends that support it (C and C++), instrument all risky
6008 pointer/array dereferencing operations, some standard library
6009 string/heap functions, and some other associated constructs with
6010 range/validity tests. Modules so instrumented should be immune to
6011 buffer overflows, invalid heap use, and some other classes of C/C++
6012 programming errors. The instrumentation relies on a separate runtime
6013 library (@file{libmudflap}), which will be linked into a program if
6014 @option{-fmudflap} is given at link time. Run-time behavior of the
6015 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
6016 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
6019 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
6020 link if your program is multi-threaded. Use @option{-fmudflapir}, in
6021 addition to @option{-fmudflap} or @option{-fmudflapth}, if
6022 instrumentation should ignore pointer reads. This produces less
6023 instrumentation (and therefore faster execution) and still provides
6024 some protection against outright memory corrupting writes, but allows
6025 erroneously read data to propagate within a program.
6027 @item -fthread-jumps
6028 @opindex fthread-jumps
6029 Perform optimizations where we check to see if a jump branches to a
6030 location where another comparison subsumed by the first is found. If
6031 so, the first branch is redirected to either the destination of the
6032 second branch or a point immediately following it, depending on whether
6033 the condition is known to be true or false.
6035 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6037 @item -fsplit-wide-types
6038 @opindex fsplit-wide-types
6039 When using a type that occupies multiple registers, such as @code{long
6040 long} on a 32-bit system, split the registers apart and allocate them
6041 independently. This normally generates better code for those types,
6042 but may make debugging more difficult.
6044 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
6047 @item -fcse-follow-jumps
6048 @opindex fcse-follow-jumps
6049 In common subexpression elimination (CSE), scan through jump instructions
6050 when the target of the jump is not reached by any other path. For
6051 example, when CSE encounters an @code{if} statement with an
6052 @code{else} clause, CSE will follow the jump when the condition
6055 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6057 @item -fcse-skip-blocks
6058 @opindex fcse-skip-blocks
6059 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
6060 follow jumps which conditionally skip over blocks. When CSE
6061 encounters a simple @code{if} statement with no else clause,
6062 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6063 body of the @code{if}.
6065 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6067 @item -frerun-cse-after-loop
6068 @opindex frerun-cse-after-loop
6069 Re-run common subexpression elimination after loop optimizations has been
6072 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6076 Perform a global common subexpression elimination pass.
6077 This pass also performs global constant and copy propagation.
6079 @emph{Note:} When compiling a program using computed gotos, a GCC
6080 extension, you may get better runtime performance if you disable
6081 the global common subexpression elimination pass by adding
6082 @option{-fno-gcse} to the command line.
6084 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6088 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
6089 attempt to move loads which are only killed by stores into themselves. This
6090 allows a loop containing a load/store sequence to be changed to a load outside
6091 the loop, and a copy/store within the loop.
6093 Enabled by default when gcse is enabled.
6097 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6098 global common subexpression elimination. This pass will attempt to move
6099 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6100 loops containing a load/store sequence can be changed to a load before
6101 the loop and a store after the loop.
6103 Not enabled at any optimization level.
6107 When @option{-fgcse-las} is enabled, the global common subexpression
6108 elimination pass eliminates redundant loads that come after stores to the
6109 same memory location (both partial and full redundancies).
6111 Not enabled at any optimization level.
6113 @item -fgcse-after-reload
6114 @opindex fgcse-after-reload
6115 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6116 pass is performed after reload. The purpose of this pass is to cleanup
6119 @item -funsafe-loop-optimizations
6120 @opindex funsafe-loop-optimizations
6121 If given, the loop optimizer will assume that loop indices do not
6122 overflow, and that the loops with nontrivial exit condition are not
6123 infinite. This enables a wider range of loop optimizations even if
6124 the loop optimizer itself cannot prove that these assumptions are valid.
6125 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
6126 if it finds this kind of loop.
6128 @item -fcrossjumping
6129 @opindex fcrossjumping
6130 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6131 resulting code may or may not perform better than without cross-jumping.
6133 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6135 @item -fauto-inc-dec
6136 @opindex fauto-inc-dec
6137 Combine increments or decrements of addresses with memory accesses.
6138 This pass is always skipped on architectures that do not have
6139 instructions to support this. Enabled by default at @option{-O} and
6140 higher on architectures that support this.
6144 Perform dead code elimination (DCE) on RTL@.
6145 Enabled by default at @option{-O} and higher.
6149 Perform dead store elimination (DSE) on RTL@.
6150 Enabled by default at @option{-O} and higher.
6152 @item -fif-conversion
6153 @opindex fif-conversion
6154 Attempt to transform conditional jumps into branch-less equivalents. This
6155 include use of conditional moves, min, max, set flags and abs instructions, and
6156 some tricks doable by standard arithmetics. The use of conditional execution
6157 on chips where it is available is controlled by @code{if-conversion2}.
6159 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6161 @item -fif-conversion2
6162 @opindex fif-conversion2
6163 Use conditional execution (where available) to transform conditional jumps into
6164 branch-less equivalents.
6166 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6168 @item -fdelete-null-pointer-checks
6169 @opindex fdelete-null-pointer-checks
6170 Assume that programs cannot safely dereference null pointers, and that
6171 no code or data element resides there. This enables simple constant
6172 folding optimizations at all optimization levels. In addition, other
6173 optimization passes in GCC use this flag to control global dataflow
6174 analyses that eliminate useless checks for null pointers; these assume
6175 that if a pointer is checked after it has already been dereferenced,
6178 Note however that in some environments this assumption is not true.
6179 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6180 for programs which depend on that behavior.
6182 Some targets, especially embedded ones, disable this option at all levels.
6183 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6184 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6185 are enabled independently at different optimization levels.
6187 @item -fexpensive-optimizations
6188 @opindex fexpensive-optimizations
6189 Perform a number of minor optimizations that are relatively expensive.
6191 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6193 @item -foptimize-register-move
6195 @opindex foptimize-register-move
6197 Attempt to reassign register numbers in move instructions and as
6198 operands of other simple instructions in order to maximize the amount of
6199 register tying. This is especially helpful on machines with two-operand
6202 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6205 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6207 @item -fira-algorithm=@var{algorithm}
6208 Use specified coloring algorithm for the integrated register
6209 allocator. The @var{algorithm} argument should be @code{priority} or
6210 @code{CB}. The first algorithm specifies Chow's priority coloring,
6211 the second one specifies Chaitin-Briggs coloring. The second
6212 algorithm can be unimplemented for some architectures. If it is
6213 implemented, it is the default because Chaitin-Briggs coloring as a
6214 rule generates a better code.
6216 @item -fira-region=@var{region}
6217 Use specified regions for the integrated register allocator. The
6218 @var{region} argument should be one of @code{all}, @code{mixed}, or
6219 @code{one}. The first value means using all loops as register
6220 allocation regions, the second value which is the default means using
6221 all loops except for loops with small register pressure as the
6222 regions, and third one means using all function as a single region.
6223 The first value can give best result for machines with small size and
6224 irregular register set, the third one results in faster and generates
6225 decent code and the smallest size code, and the default value usually
6226 give the best results in most cases and for most architectures.
6228 @item -fira-coalesce
6229 @opindex fira-coalesce
6230 Do optimistic register coalescing. This option might be profitable for
6231 architectures with big regular register files.
6233 @item -fira-loop-pressure
6234 @opindex fira-loop-pressure
6235 Use IRA to evaluate register pressure in loops for decision to move
6236 loop invariants. Usage of this option usually results in generation
6237 of faster and smaller code on machines with big register files (>= 32
6238 registers) but it can slow compiler down.
6240 This option is enabled at level @option{-O3} for some targets.
6242 @item -fno-ira-share-save-slots
6243 @opindex fno-ira-share-save-slots
6244 Switch off sharing stack slots used for saving call used hard
6245 registers living through a call. Each hard register will get a
6246 separate stack slot and as a result function stack frame will be
6249 @item -fno-ira-share-spill-slots
6250 @opindex fno-ira-share-spill-slots
6251 Switch off sharing stack slots allocated for pseudo-registers. Each
6252 pseudo-register which did not get a hard register will get a separate
6253 stack slot and as a result function stack frame will be bigger.
6255 @item -fira-verbose=@var{n}
6256 @opindex fira-verbose
6257 Set up how verbose dump file for the integrated register allocator
6258 will be. Default value is 5. If the value is greater or equal to 10,
6259 the dump file will be stderr as if the value were @var{n} minus 10.
6261 @item -fdelayed-branch
6262 @opindex fdelayed-branch
6263 If supported for the target machine, attempt to reorder instructions
6264 to exploit instruction slots available after delayed branch
6267 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6269 @item -fschedule-insns
6270 @opindex fschedule-insns
6271 If supported for the target machine, attempt to reorder instructions to
6272 eliminate execution stalls due to required data being unavailable. This
6273 helps machines that have slow floating point or memory load instructions
6274 by allowing other instructions to be issued until the result of the load
6275 or floating point instruction is required.
6277 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6279 @item -fschedule-insns2
6280 @opindex fschedule-insns2
6281 Similar to @option{-fschedule-insns}, but requests an additional pass of
6282 instruction scheduling after register allocation has been done. This is
6283 especially useful on machines with a relatively small number of
6284 registers and where memory load instructions take more than one cycle.
6286 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6288 @item -fno-sched-interblock
6289 @opindex fno-sched-interblock
6290 Don't schedule instructions across basic blocks. This is normally
6291 enabled by default when scheduling before register allocation, i.e.@:
6292 with @option{-fschedule-insns} or at @option{-O2} or higher.
6294 @item -fno-sched-spec
6295 @opindex fno-sched-spec
6296 Don't allow speculative motion of non-load instructions. This is normally
6297 enabled by default when scheduling before register allocation, i.e.@:
6298 with @option{-fschedule-insns} or at @option{-O2} or higher.
6300 @item -fsched-pressure
6301 @opindex fsched-pressure
6302 Enable register pressure sensitive insn scheduling before the register
6303 allocation. This only makes sense when scheduling before register
6304 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
6305 @option{-O2} or higher. Usage of this option can improve the
6306 generated code and decrease its size by preventing register pressure
6307 increase above the number of available hard registers and as a
6308 consequence register spills in the register allocation.
6310 @item -fsched-spec-load
6311 @opindex fsched-spec-load
6312 Allow speculative motion of some load instructions. This only makes
6313 sense when scheduling before register allocation, i.e.@: with
6314 @option{-fschedule-insns} or at @option{-O2} or higher.
6316 @item -fsched-spec-load-dangerous
6317 @opindex fsched-spec-load-dangerous
6318 Allow speculative motion of more load instructions. This only makes
6319 sense when scheduling before register allocation, i.e.@: with
6320 @option{-fschedule-insns} or at @option{-O2} or higher.
6322 @item -fsched-stalled-insns
6323 @itemx -fsched-stalled-insns=@var{n}
6324 @opindex fsched-stalled-insns
6325 Define how many insns (if any) can be moved prematurely from the queue
6326 of stalled insns into the ready list, during the second scheduling pass.
6327 @option{-fno-sched-stalled-insns} means that no insns will be moved
6328 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6329 on how many queued insns can be moved prematurely.
6330 @option{-fsched-stalled-insns} without a value is equivalent to
6331 @option{-fsched-stalled-insns=1}.
6333 @item -fsched-stalled-insns-dep
6334 @itemx -fsched-stalled-insns-dep=@var{n}
6335 @opindex fsched-stalled-insns-dep
6336 Define how many insn groups (cycles) will be examined for a dependency
6337 on a stalled insn that is candidate for premature removal from the queue
6338 of stalled insns. This has an effect only during the second scheduling pass,
6339 and only if @option{-fsched-stalled-insns} is used.
6340 @option{-fno-sched-stalled-insns-dep} is equivalent to
6341 @option{-fsched-stalled-insns-dep=0}.
6342 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6343 @option{-fsched-stalled-insns-dep=1}.
6345 @item -fsched2-use-superblocks
6346 @opindex fsched2-use-superblocks
6347 When scheduling after register allocation, do use superblock scheduling
6348 algorithm. Superblock scheduling allows motion across basic block boundaries
6349 resulting on faster schedules. This option is experimental, as not all machine
6350 descriptions used by GCC model the CPU closely enough to avoid unreliable
6351 results from the algorithm.
6353 This only makes sense when scheduling after register allocation, i.e.@: with
6354 @option{-fschedule-insns2} or at @option{-O2} or higher.
6356 @item -fsched-group-heuristic
6357 @opindex fsched-group-heuristic
6358 Enable the group heuristic in the scheduler. This heuristic favors
6359 the instruction that belongs to a schedule group. This is enabled
6360 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6361 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6363 @item -fsched-critical-path-heuristic
6364 @opindex fsched-critical-path-heuristic
6365 Enable the critical-path heuristic in the scheduler. This heuristic favors
6366 instructions on the critical path. This is enabled by default when
6367 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6368 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6370 @item -fsched-spec-insn-heuristic
6371 @opindex fsched-spec-insn-heuristic
6372 Enable the speculative instruction heuristic in the scheduler. This
6373 heuristic favors speculative instructions with greater dependency weakness.
6374 This is enabled by default when scheduling is enabled, i.e.@:
6375 with @option{-fschedule-insns} or @option{-fschedule-insns2}
6376 or at @option{-O2} or higher.
6378 @item -fsched-rank-heuristic
6379 @opindex fsched-rank-heuristic
6380 Enable the rank heuristic in the scheduler. This heuristic favors
6381 the instruction belonging to a basic block with greater size or frequency.
6382 This is enabled by default when scheduling is enabled, i.e.@:
6383 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6384 at @option{-O2} or higher.
6386 @item -fsched-last-insn-heuristic
6387 @opindex fsched-last-insn-heuristic
6388 Enable the last-instruction heuristic in the scheduler. This heuristic
6389 favors the instruction that is less dependent on the last instruction
6390 scheduled. This is enabled by default when scheduling is enabled,
6391 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6392 at @option{-O2} or higher.
6394 @item -fsched-dep-count-heuristic
6395 @opindex fsched-dep-count-heuristic
6396 Enable the dependent-count heuristic in the scheduler. This heuristic
6397 favors the instruction that has more instructions depending on it.
6398 This is enabled by default when scheduling is enabled, i.e.@:
6399 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6400 at @option{-O2} or higher.
6402 @item -fsched2-use-traces
6403 @opindex fsched2-use-traces
6404 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
6405 allocation and additionally perform code duplication in order to increase the
6406 size of superblocks using tracer pass. See @option{-ftracer} for details on
6409 This mode should produce faster but significantly longer programs. Also
6410 without @option{-fbranch-probabilities} the traces constructed may not
6411 match the reality and hurt the performance. This only makes
6412 sense when scheduling after register allocation, i.e.@: with
6413 @option{-fschedule-insns2} or at @option{-O2} or higher.
6415 @item -freschedule-modulo-scheduled-loops
6416 @opindex freschedule-modulo-scheduled-loops
6417 The modulo scheduling comes before the traditional scheduling, if a loop
6418 was modulo scheduled we may want to prevent the later scheduling passes
6419 from changing its schedule, we use this option to control that.
6421 @item -fselective-scheduling
6422 @opindex fselective-scheduling
6423 Schedule instructions using selective scheduling algorithm. Selective
6424 scheduling runs instead of the first scheduler pass.
6426 @item -fselective-scheduling2
6427 @opindex fselective-scheduling2
6428 Schedule instructions using selective scheduling algorithm. Selective
6429 scheduling runs instead of the second scheduler pass.
6431 @item -fsel-sched-pipelining
6432 @opindex fsel-sched-pipelining
6433 Enable software pipelining of innermost loops during selective scheduling.
6434 This option has no effect until one of @option{-fselective-scheduling} or
6435 @option{-fselective-scheduling2} is turned on.
6437 @item -fsel-sched-pipelining-outer-loops
6438 @opindex fsel-sched-pipelining-outer-loops
6439 When pipelining loops during selective scheduling, also pipeline outer loops.
6440 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6442 @item -fcaller-saves
6443 @opindex fcaller-saves
6444 Enable values to be allocated in registers that will be clobbered by
6445 function calls, by emitting extra instructions to save and restore the
6446 registers around such calls. Such allocation is done only when it
6447 seems to result in better code than would otherwise be produced.
6449 This option is always enabled by default on certain machines, usually
6450 those which have no call-preserved registers to use instead.
6452 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6454 @item -fconserve-stack
6455 @opindex fconserve-stack
6456 Attempt to minimize stack usage. The compiler will attempt to use less
6457 stack space, even if that makes the program slower. This option
6458 implies setting the @option{large-stack-frame} parameter to 100
6459 and the @option{large-stack-frame-growth} parameter to 400.
6461 @item -ftree-reassoc
6462 @opindex ftree-reassoc
6463 Perform reassociation on trees. This flag is enabled by default
6464 at @option{-O} and higher.
6468 Perform partial redundancy elimination (PRE) on trees. This flag is
6469 enabled by default at @option{-O2} and @option{-O3}.
6471 @item -ftree-forwprop
6472 @opindex ftree-forwprop
6473 Perform forward propagation on trees. This flag is enabled by default
6474 at @option{-O} and higher.
6478 Perform full redundancy elimination (FRE) on trees. The difference
6479 between FRE and PRE is that FRE only considers expressions
6480 that are computed on all paths leading to the redundant computation.
6481 This analysis is faster than PRE, though it exposes fewer redundancies.
6482 This flag is enabled by default at @option{-O} and higher.
6484 @item -ftree-phiprop
6485 @opindex ftree-phiprop
6486 Perform hoisting of loads from conditional pointers on trees. This
6487 pass is enabled by default at @option{-O} and higher.
6489 @item -ftree-copy-prop
6490 @opindex ftree-copy-prop
6491 Perform copy propagation on trees. This pass eliminates unnecessary
6492 copy operations. This flag is enabled by default at @option{-O} and
6495 @item -fipa-pure-const
6496 @opindex fipa-pure-const
6497 Discover which functions are pure or constant.
6498 Enabled by default at @option{-O} and higher.
6500 @item -fipa-reference
6501 @opindex fipa-reference
6502 Discover which static variables do not escape cannot escape the
6504 Enabled by default at @option{-O} and higher.
6506 @item -fipa-struct-reorg
6507 @opindex fipa-struct-reorg
6508 Perform structure reorganization optimization, that change C-like structures
6509 layout in order to better utilize spatial locality. This transformation is
6510 affective for programs containing arrays of structures. Available in two
6511 compilation modes: profile-based (enabled with @option{-fprofile-generate})
6512 or static (which uses built-in heuristics). Require @option{-fipa-type-escape}
6513 to provide the safety of this transformation. It works only in whole program
6514 mode, so it requires @option{-fwhole-program} and @option{-combine} to be
6515 enabled. Structures considered @samp{cold} by this transformation are not
6516 affected (see @option{--param struct-reorg-cold-struct-ratio=@var{value}}).
6518 With this flag, the program debug info reflects a new structure layout.
6522 Perform interprocedural pointer analysis. This option is experimental
6523 and does not affect generated code.
6527 Perform interprocedural constant propagation.
6528 This optimization analyzes the program to determine when values passed
6529 to functions are constants and then optimizes accordingly.
6530 This optimization can substantially increase performance
6531 if the application has constants passed to functions.
6532 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
6534 @item -fipa-cp-clone
6535 @opindex fipa-cp-clone
6536 Perform function cloning to make interprocedural constant propagation stronger.
6537 When enabled, interprocedural constant propagation will perform function cloning
6538 when externally visible function can be called with constant arguments.
6539 Because this optimization can create multiple copies of functions,
6540 it may significantly increase code size
6541 (see @option{--param ipcp-unit-growth=@var{value}}).
6542 This flag is enabled by default at @option{-O3}.
6544 @item -fipa-matrix-reorg
6545 @opindex fipa-matrix-reorg
6546 Perform matrix flattening and transposing.
6547 Matrix flattening tries to replace an @math{m}-dimensional matrix
6548 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
6549 This reduces the level of indirection needed for accessing the elements
6550 of the matrix. The second optimization is matrix transposing that
6551 attempts to change the order of the matrix's dimensions in order to
6552 improve cache locality.
6553 Both optimizations need the @option{-fwhole-program} flag.
6554 Transposing is enabled only if profiling information is available.
6558 Perform forward store motion on trees. This flag is
6559 enabled by default at @option{-O} and higher.
6563 Perform sparse conditional constant propagation (CCP) on trees. This
6564 pass only operates on local scalar variables and is enabled by default
6565 at @option{-O} and higher.
6567 @item -ftree-switch-conversion
6568 Perform conversion of simple initializations in a switch to
6569 initializations from a scalar array. This flag is enabled by default
6570 at @option{-O2} and higher.
6574 Perform dead code elimination (DCE) on trees. This flag is enabled by
6575 default at @option{-O} and higher.
6577 @item -ftree-builtin-call-dce
6578 @opindex ftree-builtin-call-dce
6579 Perform conditional dead code elimination (DCE) for calls to builtin functions
6580 that may set @code{errno} but are otherwise side-effect free. This flag is
6581 enabled by default at @option{-O2} and higher if @option{-Os} is not also
6584 @item -ftree-dominator-opts
6585 @opindex ftree-dominator-opts
6586 Perform a variety of simple scalar cleanups (constant/copy
6587 propagation, redundancy elimination, range propagation and expression
6588 simplification) based on a dominator tree traversal. This also
6589 performs jump threading (to reduce jumps to jumps). This flag is
6590 enabled by default at @option{-O} and higher.
6594 Perform dead store elimination (DSE) on trees. A dead store is a store into
6595 a memory location which will later be overwritten by another store without
6596 any intervening loads. In this case the earlier store can be deleted. This
6597 flag is enabled by default at @option{-O} and higher.
6601 Perform loop header copying on trees. This is beneficial since it increases
6602 effectiveness of code motion optimizations. It also saves one jump. This flag
6603 is enabled by default at @option{-O} and higher. It is not enabled
6604 for @option{-Os}, since it usually increases code size.
6606 @item -ftree-loop-optimize
6607 @opindex ftree-loop-optimize
6608 Perform loop optimizations on trees. This flag is enabled by default
6609 at @option{-O} and higher.
6611 @item -ftree-loop-linear
6612 @opindex ftree-loop-linear
6613 Perform linear loop transformations on tree. This flag can improve cache
6614 performance and allow further loop optimizations to take place.
6616 @item -floop-interchange
6617 Perform loop interchange transformations on loops. Interchanging two
6618 nested loops switches the inner and outer loops. For example, given a
6623 A(J, I) = A(J, I) * C
6627 loop interchange will transform the loop as if the user had written:
6631 A(J, I) = A(J, I) * C
6635 which can be beneficial when @code{N} is larger than the caches,
6636 because in Fortran, the elements of an array are stored in memory
6637 contiguously by column, and the original loop iterates over rows,
6638 potentially creating at each access a cache miss. This optimization
6639 applies to all the languages supported by GCC and is not limited to
6640 Fortran. To use this code transformation, GCC has to be configured
6641 with @option{--with-ppl} and @option{--with-cloog} to enable the
6642 Graphite loop transformation infrastructure.
6644 @item -floop-strip-mine
6645 Perform loop strip mining transformations on loops. Strip mining
6646 splits a loop into two nested loops. The outer loop has strides
6647 equal to the strip size and the inner loop has strides of the
6648 original loop within a strip. For example, given a loop like:
6654 loop strip mining will transform the loop as if the user had written:
6657 DO I = II, min (II + 3, N)
6662 This optimization applies to all the languages supported by GCC and is
6663 not limited to Fortran. To use this code transformation, GCC has to
6664 be configured with @option{--with-ppl} and @option{--with-cloog} to
6665 enable the Graphite loop transformation infrastructure.
6668 Perform loop blocking transformations on loops. Blocking strip mines
6669 each loop in the loop nest such that the memory accesses of the
6670 element loops fit inside caches. For example, given a loop like:
6674 A(J, I) = B(I) + C(J)
6678 loop blocking will transform the loop as if the user had written:
6682 DO I = II, min (II + 63, N)
6683 DO J = JJ, min (JJ + 63, M)
6684 A(J, I) = B(I) + C(J)
6690 which can be beneficial when @code{M} is larger than the caches,
6691 because the innermost loop will iterate over a smaller amount of data
6692 that can be kept in the caches. This optimization applies to all the
6693 languages supported by GCC and is not limited to Fortran. To use this
6694 code transformation, GCC has to be configured with @option{--with-ppl}
6695 and @option{--with-cloog} to enable the Graphite loop transformation
6698 @item -fgraphite-identity
6699 @opindex fgraphite-identity
6700 Enable the identity transformation for graphite. For every SCoP we generate
6701 the polyhedral representation and transform it back to gimple. Using
6702 @option{-fgraphite-identity} we can check the costs or benefits of the
6703 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
6704 are also performed by the code generator CLooG, like index splitting and
6705 dead code elimination in loops.
6707 @item -floop-parallelize-all
6708 Use the Graphite data dependence analysis to identify loops that can
6709 be parallelized. Parallelize all the loops that can be analyzed to
6710 not contain loop carried dependences without checking that it is
6711 profitable to parallelize the loops.
6713 @item -fcheck-data-deps
6714 @opindex fcheck-data-deps
6715 Compare the results of several data dependence analyzers. This option
6716 is used for debugging the data dependence analyzers.
6718 @item -ftree-loop-distribution
6719 Perform loop distribution. This flag can improve cache performance on
6720 big loop bodies and allow further loop optimizations, like
6721 parallelization or vectorization, to take place. For example, the loop
6738 @item -ftree-loop-im
6739 @opindex ftree-loop-im
6740 Perform loop invariant motion on trees. This pass moves only invariants that
6741 would be hard to handle at RTL level (function calls, operations that expand to
6742 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
6743 operands of conditions that are invariant out of the loop, so that we can use
6744 just trivial invariantness analysis in loop unswitching. The pass also includes
6747 @item -ftree-loop-ivcanon
6748 @opindex ftree-loop-ivcanon
6749 Create a canonical counter for number of iterations in the loop for that
6750 determining number of iterations requires complicated analysis. Later
6751 optimizations then may determine the number easily. Useful especially
6752 in connection with unrolling.
6756 Perform induction variable optimizations (strength reduction, induction
6757 variable merging and induction variable elimination) on trees.
6759 @item -ftree-parallelize-loops=n
6760 @opindex ftree-parallelize-loops
6761 Parallelize loops, i.e., split their iteration space to run in n threads.
6762 This is only possible for loops whose iterations are independent
6763 and can be arbitrarily reordered. The optimization is only
6764 profitable on multiprocessor machines, for loops that are CPU-intensive,
6765 rather than constrained e.g.@: by memory bandwidth. This option
6766 implies @option{-pthread}, and thus is only supported on targets
6767 that have support for @option{-pthread}.
6771 Perform function-local points-to analysis on trees. This flag is
6772 enabled by default at @option{-O} and higher.
6776 Perform scalar replacement of aggregates. This pass replaces structure
6777 references with scalars to prevent committing structures to memory too
6778 early. This flag is enabled by default at @option{-O} and higher.
6780 @item -ftree-copyrename
6781 @opindex ftree-copyrename
6782 Perform copy renaming on trees. This pass attempts to rename compiler
6783 temporaries to other variables at copy locations, usually resulting in
6784 variable names which more closely resemble the original variables. This flag
6785 is enabled by default at @option{-O} and higher.
6789 Perform temporary expression replacement during the SSA->normal phase. Single
6790 use/single def temporaries are replaced at their use location with their
6791 defining expression. This results in non-GIMPLE code, but gives the expanders
6792 much more complex trees to work on resulting in better RTL generation. This is
6793 enabled by default at @option{-O} and higher.
6795 @item -ftree-vectorize
6796 @opindex ftree-vectorize
6797 Perform loop vectorization on trees. This flag is enabled by default at
6800 @item -ftree-vect-loop-version
6801 @opindex ftree-vect-loop-version
6802 Perform loop versioning when doing loop vectorization on trees. When a loop
6803 appears to be vectorizable except that data alignment or data dependence cannot
6804 be determined at compile time then vectorized and non-vectorized versions of
6805 the loop are generated along with runtime checks for alignment or dependence
6806 to control which version is executed. This option is enabled by default
6807 except at level @option{-Os} where it is disabled.
6809 @item -fvect-cost-model
6810 @opindex fvect-cost-model
6811 Enable cost model for vectorization.
6815 Perform Value Range Propagation on trees. This is similar to the
6816 constant propagation pass, but instead of values, ranges of values are
6817 propagated. This allows the optimizers to remove unnecessary range
6818 checks like array bound checks and null pointer checks. This is
6819 enabled by default at @option{-O2} and higher. Null pointer check
6820 elimination is only done if @option{-fdelete-null-pointer-checks} is
6825 Perform tail duplication to enlarge superblock size. This transformation
6826 simplifies the control flow of the function allowing other optimizations to do
6829 @item -funroll-loops
6830 @opindex funroll-loops
6831 Unroll loops whose number of iterations can be determined at compile
6832 time or upon entry to the loop. @option{-funroll-loops} implies
6833 @option{-frerun-cse-after-loop}. This option makes code larger,
6834 and may or may not make it run faster.
6836 @item -funroll-all-loops
6837 @opindex funroll-all-loops
6838 Unroll all loops, even if their number of iterations is uncertain when
6839 the loop is entered. This usually makes programs run more slowly.
6840 @option{-funroll-all-loops} implies the same options as
6841 @option{-funroll-loops},
6843 @item -fsplit-ivs-in-unroller
6844 @opindex fsplit-ivs-in-unroller
6845 Enables expressing of values of induction variables in later iterations
6846 of the unrolled loop using the value in the first iteration. This breaks
6847 long dependency chains, thus improving efficiency of the scheduling passes.
6849 Combination of @option{-fweb} and CSE is often sufficient to obtain the
6850 same effect. However in cases the loop body is more complicated than
6851 a single basic block, this is not reliable. It also does not work at all
6852 on some of the architectures due to restrictions in the CSE pass.
6854 This optimization is enabled by default.
6856 @item -fvariable-expansion-in-unroller
6857 @opindex fvariable-expansion-in-unroller
6858 With this option, the compiler will create multiple copies of some
6859 local variables when unrolling a loop which can result in superior code.
6861 @item -fpredictive-commoning
6862 @opindex fpredictive-commoning
6863 Perform predictive commoning optimization, i.e., reusing computations
6864 (especially memory loads and stores) performed in previous
6865 iterations of loops.
6867 This option is enabled at level @option{-O3}.
6869 @item -fprefetch-loop-arrays
6870 @opindex fprefetch-loop-arrays
6871 If supported by the target machine, generate instructions to prefetch
6872 memory to improve the performance of loops that access large arrays.
6874 This option may generate better or worse code; results are highly
6875 dependent on the structure of loops within the source code.
6877 Disabled at level @option{-Os}.
6880 @itemx -fno-peephole2
6881 @opindex fno-peephole
6882 @opindex fno-peephole2
6883 Disable any machine-specific peephole optimizations. The difference
6884 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
6885 are implemented in the compiler; some targets use one, some use the
6886 other, a few use both.
6888 @option{-fpeephole} is enabled by default.
6889 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6891 @item -fno-guess-branch-probability
6892 @opindex fno-guess-branch-probability
6893 Do not guess branch probabilities using heuristics.
6895 GCC will use heuristics to guess branch probabilities if they are
6896 not provided by profiling feedback (@option{-fprofile-arcs}). These
6897 heuristics are based on the control flow graph. If some branch probabilities
6898 are specified by @samp{__builtin_expect}, then the heuristics will be
6899 used to guess branch probabilities for the rest of the control flow graph,
6900 taking the @samp{__builtin_expect} info into account. The interactions
6901 between the heuristics and @samp{__builtin_expect} can be complex, and in
6902 some cases, it may be useful to disable the heuristics so that the effects
6903 of @samp{__builtin_expect} are easier to understand.
6905 The default is @option{-fguess-branch-probability} at levels
6906 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6908 @item -freorder-blocks
6909 @opindex freorder-blocks
6910 Reorder basic blocks in the compiled function in order to reduce number of
6911 taken branches and improve code locality.
6913 Enabled at levels @option{-O2}, @option{-O3}.
6915 @item -freorder-blocks-and-partition
6916 @opindex freorder-blocks-and-partition
6917 In addition to reordering basic blocks in the compiled function, in order
6918 to reduce number of taken branches, partitions hot and cold basic blocks
6919 into separate sections of the assembly and .o files, to improve
6920 paging and cache locality performance.
6922 This optimization is automatically turned off in the presence of
6923 exception handling, for linkonce sections, for functions with a user-defined
6924 section attribute and on any architecture that does not support named
6927 @item -freorder-functions
6928 @opindex freorder-functions
6929 Reorder functions in the object file in order to
6930 improve code locality. This is implemented by using special
6931 subsections @code{.text.hot} for most frequently executed functions and
6932 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
6933 the linker so object file format must support named sections and linker must
6934 place them in a reasonable way.
6936 Also profile feedback must be available in to make this option effective. See
6937 @option{-fprofile-arcs} for details.
6939 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6941 @item -fstrict-aliasing
6942 @opindex fstrict-aliasing
6943 Allow the compiler to assume the strictest aliasing rules applicable to
6944 the language being compiled. For C (and C++), this activates
6945 optimizations based on the type of expressions. In particular, an
6946 object of one type is assumed never to reside at the same address as an
6947 object of a different type, unless the types are almost the same. For
6948 example, an @code{unsigned int} can alias an @code{int}, but not a
6949 @code{void*} or a @code{double}. A character type may alias any other
6952 @anchor{Type-punning}Pay special attention to code like this:
6965 The practice of reading from a different union member than the one most
6966 recently written to (called ``type-punning'') is common. Even with
6967 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
6968 is accessed through the union type. So, the code above will work as
6969 expected. @xref{Structures unions enumerations and bit-fields
6970 implementation}. However, this code might not:
6981 Similarly, access by taking the address, casting the resulting pointer
6982 and dereferencing the result has undefined behavior, even if the cast
6983 uses a union type, e.g.:
6987 return ((union a_union *) &d)->i;
6991 The @option{-fstrict-aliasing} option is enabled at levels
6992 @option{-O2}, @option{-O3}, @option{-Os}.
6994 @item -fstrict-overflow
6995 @opindex fstrict-overflow
6996 Allow the compiler to assume strict signed overflow rules, depending
6997 on the language being compiled. For C (and C++) this means that
6998 overflow when doing arithmetic with signed numbers is undefined, which
6999 means that the compiler may assume that it will not happen. This
7000 permits various optimizations. For example, the compiler will assume
7001 that an expression like @code{i + 10 > i} will always be true for
7002 signed @code{i}. This assumption is only valid if signed overflow is
7003 undefined, as the expression is false if @code{i + 10} overflows when
7004 using twos complement arithmetic. When this option is in effect any
7005 attempt to determine whether an operation on signed numbers will
7006 overflow must be written carefully to not actually involve overflow.
7008 This option also allows the compiler to assume strict pointer
7009 semantics: given a pointer to an object, if adding an offset to that
7010 pointer does not produce a pointer to the same object, the addition is
7011 undefined. This permits the compiler to conclude that @code{p + u >
7012 p} is always true for a pointer @code{p} and unsigned integer
7013 @code{u}. This assumption is only valid because pointer wraparound is
7014 undefined, as the expression is false if @code{p + u} overflows using
7015 twos complement arithmetic.
7017 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
7018 that integer signed overflow is fully defined: it wraps. When
7019 @option{-fwrapv} is used, there is no difference between
7020 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
7021 integers. With @option{-fwrapv} certain types of overflow are
7022 permitted. For example, if the compiler gets an overflow when doing
7023 arithmetic on constants, the overflowed value can still be used with
7024 @option{-fwrapv}, but not otherwise.
7026 The @option{-fstrict-overflow} option is enabled at levels
7027 @option{-O2}, @option{-O3}, @option{-Os}.
7029 @item -falign-functions
7030 @itemx -falign-functions=@var{n}
7031 @opindex falign-functions
7032 Align the start of functions to the next power-of-two greater than
7033 @var{n}, skipping up to @var{n} bytes. For instance,
7034 @option{-falign-functions=32} aligns functions to the next 32-byte
7035 boundary, but @option{-falign-functions=24} would align to the next
7036 32-byte boundary only if this can be done by skipping 23 bytes or less.
7038 @option{-fno-align-functions} and @option{-falign-functions=1} are
7039 equivalent and mean that functions will not be aligned.
7041 Some assemblers only support this flag when @var{n} is a power of two;
7042 in that case, it is rounded up.
7044 If @var{n} is not specified or is zero, use a machine-dependent default.
7046 Enabled at levels @option{-O2}, @option{-O3}.
7048 @item -falign-labels
7049 @itemx -falign-labels=@var{n}
7050 @opindex falign-labels
7051 Align all branch targets to a power-of-two boundary, skipping up to
7052 @var{n} bytes like @option{-falign-functions}. This option can easily
7053 make code slower, because it must insert dummy operations for when the
7054 branch target is reached in the usual flow of the code.
7056 @option{-fno-align-labels} and @option{-falign-labels=1} are
7057 equivalent and mean that labels will not be aligned.
7059 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
7060 are greater than this value, then their values are used instead.
7062 If @var{n} is not specified or is zero, use a machine-dependent default
7063 which is very likely to be @samp{1}, meaning no alignment.
7065 Enabled at levels @option{-O2}, @option{-O3}.
7068 @itemx -falign-loops=@var{n}
7069 @opindex falign-loops
7070 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7071 like @option{-falign-functions}. The hope is that the loop will be
7072 executed many times, which will make up for any execution of the dummy
7075 @option{-fno-align-loops} and @option{-falign-loops=1} are
7076 equivalent and mean that loops will not be aligned.
7078 If @var{n} is not specified or is zero, use a machine-dependent default.
7080 Enabled at levels @option{-O2}, @option{-O3}.
7083 @itemx -falign-jumps=@var{n}
7084 @opindex falign-jumps
7085 Align branch targets to a power-of-two boundary, for branch targets
7086 where the targets can only be reached by jumping, skipping up to @var{n}
7087 bytes like @option{-falign-functions}. In this case, no dummy operations
7090 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7091 equivalent and mean that loops will not be aligned.
7093 If @var{n} is not specified or is zero, use a machine-dependent default.
7095 Enabled at levels @option{-O2}, @option{-O3}.
7097 @item -funit-at-a-time
7098 @opindex funit-at-a-time
7099 This option is left for compatibility reasons. @option{-funit-at-a-time}
7100 has no effect, while @option{-fno-unit-at-a-time} implies
7101 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7105 @item -fno-toplevel-reorder
7106 @opindex fno-toplevel-reorder
7107 Do not reorder top-level functions, variables, and @code{asm}
7108 statements. Output them in the same order that they appear in the
7109 input file. When this option is used, unreferenced static variables
7110 will not be removed. This option is intended to support existing code
7111 which relies on a particular ordering. For new code, it is better to
7114 Enabled at level @option{-O0}. When disabled explicitly, it also imply
7115 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
7120 Constructs webs as commonly used for register allocation purposes and assign
7121 each web individual pseudo register. This allows the register allocation pass
7122 to operate on pseudos directly, but also strengthens several other optimization
7123 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7124 however, make debugging impossible, since variables will no longer stay in a
7127 Enabled by default with @option{-funroll-loops}.
7129 @item -fwhole-program
7130 @opindex fwhole-program
7131 Assume that the current compilation unit represents the whole program being
7132 compiled. All public functions and variables with the exception of @code{main}
7133 and those merged by attribute @code{externally_visible} become static functions
7134 and in effect are optimized more aggressively by interprocedural optimizers.
7135 While this option is equivalent to proper use of the @code{static} keyword for
7136 programs consisting of a single file, in combination with option
7137 @option{-combine}, @option{-flto} or @option{-fwhopr} this flag can be used to
7138 compile many smaller scale programs since the functions and variables become
7139 local for the whole combined compilation unit, not for the single source file
7142 This option implies @option{-fwhole-file} for Fortran programs.
7146 This option runs the standard link-time optimizer. When invoked
7147 with source code, it generates GIMPLE (one of GCC's internal
7148 representations) and writes it to special ELF sections in the object
7149 file. When the object files are linked together, all the function
7150 bodies are read from these ELF sections and instantiated as if they
7151 had been part of the same translation unit.
7153 To use the link-timer optimizer, @option{-flto} needs to be specified at
7154 compile time and during the final link. For example,
7157 gcc -c -O2 -flto foo.c
7158 gcc -c -O2 -flto bar.c
7159 gcc -o myprog -flto -O2 foo.o bar.o
7162 The first two invocations to GCC will save a bytecode representation
7163 of GIMPLE into special ELF sections inside @file{foo.o} and
7164 @file{bar.o}. The final invocation will read the GIMPLE bytecode from
7165 @file{foo.o} and @file{bar.o}, merge the two files into a single
7166 internal image, and compile the result as usual. Since both
7167 @file{foo.o} and @file{bar.o} are merged into a single image, this
7168 causes all the inter-procedural analyses and optimizations in GCC to
7169 work across the two files as if they were a single one. This means,
7170 for example, that the inliner will be able to inline functions in
7171 @file{bar.o} into functions in @file{foo.o} and vice-versa.
7173 Another (simpler) way to enable link-time optimization is,
7176 gcc -o myprog -flto -O2 foo.c bar.c
7179 The above will generate bytecode for @file{foo.c} and @file{bar.c},
7180 merge them together into a single GIMPLE representation and optimize
7181 them as usual to produce @file{myprog}.
7183 The only important thing to keep in mind is that to enable link-time
7184 optimizations the @option{-flto} flag needs to be passed to both the
7185 compile and the link commands.
7187 Note that when a file is compiled with @option{-flto}, the generated
7188 object file will be larger than a regular object file because it will
7189 contain GIMPLE bytecodes and the usual final code. This means that
7190 object files with LTO information can be linked as a normal object
7191 file. So, in the previous example, if the final link is done with
7194 gcc -o myprog foo.o bar.o
7197 The only difference will be that no inter-procedural optimizations
7198 will be applied to produce @file{myprog}. The two object files
7199 @file{foo.o} and @file{bar.o} will be simply sent to the regular
7202 Additionally, the optimization flags used to compile individual files
7203 are not necessarily related to those used at link-time. For instance,
7206 gcc -c -O0 -flto foo.c
7207 gcc -c -O0 -flto bar.c
7208 gcc -o myprog -flto -O3 foo.o bar.o
7211 This will produce individual object files with unoptimized assembler
7212 code, but the resulting binary @file{myprog} will be optimized at
7213 @option{-O3}. Now, if the final binary is generated without
7214 @option{-flto}, then @file{myprog} will not be optimized.
7216 When producing the final binary with @option{-flto}, GCC will only
7217 apply link-time optimizations to those files that contain bytecode.
7218 Therefore, you can mix and match object files and libraries with
7219 GIMPLE bytecodes and final object code. GCC will automatically select
7220 which files to optimize in LTO mode and which files to link without
7223 There are some code generation flags that GCC will preserve when
7224 generating bytecodes, as they need to be used during the final link
7225 stage. Currently, the following options are saved into the GIMPLE
7226 bytecode files: @option{-fPIC}, @option{-fcommon} and all the
7227 @option{-m} target flags.
7229 At link time, these options are read-in and reapplied. Note that the
7230 current implementation makes no attempt at recognizing conflicting
7231 values for these options. If two or more files have a conflicting
7232 value (e.g., one file is compiled with @option{-fPIC} and another
7233 isn't), the compiler will simply use the last value read from the
7234 bytecode files. It is recommended, then, that all the files
7235 participating in the same link be compiled with the same options.
7237 Another feature of LTO is that it is possible to apply interprocedural
7238 optimizations on files written in different languages. This requires
7239 some support in the language front end. Currently, the C, C++ and
7240 Fortran front ends are capable of emitting GIMPLE bytecodes, so
7241 something like this should work
7246 gfortran -c -flto baz.f90
7247 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
7250 Notice that the final link is done with @command{g++} to get the C++
7251 runtime libraries and @option{-lgfortran} is added to get the Fortran
7252 runtime libraries. In general, when mixing languages in LTO mode, you
7253 should use the same link command used when mixing languages in a
7254 regular (non-LTO) compilation. This means that if your build process
7255 was mixing languages before, all you need to add is @option{-flto} to
7256 all the compile and link commands.
7258 If object files containing GIMPLE bytecode are stored in a library
7259 archive, say @file{libfoo.a}, it is possible to extract and use them
7260 in an LTO link if you are using @command{gold} as the linker (which,
7261 in turn requires GCC to be configured with @option{--enable-gold}).
7262 To enable this feature, use the flag @option{-use-linker-plugin} at
7266 gcc -o myprog -O2 -flto -use-linker-plugin a.o b.o -lfoo
7269 With the linker plugin enabled, @command{gold} will extract the needed
7270 GIMPLE files from @file{libfoo.a} and pass them on to the running GCC
7271 to make them part of the aggregated GIMPLE image to be optimized.
7273 If you are not using @command{gold} and/or do not specify
7274 @option{-use-linker-plugin} then the objects inside @file{libfoo.a}
7275 will be extracted and linked as usual, but they will not participate
7276 in the LTO optimization process.
7278 Link time optimizations do not require the presence of the whole
7279 program to operate. If the program does not require any symbols to
7280 be exported, it is possible to combine @option{-flto} and
7281 @option{-fwhopr} with @option{-fwhole-program} to allow the
7282 interprocedural optimizers to use more aggressive assumptions which
7283 may lead to improved optimization opportunities.
