1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
2 @c 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009
3 @c Free Software Foundation, Inc.
4 @c This is part of the GCC manual.
5 @c For copying conditions, see the file gcc.texi.
12 @c man begin COPYRIGHT
13 Copyright @copyright{} 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
14 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009
15 Free Software Foundation, Inc.
17 Permission is granted to copy, distribute and/or modify this document
18 under the terms of the GNU Free Documentation License, Version 1.2 or
19 any later version published by the Free Software Foundation; with the
20 Invariant Sections being ``GNU General Public License'' and ``Funding
21 Free Software'', the Front-Cover texts being (a) (see below), and with
22 the Back-Cover Texts being (b) (see below). A copy of the license is
23 included in the gfdl(7) man page.
25 (a) The FSF's Front-Cover Text is:
29 (b) The FSF's Back-Cover Text is:
31 You have freedom to copy and modify this GNU Manual, like GNU
32 software. Copies published by the Free Software Foundation raise
33 funds for GNU development.
35 @c Set file name and title for the man page.
37 @settitle GNU project C and C++ compiler
39 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
40 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
41 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
42 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
43 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
44 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
45 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
47 Only the most useful options are listed here; see below for the
48 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
51 gpl(7), gfdl(7), fsf-funding(7),
52 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
53 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
54 @file{ld}, @file{binutils} and @file{gdb}.
57 For instructions on reporting bugs, see
61 See the Info entry for @command{gcc}, or
62 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
63 for contributors to GCC@.
68 @chapter GCC Command Options
69 @cindex GCC command options
70 @cindex command options
71 @cindex options, GCC command
73 @c man begin DESCRIPTION
74 When you invoke GCC, it normally does preprocessing, compilation,
75 assembly and linking. The ``overall options'' allow you to stop this
76 process at an intermediate stage. For example, the @option{-c} option
77 says not to run the linker. Then the output consists of object files
78 output by the assembler.
80 Other options are passed on to one stage of processing. Some options
81 control the preprocessor and others the compiler itself. Yet other
82 options control the assembler and linker; most of these are not
83 documented here, since you rarely need to use any of them.
85 @cindex C compilation options
86 Most of the command line options that you can use with GCC are useful
87 for C programs; when an option is only useful with another language
88 (usually C++), the explanation says so explicitly. If the description
89 for a particular option does not mention a source language, you can use
90 that option with all supported languages.
92 @cindex C++ compilation options
93 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
94 options for compiling C++ programs.
96 @cindex grouping options
97 @cindex options, grouping
98 The @command{gcc} program accepts options and file names as operands. Many
99 options have multi-letter names; therefore multiple single-letter options
100 may @emph{not} be grouped: @option{-dv} is very different from @w{@samp{-d
103 @cindex order of options
104 @cindex options, order
105 You can mix options and other arguments. For the most part, the order
106 you use doesn't matter. Order does matter when you use several
107 options of the same kind; for example, if you specify @option{-L} more
108 than once, the directories are searched in the order specified. Also,
109 the placement of the @option{-l} option is significant.
111 Many options have long names starting with @samp{-f} or with
112 @samp{-W}---for example,
113 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
114 these have both positive and negative forms; the negative form of
115 @option{-ffoo} would be @option{-fno-foo}. This manual documents
116 only one of these two forms, whichever one is not the default.
120 @xref{Option Index}, for an index to GCC's options.
123 * Option Summary:: Brief list of all options, without explanations.
124 * Overall Options:: Controlling the kind of output:
125 an executable, object files, assembler files,
126 or preprocessed source.
127 * Invoking G++:: Compiling C++ programs.
128 * C Dialect Options:: Controlling the variant of C language compiled.
129 * C++ Dialect Options:: Variations on C++.
130 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
132 * Language Independent Options:: Controlling how diagnostics should be
134 * Warning Options:: How picky should the compiler be?
135 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
136 * Optimize Options:: How much optimization?
137 * Preprocessor Options:: Controlling header files and macro definitions.
138 Also, getting dependency information for Make.
139 * Assembler Options:: Passing options to the assembler.
140 * Link Options:: Specifying libraries and so on.
141 * Directory Options:: Where to find header files and libraries.
142 Where to find the compiler executable files.
143 * Spec Files:: How to pass switches to sub-processes.
144 * Target Options:: Running a cross-compiler, or an old version of GCC.
145 * Submodel Options:: Specifying minor hardware or convention variations,
146 such as 68010 vs 68020.
147 * Code Gen Options:: Specifying conventions for function calls, data layout
149 * Environment Variables:: Env vars that affect GCC.
150 * Precompiled Headers:: Compiling a header once, and using it many times.
156 @section Option Summary
158 Here is a summary of all the options, grouped by type. Explanations are
159 in the following sections.
162 @item Overall Options
163 @xref{Overall Options,,Options Controlling the Kind of Output}.
164 @gccoptlist{-c -S -E -o @var{file} -combine -no-canonical-prefixes @gol
165 -pipe -pass-exit-codes @gol
166 -x @var{language} -v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help @gol
167 --version -wrapper@@@var{file} -fplugin=@var{file} -fplugin-arg-@var{name}=@var{arg}}
169 @item C Language Options
170 @xref{C Dialect Options,,Options Controlling C Dialect}.
171 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
172 -aux-info @var{filename} @gol
173 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
174 -fhosted -ffreestanding -fopenmp -fms-extensions @gol
175 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
176 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
177 -fsigned-bitfields -fsigned-char @gol
178 -funsigned-bitfields -funsigned-char}
180 @item C++ Language Options
181 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
182 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
183 -fconserve-space -ffriend-injection @gol
184 -fno-elide-constructors @gol
185 -fno-enforce-eh-specs @gol
186 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
187 -fno-implicit-templates @gol
188 -fno-implicit-inline-templates @gol
189 -fno-implement-inlines -fms-extensions @gol
190 -fno-nonansi-builtins -fno-operator-names @gol
191 -fno-optional-diags -fpermissive @gol
192 -fno-pretty-templates @gol
193 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
194 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
195 -fno-default-inline -fvisibility-inlines-hidden @gol
196 -fvisibility-ms-compat @gol
197 -Wabi -Wctor-dtor-privacy @gol
198 -Wnon-virtual-dtor -Wreorder @gol
199 -Weffc++ -Wstrict-null-sentinel @gol
200 -Wno-non-template-friend -Wold-style-cast @gol
201 -Woverloaded-virtual -Wno-pmf-conversions @gol
204 @item Objective-C and Objective-C++ Language Options
205 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
206 Objective-C and Objective-C++ Dialects}.
207 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
208 -fgnu-runtime -fnext-runtime @gol
209 -fno-nil-receivers @gol
210 -fobjc-call-cxx-cdtors @gol
211 -fobjc-direct-dispatch @gol
212 -fobjc-exceptions @gol
214 -freplace-objc-classes @gol
217 -Wassign-intercept @gol
218 -Wno-protocol -Wselector @gol
219 -Wstrict-selector-match @gol
220 -Wundeclared-selector}
222 @item Language Independent Options
223 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
224 @gccoptlist{-fmessage-length=@var{n} @gol
225 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
226 -fdiagnostics-show-option}
228 @item Warning Options
229 @xref{Warning Options,,Options to Request or Suppress Warnings}.
230 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
231 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
232 -Wno-attributes -Wno-builtin-macro-redefined @gol
233 -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual @gol
234 -Wchar-subscripts -Wclobbered -Wcomment @gol
235 -Wconversion -Wcoverage-mismatch -Wno-deprecated @gol
236 -Wno-deprecated-declarations -Wdisabled-optimization @gol
237 -Wno-div-by-zero -Wempty-body -Wenum-compare -Wno-endif-labels @gol
238 -Werror -Werror=* @gol
239 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
240 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
241 -Wformat-security -Wformat-y2k @gol
242 -Wframe-larger-than=@var{len} -Wjump-misses-init -Wignored-qualifiers @gol
243 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
244 -Winit-self -Winline @gol
245 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
246 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
247 -Wlogical-op -Wlong-long @gol
248 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
249 -Wmissing-format-attribute -Wmissing-include-dirs @gol
250 -Wmissing-noreturn -Wno-mudflap @gol
251 -Wno-multichar -Wnonnull -Wno-overflow @gol
252 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
253 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
254 -Wpointer-arith -Wno-pointer-to-int-cast @gol
255 -Wredundant-decls @gol
256 -Wreturn-type -Wsequence-point -Wshadow @gol
257 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
258 -Wstrict-aliasing -Wstrict-aliasing=n @gol
259 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
260 -Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand @gol
261 -Wsystem-headers -Wtrigraphs -Wtype-limits -Wundef -Wuninitialized @gol
262 -Wunknown-pragmas -Wno-pragmas -Wunreachable-code @gol
263 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
264 -Wunused-label -Wunused-parameter -Wno-unused-result -Wunused-value -Wunused-variable @gol
265 -Wvariadic-macros -Wvla @gol
266 -Wvolatile-register-var -Wwrite-strings}
268 @item C and Objective-C-only Warning Options
269 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
270 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
271 -Wold-style-declaration -Wold-style-definition @gol
272 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
273 -Wdeclaration-after-statement -Wpointer-sign}
275 @item Debugging Options
276 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
277 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
278 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
279 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
280 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
281 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
282 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
283 -fdump-statistics @gol
285 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
286 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
287 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
289 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
290 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
291 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
292 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
293 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
294 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
295 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
296 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
297 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
298 -fdump-tree-nrv -fdump-tree-vect @gol
299 -fdump-tree-sink @gol
300 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
301 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
302 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
303 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
304 -ftree-vectorizer-verbose=@var{n} @gol
305 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
306 -fdump-final-insns=@var{file} @gol
307 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
308 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
309 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
310 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
311 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
312 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
313 -ftest-coverage -ftime-report -fvar-tracking @gol
314 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
315 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
316 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
317 -fdebug-prefix-map=@var{old}=@var{new} @gol
318 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
319 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
320 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
321 -print-multi-directory -print-multi-lib @gol
322 -print-prog-name=@var{program} -print-search-dirs -Q @gol
323 -print-sysroot -print-sysroot-headers-suffix @gol
324 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
326 @item Optimization Options
327 @xref{Optimize Options,,Options that Control Optimization}.
329 -falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
330 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
331 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
332 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
333 -fcheck-data-deps -fconserve-stack -fcprop-registers -fcrossjumping @gol
334 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules -fcx-limited-range @gol
335 -fdata-sections -fdce -fdce @gol
336 -fdelayed-branch -fdelete-null-pointer-checks -fdse -fdse @gol
337 -fearly-inlining -fexpensive-optimizations -ffast-math @gol
338 -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
339 -fforward-propagate -ffunction-sections @gol
340 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm @gol
341 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
342 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
343 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg -fipa-pta @gol
344 -fipa-pure-const -fipa-reference -fipa-struct-reorg @gol
345 -fipa-type-escape -fira-algorithm=@var{algorithm} @gol
346 -fira-region=@var{region} -fira-coalesce -fno-ira-share-save-slots @gol
347 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
348 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
349 -floop-block -floop-interchange -floop-strip-mine @gol
350 -fmerge-all-constants -fmerge-constants -fmodulo-sched @gol
351 -fmodulo-sched-allow-regmoves -fmove-loop-invariants -fmudflap @gol
352 -fmudflapir -fmudflapth -fno-branch-count-reg -fno-default-inline @gol
353 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
354 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
355 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
356 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
357 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
358 -fpeel-loops -fpredictive-commoning -fprefetch-loop-arrays @gol
359 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
360 -fprofile-generate=@var{path} @gol
361 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
362 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
363 -freorder-blocks-and-partition -freorder-functions @gol
364 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
365 -frounding-math -fsched2-use-superblocks @gol
366 -fsched2-use-traces -fsched-spec-load -fsched-spec-load-dangerous @gol
367 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
368 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
369 -fsched-spec-insn-heuristic -fsched-reg-pressure-heuristic @gol
370 -fsched-rank-heuristic -fsched-last-insn-heuristic @gol
371 -fsched-dep-count-heuristic @gol
372 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
373 -fselective-scheduling -fselective-scheduling2 @gol
374 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
375 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
376 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
377 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer @gol
378 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
379 -ftree-copyrename -ftree-dce @gol
380 -ftree-dominator-opts -ftree-dse -ftree-forwprop -ftree-fre -ftree-loop-im @gol
381 -ftree-phiprop -ftree-loop-distribution @gol
382 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
383 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-pta -ftree-reassoc @gol
384 -ftree-sink -ftree-sra -ftree-switch-conversion @gol
385 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
386 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
387 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
388 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
390 --param @var{name}=@var{value}
391 -O -O0 -O1 -O2 -O3 -Os}
393 @item Preprocessor Options
394 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
395 @gccoptlist{-A@var{question}=@var{answer} @gol
396 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
397 -C -dD -dI -dM -dN @gol
398 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
399 -idirafter @var{dir} @gol
400 -include @var{file} -imacros @var{file} @gol
401 -iprefix @var{file} -iwithprefix @var{dir} @gol
402 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
403 -imultilib @var{dir} -isysroot @var{dir} @gol
404 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
405 -P -fworking-directory -remap @gol
406 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
407 -Xpreprocessor @var{option}}
409 @item Assembler Option
410 @xref{Assembler Options,,Passing Options to the Assembler}.
411 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
414 @xref{Link Options,,Options for Linking}.
415 @gccoptlist{@var{object-file-name} -l@var{library} @gol
416 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
417 -s -static -static-libgcc -static-libstdc++ -shared @gol
418 -shared-libgcc -symbolic @gol
419 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
422 @item Directory Options
423 @xref{Directory Options,,Options for Directory Search}.
424 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir}
425 -specs=@var{file} -I- --sysroot=@var{dir}}
428 @c I wrote this xref this way to avoid overfull hbox. -- rms
429 @xref{Target Options}.
430 @gccoptlist{-V @var{version} -b @var{machine}}
432 @item Machine Dependent Options
433 @xref{Submodel Options,,Hardware Models and Configurations}.
434 @c This list is ordered alphanumerically by subsection name.
435 @c Try and put the significant identifier (CPU or system) first,
436 @c so users have a clue at guessing where the ones they want will be.