7285 Regarding portability: the current implementation of LTO makes no
7286 attempt at generating bytecode that can be ported between different
7287 types of hosts. The bytecode files are versioned and there is a
7288 strict version check, so bytecode files generated in one version of
7289 GCC will not work with an older/newer version of GCC.
7291 This option is disabled by default.
7295 This option is identical in functionality to @option{-flto} but it
7296 differs in how the final link stage is executed. Instead of loading
7297 all the function bodies in memory, the callgraph is analyzed and
7298 optimization decisions are made (whole program analysis or WPA). Once
7299 optimization decisions are made, the callgraph is partitioned and the
7300 different sections are compiled separately (local transformations or
7301 LTRANS)@. This process allows optimizations on very large programs
7302 that otherwise would not fit in memory. This option enables
7303 @option{-fwpa} and @option{-fltrans} automatically.
7305 Disabled by default.
7309 This is an internal option used by GCC when compiling with
7310 @option{-fwhopr}. You should never need to use it.
7312 This option runs the link-time optimizer in the whole-program-analysis
7313 (WPA) mode, which reads in summary information from all inputs and
7314 performs a whole-program analysis based on summary information only.
7315 It generates object files for subsequent runs of the link-time
7316 optimizer where individual object files are optimized using both
7317 summary information from the WPA mode and the actual function bodies.
7318 It then drives the LTRANS phase.
7320 Disabled by default.
7324 This is an internal option used by GCC when compiling with
7325 @option{-fwhopr}. You should never need to use it.
7327 This option runs the link-time optimizer in the local-transformation (LTRANS)
7328 mode, which reads in output from a previous run of the LTO in WPA mode.
7329 In the LTRANS mode, LTO optimizes an object and produces the final assembly.
7331 Disabled by default.
7333 @item -fltrans-output-list=@var{file}
7334 @opindex fltrans-output-list
7335 This is an internal option used by GCC when compiling with
7336 @option{-fwhopr}. You should never need to use it.
7338 This option specifies a file to which the names of LTRANS output files are
7339 written. This option is only meaningful in conjunction with @option{-fwpa}.
7341 Disabled by default.
7343 @item -flto-compression-level=@var{n}
7344 This option specifies the level of compression used for intermediate
7345 language written to LTO object files, and is only meaningful in
7346 conjunction with LTO mode (@option{-fwhopr}, @option{-flto}). Valid
7347 values are 0 (no compression) to 9 (maximum compression). Values
7348 outside this range are clamped to either 0 or 9. If the option is not
7349 given, a default balanced compression setting is used.
7352 Prints a report with internal details on the workings of the link-time
7353 optimizer. The contents of this report vary from version to version,
7354 it is meant to be useful to GCC developers when processing object
7355 files in LTO mode (via @option{-fwhopr} or @option{-flto}).
7357 Disabled by default.
7359 @item -use-linker-plugin
7360 Enables the extraction of objects with GIMPLE bytecode information
7361 from library archives. This option relies on features available only
7362 in @command{gold}, so to use this you must configure GCC with
7363 @option{--enable-gold}. See @option{-flto} for a description on the
7364 effect of this flag and how to use it.
7366 Disabled by default.
7368 @item -fcprop-registers
7369 @opindex fcprop-registers
7370 After register allocation and post-register allocation instruction splitting,
7371 we perform a copy-propagation pass to try to reduce scheduling dependencies
7372 and occasionally eliminate the copy.
7374 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7376 @item -fprofile-correction
7377 @opindex fprofile-correction
7378 Profiles collected using an instrumented binary for multi-threaded programs may
7379 be inconsistent due to missed counter updates. When this option is specified,
7380 GCC will use heuristics to correct or smooth out such inconsistencies. By
7381 default, GCC will emit an error message when an inconsistent profile is detected.
7383 @item -fprofile-dir=@var{path}
7384 @opindex fprofile-dir
7386 Set the directory to search the profile data files in to @var{path}.
7387 This option affects only the profile data generated by
7388 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
7389 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
7390 and its related options.
7391 By default, GCC will use the current directory as @var{path}
7392 thus the profile data file will appear in the same directory as the object file.
7394 @item -fprofile-generate
7395 @itemx -fprofile-generate=@var{path}
7396 @opindex fprofile-generate
7398 Enable options usually used for instrumenting application to produce
7399 profile useful for later recompilation with profile feedback based
7400 optimization. You must use @option{-fprofile-generate} both when
7401 compiling and when linking your program.
7403 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
7405 If @var{path} is specified, GCC will look at the @var{path} to find
7406 the profile feedback data files. See @option{-fprofile-dir}.
7409 @itemx -fprofile-use=@var{path}
7410 @opindex fprofile-use
7411 Enable profile feedback directed optimizations, and optimizations
7412 generally profitable only with profile feedback available.
7414 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
7415 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
7417 By default, GCC emits an error message if the feedback profiles do not
7418 match the source code. This error can be turned into a warning by using
7419 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
7422 If @var{path} is specified, GCC will look at the @var{path} to find
7423 the profile feedback data files. See @option{-fprofile-dir}.
7426 The following options control compiler behavior regarding floating
7427 point arithmetic. These options trade off between speed and
7428 correctness. All must be specifically enabled.
7432 @opindex ffloat-store
7433 Do not store floating point variables in registers, and inhibit other
7434 options that might change whether a floating point value is taken from a
7437 @cindex floating point precision
7438 This option prevents undesirable excess precision on machines such as
7439 the 68000 where the floating registers (of the 68881) keep more
7440 precision than a @code{double} is supposed to have. Similarly for the
7441 x86 architecture. For most programs, the excess precision does only
7442 good, but a few programs rely on the precise definition of IEEE floating
7443 point. Use @option{-ffloat-store} for such programs, after modifying
7444 them to store all pertinent intermediate computations into variables.
7446 @item -fexcess-precision=@var{style}
7447 @opindex fexcess-precision
7448 This option allows further control over excess precision on machines
7449 where floating-point registers have more precision than the IEEE
7450 @code{float} and @code{double} types and the processor does not
7451 support operations rounding to those types. By default,
7452 @option{-fexcess-precision=fast} is in effect; this means that
7453 operations are carried out in the precision of the registers and that
7454 it is unpredictable when rounding to the types specified in the source
7455 code takes place. When compiling C, if
7456 @option{-fexcess-precision=standard} is specified then excess
7457 precision will follow the rules specified in ISO C99; in particular,
7458 both casts and assignments cause values to be rounded to their
7459 semantic types (whereas @option{-ffloat-store} only affects
7460 assignments). This option is enabled by default for C if a strict
7461 conformance option such as @option{-std=c99} is used.
7464 @option{-fexcess-precision=standard} is not implemented for languages
7465 other than C, and has no effect if
7466 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
7467 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
7468 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
7469 semantics apply without excess precision, and in the latter, rounding
7474 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
7475 @option{-ffinite-math-only}, @option{-fno-rounding-math},
7476 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
7478 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
7480 This option is not turned on by any @option{-O} option since
7481 it can result in incorrect output for programs which depend on
7482 an exact implementation of IEEE or ISO rules/specifications for
7483 math functions. It may, however, yield faster code for programs
7484 that do not require the guarantees of these specifications.
7486 @item -fno-math-errno
7487 @opindex fno-math-errno
7488 Do not set ERRNO after calling math functions that are executed
7489 with a single instruction, e.g., sqrt. A program that relies on
7490 IEEE exceptions for math error handling may want to use this flag
7491 for speed while maintaining IEEE arithmetic compatibility.
7493 This option is not turned on by any @option{-O} option since
7494 it can result in incorrect output for programs which depend on
7495 an exact implementation of IEEE or ISO rules/specifications for
7496 math functions. It may, however, yield faster code for programs
7497 that do not require the guarantees of these specifications.
7499 The default is @option{-fmath-errno}.
7501 On Darwin systems, the math library never sets @code{errno}. There is
7502 therefore no reason for the compiler to consider the possibility that
7503 it might, and @option{-fno-math-errno} is the default.
7505 @item -funsafe-math-optimizations
7506 @opindex funsafe-math-optimizations
7508 Allow optimizations for floating-point arithmetic that (a) assume
7509 that arguments and results are valid and (b) may violate IEEE or
7510 ANSI standards. When used at link-time, it may include libraries
7511 or startup files that change the default FPU control word or other
7512 similar optimizations.
7514 This option is not turned on by any @option{-O} option since
7515 it can result in incorrect output for programs which depend on
7516 an exact implementation of IEEE or ISO rules/specifications for
7517 math functions. It may, however, yield faster code for programs
7518 that do not require the guarantees of these specifications.
7519 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
7520 @option{-fassociative-math} and @option{-freciprocal-math}.
7522 The default is @option{-fno-unsafe-math-optimizations}.
7524 @item -fassociative-math
7525 @opindex fassociative-math
7527 Allow re-association of operands in series of floating-point operations.
7528 This violates the ISO C and C++ language standard by possibly changing
7529 computation result. NOTE: re-ordering may change the sign of zero as
7530 well as ignore NaNs and inhibit or create underflow or overflow (and
7531 thus cannot be used on a code which relies on rounding behavior like
7532 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
7533 and thus may not be used when ordered comparisons are required.
7534 This option requires that both @option{-fno-signed-zeros} and
7535 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
7536 much sense with @option{-frounding-math}.
7538 The default is @option{-fno-associative-math}.
7540 @item -freciprocal-math
7541 @opindex freciprocal-math
7543 Allow the reciprocal of a value to be used instead of dividing by
7544 the value if this enables optimizations. For example @code{x / y}
7545 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
7546 is subject to common subexpression elimination. Note that this loses
7547 precision and increases the number of flops operating on the value.
7549 The default is @option{-fno-reciprocal-math}.
7551 @item -ffinite-math-only
7552 @opindex ffinite-math-only
7553 Allow optimizations for floating-point arithmetic that assume
7554 that arguments and results are not NaNs or +-Infs.
7556 This option is not turned on by any @option{-O} option since
7557 it can result in incorrect output for programs which depend on
7558 an exact implementation of IEEE or ISO rules/specifications for
7559 math functions. It may, however, yield faster code for programs
7560 that do not require the guarantees of these specifications.
7562 The default is @option{-fno-finite-math-only}.
7564 @item -fno-signed-zeros
7565 @opindex fno-signed-zeros
7566 Allow optimizations for floating point arithmetic that ignore the
7567 signedness of zero. IEEE arithmetic specifies the behavior of
7568 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
7569 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
7570 This option implies that the sign of a zero result isn't significant.
7572 The default is @option{-fsigned-zeros}.
7574 @item -fno-trapping-math
7575 @opindex fno-trapping-math
7576 Compile code assuming that floating-point operations cannot generate
7577 user-visible traps. These traps include division by zero, overflow,
7578 underflow, inexact result and invalid operation. This option requires
7579 that @option{-fno-signaling-nans} be in effect. Setting this option may
7580 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
7582 This option should never be turned on by any @option{-O} option since
7583 it can result in incorrect output for programs which depend on
7584 an exact implementation of IEEE or ISO rules/specifications for
7587 The default is @option{-ftrapping-math}.
7589 @item -frounding-math
7590 @opindex frounding-math
7591 Disable transformations and optimizations that assume default floating
7592 point rounding behavior. This is round-to-zero for all floating point
7593 to integer conversions, and round-to-nearest for all other arithmetic
7594 truncations. This option should be specified for programs that change
7595 the FP rounding mode dynamically, or that may be executed with a
7596 non-default rounding mode. This option disables constant folding of
7597 floating point expressions at compile-time (which may be affected by
7598 rounding mode) and arithmetic transformations that are unsafe in the
7599 presence of sign-dependent rounding modes.
7601 The default is @option{-fno-rounding-math}.
7603 This option is experimental and does not currently guarantee to
7604 disable all GCC optimizations that are affected by rounding mode.
7605 Future versions of GCC may provide finer control of this setting
7606 using C99's @code{FENV_ACCESS} pragma. This command line option
7607 will be used to specify the default state for @code{FENV_ACCESS}.
7609 @item -fsignaling-nans
7610 @opindex fsignaling-nans
7611 Compile code assuming that IEEE signaling NaNs may generate user-visible
7612 traps during floating-point operations. Setting this option disables
7613 optimizations that may change the number of exceptions visible with
7614 signaling NaNs. This option implies @option{-ftrapping-math}.
7616 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
7619 The default is @option{-fno-signaling-nans}.
7621 This option is experimental and does not currently guarantee to
7622 disable all GCC optimizations that affect signaling NaN behavior.
7624 @item -fsingle-precision-constant
7625 @opindex fsingle-precision-constant
7626 Treat floating point constant as single precision constant instead of
7627 implicitly converting it to double precision constant.
7629 @item -fcx-limited-range
7630 @opindex fcx-limited-range
7631 When enabled, this option states that a range reduction step is not
7632 needed when performing complex division. Also, there is no checking
7633 whether the result of a complex multiplication or division is @code{NaN
7634 + I*NaN}, with an attempt to rescue the situation in that case. The
7635 default is @option{-fno-cx-limited-range}, but is enabled by
7636 @option{-ffast-math}.
7638 This option controls the default setting of the ISO C99
7639 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
7642 @item -fcx-fortran-rules
7643 @opindex fcx-fortran-rules
7644 Complex multiplication and division follow Fortran rules. Range
7645 reduction is done as part of complex division, but there is no checking
7646 whether the result of a complex multiplication or division is @code{NaN
7647 + I*NaN}, with an attempt to rescue the situation in that case.
7649 The default is @option{-fno-cx-fortran-rules}.
7653 The following options control optimizations that may improve
7654 performance, but are not enabled by any @option{-O} options. This
7655 section includes experimental options that may produce broken code.
7658 @item -fbranch-probabilities
7659 @opindex fbranch-probabilities
7660 After running a program compiled with @option{-fprofile-arcs}
7661 (@pxref{Debugging Options,, Options for Debugging Your Program or
7662 @command{gcc}}), you can compile it a second time using
7663 @option{-fbranch-probabilities}, to improve optimizations based on
7664 the number of times each branch was taken. When the program
7665 compiled with @option{-fprofile-arcs} exits it saves arc execution
7666 counts to a file called @file{@var{sourcename}.gcda} for each source
7667 file. The information in this data file is very dependent on the
7668 structure of the generated code, so you must use the same source code
7669 and the same optimization options for both compilations.
7671 With @option{-fbranch-probabilities}, GCC puts a
7672 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
7673 These can be used to improve optimization. Currently, they are only
7674 used in one place: in @file{reorg.c}, instead of guessing which path a
7675 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
7676 exactly determine which path is taken more often.
7678 @item -fprofile-values
7679 @opindex fprofile-values
7680 If combined with @option{-fprofile-arcs}, it adds code so that some
7681 data about values of expressions in the program is gathered.
7683 With @option{-fbranch-probabilities}, it reads back the data gathered
7684 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
7685 notes to instructions for their later usage in optimizations.
7687 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
7691 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
7692 a code to gather information about values of expressions.
7694 With @option{-fbranch-probabilities}, it reads back the data gathered
7695 and actually performs the optimizations based on them.
7696 Currently the optimizations include specialization of division operation
7697 using the knowledge about the value of the denominator.
7699 @item -frename-registers
7700 @opindex frename-registers
7701 Attempt to avoid false dependencies in scheduled code by making use
7702 of registers left over after register allocation. This optimization
7703 will most benefit processors with lots of registers. Depending on the
7704 debug information format adopted by the target, however, it can
7705 make debugging impossible, since variables will no longer stay in
7706 a ``home register''.
7708 Enabled by default with @option{-funroll-loops}.
7712 Perform tail duplication to enlarge superblock size. This transformation
7713 simplifies the control flow of the function allowing other optimizations to do
7716 Enabled with @option{-fprofile-use}.
7718 @item -funroll-loops
7719 @opindex funroll-loops
7720 Unroll loops whose number of iterations can be determined at compile time or
7721 upon entry to the loop. @option{-funroll-loops} implies
7722 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
7723 It also turns on complete loop peeling (i.e.@: complete removal of loops with
7724 small constant number of iterations). This option makes code larger, and may
7725 or may not make it run faster.
7727 Enabled with @option{-fprofile-use}.
7729 @item -funroll-all-loops
7730 @opindex funroll-all-loops
7731 Unroll all loops, even if their number of iterations is uncertain when
7732 the loop is entered. This usually makes programs run more slowly.
7733 @option{-funroll-all-loops} implies the same options as
7734 @option{-funroll-loops}.
7737 @opindex fpeel-loops
7738 Peels the loops for that there is enough information that they do not
7739 roll much (from profile feedback). It also turns on complete loop peeling
7740 (i.e.@: complete removal of loops with small constant number of iterations).
7742 Enabled with @option{-fprofile-use}.
7744 @item -fmove-loop-invariants
7745 @opindex fmove-loop-invariants
7746 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
7747 at level @option{-O1}
7749 @item -funswitch-loops
7750 @opindex funswitch-loops
7751 Move branches with loop invariant conditions out of the loop, with duplicates
7752 of the loop on both branches (modified according to result of the condition).
7754 @item -ffunction-sections
7755 @itemx -fdata-sections
7756 @opindex ffunction-sections
7757 @opindex fdata-sections
7758 Place each function or data item into its own section in the output
7759 file if the target supports arbitrary sections. The name of the
7760 function or the name of the data item determines the section's name
7763 Use these options on systems where the linker can perform optimizations
7764 to improve locality of reference in the instruction space. Most systems
7765 using the ELF object format and SPARC processors running Solaris 2 have
7766 linkers with such optimizations. AIX may have these optimizations in
7769 Only use these options when there are significant benefits from doing
7770 so. When you specify these options, the assembler and linker will
7771 create larger object and executable files and will also be slower.
7772 You will not be able to use @code{gprof} on all systems if you
7773 specify this option and you may have problems with debugging if
7774 you specify both this option and @option{-g}.
7776 @item -fbranch-target-load-optimize
7777 @opindex fbranch-target-load-optimize
7778 Perform branch target register load optimization before prologue / epilogue
7780 The use of target registers can typically be exposed only during reload,
7781 thus hoisting loads out of loops and doing inter-block scheduling needs
7782 a separate optimization pass.
7784 @item -fbranch-target-load-optimize2
7785 @opindex fbranch-target-load-optimize2
7786 Perform branch target register load optimization after prologue / epilogue
7789 @item -fbtr-bb-exclusive
7790 @opindex fbtr-bb-exclusive
7791 When performing branch target register load optimization, don't reuse
7792 branch target registers in within any basic block.
7794 @item -fstack-protector
7795 @opindex fstack-protector
7796 Emit extra code to check for buffer overflows, such as stack smashing
7797 attacks. This is done by adding a guard variable to functions with
7798 vulnerable objects. This includes functions that call alloca, and
7799 functions with buffers larger than 8 bytes. The guards are initialized
7800 when a function is entered and then checked when the function exits.
7801 If a guard check fails, an error message is printed and the program exits.
7803 @item -fstack-protector-all
7804 @opindex fstack-protector-all
7805 Like @option{-fstack-protector} except that all functions are protected.
7807 @item -fsection-anchors
7808 @opindex fsection-anchors
7809 Try to reduce the number of symbolic address calculations by using
7810 shared ``anchor'' symbols to address nearby objects. This transformation
7811 can help to reduce the number of GOT entries and GOT accesses on some
7814 For example, the implementation of the following function @code{foo}:
7818 int foo (void) @{ return a + b + c; @}
7821 would usually calculate the addresses of all three variables, but if you
7822 compile it with @option{-fsection-anchors}, it will access the variables
7823 from a common anchor point instead. The effect is similar to the
7824 following pseudocode (which isn't valid C):
7829 register int *xr = &x;
7830 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
7834 Not all targets support this option.
7836 @item --param @var{name}=@var{value}
7838 In some places, GCC uses various constants to control the amount of
7839 optimization that is done. For example, GCC will not inline functions
7840 that contain more that a certain number of instructions. You can
7841 control some of these constants on the command-line using the
7842 @option{--param} option.
7844 The names of specific parameters, and the meaning of the values, are
7845 tied to the internals of the compiler, and are subject to change
7846 without notice in future releases.
7848 In each case, the @var{value} is an integer. The allowable choices for
7849 @var{name} are given in the following table:
7852 @item struct-reorg-cold-struct-ratio
7853 The threshold ratio (as a percentage) between a structure frequency
7854 and the frequency of the hottest structure in the program. This parameter
7855 is used by struct-reorg optimization enabled by @option{-fipa-struct-reorg}.
7856 We say that if the ratio of a structure frequency, calculated by profiling,
7857 to the hottest structure frequency in the program is less than this
7858 parameter, then structure reorganization is not applied to this structure.
7861 @item predictable-branch-cost-outcome
7862 When branch is predicted to be taken with probability lower than this threshold
7863 (in percent), then it is considered well predictable. The default is 10.
7865 @item max-crossjump-edges
7866 The maximum number of incoming edges to consider for crossjumping.
7867 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
7868 the number of edges incoming to each block. Increasing values mean
7869 more aggressive optimization, making the compile time increase with
7870 probably small improvement in executable size.
7872 @item min-crossjump-insns
7873 The minimum number of instructions which must be matched at the end
7874 of two blocks before crossjumping will be performed on them. This
7875 value is ignored in the case where all instructions in the block being
7876 crossjumped from are matched. The default value is 5.
7878 @item max-grow-copy-bb-insns
7879 The maximum code size expansion factor when copying basic blocks
7880 instead of jumping. The expansion is relative to a jump instruction.
7881 The default value is 8.
7883 @item max-goto-duplication-insns
7884 The maximum number of instructions to duplicate to a block that jumps
7885 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
7886 passes, GCC factors computed gotos early in the compilation process,
7887 and unfactors them as late as possible. Only computed jumps at the
7888 end of a basic blocks with no more than max-goto-duplication-insns are
7889 unfactored. The default value is 8.
7891 @item max-delay-slot-insn-search
7892 The maximum number of instructions to consider when looking for an
7893 instruction to fill a delay slot. If more than this arbitrary number of
7894 instructions is searched, the time savings from filling the delay slot
7895 will be minimal so stop searching. Increasing values mean more
7896 aggressive optimization, making the compile time increase with probably
7897 small improvement in executable run time.
7899 @item max-delay-slot-live-search
7900 When trying to fill delay slots, the maximum number of instructions to
7901 consider when searching for a block with valid live register
7902 information. Increasing this arbitrarily chosen value means more
7903 aggressive optimization, increasing the compile time. This parameter
7904 should be removed when the delay slot code is rewritten to maintain the
7907 @item max-gcse-memory
7908 The approximate maximum amount of memory that will be allocated in
7909 order to perform the global common subexpression elimination
7910 optimization. If more memory than specified is required, the
7911 optimization will not be done.
7913 @item max-pending-list-length
7914 The maximum number of pending dependencies scheduling will allow
7915 before flushing the current state and starting over. Large functions
7916 with few branches or calls can create excessively large lists which
7917 needlessly consume memory and resources.
7919 @item max-inline-insns-single
7920 Several parameters control the tree inliner used in gcc.
7921 This number sets the maximum number of instructions (counted in GCC's
7922 internal representation) in a single function that the tree inliner
7923 will consider for inlining. This only affects functions declared
7924 inline and methods implemented in a class declaration (C++).
7925 The default value is 300.
7927 @item max-inline-insns-auto
7928 When you use @option{-finline-functions} (included in @option{-O3}),
7929 a lot of functions that would otherwise not be considered for inlining
7930 by the compiler will be investigated. To those functions, a different
7931 (more restrictive) limit compared to functions declared inline can
7933 The default value is 50.
7935 @item large-function-insns
7936 The limit specifying really large functions. For functions larger than this
7937 limit after inlining, inlining is constrained by
7938 @option{--param large-function-growth}. This parameter is useful primarily
7939 to avoid extreme compilation time caused by non-linear algorithms used by the
7941 The default value is 2700.
7943 @item large-function-growth
7944 Specifies maximal growth of large function caused by inlining in percents.
7945 The default value is 100 which limits large function growth to 2.0 times
7948 @item large-unit-insns
7949 The limit specifying large translation unit. Growth caused by inlining of
7950 units larger than this limit is limited by @option{--param inline-unit-growth}.
7951 For small units this might be too tight (consider unit consisting of function A
7952 that is inline and B that just calls A three time. If B is small relative to
7953 A, the growth of unit is 300\% and yet such inlining is very sane. For very
7954 large units consisting of small inlineable functions however the overall unit
7955 growth limit is needed to avoid exponential explosion of code size. Thus for
7956 smaller units, the size is increased to @option{--param large-unit-insns}
7957 before applying @option{--param inline-unit-growth}. The default is 10000
7959 @item inline-unit-growth
7960 Specifies maximal overall growth of the compilation unit caused by inlining.
7961 The default value is 30 which limits unit growth to 1.3 times the original
7964 @item ipcp-unit-growth
7965 Specifies maximal overall growth of the compilation unit caused by
7966 interprocedural constant propagation. The default value is 10 which limits
7967 unit growth to 1.1 times the original size.
7969 @item large-stack-frame
7970 The limit specifying large stack frames. While inlining the algorithm is trying
7971 to not grow past this limit too much. Default value is 256 bytes.
7973 @item large-stack-frame-growth
7974 Specifies maximal growth of large stack frames caused by inlining in percents.
7975 The default value is 1000 which limits large stack frame growth to 11 times
7978 @item max-inline-insns-recursive
7979 @itemx max-inline-insns-recursive-auto
7980 Specifies maximum number of instructions out-of-line copy of self recursive inline
7981 function can grow into by performing recursive inlining.
7983 For functions declared inline @option{--param max-inline-insns-recursive} is
7984 taken into account. For function not declared inline, recursive inlining
7985 happens only when @option{-finline-functions} (included in @option{-O3}) is
7986 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
7987 default value is 450.
7989 @item max-inline-recursive-depth
7990 @itemx max-inline-recursive-depth-auto
7991 Specifies maximum recursion depth used by the recursive inlining.
7993 For functions declared inline @option{--param max-inline-recursive-depth} is
7994 taken into account. For function not declared inline, recursive inlining
7995 happens only when @option{-finline-functions} (included in @option{-O3}) is
7996 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
7999 @item min-inline-recursive-probability
8000 Recursive inlining is profitable only for function having deep recursion
8001 in average and can hurt for function having little recursion depth by
8002 increasing the prologue size or complexity of function body to other
8005 When profile feedback is available (see @option{-fprofile-generate}) the actual
8006 recursion depth can be guessed from probability that function will recurse via
8007 given call expression. This parameter limits inlining only to call expression
8008 whose probability exceeds given threshold (in percents). The default value is
8011 @item early-inlining-insns
8012 Specify growth that early inliner can make. In effect it increases amount of
8013 inlining for code having large abstraction penalty. The default value is 8.
8015 @item max-early-inliner-iterations
8016 @itemx max-early-inliner-iterations
8017 Limit of iterations of early inliner. This basically bounds number of nested
8018 indirect calls early inliner can resolve. Deeper chains are still handled by
8021 @item min-vect-loop-bound
8022 The minimum number of iterations under which a loop will not get vectorized
8023 when @option{-ftree-vectorize} is used. The number of iterations after
8024 vectorization needs to be greater than the value specified by this option
8025 to allow vectorization. The default value is 0.
8027 @item max-unrolled-insns
8028 The maximum number of instructions that a loop should have if that loop
8029 is unrolled, and if the loop is unrolled, it determines how many times
8030 the loop code is unrolled.
8032 @item max-average-unrolled-insns
8033 The maximum number of instructions biased by probabilities of their execution
8034 that a loop should have if that loop is unrolled, and if the loop is unrolled,
8035 it determines how many times the loop code is unrolled.
8037 @item max-unroll-times
8038 The maximum number of unrollings of a single loop.
8040 @item max-peeled-insns
8041 The maximum number of instructions that a loop should have if that loop
8042 is peeled, and if the loop is peeled, it determines how many times
8043 the loop code is peeled.
8045 @item max-peel-times
8046 The maximum number of peelings of a single loop.
8048 @item max-completely-peeled-insns
8049 The maximum number of insns of a completely peeled loop.
8051 @item max-completely-peel-times
8052 The maximum number of iterations of a loop to be suitable for complete peeling.
8054 @item max-unswitch-insns
8055 The maximum number of insns of an unswitched loop.
8057 @item max-unswitch-level
8058 The maximum number of branches unswitched in a single loop.
8061 The minimum cost of an expensive expression in the loop invariant motion.
8063 @item iv-consider-all-candidates-bound
8064 Bound on number of candidates for induction variables below that
8065 all candidates are considered for each use in induction variable
8066 optimizations. Only the most relevant candidates are considered
8067 if there are more candidates, to avoid quadratic time complexity.
8069 @item iv-max-considered-uses
8070 The induction variable optimizations give up on loops that contain more
8071 induction variable uses.
8073 @item iv-always-prune-cand-set-bound
8074 If number of candidates in the set is smaller than this value,
8075 we always try to remove unnecessary ivs from the set during its
8076 optimization when a new iv is added to the set.
8078 @item scev-max-expr-size
8079 Bound on size of expressions used in the scalar evolutions analyzer.
8080 Large expressions slow the analyzer.
8082 @item omega-max-vars
8083 The maximum number of variables in an Omega constraint system.
8084 The default value is 128.
8086 @item omega-max-geqs
8087 The maximum number of inequalities in an Omega constraint system.
8088 The default value is 256.
8091 The maximum number of equalities in an Omega constraint system.
8092 The default value is 128.
8094 @item omega-max-wild-cards
8095 The maximum number of wildcard variables that the Omega solver will
8096 be able to insert. The default value is 18.
8098 @item omega-hash-table-size
8099 The size of the hash table in the Omega solver. The default value is
8102 @item omega-max-keys
8103 The maximal number of keys used by the Omega solver. The default
8106 @item omega-eliminate-redundant-constraints
8107 When set to 1, use expensive methods to eliminate all redundant
8108 constraints. The default value is 0.
8110 @item vect-max-version-for-alignment-checks
8111 The maximum number of runtime checks that can be performed when
8112 doing loop versioning for alignment in the vectorizer. See option
8113 ftree-vect-loop-version for more information.
8115 @item vect-max-version-for-alias-checks
8116 The maximum number of runtime checks that can be performed when
8117 doing loop versioning for alias in the vectorizer. See option
8118 ftree-vect-loop-version for more information.
8120 @item max-iterations-to-track
8122 The maximum number of iterations of a loop the brute force algorithm
8123 for analysis of # of iterations of the loop tries to evaluate.
8125 @item hot-bb-count-fraction
8126 Select fraction of the maximal count of repetitions of basic block in program
8127 given basic block needs to have to be considered hot.
8129 @item hot-bb-frequency-fraction
8130 Select fraction of the maximal frequency of executions of basic block in
8131 function given basic block needs to have to be considered hot
8133 @item max-predicted-iterations
8134 The maximum number of loop iterations we predict statically. This is useful
8135 in cases where function contain single loop with known bound and other loop
8136 with unknown. We predict the known number of iterations correctly, while
8137 the unknown number of iterations average to roughly 10. This means that the
8138 loop without bounds would appear artificially cold relative to the other one.
8140 @item align-threshold
8142 Select fraction of the maximal frequency of executions of basic block in
8143 function given basic block will get aligned.
8145 @item align-loop-iterations
8147 A loop expected to iterate at lest the selected number of iterations will get
8150 @item tracer-dynamic-coverage
8151 @itemx tracer-dynamic-coverage-feedback
8153 This value is used to limit superblock formation once the given percentage of
8154 executed instructions is covered. This limits unnecessary code size
8157 The @option{tracer-dynamic-coverage-feedback} is used only when profile
8158 feedback is available. The real profiles (as opposed to statically estimated
8159 ones) are much less balanced allowing the threshold to be larger value.
8161 @item tracer-max-code-growth
8162 Stop tail duplication once code growth has reached given percentage. This is
8163 rather hokey argument, as most of the duplicates will be eliminated later in
8164 cross jumping, so it may be set to much higher values than is the desired code
8167 @item tracer-min-branch-ratio
8169 Stop reverse growth when the reverse probability of best edge is less than this
8170 threshold (in percent).
8172 @item tracer-min-branch-ratio
8173 @itemx tracer-min-branch-ratio-feedback
8175 Stop forward growth if the best edge do have probability lower than this
8178 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
8179 compilation for profile feedback and one for compilation without. The value
8180 for compilation with profile feedback needs to be more conservative (higher) in
8181 order to make tracer effective.
8183 @item max-cse-path-length
8185 Maximum number of basic blocks on path that cse considers. The default is 10.
8188 The maximum instructions CSE process before flushing. The default is 1000.
8190 @item ggc-min-expand
8192 GCC uses a garbage collector to manage its own memory allocation. This
8193 parameter specifies the minimum percentage by which the garbage
8194 collector's heap should be allowed to expand between collections.
8195 Tuning this may improve compilation speed; it has no effect on code
8198 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
8199 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
8200 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
8201 GCC is not able to calculate RAM on a particular platform, the lower
8202 bound of 30% is used. Setting this parameter and
8203 @option{ggc-min-heapsize} to zero causes a full collection to occur at
8204 every opportunity. This is extremely slow, but can be useful for
8207 @item ggc-min-heapsize
8209 Minimum size of the garbage collector's heap before it begins bothering
8210 to collect garbage. The first collection occurs after the heap expands
8211 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
8212 tuning this may improve compilation speed, and has no effect on code
8215 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
8216 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
8217 with a lower bound of 4096 (four megabytes) and an upper bound of
8218 131072 (128 megabytes). If GCC is not able to calculate RAM on a
8219 particular platform, the lower bound is used. Setting this parameter
8220 very large effectively disables garbage collection. Setting this
8221 parameter and @option{ggc-min-expand} to zero causes a full collection
8222 to occur at every opportunity.
8224 @item max-reload-search-insns
8225 The maximum number of instruction reload should look backward for equivalent
8226 register. Increasing values mean more aggressive optimization, making the
8227 compile time increase with probably slightly better performance. The default
8230 @item max-cselib-memory-locations
8231 The maximum number of memory locations cselib should take into account.
8232 Increasing values mean more aggressive optimization, making the compile time
8233 increase with probably slightly better performance. The default value is 500.
8235 @item reorder-blocks-duplicate
8236 @itemx reorder-blocks-duplicate-feedback
8238 Used by basic block reordering pass to decide whether to use unconditional
8239 branch or duplicate the code on its destination. Code is duplicated when its
8240 estimated size is smaller than this value multiplied by the estimated size of
8241 unconditional jump in the hot spots of the program.
8243 The @option{reorder-block-duplicate-feedback} is used only when profile
8244 feedback is available and may be set to higher values than
8245 @option{reorder-block-duplicate} since information about the hot spots is more
8248 @item max-sched-ready-insns
8249 The maximum number of instructions ready to be issued the scheduler should
8250 consider at any given time during the first scheduling pass. Increasing
8251 values mean more thorough searches, making the compilation time increase
8252 with probably little benefit. The default value is 100.
8254 @item max-sched-region-blocks
8255 The maximum number of blocks in a region to be considered for
8256 interblock scheduling. The default value is 10.
8258 @item max-pipeline-region-blocks
8259 The maximum number of blocks in a region to be considered for
8260 pipelining in the selective scheduler. The default value is 15.
8262 @item max-sched-region-insns
8263 The maximum number of insns in a region to be considered for
8264 interblock scheduling. The default value is 100.
8266 @item max-pipeline-region-insns
8267 The maximum number of insns in a region to be considered for
8268 pipelining in the selective scheduler. The default value is 200.
8271 The minimum probability (in percents) of reaching a source block
8272 for interblock speculative scheduling. The default value is 40.
8274 @item max-sched-extend-regions-iters
8275 The maximum number of iterations through CFG to extend regions.
8276 0 - disable region extension,
8277 N - do at most N iterations.
8278 The default value is 0.
8280 @item max-sched-insn-conflict-delay
8281 The maximum conflict delay for an insn to be considered for speculative motion.
8282 The default value is 3.
8284 @item sched-spec-prob-cutoff
8285 The minimal probability of speculation success (in percents), so that
8286 speculative insn will be scheduled.
8287 The default value is 40.
8289 @item sched-mem-true-dep-cost
8290 Minimal distance (in CPU cycles) between store and load targeting same
8291 memory locations. The default value is 1.
8293 @item selsched-max-lookahead
8294 The maximum size of the lookahead window of selective scheduling. It is a
8295 depth of search for available instructions.
8296 The default value is 50.
8298 @item selsched-max-sched-times
8299 The maximum number of times that an instruction will be scheduled during
8300 selective scheduling. This is the limit on the number of iterations
8301 through which the instruction may be pipelined. The default value is 2.
8303 @item selsched-max-insns-to-rename
8304 The maximum number of best instructions in the ready list that are considered
8305 for renaming in the selective scheduler. The default value is 2.
8307 @item max-last-value-rtl
8308 The maximum size measured as number of RTLs that can be recorded in an expression
8309 in combiner for a pseudo register as last known value of that register. The default
8312 @item integer-share-limit
8313 Small integer constants can use a shared data structure, reducing the
8314 compiler's memory usage and increasing its speed. This sets the maximum
8315 value of a shared integer constant. The default value is 256.
8317 @item min-virtual-mappings
8318 Specifies the minimum number of virtual mappings in the incremental
8319 SSA updater that should be registered to trigger the virtual mappings
8320 heuristic defined by virtual-mappings-ratio. The default value is
8323 @item virtual-mappings-ratio
8324 If the number of virtual mappings is virtual-mappings-ratio bigger
8325 than the number of virtual symbols to be updated, then the incremental
8326 SSA updater switches to a full update for those symbols. The default
8329 @item ssp-buffer-size
8330 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
8331 protection when @option{-fstack-protection} is used.
8333 @item max-jump-thread-duplication-stmts
8334 Maximum number of statements allowed in a block that needs to be
8335 duplicated when threading jumps.
8337 @item max-fields-for-field-sensitive
8338 Maximum number of fields in a structure we will treat in
8339 a field sensitive manner during pointer analysis. The default is zero
8340 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
8342 @item prefetch-latency
8343 Estimate on average number of instructions that are executed before
8344 prefetch finishes. The distance we prefetch ahead is proportional
8345 to this constant. Increasing this number may also lead to less
8346 streams being prefetched (see @option{simultaneous-prefetches}).
8348 @item simultaneous-prefetches
8349 Maximum number of prefetches that can run at the same time.
8351 @item l1-cache-line-size
8352 The size of cache line in L1 cache, in bytes.
8355 The size of L1 cache, in kilobytes.
8358 The size of L2 cache, in kilobytes.
8360 @item min-insn-to-prefetch-ratio
8361 The minimum ratio between the number of instructions and the
8362 number of prefetches to enable prefetching in a loop with an
8365 @item prefetch-min-insn-to-mem-ratio
8366 The minimum ratio between the number of instructions and the
8367 number of memory references to enable prefetching in a loop.
8369 @item use-canonical-types
8370 Whether the compiler should use the ``canonical'' type system. By
8371 default, this should always be 1, which uses a more efficient internal
8372 mechanism for comparing types in C++ and Objective-C++. However, if
8373 bugs in the canonical type system are causing compilation failures,
8374 set this value to 0 to disable canonical types.
8376 @item switch-conversion-max-branch-ratio
8377 Switch initialization conversion will refuse to create arrays that are
8378 bigger than @option{switch-conversion-max-branch-ratio} times the number of
8379 branches in the switch.
8381 @item max-partial-antic-length
8382 Maximum length of the partial antic set computed during the tree
8383 partial redundancy elimination optimization (@option{-ftree-pre}) when
8384 optimizing at @option{-O3} and above. For some sorts of source code
8385 the enhanced partial redundancy elimination optimization can run away,
8386 consuming all of the memory available on the host machine. This
8387 parameter sets a limit on the length of the sets that are computed,
8388 which prevents the runaway behavior. Setting a value of 0 for
8389 this parameter will allow an unlimited set length.
8391 @item sccvn-max-scc-size
8392 Maximum size of a strongly connected component (SCC) during SCCVN
8393 processing. If this limit is hit, SCCVN processing for the whole
8394 function will not be done and optimizations depending on it will
8395 be disabled. The default maximum SCC size is 10000.
8397 @item ira-max-loops-num
8398 IRA uses a regional register allocation by default. If a function
8399 contains loops more than number given by the parameter, only at most
8400 given number of the most frequently executed loops will form regions
8401 for the regional register allocation. The default value of the
8404 @item ira-max-conflict-table-size
8405 Although IRA uses a sophisticated algorithm of compression conflict
8406 table, the table can be still big for huge functions. If the conflict
8407 table for a function could be more than size in MB given by the
8408 parameter, the conflict table is not built and faster, simpler, and
8409 lower quality register allocation algorithm will be used. The
8410 algorithm do not use pseudo-register conflicts. The default value of
8411 the parameter is 2000.
8413 @item ira-loop-reserved-regs
8414 IRA can be used to evaluate more accurate register pressure in loops
8415 for decision to move loop invariants (see @option{-O3}). The number
8416 of available registers reserved for some other purposes is described
8417 by this parameter. The default value of the parameter is 2 which is
8418 minimal number of registers needed for execution of typical
8419 instruction. This value is the best found from numerous experiments.
8421 @item loop-invariant-max-bbs-in-loop
8422 Loop invariant motion can be very expensive, both in compile time and
8423 in amount of needed compile time memory, with very large loops. Loops
8424 with more basic blocks than this parameter won't have loop invariant
8425 motion optimization performed on them. The default value of the
8426 parameter is 1000 for -O1 and 10000 for -O2 and above.
8428 @item min-nondebug-insn-uid
8429 Use uids starting at this parameter for nondebug insns. The range below
8430 the parameter is reserved exclusively for debug insns created by
8431 @option{-fvar-tracking-assignments}, but debug insns may get
8432 (non-overlapping) uids above it if the reserved range is exhausted.
8434 @item ipa-sra-ptr-growth-factor
8435 IPA-SRA will replace a pointer to an aggregate with one or more new
8436 parameters only when their cumulative size is less or equal to
8437 @option{ipa-sra-ptr-growth-factor} times the size of the original
8443 @node Preprocessor Options
8444 @section Options Controlling the Preprocessor
8445 @cindex preprocessor options
8446 @cindex options, preprocessor
8448 These options control the C preprocessor, which is run on each C source
8449 file before actual compilation.
8451 If you use the @option{-E} option, nothing is done except preprocessing.
8452 Some of these options make sense only together with @option{-E} because
8453 they cause the preprocessor output to be unsuitable for actual
8457 @item -Wp,@var{option}
8459 You can use @option{-Wp,@var{option}} to bypass the compiler driver
8460 and pass @var{option} directly through to the preprocessor. If
8461 @var{option} contains commas, it is split into multiple options at the
8462 commas. However, many options are modified, translated or interpreted
8463 by the compiler driver before being passed to the preprocessor, and
8464 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
8465 interface is undocumented and subject to change, so whenever possible
8466 you should avoid using @option{-Wp} and let the driver handle the
8469 @item -Xpreprocessor @var{option}
8470 @opindex Xpreprocessor
8471 Pass @var{option} as an option to the preprocessor. You can use this to
8472 supply system-specific preprocessor options which GCC does not know how to
8475 If you want to pass an option that takes an argument, you must use
8476 @option{-Xpreprocessor} twice, once for the option and once for the argument.
8479 @include cppopts.texi
8481 @node Assembler Options
8482 @section Passing Options to the Assembler
8484 @c prevent bad page break with this line
8485 You can pass options to the assembler.
8488 @item -Wa,@var{option}
8490 Pass @var{option} as an option to the assembler. If @var{option}
8491 contains commas, it is split into multiple options at the commas.
8493 @item -Xassembler @var{option}
8495 Pass @var{option} as an option to the assembler. You can use this to
8496 supply system-specific assembler options which GCC does not know how to
8499 If you want to pass an option that takes an argument, you must use
8500 @option{-Xassembler} twice, once for the option and once for the argument.
8505 @section Options for Linking
8506 @cindex link options
8507 @cindex options, linking
8509 These options come into play when the compiler links object files into
8510 an executable output file. They are meaningless if the compiler is
8511 not doing a link step.
8515 @item @var{object-file-name}
8516 A file name that does not end in a special recognized suffix is
8517 considered to name an object file or library. (Object files are
8518 distinguished from libraries by the linker according to the file
8519 contents.) If linking is done, these object files are used as input
8528 If any of these options is used, then the linker is not run, and
8529 object file names should not be used as arguments. @xref{Overall
8533 @item -l@var{library}
8534 @itemx -l @var{library}
8536 Search the library named @var{library} when linking. (The second
8537 alternative with the library as a separate argument is only for
8538 POSIX compliance and is not recommended.)
8540 It makes a difference where in the command you write this option; the
8541 linker searches and processes libraries and object files in the order they
8542 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
8543 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
8544 to functions in @samp{z}, those functions may not be loaded.
8546 The linker searches a standard list of directories for the library,
8547 which is actually a file named @file{lib@var{library}.a}. The linker
8548 then uses this file as if it had been specified precisely by name.
8550 The directories searched include several standard system directories
8551 plus any that you specify with @option{-L}.
8553 Normally the files found this way are library files---archive files
8554 whose members are object files. The linker handles an archive file by
8555 scanning through it for members which define symbols that have so far
8556 been referenced but not defined. But if the file that is found is an
8557 ordinary object file, it is linked in the usual fashion. The only
8558 difference between using an @option{-l} option and specifying a file name
8559 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
8560 and searches several directories.
8564 You need this special case of the @option{-l} option in order to
8565 link an Objective-C or Objective-C++ program.
8568 @opindex nostartfiles
8569 Do not use the standard system startup files when linking.
8570 The standard system libraries are used normally, unless @option{-nostdlib}
8571 or @option{-nodefaultlibs} is used.
8573 @item -nodefaultlibs
8574 @opindex nodefaultlibs
8575 Do not use the standard system libraries when linking.
8576 Only the libraries you specify will be passed to the linker, options
8577 specifying linkage of the system libraries, such as @code{-static-libgcc}
8578 or @code{-shared-libgcc}, will be ignored.
8579 The standard startup files are used normally, unless @option{-nostartfiles}
8580 is used. The compiler may generate calls to @code{memcmp},
8581 @code{memset}, @code{memcpy} and @code{memmove}.
8582 These entries are usually resolved by entries in
8583 libc. These entry points should be supplied through some other
8584 mechanism when this option is specified.
8588 Do not use the standard system startup files or libraries when linking.
8589 No startup files and only the libraries you specify will be passed to
8590 the linker, options specifying linkage of the system libraries, such as
8591 @code{-static-libgcc} or @code{-shared-libgcc}, will be ignored.
8592 The compiler may generate calls to @code{memcmp}, @code{memset},
8593 @code{memcpy} and @code{memmove}.
8594 These entries are usually resolved by entries in
8595 libc. These entry points should be supplied through some other
8596 mechanism when this option is specified.
8598 @cindex @option{-lgcc}, use with @option{-nostdlib}
8599 @cindex @option{-nostdlib} and unresolved references
8600 @cindex unresolved references and @option{-nostdlib}
8601 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
8602 @cindex @option{-nodefaultlibs} and unresolved references
8603 @cindex unresolved references and @option{-nodefaultlibs}
8604 One of the standard libraries bypassed by @option{-nostdlib} and
8605 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
8606 that GCC uses to overcome shortcomings of particular machines, or special
8607 needs for some languages.
8608 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
8609 Collection (GCC) Internals},
8610 for more discussion of @file{libgcc.a}.)
8611 In most cases, you need @file{libgcc.a} even when you want to avoid
8612 other standard libraries. In other words, when you specify @option{-nostdlib}
8613 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
8614 This ensures that you have no unresolved references to internal GCC
8615 library subroutines. (For example, @samp{__main}, used to ensure C++
8616 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
8617 GNU Compiler Collection (GCC) Internals}.)
8621 Produce a position independent executable on targets which support it.
8622 For predictable results, you must also specify the same set of options
8623 that were used to generate code (@option{-fpie}, @option{-fPIE},
8624 or model suboptions) when you specify this option.
8628 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
8629 that support it. This instructs the linker to add all symbols, not
8630 only used ones, to the dynamic symbol table. This option is needed
8631 for some uses of @code{dlopen} or to allow obtaining backtraces
8632 from within a program.
8636 Remove all symbol table and relocation information from the executable.
8640 On systems that support dynamic linking, this prevents linking with the shared
8641 libraries. On other systems, this option has no effect.
8645 Produce a shared object which can then be linked with other objects to
8646 form an executable. Not all systems support this option. For predictable
8647 results, you must also specify the same set of options that were used to
8648 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
8649 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
8650 needs to build supplementary stub code for constructors to work. On
8651 multi-libbed systems, @samp{gcc -shared} must select the correct support
8652 libraries to link against. Failing to supply the correct flags may lead
8653 to subtle defects. Supplying them in cases where they are not necessary
8656 @item -shared-libgcc
8657 @itemx -static-libgcc
8658 @opindex shared-libgcc
8659 @opindex static-libgcc
8660 On systems that provide @file{libgcc} as a shared library, these options
8661 force the use of either the shared or static version respectively.
8662 If no shared version of @file{libgcc} was built when the compiler was
8663 configured, these options have no effect.
8665 There are several situations in which an application should use the
8666 shared @file{libgcc} instead of the static version. The most common
8667 of these is when the application wishes to throw and catch exceptions
8668 across different shared libraries. In that case, each of the libraries
8669 as well as the application itself should use the shared @file{libgcc}.
8671 Therefore, the G++ and GCJ drivers automatically add
8672 @option{-shared-libgcc} whenever you build a shared library or a main
8673 executable, because C++ and Java programs typically use exceptions, so
8674 this is the right thing to do.
8676 If, instead, you use the GCC driver to create shared libraries, you may
8677 find that they will not always be linked with the shared @file{libgcc}.
8678 If GCC finds, at its configuration time, that you have a non-GNU linker
8679 or a GNU linker that does not support option @option{--eh-frame-hdr},
8680 it will link the shared version of @file{libgcc} into shared libraries
8681 by default. Otherwise, it will take advantage of the linker and optimize
8682 away the linking with the shared version of @file{libgcc}, linking with
8683 the static version of libgcc by default. This allows exceptions to
8684 propagate through such shared libraries, without incurring relocation
8685 costs at library load time.
8687 However, if a library or main executable is supposed to throw or catch
8688 exceptions, you must link it using the G++ or GCJ driver, as appropriate
8689 for the languages used in the program, or using the option
8690 @option{-shared-libgcc}, such that it is linked with the shared
8693 @item -static-libstdc++
8694 When the @command{g++} program is used to link a C++ program, it will
8695 normally automatically link against @option{libstdc++}. If
8696 @file{libstdc++} is available as a shared library, and the
8697 @option{-static} option is not used, then this will link against the
8698 shared version of @file{libstdc++}. That is normally fine. However, it
8699 is sometimes useful to freeze the version of @file{libstdc++} used by
8700 the program without going all the way to a fully static link. The
8701 @option{-static-libstdc++} option directs the @command{g++} driver to
8702 link @file{libstdc++} statically, without necessarily linking other
8703 libraries statically.
8707 Bind references to global symbols when building a shared object. Warn
8708 about any unresolved references (unless overridden by the link editor
8709 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
8712 @item -T @var{script}
8714 @cindex linker script
8715 Use @var{script} as the linker script. This option is supported by most
8716 systems using the GNU linker. On some targets, such as bare-board
8717 targets without an operating system, the @option{-T} option may be required
8718 when linking to avoid references to undefined symbols.
8720 @item -Xlinker @var{option}
8722 Pass @var{option} as an option to the linker. You can use this to
8723 supply system-specific linker options which GCC does not know how to
8726 If you want to pass an option that takes a separate argument, you must use
8727 @option{-Xlinker} twice, once for the option and once for the argument.
8728 For example, to pass @option{-assert definitions}, you must write
8729 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
8730 @option{-Xlinker "-assert definitions"}, because this passes the entire
8731 string as a single argument, which is not what the linker expects.
8733 When using the GNU linker, it is usually more convenient to pass
8734 arguments to linker options using the @option{@var{option}=@var{value}}
8735 syntax than as separate arguments. For example, you can specify
8736 @samp{-Xlinker -Map=output.map} rather than
8737 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
8738 this syntax for command-line options.
8740 @item -Wl,@var{option}
8742 Pass @var{option} as an option to the linker. If @var{option} contains
8743 commas, it is split into multiple options at the commas. You can use this
8744 syntax to pass an argument to the option.
8745 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
8746 linker. When using the GNU linker, you can also get the same effect with
8747 @samp{-Wl,-Map=output.map}.
8749 @item -u @var{symbol}
8751 Pretend the symbol @var{symbol} is undefined, to force linking of
8752 library modules to define it. You can use @option{-u} multiple times with
8753 different symbols to force loading of additional library modules.
8756 @node Directory Options
8757 @section Options for Directory Search
8758 @cindex directory options
8759 @cindex options, directory search
8762 These options specify directories to search for header files, for
8763 libraries and for parts of the compiler:
8768 Add the directory @var{dir} to the head of the list of directories to be
8769 searched for header files. This can be used to override a system header
8770 file, substituting your own version, since these directories are
8771 searched before the system header file directories. However, you should
8772 not use this option to add directories that contain vendor-supplied
8773 system header files (use @option{-isystem} for that). If you use more than
8774 one @option{-I} option, the directories are scanned in left-to-right
8775 order; the standard system directories come after.
8777 If a standard system include directory, or a directory specified with
8778 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
8779 option will be ignored. The directory will still be searched but as a
8780 system directory at its normal position in the system include chain.
8781 This is to ensure that GCC's procedure to fix buggy system headers and
8782 the ordering for the include_next directive are not inadvertently changed.
8783 If you really need to change the search order for system directories,
8784 use the @option{-nostdinc} and/or @option{-isystem} options.
8786 @item -iquote@var{dir}
8788 Add the directory @var{dir} to the head of the list of directories to
8789 be searched for header files only for the case of @samp{#include
8790 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
8791 otherwise just like @option{-I}.
8795 Add directory @var{dir} to the list of directories to be searched
8798 @item -B@var{prefix}
8800 This option specifies where to find the executables, libraries,
8801 include files, and data files of the compiler itself.
8803 The compiler driver program runs one or more of the subprograms
8804 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
8805 @var{prefix} as a prefix for each program it tries to run, both with and
8806 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
8808 For each subprogram to be run, the compiler driver first tries the
8809 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
8810 was not specified, the driver tries two standard prefixes, which are
8811 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
8812 those results in a file name that is found, the unmodified program
8813 name is searched for using the directories specified in your
8814 @env{PATH} environment variable.
8816 The compiler will check to see if the path provided by the @option{-B}
8817 refers to a directory, and if necessary it will add a directory
8818 separator character at the end of the path.
8820 @option{-B} prefixes that effectively specify directory names also apply
8821 to libraries in the linker, because the compiler translates these
8822 options into @option{-L} options for the linker. They also apply to
8823 includes files in the preprocessor, because the compiler translates these
8824 options into @option{-isystem} options for the preprocessor. In this case,
8825 the compiler appends @samp{include} to the prefix.
8827 The run-time support file @file{libgcc.a} can also be searched for using
8828 the @option{-B} prefix, if needed. If it is not found there, the two
8829 standard prefixes above are tried, and that is all. The file is left
8830 out of the link if it is not found by those means.
8832 Another way to specify a prefix much like the @option{-B} prefix is to use
8833 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
8836 As a special kludge, if the path provided by @option{-B} is
8837 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
8838 9, then it will be replaced by @file{[dir/]include}. This is to help
8839 with boot-strapping the compiler.
8841 @item -specs=@var{file}
8843 Process @var{file} after the compiler reads in the standard @file{specs}
8844 file, in order to override the defaults that the @file{gcc} driver
8845 program uses when determining what switches to pass to @file{cc1},
8846 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
8847 @option{-specs=@var{file}} can be specified on the command line, and they
8848 are processed in order, from left to right.
8850 @item --sysroot=@var{dir}
8852 Use @var{dir} as the logical root directory for headers and libraries.
8853 For example, if the compiler would normally search for headers in
8854 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
8855 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
8857 If you use both this option and the @option{-isysroot} option, then
8858 the @option{--sysroot} option will apply to libraries, but the
8859 @option{-isysroot} option will apply to header files.
8861 The GNU linker (beginning with version 2.16) has the necessary support
8862 for this option. If your linker does not support this option, the
8863 header file aspect of @option{--sysroot} will still work, but the
8864 library aspect will not.
8868 This option has been deprecated. Please use @option{-iquote} instead for
8869 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
8870 Any directories you specify with @option{-I} options before the @option{-I-}
8871 option are searched only for the case of @samp{#include "@var{file}"};
8872 they are not searched for @samp{#include <@var{file}>}.
8874 If additional directories are specified with @option{-I} options after
8875 the @option{-I-}, these directories are searched for all @samp{#include}
8876 directives. (Ordinarily @emph{all} @option{-I} directories are used
8879 In addition, the @option{-I-} option inhibits the use of the current
8880 directory (where the current input file came from) as the first search
8881 directory for @samp{#include "@var{file}"}. There is no way to
8882 override this effect of @option{-I-}. With @option{-I.} you can specify
8883 searching the directory which was current when the compiler was
8884 invoked. That is not exactly the same as what the preprocessor does
8885 by default, but it is often satisfactory.
8887 @option{-I-} does not inhibit the use of the standard system directories
8888 for header files. Thus, @option{-I-} and @option{-nostdinc} are
8895 @section Specifying subprocesses and the switches to pass to them
8898 @command{gcc} is a driver program. It performs its job by invoking a
8899 sequence of other programs to do the work of compiling, assembling and
8900 linking. GCC interprets its command-line parameters and uses these to
8901 deduce which programs it should invoke, and which command-line options
8902 it ought to place on their command lines. This behavior is controlled
8903 by @dfn{spec strings}. In most cases there is one spec string for each
8904 program that GCC can invoke, but a few programs have multiple spec
8905 strings to control their behavior. The spec strings built into GCC can
8906 be overridden by using the @option{-specs=} command-line switch to specify
8909 @dfn{Spec files} are plaintext files that are used to construct spec
8910 strings. They consist of a sequence of directives separated by blank
8911 lines. The type of directive is determined by the first non-whitespace
8912 character on the line and it can be one of the following:
8915 @item %@var{command}
8916 Issues a @var{command} to the spec file processor. The commands that can
8920 @item %include <@var{file}>
8922 Search for @var{file} and insert its text at the current point in the
8925 @item %include_noerr <@var{file}>
8926 @cindex %include_noerr
8927 Just like @samp{%include}, but do not generate an error message if the include
8928 file cannot be found.
8930 @item %rename @var{old_name} @var{new_name}
8932 Rename the spec string @var{old_name} to @var{new_name}.
8936 @item *[@var{spec_name}]:
8937 This tells the compiler to create, override or delete the named spec
8938 string. All lines after this directive up to the next directive or
8939 blank line are considered to be the text for the spec string. If this
8940 results in an empty string then the spec will be deleted. (Or, if the
8941 spec did not exist, then nothing will happened.) Otherwise, if the spec
8942 does not currently exist a new spec will be created. If the spec does
8943 exist then its contents will be overridden by the text of this
8944 directive, unless the first character of that text is the @samp{+}
8945 character, in which case the text will be appended to the spec.
8947 @item [@var{suffix}]:
8948 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
8949 and up to the next directive or blank line are considered to make up the
8950 spec string for the indicated suffix. When the compiler encounters an
8951 input file with the named suffix, it will processes the spec string in
8952 order to work out how to compile that file. For example:
8959 This says that any input file whose name ends in @samp{.ZZ} should be
8960 passed to the program @samp{z-compile}, which should be invoked with the
8961 command-line switch @option{-input} and with the result of performing the
8962 @samp{%i} substitution. (See below.)
8964 As an alternative to providing a spec string, the text that follows a
8965 suffix directive can be one of the following:
8968 @item @@@var{language}
8969 This says that the suffix is an alias for a known @var{language}. This is
8970 similar to using the @option{-x} command-line switch to GCC to specify a
8971 language explicitly. For example:
8978 Says that .ZZ files are, in fact, C++ source files.
8981 This causes an error messages saying:
8984 @var{name} compiler not installed on this system.
8988 GCC already has an extensive list of suffixes built into it.
8989 This directive will add an entry to the end of the list of suffixes, but
8990 since the list is searched from the end backwards, it is effectively
8991 possible to override earlier entries using this technique.
8995 GCC has the following spec strings built into it. Spec files can
8996 override these strings or create their own. Note that individual
8997 targets can also add their own spec strings to this list.
9000 asm Options to pass to the assembler
9001 asm_final Options to pass to the assembler post-processor
9002 cpp Options to pass to the C preprocessor
9003 cc1 Options to pass to the C compiler
9004 cc1plus Options to pass to the C++ compiler
9005 endfile Object files to include at the end of the link
9006 link Options to pass to the linker
9007 lib Libraries to include on the command line to the linker
9008 libgcc Decides which GCC support library to pass to the linker
9009 linker Sets the name of the linker
9010 predefines Defines to be passed to the C preprocessor
9011 signed_char Defines to pass to CPP to say whether @code{char} is signed
9013 startfile Object files to include at the start of the link
9016 Here is a small example of a spec file:
9022 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
9025 This example renames the spec called @samp{lib} to @samp{old_lib} and
9026 then overrides the previous definition of @samp{lib} with a new one.
9027 The new definition adds in some extra command-line options before
9028 including the text of the old definition.
9030 @dfn{Spec strings} are a list of command-line options to be passed to their
9031 corresponding program. In addition, the spec strings can contain
9032 @samp{%}-prefixed sequences to substitute variable text or to
9033 conditionally insert text into the command line. Using these constructs
9034 it is possible to generate quite complex command lines.
9036 Here is a table of all defined @samp{%}-sequences for spec
9037 strings. Note that spaces are not generated automatically around the
9038 results of expanding these sequences. Therefore you can concatenate them
9039 together or combine them with constant text in a single argument.
9043 Substitute one @samp{%} into the program name or argument.
9046 Substitute the name of the input file being processed.
9049 Substitute the basename of the input file being processed.
9050 This is the substring up to (and not including) the last period
9051 and not including the directory.
9054 This is the same as @samp{%b}, but include the file suffix (text after
9058 Marks the argument containing or following the @samp{%d} as a
9059 temporary file name, so that that file will be deleted if GCC exits
9060 successfully. Unlike @samp{%g}, this contributes no text to the
9063 @item %g@var{suffix}
9064 Substitute a file name that has suffix @var{suffix} and is chosen
9065 once per compilation, and mark the argument in the same way as
9066 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
9067 name is now chosen in a way that is hard to predict even when previously
9068 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
9069 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
9070 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
9071 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
9072 was simply substituted with a file name chosen once per compilation,
9073 without regard to any appended suffix (which was therefore treated
9074 just like ordinary text), making such attacks more likely to succeed.
9076 @item %u@var{suffix}
9077 Like @samp{%g}, but generates a new temporary file name even if
9078 @samp{%u@var{suffix}} was already seen.
9080 @item %U@var{suffix}
9081 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
9082 new one if there is no such last file name. In the absence of any
9083 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
9084 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
9085 would involve the generation of two distinct file names, one
9086 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
9087 simply substituted with a file name chosen for the previous @samp{%u},
9088 without regard to any appended suffix.
9090 @item %j@var{suffix}
9091 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
9092 writable, and if save-temps is off; otherwise, substitute the name
9093 of a temporary file, just like @samp{%u}. This temporary file is not
9094 meant for communication between processes, but rather as a junk
9097 @item %|@var{suffix}
9098 @itemx %m@var{suffix}
9099 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
9100 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
9101 all. These are the two most common ways to instruct a program that it
9102 should read from standard input or write to standard output. If you
9103 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
9104 construct: see for example @file{f/lang-specs.h}.
9106 @item %.@var{SUFFIX}
9107 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
9108 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
9109 terminated by the next space or %.
9112 Marks the argument containing or following the @samp{%w} as the
9113 designated output file of this compilation. This puts the argument
9114 into the sequence of arguments that @samp{%o} will substitute later.
9117 Substitutes the names of all the output files, with spaces
9118 automatically placed around them. You should write spaces
9119 around the @samp{%o} as well or the results are undefined.
9120 @samp{%o} is for use in the specs for running the linker.
9121 Input files whose names have no recognized suffix are not compiled
9122 at all, but they are included among the output files, so they will
9126 Substitutes the suffix for object files. Note that this is
9127 handled specially when it immediately follows @samp{%g, %u, or %U},
9128 because of the need for those to form complete file names. The
9129 handling is such that @samp{%O} is treated exactly as if it had already
9130 been substituted, except that @samp{%g, %u, and %U} do not currently
9131 support additional @var{suffix} characters following @samp{%O} as they would
9132 following, for example, @samp{.o}.
9135 Substitutes the standard macro predefinitions for the
9136 current target machine. Use this when running @code{cpp}.
9139 Like @samp{%p}, but puts @samp{__} before and after the name of each
9140 predefined macro, except for macros that start with @samp{__} or with
9141 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
9145 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
9146 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
9147 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
9148 and @option{-imultilib} as necessary.
9151 Current argument is the name of a library or startup file of some sort.
9152 Search for that file in a standard list of directories and substitute
9153 the full name found. The current working directory is included in the
9154 list of directories scanned.
9157 Current argument is the name of a linker script. Search for that file
9158 in the current list of directories to scan for libraries. If the file
9159 is located insert a @option{--script} option into the command line
9160 followed by the full path name found. If the file is not found then
9161 generate an error message. Note: the current working directory is not
9165 Print @var{str} as an error message. @var{str} is terminated by a newline.
9166 Use this when inconsistent options are detected.
9169 Substitute the contents of spec string @var{name} at this point.
9172 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
9174 @item %x@{@var{option}@}
9175 Accumulate an option for @samp{%X}.
9178 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
9182 Output the accumulated assembler options specified by @option{-Wa}.
9185 Output the accumulated preprocessor options specified by @option{-Wp}.
9188 Process the @code{asm} spec. This is used to compute the
9189 switches to be passed to the assembler.
9192 Process the @code{asm_final} spec. This is a spec string for
9193 passing switches to an assembler post-processor, if such a program is
9197 Process the @code{link} spec. This is the spec for computing the
9198 command line passed to the linker. Typically it will make use of the
9199 @samp{%L %G %S %D and %E} sequences.
9202 Dump out a @option{-L} option for each directory that GCC believes might
9203 contain startup files. If the target supports multilibs then the
9204 current multilib directory will be prepended to each of these paths.
9207 Process the @code{lib} spec. This is a spec string for deciding which
9208 libraries should be included on the command line to the linker.
9211 Process the @code{libgcc} spec. This is a spec string for deciding
9212 which GCC support library should be included on the command line to the linker.
9215 Process the @code{startfile} spec. This is a spec for deciding which
9216 object files should be the first ones passed to the linker. Typically
9217 this might be a file named @file{crt0.o}.
9220 Process the @code{endfile} spec. This is a spec string that specifies
9221 the last object files that will be passed to the linker.
9224 Process the @code{cpp} spec. This is used to construct the arguments
9225 to be passed to the C preprocessor.
9228 Process the @code{cc1} spec. This is used to construct the options to be
9229 passed to the actual C compiler (@samp{cc1}).
9232 Process the @code{cc1plus} spec. This is used to construct the options to be
9233 passed to the actual C++ compiler (@samp{cc1plus}).
9236 Substitute the variable part of a matched option. See below.
9237 Note that each comma in the substituted string is replaced by
9241 Remove all occurrences of @code{-S} from the command line. Note---this
9242 command is position dependent. @samp{%} commands in the spec string
9243 before this one will see @code{-S}, @samp{%} commands in the spec string
9244 after this one will not.
9246 @item %:@var{function}(@var{args})
9247 Call the named function @var{function}, passing it @var{args}.
9248 @var{args} is first processed as a nested spec string, then split
9249 into an argument vector in the usual fashion. The function returns
9250 a string which is processed as if it had appeared literally as part
9251 of the current spec.
9253 The following built-in spec functions are provided:
9257 The @code{getenv} spec function takes two arguments: an environment
9258 variable name and a string. If the environment variable is not
9259 defined, a fatal error is issued. Otherwise, the return value is the
9260 value of the environment variable concatenated with the string. For
9261 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
9264 %:getenv(TOPDIR /include)
9267 expands to @file{/path/to/top/include}.
9269 @item @code{if-exists}
9270 The @code{if-exists} spec function takes one argument, an absolute
9271 pathname to a file. If the file exists, @code{if-exists} returns the
9272 pathname. Here is a small example of its usage:
9276 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
9279 @item @code{if-exists-else}
9280 The @code{if-exists-else} spec function is similar to the @code{if-exists}
9281 spec function, except that it takes two arguments. The first argument is
9282 an absolute pathname to a file. If the file exists, @code{if-exists-else}
9283 returns the pathname. If it does not exist, it returns the second argument.
9284 This way, @code{if-exists-else} can be used to select one file or another,
9285 based on the existence of the first. Here is a small example of its usage:
9289 crt0%O%s %:if-exists(crti%O%s) \
9290 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
9293 @item @code{replace-outfile}
9294 The @code{replace-outfile} spec function takes two arguments. It looks for the
9295 first argument in the outfiles array and replaces it with the second argument. Here
9296 is a small example of its usage:
9299 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
9302 @item @code{print-asm-header}
9303 The @code{print-asm-header} function takes no arguments and simply
9304 prints a banner like:
9310 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
9313 It is used to separate compiler options from assembler options
9314 in the @option{--target-help} output.
9318 Substitutes the @code{-S} switch, if that switch was given to GCC@.
9319 If that switch was not specified, this substitutes nothing. Note that
9320 the leading dash is omitted when specifying this option, and it is
9321 automatically inserted if the substitution is performed. Thus the spec
9322 string @samp{%@{foo@}} would match the command-line option @option{-foo}
9323 and would output the command line option @option{-foo}.
9325 @item %W@{@code{S}@}
9326 Like %@{@code{S}@} but mark last argument supplied within as a file to be
9329 @item %@{@code{S}*@}
9330 Substitutes all the switches specified to GCC whose names start
9331 with @code{-S}, but which also take an argument. This is used for
9332 switches like @option{-o}, @option{-D}, @option{-I}, etc.
9333 GCC considers @option{-o foo} as being
9334 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
9335 text, including the space. Thus two arguments would be generated.
9337 @item %@{@code{S}*&@code{T}*@}
9338 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
9339 (the order of @code{S} and @code{T} in the spec is not significant).
9340 There can be any number of ampersand-separated variables; for each the
9341 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
9343 @item %@{@code{S}:@code{X}@}
9344 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
9346 @item %@{!@code{S}:@code{X}@}
9347 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
9349 @item %@{@code{S}*:@code{X}@}
9350 Substitutes @code{X} if one or more switches whose names start with
9351 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
9352 once, no matter how many such switches appeared. However, if @code{%*}
9353 appears somewhere in @code{X}, then @code{X} will be substituted once
9354 for each matching switch, with the @code{%*} replaced by the part of
9355 that switch that matched the @code{*}.
9357 @item %@{.@code{S}:@code{X}@}
9358 Substitutes @code{X}, if processing a file with suffix @code{S}.
9360 @item %@{!.@code{S}:@code{X}@}
9361 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
9363 @item %@{,@code{S}:@code{X}@}
9364 Substitutes @code{X}, if processing a file for language @code{S}.
9366 @item %@{!,@code{S}:@code{X}@}
9367 Substitutes @code{X}, if not processing a file for language @code{S}.
9369 @item %@{@code{S}|@code{P}:@code{X}@}
9370 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
9371 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
9372 @code{*} sequences as well, although they have a stronger binding than
9373 the @samp{|}. If @code{%*} appears in @code{X}, all of the
9374 alternatives must be starred, and only the first matching alternative
9377 For example, a spec string like this:
9380 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
9383 will output the following command-line options from the following input
9384 command-line options:
9389 -d fred.c -foo -baz -boggle
9390 -d jim.d -bar -baz -boggle
9393 @item %@{S:X; T:Y; :D@}
9395 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
9396 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
9397 be as many clauses as you need. This may be combined with @code{.},
9398 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
9403 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
9404 construct may contain other nested @samp{%} constructs or spaces, or
9405 even newlines. They are processed as usual, as described above.
9406 Trailing white space in @code{X} is ignored. White space may also
9407 appear anywhere on the left side of the colon in these constructs,
9408 except between @code{.} or @code{*} and the corresponding word.
9410 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
9411 handled specifically in these constructs. If another value of
9412 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
9413 @option{-W} switch is found later in the command line, the earlier
9414 switch value is ignored, except with @{@code{S}*@} where @code{S} is
9415 just one letter, which passes all matching options.
9417 The character @samp{|} at the beginning of the predicate text is used to
9418 indicate that a command should be piped to the following command, but
9419 only if @option{-pipe} is specified.
9421 It is built into GCC which switches take arguments and which do not.
9422 (You might think it would be useful to generalize this to allow each
9423 compiler's spec to say which switches take arguments. But this cannot
9424 be done in a consistent fashion. GCC cannot even decide which input
9425 files have been specified without knowing which switches take arguments,
9426 and it must know which input files to compile in order to tell which
9429 GCC also knows implicitly that arguments starting in @option{-l} are to be
9430 treated as compiler output files, and passed to the linker in their
9431 proper position among the other output files.
9433 @c man begin OPTIONS
9435 @node Target Options
9436 @section Specifying Target Machine and Compiler Version
9437 @cindex target options
9438 @cindex cross compiling
9439 @cindex specifying machine version
9440 @cindex specifying compiler version and target machine
9441 @cindex compiler version, specifying
9442 @cindex target machine, specifying
9444 The usual way to run GCC is to run the executable called @file{gcc}, or
9445 @file{<machine>-gcc} when cross-compiling, or
9446 @file{<machine>-gcc-<version>} to run a version other than the one that
9447 was installed last. Sometimes this is inconvenient, so GCC provides
9448 options that will switch to another cross-compiler or version.
9451 @item -b @var{machine}
9453 The argument @var{machine} specifies the target machine for compilation.
9455 The value to use for @var{machine} is the same as was specified as the
9456 machine type when configuring GCC as a cross-compiler. For
9457 example, if a cross-compiler was configured with @samp{configure
9458 arm-elf}, meaning to compile for an arm processor with elf binaries,
9459 then you would specify @option{-b arm-elf} to run that cross compiler.
9460 Because there are other options beginning with @option{-b}, the
9461 configuration must contain a hyphen, or @option{-b} alone should be one
9462 argument followed by the configuration in the next argument.
9464 @item -V @var{version}
9466 The argument @var{version} specifies which version of GCC to run.
9467 This is useful when multiple versions are installed. For example,
9468 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
9471 The @option{-V} and @option{-b} options work by running the
9472 @file{<machine>-gcc-<version>} executable, so there's no real reason to
9473 use them if you can just run that directly.
9475 @node Submodel Options
9476 @section Hardware Models and Configurations
9477 @cindex submodel options
9478 @cindex specifying hardware config
9479 @cindex hardware models and configurations, specifying
9480 @cindex machine dependent options
9482 Earlier we discussed the standard option @option{-b} which chooses among
9483 different installed compilers for completely different target
9484 machines, such as VAX vs.@: 68000 vs.@: 80386.
9486 In addition, each of these target machine types can have its own
9487 special options, starting with @samp{-m}, to choose among various
9488 hardware models or configurations---for example, 68010 vs 68020,
9489 floating coprocessor or none. A single installed version of the
9490 compiler can compile for any model or configuration, according to the
9493 Some configurations of the compiler also support additional special
9494 options, usually for compatibility with other compilers on the same
9497 @c This list is ordered alphanumerically by subsection name.
9498 @c It should be the same order and spelling as these options are listed
9499 @c in Machine Dependent Options
9505 * Blackfin Options::
9509 * DEC Alpha Options::
9510 * DEC Alpha/VMS Options::
9513 * GNU/Linux Options::
9516 * i386 and x86-64 Options::
9517 * i386 and x86-64 Windows Options::
9519 * IA-64/VMS Options::
9530 * picoChip Options::
9532 * RS/6000 and PowerPC Options::
9533 * S/390 and zSeries Options::
9538 * System V Options::
9543 * Xstormy16 Options::
9549 @subsection ARC Options
9552 These options are defined for ARC implementations:
9557 Compile code for little endian mode. This is the default.
9561 Compile code for big endian mode.
9564 @opindex mmangle-cpu
9565 Prepend the name of the cpu to all public symbol names.
9566 In multiple-processor systems, there are many ARC variants with different
9567 instruction and register set characteristics. This flag prevents code
9568 compiled for one cpu to be linked with code compiled for another.
9569 No facility exists for handling variants that are ``almost identical''.
9570 This is an all or nothing option.
9572 @item -mcpu=@var{cpu}
9574 Compile code for ARC variant @var{cpu}.
9575 Which variants are supported depend on the configuration.
9576 All variants support @option{-mcpu=base}, this is the default.
9578 @item -mtext=@var{text-section}
9579 @itemx -mdata=@var{data-section}
9580 @itemx -mrodata=@var{readonly-data-section}
9584 Put functions, data, and readonly data in @var{text-section},
9585 @var{data-section}, and @var{readonly-data-section} respectively
9586 by default. This can be overridden with the @code{section} attribute.
9587 @xref{Variable Attributes}.
9589 @item -mfix-cortex-m3-ldrd
9590 @opindex mfix-cortex-m3-ldrd
9591 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
9592 with overlapping destination and base registers are used. This option avoids
9593 generating these instructions. This option is enabled by default when
9594 @option{-mcpu=cortex-m3} is specified.
9599 @subsection ARM Options
9602 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
9606 @item -mabi=@var{name}
9608 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
9609 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
9612 @opindex mapcs-frame
9613 Generate a stack frame that is compliant with the ARM Procedure Call
9614 Standard for all functions, even if this is not strictly necessary for
9615 correct execution of the code. Specifying @option{-fomit-frame-pointer}
9616 with this option will cause the stack frames not to be generated for
9617 leaf functions. The default is @option{-mno-apcs-frame}.
9621 This is a synonym for @option{-mapcs-frame}.
9624 @c not currently implemented
9625 @item -mapcs-stack-check
9626 @opindex mapcs-stack-check
9627 Generate code to check the amount of stack space available upon entry to
9628 every function (that actually uses some stack space). If there is
9629 insufficient space available then either the function
9630 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
9631 called, depending upon the amount of stack space required. The run time
9632 system is required to provide these functions. The default is
9633 @option{-mno-apcs-stack-check}, since this produces smaller code.
9635 @c not currently implemented
9637 @opindex mapcs-float
9638 Pass floating point arguments using the float point registers. This is
9639 one of the variants of the APCS@. This option is recommended if the
9640 target hardware has a floating point unit or if a lot of floating point
9641 arithmetic is going to be performed by the code. The default is
9642 @option{-mno-apcs-float}, since integer only code is slightly increased in
9643 size if @option{-mapcs-float} is used.
9645 @c not currently implemented
9646 @item -mapcs-reentrant
9647 @opindex mapcs-reentrant
9648 Generate reentrant, position independent code. The default is
9649 @option{-mno-apcs-reentrant}.
9652 @item -mthumb-interwork
9653 @opindex mthumb-interwork
9654 Generate code which supports calling between the ARM and Thumb
9655 instruction sets. Without this option the two instruction sets cannot
9656 be reliably used inside one program. The default is
9657 @option{-mno-thumb-interwork}, since slightly larger code is generated
9658 when @option{-mthumb-interwork} is specified.
9660 @item -mno-sched-prolog
9661 @opindex mno-sched-prolog
9662 Prevent the reordering of instructions in the function prolog, or the
9663 merging of those instruction with the instructions in the function's
9664 body. This means that all functions will start with a recognizable set
9665 of instructions (or in fact one of a choice from a small set of
9666 different function prologues), and this information can be used to
9667 locate the start if functions inside an executable piece of code. The
9668 default is @option{-msched-prolog}.
9670 @item -mfloat-abi=@var{name}
9672 Specifies which floating-point ABI to use. Permissible values
9673 are: @samp{soft}, @samp{softfp} and @samp{hard}.
9675 Specifying @samp{soft} causes GCC to generate output containing
9676 library calls for floating-point operations.
9677 @samp{softfp} allows the generation of code using hardware floating-point
9678 instructions, but still uses the soft-float calling conventions.
9679 @samp{hard} allows generation of floating-point instructions
9680 and uses FPU-specific calling conventions.
9682 The default depends on the specific target configuration. Note that
9683 the hard-float and soft-float ABIs are not link-compatible; you must
9684 compile your entire program with the same ABI, and link with a
9685 compatible set of libraries.
9688 @opindex mhard-float
9689 Equivalent to @option{-mfloat-abi=hard}.
9692 @opindex msoft-float
9693 Equivalent to @option{-mfloat-abi=soft}.
9695 @item -mlittle-endian
9696 @opindex mlittle-endian
9697 Generate code for a processor running in little-endian mode. This is
9698 the default for all standard configurations.
9701 @opindex mbig-endian
9702 Generate code for a processor running in big-endian mode; the default is
9703 to compile code for a little-endian processor.
9705 @item -mwords-little-endian
9706 @opindex mwords-little-endian
9707 This option only applies when generating code for big-endian processors.
9708 Generate code for a little-endian word order but a big-endian byte
9709 order. That is, a byte order of the form @samp{32107654}. Note: this
9710 option should only be used if you require compatibility with code for
9711 big-endian ARM processors generated by versions of the compiler prior to
9714 @item -mcpu=@var{name}
9716 This specifies the name of the target ARM processor. GCC uses this name
9717 to determine what kind of instructions it can emit when generating
9718 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
9719 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
9720 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
9721 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
9722 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
9724 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
9725 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
9726 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
9727 @samp{strongarm1110},
9728 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
9729 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
9730 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
9731 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
9732 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
9733 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
9734 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
9735 @samp{cortex-a8}, @samp{cortex-a9},
9736 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-m3},
9739 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9741 @item -mtune=@var{name}
9743 This option is very similar to the @option{-mcpu=} option, except that
9744 instead of specifying the actual target processor type, and hence
9745 restricting which instructions can be used, it specifies that GCC should
9746 tune the performance of the code as if the target were of the type
9747 specified in this option, but still choosing the instructions that it
9748 will generate based on the cpu specified by a @option{-mcpu=} option.
9749 For some ARM implementations better performance can be obtained by using
9752 @item -march=@var{name}
9754 This specifies the name of the target ARM architecture. GCC uses this
9755 name to determine what kind of instructions it can emit when generating
9756 assembly code. This option can be used in conjunction with or instead
9757 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
9758 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
9759 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
9760 @samp{armv6}, @samp{armv6j},
9761 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
9762 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
9763 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9765 @item -mfpu=@var{name}
9766 @itemx -mfpe=@var{number}
9767 @itemx -mfp=@var{number}
9771 This specifies what floating point hardware (or hardware emulation) is
9772 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
9773 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-d16},
9774 @samp{neon}, and @samp{neon-fp16}. @option{-mfp} and @option{-mfpe}
9775 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
9776 with older versions of GCC@.
9778 If @option{-msoft-float} is specified this specifies the format of
9779 floating point values.
9781 @item -mfp16-format=@var{name}
9782 @opindex mfp16-format
9783 Specify the format of the @code{__fp16} half-precision floating-point type.
9784 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
9785 the default is @samp{none}, in which case the @code{__fp16} type is not
9786 defined. @xref{Half-Precision}, for more information.
9788 @item -mstructure-size-boundary=@var{n}
9789 @opindex mstructure-size-boundary
9790 The size of all structures and unions will be rounded up to a multiple
9791 of the number of bits set by this option. Permissible values are 8, 32
9792 and 64. The default value varies for different toolchains. For the COFF
9793 targeted toolchain the default value is 8. A value of 64 is only allowed
9794 if the underlying ABI supports it.
9796 Specifying the larger number can produce faster, more efficient code, but
9797 can also increase the size of the program. Different values are potentially
9798 incompatible. Code compiled with one value cannot necessarily expect to
9799 work with code or libraries compiled with another value, if they exchange
9800 information using structures or unions.
9802 @item -mabort-on-noreturn
9803 @opindex mabort-on-noreturn
9804 Generate a call to the function @code{abort} at the end of a
9805 @code{noreturn} function. It will be executed if the function tries to
9809 @itemx -mno-long-calls
9810 @opindex mlong-calls
9811 @opindex mno-long-calls
9812 Tells the compiler to perform function calls by first loading the
9813 address of the function into a register and then performing a subroutine
9814 call on this register. This switch is needed if the target function
9815 will lie outside of the 64 megabyte addressing range of the offset based
9816 version of subroutine call instruction.
9818 Even if this switch is enabled, not all function calls will be turned
9819 into long calls. The heuristic is that static functions, functions
9820 which have the @samp{short-call} attribute, functions that are inside
9821 the scope of a @samp{#pragma no_long_calls} directive and functions whose
9822 definitions have already been compiled within the current compilation
9823 unit, will not be turned into long calls. The exception to this rule is
9824 that weak function definitions, functions with the @samp{long-call}
9825 attribute or the @samp{section} attribute, and functions that are within
9826 the scope of a @samp{#pragma long_calls} directive, will always be
9827 turned into long calls.
9829 This feature is not enabled by default. Specifying
9830 @option{-mno-long-calls} will restore the default behavior, as will
9831 placing the function calls within the scope of a @samp{#pragma
9832 long_calls_off} directive. Note these switches have no effect on how
9833 the compiler generates code to handle function calls via function
9836 @item -msingle-pic-base
9837 @opindex msingle-pic-base
9838 Treat the register used for PIC addressing as read-only, rather than
9839 loading it in the prologue for each function. The run-time system is
9840 responsible for initializing this register with an appropriate value
9841 before execution begins.
9843 @item -mpic-register=@var{reg}
9844 @opindex mpic-register
9845 Specify the register to be used for PIC addressing. The default is R10
9846 unless stack-checking is enabled, when R9 is used.
9848 @item -mcirrus-fix-invalid-insns
9849 @opindex mcirrus-fix-invalid-insns
9850 @opindex mno-cirrus-fix-invalid-insns
9851 Insert NOPs into the instruction stream to in order to work around
9852 problems with invalid Maverick instruction combinations. This option
9853 is only valid if the @option{-mcpu=ep9312} option has been used to
9854 enable generation of instructions for the Cirrus Maverick floating
9855 point co-processor. This option is not enabled by default, since the
9856 problem is only present in older Maverick implementations. The default
9857 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
9860 @item -mpoke-function-name
9861 @opindex mpoke-function-name
9862 Write the name of each function into the text section, directly
9863 preceding the function prologue. The generated code is similar to this:
9867 .ascii "arm_poke_function_name", 0
9870 .word 0xff000000 + (t1 - t0)
9871 arm_poke_function_name
9873 stmfd sp!, @{fp, ip, lr, pc@}
9877 When performing a stack backtrace, code can inspect the value of
9878 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
9879 location @code{pc - 12} and the top 8 bits are set, then we know that
9880 there is a function name embedded immediately preceding this location
9881 and has length @code{((pc[-3]) & 0xff000000)}.
9885 Generate code for the Thumb instruction set. The default is to
9886 use the 32-bit ARM instruction set.
9887 This option automatically enables either 16-bit Thumb-1 or
9888 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
9889 and @option{-march=@var{name}} options. This option is not passed to the
9890 assembler. If you want to force assembler files to be interpreted as Thumb code,
9891 either add a @samp{.thumb} directive to the source or pass the @option{-mthumb}
9892 option directly to the assembler by prefixing it with @option{-Wa}.
9895 @opindex mtpcs-frame
9896 Generate a stack frame that is compliant with the Thumb Procedure Call
9897 Standard for all non-leaf functions. (A leaf function is one that does
9898 not call any other functions.) The default is @option{-mno-tpcs-frame}.
9900 @item -mtpcs-leaf-frame
9901 @opindex mtpcs-leaf-frame
9902 Generate a stack frame that is compliant with the Thumb Procedure Call
9903 Standard for all leaf functions. (A leaf function is one that does
9904 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
9906 @item -mcallee-super-interworking
9907 @opindex mcallee-super-interworking
9908 Gives all externally visible functions in the file being compiled an ARM
9909 instruction set header which switches to Thumb mode before executing the
9910 rest of the function. This allows these functions to be called from
9911 non-interworking code. This option is not valid in AAPCS configurations
9912 because interworking is enabled by default.
9914 @item -mcaller-super-interworking
9915 @opindex mcaller-super-interworking
9916 Allows calls via function pointers (including virtual functions) to
9917 execute correctly regardless of whether the target code has been
9918 compiled for interworking or not. There is a small overhead in the cost
9919 of executing a function pointer if this option is enabled. This option
9920 is not valid in AAPCS configurations because interworking is enabled
9923 @item -mtp=@var{name}
9925 Specify the access model for the thread local storage pointer. The valid
9926 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
9927 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
9928 (supported in the arm6k architecture), and @option{auto}, which uses the
9929 best available method for the selected processor. The default setting is
9932 @item -mword-relocations
9933 @opindex mword-relocations
9934 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
9935 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
9936 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
9942 @subsection AVR Options
9945 These options are defined for AVR implementations:
9948 @item -mmcu=@var{mcu}
9950 Specify ATMEL AVR instruction set or MCU type.
9952 Instruction set avr1 is for the minimal AVR core, not supported by the C
9953 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
9954 attiny11, attiny12, attiny15, attiny28).
9956 Instruction set avr2 (default) is for the classic AVR core with up to
9957 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
9958 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
9959 at90c8534, at90s8535).
9961 Instruction set avr3 is for the classic AVR core with up to 128K program
9962 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
9964 Instruction set avr4 is for the enhanced AVR core with up to 8K program
9965 memory space (MCU types: atmega8, atmega83, atmega85).
9967 Instruction set avr5 is for the enhanced AVR core with up to 128K program
9968 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
9969 atmega64, atmega128, at43usb355, at94k).
9971 @item -mno-interrupts
9972 @opindex mno-interrupts
9973 Generated code is not compatible with hardware interrupts.
9974 Code size will be smaller.
9976 @item -mcall-prologues
9977 @opindex mcall-prologues
9978 Functions prologues/epilogues expanded as call to appropriate
9979 subroutines. Code size will be smaller.
9982 @opindex mtiny-stack
9983 Change only the low 8 bits of the stack pointer.
9987 Assume int to be 8 bit integer. This affects the sizes of all types: A
9988 char will be 1 byte, an int will be 1 byte, a long will be 2 bytes
9989 and long long will be 4 bytes. Please note that this option does not
9990 comply to the C standards, but it will provide you with smaller code
9994 @node Blackfin Options
9995 @subsection Blackfin Options
9996 @cindex Blackfin Options
9999 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
10001 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
10002 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
10003 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
10004 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
10005 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
10006 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
10007 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
10009 The optional @var{sirevision} specifies the silicon revision of the target
10010 Blackfin processor. Any workarounds available for the targeted silicon revision
10011 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
10012 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
10013 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
10014 hexadecimal digits representing the major and minor numbers in the silicon
10015 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
10016 is not defined. If @var{sirevision} is @samp{any}, the
10017 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
10018 If this optional @var{sirevision} is not used, GCC assumes the latest known
10019 silicon revision of the targeted Blackfin processor.