439 @gccoptlist{-EB -EL @gol
440 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
441 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
444 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
445 -mabi=@var{name} @gol
446 -mapcs-stack-check -mno-apcs-stack-check @gol
447 -mapcs-float -mno-apcs-float @gol
448 -mapcs-reentrant -mno-apcs-reentrant @gol
449 -msched-prolog -mno-sched-prolog @gol
450 -mlittle-endian -mbig-endian -mwords-little-endian @gol
451 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
452 -mfp16-format=@var{name}
453 -mthumb-interwork -mno-thumb-interwork @gol
454 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
455 -mstructure-size-boundary=@var{n} @gol
456 -mabort-on-noreturn @gol
457 -mlong-calls -mno-long-calls @gol
458 -msingle-pic-base -mno-single-pic-base @gol
459 -mpic-register=@var{reg} @gol
460 -mnop-fun-dllimport @gol
461 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
462 -mpoke-function-name @gol
464 -mtpcs-frame -mtpcs-leaf-frame @gol
465 -mcaller-super-interworking -mcallee-super-interworking @gol
467 -mword-relocations @gol
468 -mfix-cortex-m3-ldrd}
471 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
472 -mcall-prologues -mtiny-stack -mint8}
474 @emph{Blackfin Options}
475 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
476 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
477 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
478 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
479 -mno-id-shared-library -mshared-library-id=@var{n} @gol
480 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
481 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
482 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
486 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
487 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
488 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
489 -mstack-align -mdata-align -mconst-align @gol
490 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
491 -melf -maout -melinux -mlinux -sim -sim2 @gol
492 -mmul-bug-workaround -mno-mul-bug-workaround}
495 @gccoptlist{-mmac -mpush-args}
497 @emph{Darwin Options}
498 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
499 -arch_only -bind_at_load -bundle -bundle_loader @gol
500 -client_name -compatibility_version -current_version @gol
502 -dependency-file -dylib_file -dylinker_install_name @gol
503 -dynamic -dynamiclib -exported_symbols_list @gol
504 -filelist -flat_namespace -force_cpusubtype_ALL @gol
505 -force_flat_namespace -headerpad_max_install_names @gol
507 -image_base -init -install_name -keep_private_externs @gol
508 -multi_module -multiply_defined -multiply_defined_unused @gol
509 -noall_load -no_dead_strip_inits_and_terms @gol
510 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
511 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
512 -private_bundle -read_only_relocs -sectalign @gol
513 -sectobjectsymbols -whyload -seg1addr @gol
514 -sectcreate -sectobjectsymbols -sectorder @gol
515 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
516 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
517 -segprot -segs_read_only_addr -segs_read_write_addr @gol
518 -single_module -static -sub_library -sub_umbrella @gol
519 -twolevel_namespace -umbrella -undefined @gol
520 -unexported_symbols_list -weak_reference_mismatches @gol
521 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
522 -mkernel -mone-byte-bool}
524 @emph{DEC Alpha Options}
525 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
526 -mieee -mieee-with-inexact -mieee-conformant @gol
527 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
528 -mtrap-precision=@var{mode} -mbuild-constants @gol
529 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
530 -mbwx -mmax -mfix -mcix @gol
531 -mfloat-vax -mfloat-ieee @gol
532 -mexplicit-relocs -msmall-data -mlarge-data @gol
533 -msmall-text -mlarge-text @gol
534 -mmemory-latency=@var{time}}
536 @emph{DEC Alpha/VMS Options}
537 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix}}
540 @gccoptlist{-msmall-model -mno-lsim}
543 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
544 -mhard-float -msoft-float @gol
545 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
546 -mdouble -mno-double @gol
547 -mmedia -mno-media -mmuladd -mno-muladd @gol
548 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
549 -mlinked-fp -mlong-calls -malign-labels @gol
550 -mlibrary-pic -macc-4 -macc-8 @gol
551 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
552 -moptimize-membar -mno-optimize-membar @gol
553 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
554 -mvliw-branch -mno-vliw-branch @gol
555 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
556 -mno-nested-cond-exec -mtomcat-stats @gol
560 @emph{GNU/Linux Options}
561 @gccoptlist{-muclibc}
563 @emph{H8/300 Options}
564 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
567 @gccoptlist{-march=@var{architecture-type} @gol
568 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
569 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
570 -mfixed-range=@var{register-range} @gol
571 -mjump-in-delay -mlinker-opt -mlong-calls @gol
572 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
573 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
574 -mno-jump-in-delay -mno-long-load-store @gol
575 -mno-portable-runtime -mno-soft-float @gol
576 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
577 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
578 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
579 -munix=@var{unix-std} -nolibdld -static -threads}
581 @emph{i386 and x86-64 Options}
582 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
583 -mfpmath=@var{unit} @gol
584 -masm=@var{dialect} -mno-fancy-math-387 @gol
585 -mno-fp-ret-in-387 -msoft-float @gol
586 -mno-wide-multiply -mrtd -malign-double @gol
587 -mpreferred-stack-boundary=@var{num}
588 -mincoming-stack-boundary=@var{num}
589 -mcld -mcx16 -msahf -mmovbe -mcrc32 -mrecip @gol
590 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
592 -msse4a -m3dnow -mpopcnt -mabm -msse5 @gol
593 -mthreads -mno-align-stringops -minline-all-stringops @gol
594 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
595 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
596 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
597 -mveclibabi=@var{type} -mpc32 -mpc64 -mpc80 -mstackrealign @gol
598 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
599 -mcmodel=@var{code-model} -mabi=@var{name} @gol
600 -m32 -m64 -mlarge-data-threshold=@var{num} @gol
601 -mfused-madd -mno-fused-madd -msse2avx}
604 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
605 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
606 -mconstant-gp -mauto-pic -mfused-madd @gol
607 -minline-float-divide-min-latency @gol
608 -minline-float-divide-max-throughput @gol
609 -mno-inline-float-divide @gol
610 -minline-int-divide-min-latency @gol
611 -minline-int-divide-max-throughput @gol
612 -mno-inline-int-divide @gol
613 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
614 -mno-inline-sqrt @gol
615 -mdwarf2-asm -mearly-stop-bits @gol
616 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
617 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
618 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
619 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
620 -msched-spec-ldc -msched-spec-control-ldc @gol
621 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
622 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
623 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
624 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
626 @emph{M32R/D Options}
627 @gccoptlist{-m32r2 -m32rx -m32r @gol
629 -malign-loops -mno-align-loops @gol
630 -missue-rate=@var{number} @gol
631 -mbranch-cost=@var{number} @gol
632 -mmodel=@var{code-size-model-type} @gol
633 -msdata=@var{sdata-type} @gol
634 -mno-flush-func -mflush-func=@var{name} @gol
635 -mno-flush-trap -mflush-trap=@var{number} @gol
639 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
641 @emph{M680x0 Options}
642 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
643 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
644 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
645 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
646 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
647 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
648 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
649 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
652 @emph{M68hc1x Options}
653 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
654 -mauto-incdec -minmax -mlong-calls -mshort @gol
655 -msoft-reg-count=@var{count}}
658 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
659 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
660 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
661 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
662 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
665 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
666 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
667 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
668 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
672 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
673 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
674 -mips64 -mips64r2 @gol
675 -mips16 -mno-mips16 -mflip-mips16 @gol
676 -minterlink-mips16 -mno-interlink-mips16 @gol
677 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
678 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
679 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
680 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
681 -mfpu=@var{fpu-type} @gol
682 -msmartmips -mno-smartmips @gol
683 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
684 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
685 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
686 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
687 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
688 -membedded-data -mno-embedded-data @gol
689 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
690 -mcode-readable=@var{setting} @gol
691 -msplit-addresses -mno-split-addresses @gol
692 -mexplicit-relocs -mno-explicit-relocs @gol
693 -mcheck-zero-division -mno-check-zero-division @gol
694 -mdivide-traps -mdivide-breaks @gol
695 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
696 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
697 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
698 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
699 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
700 -mflush-func=@var{func} -mno-flush-func @gol
701 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
702 -mfp-exceptions -mno-fp-exceptions @gol
703 -mvr4130-align -mno-vr4130-align -msynci -mno-synci}
706 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
707 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
708 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
709 -mno-base-addresses -msingle-exit -mno-single-exit}
711 @emph{MN10300 Options}
712 @gccoptlist{-mmult-bug -mno-mult-bug @gol
713 -mam33 -mno-am33 @gol
714 -mam33-2 -mno-am33-2 @gol
715 -mreturn-pointer-on-d0 @gol
718 @emph{PDP-11 Options}
719 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
720 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
721 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
722 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
723 -mbranch-expensive -mbranch-cheap @gol
724 -msplit -mno-split -munix-asm -mdec-asm}
726 @emph{picoChip Options}
727 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N}
728 -msymbol-as-address -mno-inefficient-warnings}
730 @emph{PowerPC Options}
731 See RS/6000 and PowerPC Options.
733 @emph{RS/6000 and PowerPC Options}
734 @gccoptlist{-mcpu=@var{cpu-type} @gol
735 -mtune=@var{cpu-type} @gol
736 -mpower -mno-power -mpower2 -mno-power2 @gol
737 -mpowerpc -mpowerpc64 -mno-powerpc @gol
738 -maltivec -mno-altivec @gol
739 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
740 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
741 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mfprnd -mno-fprnd @gol
742 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
743 -mnew-mnemonics -mold-mnemonics @gol
744 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
745 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
746 -malign-power -malign-natural @gol
747 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
748 -msingle-float -mdouble-float -msimple-fpu @gol
749 -mstring -mno-string -mupdate -mno-update @gol
750 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
751 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
752 -mstrict-align -mno-strict-align -mrelocatable @gol
753 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
754 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
755 -mdynamic-no-pic -maltivec -mswdiv @gol
756 -mprioritize-restricted-insns=@var{priority} @gol
757 -msched-costly-dep=@var{dependence_type} @gol
758 -minsert-sched-nops=@var{scheme} @gol
759 -mcall-sysv -mcall-netbsd @gol
760 -maix-struct-return -msvr4-struct-return @gol
761 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
762 -misel -mno-isel @gol
763 -misel=yes -misel=no @gol
765 -mspe=yes -mspe=no @gol
767 -mgen-cell-microcode -mwarn-cell-microcode @gol
768 -mvrsave -mno-vrsave @gol
769 -mmulhw -mno-mulhw @gol
770 -mdlmzb -mno-dlmzb @gol
771 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
772 -mprototype -mno-prototype @gol
773 -msim -mmvme -mads -myellowknife -memb -msdata @gol
774 -msdata=@var{opt} -mvxworks -G @var{num} -pthread}
776 @emph{S/390 and zSeries Options}
777 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
778 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
779 -mlong-double-64 -mlong-double-128 @gol
780 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
781 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
782 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
783 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
784 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
787 @gccoptlist{-meb -mel @gol
791 -mscore5 -mscore5u -mscore7 -mscore7d}
794 @gccoptlist{-m1 -m2 -m2e @gol
795 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
797 -m4-nofpu -m4-single-only -m4-single -m4 @gol
798 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
799 -m5-64media -m5-64media-nofpu @gol
800 -m5-32media -m5-32media-nofpu @gol
801 -m5-compact -m5-compact-nofpu @gol
802 -mb -ml -mdalign -mrelax @gol
803 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
804 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
805 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
806 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
807 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
811 @gccoptlist{-mcpu=@var{cpu-type} @gol
812 -mtune=@var{cpu-type} @gol
813 -mcmodel=@var{code-model} @gol
814 -m32 -m64 -mapp-regs -mno-app-regs @gol
815 -mfaster-structs -mno-faster-structs @gol
816 -mfpu -mno-fpu -mhard-float -msoft-float @gol
817 -mhard-quad-float -msoft-quad-float @gol
818 -mimpure-text -mno-impure-text -mlittle-endian @gol
819 -mstack-bias -mno-stack-bias @gol
820 -munaligned-doubles -mno-unaligned-doubles @gol
821 -mv8plus -mno-v8plus -mvis -mno-vis
822 -threads -pthreads -pthread}
825 @gccoptlist{-mwarn-reloc -merror-reloc @gol
826 -msafe-dma -munsafe-dma @gol
828 -msmall-mem -mlarge-mem -mstdmain @gol
829 -mfixed-range=@var{register-range}}
831 @emph{System V Options}
832 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
835 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
836 -mprolog-function -mno-prolog-function -mspace @gol
837 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
838 -mapp-regs -mno-app-regs @gol
839 -mdisable-callt -mno-disable-callt @gol
845 @gccoptlist{-mg -mgnu -munix}
847 @emph{VxWorks Options}
848 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
849 -Xbind-lazy -Xbind-now}
851 @emph{x86-64 Options}
852 See i386 and x86-64 Options.
854 @emph{i386 and x86-64 Windows Options}
855 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll
856 -mnop-fun-dllimport -mthread -municode -mwin32 -mwindows}
858 @emph{Xstormy16 Options}
861 @emph{Xtensa Options}
862 @gccoptlist{-mconst16 -mno-const16 @gol
863 -mfused-madd -mno-fused-madd @gol
864 -mserialize-volatile -mno-serialize-volatile @gol
865 -mtext-section-literals -mno-text-section-literals @gol
866 -mtarget-align -mno-target-align @gol
867 -mlongcalls -mno-longcalls}
869 @emph{zSeries Options}
870 See S/390 and zSeries Options.
872 @item Code Generation Options
873 @xref{Code Gen Options,,Options for Code Generation Conventions}.
874 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
875 -ffixed-@var{reg} -fexceptions @gol
876 -fnon-call-exceptions -funwind-tables @gol
877 -fasynchronous-unwind-tables @gol
878 -finhibit-size-directive -finstrument-functions @gol
879 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
880 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
881 -fno-common -fno-ident @gol
882 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
883 -fno-jump-tables @gol
884 -frecord-gcc-switches @gol
885 -freg-struct-return -fshort-enums @gol
886 -fshort-double -fshort-wchar @gol
887 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
888 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
889 -fno-stack-limit -fargument-alias -fargument-noalias @gol
890 -fargument-noalias-global -fargument-noalias-anything @gol
891 -fleading-underscore -ftls-model=@var{model} @gol
892 -ftrapv -fwrapv -fbounds-check @gol
897 * Overall Options:: Controlling the kind of output:
898 an executable, object files, assembler files,
899 or preprocessed source.
900 * C Dialect Options:: Controlling the variant of C language compiled.
901 * C++ Dialect Options:: Variations on C++.
902 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
904 * Language Independent Options:: Controlling how diagnostics should be
906 * Warning Options:: How picky should the compiler be?
907 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
908 * Optimize Options:: How much optimization?
909 * Preprocessor Options:: Controlling header files and macro definitions.
910 Also, getting dependency information for Make.
911 * Assembler Options:: Passing options to the assembler.
912 * Link Options:: Specifying libraries and so on.
913 * Directory Options:: Where to find header files and libraries.
914 Where to find the compiler executable files.
915 * Spec Files:: How to pass switches to sub-processes.
916 * Target Options:: Running a cross-compiler, or an old version of GCC.
919 @node Overall Options
920 @section Options Controlling the Kind of Output
922 Compilation can involve up to four stages: preprocessing, compilation
923 proper, assembly and linking, always in that order. GCC is capable of
924 preprocessing and compiling several files either into several
925 assembler input files, or into one assembler input file; then each
926 assembler input file produces an object file, and linking combines all
927 the object files (those newly compiled, and those specified as input)
928 into an executable file.
930 @cindex file name suffix
931 For any given input file, the file name suffix determines what kind of
936 C source code which must be preprocessed.
939 C source code which should not be preprocessed.
942 C++ source code which should not be preprocessed.
945 Objective-C source code. Note that you must link with the @file{libobjc}
946 library to make an Objective-C program work.
949 Objective-C source code which should not be preprocessed.
953 Objective-C++ source code. Note that you must link with the @file{libobjc}
954 library to make an Objective-C++ program work. Note that @samp{.M} refers
955 to a literal capital M@.
958 Objective-C++ source code which should not be preprocessed.
961 C, C++, Objective-C or Objective-C++ header file to be turned into a
966 @itemx @var{file}.cxx
967 @itemx @var{file}.cpp
968 @itemx @var{file}.CPP
969 @itemx @var{file}.c++
971 C++ source code which must be preprocessed. Note that in @samp{.cxx},
972 the last two letters must both be literally @samp{x}. Likewise,
973 @samp{.C} refers to a literal capital C@.
977 Objective-C++ source code which must be preprocessed.
980 Objective-C++ source code which should not be preprocessed.
985 @itemx @var{file}.hxx
986 @itemx @var{file}.hpp
987 @itemx @var{file}.HPP
988 @itemx @var{file}.h++
989 @itemx @var{file}.tcc
990 C++ header file to be turned into a precompiled header.
993 @itemx @var{file}.for
994 @itemx @var{file}.ftn
995 Fixed form Fortran source code which should not be preprocessed.
998 @itemx @var{file}.FOR
999 @itemx @var{file}.fpp
1000 @itemx @var{file}.FPP
1001 @itemx @var{file}.FTN
1002 Fixed form Fortran source code which must be preprocessed (with the traditional
1005 @item @var{file}.f90
1006 @itemx @var{file}.f95
1007 @itemx @var{file}.f03
1008 @itemx @var{file}.f08
1009 Free form Fortran source code which should not be preprocessed.
1011 @item @var{file}.F90
1012 @itemx @var{file}.F95
1013 @itemx @var{file}.F03
1014 @itemx @var{file}.F08
1015 Free form Fortran source code which must be preprocessed (with the
1016 traditional preprocessor).
1018 @c FIXME: Descriptions of Java file types.
1024 @item @var{file}.ads
1025 Ada source code file which contains a library unit declaration (a
1026 declaration of a package, subprogram, or generic, or a generic
1027 instantiation), or a library unit renaming declaration (a package,
1028 generic, or subprogram renaming declaration). Such files are also
1031 @item @var{file}.adb
1032 Ada source code file containing a library unit body (a subprogram or
1033 package body). Such files are also called @dfn{bodies}.
1035 @c GCC also knows about some suffixes for languages not yet included:
1046 @itemx @var{file}.sx
1047 Assembler code which must be preprocessed.
1050 An object file to be fed straight into linking.
1051 Any file name with no recognized suffix is treated this way.
1055 You can specify the input language explicitly with the @option{-x} option:
1058 @item -x @var{language}
1059 Specify explicitly the @var{language} for the following input files
1060 (rather than letting the compiler choose a default based on the file
1061 name suffix). This option applies to all following input files until
1062 the next @option{-x} option. Possible values for @var{language} are:
1064 c c-header c-cpp-output
1065 c++ c++-header c++-cpp-output
1066 objective-c objective-c-header objective-c-cpp-output
1067 objective-c++ objective-c++-header objective-c++-cpp-output
1068 assembler assembler-with-cpp
1070 f77 f77-cpp-input f95 f95-cpp-input
1075 Turn off any specification of a language, so that subsequent files are
1076 handled according to their file name suffixes (as they are if @option{-x}
1077 has not been used at all).
1079 @item -pass-exit-codes
1080 @opindex pass-exit-codes
1081 Normally the @command{gcc} program will exit with the code of 1 if any
1082 phase of the compiler returns a non-success return code. If you specify
1083 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1084 numerically highest error produced by any phase that returned an error
1085 indication. The C, C++, and Fortran frontends return 4, if an internal
1086 compiler error is encountered.
1089 If you only want some of the stages of compilation, you can use
1090 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1091 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1092 @command{gcc} is to stop. Note that some combinations (for example,
1093 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1098 Compile or assemble the source files, but do not link. The linking
1099 stage simply is not done. The ultimate output is in the form of an
1100 object file for each source file.
1102 By default, the object file name for a source file is made by replacing
1103 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1105 Unrecognized input files, not requiring compilation or assembly, are
1110 Stop after the stage of compilation proper; do not assemble. The output
1111 is in the form of an assembler code file for each non-assembler input
1114 By default, the assembler file name for a source file is made by
1115 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1117 Input files that don't require compilation are ignored.
1121 Stop after the preprocessing stage; do not run the compiler proper. The
1122 output is in the form of preprocessed source code, which is sent to the
1125 Input files which don't require preprocessing are ignored.
1127 @cindex output file option
1130 Place output in file @var{file}. This applies regardless to whatever
1131 sort of output is being produced, whether it be an executable file,
1132 an object file, an assembler file or preprocessed C code.
1134 If @option{-o} is not specified, the default is to put an executable
1135 file in @file{a.out}, the object file for
1136 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1137 assembler file in @file{@var{source}.s}, a precompiled header file in
1138 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1143 Print (on standard error output) the commands executed to run the stages
1144 of compilation. Also print the version number of the compiler driver
1145 program and of the preprocessor and the compiler proper.
1149 Like @option{-v} except the commands are not executed and all command
1150 arguments are quoted. This is useful for shell scripts to capture the
1151 driver-generated command lines.
1155 Use pipes rather than temporary files for communication between the
1156 various stages of compilation. This fails to work on some systems where
1157 the assembler is unable to read from a pipe; but the GNU assembler has
1162 If you are compiling multiple source files, this option tells the driver
1163 to pass all the source files to the compiler at once (for those
1164 languages for which the compiler can handle this). This will allow
1165 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1166 language for which this is supported is C@. If you pass source files for
1167 multiple languages to the driver, using this option, the driver will invoke
1168 the compiler(s) that support IMA once each, passing each compiler all the
1169 source files appropriate for it. For those languages that do not support
1170 IMA this option will be ignored, and the compiler will be invoked once for
1171 each source file in that language. If you use this option in conjunction
1172 with @option{-save-temps}, the compiler will generate multiple
1174 (one for each source file), but only one (combined) @file{.o} or
1179 Print (on the standard output) a description of the command line options
1180 understood by @command{gcc}. If the @option{-v} option is also specified
1181 then @option{--help} will also be passed on to the various processes
1182 invoked by @command{gcc}, so that they can display the command line options
1183 they accept. If the @option{-Wextra} option has also been specified
1184 (prior to the @option{--help} option), then command line options which
1185 have no documentation associated with them will also be displayed.
1188 @opindex target-help
1189 Print (on the standard output) a description of target-specific command
1190 line options for each tool. For some targets extra target-specific
1191 information may also be printed.
1193 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1194 Print (on the standard output) a description of the command line
1195 options understood by the compiler that fit into all specified classes
1196 and qualifiers. These are the supported classes:
1199 @item @samp{optimizers}
1200 This will display all of the optimization options supported by the
1203 @item @samp{warnings}
1204 This will display all of the options controlling warning messages
1205 produced by the compiler.
1208 This will display target-specific options. Unlike the
1209 @option{--target-help} option however, target-specific options of the
1210 linker and assembler will not be displayed. This is because those
1211 tools do not currently support the extended @option{--help=} syntax.
1214 This will display the values recognized by the @option{--param}
1217 @item @var{language}
1218 This will display the options supported for @var{language}, where
1219 @var{language} is the name of one of the languages supported in this
1223 This will display the options that are common to all languages.
1226 These are the supported qualifiers:
1229 @item @samp{undocumented}
1230 Display only those options which are undocumented.
1233 Display options which take an argument that appears after an equal
1234 sign in the same continuous piece of text, such as:
1235 @samp{--help=target}.
1237 @item @samp{separate}
1238 Display options which take an argument that appears as a separate word
1239 following the original option, such as: @samp{-o output-file}.
1242 Thus for example to display all the undocumented target-specific
1243 switches supported by the compiler the following can be used:
1246 --help=target,undocumented
1249 The sense of a qualifier can be inverted by prefixing it with the
1250 @samp{^} character, so for example to display all binary warning
1251 options (i.e., ones that are either on or off and that do not take an
1252 argument), which have a description the following can be used:
1255 --help=warnings,^joined,^undocumented
1258 The argument to @option{--help=} should not consist solely of inverted
1261 Combining several classes is possible, although this usually
1262 restricts the output by so much that there is nothing to display. One
1263 case where it does work however is when one of the classes is
1264 @var{target}. So for example to display all the target-specific
1265 optimization options the following can be used:
1268 --help=target,optimizers
1271 The @option{--help=} option can be repeated on the command line. Each
1272 successive use will display its requested class of options, skipping
1273 those that have already been displayed.
1275 If the @option{-Q} option appears on the command line before the
1276 @option{--help=} option, then the descriptive text displayed by
1277 @option{--help=} is changed. Instead of describing the displayed
1278 options, an indication is given as to whether the option is enabled,
1279 disabled or set to a specific value (assuming that the compiler
1280 knows this at the point where the @option{--help=} option is used).
1282 Here is a truncated example from the ARM port of @command{gcc}:
1285 % gcc -Q -mabi=2 --help=target -c
1286 The following options are target specific:
1288 -mabort-on-noreturn [disabled]
1292 The output is sensitive to the effects of previous command line
1293 options, so for example it is possible to find out which optimizations
1294 are enabled at @option{-O2} by using:
1297 -Q -O2 --help=optimizers
1300 Alternatively you can discover which binary optimizations are enabled
1301 by @option{-O3} by using:
1304 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1305 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1306 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1309 @item -no-canonical-prefixes
1310 @opindex no-canonical-prefixes
1311 Do not expand any symbolic links, resolve references to @samp{/../}
1312 or @samp{/./}, or make the path absolute when generating a relative
1317 Display the version number and copyrights of the invoked GCC@.
1321 Invoke all subcommands under a wrapper program. It takes a single
1322 comma separated list as an argument, which will be used to invoke
1326 gcc -c t.c -wrapper gdb,--args
1329 This will invoke all subprograms of gcc under "gdb --args",
1330 thus cc1 invocation will be "gdb --args cc1 ...".
1332 @item -fplugin=@var{name}.so
1333 Load the plugin code in file @var{name}.so, assumed to be a
1334 shared object to be dlopen'd by the compiler. The base name of
1335 the shared object file is used to identify the plugin for the
1336 purposes of argument parsing (See
1337 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1338 Each plugin should define the callback functions specified in the
1341 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1342 Define an argument called @var{key} with a value of @var{value}
1343 for the plugin called @var{name}.
1345 @include @value{srcdir}/../libiberty/at-file.texi
1349 @section Compiling C++ Programs
1351 @cindex suffixes for C++ source
1352 @cindex C++ source file suffixes
1353 C++ source files conventionally use one of the suffixes @samp{.C},
1354 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1355 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1356 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1357 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1358 files with these names and compiles them as C++ programs even if you
1359 call the compiler the same way as for compiling C programs (usually
1360 with the name @command{gcc}).
1364 However, the use of @command{gcc} does not add the C++ library.
1365 @command{g++} is a program that calls GCC and treats @samp{.c},
1366 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1367 files unless @option{-x} is used, and automatically specifies linking
1368 against the C++ library. This program is also useful when
1369 precompiling a C header file with a @samp{.h} extension for use in C++
1370 compilations. On many systems, @command{g++} is also installed with
1371 the name @command{c++}.
1373 @cindex invoking @command{g++}
1374 When you compile C++ programs, you may specify many of the same
1375 command-line options that you use for compiling programs in any
1376 language; or command-line options meaningful for C and related
1377 languages; or options that are meaningful only for C++ programs.
1378 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1379 explanations of options for languages related to C@.
1380 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1381 explanations of options that are meaningful only for C++ programs.
1383 @node C Dialect Options
1384 @section Options Controlling C Dialect
1385 @cindex dialect options
1386 @cindex language dialect options
1387 @cindex options, dialect
1389 The following options control the dialect of C (or languages derived
1390 from C, such as C++, Objective-C and Objective-C++) that the compiler
1394 @cindex ANSI support
1398 In C mode, this is equivalent to @samp{-std=c89}. In C++ mode, it is
1399 equivalent to @samp{-std=c++98}.
1401 This turns off certain features of GCC that are incompatible with ISO
1402 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1403 such as the @code{asm} and @code{typeof} keywords, and
1404 predefined macros such as @code{unix} and @code{vax} that identify the
1405 type of system you are using. It also enables the undesirable and
1406 rarely used ISO trigraph feature. For the C compiler,
1407 it disables recognition of C++ style @samp{//} comments as well as
1408 the @code{inline} keyword.
1410 The alternate keywords @code{__asm__}, @code{__extension__},
1411 @code{__inline__} and @code{__typeof__} continue to work despite
1412 @option{-ansi}. You would not want to use them in an ISO C program, of
1413 course, but it is useful to put them in header files that might be included
1414 in compilations done with @option{-ansi}. Alternate predefined macros
1415 such as @code{__unix__} and @code{__vax__} are also available, with or
1416 without @option{-ansi}.
1418 The @option{-ansi} option does not cause non-ISO programs to be
1419 rejected gratuitously. For that, @option{-pedantic} is required in
1420 addition to @option{-ansi}. @xref{Warning Options}.
1422 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1423 option is used. Some header files may notice this macro and refrain
1424 from declaring certain functions or defining certain macros that the
1425 ISO standard doesn't call for; this is to avoid interfering with any
1426 programs that might use these names for other things.
1428 Functions that would normally be built in but do not have semantics
1429 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1430 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1431 built-in functions provided by GCC}, for details of the functions
1436 Determine the language standard. @xref{Standards,,Language Standards
1437 Supported by GCC}, for details of these standard versions. This option
1438 is currently only supported when compiling C or C++.
1440 The compiler can accept several base standards, such as @samp{c89} or
1441 @samp{c++98}, and GNU dialects of those standards, such as
1442 @samp{gnu89} or @samp{gnu++98}. By specifying a base standard, the
1443 compiler will accept all programs following that standard and those
1444 using GNU extensions that do not contradict it. For example,
1445 @samp{-std=c89} turns off certain features of GCC that are
1446 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1447 keywords, but not other GNU extensions that do not have a meaning in
1448 ISO C90, such as omitting the middle term of a @code{?:}
1449 expression. On the other hand, by specifying a GNU dialect of a
1450 standard, all features the compiler support are enabled, even when
1451 those features change the meaning of the base standard and some
1452 strict-conforming programs may be rejected. The particular standard
1453 is used by @option{-pedantic} to identify which features are GNU
1454 extensions given that version of the standard. For example
1455 @samp{-std=gnu89 -pedantic} would warn about C++ style @samp{//}
1456 comments, while @samp{-std=gnu99 -pedantic} would not.
1458 A value for this option must be provided; possible values are
1463 Support all ISO C90 programs (certain GNU extensions that conflict
1464 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1466 @item iso9899:199409
1467 ISO C90 as modified in amendment 1.
1473 ISO C99. Note that this standard is not yet fully supported; see
1474 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1475 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1478 GNU dialect of ISO C90 (including some C99 features). This
1479 is the default for C code.
1483 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1484 this will become the default. The name @samp{gnu9x} is deprecated.
1487 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1491 GNU dialect of @option{-std=c++98}. This is the default for
1495 The working draft of the upcoming ISO C++0x standard. This option
1496 enables experimental features that are likely to be included in
1497 C++0x. The working draft is constantly changing, and any feature that is
1498 enabled by this flag may be removed from future versions of GCC if it is
1499 not part of the C++0x standard.
1502 GNU dialect of @option{-std=c++0x}. This option enables
1503 experimental features that may be removed in future versions of GCC.
1506 @item -fgnu89-inline
1507 @opindex fgnu89-inline
1508 The option @option{-fgnu89-inline} tells GCC to use the traditional
1509 GNU semantics for @code{inline} functions when in C99 mode.
1510 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1511 is accepted and ignored by GCC versions 4.1.3 up to but not including
1512 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1513 C99 mode. Using this option is roughly equivalent to adding the
1514 @code{gnu_inline} function attribute to all inline functions
1515 (@pxref{Function Attributes}).
1517 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1518 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1519 specifies the default behavior). This option was first supported in
1520 GCC 4.3. This option is not supported in C89 or gnu89 mode.
1522 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1523 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1524 in effect for @code{inline} functions. @xref{Common Predefined
1525 Macros,,,cpp,The C Preprocessor}.
1527 @item -aux-info @var{filename}
1529 Output to the given filename prototyped declarations for all functions
1530 declared and/or defined in a translation unit, including those in header
1531 files. This option is silently ignored in any language other than C@.
1533 Besides declarations, the file indicates, in comments, the origin of
1534 each declaration (source file and line), whether the declaration was
1535 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1536 @samp{O} for old, respectively, in the first character after the line
1537 number and the colon), and whether it came from a declaration or a
1538 definition (@samp{C} or @samp{F}, respectively, in the following
1539 character). In the case of function definitions, a K&R-style list of
1540 arguments followed by their declarations is also provided, inside
1541 comments, after the declaration.
1545 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1546 keyword, so that code can use these words as identifiers. You can use
1547 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1548 instead. @option{-ansi} implies @option{-fno-asm}.
1550 In C++, this switch only affects the @code{typeof} keyword, since
1551 @code{asm} and @code{inline} are standard keywords. You may want to
1552 use the @option{-fno-gnu-keywords} flag instead, which has the same
1553 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1554 switch only affects the @code{asm} and @code{typeof} keywords, since
1555 @code{inline} is a standard keyword in ISO C99.
1558 @itemx -fno-builtin-@var{function}
1559 @opindex fno-builtin
1560 @cindex built-in functions
1561 Don't recognize built-in functions that do not begin with
1562 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1563 functions provided by GCC}, for details of the functions affected,
1564 including those which are not built-in functions when @option{-ansi} or
1565 @option{-std} options for strict ISO C conformance are used because they
1566 do not have an ISO standard meaning.
1568 GCC normally generates special code to handle certain built-in functions
1569 more efficiently; for instance, calls to @code{alloca} may become single
1570 instructions that adjust the stack directly, and calls to @code{memcpy}
1571 may become inline copy loops. The resulting code is often both smaller
1572 and faster, but since the function calls no longer appear as such, you
1573 cannot set a breakpoint on those calls, nor can you change the behavior
1574 of the functions by linking with a different library. In addition,
1575 when a function is recognized as a built-in function, GCC may use
1576 information about that function to warn about problems with calls to
1577 that function, or to generate more efficient code, even if the
1578 resulting code still contains calls to that function. For example,
1579 warnings are given with @option{-Wformat} for bad calls to
1580 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1581 known not to modify global memory.
1583 With the @option{-fno-builtin-@var{function}} option
1584 only the built-in function @var{function} is
1585 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1586 function is named that is not built-in in this version of GCC, this
1587 option is ignored. There is no corresponding
1588 @option{-fbuiltin-@var{function}} option; if you wish to enable
1589 built-in functions selectively when using @option{-fno-builtin} or
1590 @option{-ffreestanding}, you may define macros such as:
1593 #define abs(n) __builtin_abs ((n))
1594 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1599 @cindex hosted environment
1601 Assert that compilation takes place in a hosted environment. This implies
1602 @option{-fbuiltin}. A hosted environment is one in which the
1603 entire standard library is available, and in which @code{main} has a return
1604 type of @code{int}. Examples are nearly everything except a kernel.
1605 This is equivalent to @option{-fno-freestanding}.
1607 @item -ffreestanding
1608 @opindex ffreestanding
1609 @cindex hosted environment
1611 Assert that compilation takes place in a freestanding environment. This
1612 implies @option{-fno-builtin}. A freestanding environment
1613 is one in which the standard library may not exist, and program startup may
1614 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1615 This is equivalent to @option{-fno-hosted}.
1617 @xref{Standards,,Language Standards Supported by GCC}, for details of
1618 freestanding and hosted environments.
1622 @cindex openmp parallel
1623 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1624 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1625 compiler generates parallel code according to the OpenMP Application
1626 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1627 implies @option{-pthread}, and thus is only supported on targets that
1628 have support for @option{-pthread}.
1630 @item -fms-extensions
1631 @opindex fms-extensions
1632 Accept some non-standard constructs used in Microsoft header files.
1634 Some cases of unnamed fields in structures and unions are only
1635 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1636 fields within structs/unions}, for details.
1640 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1641 options for strict ISO C conformance) implies @option{-trigraphs}.
1643 @item -no-integrated-cpp
1644 @opindex no-integrated-cpp
1645 Performs a compilation in two passes: preprocessing and compiling. This
1646 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1647 @option{-B} option. The user supplied compilation step can then add in
1648 an additional preprocessing step after normal preprocessing but before
1649 compiling. The default is to use the integrated cpp (internal cpp)
1651 The semantics of this option will change if "cc1", "cc1plus", and
1652 "cc1obj" are merged.
1654 @cindex traditional C language
1655 @cindex C language, traditional
1657 @itemx -traditional-cpp
1658 @opindex traditional-cpp
1659 @opindex traditional
1660 Formerly, these options caused GCC to attempt to emulate a pre-standard
1661 C compiler. They are now only supported with the @option{-E} switch.
1662 The preprocessor continues to support a pre-standard mode. See the GNU
1663 CPP manual for details.
1665 @item -fcond-mismatch
1666 @opindex fcond-mismatch
1667 Allow conditional expressions with mismatched types in the second and
1668 third arguments. The value of such an expression is void. This option
1669 is not supported for C++.
1671 @item -flax-vector-conversions
1672 @opindex flax-vector-conversions
1673 Allow implicit conversions between vectors with differing numbers of
1674 elements and/or incompatible element types. This option should not be
1677 @item -funsigned-char
1678 @opindex funsigned-char
1679 Let the type @code{char} be unsigned, like @code{unsigned char}.
1681 Each kind of machine has a default for what @code{char} should
1682 be. It is either like @code{unsigned char} by default or like
1683 @code{signed char} by default.
1685 Ideally, a portable program should always use @code{signed char} or
1686 @code{unsigned char} when it depends on the signedness of an object.
1687 But many programs have been written to use plain @code{char} and
1688 expect it to be signed, or expect it to be unsigned, depending on the
1689 machines they were written for. This option, and its inverse, let you
1690 make such a program work with the opposite default.
1692 The type @code{char} is always a distinct type from each of
1693 @code{signed char} or @code{unsigned char}, even though its behavior
1694 is always just like one of those two.
1697 @opindex fsigned-char
1698 Let the type @code{char} be signed, like @code{signed char}.
1700 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1701 the negative form of @option{-funsigned-char}. Likewise, the option
1702 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1704 @item -fsigned-bitfields
1705 @itemx -funsigned-bitfields
1706 @itemx -fno-signed-bitfields
1707 @itemx -fno-unsigned-bitfields
1708 @opindex fsigned-bitfields
1709 @opindex funsigned-bitfields
1710 @opindex fno-signed-bitfields
1711 @opindex fno-unsigned-bitfields
1712 These options control whether a bit-field is signed or unsigned, when the
1713 declaration does not use either @code{signed} or @code{unsigned}. By
1714 default, such a bit-field is signed, because this is consistent: the
1715 basic integer types such as @code{int} are signed types.
1718 @node C++ Dialect Options
1719 @section Options Controlling C++ Dialect
1721 @cindex compiler options, C++
1722 @cindex C++ options, command line
1723 @cindex options, C++
1724 This section describes the command-line options that are only meaningful
1725 for C++ programs; but you can also use most of the GNU compiler options
1726 regardless of what language your program is in. For example, you
1727 might compile a file @code{firstClass.C} like this:
1730 g++ -g -frepo -O -c firstClass.C
1734 In this example, only @option{-frepo} is an option meant
1735 only for C++ programs; you can use the other options with any
1736 language supported by GCC@.
1738 Here is a list of options that are @emph{only} for compiling C++ programs:
1742 @item -fabi-version=@var{n}
1743 @opindex fabi-version
1744 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1745 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1746 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1747 the version that conforms most closely to the C++ ABI specification.
1748 Therefore, the ABI obtained using version 0 will change as ABI bugs
1751 The default is version 2.
1753 @item -fno-access-control
1754 @opindex fno-access-control
1755 Turn off all access checking. This switch is mainly useful for working
1756 around bugs in the access control code.
1760 Check that the pointer returned by @code{operator new} is non-null
1761 before attempting to modify the storage allocated. This check is
1762 normally unnecessary because the C++ standard specifies that
1763 @code{operator new} will only return @code{0} if it is declared
1764 @samp{throw()}, in which case the compiler will always check the
1765 return value even without this option. In all other cases, when
1766 @code{operator new} has a non-empty exception specification, memory
1767 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1768 @samp{new (nothrow)}.
1770 @item -fconserve-space
1771 @opindex fconserve-space
1772 Put uninitialized or runtime-initialized global variables into the
1773 common segment, as C does. This saves space in the executable at the
1774 cost of not diagnosing duplicate definitions. If you compile with this
1775 flag and your program mysteriously crashes after @code{main()} has
1776 completed, you may have an object that is being destroyed twice because
1777 two definitions were merged.
1779 This option is no longer useful on most targets, now that support has
1780 been added for putting variables into BSS without making them common.
1782 @item -ffriend-injection
1783 @opindex ffriend-injection
1784 Inject friend functions into the enclosing namespace, so that they are
1785 visible outside the scope of the class in which they are declared.
1786 Friend functions were documented to work this way in the old Annotated
1787 C++ Reference Manual, and versions of G++ before 4.1 always worked
1788 that way. However, in ISO C++ a friend function which is not declared
1789 in an enclosing scope can only be found using argument dependent
1790 lookup. This option causes friends to be injected as they were in
1793 This option is for compatibility, and may be removed in a future
1796 @item -fno-elide-constructors
1797 @opindex fno-elide-constructors
1798 The C++ standard allows an implementation to omit creating a temporary
1799 which is only used to initialize another object of the same type.
1800 Specifying this option disables that optimization, and forces G++ to
1801 call the copy constructor in all cases.
1803 @item -fno-enforce-eh-specs
1804 @opindex fno-enforce-eh-specs
1805 Don't generate code to check for violation of exception specifications
1806 at runtime. This option violates the C++ standard, but may be useful
1807 for reducing code size in production builds, much like defining
1808 @samp{NDEBUG}. This does not give user code permission to throw
1809 exceptions in violation of the exception specifications; the compiler
1810 will still optimize based on the specifications, so throwing an
1811 unexpected exception will result in undefined behavior.
1814 @itemx -fno-for-scope
1816 @opindex fno-for-scope
1817 If @option{-ffor-scope} is specified, the scope of variables declared in
1818 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1819 as specified by the C++ standard.
1820 If @option{-fno-for-scope} is specified, the scope of variables declared in
1821 a @i{for-init-statement} extends to the end of the enclosing scope,
1822 as was the case in old versions of G++, and other (traditional)
1823 implementations of C++.
1825 The default if neither flag is given to follow the standard,
1826 but to allow and give a warning for old-style code that would
1827 otherwise be invalid, or have different behavior.
1829 @item -fno-gnu-keywords
1830 @opindex fno-gnu-keywords
1831 Do not recognize @code{typeof} as a keyword, so that code can use this
1832 word as an identifier. You can use the keyword @code{__typeof__} instead.
1833 @option{-ansi} implies @option{-fno-gnu-keywords}.