10021 Support for @samp{bf561} is incomplete. For @samp{bf561},
10022 Only the processor macro is defined.
10023 Without this option, @samp{bf532} is used as the processor by default.
10024 The corresponding predefined processor macros for @var{cpu} is to
10025 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
10026 provided by libgloss to be linked in if @option{-msim} is not given.
10030 Specifies that the program will be run on the simulator. This causes
10031 the simulator BSP provided by libgloss to be linked in. This option
10032 has effect only for @samp{bfin-elf} toolchain.
10033 Certain other options, such as @option{-mid-shared-library} and
10034 @option{-mfdpic}, imply @option{-msim}.
10036 @item -momit-leaf-frame-pointer
10037 @opindex momit-leaf-frame-pointer
10038 Don't keep the frame pointer in a register for leaf functions. This
10039 avoids the instructions to save, set up and restore frame pointers and
10040 makes an extra register available in leaf functions. The option
10041 @option{-fomit-frame-pointer} removes the frame pointer for all functions
10042 which might make debugging harder.
10044 @item -mspecld-anomaly
10045 @opindex mspecld-anomaly
10046 When enabled, the compiler will ensure that the generated code does not
10047 contain speculative loads after jump instructions. If this option is used,
10048 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
10050 @item -mno-specld-anomaly
10051 @opindex mno-specld-anomaly
10052 Don't generate extra code to prevent speculative loads from occurring.
10054 @item -mcsync-anomaly
10055 @opindex mcsync-anomaly
10056 When enabled, the compiler will ensure that the generated code does not
10057 contain CSYNC or SSYNC instructions too soon after conditional branches.
10058 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
10060 @item -mno-csync-anomaly
10061 @opindex mno-csync-anomaly
10062 Don't generate extra code to prevent CSYNC or SSYNC instructions from
10063 occurring too soon after a conditional branch.
10067 When enabled, the compiler is free to take advantage of the knowledge that
10068 the entire program fits into the low 64k of memory.
10071 @opindex mno-low-64k
10072 Assume that the program is arbitrarily large. This is the default.
10074 @item -mstack-check-l1
10075 @opindex mstack-check-l1
10076 Do stack checking using information placed into L1 scratchpad memory by the
10079 @item -mid-shared-library
10080 @opindex mid-shared-library
10081 Generate code that supports shared libraries via the library ID method.
10082 This allows for execute in place and shared libraries in an environment
10083 without virtual memory management. This option implies @option{-fPIC}.
10084 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10086 @item -mno-id-shared-library
10087 @opindex mno-id-shared-library
10088 Generate code that doesn't assume ID based shared libraries are being used.
10089 This is the default.
10091 @item -mleaf-id-shared-library
10092 @opindex mleaf-id-shared-library
10093 Generate code that supports shared libraries via the library ID method,
10094 but assumes that this library or executable won't link against any other
10095 ID shared libraries. That allows the compiler to use faster code for jumps
10098 @item -mno-leaf-id-shared-library
10099 @opindex mno-leaf-id-shared-library
10100 Do not assume that the code being compiled won't link against any ID shared
10101 libraries. Slower code will be generated for jump and call insns.
10103 @item -mshared-library-id=n
10104 @opindex mshared-library-id
10105 Specified the identification number of the ID based shared library being
10106 compiled. Specifying a value of 0 will generate more compact code, specifying
10107 other values will force the allocation of that number to the current
10108 library but is no more space or time efficient than omitting this option.
10112 Generate code that allows the data segment to be located in a different
10113 area of memory from the text segment. This allows for execute in place in
10114 an environment without virtual memory management by eliminating relocations
10115 against the text section.
10117 @item -mno-sep-data
10118 @opindex mno-sep-data
10119 Generate code that assumes that the data segment follows the text segment.
10120 This is the default.
10123 @itemx -mno-long-calls
10124 @opindex mlong-calls
10125 @opindex mno-long-calls
10126 Tells the compiler to perform function calls by first loading the
10127 address of the function into a register and then performing a subroutine
10128 call on this register. This switch is needed if the target function
10129 will lie outside of the 24 bit addressing range of the offset based
10130 version of subroutine call instruction.
10132 This feature is not enabled by default. Specifying
10133 @option{-mno-long-calls} will restore the default behavior. Note these
10134 switches have no effect on how the compiler generates code to handle
10135 function calls via function pointers.
10139 Link with the fast floating-point library. This library relaxes some of
10140 the IEEE floating-point standard's rules for checking inputs against
10141 Not-a-Number (NAN), in the interest of performance.
10144 @opindex minline-plt
10145 Enable inlining of PLT entries in function calls to functions that are
10146 not known to bind locally. It has no effect without @option{-mfdpic}.
10149 @opindex mmulticore
10150 Build standalone application for multicore Blackfin processor. Proper
10151 start files and link scripts will be used to support multicore.
10152 This option defines @code{__BFIN_MULTICORE}. It can only be used with
10153 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
10154 @option{-mcorea} or @option{-mcoreb}. If it's used without
10155 @option{-mcorea} or @option{-mcoreb}, single application/dual core
10156 programming model is used. In this model, the main function of Core B
10157 should be named as coreb_main. If it's used with @option{-mcorea} or
10158 @option{-mcoreb}, one application per core programming model is used.
10159 If this option is not used, single core application programming
10164 Build standalone application for Core A of BF561 when using
10165 one application per core programming model. Proper start files
10166 and link scripts will be used to support Core A. This option
10167 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
10171 Build standalone application for Core B of BF561 when using
10172 one application per core programming model. Proper start files
10173 and link scripts will be used to support Core B. This option
10174 defines @code{__BFIN_COREB}. When this option is used, coreb_main
10175 should be used instead of main. It must be used with
10176 @option{-mmulticore}.
10180 Build standalone application for SDRAM. Proper start files and
10181 link scripts will be used to put the application into SDRAM.
10182 Loader should initialize SDRAM before loading the application
10183 into SDRAM. This option defines @code{__BFIN_SDRAM}.
10187 Assume that ICPLBs are enabled at runtime. This has an effect on certain
10188 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
10189 are enabled; for standalone applications the default is off.
10193 @subsection CRIS Options
10194 @cindex CRIS Options
10196 These options are defined specifically for the CRIS ports.
10199 @item -march=@var{architecture-type}
10200 @itemx -mcpu=@var{architecture-type}
10203 Generate code for the specified architecture. The choices for
10204 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
10205 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
10206 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
10209 @item -mtune=@var{architecture-type}
10211 Tune to @var{architecture-type} everything applicable about the generated
10212 code, except for the ABI and the set of available instructions. The
10213 choices for @var{architecture-type} are the same as for
10214 @option{-march=@var{architecture-type}}.
10216 @item -mmax-stack-frame=@var{n}
10217 @opindex mmax-stack-frame
10218 Warn when the stack frame of a function exceeds @var{n} bytes.
10224 The options @option{-metrax4} and @option{-metrax100} are synonyms for
10225 @option{-march=v3} and @option{-march=v8} respectively.
10227 @item -mmul-bug-workaround
10228 @itemx -mno-mul-bug-workaround
10229 @opindex mmul-bug-workaround
10230 @opindex mno-mul-bug-workaround
10231 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
10232 models where it applies. This option is active by default.
10236 Enable CRIS-specific verbose debug-related information in the assembly
10237 code. This option also has the effect to turn off the @samp{#NO_APP}
10238 formatted-code indicator to the assembler at the beginning of the
10243 Do not use condition-code results from previous instruction; always emit
10244 compare and test instructions before use of condition codes.
10246 @item -mno-side-effects
10247 @opindex mno-side-effects
10248 Do not emit instructions with side-effects in addressing modes other than
10251 @item -mstack-align
10252 @itemx -mno-stack-align
10253 @itemx -mdata-align
10254 @itemx -mno-data-align
10255 @itemx -mconst-align
10256 @itemx -mno-const-align
10257 @opindex mstack-align
10258 @opindex mno-stack-align
10259 @opindex mdata-align
10260 @opindex mno-data-align
10261 @opindex mconst-align
10262 @opindex mno-const-align
10263 These options (no-options) arranges (eliminate arrangements) for the
10264 stack-frame, individual data and constants to be aligned for the maximum
10265 single data access size for the chosen CPU model. The default is to
10266 arrange for 32-bit alignment. ABI details such as structure layout are
10267 not affected by these options.
10275 Similar to the stack- data- and const-align options above, these options
10276 arrange for stack-frame, writable data and constants to all be 32-bit,
10277 16-bit or 8-bit aligned. The default is 32-bit alignment.
10279 @item -mno-prologue-epilogue
10280 @itemx -mprologue-epilogue
10281 @opindex mno-prologue-epilogue
10282 @opindex mprologue-epilogue
10283 With @option{-mno-prologue-epilogue}, the normal function prologue and
10284 epilogue that sets up the stack-frame are omitted and no return
10285 instructions or return sequences are generated in the code. Use this
10286 option only together with visual inspection of the compiled code: no
10287 warnings or errors are generated when call-saved registers must be saved,
10288 or storage for local variable needs to be allocated.
10292 @opindex mno-gotplt
10294 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
10295 instruction sequences that load addresses for functions from the PLT part
10296 of the GOT rather than (traditional on other architectures) calls to the
10297 PLT@. The default is @option{-mgotplt}.
10301 Legacy no-op option only recognized with the cris-axis-elf and
10302 cris-axis-linux-gnu targets.
10306 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
10310 This option, recognized for the cris-axis-elf arranges
10311 to link with input-output functions from a simulator library. Code,
10312 initialized data and zero-initialized data are allocated consecutively.
10316 Like @option{-sim}, but pass linker options to locate initialized data at
10317 0x40000000 and zero-initialized data at 0x80000000.
10321 @subsection CRX Options
10322 @cindex CRX Options
10324 These options are defined specifically for the CRX ports.
10330 Enable the use of multiply-accumulate instructions. Disabled by default.
10333 @opindex mpush-args
10334 Push instructions will be used to pass outgoing arguments when functions
10335 are called. Enabled by default.
10338 @node Darwin Options
10339 @subsection Darwin Options
10340 @cindex Darwin options
10342 These options are defined for all architectures running the Darwin operating
10345 FSF GCC on Darwin does not create ``fat'' object files; it will create
10346 an object file for the single architecture that it was built to
10347 target. Apple's GCC on Darwin does create ``fat'' files if multiple
10348 @option{-arch} options are used; it does so by running the compiler or
10349 linker multiple times and joining the results together with
10352 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
10353 @samp{i686}) is determined by the flags that specify the ISA
10354 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
10355 @option{-force_cpusubtype_ALL} option can be used to override this.
10357 The Darwin tools vary in their behavior when presented with an ISA
10358 mismatch. The assembler, @file{as}, will only permit instructions to
10359 be used that are valid for the subtype of the file it is generating,
10360 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
10361 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
10362 and print an error if asked to create a shared library with a less
10363 restrictive subtype than its input files (for instance, trying to put
10364 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
10365 for executables, @file{ld}, will quietly give the executable the most
10366 restrictive subtype of any of its input files.
10371 Add the framework directory @var{dir} to the head of the list of
10372 directories to be searched for header files. These directories are
10373 interleaved with those specified by @option{-I} options and are
10374 scanned in a left-to-right order.
10376 A framework directory is a directory with frameworks in it. A
10377 framework is a directory with a @samp{"Headers"} and/or
10378 @samp{"PrivateHeaders"} directory contained directly in it that ends
10379 in @samp{".framework"}. The name of a framework is the name of this
10380 directory excluding the @samp{".framework"}. Headers associated with
10381 the framework are found in one of those two directories, with
10382 @samp{"Headers"} being searched first. A subframework is a framework
10383 directory that is in a framework's @samp{"Frameworks"} directory.
10384 Includes of subframework headers can only appear in a header of a
10385 framework that contains the subframework, or in a sibling subframework
10386 header. Two subframeworks are siblings if they occur in the same
10387 framework. A subframework should not have the same name as a
10388 framework, a warning will be issued if this is violated. Currently a
10389 subframework cannot have subframeworks, in the future, the mechanism
10390 may be extended to support this. The standard frameworks can be found
10391 in @samp{"/System/Library/Frameworks"} and
10392 @samp{"/Library/Frameworks"}. An example include looks like
10393 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
10394 the name of the framework and header.h is found in the
10395 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
10397 @item -iframework@var{dir}
10398 @opindex iframework
10399 Like @option{-F} except the directory is a treated as a system
10400 directory. The main difference between this @option{-iframework} and
10401 @option{-F} is that with @option{-iframework} the compiler does not
10402 warn about constructs contained within header files found via
10403 @var{dir}. This option is valid only for the C family of languages.
10407 Emit debugging information for symbols that are used. For STABS
10408 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
10409 This is by default ON@.
10413 Emit debugging information for all symbols and types.
10415 @item -mmacosx-version-min=@var{version}
10416 The earliest version of MacOS X that this executable will run on
10417 is @var{version}. Typical values of @var{version} include @code{10.1},
10418 @code{10.2}, and @code{10.3.9}.
10420 If the compiler was built to use the system's headers by default,
10421 then the default for this option is the system version on which the
10422 compiler is running, otherwise the default is to make choices which
10423 are compatible with as many systems and code bases as possible.
10427 Enable kernel development mode. The @option{-mkernel} option sets
10428 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
10429 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
10430 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
10431 applicable. This mode also sets @option{-mno-altivec},
10432 @option{-msoft-float}, @option{-fno-builtin} and
10433 @option{-mlong-branch} for PowerPC targets.
10435 @item -mone-byte-bool
10436 @opindex mone-byte-bool
10437 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
10438 By default @samp{sizeof(bool)} is @samp{4} when compiling for
10439 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
10440 option has no effect on x86.
10442 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
10443 to generate code that is not binary compatible with code generated
10444 without that switch. Using this switch may require recompiling all
10445 other modules in a program, including system libraries. Use this
10446 switch to conform to a non-default data model.
10448 @item -mfix-and-continue
10449 @itemx -ffix-and-continue
10450 @itemx -findirect-data
10451 @opindex mfix-and-continue
10452 @opindex ffix-and-continue
10453 @opindex findirect-data
10454 Generate code suitable for fast turn around development. Needed to
10455 enable gdb to dynamically load @code{.o} files into already running
10456 programs. @option{-findirect-data} and @option{-ffix-and-continue}
10457 are provided for backwards compatibility.
10461 Loads all members of static archive libraries.
10462 See man ld(1) for more information.
10464 @item -arch_errors_fatal
10465 @opindex arch_errors_fatal
10466 Cause the errors having to do with files that have the wrong architecture
10469 @item -bind_at_load
10470 @opindex bind_at_load
10471 Causes the output file to be marked such that the dynamic linker will
10472 bind all undefined references when the file is loaded or launched.
10476 Produce a Mach-o bundle format file.
10477 See man ld(1) for more information.
10479 @item -bundle_loader @var{executable}
10480 @opindex bundle_loader
10481 This option specifies the @var{executable} that will be loading the build
10482 output file being linked. See man ld(1) for more information.
10485 @opindex dynamiclib
10486 When passed this option, GCC will produce a dynamic library instead of
10487 an executable when linking, using the Darwin @file{libtool} command.
10489 @item -force_cpusubtype_ALL
10490 @opindex force_cpusubtype_ALL
10491 This causes GCC's output file to have the @var{ALL} subtype, instead of
10492 one controlled by the @option{-mcpu} or @option{-march} option.
10494 @item -allowable_client @var{client_name}
10495 @itemx -client_name
10496 @itemx -compatibility_version
10497 @itemx -current_version
10499 @itemx -dependency-file
10501 @itemx -dylinker_install_name
10503 @itemx -exported_symbols_list
10505 @itemx -flat_namespace
10506 @itemx -force_flat_namespace
10507 @itemx -headerpad_max_install_names
10510 @itemx -install_name
10511 @itemx -keep_private_externs
10512 @itemx -multi_module
10513 @itemx -multiply_defined
10514 @itemx -multiply_defined_unused
10516 @itemx -no_dead_strip_inits_and_terms
10517 @itemx -nofixprebinding
10518 @itemx -nomultidefs
10520 @itemx -noseglinkedit
10521 @itemx -pagezero_size
10523 @itemx -prebind_all_twolevel_modules
10524 @itemx -private_bundle
10525 @itemx -read_only_relocs
10527 @itemx -sectobjectsymbols
10531 @itemx -sectobjectsymbols
10534 @itemx -segs_read_only_addr
10535 @itemx -segs_read_write_addr
10536 @itemx -seg_addr_table
10537 @itemx -seg_addr_table_filename
10538 @itemx -seglinkedit
10540 @itemx -segs_read_only_addr
10541 @itemx -segs_read_write_addr
10542 @itemx -single_module
10544 @itemx -sub_library
10545 @itemx -sub_umbrella
10546 @itemx -twolevel_namespace
10549 @itemx -unexported_symbols_list
10550 @itemx -weak_reference_mismatches
10551 @itemx -whatsloaded
10552 @opindex allowable_client
10553 @opindex client_name
10554 @opindex compatibility_version
10555 @opindex current_version
10556 @opindex dead_strip
10557 @opindex dependency-file
10558 @opindex dylib_file
10559 @opindex dylinker_install_name
10561 @opindex exported_symbols_list
10563 @opindex flat_namespace
10564 @opindex force_flat_namespace
10565 @opindex headerpad_max_install_names
10566 @opindex image_base
10568 @opindex install_name
10569 @opindex keep_private_externs
10570 @opindex multi_module
10571 @opindex multiply_defined
10572 @opindex multiply_defined_unused
10573 @opindex noall_load
10574 @opindex no_dead_strip_inits_and_terms
10575 @opindex nofixprebinding
10576 @opindex nomultidefs
10578 @opindex noseglinkedit
10579 @opindex pagezero_size
10581 @opindex prebind_all_twolevel_modules
10582 @opindex private_bundle
10583 @opindex read_only_relocs
10585 @opindex sectobjectsymbols
10588 @opindex sectcreate
10589 @opindex sectobjectsymbols
10592 @opindex segs_read_only_addr
10593 @opindex segs_read_write_addr
10594 @opindex seg_addr_table
10595 @opindex seg_addr_table_filename
10596 @opindex seglinkedit
10598 @opindex segs_read_only_addr
10599 @opindex segs_read_write_addr
10600 @opindex single_module
10602 @opindex sub_library
10603 @opindex sub_umbrella
10604 @opindex twolevel_namespace
10607 @opindex unexported_symbols_list
10608 @opindex weak_reference_mismatches
10609 @opindex whatsloaded
10610 These options are passed to the Darwin linker. The Darwin linker man page
10611 describes them in detail.
10614 @node DEC Alpha Options
10615 @subsection DEC Alpha Options
10617 These @samp{-m} options are defined for the DEC Alpha implementations:
10620 @item -mno-soft-float
10621 @itemx -msoft-float
10622 @opindex mno-soft-float
10623 @opindex msoft-float
10624 Use (do not use) the hardware floating-point instructions for
10625 floating-point operations. When @option{-msoft-float} is specified,
10626 functions in @file{libgcc.a} will be used to perform floating-point
10627 operations. Unless they are replaced by routines that emulate the
10628 floating-point operations, or compiled in such a way as to call such
10629 emulations routines, these routines will issue floating-point
10630 operations. If you are compiling for an Alpha without floating-point
10631 operations, you must ensure that the library is built so as not to call
10634 Note that Alpha implementations without floating-point operations are
10635 required to have floating-point registers.
10638 @itemx -mno-fp-regs
10640 @opindex mno-fp-regs
10641 Generate code that uses (does not use) the floating-point register set.
10642 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
10643 register set is not used, floating point operands are passed in integer
10644 registers as if they were integers and floating-point results are passed
10645 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
10646 so any function with a floating-point argument or return value called by code
10647 compiled with @option{-mno-fp-regs} must also be compiled with that
10650 A typical use of this option is building a kernel that does not use,
10651 and hence need not save and restore, any floating-point registers.
10655 The Alpha architecture implements floating-point hardware optimized for
10656 maximum performance. It is mostly compliant with the IEEE floating
10657 point standard. However, for full compliance, software assistance is
10658 required. This option generates code fully IEEE compliant code
10659 @emph{except} that the @var{inexact-flag} is not maintained (see below).
10660 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
10661 defined during compilation. The resulting code is less efficient but is
10662 able to correctly support denormalized numbers and exceptional IEEE
10663 values such as not-a-number and plus/minus infinity. Other Alpha
10664 compilers call this option @option{-ieee_with_no_inexact}.
10666 @item -mieee-with-inexact
10667 @opindex mieee-with-inexact
10668 This is like @option{-mieee} except the generated code also maintains
10669 the IEEE @var{inexact-flag}. Turning on this option causes the
10670 generated code to implement fully-compliant IEEE math. In addition to
10671 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
10672 macro. On some Alpha implementations the resulting code may execute
10673 significantly slower than the code generated by default. Since there is
10674 very little code that depends on the @var{inexact-flag}, you should
10675 normally not specify this option. Other Alpha compilers call this
10676 option @option{-ieee_with_inexact}.
10678 @item -mfp-trap-mode=@var{trap-mode}
10679 @opindex mfp-trap-mode
10680 This option controls what floating-point related traps are enabled.
10681 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
10682 The trap mode can be set to one of four values:
10686 This is the default (normal) setting. The only traps that are enabled
10687 are the ones that cannot be disabled in software (e.g., division by zero
10691 In addition to the traps enabled by @samp{n}, underflow traps are enabled
10695 Like @samp{u}, but the instructions are marked to be safe for software
10696 completion (see Alpha architecture manual for details).
10699 Like @samp{su}, but inexact traps are enabled as well.
10702 @item -mfp-rounding-mode=@var{rounding-mode}
10703 @opindex mfp-rounding-mode
10704 Selects the IEEE rounding mode. Other Alpha compilers call this option
10705 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
10710 Normal IEEE rounding mode. Floating point numbers are rounded towards
10711 the nearest machine number or towards the even machine number in case
10715 Round towards minus infinity.
10718 Chopped rounding mode. Floating point numbers are rounded towards zero.
10721 Dynamic rounding mode. A field in the floating point control register
10722 (@var{fpcr}, see Alpha architecture reference manual) controls the
10723 rounding mode in effect. The C library initializes this register for
10724 rounding towards plus infinity. Thus, unless your program modifies the
10725 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
10728 @item -mtrap-precision=@var{trap-precision}
10729 @opindex mtrap-precision
10730 In the Alpha architecture, floating point traps are imprecise. This
10731 means without software assistance it is impossible to recover from a
10732 floating trap and program execution normally needs to be terminated.
10733 GCC can generate code that can assist operating system trap handlers
10734 in determining the exact location that caused a floating point trap.
10735 Depending on the requirements of an application, different levels of
10736 precisions can be selected:
10740 Program precision. This option is the default and means a trap handler
10741 can only identify which program caused a floating point exception.
10744 Function precision. The trap handler can determine the function that
10745 caused a floating point exception.
10748 Instruction precision. The trap handler can determine the exact
10749 instruction that caused a floating point exception.
10752 Other Alpha compilers provide the equivalent options called
10753 @option{-scope_safe} and @option{-resumption_safe}.
10755 @item -mieee-conformant
10756 @opindex mieee-conformant
10757 This option marks the generated code as IEEE conformant. You must not
10758 use this option unless you also specify @option{-mtrap-precision=i} and either
10759 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
10760 is to emit the line @samp{.eflag 48} in the function prologue of the
10761 generated assembly file. Under DEC Unix, this has the effect that
10762 IEEE-conformant math library routines will be linked in.
10764 @item -mbuild-constants
10765 @opindex mbuild-constants
10766 Normally GCC examines a 32- or 64-bit integer constant to
10767 see if it can construct it from smaller constants in two or three
10768 instructions. If it cannot, it will output the constant as a literal and
10769 generate code to load it from the data segment at runtime.
10771 Use this option to require GCC to construct @emph{all} integer constants
10772 using code, even if it takes more instructions (the maximum is six).
10774 You would typically use this option to build a shared library dynamic
10775 loader. Itself a shared library, it must relocate itself in memory
10776 before it can find the variables and constants in its own data segment.
10782 Select whether to generate code to be assembled by the vendor-supplied
10783 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
10801 Indicate whether GCC should generate code to use the optional BWX,
10802 CIX, FIX and MAX instruction sets. The default is to use the instruction
10803 sets supported by the CPU type specified via @option{-mcpu=} option or that
10804 of the CPU on which GCC was built if none was specified.
10807 @itemx -mfloat-ieee
10808 @opindex mfloat-vax
10809 @opindex mfloat-ieee
10810 Generate code that uses (does not use) VAX F and G floating point
10811 arithmetic instead of IEEE single and double precision.
10813 @item -mexplicit-relocs
10814 @itemx -mno-explicit-relocs
10815 @opindex mexplicit-relocs
10816 @opindex mno-explicit-relocs
10817 Older Alpha assemblers provided no way to generate symbol relocations
10818 except via assembler macros. Use of these macros does not allow
10819 optimal instruction scheduling. GNU binutils as of version 2.12
10820 supports a new syntax that allows the compiler to explicitly mark
10821 which relocations should apply to which instructions. This option
10822 is mostly useful for debugging, as GCC detects the capabilities of
10823 the assembler when it is built and sets the default accordingly.
10826 @itemx -mlarge-data
10827 @opindex msmall-data
10828 @opindex mlarge-data
10829 When @option{-mexplicit-relocs} is in effect, static data is
10830 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
10831 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
10832 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
10833 16-bit relocations off of the @code{$gp} register. This limits the
10834 size of the small data area to 64KB, but allows the variables to be
10835 directly accessed via a single instruction.
10837 The default is @option{-mlarge-data}. With this option the data area
10838 is limited to just below 2GB@. Programs that require more than 2GB of
10839 data must use @code{malloc} or @code{mmap} to allocate the data in the
10840 heap instead of in the program's data segment.
10842 When generating code for shared libraries, @option{-fpic} implies
10843 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
10846 @itemx -mlarge-text
10847 @opindex msmall-text
10848 @opindex mlarge-text
10849 When @option{-msmall-text} is used, the compiler assumes that the
10850 code of the entire program (or shared library) fits in 4MB, and is
10851 thus reachable with a branch instruction. When @option{-msmall-data}
10852 is used, the compiler can assume that all local symbols share the
10853 same @code{$gp} value, and thus reduce the number of instructions
10854 required for a function call from 4 to 1.
10856 The default is @option{-mlarge-text}.
10858 @item -mcpu=@var{cpu_type}
10860 Set the instruction set and instruction scheduling parameters for
10861 machine type @var{cpu_type}. You can specify either the @samp{EV}
10862 style name or the corresponding chip number. GCC supports scheduling
10863 parameters for the EV4, EV5 and EV6 family of processors and will
10864 choose the default values for the instruction set from the processor
10865 you specify. If you do not specify a processor type, GCC will default
10866 to the processor on which the compiler was built.
10868 Supported values for @var{cpu_type} are
10874 Schedules as an EV4 and has no instruction set extensions.
10878 Schedules as an EV5 and has no instruction set extensions.
10882 Schedules as an EV5 and supports the BWX extension.
10887 Schedules as an EV5 and supports the BWX and MAX extensions.
10891 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
10895 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
10898 Native Linux/GNU toolchains also support the value @samp{native},
10899 which selects the best architecture option for the host processor.
10900 @option{-mcpu=native} has no effect if GCC does not recognize
10903 @item -mtune=@var{cpu_type}
10905 Set only the instruction scheduling parameters for machine type
10906 @var{cpu_type}. The instruction set is not changed.
10908 Native Linux/GNU toolchains also support the value @samp{native},
10909 which selects the best architecture option for the host processor.
10910 @option{-mtune=native} has no effect if GCC does not recognize
10913 @item -mmemory-latency=@var{time}
10914 @opindex mmemory-latency
10915 Sets the latency the scheduler should assume for typical memory
10916 references as seen by the application. This number is highly
10917 dependent on the memory access patterns used by the application
10918 and the size of the external cache on the machine.
10920 Valid options for @var{time} are
10924 A decimal number representing clock cycles.
10930 The compiler contains estimates of the number of clock cycles for
10931 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
10932 (also called Dcache, Scache, and Bcache), as well as to main memory.
10933 Note that L3 is only valid for EV5.
10938 @node DEC Alpha/VMS Options
10939 @subsection DEC Alpha/VMS Options
10941 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
10944 @item -mvms-return-codes
10945 @opindex mvms-return-codes
10946 Return VMS condition codes from main. The default is to return POSIX
10947 style condition (e.g.@: error) codes.
10949 @item -mdebug-main=@var{prefix}
10950 @opindex mdebug-main=@var{prefix}
10951 Flag the first routine whose name starts with @var{prefix} as the main
10952 routine for the debugger.
10956 Default to 64bit memory allocation routines.
10960 @subsection FR30 Options
10961 @cindex FR30 Options
10963 These options are defined specifically for the FR30 port.
10967 @item -msmall-model
10968 @opindex msmall-model
10969 Use the small address space model. This can produce smaller code, but
10970 it does assume that all symbolic values and addresses will fit into a
10975 Assume that run-time support has been provided and so there is no need
10976 to include the simulator library (@file{libsim.a}) on the linker
10982 @subsection FRV Options
10983 @cindex FRV Options
10989 Only use the first 32 general purpose registers.
10994 Use all 64 general purpose registers.
10999 Use only the first 32 floating point registers.
11004 Use all 64 floating point registers
11007 @opindex mhard-float
11009 Use hardware instructions for floating point operations.
11012 @opindex msoft-float
11014 Use library routines for floating point operations.
11019 Dynamically allocate condition code registers.
11024 Do not try to dynamically allocate condition code registers, only
11025 use @code{icc0} and @code{fcc0}.
11030 Change ABI to use double word insns.
11035 Do not use double word instructions.
11040 Use floating point double instructions.
11043 @opindex mno-double
11045 Do not use floating point double instructions.
11050 Use media instructions.
11055 Do not use media instructions.
11060 Use multiply and add/subtract instructions.
11063 @opindex mno-muladd
11065 Do not use multiply and add/subtract instructions.
11070 Select the FDPIC ABI, that uses function descriptors to represent
11071 pointers to functions. Without any PIC/PIE-related options, it
11072 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
11073 assumes GOT entries and small data are within a 12-bit range from the
11074 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
11075 are computed with 32 bits.
11076 With a @samp{bfin-elf} target, this option implies @option{-msim}.
11079 @opindex minline-plt
11081 Enable inlining of PLT entries in function calls to functions that are
11082 not known to bind locally. It has no effect without @option{-mfdpic}.
11083 It's enabled by default if optimizing for speed and compiling for
11084 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
11085 optimization option such as @option{-O3} or above is present in the
11091 Assume a large TLS segment when generating thread-local code.
11096 Do not assume a large TLS segment when generating thread-local code.
11101 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
11102 that is known to be in read-only sections. It's enabled by default,
11103 except for @option{-fpic} or @option{-fpie}: even though it may help
11104 make the global offset table smaller, it trades 1 instruction for 4.
11105 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
11106 one of which may be shared by multiple symbols, and it avoids the need
11107 for a GOT entry for the referenced symbol, so it's more likely to be a
11108 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
11110 @item -multilib-library-pic
11111 @opindex multilib-library-pic
11113 Link with the (library, not FD) pic libraries. It's implied by
11114 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
11115 @option{-fpic} without @option{-mfdpic}. You should never have to use
11119 @opindex mlinked-fp
11121 Follow the EABI requirement of always creating a frame pointer whenever
11122 a stack frame is allocated. This option is enabled by default and can
11123 be disabled with @option{-mno-linked-fp}.
11126 @opindex mlong-calls
11128 Use indirect addressing to call functions outside the current
11129 compilation unit. This allows the functions to be placed anywhere
11130 within the 32-bit address space.
11132 @item -malign-labels
11133 @opindex malign-labels
11135 Try to align labels to an 8-byte boundary by inserting nops into the
11136 previous packet. This option only has an effect when VLIW packing
11137 is enabled. It doesn't create new packets; it merely adds nops to
11140 @item -mlibrary-pic
11141 @opindex mlibrary-pic
11143 Generate position-independent EABI code.
11148 Use only the first four media accumulator registers.
11153 Use all eight media accumulator registers.
11158 Pack VLIW instructions.
11163 Do not pack VLIW instructions.
11166 @opindex mno-eflags
11168 Do not mark ABI switches in e_flags.
11171 @opindex mcond-move
11173 Enable the use of conditional-move instructions (default).
11175 This switch is mainly for debugging the compiler and will likely be removed
11176 in a future version.
11178 @item -mno-cond-move
11179 @opindex mno-cond-move
11181 Disable the use of conditional-move instructions.
11183 This switch is mainly for debugging the compiler and will likely be removed
11184 in a future version.
11189 Enable the use of conditional set instructions (default).
11191 This switch is mainly for debugging the compiler and will likely be removed
11192 in a future version.
11197 Disable the use of conditional set instructions.
11199 This switch is mainly for debugging the compiler and will likely be removed
11200 in a future version.
11203 @opindex mcond-exec
11205 Enable the use of conditional execution (default).
11207 This switch is mainly for debugging the compiler and will likely be removed
11208 in a future version.
11210 @item -mno-cond-exec
11211 @opindex mno-cond-exec
11213 Disable the use of conditional execution.
11215 This switch is mainly for debugging the compiler and will likely be removed
11216 in a future version.
11218 @item -mvliw-branch
11219 @opindex mvliw-branch
11221 Run a pass to pack branches into VLIW instructions (default).
11223 This switch is mainly for debugging the compiler and will likely be removed
11224 in a future version.
11226 @item -mno-vliw-branch
11227 @opindex mno-vliw-branch
11229 Do not run a pass to pack branches into VLIW instructions.
11231 This switch is mainly for debugging the compiler and will likely be removed
11232 in a future version.
11234 @item -mmulti-cond-exec
11235 @opindex mmulti-cond-exec
11237 Enable optimization of @code{&&} and @code{||} in conditional execution
11240 This switch is mainly for debugging the compiler and will likely be removed
11241 in a future version.
11243 @item -mno-multi-cond-exec
11244 @opindex mno-multi-cond-exec
11246 Disable optimization of @code{&&} and @code{||} in conditional execution.
11248 This switch is mainly for debugging the compiler and will likely be removed
11249 in a future version.
11251 @item -mnested-cond-exec
11252 @opindex mnested-cond-exec
11254 Enable nested conditional execution optimizations (default).
11256 This switch is mainly for debugging the compiler and will likely be removed
11257 in a future version.
11259 @item -mno-nested-cond-exec
11260 @opindex mno-nested-cond-exec
11262 Disable nested conditional execution optimizations.
11264 This switch is mainly for debugging the compiler and will likely be removed
11265 in a future version.
11267 @item -moptimize-membar
11268 @opindex moptimize-membar
11270 This switch removes redundant @code{membar} instructions from the
11271 compiler generated code. It is enabled by default.
11273 @item -mno-optimize-membar
11274 @opindex mno-optimize-membar
11276 This switch disables the automatic removal of redundant @code{membar}
11277 instructions from the generated code.
11279 @item -mtomcat-stats
11280 @opindex mtomcat-stats
11282 Cause gas to print out tomcat statistics.
11284 @item -mcpu=@var{cpu}
11287 Select the processor type for which to generate code. Possible values are
11288 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
11289 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
11293 @node GNU/Linux Options
11294 @subsection GNU/Linux Options
11296 These @samp{-m} options are defined for GNU/Linux targets:
11301 Use the GNU C library instead of uClibc. This is the default except
11302 on @samp{*-*-linux-*uclibc*} targets.
11306 Use uClibc instead of the GNU C library. This is the default on
11307 @samp{*-*-linux-*uclibc*} targets.
11310 @node H8/300 Options
11311 @subsection H8/300 Options
11313 These @samp{-m} options are defined for the H8/300 implementations:
11318 Shorten some address references at link time, when possible; uses the
11319 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
11320 ld, Using ld}, for a fuller description.
11324 Generate code for the H8/300H@.
11328 Generate code for the H8S@.
11332 Generate code for the H8S and H8/300H in the normal mode. This switch
11333 must be used either with @option{-mh} or @option{-ms}.
11337 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
11341 Make @code{int} data 32 bits by default.
11344 @opindex malign-300
11345 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
11346 The default for the H8/300H and H8S is to align longs and floats on 4
11348 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
11349 This option has no effect on the H8/300.
11353 @subsection HPPA Options
11354 @cindex HPPA Options
11356 These @samp{-m} options are defined for the HPPA family of computers:
11359 @item -march=@var{architecture-type}
11361 Generate code for the specified architecture. The choices for
11362 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
11363 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
11364 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
11365 architecture option for your machine. Code compiled for lower numbered
11366 architectures will run on higher numbered architectures, but not the
11369 @item -mpa-risc-1-0
11370 @itemx -mpa-risc-1-1
11371 @itemx -mpa-risc-2-0
11372 @opindex mpa-risc-1-0
11373 @opindex mpa-risc-1-1
11374 @opindex mpa-risc-2-0
11375 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
11378 @opindex mbig-switch
11379 Generate code suitable for big switch tables. Use this option only if
11380 the assembler/linker complain about out of range branches within a switch
11383 @item -mjump-in-delay
11384 @opindex mjump-in-delay
11385 Fill delay slots of function calls with unconditional jump instructions
11386 by modifying the return pointer for the function call to be the target
11387 of the conditional jump.
11389 @item -mdisable-fpregs
11390 @opindex mdisable-fpregs
11391 Prevent floating point registers from being used in any manner. This is
11392 necessary for compiling kernels which perform lazy context switching of
11393 floating point registers. If you use this option and attempt to perform
11394 floating point operations, the compiler will abort.
11396 @item -mdisable-indexing
11397 @opindex mdisable-indexing
11398 Prevent the compiler from using indexing address modes. This avoids some
11399 rather obscure problems when compiling MIG generated code under MACH@.
11401 @item -mno-space-regs
11402 @opindex mno-space-regs
11403 Generate code that assumes the target has no space registers. This allows
11404 GCC to generate faster indirect calls and use unscaled index address modes.
11406 Such code is suitable for level 0 PA systems and kernels.
11408 @item -mfast-indirect-calls
11409 @opindex mfast-indirect-calls
11410 Generate code that assumes calls never cross space boundaries. This
11411 allows GCC to emit code which performs faster indirect calls.
11413 This option will not work in the presence of shared libraries or nested
11416 @item -mfixed-range=@var{register-range}
11417 @opindex mfixed-range
11418 Generate code treating the given register range as fixed registers.
11419 A fixed register is one that the register allocator can not use. This is
11420 useful when compiling kernel code. A register range is specified as
11421 two registers separated by a dash. Multiple register ranges can be
11422 specified separated by a comma.
11424 @item -mlong-load-store
11425 @opindex mlong-load-store
11426 Generate 3-instruction load and store sequences as sometimes required by
11427 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
11430 @item -mportable-runtime
11431 @opindex mportable-runtime
11432 Use the portable calling conventions proposed by HP for ELF systems.
11436 Enable the use of assembler directives only GAS understands.
11438 @item -mschedule=@var{cpu-type}
11440 Schedule code according to the constraints for the machine type
11441 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
11442 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
11443 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
11444 proper scheduling option for your machine. The default scheduling is
11448 @opindex mlinker-opt
11449 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
11450 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
11451 linkers in which they give bogus error messages when linking some programs.
11454 @opindex msoft-float
11455 Generate output containing library calls for floating point.
11456 @strong{Warning:} the requisite libraries are not available for all HPPA
11457 targets. Normally the facilities of the machine's usual C compiler are
11458 used, but this cannot be done directly in cross-compilation. You must make
11459 your own arrangements to provide suitable library functions for
11462 @option{-msoft-float} changes the calling convention in the output file;
11463 therefore, it is only useful if you compile @emph{all} of a program with
11464 this option. In particular, you need to compile @file{libgcc.a}, the
11465 library that comes with GCC, with @option{-msoft-float} in order for
11470 Generate the predefine, @code{_SIO}, for server IO@. The default is
11471 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
11472 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
11473 options are available under HP-UX and HI-UX@.
11477 Use GNU ld specific options. This passes @option{-shared} to ld when
11478 building a shared library. It is the default when GCC is configured,
11479 explicitly or implicitly, with the GNU linker. This option does not
11480 have any affect on which ld is called, it only changes what parameters
11481 are passed to that ld. The ld that is called is determined by the
11482 @option{--with-ld} configure option, GCC's program search path, and
11483 finally by the user's @env{PATH}. The linker used by GCC can be printed
11484 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
11485 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11489 Use HP ld specific options. This passes @option{-b} to ld when building
11490 a shared library and passes @option{+Accept TypeMismatch} to ld on all
11491 links. It is the default when GCC is configured, explicitly or
11492 implicitly, with the HP linker. This option does not have any affect on
11493 which ld is called, it only changes what parameters are passed to that
11494 ld. The ld that is called is determined by the @option{--with-ld}
11495 configure option, GCC's program search path, and finally by the user's
11496 @env{PATH}. The linker used by GCC can be printed using @samp{which
11497 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
11498 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11501 @opindex mno-long-calls
11502 Generate code that uses long call sequences. This ensures that a call
11503 is always able to reach linker generated stubs. The default is to generate
11504 long calls only when the distance from the call site to the beginning
11505 of the function or translation unit, as the case may be, exceeds a
11506 predefined limit set by the branch type being used. The limits for
11507 normal calls are 7,600,000 and 240,000 bytes, respectively for the
11508 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
11511 Distances are measured from the beginning of functions when using the
11512 @option{-ffunction-sections} option, or when using the @option{-mgas}
11513 and @option{-mno-portable-runtime} options together under HP-UX with
11516 It is normally not desirable to use this option as it will degrade
11517 performance. However, it may be useful in large applications,
11518 particularly when partial linking is used to build the application.
11520 The types of long calls used depends on the capabilities of the
11521 assembler and linker, and the type of code being generated. The
11522 impact on systems that support long absolute calls, and long pic
11523 symbol-difference or pc-relative calls should be relatively small.
11524 However, an indirect call is used on 32-bit ELF systems in pic code
11525 and it is quite long.
11527 @item -munix=@var{unix-std}
11529 Generate compiler predefines and select a startfile for the specified
11530 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
11531 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
11532 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
11533 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
11534 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
11537 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
11538 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
11539 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
11540 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
11541 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
11542 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
11544 It is @emph{important} to note that this option changes the interfaces
11545 for various library routines. It also affects the operational behavior
11546 of the C library. Thus, @emph{extreme} care is needed in using this
11549 Library code that is intended to operate with more than one UNIX
11550 standard must test, set and restore the variable @var{__xpg4_extended_mask}
11551 as appropriate. Most GNU software doesn't provide this capability.
11555 Suppress the generation of link options to search libdld.sl when the
11556 @option{-static} option is specified on HP-UX 10 and later.
11560 The HP-UX implementation of setlocale in libc has a dependency on
11561 libdld.sl. There isn't an archive version of libdld.sl. Thus,
11562 when the @option{-static} option is specified, special link options
11563 are needed to resolve this dependency.
11565 On HP-UX 10 and later, the GCC driver adds the necessary options to
11566 link with libdld.sl when the @option{-static} option is specified.
11567 This causes the resulting binary to be dynamic. On the 64-bit port,
11568 the linkers generate dynamic binaries by default in any case. The
11569 @option{-nolibdld} option can be used to prevent the GCC driver from
11570 adding these link options.
11574 Add support for multithreading with the @dfn{dce thread} library
11575 under HP-UX@. This option sets flags for both the preprocessor and
11579 @node i386 and x86-64 Options
11580 @subsection Intel 386 and AMD x86-64 Options
11581 @cindex i386 Options
11582 @cindex x86-64 Options
11583 @cindex Intel 386 Options
11584 @cindex AMD x86-64 Options
11586 These @samp{-m} options are defined for the i386 and x86-64 family of
11590 @item -mtune=@var{cpu-type}
11592 Tune to @var{cpu-type} everything applicable about the generated code, except
11593 for the ABI and the set of available instructions. The choices for
11594 @var{cpu-type} are:
11597 Produce code optimized for the most common IA32/AMD64/EM64T processors.
11598 If you know the CPU on which your code will run, then you should use
11599 the corresponding @option{-mtune} option instead of
11600 @option{-mtune=generic}. But, if you do not know exactly what CPU users
11601 of your application will have, then you should use this option.
11603 As new processors are deployed in the marketplace, the behavior of this
11604 option will change. Therefore, if you upgrade to a newer version of
11605 GCC, the code generated option will change to reflect the processors
11606 that were most common when that version of GCC was released.
11608 There is no @option{-march=generic} option because @option{-march}
11609 indicates the instruction set the compiler can use, and there is no
11610 generic instruction set applicable to all processors. In contrast,
11611 @option{-mtune} indicates the processor (or, in this case, collection of
11612 processors) for which the code is optimized.
11614 This selects the CPU to tune for at compilation time by determining
11615 the processor type of the compiling machine. Using @option{-mtune=native}
11616 will produce code optimized for the local machine under the constraints
11617 of the selected instruction set. Using @option{-march=native} will
11618 enable all instruction subsets supported by the local machine (hence
11619 the result might not run on different machines).