1835 @item -fno-implicit-templates
1836 @opindex fno-implicit-templates
1837 Never emit code for non-inline templates which are instantiated
1838 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1839 @xref{Template Instantiation}, for more information.
1841 @item -fno-implicit-inline-templates
1842 @opindex fno-implicit-inline-templates
1843 Don't emit code for implicit instantiations of inline templates, either.
1844 The default is to handle inlines differently so that compiles with and
1845 without optimization will need the same set of explicit instantiations.
1847 @item -fno-implement-inlines
1848 @opindex fno-implement-inlines
1849 To save space, do not emit out-of-line copies of inline functions
1850 controlled by @samp{#pragma implementation}. This will cause linker
1851 errors if these functions are not inlined everywhere they are called.
1853 @item -fms-extensions
1854 @opindex fms-extensions
1855 Disable pedantic warnings about constructs used in MFC, such as implicit
1856 int and getting a pointer to member function via non-standard syntax.
1858 @item -fno-nonansi-builtins
1859 @opindex fno-nonansi-builtins
1860 Disable built-in declarations of functions that are not mandated by
1861 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1862 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1864 @item -fno-operator-names
1865 @opindex fno-operator-names
1866 Do not treat the operator name keywords @code{and}, @code{bitand},
1867 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1868 synonyms as keywords.
1870 @item -fno-optional-diags
1871 @opindex fno-optional-diags
1872 Disable diagnostics that the standard says a compiler does not need to
1873 issue. Currently, the only such diagnostic issued by G++ is the one for
1874 a name having multiple meanings within a class.
1877 @opindex fpermissive
1878 Downgrade some diagnostics about nonconformant code from errors to
1879 warnings. Thus, using @option{-fpermissive} will allow some
1880 nonconforming code to compile.
1882 @item -fno-pretty-templates
1883 @opindex fno-pretty-templates
1884 When an error message refers to a specialization of a function
1885 template, the compiler will normally print the signature of the
1886 template followed by the template arguments and any typedefs or
1887 typenames in the signature (e.g. @code{void f(T) [with T = int]}
1888 rather than @code{void f(int)}) so that it's clear which template is
1889 involved. When an error message refers to a specialization of a class
1890 template, the compiler will omit any template arguments which match
1891 the default template arguments for that template. If either of these
1892 behaviors make it harder to understand the error message rather than
1893 easier, using @option{-fno-pretty-templates} will disable them.
1897 Enable automatic template instantiation at link time. This option also
1898 implies @option{-fno-implicit-templates}. @xref{Template
1899 Instantiation}, for more information.
1903 Disable generation of information about every class with virtual
1904 functions for use by the C++ runtime type identification features
1905 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1906 of the language, you can save some space by using this flag. Note that
1907 exception handling uses the same information, but it will generate it as
1908 needed. The @samp{dynamic_cast} operator can still be used for casts that
1909 do not require runtime type information, i.e.@: casts to @code{void *} or to
1910 unambiguous base classes.
1914 Emit statistics about front-end processing at the end of the compilation.
1915 This information is generally only useful to the G++ development team.
1917 @item -ftemplate-depth-@var{n}
1918 @opindex ftemplate-depth
1919 Set the maximum instantiation depth for template classes to @var{n}.
1920 A limit on the template instantiation depth is needed to detect
1921 endless recursions during template class instantiation. ANSI/ISO C++
1922 conforming programs must not rely on a maximum depth greater than 17.
1924 @item -fno-threadsafe-statics
1925 @opindex fno-threadsafe-statics
1926 Do not emit the extra code to use the routines specified in the C++
1927 ABI for thread-safe initialization of local statics. You can use this
1928 option to reduce code size slightly in code that doesn't need to be
1931 @item -fuse-cxa-atexit
1932 @opindex fuse-cxa-atexit
1933 Register destructors for objects with static storage duration with the
1934 @code{__cxa_atexit} function rather than the @code{atexit} function.
1935 This option is required for fully standards-compliant handling of static
1936 destructors, but will only work if your C library supports
1937 @code{__cxa_atexit}.
1939 @item -fno-use-cxa-get-exception-ptr
1940 @opindex fno-use-cxa-get-exception-ptr
1941 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
1942 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
1943 if the runtime routine is not available.
1945 @item -fvisibility-inlines-hidden
1946 @opindex fvisibility-inlines-hidden
1947 This switch declares that the user does not attempt to compare
1948 pointers to inline methods where the addresses of the two functions
1949 were taken in different shared objects.
1951 The effect of this is that GCC may, effectively, mark inline methods with
1952 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1953 appear in the export table of a DSO and do not require a PLT indirection
1954 when used within the DSO@. Enabling this option can have a dramatic effect
1955 on load and link times of a DSO as it massively reduces the size of the
1956 dynamic export table when the library makes heavy use of templates.
1958 The behavior of this switch is not quite the same as marking the
1959 methods as hidden directly, because it does not affect static variables
1960 local to the function or cause the compiler to deduce that
1961 the function is defined in only one shared object.
1963 You may mark a method as having a visibility explicitly to negate the
1964 effect of the switch for that method. For example, if you do want to
1965 compare pointers to a particular inline method, you might mark it as
1966 having default visibility. Marking the enclosing class with explicit
1967 visibility will have no effect.
1969 Explicitly instantiated inline methods are unaffected by this option
1970 as their linkage might otherwise cross a shared library boundary.
1971 @xref{Template Instantiation}.
1973 @item -fvisibility-ms-compat
1974 @opindex fvisibility-ms-compat
1975 This flag attempts to use visibility settings to make GCC's C++
1976 linkage model compatible with that of Microsoft Visual Studio.
1978 The flag makes these changes to GCC's linkage model:
1982 It sets the default visibility to @code{hidden}, like
1983 @option{-fvisibility=hidden}.
1986 Types, but not their members, are not hidden by default.
1989 The One Definition Rule is relaxed for types without explicit
1990 visibility specifications which are defined in more than one different
1991 shared object: those declarations are permitted if they would have
1992 been permitted when this option was not used.
1995 In new code it is better to use @option{-fvisibility=hidden} and
1996 export those classes which are intended to be externally visible.
1997 Unfortunately it is possible for code to rely, perhaps accidentally,
1998 on the Visual Studio behavior.
2000 Among the consequences of these changes are that static data members
2001 of the same type with the same name but defined in different shared
2002 objects will be different, so changing one will not change the other;
2003 and that pointers to function members defined in different shared
2004 objects may not compare equal. When this flag is given, it is a
2005 violation of the ODR to define types with the same name differently.
2009 Do not use weak symbol support, even if it is provided by the linker.
2010 By default, G++ will use weak symbols if they are available. This
2011 option exists only for testing, and should not be used by end-users;
2012 it will result in inferior code and has no benefits. This option may
2013 be removed in a future release of G++.
2017 Do not search for header files in the standard directories specific to
2018 C++, but do still search the other standard directories. (This option
2019 is used when building the C++ library.)
2022 In addition, these optimization, warning, and code generation options
2023 have meanings only for C++ programs:
2026 @item -fno-default-inline
2027 @opindex fno-default-inline
2028 Do not assume @samp{inline} for functions defined inside a class scope.
2029 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2030 functions will have linkage like inline functions; they just won't be
2033 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2036 Warn when G++ generates code that is probably not compatible with the
2037 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2038 all such cases, there are probably some cases that are not warned about,
2039 even though G++ is generating incompatible code. There may also be
2040 cases where warnings are emitted even though the code that is generated
2043 You should rewrite your code to avoid these warnings if you are
2044 concerned about the fact that code generated by G++ may not be binary
2045 compatible with code generated by other compilers.
2047 The known incompatibilities at this point include:
2052 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2053 pack data into the same byte as a base class. For example:
2056 struct A @{ virtual void f(); int f1 : 1; @};
2057 struct B : public A @{ int f2 : 1; @};
2061 In this case, G++ will place @code{B::f2} into the same byte
2062 as@code{A::f1}; other compilers will not. You can avoid this problem
2063 by explicitly padding @code{A} so that its size is a multiple of the
2064 byte size on your platform; that will cause G++ and other compilers to
2065 layout @code{B} identically.
2068 Incorrect handling of tail-padding for virtual bases. G++ does not use
2069 tail padding when laying out virtual bases. For example:
2072 struct A @{ virtual void f(); char c1; @};
2073 struct B @{ B(); char c2; @};
2074 struct C : public A, public virtual B @{@};
2078 In this case, G++ will not place @code{B} into the tail-padding for
2079 @code{A}; other compilers will. You can avoid this problem by
2080 explicitly padding @code{A} so that its size is a multiple of its
2081 alignment (ignoring virtual base classes); that will cause G++ and other
2082 compilers to layout @code{C} identically.
2085 Incorrect handling of bit-fields with declared widths greater than that
2086 of their underlying types, when the bit-fields appear in a union. For
2090 union U @{ int i : 4096; @};
2094 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2095 union too small by the number of bits in an @code{int}.
2098 Empty classes can be placed at incorrect offsets. For example:
2108 struct C : public B, public A @{@};
2112 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2113 it should be placed at offset zero. G++ mistakenly believes that the
2114 @code{A} data member of @code{B} is already at offset zero.
2117 Names of template functions whose types involve @code{typename} or
2118 template template parameters can be mangled incorrectly.
2121 template <typename Q>
2122 void f(typename Q::X) @{@}
2124 template <template <typename> class Q>
2125 void f(typename Q<int>::X) @{@}
2129 Instantiations of these templates may be mangled incorrectly.
2133 It also warns psABI related changes. The known psABI changes at this
2139 For SYSV/x86-64, when passing union with long double, it is changed to
2140 pass in memory as specified in psABI. For example:
2150 @code{union U} will always be passed in memory.
2154 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2155 @opindex Wctor-dtor-privacy
2156 @opindex Wno-ctor-dtor-privacy
2157 Warn when a class seems unusable because all the constructors or
2158 destructors in that class are private, and it has neither friends nor
2159 public static member functions.
2161 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2162 @opindex Wnon-virtual-dtor
2163 @opindex Wno-non-virtual-dtor
2164 Warn when a class has virtual functions and accessible non-virtual
2165 destructor, in which case it would be possible but unsafe to delete
2166 an instance of a derived class through a pointer to the base class.
2167 This warning is also enabled if -Weffc++ is specified.
2169 @item -Wreorder @r{(C++ and Objective-C++ only)}
2171 @opindex Wno-reorder
2172 @cindex reordering, warning
2173 @cindex warning for reordering of member initializers
2174 Warn when the order of member initializers given in the code does not
2175 match the order in which they must be executed. For instance:
2181 A(): j (0), i (1) @{ @}
2185 The compiler will rearrange the member initializers for @samp{i}
2186 and @samp{j} to match the declaration order of the members, emitting
2187 a warning to that effect. This warning is enabled by @option{-Wall}.
2190 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2193 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2196 Warn about violations of the following style guidelines from Scott Meyers'
2197 @cite{Effective C++} book:
2201 Item 11: Define a copy constructor and an assignment operator for classes
2202 with dynamically allocated memory.
2205 Item 12: Prefer initialization to assignment in constructors.
2208 Item 14: Make destructors virtual in base classes.
2211 Item 15: Have @code{operator=} return a reference to @code{*this}.
2214 Item 23: Don't try to return a reference when you must return an object.
2218 Also warn about violations of the following style guidelines from
2219 Scott Meyers' @cite{More Effective C++} book:
2223 Item 6: Distinguish between prefix and postfix forms of increment and
2224 decrement operators.
2227 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2231 When selecting this option, be aware that the standard library
2232 headers do not obey all of these guidelines; use @samp{grep -v}
2233 to filter out those warnings.
2235 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2236 @opindex Wstrict-null-sentinel
2237 @opindex Wno-strict-null-sentinel
2238 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2239 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2240 to @code{__null}. Although it is a null pointer constant not a null pointer,
2241 it is guaranteed to be of the same size as a pointer. But this use is
2242 not portable across different compilers.
2244 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2245 @opindex Wno-non-template-friend
2246 @opindex Wnon-template-friend
2247 Disable warnings when non-templatized friend functions are declared
2248 within a template. Since the advent of explicit template specification
2249 support in G++, if the name of the friend is an unqualified-id (i.e.,
2250 @samp{friend foo(int)}), the C++ language specification demands that the
2251 friend declare or define an ordinary, nontemplate function. (Section
2252 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2253 could be interpreted as a particular specialization of a templatized
2254 function. Because this non-conforming behavior is no longer the default
2255 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2256 check existing code for potential trouble spots and is on by default.
2257 This new compiler behavior can be turned off with
2258 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2259 but disables the helpful warning.
2261 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2262 @opindex Wold-style-cast
2263 @opindex Wno-old-style-cast
2264 Warn if an old-style (C-style) cast to a non-void type is used within
2265 a C++ program. The new-style casts (@samp{dynamic_cast},
2266 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2267 less vulnerable to unintended effects and much easier to search for.
2269 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2270 @opindex Woverloaded-virtual
2271 @opindex Wno-overloaded-virtual
2272 @cindex overloaded virtual fn, warning
2273 @cindex warning for overloaded virtual fn
2274 Warn when a function declaration hides virtual functions from a
2275 base class. For example, in:
2282 struct B: public A @{
2287 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2295 will fail to compile.
2297 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2298 @opindex Wno-pmf-conversions
2299 @opindex Wpmf-conversions
2300 Disable the diagnostic for converting a bound pointer to member function
2303 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2304 @opindex Wsign-promo
2305 @opindex Wno-sign-promo
2306 Warn when overload resolution chooses a promotion from unsigned or
2307 enumerated type to a signed type, over a conversion to an unsigned type of
2308 the same size. Previous versions of G++ would try to preserve
2309 unsignedness, but the standard mandates the current behavior.
2314 A& operator = (int);
2324 In this example, G++ will synthesize a default @samp{A& operator =
2325 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2328 @node Objective-C and Objective-C++ Dialect Options
2329 @section Options Controlling Objective-C and Objective-C++ Dialects
2331 @cindex compiler options, Objective-C and Objective-C++
2332 @cindex Objective-C and Objective-C++ options, command line
2333 @cindex options, Objective-C and Objective-C++
2334 (NOTE: This manual does not describe the Objective-C and Objective-C++
2335 languages themselves. See @xref{Standards,,Language Standards
2336 Supported by GCC}, for references.)
2338 This section describes the command-line options that are only meaningful
2339 for Objective-C and Objective-C++ programs, but you can also use most of
2340 the language-independent GNU compiler options.
2341 For example, you might compile a file @code{some_class.m} like this:
2344 gcc -g -fgnu-runtime -O -c some_class.m
2348 In this example, @option{-fgnu-runtime} is an option meant only for
2349 Objective-C and Objective-C++ programs; you can use the other options with
2350 any language supported by GCC@.
2352 Note that since Objective-C is an extension of the C language, Objective-C
2353 compilations may also use options specific to the C front-end (e.g.,
2354 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2355 C++-specific options (e.g., @option{-Wabi}).
2357 Here is a list of options that are @emph{only} for compiling Objective-C
2358 and Objective-C++ programs:
2361 @item -fconstant-string-class=@var{class-name}
2362 @opindex fconstant-string-class
2363 Use @var{class-name} as the name of the class to instantiate for each
2364 literal string specified with the syntax @code{@@"@dots{}"}. The default
2365 class name is @code{NXConstantString} if the GNU runtime is being used, and
2366 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2367 @option{-fconstant-cfstrings} option, if also present, will override the
2368 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2369 to be laid out as constant CoreFoundation strings.
2372 @opindex fgnu-runtime
2373 Generate object code compatible with the standard GNU Objective-C
2374 runtime. This is the default for most types of systems.
2376 @item -fnext-runtime
2377 @opindex fnext-runtime
2378 Generate output compatible with the NeXT runtime. This is the default
2379 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2380 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2383 @item -fno-nil-receivers
2384 @opindex fno-nil-receivers
2385 Assume that all Objective-C message dispatches (e.g.,
2386 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
2387 is not @code{nil}. This allows for more efficient entry points in the runtime
2388 to be used. Currently, this option is only available in conjunction with
2389 the NeXT runtime on Mac OS X 10.3 and later.
2391 @item -fobjc-call-cxx-cdtors
2392 @opindex fobjc-call-cxx-cdtors
2393 For each Objective-C class, check if any of its instance variables is a
2394 C++ object with a non-trivial default constructor. If so, synthesize a
2395 special @code{- (id) .cxx_construct} instance method that will run
2396 non-trivial default constructors on any such instance variables, in order,
2397 and then return @code{self}. Similarly, check if any instance variable
2398 is a C++ object with a non-trivial destructor, and if so, synthesize a
2399 special @code{- (void) .cxx_destruct} method that will run
2400 all such default destructors, in reverse order.
2402 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
2403 thusly generated will only operate on instance variables declared in the
2404 current Objective-C class, and not those inherited from superclasses. It
2405 is the responsibility of the Objective-C runtime to invoke all such methods
2406 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
2407 will be invoked by the runtime immediately after a new object
2408 instance is allocated; the @code{- (void) .cxx_destruct} methods will
2409 be invoked immediately before the runtime deallocates an object instance.
2411 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2412 support for invoking the @code{- (id) .cxx_construct} and
2413 @code{- (void) .cxx_destruct} methods.
2415 @item -fobjc-direct-dispatch
2416 @opindex fobjc-direct-dispatch
2417 Allow fast jumps to the message dispatcher. On Darwin this is
2418 accomplished via the comm page.
2420 @item -fobjc-exceptions
2421 @opindex fobjc-exceptions
2422 Enable syntactic support for structured exception handling in Objective-C,
2423 similar to what is offered by C++ and Java. This option is
2424 unavailable in conjunction with the NeXT runtime on Mac OS X 10.2 and
2433 @@catch (AnObjCClass *exc) @{
2440 @@catch (AnotherClass *exc) @{
2443 @@catch (id allOthers) @{
2453 The @code{@@throw} statement may appear anywhere in an Objective-C or
2454 Objective-C++ program; when used inside of a @code{@@catch} block, the
2455 @code{@@throw} may appear without an argument (as shown above), in which case
2456 the object caught by the @code{@@catch} will be rethrown.
2458 Note that only (pointers to) Objective-C objects may be thrown and
2459 caught using this scheme. When an object is thrown, it will be caught
2460 by the nearest @code{@@catch} clause capable of handling objects of that type,
2461 analogously to how @code{catch} blocks work in C++ and Java. A
2462 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2463 any and all Objective-C exceptions not caught by previous @code{@@catch}
2466 The @code{@@finally} clause, if present, will be executed upon exit from the
2467 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2468 regardless of whether any exceptions are thrown, caught or rethrown
2469 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2470 of the @code{finally} clause in Java.
2472 There are several caveats to using the new exception mechanism:
2476 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2477 idioms provided by the @code{NSException} class, the new
2478 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2479 systems, due to additional functionality needed in the (NeXT) Objective-C
2483 As mentioned above, the new exceptions do not support handling
2484 types other than Objective-C objects. Furthermore, when used from
2485 Objective-C++, the Objective-C exception model does not interoperate with C++
2486 exceptions at this time. This means you cannot @code{@@throw} an exception
2487 from Objective-C and @code{catch} it in C++, or vice versa
2488 (i.e., @code{throw @dots{} @@catch}).
2491 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2492 blocks for thread-safe execution:
2495 @@synchronized (ObjCClass *guard) @{
2500 Upon entering the @code{@@synchronized} block, a thread of execution shall
2501 first check whether a lock has been placed on the corresponding @code{guard}
2502 object by another thread. If it has, the current thread shall wait until
2503 the other thread relinquishes its lock. Once @code{guard} becomes available,
2504 the current thread will place its own lock on it, execute the code contained in
2505 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2506 making @code{guard} available to other threads).
2508 Unlike Java, Objective-C does not allow for entire methods to be marked
2509 @code{@@synchronized}. Note that throwing exceptions out of
2510 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2511 to be unlocked properly.
2515 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2517 @item -freplace-objc-classes
2518 @opindex freplace-objc-classes
2519 Emit a special marker instructing @command{ld(1)} not to statically link in
2520 the resulting object file, and allow @command{dyld(1)} to load it in at
2521 run time instead. This is used in conjunction with the Fix-and-Continue
2522 debugging mode, where the object file in question may be recompiled and
2523 dynamically reloaded in the course of program execution, without the need
2524 to restart the program itself. Currently, Fix-and-Continue functionality
2525 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2530 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2531 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2532 compile time) with static class references that get initialized at load time,
2533 which improves run-time performance. Specifying the @option{-fzero-link} flag
2534 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2535 to be retained. This is useful in Zero-Link debugging mode, since it allows
2536 for individual class implementations to be modified during program execution.
2540 Dump interface declarations for all classes seen in the source file to a
2541 file named @file{@var{sourcename}.decl}.
2543 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2544 @opindex Wassign-intercept
2545 @opindex Wno-assign-intercept
2546 Warn whenever an Objective-C assignment is being intercepted by the
2549 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2550 @opindex Wno-protocol
2552 If a class is declared to implement a protocol, a warning is issued for
2553 every method in the protocol that is not implemented by the class. The
2554 default behavior is to issue a warning for every method not explicitly
2555 implemented in the class, even if a method implementation is inherited
2556 from the superclass. If you use the @option{-Wno-protocol} option, then
2557 methods inherited from the superclass are considered to be implemented,
2558 and no warning is issued for them.
2560 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2562 @opindex Wno-selector
2563 Warn if multiple methods of different types for the same selector are
2564 found during compilation. The check is performed on the list of methods
2565 in the final stage of compilation. Additionally, a check is performed
2566 for each selector appearing in a @code{@@selector(@dots{})}
2567 expression, and a corresponding method for that selector has been found
2568 during compilation. Because these checks scan the method table only at
2569 the end of compilation, these warnings are not produced if the final
2570 stage of compilation is not reached, for example because an error is
2571 found during compilation, or because the @option{-fsyntax-only} option is
2574 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2575 @opindex Wstrict-selector-match
2576 @opindex Wno-strict-selector-match
2577 Warn if multiple methods with differing argument and/or return types are
2578 found for a given selector when attempting to send a message using this
2579 selector to a receiver of type @code{id} or @code{Class}. When this flag
2580 is off (which is the default behavior), the compiler will omit such warnings
2581 if any differences found are confined to types which share the same size
2584 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2585 @opindex Wundeclared-selector
2586 @opindex Wno-undeclared-selector
2587 Warn if a @code{@@selector(@dots{})} expression referring to an
2588 undeclared selector is found. A selector is considered undeclared if no
2589 method with that name has been declared before the
2590 @code{@@selector(@dots{})} expression, either explicitly in an
2591 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2592 an @code{@@implementation} section. This option always performs its
2593 checks as soon as a @code{@@selector(@dots{})} expression is found,
2594 while @option{-Wselector} only performs its checks in the final stage of
2595 compilation. This also enforces the coding style convention
2596 that methods and selectors must be declared before being used.
2598 @item -print-objc-runtime-info
2599 @opindex print-objc-runtime-info
2600 Generate C header describing the largest structure that is passed by
2605 @node Language Independent Options
2606 @section Options to Control Diagnostic Messages Formatting
2607 @cindex options to control diagnostics formatting
2608 @cindex diagnostic messages
2609 @cindex message formatting
2611 Traditionally, diagnostic messages have been formatted irrespective of
2612 the output device's aspect (e.g.@: its width, @dots{}). The options described
2613 below can be used to control the diagnostic messages formatting
2614 algorithm, e.g.@: how many characters per line, how often source location
2615 information should be reported. Right now, only the C++ front end can
2616 honor these options. However it is expected, in the near future, that
2617 the remaining front ends would be able to digest them correctly.
2620 @item -fmessage-length=@var{n}
2621 @opindex fmessage-length
2622 Try to format error messages so that they fit on lines of about @var{n}
2623 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2624 the front ends supported by GCC@. If @var{n} is zero, then no
2625 line-wrapping will be done; each error message will appear on a single
2628 @opindex fdiagnostics-show-location
2629 @item -fdiagnostics-show-location=once
2630 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2631 reporter to emit @emph{once} source location information; that is, in
2632 case the message is too long to fit on a single physical line and has to
2633 be wrapped, the source location won't be emitted (as prefix) again,
2634 over and over, in subsequent continuation lines. This is the default
2637 @item -fdiagnostics-show-location=every-line
2638 Only meaningful in line-wrapping mode. Instructs the diagnostic
2639 messages reporter to emit the same source location information (as
2640 prefix) for physical lines that result from the process of breaking
2641 a message which is too long to fit on a single line.
2643 @item -fdiagnostics-show-option
2644 @opindex fdiagnostics-show-option
2645 This option instructs the diagnostic machinery to add text to each
2646 diagnostic emitted, which indicates which command line option directly
2647 controls that diagnostic, when such an option is known to the
2648 diagnostic machinery.
2650 @item -Wcoverage-mismatch
2651 @opindex Wcoverage-mismatch
2652 Warn if feedback profiles do not match when using the
2653 @option{-fprofile-use} option.
2654 If a source file was changed between @option{-fprofile-gen} and
2655 @option{-fprofile-use}, the files with the profile feedback can fail
2656 to match the source file and GCC can not use the profile feedback
2657 information. By default, GCC emits an error message in this case.
2658 The option @option{-Wcoverage-mismatch} emits a warning instead of an
2659 error. GCC does not use appropriate feedback profiles, so using this
2660 option can result in poorly optimized code. This option is useful
2661 only in the case of very minor changes such as bug fixes to an
2666 @node Warning Options
2667 @section Options to Request or Suppress Warnings
2668 @cindex options to control warnings
2669 @cindex warning messages
2670 @cindex messages, warning
2671 @cindex suppressing warnings
2673 Warnings are diagnostic messages that report constructions which
2674 are not inherently erroneous but which are risky or suggest there
2675 may have been an error.
2677 The following language-independent options do not enable specific
2678 warnings but control the kinds of diagnostics produced by GCC.
2681 @cindex syntax checking
2683 @opindex fsyntax-only
2684 Check the code for syntax errors, but don't do anything beyond that.
2688 Inhibit all warning messages.
2693 Make all warnings into errors.
2698 Make the specified warning into an error. The specifier for a warning
2699 is appended, for example @option{-Werror=switch} turns the warnings
2700 controlled by @option{-Wswitch} into errors. This switch takes a
2701 negative form, to be used to negate @option{-Werror} for specific
2702 warnings, for example @option{-Wno-error=switch} makes
2703 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2704 is in effect. You can use the @option{-fdiagnostics-show-option}
2705 option to have each controllable warning amended with the option which
2706 controls it, to determine what to use with this option.
2708 Note that specifying @option{-Werror=}@var{foo} automatically implies
2709 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2712 @item -Wfatal-errors
2713 @opindex Wfatal-errors
2714 @opindex Wno-fatal-errors
2715 This option causes the compiler to abort compilation on the first error
2716 occurred rather than trying to keep going and printing further error
2721 You can request many specific warnings with options beginning
2722 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2723 implicit declarations. Each of these specific warning options also
2724 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2725 example, @option{-Wno-implicit}. This manual lists only one of the
2726 two forms, whichever is not the default. For further,
2727 language-specific options also refer to @ref{C++ Dialect Options} and
2728 @ref{Objective-C and Objective-C++ Dialect Options}.
2733 Issue all the warnings demanded by strict ISO C and ISO C++;
2734 reject all programs that use forbidden extensions, and some other
2735 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2736 version of the ISO C standard specified by any @option{-std} option used.
2738 Valid ISO C and ISO C++ programs should compile properly with or without
2739 this option (though a rare few will require @option{-ansi} or a
2740 @option{-std} option specifying the required version of ISO C)@. However,
2741 without this option, certain GNU extensions and traditional C and C++
2742 features are supported as well. With this option, they are rejected.
2744 @option{-pedantic} does not cause warning messages for use of the
2745 alternate keywords whose names begin and end with @samp{__}. Pedantic
2746 warnings are also disabled in the expression that follows
2747 @code{__extension__}. However, only system header files should use
2748 these escape routes; application programs should avoid them.
2749 @xref{Alternate Keywords}.
2751 Some users try to use @option{-pedantic} to check programs for strict ISO
2752 C conformance. They soon find that it does not do quite what they want:
2753 it finds some non-ISO practices, but not all---only those for which
2754 ISO C @emph{requires} a diagnostic, and some others for which
2755 diagnostics have been added.
2757 A feature to report any failure to conform to ISO C might be useful in
2758 some instances, but would require considerable additional work and would
2759 be quite different from @option{-pedantic}. We don't have plans to
2760 support such a feature in the near future.
2762 Where the standard specified with @option{-std} represents a GNU
2763 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2764 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2765 extended dialect is based. Warnings from @option{-pedantic} are given
2766 where they are required by the base standard. (It would not make sense
2767 for such warnings to be given only for features not in the specified GNU
2768 C dialect, since by definition the GNU dialects of C include all
2769 features the compiler supports with the given option, and there would be
2770 nothing to warn about.)
2772 @item -pedantic-errors
2773 @opindex pedantic-errors
2774 Like @option{-pedantic}, except that errors are produced rather than
2780 This enables all the warnings about constructions that some users
2781 consider questionable, and that are easy to avoid (or modify to
2782 prevent the warning), even in conjunction with macros. This also
2783 enables some language-specific warnings described in @ref{C++ Dialect
2784 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2786 @option{-Wall} turns on the following warning flags:
2788 @gccoptlist{-Waddress @gol
2789 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2791 -Wchar-subscripts @gol
2792 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
2794 -Wimplicit-function-declaration @gol
2797 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2798 -Wmissing-braces @gol
2804 -Wsequence-point @gol
2805 -Wsign-compare @r{(only in C++)} @gol
2806 -Wstrict-aliasing @gol
2807 -Wstrict-overflow=1 @gol
2810 -Wuninitialized @gol
2811 -Wunknown-pragmas @gol
2812 -Wunused-function @gol
2815 -Wunused-variable @gol
2816 -Wvolatile-register-var @gol
2819 Note that some warning flags are not implied by @option{-Wall}. Some of
2820 them warn about constructions that users generally do not consider
2821 questionable, but which occasionally you might wish to check for;
2822 others warn about constructions that are necessary or hard to avoid in
2823 some cases, and there is no simple way to modify the code to suppress
2824 the warning. Some of them are enabled by @option{-Wextra} but many of
2825 them must be enabled individually.
2831 This enables some extra warning flags that are not enabled by
2832 @option{-Wall}. (This option used to be called @option{-W}. The older
2833 name is still supported, but the newer name is more descriptive.)
2835 @gccoptlist{-Wclobbered @gol
2837 -Wignored-qualifiers @gol
2838 -Wmissing-field-initializers @gol
2839 -Wmissing-parameter-type @r{(C only)} @gol
2840 -Wold-style-declaration @r{(C only)} @gol
2841 -Woverride-init @gol
2844 -Wuninitialized @gol
2845 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2848 The option @option{-Wextra} also prints warning messages for the
2854 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2855 @samp{>}, or @samp{>=}.
2858 (C++ only) An enumerator and a non-enumerator both appear in a
2859 conditional expression.
2862 (C++ only) Ambiguous virtual bases.
2865 (C++ only) Subscripting an array which has been declared @samp{register}.
2868 (C++ only) Taking the address of a variable which has been declared
2872 (C++ only) A base class is not initialized in a derived class' copy
2877 @item -Wchar-subscripts
2878 @opindex Wchar-subscripts
2879 @opindex Wno-char-subscripts
2880 Warn if an array subscript has type @code{char}. This is a common cause
2881 of error, as programmers often forget that this type is signed on some
2883 This warning is enabled by @option{-Wall}.
2887 @opindex Wno-comment
2888 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2889 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2890 This warning is enabled by @option{-Wall}.
2895 @opindex ffreestanding
2896 @opindex fno-builtin
2897 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2898 the arguments supplied have types appropriate to the format string
2899 specified, and that the conversions specified in the format string make
2900 sense. This includes standard functions, and others specified by format
2901 attributes (@pxref{Function Attributes}), in the @code{printf},
2902 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2903 not in the C standard) families (or other target-specific families).
2904 Which functions are checked without format attributes having been
2905 specified depends on the standard version selected, and such checks of
2906 functions without the attribute specified are disabled by
2907 @option{-ffreestanding} or @option{-fno-builtin}.
2909 The formats are checked against the format features supported by GNU
2910 libc version 2.2. These include all ISO C90 and C99 features, as well
2911 as features from the Single Unix Specification and some BSD and GNU
2912 extensions. Other library implementations may not support all these
2913 features; GCC does not support warning about features that go beyond a
2914 particular library's limitations. However, if @option{-pedantic} is used
2915 with @option{-Wformat}, warnings will be given about format features not
2916 in the selected standard version (but not for @code{strfmon} formats,
2917 since those are not in any version of the C standard). @xref{C Dialect
2918 Options,,Options Controlling C Dialect}.
2920 Since @option{-Wformat} also checks for null format arguments for
2921 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2923 @option{-Wformat} is included in @option{-Wall}. For more control over some
2924 aspects of format checking, the options @option{-Wformat-y2k},
2925 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2926 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2927 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2930 @opindex Wformat-y2k
2931 @opindex Wno-format-y2k
2932 If @option{-Wformat} is specified, also warn about @code{strftime}
2933 formats which may yield only a two-digit year.
2935 @item -Wno-format-contains-nul
2936 @opindex Wno-format-contains-nul
2937 @opindex Wformat-contains-nul
2938 If @option{-Wformat} is specified, do not warn about format strings that
2941 @item -Wno-format-extra-args
2942 @opindex Wno-format-extra-args
2943 @opindex Wformat-extra-args
2944 If @option{-Wformat} is specified, do not warn about excess arguments to a
2945 @code{printf} or @code{scanf} format function. The C standard specifies
2946 that such arguments are ignored.
2948 Where the unused arguments lie between used arguments that are
2949 specified with @samp{$} operand number specifications, normally
2950 warnings are still given, since the implementation could not know what
2951 type to pass to @code{va_arg} to skip the unused arguments. However,
2952 in the case of @code{scanf} formats, this option will suppress the
2953 warning if the unused arguments are all pointers, since the Single
2954 Unix Specification says that such unused arguments are allowed.
2956 @item -Wno-format-zero-length @r{(C and Objective-C only)}
2957 @opindex Wno-format-zero-length
2958 @opindex Wformat-zero-length
2959 If @option{-Wformat} is specified, do not warn about zero-length formats.
2960 The C standard specifies that zero-length formats are allowed.
2962 @item -Wformat-nonliteral
2963 @opindex Wformat-nonliteral
2964 @opindex Wno-format-nonliteral
2965 If @option{-Wformat} is specified, also warn if the format string is not a
2966 string literal and so cannot be checked, unless the format function
2967 takes its format arguments as a @code{va_list}.
2969 @item -Wformat-security
2970 @opindex Wformat-security
2971 @opindex Wno-format-security
2972 If @option{-Wformat} is specified, also warn about uses of format
2973 functions that represent possible security problems. At present, this
2974 warns about calls to @code{printf} and @code{scanf} functions where the
2975 format string is not a string literal and there are no format arguments,
2976 as in @code{printf (foo);}. This may be a security hole if the format
2977 string came from untrusted input and contains @samp{%n}. (This is
2978 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2979 in future warnings may be added to @option{-Wformat-security} that are not
2980 included in @option{-Wformat-nonliteral}.)
2984 @opindex Wno-format=2
2985 Enable @option{-Wformat} plus format checks not included in
2986 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2987 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2989 @item -Wnonnull @r{(C and Objective-C only)}
2991 @opindex Wno-nonnull
2992 Warn about passing a null pointer for arguments marked as
2993 requiring a non-null value by the @code{nonnull} function attribute.
2995 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2996 can be disabled with the @option{-Wno-nonnull} option.
2998 @item -Wjump-misses-init @r{(C, Objective-C only)}
2999 @opindex Wjump-misses-init
3000 @opindex Wno-jump-misses-init
3001 Warn if a @code{goto} statement or a @code{switch} statement jumps
3002 forward across the initialization of a variable, or jumps backward to a
3003 label after the variable has been initialized. This only warns about
3004 variables which are initialized when they are declared. This warning is
3005 only supported for C and Objective C; in C++ this sort of branch is an
3008 @option{-Wjump-misses-init} is included in @option{-Wall} and
3009 @option{-Wc++-compat}. It can be disabled with the
3010 @option{-Wno-jump-misses-init} option.
3012 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3014 @opindex Wno-init-self
3015 Warn about uninitialized variables which are initialized with themselves.
3016 Note this option can only be used with the @option{-Wuninitialized} option.
3018 For example, GCC will warn about @code{i} being uninitialized in the
3019 following snippet only when @option{-Winit-self} has been specified:
3030 @item -Wimplicit-int @r{(C and Objective-C only)}
3031 @opindex Wimplicit-int
3032 @opindex Wno-implicit-int
3033 Warn when a declaration does not specify a type.
3034 This warning is enabled by @option{-Wall}.
3036 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3037 @opindex Wimplicit-function-declaration
3038 @opindex Wno-implicit-function-declaration
3039 Give a warning whenever a function is used before being declared. In
3040 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3041 enabled by default and it is made into an error by
3042 @option{-pedantic-errors}. This warning is also enabled by
3047 @opindex Wno-implicit
3048 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3049 This warning is enabled by @option{-Wall}.
3051 @item -Wignored-qualifiers @r{(C and C++ only)}
3052 @opindex Wignored-qualifiers
3053 @opindex Wno-ignored-qualifiers
3054 Warn if the return type of a function has a type qualifier
3055 such as @code{const}. For ISO C such a type qualifier has no effect,
3056 since the value returned by a function is not an lvalue.
3057 For C++, the warning is only emitted for scalar types or @code{void}.
3058 ISO C prohibits qualified @code{void} return types on function
3059 definitions, so such return types always receive a warning
3060 even without this option.