11621 Original Intel's i386 CPU@.
11623 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
11624 @item i586, pentium
11625 Intel Pentium CPU with no MMX support.
11627 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
11629 Intel PentiumPro CPU@.
11631 Same as @code{generic}, but when used as @code{march} option, PentiumPro
11632 instruction set will be used, so the code will run on all i686 family chips.
11634 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
11635 @item pentium3, pentium3m
11636 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
11639 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
11640 support. Used by Centrino notebooks.
11641 @item pentium4, pentium4m
11642 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
11644 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
11647 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
11648 SSE2 and SSE3 instruction set support.
11650 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
11651 instruction set support.
11653 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
11654 instruction set support.
11656 AMD K6 CPU with MMX instruction set support.
11658 Improved versions of AMD K6 CPU with MMX and 3dNOW!@: instruction set support.
11659 @item athlon, athlon-tbird
11660 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and SSE prefetch instructions
11662 @item athlon-4, athlon-xp, athlon-mp
11663 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and full SSE
11664 instruction set support.
11665 @item k8, opteron, athlon64, athlon-fx
11666 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
11667 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW!@: and 64-bit instruction set extensions.)
11668 @item k8-sse3, opteron-sse3, athlon64-sse3
11669 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
11670 @item amdfam10, barcelona
11671 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
11672 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3dNOW!, enhanced 3dNOW!, ABM and 64-bit
11673 instruction set extensions.)
11675 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
11678 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!@:
11679 instruction set support.
11681 Via C3 CPU with MMX and 3dNOW!@: instruction set support. (No scheduling is
11682 implemented for this chip.)
11684 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
11685 implemented for this chip.)
11687 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
11690 While picking a specific @var{cpu-type} will schedule things appropriately
11691 for that particular chip, the compiler will not generate any code that
11692 does not run on the i386 without the @option{-march=@var{cpu-type}} option
11695 @item -march=@var{cpu-type}
11697 Generate instructions for the machine type @var{cpu-type}. The choices
11698 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
11699 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
11701 @item -mcpu=@var{cpu-type}
11703 A deprecated synonym for @option{-mtune}.
11705 @item -mfpmath=@var{unit}
11707 Generate floating point arithmetics for selected unit @var{unit}. The choices
11708 for @var{unit} are:
11712 Use the standard 387 floating point coprocessor present majority of chips and
11713 emulated otherwise. Code compiled with this option will run almost everywhere.
11714 The temporary results are computed in 80bit precision instead of precision
11715 specified by the type resulting in slightly different results compared to most
11716 of other chips. See @option{-ffloat-store} for more detailed description.
11718 This is the default choice for i386 compiler.
11721 Use scalar floating point instructions present in the SSE instruction set.
11722 This instruction set is supported by Pentium3 and newer chips, in the AMD line
11723 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
11724 instruction set supports only single precision arithmetics, thus the double and
11725 extended precision arithmetics is still done using 387. Later version, present
11726 only in Pentium4 and the future AMD x86-64 chips supports double precision
11729 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
11730 or @option{-msse2} switches to enable SSE extensions and make this option
11731 effective. For the x86-64 compiler, these extensions are enabled by default.
11733 The resulting code should be considerably faster in the majority of cases and avoid
11734 the numerical instability problems of 387 code, but may break some existing
11735 code that expects temporaries to be 80bit.
11737 This is the default choice for the x86-64 compiler.
11742 Attempt to utilize both instruction sets at once. This effectively double the
11743 amount of available registers and on chips with separate execution units for
11744 387 and SSE the execution resources too. Use this option with care, as it is
11745 still experimental, because the GCC register allocator does not model separate
11746 functional units well resulting in instable performance.
11749 @item -masm=@var{dialect}
11750 @opindex masm=@var{dialect}
11751 Output asm instructions using selected @var{dialect}. Supported
11752 choices are @samp{intel} or @samp{att} (the default one). Darwin does
11753 not support @samp{intel}.
11756 @itemx -mno-ieee-fp
11758 @opindex mno-ieee-fp
11759 Control whether or not the compiler uses IEEE floating point
11760 comparisons. These handle correctly the case where the result of a
11761 comparison is unordered.
11764 @opindex msoft-float
11765 Generate output containing library calls for floating point.
11766 @strong{Warning:} the requisite libraries are not part of GCC@.
11767 Normally the facilities of the machine's usual C compiler are used, but
11768 this can't be done directly in cross-compilation. You must make your
11769 own arrangements to provide suitable library functions for
11772 On machines where a function returns floating point results in the 80387
11773 register stack, some floating point opcodes may be emitted even if
11774 @option{-msoft-float} is used.
11776 @item -mno-fp-ret-in-387
11777 @opindex mno-fp-ret-in-387
11778 Do not use the FPU registers for return values of functions.
11780 The usual calling convention has functions return values of types
11781 @code{float} and @code{double} in an FPU register, even if there
11782 is no FPU@. The idea is that the operating system should emulate
11785 The option @option{-mno-fp-ret-in-387} causes such values to be returned
11786 in ordinary CPU registers instead.
11788 @item -mno-fancy-math-387
11789 @opindex mno-fancy-math-387
11790 Some 387 emulators do not support the @code{sin}, @code{cos} and
11791 @code{sqrt} instructions for the 387. Specify this option to avoid
11792 generating those instructions. This option is the default on FreeBSD,
11793 OpenBSD and NetBSD@. This option is overridden when @option{-march}
11794 indicates that the target cpu will always have an FPU and so the
11795 instruction will not need emulation. As of revision 2.6.1, these
11796 instructions are not generated unless you also use the
11797 @option{-funsafe-math-optimizations} switch.
11799 @item -malign-double
11800 @itemx -mno-align-double
11801 @opindex malign-double
11802 @opindex mno-align-double
11803 Control whether GCC aligns @code{double}, @code{long double}, and
11804 @code{long long} variables on a two word boundary or a one word
11805 boundary. Aligning @code{double} variables on a two word boundary will
11806 produce code that runs somewhat faster on a @samp{Pentium} at the
11807 expense of more memory.
11809 On x86-64, @option{-malign-double} is enabled by default.
11811 @strong{Warning:} if you use the @option{-malign-double} switch,
11812 structures containing the above types will be aligned differently than
11813 the published application binary interface specifications for the 386
11814 and will not be binary compatible with structures in code compiled
11815 without that switch.
11817 @item -m96bit-long-double
11818 @itemx -m128bit-long-double
11819 @opindex m96bit-long-double
11820 @opindex m128bit-long-double
11821 These switches control the size of @code{long double} type. The i386
11822 application binary interface specifies the size to be 96 bits,
11823 so @option{-m96bit-long-double} is the default in 32 bit mode.
11825 Modern architectures (Pentium and newer) would prefer @code{long double}
11826 to be aligned to an 8 or 16 byte boundary. In arrays or structures
11827 conforming to the ABI, this would not be possible. So specifying a
11828 @option{-m128bit-long-double} will align @code{long double}
11829 to a 16 byte boundary by padding the @code{long double} with an additional
11832 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
11833 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
11835 Notice that neither of these options enable any extra precision over the x87
11836 standard of 80 bits for a @code{long double}.
11838 @strong{Warning:} if you override the default value for your target ABI, the
11839 structures and arrays containing @code{long double} variables will change
11840 their size as well as function calling convention for function taking
11841 @code{long double} will be modified. Hence they will not be binary
11842 compatible with arrays or structures in code compiled without that switch.
11844 @item -mlarge-data-threshold=@var{number}
11845 @opindex mlarge-data-threshold=@var{number}
11846 When @option{-mcmodel=medium} is specified, the data greater than
11847 @var{threshold} are placed in large data section. This value must be the
11848 same across all object linked into the binary and defaults to 65535.
11852 Use a different function-calling convention, in which functions that
11853 take a fixed number of arguments return with the @code{ret} @var{num}
11854 instruction, which pops their arguments while returning. This saves one
11855 instruction in the caller since there is no need to pop the arguments
11858 You can specify that an individual function is called with this calling
11859 sequence with the function attribute @samp{stdcall}. You can also
11860 override the @option{-mrtd} option by using the function attribute
11861 @samp{cdecl}. @xref{Function Attributes}.
11863 @strong{Warning:} this calling convention is incompatible with the one
11864 normally used on Unix, so you cannot use it if you need to call
11865 libraries compiled with the Unix compiler.
11867 Also, you must provide function prototypes for all functions that
11868 take variable numbers of arguments (including @code{printf});
11869 otherwise incorrect code will be generated for calls to those
11872 In addition, seriously incorrect code will result if you call a
11873 function with too many arguments. (Normally, extra arguments are
11874 harmlessly ignored.)
11876 @item -mregparm=@var{num}
11878 Control how many registers are used to pass integer arguments. By
11879 default, no registers are used to pass arguments, and at most 3
11880 registers can be used. You can control this behavior for a specific
11881 function by using the function attribute @samp{regparm}.
11882 @xref{Function Attributes}.
11884 @strong{Warning:} if you use this switch, and
11885 @var{num} is nonzero, then you must build all modules with the same
11886 value, including any libraries. This includes the system libraries and
11890 @opindex msseregparm
11891 Use SSE register passing conventions for float and double arguments
11892 and return values. You can control this behavior for a specific
11893 function by using the function attribute @samp{sseregparm}.
11894 @xref{Function Attributes}.
11896 @strong{Warning:} if you use this switch then you must build all
11897 modules with the same value, including any libraries. This includes
11898 the system libraries and startup modules.
11907 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
11908 is specified, the significands of results of floating-point operations are
11909 rounded to 24 bits (single precision); @option{-mpc64} rounds the
11910 significands of results of floating-point operations to 53 bits (double
11911 precision) and @option{-mpc80} rounds the significands of results of
11912 floating-point operations to 64 bits (extended double precision), which is
11913 the default. When this option is used, floating-point operations in higher
11914 precisions are not available to the programmer without setting the FPU
11915 control word explicitly.
11917 Setting the rounding of floating-point operations to less than the default
11918 80 bits can speed some programs by 2% or more. Note that some mathematical
11919 libraries assume that extended precision (80 bit) floating-point operations
11920 are enabled by default; routines in such libraries could suffer significant
11921 loss of accuracy, typically through so-called "catastrophic cancellation",
11922 when this option is used to set the precision to less than extended precision.
11924 @item -mstackrealign
11925 @opindex mstackrealign
11926 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
11927 option will generate an alternate prologue and epilogue that realigns the
11928 runtime stack if necessary. This supports mixing legacy codes that keep
11929 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
11930 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
11931 applicable to individual functions.
11933 @item -mpreferred-stack-boundary=@var{num}
11934 @opindex mpreferred-stack-boundary
11935 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
11936 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
11937 the default is 4 (16 bytes or 128 bits).
11939 @item -mincoming-stack-boundary=@var{num}
11940 @opindex mincoming-stack-boundary
11941 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
11942 boundary. If @option{-mincoming-stack-boundary} is not specified,
11943 the one specified by @option{-mpreferred-stack-boundary} will be used.
11945 On Pentium and PentiumPro, @code{double} and @code{long double} values
11946 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
11947 suffer significant run time performance penalties. On Pentium III, the
11948 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
11949 properly if it is not 16 byte aligned.
11951 To ensure proper alignment of this values on the stack, the stack boundary
11952 must be as aligned as that required by any value stored on the stack.
11953 Further, every function must be generated such that it keeps the stack
11954 aligned. Thus calling a function compiled with a higher preferred
11955 stack boundary from a function compiled with a lower preferred stack
11956 boundary will most likely misalign the stack. It is recommended that
11957 libraries that use callbacks always use the default setting.
11959 This extra alignment does consume extra stack space, and generally
11960 increases code size. Code that is sensitive to stack space usage, such
11961 as embedded systems and operating system kernels, may want to reduce the
11962 preferred alignment to @option{-mpreferred-stack-boundary=2}.
12002 These switches enable or disable the use of instructions in the MMX,
12003 SSE, SSE2, SSE3, SSSE3, SSE4.1, AVX, AES, PCLMUL, SSE4A, FMA4, ABM or
12004 3DNow!@: extended instruction sets.
12005 These extensions are also available as built-in functions: see
12006 @ref{X86 Built-in Functions}, for details of the functions enabled and
12007 disabled by these switches.
12009 To have SSE/SSE2 instructions generated automatically from floating-point
12010 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
12012 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
12013 generates new AVX instructions or AVX equivalence for all SSEx instructions
12016 These options will enable GCC to use these extended instructions in
12017 generated code, even without @option{-mfpmath=sse}. Applications which
12018 perform runtime CPU detection must compile separate files for each
12019 supported architecture, using the appropriate flags. In particular,
12020 the file containing the CPU detection code should be compiled without
12025 This option instructs GCC to emit a @code{cld} instruction in the prologue
12026 of functions that use string instructions. String instructions depend on
12027 the DF flag to select between autoincrement or autodecrement mode. While the
12028 ABI specifies the DF flag to be cleared on function entry, some operating
12029 systems violate this specification by not clearing the DF flag in their
12030 exception dispatchers. The exception handler can be invoked with the DF flag
12031 set which leads to wrong direction mode, when string instructions are used.
12032 This option can be enabled by default on 32-bit x86 targets by configuring
12033 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
12034 instructions can be suppressed with the @option{-mno-cld} compiler option
12039 This option will enable GCC to use CMPXCHG16B instruction in generated code.
12040 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
12041 data types. This is useful for high resolution counters that could be updated
12042 by multiple processors (or cores). This instruction is generated as part of
12043 atomic built-in functions: see @ref{Atomic Builtins} for details.
12047 This option will enable GCC to use SAHF instruction in generated 64-bit code.
12048 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
12049 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
12050 SAHF are load and store instructions, respectively, for certain status flags.
12051 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
12052 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
12056 This option will enable GCC to use movbe instruction to implement
12057 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
12061 This option will enable built-in functions, @code{__builtin_ia32_crc32qi},
12062 @code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and
12063 @code{__builtin_ia32_crc32di} to generate the crc32 machine instruction.
12067 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
12068 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
12069 to increase precision instead of DIVSS and SQRTSS (and their vectorized
12070 variants) for single precision floating point arguments. These instructions
12071 are generated only when @option{-funsafe-math-optimizations} is enabled
12072 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
12073 Note that while the throughput of the sequence is higher than the throughput
12074 of the non-reciprocal instruction, the precision of the sequence can be
12075 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
12077 @item -mveclibabi=@var{type}
12078 @opindex mveclibabi
12079 Specifies the ABI type to use for vectorizing intrinsics using an
12080 external library. Supported types are @code{svml} for the Intel short
12081 vector math library and @code{acml} for the AMD math core library style
12082 of interfacing. GCC will currently emit calls to @code{vmldExp2},
12083 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
12084 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
12085 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
12086 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
12087 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
12088 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
12089 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
12090 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
12091 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
12092 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
12093 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
12094 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
12095 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
12096 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
12097 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
12098 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
12099 compatible library will have to be specified at link time.
12101 @item -mabi=@var{name}
12103 Generate code for the specified calling convention. Permissible values
12104 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
12105 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
12106 ABI when targeting Windows. On all other systems, the default is the
12107 SYSV ABI. You can control this behavior for a specific function by
12108 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
12109 @xref{Function Attributes}.
12112 @itemx -mno-push-args
12113 @opindex mpush-args
12114 @opindex mno-push-args
12115 Use PUSH operations to store outgoing parameters. This method is shorter
12116 and usually equally fast as method using SUB/MOV operations and is enabled
12117 by default. In some cases disabling it may improve performance because of
12118 improved scheduling and reduced dependencies.
12120 @item -maccumulate-outgoing-args
12121 @opindex maccumulate-outgoing-args
12122 If enabled, the maximum amount of space required for outgoing arguments will be
12123 computed in the function prologue. This is faster on most modern CPUs
12124 because of reduced dependencies, improved scheduling and reduced stack usage
12125 when preferred stack boundary is not equal to 2. The drawback is a notable
12126 increase in code size. This switch implies @option{-mno-push-args}.
12130 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
12131 on thread-safe exception handling must compile and link all code with the
12132 @option{-mthreads} option. When compiling, @option{-mthreads} defines
12133 @option{-D_MT}; when linking, it links in a special thread helper library
12134 @option{-lmingwthrd} which cleans up per thread exception handling data.
12136 @item -mno-align-stringops
12137 @opindex mno-align-stringops
12138 Do not align destination of inlined string operations. This switch reduces
12139 code size and improves performance in case the destination is already aligned,
12140 but GCC doesn't know about it.
12142 @item -minline-all-stringops
12143 @opindex minline-all-stringops
12144 By default GCC inlines string operations only when destination is known to be
12145 aligned at least to 4 byte boundary. This enables more inlining, increase code
12146 size, but may improve performance of code that depends on fast memcpy, strlen
12147 and memset for short lengths.
12149 @item -minline-stringops-dynamically
12150 @opindex minline-stringops-dynamically
12151 For string operation of unknown size, inline runtime checks so for small
12152 blocks inline code is used, while for large blocks library call is used.
12154 @item -mstringop-strategy=@var{alg}
12155 @opindex mstringop-strategy=@var{alg}
12156 Overwrite internal decision heuristic about particular algorithm to inline
12157 string operation with. The allowed values are @code{rep_byte},
12158 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
12159 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
12160 expanding inline loop, @code{libcall} for always expanding library call.
12162 @item -momit-leaf-frame-pointer
12163 @opindex momit-leaf-frame-pointer
12164 Don't keep the frame pointer in a register for leaf functions. This
12165 avoids the instructions to save, set up and restore frame pointers and
12166 makes an extra register available in leaf functions. The option
12167 @option{-fomit-frame-pointer} removes the frame pointer for all functions
12168 which might make debugging harder.
12170 @item -mtls-direct-seg-refs
12171 @itemx -mno-tls-direct-seg-refs
12172 @opindex mtls-direct-seg-refs
12173 Controls whether TLS variables may be accessed with offsets from the
12174 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
12175 or whether the thread base pointer must be added. Whether or not this
12176 is legal depends on the operating system, and whether it maps the
12177 segment to cover the entire TLS area.
12179 For systems that use GNU libc, the default is on.
12182 @itemx -mno-sse2avx
12184 Specify that the assembler should encode SSE instructions with VEX
12185 prefix. The option @option{-mavx} turns this on by default.
12188 These @samp{-m} switches are supported in addition to the above
12189 on AMD x86-64 processors in 64-bit environments.
12196 Generate code for a 32-bit or 64-bit environment.
12197 The 32-bit environment sets int, long and pointer to 32 bits and
12198 generates code that runs on any i386 system.
12199 The 64-bit environment sets int to 32 bits and long and pointer
12200 to 64 bits and generates code for AMD's x86-64 architecture. For
12201 darwin only the -m64 option turns off the @option{-fno-pic} and
12202 @option{-mdynamic-no-pic} options.
12204 @item -mno-red-zone
12205 @opindex mno-red-zone
12206 Do not use a so called red zone for x86-64 code. The red zone is mandated
12207 by the x86-64 ABI, it is a 128-byte area beyond the location of the
12208 stack pointer that will not be modified by signal or interrupt handlers
12209 and therefore can be used for temporary data without adjusting the stack
12210 pointer. The flag @option{-mno-red-zone} disables this red zone.
12212 @item -mcmodel=small
12213 @opindex mcmodel=small
12214 Generate code for the small code model: the program and its symbols must
12215 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
12216 Programs can be statically or dynamically linked. This is the default
12219 @item -mcmodel=kernel
12220 @opindex mcmodel=kernel
12221 Generate code for the kernel code model. The kernel runs in the
12222 negative 2 GB of the address space.
12223 This model has to be used for Linux kernel code.
12225 @item -mcmodel=medium
12226 @opindex mcmodel=medium
12227 Generate code for the medium model: The program is linked in the lower 2
12228 GB of the address space. Small symbols are also placed there. Symbols
12229 with sizes larger than @option{-mlarge-data-threshold} are put into
12230 large data or bss sections and can be located above 2GB. Programs can
12231 be statically or dynamically linked.
12233 @item -mcmodel=large
12234 @opindex mcmodel=large
12235 Generate code for the large model: This model makes no assumptions
12236 about addresses and sizes of sections.
12239 @node IA-64 Options
12240 @subsection IA-64 Options
12241 @cindex IA-64 Options
12243 These are the @samp{-m} options defined for the Intel IA-64 architecture.
12247 @opindex mbig-endian
12248 Generate code for a big endian target. This is the default for HP-UX@.
12250 @item -mlittle-endian
12251 @opindex mlittle-endian
12252 Generate code for a little endian target. This is the default for AIX5
12258 @opindex mno-gnu-as
12259 Generate (or don't) code for the GNU assembler. This is the default.
12260 @c Also, this is the default if the configure option @option{--with-gnu-as}
12266 @opindex mno-gnu-ld
12267 Generate (or don't) code for the GNU linker. This is the default.
12268 @c Also, this is the default if the configure option @option{--with-gnu-ld}
12273 Generate code that does not use a global pointer register. The result
12274 is not position independent code, and violates the IA-64 ABI@.
12276 @item -mvolatile-asm-stop
12277 @itemx -mno-volatile-asm-stop
12278 @opindex mvolatile-asm-stop
12279 @opindex mno-volatile-asm-stop
12280 Generate (or don't) a stop bit immediately before and after volatile asm
12283 @item -mregister-names
12284 @itemx -mno-register-names
12285 @opindex mregister-names
12286 @opindex mno-register-names
12287 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
12288 the stacked registers. This may make assembler output more readable.
12294 Disable (or enable) optimizations that use the small data section. This may
12295 be useful for working around optimizer bugs.
12297 @item -mconstant-gp
12298 @opindex mconstant-gp
12299 Generate code that uses a single constant global pointer value. This is
12300 useful when compiling kernel code.
12304 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
12305 This is useful when compiling firmware code.
12307 @item -minline-float-divide-min-latency
12308 @opindex minline-float-divide-min-latency
12309 Generate code for inline divides of floating point values
12310 using the minimum latency algorithm.
12312 @item -minline-float-divide-max-throughput
12313 @opindex minline-float-divide-max-throughput
12314 Generate code for inline divides of floating point values
12315 using the maximum throughput algorithm.
12317 @item -mno-inline-float-divide
12318 @opindex mno-inline-float-divide
12319 Do not generate inline code for divides of floating point values.
12321 @item -minline-int-divide-min-latency
12322 @opindex minline-int-divide-min-latency
12323 Generate code for inline divides of integer values
12324 using the minimum latency algorithm.
12326 @item -minline-int-divide-max-throughput
12327 @opindex minline-int-divide-max-throughput
12328 Generate code for inline divides of integer values
12329 using the maximum throughput algorithm.
12331 @item -mno-inline-int-divide
12332 @opindex mno-inline-int-divide
12333 Do not generate inline code for divides of integer values.
12335 @item -minline-sqrt-min-latency
12336 @opindex minline-sqrt-min-latency
12337 Generate code for inline square roots
12338 using the minimum latency algorithm.
12340 @item -minline-sqrt-max-throughput
12341 @opindex minline-sqrt-max-throughput
12342 Generate code for inline square roots
12343 using the maximum throughput algorithm.
12345 @item -mno-inline-sqrt
12346 @opindex mno-inline-sqrt
12347 Do not generate inline code for sqrt.
12350 @itemx -mno-fused-madd
12351 @opindex mfused-madd
12352 @opindex mno-fused-madd
12353 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
12354 instructions. The default is to use these instructions.
12356 @item -mno-dwarf2-asm
12357 @itemx -mdwarf2-asm
12358 @opindex mno-dwarf2-asm
12359 @opindex mdwarf2-asm
12360 Don't (or do) generate assembler code for the DWARF2 line number debugging
12361 info. This may be useful when not using the GNU assembler.
12363 @item -mearly-stop-bits
12364 @itemx -mno-early-stop-bits
12365 @opindex mearly-stop-bits
12366 @opindex mno-early-stop-bits
12367 Allow stop bits to be placed earlier than immediately preceding the
12368 instruction that triggered the stop bit. This can improve instruction
12369 scheduling, but does not always do so.
12371 @item -mfixed-range=@var{register-range}
12372 @opindex mfixed-range
12373 Generate code treating the given register range as fixed registers.
12374 A fixed register is one that the register allocator can not use. This is
12375 useful when compiling kernel code. A register range is specified as
12376 two registers separated by a dash. Multiple register ranges can be
12377 specified separated by a comma.
12379 @item -mtls-size=@var{tls-size}
12381 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
12384 @item -mtune=@var{cpu-type}
12386 Tune the instruction scheduling for a particular CPU, Valid values are
12387 itanium, itanium1, merced, itanium2, and mckinley.
12393 Generate code for a 32-bit or 64-bit environment.
12394 The 32-bit environment sets int, long and pointer to 32 bits.
12395 The 64-bit environment sets int to 32 bits and long and pointer
12396 to 64 bits. These are HP-UX specific flags.
12398 @item -mno-sched-br-data-spec
12399 @itemx -msched-br-data-spec
12400 @opindex mno-sched-br-data-spec
12401 @opindex msched-br-data-spec
12402 (Dis/En)able data speculative scheduling before reload.
12403 This will result in generation of the ld.a instructions and
12404 the corresponding check instructions (ld.c / chk.a).
12405 The default is 'disable'.
12407 @item -msched-ar-data-spec
12408 @itemx -mno-sched-ar-data-spec
12409 @opindex msched-ar-data-spec
12410 @opindex mno-sched-ar-data-spec
12411 (En/Dis)able data speculative scheduling after reload.
12412 This will result in generation of the ld.a instructions and
12413 the corresponding check instructions (ld.c / chk.a).
12414 The default is 'enable'.
12416 @item -mno-sched-control-spec
12417 @itemx -msched-control-spec
12418 @opindex mno-sched-control-spec
12419 @opindex msched-control-spec
12420 (Dis/En)able control speculative scheduling. This feature is
12421 available only during region scheduling (i.e.@: before reload).
12422 This will result in generation of the ld.s instructions and
12423 the corresponding check instructions chk.s .
12424 The default is 'disable'.
12426 @item -msched-br-in-data-spec
12427 @itemx -mno-sched-br-in-data-spec
12428 @opindex msched-br-in-data-spec
12429 @opindex mno-sched-br-in-data-spec
12430 (En/Dis)able speculative scheduling of the instructions that
12431 are dependent on the data speculative loads before reload.
12432 This is effective only with @option{-msched-br-data-spec} enabled.
12433 The default is 'enable'.
12435 @item -msched-ar-in-data-spec
12436 @itemx -mno-sched-ar-in-data-spec
12437 @opindex msched-ar-in-data-spec
12438 @opindex mno-sched-ar-in-data-spec
12439 (En/Dis)able speculative scheduling of the instructions that
12440 are dependent on the data speculative loads after reload.
12441 This is effective only with @option{-msched-ar-data-spec} enabled.
12442 The default is 'enable'.
12444 @item -msched-in-control-spec
12445 @itemx -mno-sched-in-control-spec
12446 @opindex msched-in-control-spec
12447 @opindex mno-sched-in-control-spec
12448 (En/Dis)able speculative scheduling of the instructions that
12449 are dependent on the control speculative loads.
12450 This is effective only with @option{-msched-control-spec} enabled.
12451 The default is 'enable'.
12453 @item -mno-sched-prefer-non-data-spec-insns
12454 @itemx -msched-prefer-non-data-spec-insns
12455 @opindex mno-sched-prefer-non-data-spec-insns
12456 @opindex msched-prefer-non-data-spec-insns
12457 If enabled, data speculative instructions will be chosen for schedule
12458 only if there are no other choices at the moment. This will make
12459 the use of the data speculation much more conservative.
12460 The default is 'disable'.
12462 @item -mno-sched-prefer-non-control-spec-insns
12463 @itemx -msched-prefer-non-control-spec-insns
12464 @opindex mno-sched-prefer-non-control-spec-insns
12465 @opindex msched-prefer-non-control-spec-insns
12466 If enabled, control speculative instructions will be chosen for schedule
12467 only if there are no other choices at the moment. This will make
12468 the use of the control speculation much more conservative.
12469 The default is 'disable'.
12471 @item -mno-sched-count-spec-in-critical-path
12472 @itemx -msched-count-spec-in-critical-path
12473 @opindex mno-sched-count-spec-in-critical-path
12474 @opindex msched-count-spec-in-critical-path
12475 If enabled, speculative dependencies will be considered during
12476 computation of the instructions priorities. This will make the use of the
12477 speculation a bit more conservative.
12478 The default is 'disable'.
12480 @item -msched-spec-ldc
12481 @opindex msched-spec-ldc
12482 Use a simple data speculation check. This option is on by default.
12484 @item -msched-control-spec-ldc
12485 @opindex msched-spec-ldc
12486 Use a simple check for control speculation. This option is on by default.
12488 @item -msched-stop-bits-after-every-cycle
12489 @opindex msched-stop-bits-after-every-cycle
12490 Place a stop bit after every cycle when scheduling. This option is on
12493 @item -msched-fp-mem-deps-zero-cost
12494 @opindex msched-fp-mem-deps-zero-cost
12495 Assume that floating-point stores and loads are not likely to cause a conflict
12496 when placed into the same instruction group. This option is disabled by
12499 @item -msel-sched-dont-check-control-spec
12500 @opindex msel-sched-dont-check-control-spec
12501 Generate checks for control speculation in selective scheduling.
12502 This flag is disabled by default.
12504 @item -msched-max-memory-insns=@var{max-insns}
12505 @opindex msched-max-memory-insns
12506 Limit on the number of memory insns per instruction group, giving lower
12507 priority to subsequent memory insns attempting to schedule in the same
12508 instruction group. Frequently useful to prevent cache bank conflicts.
12509 The default value is 1.
12511 @item -msched-max-memory-insns-hard-limit
12512 @opindex msched-max-memory-insns-hard-limit
12513 Disallow more than `msched-max-memory-insns' in instruction group.
12514 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
12515 when limit is reached but may still schedule memory operations.
12519 @node IA-64/VMS Options
12520 @subsection IA-64/VMS Options
12522 These @samp{-m} options are defined for the IA-64/VMS implementations:
12525 @item -mvms-return-codes
12526 @opindex mvms-return-codes
12527 Return VMS condition codes from main. The default is to return POSIX
12528 style condition (e.g.@ error) codes.
12530 @item -mdebug-main=@var{prefix}
12531 @opindex mdebug-main=@var{prefix}
12532 Flag the first routine whose name starts with @var{prefix} as the main
12533 routine for the debugger.
12537 Default to 64bit memory allocation routines.
12541 @subsection M32C Options
12542 @cindex M32C options
12545 @item -mcpu=@var{name}
12547 Select the CPU for which code is generated. @var{name} may be one of
12548 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
12549 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
12550 the M32C/80 series.
12554 Specifies that the program will be run on the simulator. This causes
12555 an alternate runtime library to be linked in which supports, for
12556 example, file I/O@. You must not use this option when generating
12557 programs that will run on real hardware; you must provide your own
12558 runtime library for whatever I/O functions are needed.
12560 @item -memregs=@var{number}
12562 Specifies the number of memory-based pseudo-registers GCC will use
12563 during code generation. These pseudo-registers will be used like real
12564 registers, so there is a tradeoff between GCC's ability to fit the
12565 code into available registers, and the performance penalty of using
12566 memory instead of registers. Note that all modules in a program must
12567 be compiled with the same value for this option. Because of that, you
12568 must not use this option with the default runtime libraries gcc
12573 @node M32R/D Options
12574 @subsection M32R/D Options
12575 @cindex M32R/D options
12577 These @option{-m} options are defined for Renesas M32R/D architectures:
12582 Generate code for the M32R/2@.
12586 Generate code for the M32R/X@.
12590 Generate code for the M32R@. This is the default.
12592 @item -mmodel=small
12593 @opindex mmodel=small
12594 Assume all objects live in the lower 16MB of memory (so that their addresses
12595 can be loaded with the @code{ld24} instruction), and assume all subroutines
12596 are reachable with the @code{bl} instruction.
12597 This is the default.
12599 The addressability of a particular object can be set with the
12600 @code{model} attribute.
12602 @item -mmodel=medium
12603 @opindex mmodel=medium
12604 Assume objects may be anywhere in the 32-bit address space (the compiler
12605 will generate @code{seth/add3} instructions to load their addresses), and
12606 assume all subroutines are reachable with the @code{bl} instruction.
12608 @item -mmodel=large
12609 @opindex mmodel=large
12610 Assume objects may be anywhere in the 32-bit address space (the compiler
12611 will generate @code{seth/add3} instructions to load their addresses), and
12612 assume subroutines may not be reachable with the @code{bl} instruction
12613 (the compiler will generate the much slower @code{seth/add3/jl}
12614 instruction sequence).
12617 @opindex msdata=none
12618 Disable use of the small data area. Variables will be put into
12619 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
12620 @code{section} attribute has been specified).
12621 This is the default.
12623 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
12624 Objects may be explicitly put in the small data area with the
12625 @code{section} attribute using one of these sections.
12627 @item -msdata=sdata
12628 @opindex msdata=sdata
12629 Put small global and static data in the small data area, but do not
12630 generate special code to reference them.
12633 @opindex msdata=use
12634 Put small global and static data in the small data area, and generate
12635 special instructions to reference them.
12639 @cindex smaller data references
12640 Put global and static objects less than or equal to @var{num} bytes
12641 into the small data or bss sections instead of the normal data or bss
12642 sections. The default value of @var{num} is 8.
12643 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
12644 for this option to have any effect.
12646 All modules should be compiled with the same @option{-G @var{num}} value.
12647 Compiling with different values of @var{num} may or may not work; if it
12648 doesn't the linker will give an error message---incorrect code will not be
12653 Makes the M32R specific code in the compiler display some statistics
12654 that might help in debugging programs.
12656 @item -malign-loops
12657 @opindex malign-loops
12658 Align all loops to a 32-byte boundary.
12660 @item -mno-align-loops
12661 @opindex mno-align-loops
12662 Do not enforce a 32-byte alignment for loops. This is the default.
12664 @item -missue-rate=@var{number}
12665 @opindex missue-rate=@var{number}
12666 Issue @var{number} instructions per cycle. @var{number} can only be 1
12669 @item -mbranch-cost=@var{number}
12670 @opindex mbranch-cost=@var{number}
12671 @var{number} can only be 1 or 2. If it is 1 then branches will be
12672 preferred over conditional code, if it is 2, then the opposite will
12675 @item -mflush-trap=@var{number}
12676 @opindex mflush-trap=@var{number}
12677 Specifies the trap number to use to flush the cache. The default is
12678 12. Valid numbers are between 0 and 15 inclusive.
12680 @item -mno-flush-trap
12681 @opindex mno-flush-trap
12682 Specifies that the cache cannot be flushed by using a trap.
12684 @item -mflush-func=@var{name}
12685 @opindex mflush-func=@var{name}
12686 Specifies the name of the operating system function to call to flush
12687 the cache. The default is @emph{_flush_cache}, but a function call
12688 will only be used if a trap is not available.
12690 @item -mno-flush-func
12691 @opindex mno-flush-func
12692 Indicates that there is no OS function for flushing the cache.
12696 @node M680x0 Options
12697 @subsection M680x0 Options
12698 @cindex M680x0 options
12700 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
12701 The default settings depend on which architecture was selected when
12702 the compiler was configured; the defaults for the most common choices
12706 @item -march=@var{arch}
12708 Generate code for a specific M680x0 or ColdFire instruction set
12709 architecture. Permissible values of @var{arch} for M680x0
12710 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
12711 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
12712 architectures are selected according to Freescale's ISA classification
12713 and the permissible values are: @samp{isaa}, @samp{isaaplus},
12714 @samp{isab} and @samp{isac}.
12716 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
12717 code for a ColdFire target. The @var{arch} in this macro is one of the
12718 @option{-march} arguments given above.
12720 When used together, @option{-march} and @option{-mtune} select code
12721 that runs on a family of similar processors but that is optimized
12722 for a particular microarchitecture.
12724 @item -mcpu=@var{cpu}
12726 Generate code for a specific M680x0 or ColdFire processor.
12727 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
12728 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
12729 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
12730 below, which also classifies the CPUs into families:
12732 @multitable @columnfractions 0.20 0.80
12733 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
12734 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
12735 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
12736 @item @samp{5206e} @tab @samp{5206e}
12737 @item @samp{5208} @tab @samp{5207} @samp{5208}
12738 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
12739 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
12740 @item @samp{5216} @tab @samp{5214} @samp{5216}
12741 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
12742 @item @samp{5225} @tab @samp{5224} @samp{5225}
12743 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
12744 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
12745 @item @samp{5249} @tab @samp{5249}
12746 @item @samp{5250} @tab @samp{5250}
12747 @item @samp{5271} @tab @samp{5270} @samp{5271}
12748 @item @samp{5272} @tab @samp{5272}
12749 @item @samp{5275} @tab @samp{5274} @samp{5275}
12750 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
12751 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
12752 @item @samp{5307} @tab @samp{5307}
12753 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
12754 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
12755 @item @samp{5407} @tab @samp{5407}
12756 @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}
12759 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
12760 @var{arch} is compatible with @var{cpu}. Other combinations of
12761 @option{-mcpu} and @option{-march} are rejected.
12763 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
12764 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
12765 where the value of @var{family} is given by the table above.
12767 @item -mtune=@var{tune}
12769 Tune the code for a particular microarchitecture, within the
12770 constraints set by @option{-march} and @option{-mcpu}.
12771 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
12772 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
12773 and @samp{cpu32}. The ColdFire microarchitectures
12774 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
12776 You can also use @option{-mtune=68020-40} for code that needs
12777 to run relatively well on 68020, 68030 and 68040 targets.
12778 @option{-mtune=68020-60} is similar but includes 68060 targets
12779 as well. These two options select the same tuning decisions as
12780 @option{-m68020-40} and @option{-m68020-60} respectively.
12782 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
12783 when tuning for 680x0 architecture @var{arch}. It also defines
12784 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
12785 option is used. If gcc is tuning for a range of architectures,
12786 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
12787 it defines the macros for every architecture in the range.
12789 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
12790 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
12791 of the arguments given above.
12797 Generate output for a 68000. This is the default
12798 when the compiler is configured for 68000-based systems.
12799 It is equivalent to @option{-march=68000}.
12801 Use this option for microcontrollers with a 68000 or EC000 core,
12802 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
12806 Generate output for a 68010. This is the default
12807 when the compiler is configured for 68010-based systems.
12808 It is equivalent to @option{-march=68010}.
12814 Generate output for a 68020. This is the default
12815 when the compiler is configured for 68020-based systems.
12816 It is equivalent to @option{-march=68020}.
12820 Generate output for a 68030. This is the default when the compiler is
12821 configured for 68030-based systems. It is equivalent to
12822 @option{-march=68030}.
12826 Generate output for a 68040. This is the default when the compiler is
12827 configured for 68040-based systems. It is equivalent to
12828 @option{-march=68040}.
12830 This option inhibits the use of 68881/68882 instructions that have to be
12831 emulated by software on the 68040. Use this option if your 68040 does not
12832 have code to emulate those instructions.
12836 Generate output for a 68060. This is the default when the compiler is
12837 configured for 68060-based systems. It is equivalent to
12838 @option{-march=68060}.
12840 This option inhibits the use of 68020 and 68881/68882 instructions that
12841 have to be emulated by software on the 68060. Use this option if your 68060
12842 does not have code to emulate those instructions.
12846 Generate output for a CPU32. This is the default
12847 when the compiler is configured for CPU32-based systems.
12848 It is equivalent to @option{-march=cpu32}.
12850 Use this option for microcontrollers with a
12851 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
12852 68336, 68340, 68341, 68349 and 68360.
12856 Generate output for a 520X ColdFire CPU@. This is the default
12857 when the compiler is configured for 520X-based systems.
12858 It is equivalent to @option{-mcpu=5206}, and is now deprecated
12859 in favor of that option.
12861 Use this option for microcontroller with a 5200 core, including
12862 the MCF5202, MCF5203, MCF5204 and MCF5206.
12866 Generate output for a 5206e ColdFire CPU@. The option is now
12867 deprecated in favor of the equivalent @option{-mcpu=5206e}.
12871 Generate output for a member of the ColdFire 528X family.
12872 The option is now deprecated in favor of the equivalent
12873 @option{-mcpu=528x}.
12877 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
12878 in favor of the equivalent @option{-mcpu=5307}.
12882 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
12883 in favor of the equivalent @option{-mcpu=5407}.
12887 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
12888 This includes use of hardware floating point instructions.
12889 The option is equivalent to @option{-mcpu=547x}, and is now
12890 deprecated in favor of that option.
12894 Generate output for a 68040, without using any of the new instructions.
12895 This results in code which can run relatively efficiently on either a
12896 68020/68881 or a 68030 or a 68040. The generated code does use the
12897 68881 instructions that are emulated on the 68040.
12899 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
12903 Generate output for a 68060, without using any of the new instructions.
12904 This results in code which can run relatively efficiently on either a
12905 68020/68881 or a 68030 or a 68040. The generated code does use the
12906 68881 instructions that are emulated on the 68060.
12908 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
12912 @opindex mhard-float
12914 Generate floating-point instructions. This is the default for 68020
12915 and above, and for ColdFire devices that have an FPU@. It defines the
12916 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
12917 on ColdFire targets.
12920 @opindex msoft-float
12921 Do not generate floating-point instructions; use library calls instead.
12922 This is the default for 68000, 68010, and 68832 targets. It is also
12923 the default for ColdFire devices that have no FPU.
12929 Generate (do not generate) ColdFire hardware divide and remainder
12930 instructions. If @option{-march} is used without @option{-mcpu},
12931 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
12932 architectures. Otherwise, the default is taken from the target CPU
12933 (either the default CPU, or the one specified by @option{-mcpu}). For
12934 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
12935 @option{-mcpu=5206e}.
12937 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
12941 Consider type @code{int} to be 16 bits wide, like @code{short int}.
12942 Additionally, parameters passed on the stack are also aligned to a
12943 16-bit boundary even on targets whose API mandates promotion to 32-bit.
12947 Do not consider type @code{int} to be 16 bits wide. This is the default.
12950 @itemx -mno-bitfield
12951 @opindex mnobitfield
12952 @opindex mno-bitfield
12953 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
12954 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
12958 Do use the bit-field instructions. The @option{-m68020} option implies
12959 @option{-mbitfield}. This is the default if you use a configuration
12960 designed for a 68020.
12964 Use a different function-calling convention, in which functions
12965 that take a fixed number of arguments return with the @code{rtd}
12966 instruction, which pops their arguments while returning. This
12967 saves one instruction in the caller since there is no need to pop
12968 the arguments there.
12970 This calling convention is incompatible with the one normally
12971 used on Unix, so you cannot use it if you need to call libraries
12972 compiled with the Unix compiler.
12974 Also, you must provide function prototypes for all functions that
12975 take variable numbers of arguments (including @code{printf});
12976 otherwise incorrect code will be generated for calls to those
12979 In addition, seriously incorrect code will result if you call a
12980 function with too many arguments. (Normally, extra arguments are
12981 harmlessly ignored.)
12983 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
12984 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
12988 Do not use the calling conventions selected by @option{-mrtd}.
12989 This is the default.
12992 @itemx -mno-align-int
12993 @opindex malign-int
12994 @opindex mno-align-int
12995 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
12996 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
12997 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
12998 Aligning variables on 32-bit boundaries produces code that runs somewhat
12999 faster on processors with 32-bit busses at the expense of more memory.
13001 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
13002 align structures containing the above types differently than
13003 most published application binary interface specifications for the m68k.
13007 Use the pc-relative addressing mode of the 68000 directly, instead of
13008 using a global offset table. At present, this option implies @option{-fpic},
13009 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
13010 not presently supported with @option{-mpcrel}, though this could be supported for
13011 68020 and higher processors.
13013 @item -mno-strict-align
13014 @itemx -mstrict-align
13015 @opindex mno-strict-align
13016 @opindex mstrict-align
13017 Do not (do) assume that unaligned memory references will be handled by
13021 Generate code that allows the data segment to be located in a different
13022 area of memory from the text segment. This allows for execute in place in
13023 an environment without virtual memory management. This option implies
13026 @item -mno-sep-data
13027 Generate code that assumes that the data segment follows the text segment.
13028 This is the default.
13030 @item -mid-shared-library
13031 Generate code that supports shared libraries via the library ID method.
13032 This allows for execute in place and shared libraries in an environment
13033 without virtual memory management. This option implies @option{-fPIC}.
13035 @item -mno-id-shared-library
13036 Generate code that doesn't assume ID based shared libraries are being used.
13037 This is the default.
13039 @item -mshared-library-id=n
13040 Specified the identification number of the ID based shared library being
13041 compiled. Specifying a value of 0 will generate more compact code, specifying
13042 other values will force the allocation of that number to the current
13043 library but is no more space or time efficient than omitting this option.
13049 When generating position-independent code for ColdFire, generate code
13050 that works if the GOT has more than 8192 entries. This code is
13051 larger and slower than code generated without this option. On M680x0
13052 processors, this option is not needed; @option{-fPIC} suffices.
13054 GCC normally uses a single instruction to load values from the GOT@.
13055 While this is relatively efficient, it only works if the GOT
13056 is smaller than about 64k. Anything larger causes the linker
13057 to report an error such as:
13059 @cindex relocation truncated to fit (ColdFire)
13061 relocation truncated to fit: R_68K_GOT16O foobar
13064 If this happens, you should recompile your code with @option{-mxgot}.
13065 It should then work with very large GOTs. However, code generated with
13066 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
13067 the value of a global symbol.