3062 This warning is also enabled by @option{-Wextra}.
3067 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3068 a function with external linkage, returning int, taking either zero
3069 arguments, two, or three arguments of appropriate types. This warning
3070 is enabled by default in C++ and is enabled by either @option{-Wall}
3071 or @option{-pedantic}.
3073 @item -Wmissing-braces
3074 @opindex Wmissing-braces
3075 @opindex Wno-missing-braces
3076 Warn if an aggregate or union initializer is not fully bracketed. In
3077 the following example, the initializer for @samp{a} is not fully
3078 bracketed, but that for @samp{b} is fully bracketed.
3081 int a[2][2] = @{ 0, 1, 2, 3 @};
3082 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3085 This warning is enabled by @option{-Wall}.
3087 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3088 @opindex Wmissing-include-dirs
3089 @opindex Wno-missing-include-dirs
3090 Warn if a user-supplied include directory does not exist.
3093 @opindex Wparentheses
3094 @opindex Wno-parentheses
3095 Warn if parentheses are omitted in certain contexts, such
3096 as when there is an assignment in a context where a truth value
3097 is expected, or when operators are nested whose precedence people
3098 often get confused about.
3100 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3101 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3102 interpretation from that of ordinary mathematical notation.
3104 Also warn about constructions where there may be confusion to which
3105 @code{if} statement an @code{else} branch belongs. Here is an example of
3120 In C/C++, every @code{else} branch belongs to the innermost possible
3121 @code{if} statement, which in this example is @code{if (b)}. This is
3122 often not what the programmer expected, as illustrated in the above
3123 example by indentation the programmer chose. When there is the
3124 potential for this confusion, GCC will issue a warning when this flag
3125 is specified. To eliminate the warning, add explicit braces around
3126 the innermost @code{if} statement so there is no way the @code{else}
3127 could belong to the enclosing @code{if}. The resulting code would
3144 This warning is enabled by @option{-Wall}.
3146 @item -Wsequence-point
3147 @opindex Wsequence-point
3148 @opindex Wno-sequence-point
3149 Warn about code that may have undefined semantics because of violations
3150 of sequence point rules in the C and C++ standards.
3152 The C and C++ standards defines the order in which expressions in a C/C++
3153 program are evaluated in terms of @dfn{sequence points}, which represent
3154 a partial ordering between the execution of parts of the program: those
3155 executed before the sequence point, and those executed after it. These
3156 occur after the evaluation of a full expression (one which is not part
3157 of a larger expression), after the evaluation of the first operand of a
3158 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3159 function is called (but after the evaluation of its arguments and the
3160 expression denoting the called function), and in certain other places.
3161 Other than as expressed by the sequence point rules, the order of
3162 evaluation of subexpressions of an expression is not specified. All
3163 these rules describe only a partial order rather than a total order,
3164 since, for example, if two functions are called within one expression
3165 with no sequence point between them, the order in which the functions
3166 are called is not specified. However, the standards committee have
3167 ruled that function calls do not overlap.
3169 It is not specified when between sequence points modifications to the
3170 values of objects take effect. Programs whose behavior depends on this
3171 have undefined behavior; the C and C++ standards specify that ``Between
3172 the previous and next sequence point an object shall have its stored
3173 value modified at most once by the evaluation of an expression.
3174 Furthermore, the prior value shall be read only to determine the value
3175 to be stored.''. If a program breaks these rules, the results on any
3176 particular implementation are entirely unpredictable.
3178 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3179 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3180 diagnosed by this option, and it may give an occasional false positive
3181 result, but in general it has been found fairly effective at detecting
3182 this sort of problem in programs.
3184 The standard is worded confusingly, therefore there is some debate
3185 over the precise meaning of the sequence point rules in subtle cases.
3186 Links to discussions of the problem, including proposed formal
3187 definitions, may be found on the GCC readings page, at
3188 @w{@uref{http://gcc.gnu.org/readings.html}}.
3190 This warning is enabled by @option{-Wall} for C and C++.
3193 @opindex Wreturn-type
3194 @opindex Wno-return-type
3195 Warn whenever a function is defined with a return-type that defaults
3196 to @code{int}. Also warn about any @code{return} statement with no
3197 return-value in a function whose return-type is not @code{void}
3198 (falling off the end of the function body is considered returning
3199 without a value), and about a @code{return} statement with an
3200 expression in a function whose return-type is @code{void}.
3202 For C++, a function without return type always produces a diagnostic
3203 message, even when @option{-Wno-return-type} is specified. The only
3204 exceptions are @samp{main} and functions defined in system headers.
3206 This warning is enabled by @option{-Wall}.
3211 Warn whenever a @code{switch} statement has an index of enumerated type
3212 and lacks a @code{case} for one or more of the named codes of that
3213 enumeration. (The presence of a @code{default} label prevents this
3214 warning.) @code{case} labels outside the enumeration range also
3215 provoke warnings when this option is used (even if there is a
3216 @code{default} label).
3217 This warning is enabled by @option{-Wall}.
3219 @item -Wswitch-default
3220 @opindex Wswitch-default
3221 @opindex Wno-switch-default
3222 Warn whenever a @code{switch} statement does not have a @code{default}
3226 @opindex Wswitch-enum
3227 @opindex Wno-switch-enum
3228 Warn whenever a @code{switch} statement has an index of enumerated type
3229 and lacks a @code{case} for one or more of the named codes of that
3230 enumeration. @code{case} labels outside the enumeration range also
3231 provoke warnings when this option is used. The only difference
3232 between @option{-Wswitch} and this option is that this option gives a
3233 warning about an omitted enumeration code even if there is a
3234 @code{default} label.
3236 @item -Wsync-nand @r{(C and C++ only)}
3238 @opindex Wno-sync-nand
3239 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3240 built-in functions are used. These functions changed semantics in GCC 4.4.
3244 @opindex Wno-trigraphs
3245 Warn if any trigraphs are encountered that might change the meaning of
3246 the program (trigraphs within comments are not warned about).
3247 This warning is enabled by @option{-Wall}.
3249 @item -Wunused-function
3250 @opindex Wunused-function
3251 @opindex Wno-unused-function
3252 Warn whenever a static function is declared but not defined or a
3253 non-inline static function is unused.
3254 This warning is enabled by @option{-Wall}.
3256 @item -Wunused-label
3257 @opindex Wunused-label
3258 @opindex Wno-unused-label
3259 Warn whenever a label is declared but not used.
3260 This warning is enabled by @option{-Wall}.
3262 To suppress this warning use the @samp{unused} attribute
3263 (@pxref{Variable Attributes}).
3265 @item -Wunused-parameter
3266 @opindex Wunused-parameter
3267 @opindex Wno-unused-parameter
3268 Warn whenever a function parameter is unused aside from its declaration.
3270 To suppress this warning use the @samp{unused} attribute
3271 (@pxref{Variable Attributes}).
3273 @item -Wno-unused-result
3274 @opindex Wunused-result
3275 @opindex Wno-unused-result
3276 Do not warn if a caller of a function marked with attribute
3277 @code{warn_unused_result} (@pxref{Variable Attributes}) does not use
3278 its return value. The default is @option{-Wunused-result}.
3280 @item -Wunused-variable
3281 @opindex Wunused-variable
3282 @opindex Wno-unused-variable
3283 Warn whenever a local variable or non-constant static variable is unused
3284 aside from its declaration.
3285 This warning is enabled by @option{-Wall}.
3287 To suppress this warning use the @samp{unused} attribute
3288 (@pxref{Variable Attributes}).
3290 @item -Wunused-value
3291 @opindex Wunused-value
3292 @opindex Wno-unused-value
3293 Warn whenever a statement computes a result that is explicitly not
3294 used. To suppress this warning cast the unused expression to
3295 @samp{void}. This includes an expression-statement or the left-hand
3296 side of a comma expression that contains no side effects. For example,
3297 an expression such as @samp{x[i,j]} will cause a warning, while
3298 @samp{x[(void)i,j]} will not.
3300 This warning is enabled by @option{-Wall}.
3305 All the above @option{-Wunused} options combined.
3307 In order to get a warning about an unused function parameter, you must
3308 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3309 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3311 @item -Wuninitialized
3312 @opindex Wuninitialized
3313 @opindex Wno-uninitialized
3314 Warn if an automatic variable is used without first being initialized
3315 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3316 warn if a non-static reference or non-static @samp{const} member
3317 appears in a class without constructors.
3319 If you want to warn about code which uses the uninitialized value of the
3320 variable in its own initializer, use the @option{-Winit-self} option.
3322 These warnings occur for individual uninitialized or clobbered
3323 elements of structure, union or array variables as well as for
3324 variables which are uninitialized or clobbered as a whole. They do
3325 not occur for variables or elements declared @code{volatile}. Because
3326 these warnings depend on optimization, the exact variables or elements
3327 for which there are warnings will depend on the precise optimization
3328 options and version of GCC used.
3330 Note that there may be no warning about a variable that is used only
3331 to compute a value that itself is never used, because such
3332 computations may be deleted by data flow analysis before the warnings
3335 These warnings are made optional because GCC is not smart
3336 enough to see all the reasons why the code might be correct
3337 despite appearing to have an error. Here is one example of how
3358 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3359 always initialized, but GCC doesn't know this. Here is
3360 another common case:
3365 if (change_y) save_y = y, y = new_y;
3367 if (change_y) y = save_y;
3372 This has no bug because @code{save_y} is used only if it is set.
3374 @cindex @code{longjmp} warnings
3375 This option also warns when a non-volatile automatic variable might be
3376 changed by a call to @code{longjmp}. These warnings as well are possible
3377 only in optimizing compilation.
3379 The compiler sees only the calls to @code{setjmp}. It cannot know
3380 where @code{longjmp} will be called; in fact, a signal handler could
3381 call it at any point in the code. As a result, you may get a warning
3382 even when there is in fact no problem because @code{longjmp} cannot
3383 in fact be called at the place which would cause a problem.
3385 Some spurious warnings can be avoided if you declare all the functions
3386 you use that never return as @code{noreturn}. @xref{Function
3389 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3391 @item -Wunknown-pragmas
3392 @opindex Wunknown-pragmas
3393 @opindex Wno-unknown-pragmas
3394 @cindex warning for unknown pragmas
3395 @cindex unknown pragmas, warning
3396 @cindex pragmas, warning of unknown
3397 Warn when a #pragma directive is encountered which is not understood by
3398 GCC@. If this command line option is used, warnings will even be issued
3399 for unknown pragmas in system header files. This is not the case if
3400 the warnings were only enabled by the @option{-Wall} command line option.
3403 @opindex Wno-pragmas
3405 Do not warn about misuses of pragmas, such as incorrect parameters,
3406 invalid syntax, or conflicts between pragmas. See also
3407 @samp{-Wunknown-pragmas}.
3409 @item -Wstrict-aliasing
3410 @opindex Wstrict-aliasing
3411 @opindex Wno-strict-aliasing
3412 This option is only active when @option{-fstrict-aliasing} is active.
3413 It warns about code which might break the strict aliasing rules that the
3414 compiler is using for optimization. The warning does not catch all
3415 cases, but does attempt to catch the more common pitfalls. It is
3416 included in @option{-Wall}.
3417 It is equivalent to @option{-Wstrict-aliasing=3}
3419 @item -Wstrict-aliasing=n
3420 @opindex Wstrict-aliasing=n
3421 @opindex Wno-strict-aliasing=n
3422 This option is only active when @option{-fstrict-aliasing} is active.
3423 It warns about code which might break the strict aliasing rules that the
3424 compiler is using for optimization.
3425 Higher levels correspond to higher accuracy (fewer false positives).
3426 Higher levels also correspond to more effort, similar to the way -O works.
3427 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3430 Level 1: Most aggressive, quick, least accurate.
3431 Possibly useful when higher levels
3432 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3433 false negatives. However, it has many false positives.
3434 Warns for all pointer conversions between possibly incompatible types,
3435 even if never dereferenced. Runs in the frontend only.
3437 Level 2: Aggressive, quick, not too precise.
3438 May still have many false positives (not as many as level 1 though),
3439 and few false negatives (but possibly more than level 1).
3440 Unlike level 1, it only warns when an address is taken. Warns about
3441 incomplete types. Runs in the frontend only.
3443 Level 3 (default for @option{-Wstrict-aliasing}):
3444 Should have very few false positives and few false
3445 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3446 Takes care of the common punn+dereference pattern in the frontend:
3447 @code{*(int*)&some_float}.
3448 If optimization is enabled, it also runs in the backend, where it deals
3449 with multiple statement cases using flow-sensitive points-to information.
3450 Only warns when the converted pointer is dereferenced.
3451 Does not warn about incomplete types.
3453 @item -Wstrict-overflow
3454 @itemx -Wstrict-overflow=@var{n}
3455 @opindex Wstrict-overflow
3456 @opindex Wno-strict-overflow
3457 This option is only active when @option{-fstrict-overflow} is active.
3458 It warns about cases where the compiler optimizes based on the
3459 assumption that signed overflow does not occur. Note that it does not
3460 warn about all cases where the code might overflow: it only warns
3461 about cases where the compiler implements some optimization. Thus
3462 this warning depends on the optimization level.
3464 An optimization which assumes that signed overflow does not occur is
3465 perfectly safe if the values of the variables involved are such that
3466 overflow never does, in fact, occur. Therefore this warning can
3467 easily give a false positive: a warning about code which is not
3468 actually a problem. To help focus on important issues, several
3469 warning levels are defined. No warnings are issued for the use of
3470 undefined signed overflow when estimating how many iterations a loop
3471 will require, in particular when determining whether a loop will be
3475 @item -Wstrict-overflow=1
3476 Warn about cases which are both questionable and easy to avoid. For
3477 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3478 compiler will simplify this to @code{1}. This level of
3479 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3480 are not, and must be explicitly requested.
3482 @item -Wstrict-overflow=2
3483 Also warn about other cases where a comparison is simplified to a
3484 constant. For example: @code{abs (x) >= 0}. This can only be
3485 simplified when @option{-fstrict-overflow} is in effect, because
3486 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3487 zero. @option{-Wstrict-overflow} (with no level) is the same as
3488 @option{-Wstrict-overflow=2}.
3490 @item -Wstrict-overflow=3
3491 Also warn about other cases where a comparison is simplified. For
3492 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3494 @item -Wstrict-overflow=4
3495 Also warn about other simplifications not covered by the above cases.
3496 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3498 @item -Wstrict-overflow=5
3499 Also warn about cases where the compiler reduces the magnitude of a
3500 constant involved in a comparison. For example: @code{x + 2 > y} will
3501 be simplified to @code{x + 1 >= y}. This is reported only at the
3502 highest warning level because this simplification applies to many
3503 comparisons, so this warning level will give a very large number of
3507 @item -Warray-bounds
3508 @opindex Wno-array-bounds
3509 @opindex Warray-bounds
3510 This option is only active when @option{-ftree-vrp} is active
3511 (default for -O2 and above). It warns about subscripts to arrays
3512 that are always out of bounds. This warning is enabled by @option{-Wall}.
3514 @item -Wno-div-by-zero
3515 @opindex Wno-div-by-zero
3516 @opindex Wdiv-by-zero
3517 Do not warn about compile-time integer division by zero. Floating point
3518 division by zero is not warned about, as it can be a legitimate way of
3519 obtaining infinities and NaNs.
3521 @item -Wsystem-headers
3522 @opindex Wsystem-headers
3523 @opindex Wno-system-headers
3524 @cindex warnings from system headers
3525 @cindex system headers, warnings from
3526 Print warning messages for constructs found in system header files.
3527 Warnings from system headers are normally suppressed, on the assumption
3528 that they usually do not indicate real problems and would only make the
3529 compiler output harder to read. Using this command line option tells
3530 GCC to emit warnings from system headers as if they occurred in user
3531 code. However, note that using @option{-Wall} in conjunction with this
3532 option will @emph{not} warn about unknown pragmas in system
3533 headers---for that, @option{-Wunknown-pragmas} must also be used.
3536 @opindex Wfloat-equal
3537 @opindex Wno-float-equal
3538 Warn if floating point values are used in equality comparisons.
3540 The idea behind this is that sometimes it is convenient (for the
3541 programmer) to consider floating-point values as approximations to
3542 infinitely precise real numbers. If you are doing this, then you need
3543 to compute (by analyzing the code, or in some other way) the maximum or
3544 likely maximum error that the computation introduces, and allow for it
3545 when performing comparisons (and when producing output, but that's a
3546 different problem). In particular, instead of testing for equality, you
3547 would check to see whether the two values have ranges that overlap; and
3548 this is done with the relational operators, so equality comparisons are
3551 @item -Wtraditional @r{(C and Objective-C only)}
3552 @opindex Wtraditional
3553 @opindex Wno-traditional
3554 Warn about certain constructs that behave differently in traditional and
3555 ISO C@. Also warn about ISO C constructs that have no traditional C
3556 equivalent, and/or problematic constructs which should be avoided.
3560 Macro parameters that appear within string literals in the macro body.
3561 In traditional C macro replacement takes place within string literals,
3562 but does not in ISO C@.
3565 In traditional C, some preprocessor directives did not exist.
3566 Traditional preprocessors would only consider a line to be a directive
3567 if the @samp{#} appeared in column 1 on the line. Therefore
3568 @option{-Wtraditional} warns about directives that traditional C
3569 understands but would ignore because the @samp{#} does not appear as the
3570 first character on the line. It also suggests you hide directives like
3571 @samp{#pragma} not understood by traditional C by indenting them. Some
3572 traditional implementations would not recognize @samp{#elif}, so it
3573 suggests avoiding it altogether.
3576 A function-like macro that appears without arguments.
3579 The unary plus operator.
3582 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3583 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3584 constants.) Note, these suffixes appear in macros defined in the system
3585 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3586 Use of these macros in user code might normally lead to spurious
3587 warnings, however GCC's integrated preprocessor has enough context to
3588 avoid warning in these cases.
3591 A function declared external in one block and then used after the end of
3595 A @code{switch} statement has an operand of type @code{long}.
3598 A non-@code{static} function declaration follows a @code{static} one.
3599 This construct is not accepted by some traditional C compilers.
3602 The ISO type of an integer constant has a different width or
3603 signedness from its traditional type. This warning is only issued if
3604 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3605 typically represent bit patterns, are not warned about.
3608 Usage of ISO string concatenation is detected.
3611 Initialization of automatic aggregates.
3614 Identifier conflicts with labels. Traditional C lacks a separate
3615 namespace for labels.
3618 Initialization of unions. If the initializer is zero, the warning is
3619 omitted. This is done under the assumption that the zero initializer in
3620 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3621 initializer warnings and relies on default initialization to zero in the
3625 Conversions by prototypes between fixed/floating point values and vice
3626 versa. The absence of these prototypes when compiling with traditional
3627 C would cause serious problems. This is a subset of the possible
3628 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3631 Use of ISO C style function definitions. This warning intentionally is
3632 @emph{not} issued for prototype declarations or variadic functions
3633 because these ISO C features will appear in your code when using
3634 libiberty's traditional C compatibility macros, @code{PARAMS} and
3635 @code{VPARAMS}. This warning is also bypassed for nested functions
3636 because that feature is already a GCC extension and thus not relevant to
3637 traditional C compatibility.
3640 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3641 @opindex Wtraditional-conversion
3642 @opindex Wno-traditional-conversion
3643 Warn if a prototype causes a type conversion that is different from what
3644 would happen to the same argument in the absence of a prototype. This
3645 includes conversions of fixed point to floating and vice versa, and
3646 conversions changing the width or signedness of a fixed point argument
3647 except when the same as the default promotion.
3649 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3650 @opindex Wdeclaration-after-statement
3651 @opindex Wno-declaration-after-statement
3652 Warn when a declaration is found after a statement in a block. This
3653 construct, known from C++, was introduced with ISO C99 and is by default
3654 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3655 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3660 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3662 @item -Wno-endif-labels
3663 @opindex Wno-endif-labels
3664 @opindex Wendif-labels
3665 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3670 Warn whenever a local variable shadows another local variable, parameter or
3671 global variable or whenever a built-in function is shadowed.
3673 @item -Wlarger-than=@var{len}
3674 @opindex Wlarger-than=@var{len}
3675 @opindex Wlarger-than-@var{len}
3676 Warn whenever an object of larger than @var{len} bytes is defined.
3678 @item -Wframe-larger-than=@var{len}
3679 @opindex Wframe-larger-than
3680 Warn if the size of a function frame is larger than @var{len} bytes.
3681 The computation done to determine the stack frame size is approximate
3682 and not conservative.
3683 The actual requirements may be somewhat greater than @var{len}
3684 even if you do not get a warning. In addition, any space allocated
3685 via @code{alloca}, variable-length arrays, or related constructs
3686 is not included by the compiler when determining
3687 whether or not to issue a warning.
3689 @item -Wunsafe-loop-optimizations
3690 @opindex Wunsafe-loop-optimizations
3691 @opindex Wno-unsafe-loop-optimizations
3692 Warn if the loop cannot be optimized because the compiler could not
3693 assume anything on the bounds of the loop indices. With
3694 @option{-funsafe-loop-optimizations} warn if the compiler made
3697 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
3698 @opindex Wno-pedantic-ms-format
3699 @opindex Wpedantic-ms-format
3700 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
3701 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
3702 depending on the MS runtime, when you are using the options @option{-Wformat}
3703 and @option{-pedantic} without gnu-extensions.
3705 @item -Wpointer-arith
3706 @opindex Wpointer-arith
3707 @opindex Wno-pointer-arith
3708 Warn about anything that depends on the ``size of'' a function type or
3709 of @code{void}. GNU C assigns these types a size of 1, for
3710 convenience in calculations with @code{void *} pointers and pointers
3711 to functions. In C++, warn also when an arithmetic operation involves
3712 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3715 @opindex Wtype-limits
3716 @opindex Wno-type-limits
3717 Warn if a comparison is always true or always false due to the limited
3718 range of the data type, but do not warn for constant expressions. For
3719 example, warn if an unsigned variable is compared against zero with
3720 @samp{<} or @samp{>=}. This warning is also enabled by
3723 @item -Wbad-function-cast @r{(C and Objective-C only)}
3724 @opindex Wbad-function-cast
3725 @opindex Wno-bad-function-cast
3726 Warn whenever a function call is cast to a non-matching type.
3727 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3729 @item -Wc++-compat @r{(C and Objective-C only)}
3730 Warn about ISO C constructs that are outside of the common subset of
3731 ISO C and ISO C++, e.g.@: request for implicit conversion from
3732 @code{void *} to a pointer to non-@code{void} type.
3734 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3735 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3736 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3737 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3741 @opindex Wno-cast-qual
3742 Warn whenever a pointer is cast so as to remove a type qualifier from
3743 the target type. For example, warn if a @code{const char *} is cast
3744 to an ordinary @code{char *}.
3746 Also warn when making a cast which introduces a type qualifier in an
3747 unsafe way. For example, casting @code{char **} to @code{const char **}
3748 is unsafe, as in this example:
3751 /* p is char ** value. */
3752 const char **q = (const char **) p;
3753 /* Assignment of readonly string to const char * is OK. */
3755 /* Now char** pointer points to read-only memory. */
3760 @opindex Wcast-align
3761 @opindex Wno-cast-align
3762 Warn whenever a pointer is cast such that the required alignment of the
3763 target is increased. For example, warn if a @code{char *} is cast to
3764 an @code{int *} on machines where integers can only be accessed at
3765 two- or four-byte boundaries.
3767 @item -Wwrite-strings
3768 @opindex Wwrite-strings
3769 @opindex Wno-write-strings
3770 When compiling C, give string constants the type @code{const
3771 char[@var{length}]} so that copying the address of one into a
3772 non-@code{const} @code{char *} pointer will get a warning. These
3773 warnings will help you find at compile time code that can try to write
3774 into a string constant, but only if you have been very careful about
3775 using @code{const} in declarations and prototypes. Otherwise, it will
3776 just be a nuisance. This is why we did not make @option{-Wall} request
3779 When compiling C++, warn about the deprecated conversion from string
3780 literals to @code{char *}. This warning is enabled by default for C++
3785 @opindex Wno-clobbered
3786 Warn for variables that might be changed by @samp{longjmp} or
3787 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
3790 @opindex Wconversion
3791 @opindex Wno-conversion
3792 Warn for implicit conversions that may alter a value. This includes
3793 conversions between real and integer, like @code{abs (x)} when
3794 @code{x} is @code{double}; conversions between signed and unsigned,
3795 like @code{unsigned ui = -1}; and conversions to smaller types, like
3796 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
3797 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
3798 changed by the conversion like in @code{abs (2.0)}. Warnings about
3799 conversions between signed and unsigned integers can be disabled by
3800 using @option{-Wno-sign-conversion}.
3802 For C++, also warn for conversions between @code{NULL} and non-pointer
3803 types; confusing overload resolution for user-defined conversions; and
3804 conversions that will never use a type conversion operator:
3805 conversions to @code{void}, the same type, a base class or a reference
3806 to them. Warnings about conversions between signed and unsigned
3807 integers are disabled by default in C++ unless
3808 @option{-Wsign-conversion} is explicitly enabled.
3811 @opindex Wempty-body
3812 @opindex Wno-empty-body
3813 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
3814 while} statement. This warning is also enabled by @option{-Wextra}.
3816 @item -Wenum-compare
3817 @opindex Wenum-compare
3818 @opindex Wno-enum-compare
3819 Warn about a comparison between values of different enum types. In C++
3820 this warning is enabled by default. In C this warning is enabled by
3823 @item -Wsign-compare
3824 @opindex Wsign-compare
3825 @opindex Wno-sign-compare
3826 @cindex warning for comparison of signed and unsigned values
3827 @cindex comparison of signed and unsigned values, warning
3828 @cindex signed and unsigned values, comparison warning
3829 Warn when a comparison between signed and unsigned values could produce
3830 an incorrect result when the signed value is converted to unsigned.
3831 This warning is also enabled by @option{-Wextra}; to get the other warnings
3832 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3834 @item -Wsign-conversion
3835 @opindex Wsign-conversion
3836 @opindex Wno-sign-conversion
3837 Warn for implicit conversions that may change the sign of an integer
3838 value, like assigning a signed integer expression to an unsigned
3839 integer variable. An explicit cast silences the warning. In C, this
3840 option is enabled also by @option{-Wconversion}.
3844 @opindex Wno-address
3845 Warn about suspicious uses of memory addresses. These include using
3846 the address of a function in a conditional expression, such as
3847 @code{void func(void); if (func)}, and comparisons against the memory
3848 address of a string literal, such as @code{if (x == "abc")}. Such
3849 uses typically indicate a programmer error: the address of a function
3850 always evaluates to true, so their use in a conditional usually
3851 indicate that the programmer forgot the parentheses in a function
3852 call; and comparisons against string literals result in unspecified
3853 behavior and are not portable in C, so they usually indicate that the
3854 programmer intended to use @code{strcmp}. This warning is enabled by
3858 @opindex Wlogical-op
3859 @opindex Wno-logical-op
3860 Warn about suspicious uses of logical operators in expressions.
3861 This includes using logical operators in contexts where a
3862 bit-wise operator is likely to be expected.
3864 @item -Waggregate-return
3865 @opindex Waggregate-return
3866 @opindex Wno-aggregate-return
3867 Warn if any functions that return structures or unions are defined or
3868 called. (In languages where you can return an array, this also elicits
3871 @item -Wno-attributes
3872 @opindex Wno-attributes
3873 @opindex Wattributes
3874 Do not warn if an unexpected @code{__attribute__} is used, such as
3875 unrecognized attributes, function attributes applied to variables,
3876 etc. This will not stop errors for incorrect use of supported
3879 @item -Wno-builtin-macro-redefined
3880 @opindex Wno-builtin-macro-redefined
3881 @opindex Wbuiltin-macro-redefined
3882 Do not warn if certain built-in macros are redefined. This suppresses
3883 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
3884 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
3886 @item -Wstrict-prototypes @r{(C and Objective-C only)}
3887 @opindex Wstrict-prototypes
3888 @opindex Wno-strict-prototypes
3889 Warn if a function is declared or defined without specifying the
3890 argument types. (An old-style function definition is permitted without
3891 a warning if preceded by a declaration which specifies the argument
3894 @item -Wold-style-declaration @r{(C and Objective-C only)}
3895 @opindex Wold-style-declaration
3896 @opindex Wno-old-style-declaration
3897 Warn for obsolescent usages, according to the C Standard, in a
3898 declaration. For example, warn if storage-class specifiers like
3899 @code{static} are not the first things in a declaration. This warning
3900 is also enabled by @option{-Wextra}.
3902 @item -Wold-style-definition @r{(C and Objective-C only)}
3903 @opindex Wold-style-definition
3904 @opindex Wno-old-style-definition
3905 Warn if an old-style function definition is used. A warning is given
3906 even if there is a previous prototype.
3908 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
3909 @opindex Wmissing-parameter-type
3910 @opindex Wno-missing-parameter-type
3911 A function parameter is declared without a type specifier in K&R-style
3918 This warning is also enabled by @option{-Wextra}.
3920 @item -Wmissing-prototypes @r{(C and Objective-C only)}
3921 @opindex Wmissing-prototypes
3922 @opindex Wno-missing-prototypes
3923 Warn if a global function is defined without a previous prototype
3924 declaration. This warning is issued even if the definition itself
3925 provides a prototype. The aim is to detect global functions that fail
3926 to be declared in header files.
3928 @item -Wmissing-declarations
3929 @opindex Wmissing-declarations
3930 @opindex Wno-missing-declarations
3931 Warn if a global function is defined without a previous declaration.
3932 Do so even if the definition itself provides a prototype.
3933 Use this option to detect global functions that are not declared in
3934 header files. In C++, no warnings are issued for function templates,
3935 or for inline functions, or for functions in anonymous namespaces.
3937 @item -Wmissing-field-initializers
3938 @opindex Wmissing-field-initializers
3939 @opindex Wno-missing-field-initializers
3943 Warn if a structure's initializer has some fields missing. For
3944 example, the following code would cause such a warning, because
3945 @code{x.h} is implicitly zero:
3948 struct s @{ int f, g, h; @};
3949 struct s x = @{ 3, 4 @};
3952 This option does not warn about designated initializers, so the following
3953 modification would not trigger a warning:
3956 struct s @{ int f, g, h; @};
3957 struct s x = @{ .f = 3, .g = 4 @};
3960 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3961 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3963 @item -Wmissing-noreturn
3964 @opindex Wmissing-noreturn
3965 @opindex Wno-missing-noreturn
3966 Warn about functions which might be candidates for attribute @code{noreturn}.
3967 Note these are only possible candidates, not absolute ones. Care should
3968 be taken to manually verify functions actually do not ever return before
3969 adding the @code{noreturn} attribute, otherwise subtle code generation
3970 bugs could be introduced. You will not get a warning for @code{main} in
3971 hosted C environments.
3973 @item -Wmissing-format-attribute
3974 @opindex Wmissing-format-attribute
3975 @opindex Wno-missing-format-attribute
3978 Warn about function pointers which might be candidates for @code{format}
3979 attributes. Note these are only possible candidates, not absolute ones.
3980 GCC will guess that function pointers with @code{format} attributes that
3981 are used in assignment, initialization, parameter passing or return
3982 statements should have a corresponding @code{format} attribute in the
3983 resulting type. I.e.@: the left-hand side of the assignment or
3984 initialization, the type of the parameter variable, or the return type
3985 of the containing function respectively should also have a @code{format}
3986 attribute to avoid the warning.
3988 GCC will also warn about function definitions which might be
3989 candidates for @code{format} attributes. Again, these are only
3990 possible candidates. GCC will guess that @code{format} attributes
3991 might be appropriate for any function that calls a function like
3992 @code{vprintf} or @code{vscanf}, but this might not always be the
3993 case, and some functions for which @code{format} attributes are
3994 appropriate may not be detected.
3996 @item -Wno-multichar
3997 @opindex Wno-multichar
3999 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4000 Usually they indicate a typo in the user's code, as they have
4001 implementation-defined values, and should not be used in portable code.
4003 @item -Wnormalized=<none|id|nfc|nfkc>
4004 @opindex Wnormalized=
4007 @cindex character set, input normalization
4008 In ISO C and ISO C++, two identifiers are different if they are
4009 different sequences of characters. However, sometimes when characters
4010 outside the basic ASCII character set are used, you can have two
4011 different character sequences that look the same. To avoid confusion,
4012 the ISO 10646 standard sets out some @dfn{normalization rules} which
4013 when applied ensure that two sequences that look the same are turned into
4014 the same sequence. GCC can warn you if you are using identifiers which
4015 have not been normalized; this option controls that warning.
4017 There are four levels of warning that GCC supports. The default is
4018 @option{-Wnormalized=nfc}, which warns about any identifier which is
4019 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4020 recommended form for most uses.
4022 Unfortunately, there are some characters which ISO C and ISO C++ allow
4023 in identifiers that when turned into NFC aren't allowable as
4024 identifiers. That is, there's no way to use these symbols in portable
4025 ISO C or C++ and have all your identifiers in NFC@.
4026 @option{-Wnormalized=id} suppresses the warning for these characters.
4027 It is hoped that future versions of the standards involved will correct
4028 this, which is why this option is not the default.
4030 You can switch the warning off for all characters by writing
4031 @option{-Wnormalized=none}. You would only want to do this if you
4032 were using some other normalization scheme (like ``D''), because
4033 otherwise you can easily create bugs that are literally impossible to see.
4035 Some characters in ISO 10646 have distinct meanings but look identical
4036 in some fonts or display methodologies, especially once formatting has
4037 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4038 LETTER N'', will display just like a regular @code{n} which has been
4039 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4040 normalization scheme to convert all these into a standard form as
4041 well, and GCC will warn if your code is not in NFKC if you use
4042 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4043 about every identifier that contains the letter O because it might be
4044 confused with the digit 0, and so is not the default, but may be
4045 useful as a local coding convention if the programming environment is
4046 unable to be fixed to display these characters distinctly.
4048 @item -Wno-deprecated
4049 @opindex Wno-deprecated
4050 @opindex Wdeprecated
4051 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4053 @item -Wno-deprecated-declarations
4054 @opindex Wno-deprecated-declarations
4055 @opindex Wdeprecated-declarations
4056 Do not warn about uses of functions (@pxref{Function Attributes}),
4057 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4058 Attributes}) marked as deprecated by using the @code{deprecated}
4062 @opindex Wno-overflow
4064 Do not warn about compile-time overflow in constant expressions.
4066 @item -Woverride-init @r{(C and Objective-C only)}
4067 @opindex Woverride-init
4068 @opindex Wno-override-init
4072 Warn if an initialized field without side effects is overridden when
4073 using designated initializers (@pxref{Designated Inits, , Designated
4076 This warning is included in @option{-Wextra}. To get other
4077 @option{-Wextra} warnings without this one, use @samp{-Wextra
4078 -Wno-override-init}.
4083 Warn if a structure is given the packed attribute, but the packed
4084 attribute has no effect on the layout or size of the structure.
4085 Such structures may be mis-aligned for little benefit. For
4086 instance, in this code, the variable @code{f.x} in @code{struct bar}
4087 will be misaligned even though @code{struct bar} does not itself
4088 have the packed attribute:
4095 @} __attribute__((packed));
4103 @item -Wpacked-bitfield-compat
4104 @opindex Wpacked-bitfield-compat
4105 @opindex Wno-packed-bitfield-compat
4106 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4107 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4108 the change can lead to differences in the structure layout. GCC
4109 informs you when the offset of such a field has changed in GCC 4.4.
4110 For example there is no longer a 4-bit padding between field @code{a}
4111 and @code{b} in this structure:
4118 @} __attribute__ ((packed));
4121 This warning is enabled by default. Use
4122 @option{-Wno-packed-bitfield-compat} to disable this warning.
4127 Warn if padding is included in a structure, either to align an element
4128 of the structure or to align the whole structure. Sometimes when this
4129 happens it is possible to rearrange the fields of the structure to
4130 reduce the padding and so make the structure smaller.
4132 @item -Wredundant-decls
4133 @opindex Wredundant-decls
4134 @opindex Wno-redundant-decls
4135 Warn if anything is declared more than once in the same scope, even in
4136 cases where multiple declaration is valid and changes nothing.
4138 @item -Wnested-externs @r{(C and Objective-C only)}
4139 @opindex Wnested-externs
4140 @opindex Wno-nested-externs
4141 Warn if an @code{extern} declaration is encountered within a function.
4143 @item -Wunreachable-code
4144 @opindex Wunreachable-code
4145 @opindex Wno-unreachable-code
4146 Warn if the compiler detects that code will never be executed.
4148 This option is intended to warn when the compiler detects that at
4149 least a whole line of source code will never be executed, because
4150 some condition is never satisfied or because it is after a
4151 procedure that never returns.
4153 It is possible for this option to produce a warning even though there
4154 are circumstances under which part of the affected line can be executed,
4155 so care should be taken when removing apparently-unreachable code.
4157 For instance, when a function is inlined, a warning may mean that the
4158 line is unreachable in only one inlined copy of the function.
4160 This option is not made part of @option{-Wall} because in a debugging
4161 version of a program there is often substantial code which checks
4162 correct functioning of the program and is, hopefully, unreachable
4163 because the program does work. Another common use of unreachable
4164 code is to provide behavior which is selectable at compile-time.
4169 Warn if a function can not be inlined and it was declared as inline.
4170 Even with this option, the compiler will not warn about failures to
4171 inline functions declared in system headers.
4173 The compiler uses a variety of heuristics to determine whether or not
4174 to inline a function. For example, the compiler takes into account
4175 the size of the function being inlined and the amount of inlining
4176 that has already been done in the current function. Therefore,
4177 seemingly insignificant changes in the source program can cause the
4178 warnings produced by @option{-Winline} to appear or disappear.