13069 Note that some linkers, including newer versions of the GNU linker,
13070 can create multiple GOTs and sort GOT entries. If you have such a linker,
13071 you should only need to use @option{-mxgot} when compiling a single
13072 object file that accesses more than 8192 GOT entries. Very few do.
13074 These options have no effect unless GCC is generating
13075 position-independent code.
13079 @node M68hc1x Options
13080 @subsection M68hc1x Options
13081 @cindex M68hc1x options
13083 These are the @samp{-m} options defined for the 68hc11 and 68hc12
13084 microcontrollers. The default values for these options depends on
13085 which style of microcontroller was selected when the compiler was configured;
13086 the defaults for the most common choices are given below.
13093 Generate output for a 68HC11. This is the default
13094 when the compiler is configured for 68HC11-based systems.
13100 Generate output for a 68HC12. This is the default
13101 when the compiler is configured for 68HC12-based systems.
13107 Generate output for a 68HCS12.
13109 @item -mauto-incdec
13110 @opindex mauto-incdec
13111 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
13118 Enable the use of 68HC12 min and max instructions.
13121 @itemx -mno-long-calls
13122 @opindex mlong-calls
13123 @opindex mno-long-calls
13124 Treat all calls as being far away (near). If calls are assumed to be
13125 far away, the compiler will use the @code{call} instruction to
13126 call a function and the @code{rtc} instruction for returning.
13130 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13132 @item -msoft-reg-count=@var{count}
13133 @opindex msoft-reg-count
13134 Specify the number of pseudo-soft registers which are used for the
13135 code generation. The maximum number is 32. Using more pseudo-soft
13136 register may or may not result in better code depending on the program.
13137 The default is 4 for 68HC11 and 2 for 68HC12.
13141 @node MCore Options
13142 @subsection MCore Options
13143 @cindex MCore options
13145 These are the @samp{-m} options defined for the Motorola M*Core
13151 @itemx -mno-hardlit
13153 @opindex mno-hardlit
13154 Inline constants into the code stream if it can be done in two
13155 instructions or less.
13161 Use the divide instruction. (Enabled by default).
13163 @item -mrelax-immediate
13164 @itemx -mno-relax-immediate
13165 @opindex mrelax-immediate
13166 @opindex mno-relax-immediate
13167 Allow arbitrary sized immediates in bit operations.
13169 @item -mwide-bitfields
13170 @itemx -mno-wide-bitfields
13171 @opindex mwide-bitfields
13172 @opindex mno-wide-bitfields
13173 Always treat bit-fields as int-sized.
13175 @item -m4byte-functions
13176 @itemx -mno-4byte-functions
13177 @opindex m4byte-functions
13178 @opindex mno-4byte-functions
13179 Force all functions to be aligned to a four byte boundary.
13181 @item -mcallgraph-data
13182 @itemx -mno-callgraph-data
13183 @opindex mcallgraph-data
13184 @opindex mno-callgraph-data
13185 Emit callgraph information.
13188 @itemx -mno-slow-bytes
13189 @opindex mslow-bytes
13190 @opindex mno-slow-bytes
13191 Prefer word access when reading byte quantities.
13193 @item -mlittle-endian
13194 @itemx -mbig-endian
13195 @opindex mlittle-endian
13196 @opindex mbig-endian
13197 Generate code for a little endian target.
13203 Generate code for the 210 processor.
13207 Assume that run-time support has been provided and so omit the
13208 simulator library (@file{libsim.a)} from the linker command line.
13210 @item -mstack-increment=@var{size}
13211 @opindex mstack-increment
13212 Set the maximum amount for a single stack increment operation. Large
13213 values can increase the speed of programs which contain functions
13214 that need a large amount of stack space, but they can also trigger a
13215 segmentation fault if the stack is extended too much. The default
13221 @subsection MeP Options
13222 @cindex MeP options
13228 Enables the @code{abs} instruction, which is the absolute difference
13229 between two registers.
13233 Enables all the optional instructions - average, multiply, divide, bit
13234 operations, leading zero, absolute difference, min/max, clip, and
13240 Enables the @code{ave} instruction, which computes the average of two
13243 @item -mbased=@var{n}
13245 Variables of size @var{n} bytes or smaller will be placed in the
13246 @code{.based} section by default. Based variables use the @code{$tp}
13247 register as a base register, and there is a 128 byte limit to the
13248 @code{.based} section.
13252 Enables the bit operation instructions - bit test (@code{btstm}), set
13253 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
13254 test-and-set (@code{tas}).
13256 @item -mc=@var{name}
13258 Selects which section constant data will be placed in. @var{name} may
13259 be @code{tiny}, @code{near}, or @code{far}.
13263 Enables the @code{clip} instruction. Note that @code{-mclip} is not
13264 useful unless you also provide @code{-mminmax}.
13266 @item -mconfig=@var{name}
13268 Selects one of the build-in core configurations. Each MeP chip has
13269 one or more modules in it; each module has a core CPU and a variety of
13270 coprocessors, optional instructions, and peripherals. The
13271 @code{MeP-Integrator} tool, not part of GCC, provides these
13272 configurations through this option; using this option is the same as
13273 using all the corresponding command line options. The default
13274 configuration is @code{default}.
13278 Enables the coprocessor instructions. By default, this is a 32-bit
13279 coprocessor. Note that the coprocessor is normally enabled via the
13280 @code{-mconfig=} option.
13284 Enables the 32-bit coprocessor's instructions.
13288 Enables the 64-bit coprocessor's instructions.
13292 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
13296 Causes constant variables to be placed in the @code{.near} section.
13300 Enables the @code{div} and @code{divu} instructions.
13304 Generate big-endian code.
13308 Generate little-endian code.
13310 @item -mio-volatile
13311 @opindex mio-volatile
13312 Tells the compiler that any variable marked with the @code{io}
13313 attribute is to be considered volatile.
13317 Causes variables to be assigned to the @code{.far} section by default.
13321 Enables the @code{leadz} (leading zero) instruction.
13325 Causes variables to be assigned to the @code{.near} section by default.
13329 Enables the @code{min} and @code{max} instructions.
13333 Enables the multiplication and multiply-accumulate instructions.
13337 Disables all the optional instructions enabled by @code{-mall-opts}.
13341 Enables the @code{repeat} and @code{erepeat} instructions, used for
13342 low-overhead looping.
13346 Causes all variables to default to the @code{.tiny} section. Note
13347 that there is a 65536 byte limit to this section. Accesses to these
13348 variables use the @code{%gp} base register.
13352 Enables the saturation instructions. Note that the compiler does not
13353 currently generate these itself, but this option is included for
13354 compatibility with other tools, like @code{as}.
13358 Link the SDRAM-based runtime instead of the default ROM-based runtime.
13362 Link the simulator runtime libraries.
13366 Link the simulator runtime libraries, excluding built-in support
13367 for reset and exception vectors and tables.
13371 Causes all functions to default to the @code{.far} section. Without
13372 this option, functions default to the @code{.near} section.
13374 @item -mtiny=@var{n}
13376 Variables that are @var{n} bytes or smaller will be allocated to the
13377 @code{.tiny} section. These variables use the @code{$gp} base
13378 register. The default for this option is 4, but note that there's a
13379 65536 byte limit to the @code{.tiny} section.
13384 @subsection MIPS Options
13385 @cindex MIPS options
13391 Generate big-endian code.
13395 Generate little-endian code. This is the default for @samp{mips*el-*-*}
13398 @item -march=@var{arch}
13400 Generate code that will run on @var{arch}, which can be the name of a
13401 generic MIPS ISA, or the name of a particular processor.
13403 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
13404 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
13405 The processor names are:
13406 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
13407 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
13408 @samp{5kc}, @samp{5kf},
13410 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
13411 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
13412 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
13413 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
13414 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
13415 @samp{loongson2e}, @samp{loongson2f},
13419 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
13420 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
13421 @samp{rm7000}, @samp{rm9000},
13422 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
13425 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
13426 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
13428 The special value @samp{from-abi} selects the
13429 most compatible architecture for the selected ABI (that is,
13430 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
13432 Native Linux/GNU toolchains also support the value @samp{native},
13433 which selects the best architecture option for the host processor.
13434 @option{-march=native} has no effect if GCC does not recognize
13437 In processor names, a final @samp{000} can be abbreviated as @samp{k}
13438 (for example, @samp{-march=r2k}). Prefixes are optional, and
13439 @samp{vr} may be written @samp{r}.
13441 Names of the form @samp{@var{n}f2_1} refer to processors with
13442 FPUs clocked at half the rate of the core, names of the form
13443 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
13444 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
13445 processors with FPUs clocked a ratio of 3:2 with respect to the core.
13446 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
13447 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
13448 accepted as synonyms for @samp{@var{n}f1_1}.
13450 GCC defines two macros based on the value of this option. The first
13451 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
13452 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
13453 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
13454 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
13455 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
13457 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
13458 above. In other words, it will have the full prefix and will not
13459 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
13460 the macro names the resolved architecture (either @samp{"mips1"} or
13461 @samp{"mips3"}). It names the default architecture when no
13462 @option{-march} option is given.
13464 @item -mtune=@var{arch}
13466 Optimize for @var{arch}. Among other things, this option controls
13467 the way instructions are scheduled, and the perceived cost of arithmetic
13468 operations. The list of @var{arch} values is the same as for
13471 When this option is not used, GCC will optimize for the processor
13472 specified by @option{-march}. By using @option{-march} and
13473 @option{-mtune} together, it is possible to generate code that will
13474 run on a family of processors, but optimize the code for one
13475 particular member of that family.
13477 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
13478 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
13479 @samp{-march} ones described above.
13483 Equivalent to @samp{-march=mips1}.
13487 Equivalent to @samp{-march=mips2}.
13491 Equivalent to @samp{-march=mips3}.
13495 Equivalent to @samp{-march=mips4}.
13499 Equivalent to @samp{-march=mips32}.
13503 Equivalent to @samp{-march=mips32r2}.
13507 Equivalent to @samp{-march=mips64}.
13511 Equivalent to @samp{-march=mips64r2}.
13516 @opindex mno-mips16
13517 Generate (do not generate) MIPS16 code. If GCC is targetting a
13518 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
13520 MIPS16 code generation can also be controlled on a per-function basis
13521 by means of @code{mips16} and @code{nomips16} attributes.
13522 @xref{Function Attributes}, for more information.
13524 @item -mflip-mips16
13525 @opindex mflip-mips16
13526 Generate MIPS16 code on alternating functions. This option is provided
13527 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
13528 not intended for ordinary use in compiling user code.
13530 @item -minterlink-mips16
13531 @itemx -mno-interlink-mips16
13532 @opindex minterlink-mips16
13533 @opindex mno-interlink-mips16
13534 Require (do not require) that non-MIPS16 code be link-compatible with
13537 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
13538 it must either use a call or an indirect jump. @option{-minterlink-mips16}
13539 therefore disables direct jumps unless GCC knows that the target of the
13540 jump is not MIPS16.
13552 Generate code for the given ABI@.
13554 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
13555 generates 64-bit code when you select a 64-bit architecture, but you
13556 can use @option{-mgp32} to get 32-bit code instead.
13558 For information about the O64 ABI, see
13559 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
13561 GCC supports a variant of the o32 ABI in which floating-point registers
13562 are 64 rather than 32 bits wide. You can select this combination with
13563 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
13564 and @samp{mfhc1} instructions and is therefore only supported for
13565 MIPS32R2 processors.
13567 The register assignments for arguments and return values remain the
13568 same, but each scalar value is passed in a single 64-bit register
13569 rather than a pair of 32-bit registers. For example, scalar
13570 floating-point values are returned in @samp{$f0} only, not a
13571 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
13572 remains the same, but all 64 bits are saved.
13575 @itemx -mno-abicalls
13577 @opindex mno-abicalls
13578 Generate (do not generate) code that is suitable for SVR4-style
13579 dynamic objects. @option{-mabicalls} is the default for SVR4-based
13584 Generate (do not generate) code that is fully position-independent,
13585 and that can therefore be linked into shared libraries. This option
13586 only affects @option{-mabicalls}.
13588 All @option{-mabicalls} code has traditionally been position-independent,
13589 regardless of options like @option{-fPIC} and @option{-fpic}. However,
13590 as an extension, the GNU toolchain allows executables to use absolute
13591 accesses for locally-binding symbols. It can also use shorter GP
13592 initialization sequences and generate direct calls to locally-defined
13593 functions. This mode is selected by @option{-mno-shared}.
13595 @option{-mno-shared} depends on binutils 2.16 or higher and generates
13596 objects that can only be linked by the GNU linker. However, the option
13597 does not affect the ABI of the final executable; it only affects the ABI
13598 of relocatable objects. Using @option{-mno-shared} will generally make
13599 executables both smaller and quicker.
13601 @option{-mshared} is the default.
13607 Assume (do not assume) that the static and dynamic linkers
13608 support PLTs and copy relocations. This option only affects
13609 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
13610 has no effect without @samp{-msym32}.
13612 You can make @option{-mplt} the default by configuring
13613 GCC with @option{--with-mips-plt}. The default is
13614 @option{-mno-plt} otherwise.
13620 Lift (do not lift) the usual restrictions on the size of the global
13623 GCC normally uses a single instruction to load values from the GOT@.
13624 While this is relatively efficient, it will only work if the GOT
13625 is smaller than about 64k. Anything larger will cause the linker
13626 to report an error such as:
13628 @cindex relocation truncated to fit (MIPS)
13630 relocation truncated to fit: R_MIPS_GOT16 foobar
13633 If this happens, you should recompile your code with @option{-mxgot}.
13634 It should then work with very large GOTs, although it will also be
13635 less efficient, since it will take three instructions to fetch the
13636 value of a global symbol.
13638 Note that some linkers can create multiple GOTs. If you have such a
13639 linker, you should only need to use @option{-mxgot} when a single object
13640 file accesses more than 64k's worth of GOT entries. Very few do.
13642 These options have no effect unless GCC is generating position
13647 Assume that general-purpose registers are 32 bits wide.
13651 Assume that general-purpose registers are 64 bits wide.
13655 Assume that floating-point registers are 32 bits wide.
13659 Assume that floating-point registers are 64 bits wide.
13662 @opindex mhard-float
13663 Use floating-point coprocessor instructions.
13666 @opindex msoft-float
13667 Do not use floating-point coprocessor instructions. Implement
13668 floating-point calculations using library calls instead.
13670 @item -msingle-float
13671 @opindex msingle-float
13672 Assume that the floating-point coprocessor only supports single-precision
13675 @item -mdouble-float
13676 @opindex mdouble-float
13677 Assume that the floating-point coprocessor supports double-precision
13678 operations. This is the default.
13684 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
13685 implement atomic memory built-in functions. When neither option is
13686 specified, GCC will use the instructions if the target architecture
13689 @option{-mllsc} is useful if the runtime environment can emulate the
13690 instructions and @option{-mno-llsc} can be useful when compiling for
13691 nonstandard ISAs. You can make either option the default by
13692 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
13693 respectively. @option{--with-llsc} is the default for some
13694 configurations; see the installation documentation for details.
13700 Use (do not use) revision 1 of the MIPS DSP ASE@.
13701 @xref{MIPS DSP Built-in Functions}. This option defines the
13702 preprocessor macro @samp{__mips_dsp}. It also defines
13703 @samp{__mips_dsp_rev} to 1.
13709 Use (do not use) revision 2 of the MIPS DSP ASE@.
13710 @xref{MIPS DSP Built-in Functions}. This option defines the
13711 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
13712 It also defines @samp{__mips_dsp_rev} to 2.
13715 @itemx -mno-smartmips
13716 @opindex msmartmips
13717 @opindex mno-smartmips
13718 Use (do not use) the MIPS SmartMIPS ASE.
13720 @item -mpaired-single
13721 @itemx -mno-paired-single
13722 @opindex mpaired-single
13723 @opindex mno-paired-single
13724 Use (do not use) paired-single floating-point instructions.
13725 @xref{MIPS Paired-Single Support}. This option requires
13726 hardware floating-point support to be enabled.
13732 Use (do not use) MIPS Digital Media Extension instructions.
13733 This option can only be used when generating 64-bit code and requires
13734 hardware floating-point support to be enabled.
13739 @opindex mno-mips3d
13740 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
13741 The option @option{-mips3d} implies @option{-mpaired-single}.
13747 Use (do not use) MT Multithreading instructions.
13751 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
13752 an explanation of the default and the way that the pointer size is
13757 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
13759 The default size of @code{int}s, @code{long}s and pointers depends on
13760 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
13761 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
13762 32-bit @code{long}s. Pointers are the same size as @code{long}s,
13763 or the same size as integer registers, whichever is smaller.
13769 Assume (do not assume) that all symbols have 32-bit values, regardless
13770 of the selected ABI@. This option is useful in combination with
13771 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
13772 to generate shorter and faster references to symbolic addresses.
13776 Put definitions of externally-visible data in a small data section
13777 if that data is no bigger than @var{num} bytes. GCC can then access
13778 the data more efficiently; see @option{-mgpopt} for details.
13780 The default @option{-G} option depends on the configuration.
13782 @item -mlocal-sdata
13783 @itemx -mno-local-sdata
13784 @opindex mlocal-sdata
13785 @opindex mno-local-sdata
13786 Extend (do not extend) the @option{-G} behavior to local data too,
13787 such as to static variables in C@. @option{-mlocal-sdata} is the
13788 default for all configurations.
13790 If the linker complains that an application is using too much small data,
13791 you might want to try rebuilding the less performance-critical parts with
13792 @option{-mno-local-sdata}. You might also want to build large
13793 libraries with @option{-mno-local-sdata}, so that the libraries leave
13794 more room for the main program.
13796 @item -mextern-sdata
13797 @itemx -mno-extern-sdata
13798 @opindex mextern-sdata
13799 @opindex mno-extern-sdata
13800 Assume (do not assume) that externally-defined data will be in
13801 a small data section if that data is within the @option{-G} limit.
13802 @option{-mextern-sdata} is the default for all configurations.
13804 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
13805 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
13806 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
13807 is placed in a small data section. If @var{Var} is defined by another
13808 module, you must either compile that module with a high-enough
13809 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
13810 definition. If @var{Var} is common, you must link the application
13811 with a high-enough @option{-G} setting.
13813 The easiest way of satisfying these restrictions is to compile
13814 and link every module with the same @option{-G} option. However,
13815 you may wish to build a library that supports several different
13816 small data limits. You can do this by compiling the library with
13817 the highest supported @option{-G} setting and additionally using
13818 @option{-mno-extern-sdata} to stop the library from making assumptions
13819 about externally-defined data.
13825 Use (do not use) GP-relative accesses for symbols that are known to be
13826 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
13827 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
13830 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
13831 might not hold the value of @code{_gp}. For example, if the code is
13832 part of a library that might be used in a boot monitor, programs that
13833 call boot monitor routines will pass an unknown value in @code{$gp}.
13834 (In such situations, the boot monitor itself would usually be compiled
13835 with @option{-G0}.)
13837 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
13838 @option{-mno-extern-sdata}.
13840 @item -membedded-data
13841 @itemx -mno-embedded-data
13842 @opindex membedded-data
13843 @opindex mno-embedded-data
13844 Allocate variables to the read-only data section first if possible, then
13845 next in the small data section if possible, otherwise in data. This gives
13846 slightly slower code than the default, but reduces the amount of RAM required
13847 when executing, and thus may be preferred for some embedded systems.
13849 @item -muninit-const-in-rodata
13850 @itemx -mno-uninit-const-in-rodata
13851 @opindex muninit-const-in-rodata
13852 @opindex mno-uninit-const-in-rodata
13853 Put uninitialized @code{const} variables in the read-only data section.
13854 This option is only meaningful in conjunction with @option{-membedded-data}.
13856 @item -mcode-readable=@var{setting}
13857 @opindex mcode-readable
13858 Specify whether GCC may generate code that reads from executable sections.
13859 There are three possible settings:
13862 @item -mcode-readable=yes
13863 Instructions may freely access executable sections. This is the
13866 @item -mcode-readable=pcrel
13867 MIPS16 PC-relative load instructions can access executable sections,
13868 but other instructions must not do so. This option is useful on 4KSc
13869 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
13870 It is also useful on processors that can be configured to have a dual
13871 instruction/data SRAM interface and that, like the M4K, automatically
13872 redirect PC-relative loads to the instruction RAM.
13874 @item -mcode-readable=no
13875 Instructions must not access executable sections. This option can be
13876 useful on targets that are configured to have a dual instruction/data
13877 SRAM interface but that (unlike the M4K) do not automatically redirect
13878 PC-relative loads to the instruction RAM.
13881 @item -msplit-addresses
13882 @itemx -mno-split-addresses
13883 @opindex msplit-addresses
13884 @opindex mno-split-addresses
13885 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
13886 relocation operators. This option has been superseded by
13887 @option{-mexplicit-relocs} but is retained for backwards compatibility.
13889 @item -mexplicit-relocs
13890 @itemx -mno-explicit-relocs
13891 @opindex mexplicit-relocs
13892 @opindex mno-explicit-relocs
13893 Use (do not use) assembler relocation operators when dealing with symbolic
13894 addresses. The alternative, selected by @option{-mno-explicit-relocs},
13895 is to use assembler macros instead.
13897 @option{-mexplicit-relocs} is the default if GCC was configured
13898 to use an assembler that supports relocation operators.
13900 @item -mcheck-zero-division
13901 @itemx -mno-check-zero-division
13902 @opindex mcheck-zero-division
13903 @opindex mno-check-zero-division
13904 Trap (do not trap) on integer division by zero.
13906 The default is @option{-mcheck-zero-division}.
13908 @item -mdivide-traps
13909 @itemx -mdivide-breaks
13910 @opindex mdivide-traps
13911 @opindex mdivide-breaks
13912 MIPS systems check for division by zero by generating either a
13913 conditional trap or a break instruction. Using traps results in
13914 smaller code, but is only supported on MIPS II and later. Also, some
13915 versions of the Linux kernel have a bug that prevents trap from
13916 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
13917 allow conditional traps on architectures that support them and
13918 @option{-mdivide-breaks} to force the use of breaks.
13920 The default is usually @option{-mdivide-traps}, but this can be
13921 overridden at configure time using @option{--with-divide=breaks}.
13922 Divide-by-zero checks can be completely disabled using
13923 @option{-mno-check-zero-division}.
13928 @opindex mno-memcpy
13929 Force (do not force) the use of @code{memcpy()} for non-trivial block
13930 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
13931 most constant-sized copies.
13934 @itemx -mno-long-calls
13935 @opindex mlong-calls
13936 @opindex mno-long-calls
13937 Disable (do not disable) use of the @code{jal} instruction. Calling
13938 functions using @code{jal} is more efficient but requires the caller
13939 and callee to be in the same 256 megabyte segment.
13941 This option has no effect on abicalls code. The default is
13942 @option{-mno-long-calls}.
13948 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
13949 instructions, as provided by the R4650 ISA@.
13952 @itemx -mno-fused-madd
13953 @opindex mfused-madd
13954 @opindex mno-fused-madd
13955 Enable (disable) use of the floating point multiply-accumulate
13956 instructions, when they are available. The default is
13957 @option{-mfused-madd}.
13959 When multiply-accumulate instructions are used, the intermediate
13960 product is calculated to infinite precision and is not subject to
13961 the FCSR Flush to Zero bit. This may be undesirable in some
13966 Tell the MIPS assembler to not run its preprocessor over user
13967 assembler files (with a @samp{.s} suffix) when assembling them.
13970 @itemx -mno-fix-r4000
13971 @opindex mfix-r4000
13972 @opindex mno-fix-r4000
13973 Work around certain R4000 CPU errata:
13976 A double-word or a variable shift may give an incorrect result if executed
13977 immediately after starting an integer division.
13979 A double-word or a variable shift may give an incorrect result if executed
13980 while an integer multiplication is in progress.
13982 An integer division may give an incorrect result if started in a delay slot
13983 of a taken branch or a jump.
13987 @itemx -mno-fix-r4400
13988 @opindex mfix-r4400
13989 @opindex mno-fix-r4400
13990 Work around certain R4400 CPU errata:
13993 A double-word or a variable shift may give an incorrect result if executed
13994 immediately after starting an integer division.
13998 @itemx -mno-fix-r10000
13999 @opindex mfix-r10000
14000 @opindex mno-fix-r10000
14001 Work around certain R10000 errata:
14004 @code{ll}/@code{sc} sequences may not behave atomically on revisions
14005 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
14008 This option can only be used if the target architecture supports
14009 branch-likely instructions. @option{-mfix-r10000} is the default when
14010 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
14014 @itemx -mno-fix-vr4120
14015 @opindex mfix-vr4120
14016 Work around certain VR4120 errata:
14019 @code{dmultu} does not always produce the correct result.
14021 @code{div} and @code{ddiv} do not always produce the correct result if one
14022 of the operands is negative.
14024 The workarounds for the division errata rely on special functions in
14025 @file{libgcc.a}. At present, these functions are only provided by
14026 the @code{mips64vr*-elf} configurations.
14028 Other VR4120 errata require a nop to be inserted between certain pairs of
14029 instructions. These errata are handled by the assembler, not by GCC itself.
14032 @opindex mfix-vr4130
14033 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
14034 workarounds are implemented by the assembler rather than by GCC,
14035 although GCC will avoid using @code{mflo} and @code{mfhi} if the
14036 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
14037 instructions are available instead.
14040 @itemx -mno-fix-sb1
14042 Work around certain SB-1 CPU core errata.
14043 (This flag currently works around the SB-1 revision 2
14044 ``F1'' and ``F2'' floating point errata.)
14046 @item -mr10k-cache-barrier=@var{setting}
14047 @opindex mr10k-cache-barrier
14048 Specify whether GCC should insert cache barriers to avoid the
14049 side-effects of speculation on R10K processors.
14051 In common with many processors, the R10K tries to predict the outcome
14052 of a conditional branch and speculatively executes instructions from
14053 the ``taken'' branch. It later aborts these instructions if the
14054 predicted outcome was wrong. However, on the R10K, even aborted
14055 instructions can have side effects.
14057 This problem only affects kernel stores and, depending on the system,
14058 kernel loads. As an example, a speculatively-executed store may load
14059 the target memory into cache and mark the cache line as dirty, even if
14060 the store itself is later aborted. If a DMA operation writes to the
14061 same area of memory before the ``dirty'' line is flushed, the cached
14062 data will overwrite the DMA-ed data. See the R10K processor manual
14063 for a full description, including other potential problems.
14065 One workaround is to insert cache barrier instructions before every memory
14066 access that might be speculatively executed and that might have side
14067 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
14068 controls GCC's implementation of this workaround. It assumes that
14069 aborted accesses to any byte in the following regions will not have
14074 the memory occupied by the current function's stack frame;
14077 the memory occupied by an incoming stack argument;
14080 the memory occupied by an object with a link-time-constant address.
14083 It is the kernel's responsibility to ensure that speculative
14084 accesses to these regions are indeed safe.
14086 If the input program contains a function declaration such as:
14092 then the implementation of @code{foo} must allow @code{j foo} and
14093 @code{jal foo} to be executed speculatively. GCC honors this
14094 restriction for functions it compiles itself. It expects non-GCC
14095 functions (such as hand-written assembly code) to do the same.
14097 The option has three forms:
14100 @item -mr10k-cache-barrier=load-store
14101 Insert a cache barrier before a load or store that might be
14102 speculatively executed and that might have side effects even
14105 @item -mr10k-cache-barrier=store
14106 Insert a cache barrier before a store that might be speculatively
14107 executed and that might have side effects even if aborted.
14109 @item -mr10k-cache-barrier=none
14110 Disable the insertion of cache barriers. This is the default setting.
14113 @item -mflush-func=@var{func}
14114 @itemx -mno-flush-func
14115 @opindex mflush-func
14116 Specifies the function to call to flush the I and D caches, or to not
14117 call any such function. If called, the function must take the same
14118 arguments as the common @code{_flush_func()}, that is, the address of the
14119 memory range for which the cache is being flushed, the size of the
14120 memory range, and the number 3 (to flush both caches). The default
14121 depends on the target GCC was configured for, but commonly is either
14122 @samp{_flush_func} or @samp{__cpu_flush}.
14124 @item mbranch-cost=@var{num}
14125 @opindex mbranch-cost
14126 Set the cost of branches to roughly @var{num} ``simple'' instructions.
14127 This cost is only a heuristic and is not guaranteed to produce
14128 consistent results across releases. A zero cost redundantly selects
14129 the default, which is based on the @option{-mtune} setting.
14131 @item -mbranch-likely
14132 @itemx -mno-branch-likely
14133 @opindex mbranch-likely
14134 @opindex mno-branch-likely
14135 Enable or disable use of Branch Likely instructions, regardless of the
14136 default for the selected architecture. By default, Branch Likely
14137 instructions may be generated if they are supported by the selected
14138 architecture. An exception is for the MIPS32 and MIPS64 architectures
14139 and processors which implement those architectures; for those, Branch
14140 Likely instructions will not be generated by default because the MIPS32
14141 and MIPS64 architectures specifically deprecate their use.
14143 @item -mfp-exceptions
14144 @itemx -mno-fp-exceptions
14145 @opindex mfp-exceptions
14146 Specifies whether FP exceptions are enabled. This affects how we schedule
14147 FP instructions for some processors. The default is that FP exceptions are
14150 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
14151 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
14154 @item -mvr4130-align
14155 @itemx -mno-vr4130-align
14156 @opindex mvr4130-align
14157 The VR4130 pipeline is two-way superscalar, but can only issue two
14158 instructions together if the first one is 8-byte aligned. When this
14159 option is enabled, GCC will align pairs of instructions that it
14160 thinks should execute in parallel.
14162 This option only has an effect when optimizing for the VR4130.
14163 It normally makes code faster, but at the expense of making it bigger.
14164 It is enabled by default at optimization level @option{-O3}.
14169 Enable (disable) generation of @code{synci} instructions on
14170 architectures that support it. The @code{synci} instructions (if
14171 enabled) will be generated when @code{__builtin___clear_cache()} is
14174 This option defaults to @code{-mno-synci}, but the default can be
14175 overridden by configuring with @code{--with-synci}.
14177 When compiling code for single processor systems, it is generally safe
14178 to use @code{synci}. However, on many multi-core (SMP) systems, it
14179 will not invalidate the instruction caches on all cores and may lead
14180 to undefined behavior.
14182 @item -mrelax-pic-calls
14183 @itemx -mno-relax-pic-calls
14184 @opindex mrelax-pic-calls
14185 Try to turn PIC calls that are normally dispatched via register
14186 @code{$25} into direct calls. This is only possible if the linker can
14187 resolve the destination at link-time and if the destination is within
14188 range for a direct call.
14190 @option{-mrelax-pic-calls} is the default if GCC was configured to use
14191 an assembler and a linker that supports the @code{.reloc} assembly
14192 directive and @code{-mexplicit-relocs} is in effect. With
14193 @code{-mno-explicit-relocs}, this optimization can be performed by the
14194 assembler and the linker alone without help from the compiler.
14198 @subsection MMIX Options
14199 @cindex MMIX Options
14201 These options are defined for the MMIX:
14205 @itemx -mno-libfuncs
14207 @opindex mno-libfuncs
14208 Specify that intrinsic library functions are being compiled, passing all
14209 values in registers, no matter the size.
14212 @itemx -mno-epsilon
14214 @opindex mno-epsilon
14215 Generate floating-point comparison instructions that compare with respect
14216 to the @code{rE} epsilon register.
14218 @item -mabi=mmixware
14220 @opindex mabi=mmixware
14222 Generate code that passes function parameters and return values that (in
14223 the called function) are seen as registers @code{$0} and up, as opposed to
14224 the GNU ABI which uses global registers @code{$231} and up.
14226 @item -mzero-extend
14227 @itemx -mno-zero-extend
14228 @opindex mzero-extend
14229 @opindex mno-zero-extend
14230 When reading data from memory in sizes shorter than 64 bits, use (do not
14231 use) zero-extending load instructions by default, rather than
14232 sign-extending ones.
14235 @itemx -mno-knuthdiv
14237 @opindex mno-knuthdiv
14238 Make the result of a division yielding a remainder have the same sign as
14239 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
14240 remainder follows the sign of the dividend. Both methods are
14241 arithmetically valid, the latter being almost exclusively used.
14243 @item -mtoplevel-symbols
14244 @itemx -mno-toplevel-symbols
14245 @opindex mtoplevel-symbols
14246 @opindex mno-toplevel-symbols
14247 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
14248 code can be used with the @code{PREFIX} assembly directive.
14252 Generate an executable in the ELF format, rather than the default
14253 @samp{mmo} format used by the @command{mmix} simulator.
14255 @item -mbranch-predict
14256 @itemx -mno-branch-predict
14257 @opindex mbranch-predict
14258 @opindex mno-branch-predict
14259 Use (do not use) the probable-branch instructions, when static branch
14260 prediction indicates a probable branch.
14262 @item -mbase-addresses
14263 @itemx -mno-base-addresses
14264 @opindex mbase-addresses
14265 @opindex mno-base-addresses
14266 Generate (do not generate) code that uses @emph{base addresses}. Using a
14267 base address automatically generates a request (handled by the assembler
14268 and the linker) for a constant to be set up in a global register. The
14269 register is used for one or more base address requests within the range 0
14270 to 255 from the value held in the register. The generally leads to short
14271 and fast code, but the number of different data items that can be
14272 addressed is limited. This means that a program that uses lots of static
14273 data may require @option{-mno-base-addresses}.
14275 @item -msingle-exit
14276 @itemx -mno-single-exit
14277 @opindex msingle-exit
14278 @opindex mno-single-exit
14279 Force (do not force) generated code to have a single exit point in each
14283 @node MN10300 Options
14284 @subsection MN10300 Options
14285 @cindex MN10300 options
14287 These @option{-m} options are defined for Matsushita MN10300 architectures:
14292 Generate code to avoid bugs in the multiply instructions for the MN10300
14293 processors. This is the default.
14295 @item -mno-mult-bug
14296 @opindex mno-mult-bug
14297 Do not generate code to avoid bugs in the multiply instructions for the
14298 MN10300 processors.
14302 Generate code which uses features specific to the AM33 processor.
14306 Do not generate code which uses features specific to the AM33 processor. This
14309 @item -mreturn-pointer-on-d0
14310 @opindex mreturn-pointer-on-d0
14311 When generating a function which returns a pointer, return the pointer
14312 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
14313 only in a0, and attempts to call such functions without a prototype
14314 would result in errors. Note that this option is on by default; use
14315 @option{-mno-return-pointer-on-d0} to disable it.
14319 Do not link in the C run-time initialization object file.
14323 Indicate to the linker that it should perform a relaxation optimization pass
14324 to shorten branches, calls and absolute memory addresses. This option only
14325 has an effect when used on the command line for the final link step.
14327 This option makes symbolic debugging impossible.
14330 @node PDP-11 Options
14331 @subsection PDP-11 Options
14332 @cindex PDP-11 Options
14334 These options are defined for the PDP-11:
14339 Use hardware FPP floating point. This is the default. (FIS floating
14340 point on the PDP-11/40 is not supported.)
14343 @opindex msoft-float
14344 Do not use hardware floating point.
14348 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
14352 Return floating-point results in memory. This is the default.
14356 Generate code for a PDP-11/40.
14360 Generate code for a PDP-11/45. This is the default.
14364 Generate code for a PDP-11/10.
14366 @item -mbcopy-builtin
14367 @opindex mbcopy-builtin
14368 Use inline @code{movmemhi} patterns for copying memory. This is the
14373 Do not use inline @code{movmemhi} patterns for copying memory.
14379 Use 16-bit @code{int}. This is the default.
14385 Use 32-bit @code{int}.
14388 @itemx -mno-float32
14390 @opindex mno-float32
14391 Use 64-bit @code{float}. This is the default.
14394 @itemx -mno-float64
14396 @opindex mno-float64
14397 Use 32-bit @code{float}.
14401 Use @code{abshi2} pattern. This is the default.
14405 Do not use @code{abshi2} pattern.
14407 @item -mbranch-expensive
14408 @opindex mbranch-expensive
14409 Pretend that branches are expensive. This is for experimenting with
14410 code generation only.
14412 @item -mbranch-cheap
14413 @opindex mbranch-cheap
14414 Do not pretend that branches are expensive. This is the default.
14418 Generate code for a system with split I&D@.
14422 Generate code for a system without split I&D@. This is the default.
14426 Use Unix assembler syntax. This is the default when configured for
14427 @samp{pdp11-*-bsd}.
14431 Use DEC assembler syntax. This is the default when configured for any
14432 PDP-11 target other than @samp{pdp11-*-bsd}.
14435 @node picoChip Options
14436 @subsection picoChip Options
14437 @cindex picoChip options
14439 These @samp{-m} options are defined for picoChip implementations:
14443 @item -mae=@var{ae_type}
14445 Set the instruction set, register set, and instruction scheduling
14446 parameters for array element type @var{ae_type}. Supported values
14447 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
14449 @option{-mae=ANY} selects a completely generic AE type. Code
14450 generated with this option will run on any of the other AE types. The
14451 code will not be as efficient as it would be if compiled for a specific
14452 AE type, and some types of operation (e.g., multiplication) will not
14453 work properly on all types of AE.
14455 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
14456 for compiled code, and is the default.
14458 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
14459 option may suffer from poor performance of byte (char) manipulation,
14460 since the DSP AE does not provide hardware support for byte load/stores.
14462 @item -msymbol-as-address
14463 Enable the compiler to directly use a symbol name as an address in a
14464 load/store instruction, without first loading it into a
14465 register. Typically, the use of this option will generate larger
14466 programs, which run faster than when the option isn't used. However, the
14467 results vary from program to program, so it is left as a user option,
14468 rather than being permanently enabled.
14470 @item -mno-inefficient-warnings
14471 Disables warnings about the generation of inefficient code. These
14472 warnings can be generated, for example, when compiling code which
14473 performs byte-level memory operations on the MAC AE type. The MAC AE has
14474 no hardware support for byte-level memory operations, so all byte
14475 load/stores must be synthesized from word load/store operations. This is
14476 inefficient and a warning will be generated indicating to the programmer
14477 that they should rewrite the code to avoid byte operations, or to target
14478 an AE type which has the necessary hardware support. This option enables
14479 the warning to be turned off.
14483 @node PowerPC Options
14484 @subsection PowerPC Options
14485 @cindex PowerPC options
14487 These are listed under @xref{RS/6000 and PowerPC Options}.
14489 @node RS/6000 and PowerPC Options
14490 @subsection IBM RS/6000 and PowerPC Options
14491 @cindex RS/6000 and PowerPC Options
14492 @cindex IBM RS/6000 and PowerPC Options
14494 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
14501 @itemx -mno-powerpc
14502 @itemx -mpowerpc-gpopt
14503 @itemx -mno-powerpc-gpopt
14504 @itemx -mpowerpc-gfxopt
14505 @itemx -mno-powerpc-gfxopt
14507 @itemx -mno-powerpc64
14511 @itemx -mno-popcntb
14513 @itemx -mno-popcntd
14521 @itemx -mno-hard-dfp
14525 @opindex mno-power2
14527 @opindex mno-powerpc
14528 @opindex mpowerpc-gpopt
14529 @opindex mno-powerpc-gpopt
14530 @opindex mpowerpc-gfxopt
14531 @opindex mno-powerpc-gfxopt
14532 @opindex mpowerpc64
14533 @opindex mno-powerpc64
14537 @opindex mno-popcntb
14539 @opindex mno-popcntd
14545 @opindex mno-mfpgpr
14547 @opindex mno-hard-dfp
14548 GCC supports two related instruction set architectures for the
14549 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
14550 instructions supported by the @samp{rios} chip set used in the original
14551 RS/6000 systems and the @dfn{PowerPC} instruction set is the
14552 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
14553 the IBM 4xx, 6xx, and follow-on microprocessors.
14555 Neither architecture is a subset of the other. However there is a
14556 large common subset of instructions supported by both. An MQ
14557 register is included in processors supporting the POWER architecture.
14559 You use these options to specify which instructions are available on the
14560 processor you are using. The default value of these options is
14561 determined when configuring GCC@. Specifying the
14562 @option{-mcpu=@var{cpu_type}} overrides the specification of these
14563 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
14564 rather than the options listed above.
14566 The @option{-mpower} option allows GCC to generate instructions that
14567 are found only in the POWER architecture and to use the MQ register.
14568 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
14569 to generate instructions that are present in the POWER2 architecture but
14570 not the original POWER architecture.
14572 The @option{-mpowerpc} option allows GCC to generate instructions that
14573 are found only in the 32-bit subset of the PowerPC architecture.
14574 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
14575 GCC to use the optional PowerPC architecture instructions in the
14576 General Purpose group, including floating-point square root. Specifying
14577 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
14578 use the optional PowerPC architecture instructions in the Graphics
14579 group, including floating-point select.
14581 The @option{-mmfcrf} option allows GCC to generate the move from
14582 condition register field instruction implemented on the POWER4
14583 processor and other processors that support the PowerPC V2.01
14585 The @option{-mpopcntb} option allows GCC to generate the popcount and
14586 double precision FP reciprocal estimate instruction implemented on the
14587 POWER5 processor and other processors that support the PowerPC V2.02
14589 The @option{-mpopcntd} option allows GCC to generate the popcount
14590 instruction implemented on the POWER7 processor and other processors
14591 that support the PowerPC V2.06 architecture.
14592 The @option{-mfprnd} option allows GCC to generate the FP round to
14593 integer instructions implemented on the POWER5+ processor and other
14594 processors that support the PowerPC V2.03 architecture.
14595 The @option{-mcmpb} option allows GCC to generate the compare bytes
14596 instruction implemented on the POWER6 processor and other processors
14597 that support the PowerPC V2.05 architecture.
14598 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
14599 general purpose register instructions implemented on the POWER6X
14600 processor and other processors that support the extended PowerPC V2.05
14602 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
14603 point instructions implemented on some POWER processors.
14605 The @option{-mpowerpc64} option allows GCC to generate the additional
14606 64-bit instructions that are found in the full PowerPC64 architecture
14607 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
14608 @option{-mno-powerpc64}.
14610 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
14611 will use only the instructions in the common subset of both
14612 architectures plus some special AIX common-mode calls, and will not use
14613 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
14614 permits GCC to use any instruction from either architecture and to
14615 allow use of the MQ register; specify this for the Motorola MPC601.
14617 @item -mnew-mnemonics
14618 @itemx -mold-mnemonics
14619 @opindex mnew-mnemonics
14620 @opindex mold-mnemonics
14621 Select which mnemonics to use in the generated assembler code. With
14622 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
14623 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
14624 assembler mnemonics defined for the POWER architecture. Instructions
14625 defined in only one architecture have only one mnemonic; GCC uses that
14626 mnemonic irrespective of which of these options is specified.
14628 GCC defaults to the mnemonics appropriate for the architecture in
14629 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
14630 value of these option. Unless you are building a cross-compiler, you
14631 should normally not specify either @option{-mnew-mnemonics} or
14632 @option{-mold-mnemonics}, but should instead accept the default.
14634 @item -mcpu=@var{cpu_type}
14636 Set architecture type, register usage, choice of mnemonics, and
14637 instruction scheduling parameters for machine type @var{cpu_type}.
14638 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
14639 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
14640 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
14641 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
14642 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
14643 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
14644 @samp{e300c3}, @samp{e500mc}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
14645 @samp{power}, @samp{power2}, @samp{power3}, @samp{power4},
14646 @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x}, @samp{power7},
14647 @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
14648 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
14650 @option{-mcpu=common} selects a completely generic processor. Code
14651 generated under this option will run on any POWER or PowerPC processor.
14652 GCC will use only the instructions in the common subset of both
14653 architectures, and will not use the MQ register. GCC assumes a generic
14654 processor model for scheduling purposes.
14656 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
14657 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
14658 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
14659 types, with an appropriate, generic processor model assumed for
14660 scheduling purposes.
14662 The other options specify a specific processor. Code generated under
14663 those options will run best on that processor, and may not run at all on
14666 The @option{-mcpu} options automatically enable or disable the
14669 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
14670 -mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol
14671 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
14672 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx}
14674 The particular options set for any particular CPU will vary between
14675 compiler versions, depending on what setting seems to produce optimal
14676 code for that CPU; it doesn't necessarily reflect the actual hardware's
14677 capabilities. If you wish to set an individual option to a particular
14678 value, you may specify it after the @option{-mcpu} option, like
14679 @samp{-mcpu=970 -mno-altivec}.
14681 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
14682 not enabled or disabled by the @option{-mcpu} option at present because
14683 AIX does not have full support for these options. You may still
14684 enable or disable them individually if you're sure it'll work in your
14687 @item -mtune=@var{cpu_type}
14689 Set the instruction scheduling parameters for machine type
14690 @var{cpu_type}, but do not set the architecture type, register usage, or
14691 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
14692 values for @var{cpu_type} are used for @option{-mtune} as for
14693 @option{-mcpu}. If both are specified, the code generated will use the
14694 architecture, registers, and mnemonics set by @option{-mcpu}, but the
14695 scheduling parameters set by @option{-mtune}.
14701 Generate code to compute division as reciprocal estimate and iterative
14702 refinement, creating opportunities for increased throughput. This
14703 feature requires: optional PowerPC Graphics instruction set for single
14704 precision and FRE instruction for double precision, assuming divides
14705 cannot generate user-visible traps, and the domain values not include
14706 Infinities, denormals or zero denominator.
14709 @itemx -mno-altivec
14711 @opindex mno-altivec
14712 Generate code that uses (does not use) AltiVec instructions, and also
14713 enable the use of built-in functions that allow more direct access to
14714 the AltiVec instruction set. You may also need to set
14715 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
14721 @opindex mno-vrsave
14722 Generate VRSAVE instructions when generating AltiVec code.
14724 @item -mgen-cell-microcode
14725 @opindex mgen-cell-microcode
14726 Generate Cell microcode instructions
14728 @item -mwarn-cell-microcode
14729 @opindex mwarn-cell-microcode
14730 Warning when a Cell microcode instruction is going to emitted. An example
14731 of a Cell microcode instruction is a variable shift.
14734 @opindex msecure-plt
14735 Generate code that allows ld and ld.so to build executables and shared
14736 libraries with non-exec .plt and .got sections. This is a PowerPC
14737 32-bit SYSV ABI option.
14741 Generate code that uses a BSS .plt section that ld.so fills in, and
14742 requires .plt and .got sections that are both writable and executable.
14743 This is a PowerPC 32-bit SYSV ABI option.
14749 This switch enables or disables the generation of ISEL instructions.
14751 @item -misel=@var{yes/no}
14752 This switch has been deprecated. Use @option{-misel} and
14753 @option{-mno-isel} instead.
14759 This switch enables or disables the generation of SPE simd
14765 @opindex mno-paired
14766 This switch enables or disables the generation of PAIRED simd
14769 @item -mspe=@var{yes/no}
14770 This option has been deprecated. Use @option{-mspe} and
14771 @option{-mno-spe} instead.
14777 Generate code that uses (does not use) vector/scalar (VSX)
14778 instructions, and also enable the use of built-in functions that allow
14779 more direct access to the VSX instruction set.
14781 @item -mfloat-gprs=@var{yes/single/double/no}
14782 @itemx -mfloat-gprs
14783 @opindex mfloat-gprs
14784 This switch enables or disables the generation of floating point
14785 operations on the general purpose registers for architectures that
14788 The argument @var{yes} or @var{single} enables the use of
14789 single-precision floating point operations.
14791 The argument @var{double} enables the use of single and
14792 double-precision floating point operations.
14794 The argument @var{no} disables floating point operations on the
14795 general purpose registers.
14797 This option is currently only available on the MPC854x.
14803 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
14804 targets (including GNU/Linux). The 32-bit environment sets int, long
14805 and pointer to 32 bits and generates code that runs on any PowerPC
14806 variant. The 64-bit environment sets int to 32 bits and long and
14807 pointer to 64 bits, and generates code for PowerPC64, as for
14808 @option{-mpowerpc64}.
14811 @itemx -mno-fp-in-toc
14812 @itemx -mno-sum-in-toc
14813 @itemx -mminimal-toc
14815 @opindex mno-fp-in-toc
14816 @opindex mno-sum-in-toc
14817 @opindex mminimal-toc
14818 Modify generation of the TOC (Table Of Contents), which is created for
14819 every executable file. The @option{-mfull-toc} option is selected by
14820 default. In that case, GCC will allocate at least one TOC entry for
14821 each unique non-automatic variable reference in your program. GCC
14822 will also place floating-point constants in the TOC@. However, only
14823 16,384 entries are available in the TOC@.
14825 If you receive a linker error message that saying you have overflowed
14826 the available TOC space, you can reduce the amount of TOC space used
14827 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
14828 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
14829 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
14830 generate code to calculate the sum of an address and a constant at
14831 run-time instead of putting that sum into the TOC@. You may specify one
14832 or both of these options. Each causes GCC to produce very slightly
14833 slower and larger code at the expense of conserving TOC space.
14835 If you still run out of space in the TOC even when you specify both of
14836 these options, specify @option{-mminimal-toc} instead. This option causes
14837 GCC to make only one TOC entry for every file. When you specify this
14838 option, GCC will produce code that is slower and larger but which
14839 uses extremely little TOC space. You may wish to use this option
14840 only on files that contain less frequently executed code.
14846 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
14847 @code{long} type, and the infrastructure needed to support them.
14848 Specifying @option{-maix64} implies @option{-mpowerpc64} and
14849 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
14850 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
14853 @itemx -mno-xl-compat
14854 @opindex mxl-compat
14855 @opindex mno-xl-compat
14856 Produce code that conforms more closely to IBM XL compiler semantics
14857 when using AIX-compatible ABI@. Pass floating-point arguments to
14858 prototyped functions beyond the register save area (RSA) on the stack
14859 in addition to argument FPRs. Do not assume that most significant
14860 double in 128-bit long double value is properly rounded when comparing
14861 values and converting to double. Use XL symbol names for long double
14864 The AIX calling convention was extended but not initially documented to
14865 handle an obscure K&R C case of calling a function that takes the
14866 address of its arguments with fewer arguments than declared. IBM XL
14867 compilers access floating point arguments which do not fit in the
14868 RSA from the stack when a subroutine is compiled without
14869 optimization. Because always storing floating-point arguments on the
14870 stack is inefficient and rarely needed, this option is not enabled by
14871 default and only is necessary when calling subroutines compiled by IBM
14872 XL compilers without optimization.
14876 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
14877 application written to use message passing with special startup code to
14878 enable the application to run. The system must have PE installed in the
14879 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
14880 must be overridden with the @option{-specs=} option to specify the
14881 appropriate directory location. The Parallel Environment does not
14882 support threads, so the @option{-mpe} option and the @option{-pthread}
14883 option are incompatible.
14885 @item -malign-natural
14886 @itemx -malign-power
14887 @opindex malign-natural
14888 @opindex malign-power
14889 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
14890 @option{-malign-natural} overrides the ABI-defined alignment of larger
14891 types, such as floating-point doubles, on their natural size-based boundary.
14892 The option @option{-malign-power} instructs GCC to follow the ABI-specified
14893 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
14895 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
14899 @itemx -mhard-float
14900 @opindex msoft-float
14901 @opindex mhard-float
14902 Generate code that does not use (uses) the floating-point register set.
14903 Software floating point emulation is provided if you use the
14904 @option{-msoft-float} option, and pass the option to GCC when linking.
14906 @item -msingle-float
14907 @itemx -mdouble-float
14908 @opindex msingle-float
14909 @opindex mdouble-float
14910 Generate code for single or double-precision floating point operations.
14911 @option{-mdouble-float} implies @option{-msingle-float}.
14914 @opindex msimple-fpu
14915 Do not generate sqrt and div instructions for hardware floating point unit.
14919 Specify type of floating point unit. Valid values are @var{sp_lite}
14920 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
14921 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
14922 and @var{dp_full} (equivalent to -mdouble-float).
14925 @opindex mxilinx-fpu
14926 Perform optimizations for floating point unit on Xilinx PPC 405/440.
14929 @itemx -mno-multiple
14931 @opindex mno-multiple
14932 Generate code that uses (does not use) the load multiple word
14933 instructions and the store multiple word instructions. These
14934 instructions are generated by default on POWER systems, and not
14935 generated on PowerPC systems. Do not use @option{-mmultiple} on little
14936 endian PowerPC systems, since those instructions do not work when the
14937 processor is in little endian mode. The exceptions are PPC740 and
14938 PPC750 which permit the instructions usage in little endian mode.
14943 @opindex mno-string
14944 Generate code that uses (does not use) the load string instructions
14945 and the store string word instructions to save multiple registers and
14946 do small block moves. These instructions are generated by default on
14947 POWER systems, and not generated on PowerPC systems. Do not use
14948 @option{-mstring} on little endian PowerPC systems, since those
14949 instructions do not work when the processor is in little endian mode.
14950 The exceptions are PPC740 and PPC750 which permit the instructions
14951 usage in little endian mode.
14956 @opindex mno-update
14957 Generate code that uses (does not use) the load or store instructions
14958 that update the base register to the address of the calculated memory
14959 location. These instructions are generated by default. If you use
14960 @option{-mno-update}, there is a small window between the time that the
14961 stack pointer is updated and the address of the previous frame is
14962 stored, which means code that walks the stack frame across interrupts or
14963 signals may get corrupted data.
14965 @item -mavoid-indexed-addresses
14966 @item -mno-avoid-indexed-addresses
14967 @opindex mavoid-indexed-addresses
14968 @opindex mno-avoid-indexed-addresses
14969 Generate code that tries to avoid (not avoid) the use of indexed load
14970 or store instructions. These instructions can incur a performance
14971 penalty on Power6 processors in certain situations, such as when
14972 stepping through large arrays that cross a 16M boundary. This option
14973 is enabled by default when targetting Power6 and disabled otherwise.
14976 @itemx -mno-fused-madd
14977 @opindex mfused-madd
14978 @opindex mno-fused-madd
14979 Generate code that uses (does not use) the floating point multiply and
14980 accumulate instructions. These instructions are generated by default if
14981 hardware floating is used.
14987 Generate code that uses (does not use) the half-word multiply and
14988 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
14989 These instructions are generated by default when targetting those
14996 Generate code that uses (does not use) the string-search @samp{dlmzb}
14997 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
14998 generated by default when targetting those processors.
15000 @item -mno-bit-align
15002 @opindex mno-bit-align
15003 @opindex mbit-align
15004 On System V.4 and embedded PowerPC systems do not (do) force structures
15005 and unions that contain bit-fields to be aligned to the base type of the
15008 For example, by default a structure containing nothing but 8
15009 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
15010 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
15011 the structure would be aligned to a 1 byte boundary and be one byte in
15014 @item -mno-strict-align
15015 @itemx -mstrict-align
15016 @opindex mno-strict-align
15017 @opindex mstrict-align
15018 On System V.4 and embedded PowerPC systems do not (do) assume that
15019 unaligned memory references will be handled by the system.
15021 @item -mrelocatable
15022 @itemx -mno-relocatable
15023 @opindex mrelocatable
15024 @opindex mno-relocatable
15025 On embedded PowerPC systems generate code that allows (does not allow)
15026 the program to be relocated to a different address at runtime. If you
15027 use @option{-mrelocatable} on any module, all objects linked together must
15028 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
15030 @item -mrelocatable-lib
15031 @itemx -mno-relocatable-lib
15032 @opindex mrelocatable-lib
15033 @opindex mno-relocatable-lib
15034 On embedded PowerPC systems generate code that allows (does not allow)
15035 the program to be relocated to a different address at runtime. Modules
15036 compiled with @option{-mrelocatable-lib} can be linked with either modules
15037 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
15038 with modules compiled with the @option{-mrelocatable} options.
15044 On System V.4 and embedded PowerPC systems do not (do) assume that
15045 register 2 contains a pointer to a global area pointing to the addresses
15046 used in the program.
15049 @itemx -mlittle-endian
15051 @opindex mlittle-endian
15052 On System V.4 and embedded PowerPC systems compile code for the
15053 processor in little endian mode. The @option{-mlittle-endian} option is
15054 the same as @option{-mlittle}.
15057 @itemx -mbig-endian
15059 @opindex mbig-endian
15060 On System V.4 and embedded PowerPC systems compile code for the
15061 processor in big endian mode. The @option{-mbig-endian} option is
15062 the same as @option{-mbig}.
15064 @item -mdynamic-no-pic
15065 @opindex mdynamic-no-pic
15066 On Darwin and Mac OS X systems, compile code so that it is not
15067 relocatable, but that its external references are relocatable. The
15068 resulting code is suitable for applications, but not shared
15071 @item -mprioritize-restricted-insns=@var{priority}
15072 @opindex mprioritize-restricted-insns
15073 This option controls the priority that is assigned to
15074 dispatch-slot restricted instructions during the second scheduling
15075 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
15076 @var{no/highest/second-highest} priority to dispatch slot restricted
15079 @item -msched-costly-dep=@var{dependence_type}
15080 @opindex msched-costly-dep
15081 This option controls which dependences are considered costly
15082 by the target during instruction scheduling. The argument
15083 @var{dependence_type} takes one of the following values:
15084 @var{no}: no dependence is costly,
15085 @var{all}: all dependences are costly,
15086 @var{true_store_to_load}: a true dependence from store to load is costly,
15087 @var{store_to_load}: any dependence from store to load is costly,
15088 @var{number}: any dependence which latency >= @var{number} is costly.
15090 @item -minsert-sched-nops=@var{scheme}
15091 @opindex minsert-sched-nops
15092 This option controls which nop insertion scheme will be used during
15093 the second scheduling pass. The argument @var{scheme} takes one of the
15095 @var{no}: Don't insert nops.
15096 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
15097 according to the scheduler's grouping.
15098 @var{regroup_exact}: Insert nops to force costly dependent insns into
15099 separate groups. Insert exactly as many nops as needed to force an insn
15100 to a new group, according to the estimated processor grouping.
15101 @var{number}: Insert nops to force costly dependent insns into
15102 separate groups. Insert @var{number} nops to force an insn to a new group.
15105 @opindex mcall-sysv
15106 On System V.4 and embedded PowerPC systems compile code using calling
15107 conventions that adheres to the March 1995 draft of the System V
15108 Application Binary Interface, PowerPC processor supplement. This is the
15109 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
15111 @item -mcall-sysv-eabi
15113 @opindex mcall-sysv-eabi
15114 @opindex mcall-eabi
15115 Specify both @option{-mcall-sysv} and @option{-meabi} options.
15117 @item -mcall-sysv-noeabi
15118 @opindex mcall-sysv-noeabi
15119 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
15121 @item -mcall-aixdesc
15123 On System V.4 and embedded PowerPC systems compile code for the AIX
15127 @opindex mcall-linux
15128 On System V.4 and embedded PowerPC systems compile code for the
15129 Linux-based GNU system.
15133 On System V.4 and embedded PowerPC systems compile code for the
15134 Hurd-based GNU system.
15136 @item -mcall-freebsd
15137 @opindex mcall-freebsd
15138 On System V.4 and embedded PowerPC systems compile code for the
15139 FreeBSD operating system.
15141 @item -mcall-netbsd
15142 @opindex mcall-netbsd
15143 On System V.4 and embedded PowerPC systems compile code for the
15144 NetBSD operating system.
15146 @item -mcall-openbsd
15147 @opindex mcall-netbsd
15148 On System V.4 and embedded PowerPC systems compile code for the
15149 OpenBSD operating system.
15151 @item -maix-struct-return
15152 @opindex maix-struct-return
15153 Return all structures in memory (as specified by the AIX ABI)@.
15155 @item -msvr4-struct-return
15156 @opindex msvr4-struct-return
15157 Return structures smaller than 8 bytes in registers (as specified by the
15160 @item -mabi=@var{abi-type}
15162 Extend the current ABI with a particular extension, or remove such extension.
15163 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
15164 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
15168 Extend the current ABI with SPE ABI extensions. This does not change
15169 the default ABI, instead it adds the SPE ABI extensions to the current
15173 @opindex mabi=no-spe
15174 Disable Booke SPE ABI extensions for the current ABI@.
15176 @item -mabi=ibmlongdouble
15177 @opindex mabi=ibmlongdouble
15178 Change the current ABI to use IBM extended precision long double.
15179 This is a PowerPC 32-bit SYSV ABI option.
15181 @item -mabi=ieeelongdouble
15182 @opindex mabi=ieeelongdouble
15183 Change the current ABI to use IEEE extended precision long double.
15184 This is a PowerPC 32-bit Linux ABI option.
15187 @itemx -mno-prototype
15188 @opindex mprototype
15189 @opindex mno-prototype
15190 On System V.4 and embedded PowerPC systems assume that all calls to
15191 variable argument functions are properly prototyped. Otherwise, the
15192 compiler must insert an instruction before every non prototyped call to
15193 set or clear bit 6 of the condition code register (@var{CR}) to
15194 indicate whether floating point values were passed in the floating point
15195 registers in case the function takes a variable arguments. With
15196 @option{-mprototype}, only calls to prototyped variable argument functions
15197 will set or clear the bit.
15201 On embedded PowerPC systems, assume that the startup module is called
15202 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
15203 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
15208 On embedded PowerPC systems, assume that the startup module is called
15209 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
15214 On embedded PowerPC systems, assume that the startup module is called
15215 @file{crt0.o} and the standard C libraries are @file{libads.a} and
15218 @item -myellowknife
15219 @opindex myellowknife
15220 On embedded PowerPC systems, assume that the startup module is called
15221 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
15226 On System V.4 and embedded PowerPC systems, specify that you are
15227 compiling for a VxWorks system.
15231 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
15232 header to indicate that @samp{eabi} extended relocations are used.
15238 On System V.4 and embedded PowerPC systems do (do not) adhere to the
15239 Embedded Applications Binary Interface (eabi) which is a set of
15240 modifications to the System V.4 specifications. Selecting @option{-meabi}
15241 means that the stack is aligned to an 8 byte boundary, a function
15242 @code{__eabi} is called to from @code{main} to set up the eabi
15243 environment, and the @option{-msdata} option can use both @code{r2} and
15244 @code{r13} to point to two separate small data areas. Selecting
15245 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
15246 do not call an initialization function from @code{main}, and the
15247 @option{-msdata} option will only use @code{r13} to point to a single
15248 small data area. The @option{-meabi} option is on by default if you
15249 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
15252 @opindex msdata=eabi
15253 On System V.4 and embedded PowerPC systems, put small initialized
15254 @code{const} global and static data in the @samp{.sdata2} section, which
15255 is pointed to by register @code{r2}. Put small initialized
15256 non-@code{const} global and static data in the @samp{.sdata} section,
15257 which is pointed to by register @code{r13}. Put small uninitialized
15258 global and static data in the @samp{.sbss} section, which is adjacent to
15259 the @samp{.sdata} section. The @option{-msdata=eabi} option is
15260 incompatible with the @option{-mrelocatable} option. The
15261 @option{-msdata=eabi} option also sets the @option{-memb} option.
15264 @opindex msdata=sysv
15265 On System V.4 and embedded PowerPC systems, put small global and static
15266 data in the @samp{.sdata} section, which is pointed to by register
15267 @code{r13}. Put small uninitialized global and static data in the
15268 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
15269 The @option{-msdata=sysv} option is incompatible with the
15270 @option{-mrelocatable} option.
15272 @item -msdata=default
15274 @opindex msdata=default
15276 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
15277 compile code the same as @option{-msdata=eabi}, otherwise compile code the
15278 same as @option{-msdata=sysv}.
15281 @opindex msdata=data
15282 On System V.4 and embedded PowerPC systems, put small global
15283 data in the @samp{.sdata} section. Put small uninitialized global
15284 data in the @samp{.sbss} section. Do not use register @code{r13}
15285 to address small data however. This is the default behavior unless
15286 other @option{-msdata} options are used.
15290 @opindex msdata=none
15292 On embedded PowerPC systems, put all initialized global and static data
15293 in the @samp{.data} section, and all uninitialized data in the
15294 @samp{.bss} section.
15298 @cindex smaller data references (PowerPC)
15299 @cindex .sdata/.sdata2 references (PowerPC)
15300 On embedded PowerPC systems, put global and static items less than or
15301 equal to @var{num} bytes into the small data or bss sections instead of
15302 the normal data or bss section. By default, @var{num} is 8. The
15303 @option{-G @var{num}} switch is also passed to the linker.
15304 All modules should be compiled with the same @option{-G @var{num}} value.
15307 @itemx -mno-regnames
15309 @opindex mno-regnames
15310 On System V.4 and embedded PowerPC systems do (do not) emit register
15311 names in the assembly language output using symbolic forms.
15314 @itemx -mno-longcall
15316 @opindex mno-longcall
15317 By default assume that all calls are far away so that a longer more
15318 expensive calling sequence is required. This is required for calls
15319 further than 32 megabytes (33,554,432 bytes) from the current location.
15320 A short call will be generated if the compiler knows
15321 the call cannot be that far away. This setting can be overridden by
15322 the @code{shortcall} function attribute, or by @code{#pragma
15325 Some linkers are capable of detecting out-of-range calls and generating
15326 glue code on the fly. On these systems, long calls are unnecessary and
15327 generate slower code. As of this writing, the AIX linker can do this,
15328 as can the GNU linker for PowerPC/64. It is planned to add this feature
15329 to the GNU linker for 32-bit PowerPC systems as well.
15331 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
15332 callee, L42'', plus a ``branch island'' (glue code). The two target
15333 addresses represent the callee and the ``branch island''. The
15334 Darwin/PPC linker will prefer the first address and generate a ``bl
15335 callee'' if the PPC ``bl'' instruction will reach the callee directly;
15336 otherwise, the linker will generate ``bl L42'' to call the ``branch
15337 island''. The ``branch island'' is appended to the body of the
15338 calling function; it computes the full 32-bit address of the callee
15341 On Mach-O (Darwin) systems, this option directs the compiler emit to
15342 the glue for every direct call, and the Darwin linker decides whether
15343 to use or discard it.
15345 In the future, we may cause GCC to ignore all longcall specifications
15346 when the linker is known to generate glue.
15348 @item -mtls-markers
15349 @itemx -mno-tls-markers
15350 @opindex mtls-markers
15351 @opindex mno-tls-markers
15352 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
15353 specifying the function argument. The relocation allows ld to
15354 reliably associate function call with argument setup instructions for
15355 TLS optimization, which in turn allows gcc to better schedule the
15360 Adds support for multithreading with the @dfn{pthreads} library.
15361 This option sets flags for both the preprocessor and linker.
15365 @node S/390 and zSeries Options
15366 @subsection S/390 and zSeries Options
15367 @cindex S/390 and zSeries Options
15369 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
15373 @itemx -msoft-float
15374 @opindex mhard-float
15375 @opindex msoft-float
15376 Use (do not use) the hardware floating-point instructions and registers
15377 for floating-point operations. When @option{-msoft-float} is specified,
15378 functions in @file{libgcc.a} will be used to perform floating-point
15379 operations. When @option{-mhard-float} is specified, the compiler
15380 generates IEEE floating-point instructions. This is the default.
15383 @itemx -mno-hard-dfp
15385 @opindex mno-hard-dfp
15386 Use (do not use) the hardware decimal-floating-point instructions for
15387 decimal-floating-point operations. When @option{-mno-hard-dfp} is
15388 specified, functions in @file{libgcc.a} will be used to perform
15389 decimal-floating-point operations. When @option{-mhard-dfp} is
15390 specified, the compiler generates decimal-floating-point hardware
15391 instructions. This is the default for @option{-march=z9-ec} or higher.
15393 @item -mlong-double-64
15394 @itemx -mlong-double-128
15395 @opindex mlong-double-64
15396 @opindex mlong-double-128
15397 These switches control the size of @code{long double} type. A size
15398 of 64bit makes the @code{long double} type equivalent to the @code{double}
15399 type. This is the default.
15402 @itemx -mno-backchain
15403 @opindex mbackchain
15404 @opindex mno-backchain
15405 Store (do not store) the address of the caller's frame as backchain pointer
15406 into the callee's stack frame.
15407 A backchain may be needed to allow debugging using tools that do not understand
15408 DWARF-2 call frame information.
15409 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
15410 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
15411 the backchain is placed into the topmost word of the 96/160 byte register
15414 In general, code compiled with @option{-mbackchain} is call-compatible with
15415 code compiled with @option{-mmo-backchain}; however, use of the backchain
15416 for debugging purposes usually requires that the whole binary is built with
15417 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
15418 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
15419 to build a linux kernel use @option{-msoft-float}.
15421 The default is to not maintain the backchain.
15423 @item -mpacked-stack
15424 @itemx -mno-packed-stack
15425 @opindex mpacked-stack
15426 @opindex mno-packed-stack
15427 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
15428 specified, the compiler uses the all fields of the 96/160 byte register save
15429 area only for their default purpose; unused fields still take up stack space.
15430 When @option{-mpacked-stack} is specified, register save slots are densely
15431 packed at the top of the register save area; unused space is reused for other
15432 purposes, allowing for more efficient use of the available stack space.
15433 However, when @option{-mbackchain} is also in effect, the topmost word of
15434 the save area is always used to store the backchain, and the return address
15435 register is always saved two words below the backchain.
15437 As long as the stack frame backchain is not used, code generated with
15438 @option{-mpacked-stack} is call-compatible with code generated with
15439 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
15440 S/390 or zSeries generated code that uses the stack frame backchain at run
15441 time, not just for debugging purposes. Such code is not call-compatible
15442 with code compiled with @option{-mpacked-stack}. Also, note that the
15443 combination of @option{-mbackchain},
15444 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
15445 to build a linux kernel use @option{-msoft-float}.
15447 The default is to not use the packed stack layout.
15450 @itemx -mno-small-exec
15451 @opindex msmall-exec
15452 @opindex mno-small-exec
15453 Generate (or do not generate) code using the @code{bras} instruction
15454 to do subroutine calls.
15455 This only works reliably if the total executable size does not
15456 exceed 64k. The default is to use the @code{basr} instruction instead,
15457 which does not have this limitation.
15463 When @option{-m31} is specified, generate code compliant to the
15464 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
15465 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
15466 particular to generate 64-bit instructions. For the @samp{s390}
15467 targets, the default is @option{-m31}, while the @samp{s390x}
15468 targets default to @option{-m64}.
15474 When @option{-mzarch} is specified, generate code using the
15475 instructions available on z/Architecture.
15476 When @option{-mesa} is specified, generate code using the
15477 instructions available on ESA/390. Note that @option{-mesa} is
15478 not possible with @option{-m64}.
15479 When generating code compliant to the GNU/Linux for S/390 ABI,
15480 the default is @option{-mesa}. When generating code compliant
15481 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
15487 Generate (or do not generate) code using the @code{mvcle} instruction
15488 to perform block moves. When @option{-mno-mvcle} is specified,
15489 use a @code{mvc} loop instead. This is the default unless optimizing for
15496 Print (or do not print) additional debug information when compiling.
15497 The default is to not print debug information.
15499 @item -march=@var{cpu-type}
15501 Generate code that will run on @var{cpu-type}, which is the name of a system
15502 representing a certain processor type. Possible values for
15503 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
15504 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
15505 When generating code using the instructions available on z/Architecture,
15506 the default is @option{-march=z900}. Otherwise, the default is
15507 @option{-march=g5}.
15509 @item -mtune=@var{cpu-type}
15511 Tune to @var{cpu-type} everything applicable about the generated code,
15512 except for the ABI and the set of available instructions.
15513 The list of @var{cpu-type} values is the same as for @option{-march}.
15514 The default is the value used for @option{-march}.
15517 @itemx -mno-tpf-trace
15518 @opindex mtpf-trace
15519 @opindex mno-tpf-trace
15520 Generate code that adds (does not add) in TPF OS specific branches to trace
15521 routines in the operating system. This option is off by default, even
15522 when compiling for the TPF OS@.
15525 @itemx -mno-fused-madd
15526 @opindex mfused-madd
15527 @opindex mno-fused-madd
15528 Generate code that uses (does not use) the floating point multiply and
15529 accumulate instructions. These instructions are generated by default if
15530 hardware floating point is used.
15532 @item -mwarn-framesize=@var{framesize}
15533 @opindex mwarn-framesize
15534 Emit a warning if the current function exceeds the given frame size. Because
15535 this is a compile time check it doesn't need to be a real problem when the program
15536 runs. It is intended to identify functions which most probably cause
15537 a stack overflow. It is useful to be used in an environment with limited stack
15538 size e.g.@: the linux kernel.
15540 @item -mwarn-dynamicstack
15541 @opindex mwarn-dynamicstack
15542 Emit a warning if the function calls alloca or uses dynamically
15543 sized arrays. This is generally a bad idea with a limited stack size.
15545 @item -mstack-guard=@var{stack-guard}
15546 @itemx -mstack-size=@var{stack-size}
15547 @opindex mstack-guard
15548 @opindex mstack-size
15549 If these options are provided the s390 back end emits additional instructions in
15550 the function prologue which trigger a trap if the stack size is @var{stack-guard}
15551 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
15552 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
15553 the frame size of the compiled function is chosen.
15554 These options are intended to be used to help debugging stack overflow problems.
15555 The additionally emitted code causes only little overhead and hence can also be
15556 used in production like systems without greater performance degradation. The given
15557 values have to be exact powers of 2 and @var{stack-size} has to be greater than
15558 @var{stack-guard} without exceeding 64k.
15559 In order to be efficient the extra code makes the assumption that the stack starts
15560 at an address aligned to the value given by @var{stack-size}.
15561 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
15564 @node Score Options
15565 @subsection Score Options
15566 @cindex Score Options
15568 These options are defined for Score implementations:
15573 Compile code for big endian mode. This is the default.
15577 Compile code for little endian mode.
15581 Disable generate bcnz instruction.
15585 Enable generate unaligned load and store instruction.
15589 Enable the use of multiply-accumulate instructions. Disabled by default.
15593 Specify the SCORE5 as the target architecture.
15597 Specify the SCORE5U of the target architecture.
15601 Specify the SCORE7 as the target architecture. This is the default.
15605 Specify the SCORE7D as the target architecture.
15609 @subsection SH Options
15611 These @samp{-m} options are defined for the SH implementations:
15616 Generate code for the SH1.
15620 Generate code for the SH2.
15623 Generate code for the SH2e.
15627 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
15628 that the floating-point unit is not used.
15630 @item -m2a-single-only
15631 @opindex m2a-single-only
15632 Generate code for the SH2a-FPU, in such a way that no double-precision
15633 floating point operations are used.
15636 @opindex m2a-single
15637 Generate code for the SH2a-FPU assuming the floating-point unit is in
15638 single-precision mode by default.
15642 Generate code for the SH2a-FPU assuming the floating-point unit is in
15643 double-precision mode by default.
15647 Generate code for the SH3.
15651 Generate code for the SH3e.
15655 Generate code for the SH4 without a floating-point unit.
15657 @item -m4-single-only
15658 @opindex m4-single-only
15659 Generate code for the SH4 with a floating-point unit that only
15660 supports single-precision arithmetic.
15664 Generate code for the SH4 assuming the floating-point unit is in
15665 single-precision mode by default.
15669 Generate code for the SH4.
15673 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
15674 floating-point unit is not used.
15676 @item -m4a-single-only
15677 @opindex m4a-single-only
15678 Generate code for the SH4a, in such a way that no double-precision
15679 floating point operations are used.
15682 @opindex m4a-single
15683 Generate code for the SH4a assuming the floating-point unit is in
15684 single-precision mode by default.
15688 Generate code for the SH4a.
15692 Same as @option{-m4a-nofpu}, except that it implicitly passes
15693 @option{-dsp} to the assembler. GCC doesn't generate any DSP
15694 instructions at the moment.
15698 Compile code for the processor in big endian mode.
15702 Compile code for the processor in little endian mode.
15706 Align doubles at 64-bit boundaries. Note that this changes the calling
15707 conventions, and thus some functions from the standard C library will
15708 not work unless you recompile it first with @option{-mdalign}.
15712 Shorten some address references at link time, when possible; uses the
15713 linker option @option{-relax}.
15717 Use 32-bit offsets in @code{switch} tables. The default is to use
15722 Enable the use of bit manipulation instructions on SH2A.
15726 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
15727 alignment constraints.
15731 Comply with the calling conventions defined by Renesas.
15735 Comply with the calling conventions defined by Renesas.
15739 Comply with the calling conventions defined for GCC before the Renesas
15740 conventions were available. This option is the default for all
15741 targets of the SH toolchain except for @samp{sh-symbianelf}.
15744 @opindex mnomacsave
15745 Mark the @code{MAC} register as call-clobbered, even if
15746 @option{-mhitachi} is given.
15750 Increase IEEE-compliance of floating-point code.
15751 At the moment, this is equivalent to @option{-fno-finite-math-only}.
15752 When generating 16 bit SH opcodes, getting IEEE-conforming results for
15753 comparisons of NANs / infinities incurs extra overhead in every
15754 floating point comparison, therefore the default is set to
15755 @option{-ffinite-math-only}.
15757 @item -minline-ic_invalidate
15758 @opindex minline-ic_invalidate
15759 Inline code to invalidate instruction cache entries after setting up
15760 nested function trampolines.
15761 This option has no effect if -musermode is in effect and the selected
15762 code generation option (e.g. -m4) does not allow the use of the icbi
15764 If the selected code generation option does not allow the use of the icbi
15765 instruction, and -musermode is not in effect, the inlined code will
15766 manipulate the instruction cache address array directly with an associative
15767 write. This not only requires privileged mode, but it will also
15768 fail if the cache line had been mapped via the TLB and has become unmapped.
15772 Dump instruction size and location in the assembly code.
15775 @opindex mpadstruct
15776 This option is deprecated. It pads structures to multiple of 4 bytes,
15777 which is incompatible with the SH ABI@.
15781 Optimize for space instead of speed. Implied by @option{-Os}.
15784 @opindex mprefergot
15785 When generating position-independent code, emit function calls using
15786 the Global Offset Table instead of the Procedure Linkage Table.
15790 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
15791 if the inlined code would not work in user mode.
15792 This is the default when the target is @code{sh-*-linux*}.
15794 @item -multcost=@var{number}
15795 @opindex multcost=@var{number}
15796 Set the cost to assume for a multiply insn.
15798 @item -mdiv=@var{strategy}
15799 @opindex mdiv=@var{strategy}
15800 Set the division strategy to use for SHmedia code. @var{strategy} must be
15801 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
15802 inv:call2, inv:fp .
15803 "fp" performs the operation in floating point. This has a very high latency,
15804 but needs only a few instructions, so it might be a good choice if
15805 your code has enough easily exploitable ILP to allow the compiler to
15806 schedule the floating point instructions together with other instructions.
15807 Division by zero causes a floating point exception.
15808 "inv" uses integer operations to calculate the inverse of the divisor,
15809 and then multiplies the dividend with the inverse. This strategy allows
15810 cse and hoisting of the inverse calculation. Division by zero calculates
15811 an unspecified result, but does not trap.
15812 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
15813 have been found, or if the entire operation has been hoisted to the same
15814 place, the last stages of the inverse calculation are intertwined with the
15815 final multiply to reduce the overall latency, at the expense of using a few
15816 more instructions, and thus offering fewer scheduling opportunities with
15818 "call" calls a library function that usually implements the inv:minlat
15820 This gives high code density for m5-*media-nofpu compilations.
15821 "call2" uses a different entry point of the same library function, where it
15822 assumes that a pointer to a lookup table has already been set up, which
15823 exposes the pointer load to cse / code hoisting optimizations.
15824 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
15825 code generation, but if the code stays unoptimized, revert to the "call",
15826 "call2", or "fp" strategies, respectively. Note that the
15827 potentially-trapping side effect of division by zero is carried by a
15828 separate instruction, so it is possible that all the integer instructions
15829 are hoisted out, but the marker for the side effect stays where it is.
15830 A recombination to fp operations or a call is not possible in that case.
15831 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
15832 that the inverse calculation was nor separated from the multiply, they speed
15833 up division where the dividend fits into 20 bits (plus sign where applicable),
15834 by inserting a test to skip a number of operations in this case; this test
15835 slows down the case of larger dividends. inv20u assumes the case of a such
15836 a small dividend to be unlikely, and inv20l assumes it to be likely.
15838 @item -mdivsi3_libfunc=@var{name}
15839 @opindex mdivsi3_libfunc=@var{name}
15840 Set the name of the library function used for 32 bit signed division to
15841 @var{name}. This only affect the name used in the call and inv:call
15842 division strategies, and the compiler will still expect the same
15843 sets of input/output/clobbered registers as if this option was not present.
15845 @item -mfixed-range=@var{register-range}
15846 @opindex mfixed-range
15847 Generate code treating the given register range as fixed registers.
15848 A fixed register is one that the register allocator can not use. This is
15849 useful when compiling kernel code. A register range is specified as
15850 two registers separated by a dash. Multiple register ranges can be
15851 specified separated by a comma.
15853 @item -madjust-unroll
15854 @opindex madjust-unroll
15855 Throttle unrolling to avoid thrashing target registers.
15856 This option only has an effect if the gcc code base supports the
15857 TARGET_ADJUST_UNROLL_MAX target hook.
15859 @item -mindexed-addressing
15860 @opindex mindexed-addressing
15861 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
15862 This is only safe if the hardware and/or OS implement 32 bit wrap-around
15863 semantics for the indexed addressing mode. The architecture allows the
15864 implementation of processors with 64 bit MMU, which the OS could use to
15865 get 32 bit addressing, but since no current hardware implementation supports
15866 this or any other way to make the indexed addressing mode safe to use in
15867 the 32 bit ABI, the default is -mno-indexed-addressing.
15869 @item -mgettrcost=@var{number}
15870 @opindex mgettrcost=@var{number}
15871 Set the cost assumed for the gettr instruction to @var{number}.
15872 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
15876 Assume pt* instructions won't trap. This will generally generate better
15877 scheduled code, but is unsafe on current hardware. The current architecture
15878 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
15879 This has the unintentional effect of making it unsafe to schedule ptabs /
15880 ptrel before a branch, or hoist it out of a loop. For example,
15881 __do_global_ctors, a part of libgcc that runs constructors at program
15882 startup, calls functions in a list which is delimited by @minus{}1. With the
15883 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
15884 That means that all the constructors will be run a bit quicker, but when
15885 the loop comes to the end of the list, the program crashes because ptabs
15886 loads @minus{}1 into a target register. Since this option is unsafe for any
15887 hardware implementing the current architecture specification, the default
15888 is -mno-pt-fixed. Unless the user specifies a specific cost with
15889 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
15890 this deters register allocation using target registers for storing
15893 @item -minvalid-symbols
15894 @opindex minvalid-symbols
15895 Assume symbols might be invalid. Ordinary function symbols generated by
15896 the compiler will always be valid to load with movi/shori/ptabs or
15897 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
15898 to generate symbols that will cause ptabs / ptrel to trap.
15899 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
15900 It will then prevent cross-basic-block cse, hoisting and most scheduling
15901 of symbol loads. The default is @option{-mno-invalid-symbols}.
15904 @node SPARC Options
15905 @subsection SPARC Options
15906 @cindex SPARC options
15908 These @samp{-m} options are supported on the SPARC:
15911 @item -mno-app-regs
15913 @opindex mno-app-regs
15915 Specify @option{-mapp-regs} to generate output using the global registers
15916 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
15919 To be fully SVR4 ABI compliant at the cost of some performance loss,
15920 specify @option{-mno-app-regs}. You should compile libraries and system
15921 software with this option.
15924 @itemx -mhard-float
15926 @opindex mhard-float
15927 Generate output containing floating point instructions. This is the
15931 @itemx -msoft-float
15933 @opindex msoft-float
15934 Generate output containing library calls for floating point.
15935 @strong{Warning:} the requisite libraries are not available for all SPARC
15936 targets. Normally the facilities of the machine's usual C compiler are
15937 used, but this cannot be done directly in cross-compilation. You must make
15938 your own arrangements to provide suitable library functions for
15939 cross-compilation. The embedded targets @samp{sparc-*-aout} and
15940 @samp{sparclite-*-*} do provide software floating point support.
15942 @option{-msoft-float} changes the calling convention in the output file;
15943 therefore, it is only useful if you compile @emph{all} of a program with
15944 this option. In particular, you need to compile @file{libgcc.a}, the
15945 library that comes with GCC, with @option{-msoft-float} in order for
15948 @item -mhard-quad-float
15949 @opindex mhard-quad-float
15950 Generate output containing quad-word (long double) floating point
15953 @item -msoft-quad-float
15954 @opindex msoft-quad-float
15955 Generate output containing library calls for quad-word (long double)
15956 floating point instructions. The functions called are those specified
15957 in the SPARC ABI@. This is the default.
15959 As of this writing, there are no SPARC implementations that have hardware
15960 support for the quad-word floating point instructions. They all invoke
15961 a trap handler for one of these instructions, and then the trap handler
15962 emulates the effect of the instruction. Because of the trap handler overhead,
15963 this is much slower than calling the ABI library routines. Thus the
15964 @option{-msoft-quad-float} option is the default.
15966 @item -mno-unaligned-doubles
15967 @itemx -munaligned-doubles
15968 @opindex mno-unaligned-doubles
15969 @opindex munaligned-doubles
15970 Assume that doubles have 8 byte alignment. This is the default.
15972 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
15973 alignment only if they are contained in another type, or if they have an
15974 absolute address. Otherwise, it assumes they have 4 byte alignment.
15975 Specifying this option avoids some rare compatibility problems with code
15976 generated by other compilers. It is not the default because it results
15977 in a performance loss, especially for floating point code.
15979 @item -mno-faster-structs
15980 @itemx -mfaster-structs
15981 @opindex mno-faster-structs
15982 @opindex mfaster-structs
15983 With @option{-mfaster-structs}, the compiler assumes that structures
15984 should have 8 byte alignment. This enables the use of pairs of
15985 @code{ldd} and @code{std} instructions for copies in structure
15986 assignment, in place of twice as many @code{ld} and @code{st} pairs.
15987 However, the use of this changed alignment directly violates the SPARC
15988 ABI@. Thus, it's intended only for use on targets where the developer
15989 acknowledges that their resulting code will not be directly in line with
15990 the rules of the ABI@.
15992 @item -mimpure-text
15993 @opindex mimpure-text
15994 @option{-mimpure-text}, used in addition to @option{-shared}, tells
15995 the compiler to not pass @option{-z text} to the linker when linking a
15996 shared object. Using this option, you can link position-dependent
15997 code into a shared object.
15999 @option{-mimpure-text} suppresses the ``relocations remain against
16000 allocatable but non-writable sections'' linker error message.
16001 However, the necessary relocations will trigger copy-on-write, and the
16002 shared object is not actually shared across processes. Instead of
16003 using @option{-mimpure-text}, you should compile all source code with
16004 @option{-fpic} or @option{-fPIC}.
16006 This option is only available on SunOS and Solaris.
16008 @item -mcpu=@var{cpu_type}
16010 Set the instruction set, register set, and instruction scheduling parameters
16011 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
16012 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
16013 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
16014 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
16015 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
16017 Default instruction scheduling parameters are used for values that select
16018 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
16019 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
16021 Here is a list of each supported architecture and their supported
16026 v8: supersparc, hypersparc
16027 sparclite: f930, f934, sparclite86x
16029 v9: ultrasparc, ultrasparc3, niagara, niagara2
16032 By default (unless configured otherwise), GCC generates code for the V7
16033 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
16034 additionally optimizes it for the Cypress CY7C602 chip, as used in the
16035 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
16036 SPARCStation 1, 2, IPX etc.
16038 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
16039 architecture. The only difference from V7 code is that the compiler emits
16040 the integer multiply and integer divide instructions which exist in SPARC-V8
16041 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
16042 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
16045 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
16046 the SPARC architecture. This adds the integer multiply, integer divide step
16047 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
16048 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
16049 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
16050 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
16051 MB86934 chip, which is the more recent SPARClite with FPU@.
16053 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
16054 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
16055 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
16056 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
16057 optimizes it for the TEMIC SPARClet chip.
16059 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
16060 architecture. This adds 64-bit integer and floating-point move instructions,
16061 3 additional floating-point condition code registers and conditional move
16062 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
16063 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
16064 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
16065 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
16066 @option{-mcpu=niagara}, the compiler additionally optimizes it for
16067 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
16068 additionally optimizes it for Sun UltraSPARC T2 chips.
16070 @item -mtune=@var{cpu_type}
16072 Set the instruction scheduling parameters for machine type
16073 @var{cpu_type}, but do not set the instruction set or register set that the
16074 option @option{-mcpu=@var{cpu_type}} would.
16076 The same values for @option{-mcpu=@var{cpu_type}} can be used for
16077 @option{-mtune=@var{cpu_type}}, but the only useful values are those
16078 that select a particular cpu implementation. Those are @samp{cypress},
16079 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
16080 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
16081 @samp{ultrasparc3}, @samp{niagara}, and @samp{niagara2}.
16086 @opindex mno-v8plus
16087 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
16088 difference from the V8 ABI is that the global and out registers are
16089 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
16090 mode for all SPARC-V9 processors.
16096 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
16097 Visual Instruction Set extensions. The default is @option{-mno-vis}.
16100 These @samp{-m} options are supported in addition to the above
16101 on SPARC-V9 processors in 64-bit environments:
16104 @item -mlittle-endian
16105 @opindex mlittle-endian
16106 Generate code for a processor running in little-endian mode. It is only
16107 available for a few configurations and most notably not on Solaris and Linux.
16113 Generate code for a 32-bit or 64-bit environment.
16114 The 32-bit environment sets int, long and pointer to 32 bits.
16115 The 64-bit environment sets int to 32 bits and long and pointer
16118 @item -mcmodel=medlow
16119 @opindex mcmodel=medlow
16120 Generate code for the Medium/Low code model: 64-bit addresses, programs
16121 must be linked in the low 32 bits of memory. Programs can be statically
16122 or dynamically linked.
16124 @item -mcmodel=medmid
16125 @opindex mcmodel=medmid
16126 Generate code for the Medium/Middle code model: 64-bit addresses, programs
16127 must be linked in the low 44 bits of memory, the text and data segments must
16128 be less than 2GB in size and the data segment must be located within 2GB of
16131 @item -mcmodel=medany
16132 @opindex mcmodel=medany
16133 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
16134 may be linked anywhere in memory, the text and data segments must be less
16135 than 2GB in size and the data segment must be located within 2GB of the
16138 @item -mcmodel=embmedany
16139 @opindex mcmodel=embmedany
16140 Generate code for the Medium/Anywhere code model for embedded systems:
16141 64-bit addresses, the text and data segments must be less than 2GB in
16142 size, both starting anywhere in memory (determined at link time). The
16143 global register %g4 points to the base of the data segment. Programs
16144 are statically linked and PIC is not supported.
16147 @itemx -mno-stack-bias
16148 @opindex mstack-bias
16149 @opindex mno-stack-bias
16150 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
16151 frame pointer if present, are offset by @minus{}2047 which must be added back
16152 when making stack frame references. This is the default in 64-bit mode.
16153 Otherwise, assume no such offset is present.
16156 These switches are supported in addition to the above on Solaris:
16161 Add support for multithreading using the Solaris threads library. This
16162 option sets flags for both the preprocessor and linker. This option does
16163 not affect the thread safety of object code produced by the compiler or
16164 that of libraries supplied with it.
16168 Add support for multithreading using the POSIX threads library. This
16169 option sets flags for both the preprocessor and linker. This option does
16170 not affect the thread safety of object code produced by the compiler or
16171 that of libraries supplied with it.
16175 This is a synonym for @option{-pthreads}.
16179 @subsection SPU Options
16180 @cindex SPU options
16182 These @samp{-m} options are supported on the SPU:
16186 @itemx -merror-reloc
16187 @opindex mwarn-reloc
16188 @opindex merror-reloc
16190 The loader for SPU does not handle dynamic relocations. By default, GCC
16191 will give an error when it generates code that requires a dynamic
16192 relocation. @option{-mno-error-reloc} disables the error,
16193 @option{-mwarn-reloc} will generate a warning instead.
16196 @itemx -munsafe-dma
16198 @opindex munsafe-dma
16200 Instructions which initiate or test completion of DMA must not be
16201 reordered with respect to loads and stores of the memory which is being
16202 accessed. Users typically address this problem using the volatile
16203 keyword, but that can lead to inefficient code in places where the
16204 memory is known to not change. Rather than mark the memory as volatile
16205 we treat the DMA instructions as potentially effecting all memory. With
16206 @option{-munsafe-dma} users must use the volatile keyword to protect
16209 @item -mbranch-hints
16210 @opindex mbranch-hints
16212 By default, GCC will generate a branch hint instruction to avoid
16213 pipeline stalls for always taken or probably taken branches. A hint
16214 will not be generated closer than 8 instructions away from its branch.
16215 There is little reason to disable them, except for debugging purposes,
16216 or to make an object a little bit smaller.
16220 @opindex msmall-mem
16221 @opindex mlarge-mem
16223 By default, GCC generates code assuming that addresses are never larger
16224 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
16225 a full 32 bit address.
16230 By default, GCC links against startup code that assumes the SPU-style
16231 main function interface (which has an unconventional parameter list).
16232 With @option{-mstdmain}, GCC will link your program against startup
16233 code that assumes a C99-style interface to @code{main}, including a
16234 local copy of @code{argv} strings.
16236 @item -mfixed-range=@var{register-range}
16237 @opindex mfixed-range
16238 Generate code treating the given register range as fixed registers.
16239 A fixed register is one that the register allocator can not use. This is
16240 useful when compiling kernel code. A register range is specified as
16241 two registers separated by a dash. Multiple register ranges can be
16242 specified separated by a comma.
16245 @itemx -mdual-nops=@var{n}
16246 @opindex mdual-nops
16247 By default, GCC will insert nops to increase dual issue when it expects
16248 it to increase performance. @var{n} can be a value from 0 to 10. A
16249 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
16250 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
16252 @item -mhint-max-nops=@var{n}
16253 @opindex mhint-max-nops
16254 Maximum number of nops to insert for a branch hint. A branch hint must
16255 be at least 8 instructions away from the branch it is effecting. GCC
16256 will insert up to @var{n} nops to enforce this, otherwise it will not
16257 generate the branch hint.
16259 @item -mhint-max-distance=@var{n}
16260 @opindex mhint-max-distance
16261 The encoding of the branch hint instruction limits the hint to be within
16262 256 instructions of the branch it is effecting. By default, GCC makes
16263 sure it is within 125.
16266 @opindex msafe-hints
16267 Work around a hardware bug which causes the SPU to stall indefinitely.
16268 By default, GCC will insert the @code{hbrp} instruction to make sure
16269 this stall won't happen.
16273 @node System V Options
16274 @subsection Options for System V
16276 These additional options are available on System V Release 4 for
16277 compatibility with other compilers on those systems:
16282 Create a shared object.
16283 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
16287 Identify the versions of each tool used by the compiler, in a
16288 @code{.ident} assembler directive in the output.
16292 Refrain from adding @code{.ident} directives to the output file (this is
16295 @item -YP,@var{dirs}
16297 Search the directories @var{dirs}, and no others, for libraries
16298 specified with @option{-l}.
16300 @item -Ym,@var{dir}
16302 Look in the directory @var{dir} to find the M4 preprocessor.
16303 The assembler uses this option.
16304 @c This is supposed to go with a -Yd for predefined M4 macro files, but
16305 @c the generic assembler that comes with Solaris takes just -Ym.
16309 @subsection V850 Options
16310 @cindex V850 Options
16312 These @samp{-m} options are defined for V850 implementations:
16316 @itemx -mno-long-calls
16317 @opindex mlong-calls
16318 @opindex mno-long-calls
16319 Treat all calls as being far away (near). If calls are assumed to be
16320 far away, the compiler will always load the functions address up into a
16321 register, and call indirect through the pointer.
16327 Do not optimize (do optimize) basic blocks that use the same index
16328 pointer 4 or more times to copy pointer into the @code{ep} register, and
16329 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
16330 option is on by default if you optimize.
16332 @item -mno-prolog-function
16333 @itemx -mprolog-function
16334 @opindex mno-prolog-function
16335 @opindex mprolog-function
16336 Do not use (do use) external functions to save and restore registers
16337 at the prologue and epilogue of a function. The external functions
16338 are slower, but use less code space if more than one function saves
16339 the same number of registers. The @option{-mprolog-function} option
16340 is on by default if you optimize.
16344 Try to make the code as small as possible. At present, this just turns
16345 on the @option{-mep} and @option{-mprolog-function} options.
16347 @item -mtda=@var{n}
16349 Put static or global variables whose size is @var{n} bytes or less into
16350 the tiny data area that register @code{ep} points to. The tiny data
16351 area can hold up to 256 bytes in total (128 bytes for byte references).
16353 @item -msda=@var{n}
16355 Put static or global variables whose size is @var{n} bytes or less into
16356 the small data area that register @code{gp} points to. The small data
16357 area can hold up to 64 kilobytes.
16359 @item -mzda=@var{n}
16361 Put static or global variables whose size is @var{n} bytes or less into
16362 the first 32 kilobytes of memory.
16366 Specify that the target processor is the V850.
16369 @opindex mbig-switch
16370 Generate code suitable for big switch tables. Use this option only if
16371 the assembler/linker complain about out of range branches within a switch
16376 This option will cause r2 and r5 to be used in the code generated by
16377 the compiler. This setting is the default.
16379 @item -mno-app-regs
16380 @opindex mno-app-regs
16381 This option will cause r2 and r5 to be treated as fixed registers.
16385 Specify that the target processor is the V850E1. The preprocessor
16386 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
16387 this option is used.
16391 Specify that the target processor is the V850E@. The preprocessor
16392 constant @samp{__v850e__} will be defined if this option is used.
16394 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
16395 are defined then a default target processor will be chosen and the
16396 relevant @samp{__v850*__} preprocessor constant will be defined.
16398 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
16399 defined, regardless of which processor variant is the target.
16401 @item -mdisable-callt
16402 @opindex mdisable-callt
16403 This option will suppress generation of the CALLT instruction for the
16404 v850e and v850e1 flavors of the v850 architecture. The default is
16405 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
16410 @subsection VAX Options
16411 @cindex VAX options
16413 These @samp{-m} options are defined for the VAX:
16418 Do not output certain jump instructions (@code{aobleq} and so on)
16419 that the Unix assembler for the VAX cannot handle across long
16424 Do output those jump instructions, on the assumption that you
16425 will assemble with the GNU assembler.
16429 Output code for g-format floating point numbers instead of d-format.
16432 @node VxWorks Options
16433 @subsection VxWorks Options
16434 @cindex VxWorks Options
16436 The options in this section are defined for all VxWorks targets.
16437 Options specific to the target hardware are listed with the other
16438 options for that target.
16443 GCC can generate code for both VxWorks kernels and real time processes
16444 (RTPs). This option switches from the former to the latter. It also
16445 defines the preprocessor macro @code{__RTP__}.
16448 @opindex non-static
16449 Link an RTP executable against shared libraries rather than static
16450 libraries. The options @option{-static} and @option{-shared} can
16451 also be used for RTPs (@pxref{Link Options}); @option{-static}
16458 These options are passed down to the linker. They are defined for
16459 compatibility with Diab.
16462 @opindex Xbind-lazy
16463 Enable lazy binding of function calls. This option is equivalent to
16464 @option{-Wl,-z,now} and is defined for compatibility with Diab.
16468 Disable lazy binding of function calls. This option is the default and
16469 is defined for compatibility with Diab.
16472 @node x86-64 Options
16473 @subsection x86-64 Options
16474 @cindex x86-64 options
16476 These are listed under @xref{i386 and x86-64 Options}.
16478 @node i386 and x86-64 Windows Options
16479 @subsection i386 and x86-64 Windows Options
16480 @cindex i386 and x86-64 Windows Options
16482 These additional options are available for Windows targets:
16487 This option is available for Cygwin and MinGW targets. It
16488 specifies that a console application is to be generated, by
16489 instructing the linker to set the PE header subsystem type
16490 required for console applications.
16491 This is the default behavior for Cygwin and MinGW targets.
16495 This option is available for Cygwin targets. It specifies that
16496 the Cygwin internal interface is to be used for predefined
16497 preprocessor macros, C runtime libraries and related linker
16498 paths and options. For Cygwin targets this is the default behavior.
16499 This option is deprecated and will be removed in a future release.
16502 @opindex mno-cygwin
16503 This option is available for Cygwin targets. It specifies that
16504 the MinGW internal interface is to be used instead of Cygwin's, by
16505 setting MinGW-related predefined macros and linker paths and default
16507 This option is deprecated and will be removed in a future release.
16511 This option is available for Cygwin and MinGW targets. It
16512 specifies that a DLL - a dynamic link library - is to be
16513 generated, enabling the selection of the required runtime
16514 startup object and entry point.
16516 @item -mnop-fun-dllimport
16517 @opindex mnop-fun-dllimport
16518 This option is available for Cygwin and MinGW targets. It
16519 specifies that the dllimport attribute should be ignored.
16523 This option is available for MinGW targets. It specifies
16524 that MinGW-specific thread support is to be used.
16528 This option is available for mingw-w64 targets. It specifies
16529 that the UNICODE macro is getting pre-defined and that the
16530 unicode capable runtime startup code is choosen.
16534 This option is available for Cygwin and MinGW targets. It
16535 specifies that the typical Windows pre-defined macros are to
16536 be set in the pre-processor, but does not influence the choice
16537 of runtime library/startup code.
16541 This option is available for Cygwin and MinGW targets. It
16542 specifies that a GUI application is to be generated by
16543 instructing the linker to set the PE header subsystem type
16546 @item -mpe-aligned-commons
16547 @opindex mpe-aligned-commons
16548 This option is available for Cygwin and MinGW targets. It
16549 specifies that the GNU extension to the PE file format that
16550 permits the correct alignment of COMMON variables should be
16551 used when generating code. It will be enabled by default if
16552 GCC detects that the target assembler found during configuration
16553 supports the feature.
16556 See also under @ref{i386 and x86-64 Options} for standard options.
16558 @node Xstormy16 Options
16559 @subsection Xstormy16 Options
16560 @cindex Xstormy16 Options
16562 These options are defined for Xstormy16:
16567 Choose startup files and linker script suitable for the simulator.
16570 @node Xtensa Options
16571 @subsection Xtensa Options
16572 @cindex Xtensa Options
16574 These options are supported for Xtensa targets:
16578 @itemx -mno-const16
16580 @opindex mno-const16
16581 Enable or disable use of @code{CONST16} instructions for loading
16582 constant values. The @code{CONST16} instruction is currently not a
16583 standard option from Tensilica. When enabled, @code{CONST16}
16584 instructions are always used in place of the standard @code{L32R}
16585 instructions. The use of @code{CONST16} is enabled by default only if
16586 the @code{L32R} instruction is not available.
16589 @itemx -mno-fused-madd
16590 @opindex mfused-madd
16591 @opindex mno-fused-madd
16592 Enable or disable use of fused multiply/add and multiply/subtract
16593 instructions in the floating-point option. This has no effect if the
16594 floating-point option is not also enabled. Disabling fused multiply/add
16595 and multiply/subtract instructions forces the compiler to use separate
16596 instructions for the multiply and add/subtract operations. This may be
16597 desirable in some cases where strict IEEE 754-compliant results are
16598 required: the fused multiply add/subtract instructions do not round the
16599 intermediate result, thereby producing results with @emph{more} bits of
16600 precision than specified by the IEEE standard. Disabling fused multiply
16601 add/subtract instructions also ensures that the program output is not
16602 sensitive to the compiler's ability to combine multiply and add/subtract
16605 @item -mserialize-volatile
16606 @itemx -mno-serialize-volatile
16607 @opindex mserialize-volatile
16608 @opindex mno-serialize-volatile
16609 When this option is enabled, GCC inserts @code{MEMW} instructions before
16610 @code{volatile} memory references to guarantee sequential consistency.
16611 The default is @option{-mserialize-volatile}. Use
16612 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
16614 @item -mtext-section-literals
16615 @itemx -mno-text-section-literals
16616 @opindex mtext-section-literals
16617 @opindex mno-text-section-literals
16618 Control the treatment of literal pools. The default is
16619 @option{-mno-text-section-literals}, which places literals in a separate
16620 section in the output file. This allows the literal pool to be placed
16621 in a data RAM/ROM, and it also allows the linker to combine literal
16622 pools from separate object files to remove redundant literals and
16623 improve code size. With @option{-mtext-section-literals}, the literals
16624 are interspersed in the text section in order to keep them as close as
16625 possible to their references. This may be necessary for large assembly
16628 @item -mtarget-align
16629 @itemx -mno-target-align
16630 @opindex mtarget-align
16631 @opindex mno-target-align
16632 When this option is enabled, GCC instructs the assembler to
16633 automatically align instructions to reduce branch penalties at the
16634 expense of some code density. The assembler attempts to widen density
16635 instructions to align branch targets and the instructions following call
16636 instructions. If there are not enough preceding safe density
16637 instructions to align a target, no widening will be performed. The
16638 default is @option{-mtarget-align}. These options do not affect the
16639 treatment of auto-aligned instructions like @code{LOOP}, which the
16640 assembler will always align, either by widening density instructions or
16641 by inserting no-op instructions.
16644 @itemx -mno-longcalls
16645 @opindex mlongcalls
16646 @opindex mno-longcalls
16647 When this option is enabled, GCC instructs the assembler to translate
16648 direct calls to indirect calls unless it can determine that the target
16649 of a direct call is in the range allowed by the call instruction. This
16650 translation typically occurs for calls to functions in other source
16651 files. Specifically, the assembler translates a direct @code{CALL}
16652 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
16653 The default is @option{-mno-longcalls}. This option should be used in
16654 programs where the call target can potentially be out of range. This
16655 option is implemented in the assembler, not the compiler, so the
16656 assembly code generated by GCC will still show direct call
16657 instructions---look at the disassembled object code to see the actual
16658 instructions. Note that the assembler will use an indirect call for
16659 every cross-file call, not just those that really will be out of range.
16662 @node zSeries Options
16663 @subsection zSeries Options
16664 @cindex zSeries options
16666 These are listed under @xref{S/390 and zSeries Options}.
16668 @node Code Gen Options
16669 @section Options for Code Generation Conventions
16670 @cindex code generation conventions
16671 @cindex options, code generation
16672 @cindex run-time options
16674 These machine-independent options control the interface conventions
16675 used in code generation.
16677 Most of them have both positive and negative forms; the negative form
16678 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
16679 one of the forms is listed---the one which is not the default. You
16680 can figure out the other form by either removing @samp{no-} or adding
16684 @item -fbounds-check
16685 @opindex fbounds-check
16686 For front-ends that support it, generate additional code to check that
16687 indices used to access arrays are within the declared range. This is
16688 currently only supported by the Java and Fortran front-ends, where
16689 this option defaults to true and false respectively.
16693 This option generates traps for signed overflow on addition, subtraction,
16694 multiplication operations.
16698 This option instructs the compiler to assume that signed arithmetic
16699 overflow of addition, subtraction and multiplication wraps around
16700 using twos-complement representation. This flag enables some optimizations
16701 and disables others. This option is enabled by default for the Java
16702 front-end, as required by the Java language specification.
16705 @opindex fexceptions
16706 Enable exception handling. Generates extra code needed to propagate
16707 exceptions. For some targets, this implies GCC will generate frame
16708 unwind information for all functions, which can produce significant data
16709 size overhead, although it does not affect execution. If you do not
16710 specify this option, GCC will enable it by default for languages like
16711 C++ which normally require exception handling, and disable it for
16712 languages like C that do not normally require it. However, you may need
16713 to enable this option when compiling C code that needs to interoperate
16714 properly with exception handlers written in C++. You may also wish to
16715 disable this option if you are compiling older C++ programs that don't
16716 use exception handling.
16718 @item -fnon-call-exceptions
16719 @opindex fnon-call-exceptions
16720 Generate code that allows trapping instructions to throw exceptions.
16721 Note that this requires platform-specific runtime support that does
16722 not exist everywhere. Moreover, it only allows @emph{trapping}
16723 instructions to throw exceptions, i.e.@: memory references or floating
16724 point instructions. It does not allow exceptions to be thrown from
16725 arbitrary signal handlers such as @code{SIGALRM}.
16727 @item -funwind-tables
16728 @opindex funwind-tables
16729 Similar to @option{-fexceptions}, except that it will just generate any needed
16730 static data, but will not affect the generated code in any other way.
16731 You will normally not enable this option; instead, a language processor
16732 that needs this handling would enable it on your behalf.
16734 @item -fasynchronous-unwind-tables
16735 @opindex fasynchronous-unwind-tables
16736 Generate unwind table in dwarf2 format, if supported by target machine. The
16737 table is exact at each instruction boundary, so it can be used for stack
16738 unwinding from asynchronous events (such as debugger or garbage collector).
16740 @item -fpcc-struct-return
16741 @opindex fpcc-struct-return
16742 Return ``short'' @code{struct} and @code{union} values in memory like
16743 longer ones, rather than in registers. This convention is less
16744 efficient, but it has the advantage of allowing intercallability between
16745 GCC-compiled files and files compiled with other compilers, particularly
16746 the Portable C Compiler (pcc).
16748 The precise convention for returning structures in memory depends
16749 on the target configuration macros.
16751 Short structures and unions are those whose size and alignment match
16752 that of some integer type.
16754 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
16755 switch is not binary compatible with code compiled with the
16756 @option{-freg-struct-return} switch.
16757 Use it to conform to a non-default application binary interface.
16759 @item -freg-struct-return
16760 @opindex freg-struct-return
16761 Return @code{struct} and @code{union} values in registers when possible.
16762 This is more efficient for small structures than
16763 @option{-fpcc-struct-return}.
16765 If you specify neither @option{-fpcc-struct-return} nor
16766 @option{-freg-struct-return}, GCC defaults to whichever convention is
16767 standard for the target. If there is no standard convention, GCC
16768 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
16769 the principal compiler. In those cases, we can choose the standard, and
16770 we chose the more efficient register return alternative.
16772 @strong{Warning:} code compiled with the @option{-freg-struct-return}
16773 switch is not binary compatible with code compiled with the
16774 @option{-fpcc-struct-return} switch.
16775 Use it to conform to a non-default application binary interface.
16777 @item -fshort-enums
16778 @opindex fshort-enums
16779 Allocate to an @code{enum} type only as many bytes as it needs for the
16780 declared range of possible values. Specifically, the @code{enum} type
16781 will be equivalent to the smallest integer type which has enough room.
16783 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
16784 code that is not binary compatible with code generated without that switch.
16785 Use it to conform to a non-default application binary interface.
16787 @item -fshort-double
16788 @opindex fshort-double
16789 Use the same size for @code{double} as for @code{float}.
16791 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
16792 code that is not binary compatible with code generated without that switch.
16793 Use it to conform to a non-default application binary interface.
16795 @item -fshort-wchar
16796 @opindex fshort-wchar
16797 Override the underlying type for @samp{wchar_t} to be @samp{short
16798 unsigned int} instead of the default for the target. This option is
16799 useful for building programs to run under WINE@.
16801 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
16802 code that is not binary compatible with code generated without that switch.
16803 Use it to conform to a non-default application binary interface.
16806 @opindex fno-common
16807 In C code, controls the placement of uninitialized global variables.
16808 Unix C compilers have traditionally permitted multiple definitions of
16809 such variables in different compilation units by placing the variables
16811 This is the behavior specified by @option{-fcommon}, and is the default
16812 for GCC on most targets.
16813 On the other hand, this behavior is not required by ISO C, and on some
16814 targets may carry a speed or code size penalty on variable references.
16815 The @option{-fno-common} option specifies that the compiler should place
16816 uninitialized global variables in the data section of the object file,
16817 rather than generating them as common blocks.
16818 This has the effect that if the same variable is declared
16819 (without @code{extern}) in two different compilations,
16820 you will get a multiple-definition error when you link them.
16821 In this case, you must compile with @option{-fcommon} instead.
16822 Compiling with @option{-fno-common} is useful on targets for which
16823 it provides better performance, or if you wish to verify that the
16824 program will work on other systems which always treat uninitialized
16825 variable declarations this way.
16829 Ignore the @samp{#ident} directive.
16831 @item -finhibit-size-directive
16832 @opindex finhibit-size-directive
16833 Don't output a @code{.size} assembler directive, or anything else that
16834 would cause trouble if the function is split in the middle, and the
16835 two halves are placed at locations far apart in memory. This option is
16836 used when compiling @file{crtstuff.c}; you should not need to use it
16839 @item -fverbose-asm
16840 @opindex fverbose-asm
16841 Put extra commentary information in the generated assembly code to
16842 make it more readable. This option is generally only of use to those
16843 who actually need to read the generated assembly code (perhaps while
16844 debugging the compiler itself).
16846 @option{-fno-verbose-asm}, the default, causes the
16847 extra information to be omitted and is useful when comparing two assembler
16850 @item -frecord-gcc-switches
16851 @opindex frecord-gcc-switches
16852 This switch causes the command line that was used to invoke the
16853 compiler to be recorded into the object file that is being created.
16854 This switch is only implemented on some targets and the exact format
16855 of the recording is target and binary file format dependent, but it
16856 usually takes the form of a section containing ASCII text. This
16857 switch is related to the @option{-fverbose-asm} switch, but that
16858 switch only records information in the assembler output file as
16859 comments, so it never reaches the object file.
16863 @cindex global offset table
16865 Generate position-independent code (PIC) suitable for use in a shared
16866 library, if supported for the target machine. Such code accesses all
16867 constant addresses through a global offset table (GOT)@. The dynamic
16868 loader resolves the GOT entries when the program starts (the dynamic
16869 loader is not part of GCC; it is part of the operating system). If
16870 the GOT size for the linked executable exceeds a machine-specific
16871 maximum size, you get an error message from the linker indicating that
16872 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
16873 instead. (These maximums are 8k on the SPARC and 32k
16874 on the m68k and RS/6000. The 386 has no such limit.)
16876 Position-independent code requires special support, and therefore works
16877 only on certain machines. For the 386, GCC supports PIC for System V
16878 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
16879 position-independent.
16881 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
16886 If supported for the target machine, emit position-independent code,
16887 suitable for dynamic linking and avoiding any limit on the size of the
16888 global offset table. This option makes a difference on the m68k,
16889 PowerPC and SPARC@.
16891 Position-independent code requires special support, and therefore works
16892 only on certain machines.
16894 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
16901 These options are similar to @option{-fpic} and @option{-fPIC}, but
16902 generated position independent code can be only linked into executables.
16903 Usually these options are used when @option{-pie} GCC option will be
16904 used during linking.
16906 @option{-fpie} and @option{-fPIE} both define the macros
16907 @code{__pie__} and @code{__PIE__}. The macros have the value 1
16908 for @option{-fpie} and 2 for @option{-fPIE}.
16910 @item -fno-jump-tables
16911 @opindex fno-jump-tables
16912 Do not use jump tables for switch statements even where it would be
16913 more efficient than other code generation strategies. This option is
16914 of use in conjunction with @option{-fpic} or @option{-fPIC} for
16915 building code which forms part of a dynamic linker and cannot
16916 reference the address of a jump table. On some targets, jump tables
16917 do not require a GOT and this option is not needed.
16919 @item -ffixed-@var{reg}
16921 Treat the register named @var{reg} as a fixed register; generated code
16922 should never refer to it (except perhaps as a stack pointer, frame
16923 pointer or in some other fixed role).
16925 @var{reg} must be the name of a register. The register names accepted
16926 are machine-specific and are defined in the @code{REGISTER_NAMES}
16927 macro in the machine description macro file.
16929 This flag does not have a negative form, because it specifies a
16932 @item -fcall-used-@var{reg}
16933 @opindex fcall-used
16934 Treat the register named @var{reg} as an allocable register that is
16935 clobbered by function calls. It may be allocated for temporaries or
16936 variables that do not live across a call. Functions compiled this way
16937 will not save and restore the register @var{reg}.
16939 It is an error to used this flag with the frame pointer or stack pointer.
16940 Use of this flag for other registers that have fixed pervasive roles in
16941 the machine's execution model will produce disastrous results.
16943 This flag does not have a negative form, because it specifies a
16946 @item -fcall-saved-@var{reg}
16947 @opindex fcall-saved
16948 Treat the register named @var{reg} as an allocable register saved by
16949 functions. It may be allocated even for temporaries or variables that
16950 live across a call. Functions compiled this way will save and restore
16951 the register @var{reg} if they use it.
16953 It is an error to used this flag with the frame pointer or stack pointer.
16954 Use of this flag for other registers that have fixed pervasive roles in
16955 the machine's execution model will produce disastrous results.
16957 A different sort of disaster will result from the use of this flag for
16958 a register in which function values may be returned.
16960 This flag does not have a negative form, because it specifies a
16963 @item -fpack-struct[=@var{n}]
16964 @opindex fpack-struct
16965 Without a value specified, pack all structure members together without
16966 holes. When a value is specified (which must be a small power of two), pack
16967 structure members according to this value, representing the maximum
16968 alignment (that is, objects with default alignment requirements larger than
16969 this will be output potentially unaligned at the next fitting location.
16971 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
16972 code that is not binary compatible with code generated without that switch.
16973 Additionally, it makes the code suboptimal.
16974 Use it to conform to a non-default application binary interface.
16976 @item -finstrument-functions
16977 @opindex finstrument-functions
16978 Generate instrumentation calls for entry and exit to functions. Just
16979 after function entry and just before function exit, the following
16980 profiling functions will be called with the address of the current
16981 function and its call site. (On some platforms,
16982 @code{__builtin_return_address} does not work beyond the current
16983 function, so the call site information may not be available to the
16984 profiling functions otherwise.)
16987 void __cyg_profile_func_enter (void *this_fn,
16989 void __cyg_profile_func_exit (void *this_fn,
16993 The first argument is the address of the start of the current function,
16994 which may be looked up exactly in the symbol table.
16996 This instrumentation is also done for functions expanded inline in other
16997 functions. The profiling calls will indicate where, conceptually, the
16998 inline function is entered and exited. This means that addressable
16999 versions of such functions must be available. If all your uses of a
17000 function are expanded inline, this may mean an additional expansion of
17001 code size. If you use @samp{extern inline} in your C code, an
17002 addressable version of such functions must be provided. (This is
17003 normally the case anyways, but if you get lucky and the optimizer always
17004 expands the functions inline, you might have gotten away without
17005 providing static copies.)
17007 A function may be given the attribute @code{no_instrument_function}, in
17008 which case this instrumentation will not be done. This can be used, for
17009 example, for the profiling functions listed above, high-priority
17010 interrupt routines, and any functions from which the profiling functions
17011 cannot safely be called (perhaps signal handlers, if the profiling
17012 routines generate output or allocate memory).
17014 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
17015 @opindex finstrument-functions-exclude-file-list
17017 Set the list of functions that are excluded from instrumentation (see
17018 the description of @code{-finstrument-functions}). If the file that
17019 contains a function definition matches with one of @var{file}, then
17020 that function is not instrumented. The match is done on substrings:
17021 if the @var{file} parameter is a substring of the file name, it is
17022 considered to be a match.
17025 @code{-finstrument-functions-exclude-file-list=/bits/stl,include/sys}
17026 will exclude any inline function defined in files whose pathnames
17027 contain @code{/bits/stl} or @code{include/sys}.
17029 If, for some reason, you want to include letter @code{','} in one of
17030 @var{sym}, write @code{'\,'}. For example,
17031 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
17032 (note the single quote surrounding the option).
17034 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
17035 @opindex finstrument-functions-exclude-function-list
17037 This is similar to @code{-finstrument-functions-exclude-file-list},
17038 but this option sets the list of function names to be excluded from
17039 instrumentation. The function name to be matched is its user-visible
17040 name, such as @code{vector<int> blah(const vector<int> &)}, not the
17041 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
17042 match is done on substrings: if the @var{sym} parameter is a substring
17043 of the function name, it is considered to be a match. For C99 and C++
17044 extended identifiers, the function name must be given in UTF-8, not
17045 using universal character names.
17047 @item -fstack-check
17048 @opindex fstack-check
17049 Generate code to verify that you do not go beyond the boundary of the
17050 stack. You should specify this flag if you are running in an
17051 environment with multiple threads, but only rarely need to specify it in
17052 a single-threaded environment since stack overflow is automatically
17053 detected on nearly all systems if there is only one stack.
17055 Note that this switch does not actually cause checking to be done; the
17056 operating system or the language runtime must do that. The switch causes
17057 generation of code to ensure that they see the stack being extended.
17059 You can additionally specify a string parameter: @code{no} means no
17060 checking, @code{generic} means force the use of old-style checking,
17061 @code{specific} means use the best checking method and is equivalent
17062 to bare @option{-fstack-check}.
17064 Old-style checking is a generic mechanism that requires no specific
17065 target support in the compiler but comes with the following drawbacks:
17069 Modified allocation strategy for large objects: they will always be
17070 allocated dynamically if their size exceeds a fixed threshold.
17073 Fixed limit on the size of the static frame of functions: when it is
17074 topped by a particular function, stack checking is not reliable and
17075 a warning is issued by the compiler.
17078 Inefficiency: because of both the modified allocation strategy and the
17079 generic implementation, the performances of the code are hampered.
17082 Note that old-style stack checking is also the fallback method for
17083 @code{specific} if no target support has been added in the compiler.
17085 @item -fstack-limit-register=@var{reg}
17086 @itemx -fstack-limit-symbol=@var{sym}
17087 @itemx -fno-stack-limit
17088 @opindex fstack-limit-register
17089 @opindex fstack-limit-symbol
17090 @opindex fno-stack-limit
17091 Generate code to ensure that the stack does not grow beyond a certain value,
17092 either the value of a register or the address of a symbol. If the stack
17093 would grow beyond the value, a signal is raised. For most targets,
17094 the signal is raised before the stack overruns the boundary, so
17095 it is possible to catch the signal without taking special precautions.
17097 For instance, if the stack starts at absolute address @samp{0x80000000}
17098 and grows downwards, you can use the flags
17099 @option{-fstack-limit-symbol=__stack_limit} and
17100 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
17101 of 128KB@. Note that this may only work with the GNU linker.
17103 @cindex aliasing of parameters
17104 @cindex parameters, aliased
17105 @item -fargument-alias
17106 @itemx -fargument-noalias
17107 @itemx -fargument-noalias-global
17108 @itemx -fargument-noalias-anything
17109 @opindex fargument-alias
17110 @opindex fargument-noalias
17111 @opindex fargument-noalias-global
17112 @opindex fargument-noalias-anything
17113 Specify the possible relationships among parameters and between
17114 parameters and global data.
17116 @option{-fargument-alias} specifies that arguments (parameters) may
17117 alias each other and may alias global storage.@*
17118 @option{-fargument-noalias} specifies that arguments do not alias
17119 each other, but may alias global storage.@*
17120 @option{-fargument-noalias-global} specifies that arguments do not
17121 alias each other and do not alias global storage.
17122 @option{-fargument-noalias-anything} specifies that arguments do not
17123 alias any other storage.
17125 Each language will automatically use whatever option is required by
17126 the language standard. You should not need to use these options yourself.
17128 @item -fleading-underscore
17129 @opindex fleading-underscore
17130 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
17131 change the way C symbols are represented in the object file. One use
17132 is to help link with legacy assembly code.
17134 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
17135 generate code that is not binary compatible with code generated without that
17136 switch. Use it to conform to a non-default application binary interface.
17137 Not all targets provide complete support for this switch.
17139 @item -ftls-model=@var{model}
17140 @opindex ftls-model
17141 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
17142 The @var{model} argument should be one of @code{global-dynamic},
17143 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
17145 The default without @option{-fpic} is @code{initial-exec}; with
17146 @option{-fpic} the default is @code{global-dynamic}.
17148 @item -fvisibility=@var{default|internal|hidden|protected}
17149 @opindex fvisibility
17150 Set the default ELF image symbol visibility to the specified option---all
17151 symbols will be marked with this unless overridden within the code.
17152 Using this feature can very substantially improve linking and
17153 load times of shared object libraries, produce more optimized
17154 code, provide near-perfect API export and prevent symbol clashes.
17155 It is @strong{strongly} recommended that you use this in any shared objects
17158 Despite the nomenclature, @code{default} always means public ie;
17159 available to be linked against from outside the shared object.
17160 @code{protected} and @code{internal} are pretty useless in real-world
17161 usage so the only other commonly used option will be @code{hidden}.
17162 The default if @option{-fvisibility} isn't specified is
17163 @code{default}, i.e., make every
17164 symbol public---this causes the same behavior as previous versions of
17167 A good explanation of the benefits offered by ensuring ELF
17168 symbols have the correct visibility is given by ``How To Write
17169 Shared Libraries'' by Ulrich Drepper (which can be found at
17170 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
17171 solution made possible by this option to marking things hidden when
17172 the default is public is to make the default hidden and mark things
17173 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
17174 and @code{__attribute__ ((visibility("default")))} instead of
17175 @code{__declspec(dllexport)} you get almost identical semantics with
17176 identical syntax. This is a great boon to those working with
17177 cross-platform projects.
17179 For those adding visibility support to existing code, you may find
17180 @samp{#pragma GCC visibility} of use. This works by you enclosing
17181 the declarations you wish to set visibility for with (for example)
17182 @samp{#pragma GCC visibility push(hidden)} and
17183 @samp{#pragma GCC visibility pop}.
17184 Bear in mind that symbol visibility should be viewed @strong{as
17185 part of the API interface contract} and thus all new code should
17186 always specify visibility when it is not the default ie; declarations
17187 only for use within the local DSO should @strong{always} be marked explicitly
17188 as hidden as so to avoid PLT indirection overheads---making this
17189 abundantly clear also aids readability and self-documentation of the code.
17190 Note that due to ISO C++ specification requirements, operator new and
17191 operator delete must always be of default visibility.
17193 Be aware that headers from outside your project, in particular system
17194 headers and headers from any other library you use, may not be
17195 expecting to be compiled with visibility other than the default. You
17196 may need to explicitly say @samp{#pragma GCC visibility push(default)}
17197 before including any such headers.
17199 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
17200 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
17201 no modifications. However, this means that calls to @samp{extern}
17202 functions with no explicit visibility will use the PLT, so it is more
17203 effective to use @samp{__attribute ((visibility))} and/or
17204 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
17205 declarations should be treated as hidden.
17207 Note that @samp{-fvisibility} does affect C++ vague linkage
17208 entities. This means that, for instance, an exception class that will
17209 be thrown between DSOs must be explicitly marked with default
17210 visibility so that the @samp{type_info} nodes will be unified between
17213 An overview of these techniques, their benefits and how to use them
17214 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
17220 @node Environment Variables
17221 @section Environment Variables Affecting GCC
17222 @cindex environment variables
17224 @c man begin ENVIRONMENT
17225 This section describes several environment variables that affect how GCC
17226 operates. Some of them work by specifying directories or prefixes to use
17227 when searching for various kinds of files. Some are used to specify other
17228 aspects of the compilation environment.
17230 Note that you can also specify places to search using options such as
17231 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
17232 take precedence over places specified using environment variables, which
17233 in turn take precedence over those specified by the configuration of GCC@.
17234 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
17235 GNU Compiler Collection (GCC) Internals}.
17240 @c @itemx LC_COLLATE
17242 @c @itemx LC_MONETARY
17243 @c @itemx LC_NUMERIC
17248 @c @findex LC_COLLATE
17249 @findex LC_MESSAGES
17250 @c @findex LC_MONETARY
17251 @c @findex LC_NUMERIC
17255 These environment variables control the way that GCC uses
17256 localization information that allow GCC to work with different
17257 national conventions. GCC inspects the locale categories
17258 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
17259 so. These locale categories can be set to any value supported by your
17260 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
17261 Kingdom encoded in UTF-8.
17263 The @env{LC_CTYPE} environment variable specifies character
17264 classification. GCC uses it to determine the character boundaries in
17265 a string; this is needed for some multibyte encodings that contain quote
17266 and escape characters that would otherwise be interpreted as a string
17269 The @env{LC_MESSAGES} environment variable specifies the language to
17270 use in diagnostic messages.
17272 If the @env{LC_ALL} environment variable is set, it overrides the value
17273 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
17274 and @env{LC_MESSAGES} default to the value of the @env{LANG}
17275 environment variable. If none of these variables are set, GCC
17276 defaults to traditional C English behavior.
17280 If @env{TMPDIR} is set, it specifies the directory to use for temporary
17281 files. GCC uses temporary files to hold the output of one stage of
17282 compilation which is to be used as input to the next stage: for example,
17283 the output of the preprocessor, which is the input to the compiler
17286 @item GCC_EXEC_PREFIX
17287 @findex GCC_EXEC_PREFIX
17288 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
17289 names of the subprograms executed by the compiler. No slash is added
17290 when this prefix is combined with the name of a subprogram, but you can
17291 specify a prefix that ends with a slash if you wish.
17293 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
17294 an appropriate prefix to use based on the pathname it was invoked with.
17296 If GCC cannot find the subprogram using the specified prefix, it
17297 tries looking in the usual places for the subprogram.
17299 The default value of @env{GCC_EXEC_PREFIX} is
17300 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
17301 the installed compiler. In many cases @var{prefix} is the value
17302 of @code{prefix} when you ran the @file{configure} script.
17304 Other prefixes specified with @option{-B} take precedence over this prefix.
17306 This prefix is also used for finding files such as @file{crt0.o} that are
17309 In addition, the prefix is used in an unusual way in finding the
17310 directories to search for header files. For each of the standard
17311 directories whose name normally begins with @samp{/usr/local/lib/gcc}
17312 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
17313 replacing that beginning with the specified prefix to produce an
17314 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
17315 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
17316 These alternate directories are searched first; the standard directories
17317 come next. If a standard directory begins with the configured
17318 @var{prefix} then the value of @var{prefix} is replaced by
17319 @env{GCC_EXEC_PREFIX} when looking for header files.
17321 @item COMPILER_PATH
17322 @findex COMPILER_PATH
17323 The value of @env{COMPILER_PATH} is a colon-separated list of
17324 directories, much like @env{PATH}. GCC tries the directories thus
17325 specified when searching for subprograms, if it can't find the
17326 subprograms using @env{GCC_EXEC_PREFIX}.
17329 @findex LIBRARY_PATH
17330 The value of @env{LIBRARY_PATH} is a colon-separated list of
17331 directories, much like @env{PATH}. When configured as a native compiler,
17332 GCC tries the directories thus specified when searching for special
17333 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
17334 using GCC also uses these directories when searching for ordinary
17335 libraries for the @option{-l} option (but directories specified with
17336 @option{-L} come first).
17340 @cindex locale definition
17341 This variable is used to pass locale information to the compiler. One way in
17342 which this information is used is to determine the character set to be used
17343 when character literals, string literals and comments are parsed in C and C++.
17344 When the compiler is configured to allow multibyte characters,
17345 the following values for @env{LANG} are recognized:
17349 Recognize JIS characters.
17351 Recognize SJIS characters.
17353 Recognize EUCJP characters.
17356 If @env{LANG} is not defined, or if it has some other value, then the
17357 compiler will use mblen and mbtowc as defined by the default locale to
17358 recognize and translate multibyte characters.
17362 Some additional environments variables affect the behavior of the
17365 @include cppenv.texi
17369 @node Precompiled Headers
17370 @section Using Precompiled Headers
17371 @cindex precompiled headers
17372 @cindex speed of compilation
17374 Often large projects have many header files that are included in every
17375 source file. The time the compiler takes to process these header files
17376 over and over again can account for nearly all of the time required to
17377 build the project. To make builds faster, GCC allows users to
17378 `precompile' a header file; then, if builds can use the precompiled
17379 header file they will be much faster.
17381 To create a precompiled header file, simply compile it as you would any
17382 other file, if necessary using the @option{-x} option to make the driver
17383 treat it as a C or C++ header file. You will probably want to use a
17384 tool like @command{make} to keep the precompiled header up-to-date when
17385 the headers it contains change.
17387 A precompiled header file will be searched for when @code{#include} is
17388 seen in the compilation. As it searches for the included file
17389 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
17390 compiler looks for a precompiled header in each directory just before it
17391 looks for the include file in that directory. The name searched for is
17392 the name specified in the @code{#include} with @samp{.gch} appended. If
17393 the precompiled header file can't be used, it is ignored.
17395 For instance, if you have @code{#include "all.h"}, and you have
17396 @file{all.h.gch} in the same directory as @file{all.h}, then the
17397 precompiled header file will be used if possible, and the original
17398 header will be used otherwise.
17400 Alternatively, you might decide to put the precompiled header file in a
17401 directory and use @option{-I} to ensure that directory is searched
17402 before (or instead of) the directory containing the original header.
17403 Then, if you want to check that the precompiled header file is always
17404 used, you can put a file of the same name as the original header in this
17405 directory containing an @code{#error} command.
17407 This also works with @option{-include}. So yet another way to use
17408 precompiled headers, good for projects not designed with precompiled
17409 header files in mind, is to simply take most of the header files used by
17410 a project, include them from another header file, precompile that header
17411 file, and @option{-include} the precompiled header. If the header files
17412 have guards against multiple inclusion, they will be skipped because
17413 they've already been included (in the precompiled header).
17415 If you need to precompile the same header file for different
17416 languages, targets, or compiler options, you can instead make a
17417 @emph{directory} named like @file{all.h.gch}, and put each precompiled
17418 header in the directory, perhaps using @option{-o}. It doesn't matter
17419 what you call the files in the directory, every precompiled header in
17420 the directory will be considered. The first precompiled header
17421 encountered in the directory that is valid for this compilation will
17422 be used; they're searched in no particular order.
17424 There are many other possibilities, limited only by your imagination,
17425 good sense, and the constraints of your build system.
17427 A precompiled header file can be used only when these conditions apply:
17431 Only one precompiled header can be used in a particular compilation.
17434 A precompiled header can't be used once the first C token is seen. You
17435 can have preprocessor directives before a precompiled header; you can
17436 even include a precompiled header from inside another header, so long as
17437 there are no C tokens before the @code{#include}.
17440 The precompiled header file must be produced for the same language as
17441 the current compilation. You can't use a C precompiled header for a C++
17445 The precompiled header file must have been produced by the same compiler
17446 binary as the current compilation is using.
17449 Any macros defined before the precompiled header is included must
17450 either be defined in the same way as when the precompiled header was
17451 generated, or must not affect the precompiled header, which usually
17452 means that they don't appear in the precompiled header at all.
17454 The @option{-D} option is one way to define a macro before a
17455 precompiled header is included; using a @code{#define} can also do it.
17456 There are also some options that define macros implicitly, like
17457 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
17460 @item If debugging information is output when using the precompiled
17461 header, using @option{-g} or similar, the same kind of debugging information
17462 must have been output when building the precompiled header. However,
17463 a precompiled header built using @option{-g} can be used in a compilation
17464 when no debugging information is being output.
17466 @item The same @option{-m} options must generally be used when building
17467 and using the precompiled header. @xref{Submodel Options},
17468 for any cases where this rule is relaxed.
17470 @item Each of the following options must be the same when building and using
17471 the precompiled header:
17473 @gccoptlist{-fexceptions}
17476 Some other command-line options starting with @option{-f},
17477 @option{-p}, or @option{-O} must be defined in the same way as when
17478 the precompiled header was generated. At present, it's not clear
17479 which options are safe to change and which are not; the safest choice
17480 is to use exactly the same options when generating and using the
17481 precompiled header. The following are known to be safe:
17483 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
17484 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
17485 -fsched-verbose=<number> -fschedule-insns -fvisibility= @gol
17490 For all of these except the last, the compiler will automatically
17491 ignore the precompiled header if the conditions aren't met. If you
17492 find an option combination that doesn't work and doesn't cause the
17493 precompiled header to be ignored, please consider filing a bug report,
17496 If you do use differing options when generating and using the
17497 precompiled header, the actual behavior will be a mixture of the
17498 behavior for the options. For instance, if you use @option{-g} to
17499 generate the precompiled header but not when using it, you may or may
17500 not get debugging information for routines in the precompiled header.