4180 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4181 @opindex Wno-invalid-offsetof
4182 @opindex Winvalid-offsetof
4183 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4184 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4185 to a non-POD type is undefined. In existing C++ implementations,
4186 however, @samp{offsetof} typically gives meaningful results even when
4187 applied to certain kinds of non-POD types. (Such as a simple
4188 @samp{struct} that fails to be a POD type only by virtue of having a
4189 constructor.) This flag is for users who are aware that they are
4190 writing nonportable code and who have deliberately chosen to ignore the
4193 The restrictions on @samp{offsetof} may be relaxed in a future version
4194 of the C++ standard.
4196 @item -Wno-int-to-pointer-cast @r{(C and Objective-C only)}
4197 @opindex Wno-int-to-pointer-cast
4198 @opindex Wint-to-pointer-cast
4199 Suppress warnings from casts to pointer type of an integer of a
4202 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4203 @opindex Wno-pointer-to-int-cast
4204 @opindex Wpointer-to-int-cast
4205 Suppress warnings from casts from a pointer to an integer type of a
4209 @opindex Winvalid-pch
4210 @opindex Wno-invalid-pch
4211 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4212 the search path but can't be used.
4216 @opindex Wno-long-long
4217 Warn if @samp{long long} type is used. This is enabled by either
4218 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4219 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4221 @item -Wvariadic-macros
4222 @opindex Wvariadic-macros
4223 @opindex Wno-variadic-macros
4224 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4225 alternate syntax when in pedantic ISO C99 mode. This is default.
4226 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4231 Warn if variable length array is used in the code.
4232 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4233 the variable length array.
4235 @item -Wvolatile-register-var
4236 @opindex Wvolatile-register-var
4237 @opindex Wno-volatile-register-var
4238 Warn if a register variable is declared volatile. The volatile
4239 modifier does not inhibit all optimizations that may eliminate reads
4240 and/or writes to register variables. This warning is enabled by
4243 @item -Wdisabled-optimization
4244 @opindex Wdisabled-optimization
4245 @opindex Wno-disabled-optimization
4246 Warn if a requested optimization pass is disabled. This warning does
4247 not generally indicate that there is anything wrong with your code; it
4248 merely indicates that GCC's optimizers were unable to handle the code
4249 effectively. Often, the problem is that your code is too big or too
4250 complex; GCC will refuse to optimize programs when the optimization
4251 itself is likely to take inordinate amounts of time.
4253 @item -Wpointer-sign @r{(C and Objective-C only)}
4254 @opindex Wpointer-sign
4255 @opindex Wno-pointer-sign
4256 Warn for pointer argument passing or assignment with different signedness.
4257 This option is only supported for C and Objective-C@. It is implied by
4258 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4259 @option{-Wno-pointer-sign}.
4261 @item -Wstack-protector
4262 @opindex Wstack-protector
4263 @opindex Wno-stack-protector
4264 This option is only active when @option{-fstack-protector} is active. It
4265 warns about functions that will not be protected against stack smashing.
4268 @opindex Wno-mudflap
4269 Suppress warnings about constructs that cannot be instrumented by
4272 @item -Woverlength-strings
4273 @opindex Woverlength-strings
4274 @opindex Wno-overlength-strings
4275 Warn about string constants which are longer than the ``minimum
4276 maximum'' length specified in the C standard. Modern compilers
4277 generally allow string constants which are much longer than the
4278 standard's minimum limit, but very portable programs should avoid
4279 using longer strings.
4281 The limit applies @emph{after} string constant concatenation, and does
4282 not count the trailing NUL@. In C89, the limit was 509 characters; in
4283 C99, it was raised to 4095. C++98 does not specify a normative
4284 minimum maximum, so we do not diagnose overlength strings in C++@.
4286 This option is implied by @option{-pedantic}, and can be disabled with
4287 @option{-Wno-overlength-strings}.
4289 @item -Wunsuffixed-float-constants
4290 @opindex Wunsuffixed-float-constants
4292 GCC will issue a warning for any floating constant that does not have
4293 a suffix. When used together with @option{-Wsystem-headers} it will
4294 warn about such constants in system header files. This can be useful
4295 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4296 from the decimal floating-point extension to C99.
4299 @node Debugging Options
4300 @section Options for Debugging Your Program or GCC
4301 @cindex options, debugging
4302 @cindex debugging information options
4304 GCC has various special options that are used for debugging
4305 either your program or GCC:
4310 Produce debugging information in the operating system's native format
4311 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4314 On most systems that use stabs format, @option{-g} enables use of extra
4315 debugging information that only GDB can use; this extra information
4316 makes debugging work better in GDB but will probably make other debuggers
4318 refuse to read the program. If you want to control for certain whether
4319 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4320 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4322 GCC allows you to use @option{-g} with
4323 @option{-O}. The shortcuts taken by optimized code may occasionally
4324 produce surprising results: some variables you declared may not exist
4325 at all; flow of control may briefly move where you did not expect it;
4326 some statements may not be executed because they compute constant
4327 results or their values were already at hand; some statements may
4328 execute in different places because they were moved out of loops.
4330 Nevertheless it proves possible to debug optimized output. This makes
4331 it reasonable to use the optimizer for programs that might have bugs.
4333 The following options are useful when GCC is generated with the
4334 capability for more than one debugging format.
4338 Produce debugging information for use by GDB@. This means to use the
4339 most expressive format available (DWARF 2, stabs, or the native format
4340 if neither of those are supported), including GDB extensions if at all
4345 Produce debugging information in stabs format (if that is supported),
4346 without GDB extensions. This is the format used by DBX on most BSD
4347 systems. On MIPS, Alpha and System V Release 4 systems this option
4348 produces stabs debugging output which is not understood by DBX or SDB@.
4349 On System V Release 4 systems this option requires the GNU assembler.
4351 @item -feliminate-unused-debug-symbols
4352 @opindex feliminate-unused-debug-symbols
4353 Produce debugging information in stabs format (if that is supported),
4354 for only symbols that are actually used.
4356 @item -femit-class-debug-always
4357 Instead of emitting debugging information for a C++ class in only one
4358 object file, emit it in all object files using the class. This option
4359 should be used only with debuggers that are unable to handle the way GCC
4360 normally emits debugging information for classes because using this
4361 option will increase the size of debugging information by as much as a
4366 Produce debugging information in stabs format (if that is supported),
4367 using GNU extensions understood only by the GNU debugger (GDB)@. The
4368 use of these extensions is likely to make other debuggers crash or
4369 refuse to read the program.
4373 Produce debugging information in COFF format (if that is supported).
4374 This is the format used by SDB on most System V systems prior to
4379 Produce debugging information in XCOFF format (if that is supported).
4380 This is the format used by the DBX debugger on IBM RS/6000 systems.
4384 Produce debugging information in XCOFF format (if that is supported),
4385 using GNU extensions understood only by the GNU debugger (GDB)@. The
4386 use of these extensions is likely to make other debuggers crash or
4387 refuse to read the program, and may cause assemblers other than the GNU
4388 assembler (GAS) to fail with an error.
4390 @item -gdwarf-@var{version}
4391 @opindex gdwarf-@var{version}
4392 Produce debugging information in DWARF format (if that is
4393 supported). This is the format used by DBX on IRIX 6. The value
4394 of @var{version} may be either 2 or 3; the default version is 2.
4396 Note that with DWARF version 2 some ports require, and will always
4397 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4401 Produce debugging information in VMS debug format (if that is
4402 supported). This is the format used by DEBUG on VMS systems.
4405 @itemx -ggdb@var{level}
4406 @itemx -gstabs@var{level}
4407 @itemx -gcoff@var{level}
4408 @itemx -gxcoff@var{level}
4409 @itemx -gvms@var{level}
4410 Request debugging information and also use @var{level} to specify how
4411 much information. The default level is 2.
4413 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4416 Level 1 produces minimal information, enough for making backtraces in
4417 parts of the program that you don't plan to debug. This includes
4418 descriptions of functions and external variables, but no information
4419 about local variables and no line numbers.
4421 Level 3 includes extra information, such as all the macro definitions
4422 present in the program. Some debuggers support macro expansion when
4423 you use @option{-g3}.
4425 @option{-gdwarf-2} does not accept a concatenated debug level, because
4426 GCC used to support an option @option{-gdwarf} that meant to generate
4427 debug information in version 1 of the DWARF format (which is very
4428 different from version 2), and it would have been too confusing. That
4429 debug format is long obsolete, but the option cannot be changed now.
4430 Instead use an additional @option{-g@var{level}} option to change the
4431 debug level for DWARF.
4435 Turn off generation of debug info, if leaving out this option would have
4436 generated it, or turn it on at level 2 otherwise. The position of this
4437 argument in the command line does not matter, it takes effect after all
4438 other options are processed, and it does so only once, no matter how
4439 many times it is given. This is mainly intended to be used with
4440 @option{-fcompare-debug}.
4442 @item -fdump-final-insns=@var{file}
4443 @opindex fdump-final-insns=
4444 Dump the final internal representation (RTL) to @var{file}.
4446 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4447 @opindex fcompare-debug
4448 @opindex fno-compare-debug
4449 If no error occurs during compilation, run the compiler a second time,
4450 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4451 passed to the second compilation. Dump the final internal
4452 representation in both compilations, and print an error if they differ.
4454 If the equal sign is omitted, the default @option{-gtoggle} is used.
4456 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4457 and nonzero, implicitly enables @option{-fcompare-debug}. If
4458 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4459 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4462 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4463 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4464 of the final representation and the second compilation, preventing even
4465 @env{GCC_COMPARE_DEBUG} from taking effect.
4467 To verify full coverage during @option{-fcompare-debug} testing, set
4468 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4469 which GCC will reject as an invalid option in any actual compilation
4470 (rather than preprocessing, assembly or linking). To get just a
4471 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4472 not overridden} will do.
4474 @item -fcompare-debug-second
4475 @opindex fcompare-debug-second
4476 This option is implicitly passed to the compiler for the second
4477 compilation requested by @option{-fcompare-debug}, along with options to
4478 silence warnings, and omitting other options that would cause
4479 side-effect compiler outputs to files or to the standard output. Dump
4480 files and preserved temporary files are renamed so as to contain the
4481 @code{.gk} additional extension during the second compilation, to avoid
4482 overwriting those generated by the first.
4484 When this option is passed to the compiler driver, it causes the
4485 @emph{first} compilation to be skipped, which makes it useful for little
4486 other than debugging the compiler proper.
4488 @item -feliminate-dwarf2-dups
4489 @opindex feliminate-dwarf2-dups
4490 Compress DWARF2 debugging information by eliminating duplicated
4491 information about each symbol. This option only makes sense when
4492 generating DWARF2 debugging information with @option{-gdwarf-2}.
4494 @item -femit-struct-debug-baseonly
4495 Emit debug information for struct-like types
4496 only when the base name of the compilation source file
4497 matches the base name of file in which the struct was defined.
4499 This option substantially reduces the size of debugging information,
4500 but at significant potential loss in type information to the debugger.
4501 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4502 See @option{-femit-struct-debug-detailed} for more detailed control.
4504 This option works only with DWARF 2.
4506 @item -femit-struct-debug-reduced
4507 Emit debug information for struct-like types
4508 only when the base name of the compilation source file
4509 matches the base name of file in which the type was defined,
4510 unless the struct is a template or defined in a system header.
4512 This option significantly reduces the size of debugging information,
4513 with some potential loss in type information to the debugger.
4514 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4515 See @option{-femit-struct-debug-detailed} for more detailed control.
4517 This option works only with DWARF 2.
4519 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4520 Specify the struct-like types
4521 for which the compiler will generate debug information.
4522 The intent is to reduce duplicate struct debug information
4523 between different object files within the same program.
4525 This option is a detailed version of
4526 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4527 which will serve for most needs.
4529 A specification has the syntax
4530 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4532 The optional first word limits the specification to
4533 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4534 A struct type is used directly when it is the type of a variable, member.
4535 Indirect uses arise through pointers to structs.
4536 That is, when use of an incomplete struct would be legal, the use is indirect.
4538 @samp{struct one direct; struct two * indirect;}.
4540 The optional second word limits the specification to
4541 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4542 Generic structs are a bit complicated to explain.
4543 For C++, these are non-explicit specializations of template classes,
4544 or non-template classes within the above.
4545 Other programming languages have generics,
4546 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4548 The third word specifies the source files for those
4549 structs for which the compiler will emit debug information.
4550 The values @samp{none} and @samp{any} have the normal meaning.
4551 The value @samp{base} means that
4552 the base of name of the file in which the type declaration appears
4553 must match the base of the name of the main compilation file.
4554 In practice, this means that
4555 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4556 but types declared in other header will not.
4557 The value @samp{sys} means those types satisfying @samp{base}
4558 or declared in system or compiler headers.
4560 You may need to experiment to determine the best settings for your application.
4562 The default is @samp{-femit-struct-debug-detailed=all}.
4564 This option works only with DWARF 2.
4566 @item -fno-merge-debug-strings
4567 @opindex fmerge-debug-strings
4568 @opindex fno-merge-debug-strings
4569 Direct the linker to not merge together strings in the debugging
4570 information which are identical in different object files. Merging is
4571 not supported by all assemblers or linkers. Merging decreases the size
4572 of the debug information in the output file at the cost of increasing
4573 link processing time. Merging is enabled by default.
4575 @item -fdebug-prefix-map=@var{old}=@var{new}
4576 @opindex fdebug-prefix-map
4577 When compiling files in directory @file{@var{old}}, record debugging
4578 information describing them as in @file{@var{new}} instead.
4580 @item -fno-dwarf2-cfi-asm
4581 @opindex fdwarf2-cfi-asm
4582 @opindex fno-dwarf2-cfi-asm
4583 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4584 instead of using GAS @code{.cfi_*} directives.
4586 @cindex @command{prof}
4589 Generate extra code to write profile information suitable for the
4590 analysis program @command{prof}. You must use this option when compiling
4591 the source files you want data about, and you must also use it when
4594 @cindex @command{gprof}
4597 Generate extra code to write profile information suitable for the
4598 analysis program @command{gprof}. You must use this option when compiling
4599 the source files you want data about, and you must also use it when
4604 Makes the compiler print out each function name as it is compiled, and
4605 print some statistics about each pass when it finishes.
4608 @opindex ftime-report
4609 Makes the compiler print some statistics about the time consumed by each
4610 pass when it finishes.
4613 @opindex fmem-report
4614 Makes the compiler print some statistics about permanent memory
4615 allocation when it finishes.
4617 @item -fpre-ipa-mem-report
4618 @opindex fpre-ipa-mem-report
4619 @item -fpost-ipa-mem-report
4620 @opindex fpost-ipa-mem-report
4621 Makes the compiler print some statistics about permanent memory
4622 allocation before or after interprocedural optimization.
4624 @item -fprofile-arcs
4625 @opindex fprofile-arcs
4626 Add code so that program flow @dfn{arcs} are instrumented. During
4627 execution the program records how many times each branch and call is
4628 executed and how many times it is taken or returns. When the compiled
4629 program exits it saves this data to a file called
4630 @file{@var{auxname}.gcda} for each source file. The data may be used for
4631 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4632 test coverage analysis (@option{-ftest-coverage}). Each object file's
4633 @var{auxname} is generated from the name of the output file, if
4634 explicitly specified and it is not the final executable, otherwise it is
4635 the basename of the source file. In both cases any suffix is removed
4636 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4637 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4638 @xref{Cross-profiling}.
4640 @cindex @command{gcov}
4644 This option is used to compile and link code instrumented for coverage
4645 analysis. The option is a synonym for @option{-fprofile-arcs}
4646 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4647 linking). See the documentation for those options for more details.
4652 Compile the source files with @option{-fprofile-arcs} plus optimization
4653 and code generation options. For test coverage analysis, use the
4654 additional @option{-ftest-coverage} option. You do not need to profile
4655 every source file in a program.
4658 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4659 (the latter implies the former).
4662 Run the program on a representative workload to generate the arc profile
4663 information. This may be repeated any number of times. You can run
4664 concurrent instances of your program, and provided that the file system
4665 supports locking, the data files will be correctly updated. Also
4666 @code{fork} calls are detected and correctly handled (double counting
4670 For profile-directed optimizations, compile the source files again with
4671 the same optimization and code generation options plus
4672 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4673 Control Optimization}).
4676 For test coverage analysis, use @command{gcov} to produce human readable
4677 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4678 @command{gcov} documentation for further information.
4682 With @option{-fprofile-arcs}, for each function of your program GCC
4683 creates a program flow graph, then finds a spanning tree for the graph.
4684 Only arcs that are not on the spanning tree have to be instrumented: the
4685 compiler adds code to count the number of times that these arcs are
4686 executed. When an arc is the only exit or only entrance to a block, the
4687 instrumentation code can be added to the block; otherwise, a new basic
4688 block must be created to hold the instrumentation code.
4691 @item -ftest-coverage
4692 @opindex ftest-coverage
4693 Produce a notes file that the @command{gcov} code-coverage utility
4694 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4695 show program coverage. Each source file's note file is called
4696 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4697 above for a description of @var{auxname} and instructions on how to
4698 generate test coverage data. Coverage data will match the source files
4699 more closely, if you do not optimize.
4701 @item -fdbg-cnt-list
4702 @opindex fdbg-cnt-list
4703 Print the name and the counter upperbound for all debug counters.
4705 @item -fdbg-cnt=@var{counter-value-list}
4707 Set the internal debug counter upperbound. @var{counter-value-list}
4708 is a comma-separated list of @var{name}:@var{value} pairs
4709 which sets the upperbound of each debug counter @var{name} to @var{value}.
4710 All debug counters have the initial upperbound of @var{UINT_MAX},
4711 thus dbg_cnt() returns true always unless the upperbound is set by this option.
4712 e.g. With -fdbg-cnt=dce:10,tail_call:0
4713 dbg_cnt(dce) will return true only for first 10 invocations
4714 and dbg_cnt(tail_call) will return false always.
4716 @item -d@var{letters}
4717 @itemx -fdump-rtl-@var{pass}
4719 Says to make debugging dumps during compilation at times specified by
4720 @var{letters}. This is used for debugging the RTL-based passes of the
4721 compiler. The file names for most of the dumps are made by appending
4722 a pass number and a word to the @var{dumpname}, and the files are
4723 created in the directory of the output file. @var{dumpname} is
4724 generated from the name of the output file, if explicitly specified
4725 and it is not an executable, otherwise it is the basename of the
4726 source file. These switches may have different effects when
4727 @option{-E} is used for preprocessing.
4729 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
4730 @option{-d} option @var{letters}. Here are the possible
4731 letters for use in @var{pass} and @var{letters}, and their meanings:
4735 @item -fdump-rtl-alignments
4736 @opindex fdump-rtl-alignments
4737 Dump after branch alignments have been computed.
4739 @item -fdump-rtl-asmcons
4740 @opindex fdump-rtl-asmcons
4741 Dump after fixing rtl statements that have unsatisfied in/out constraints.
4743 @item -fdump-rtl-auto_inc_dec
4744 @opindex fdump-rtl-auto_inc_dec
4745 Dump after auto-inc-dec discovery. This pass is only run on
4746 architectures that have auto inc or auto dec instructions.
4748 @item -fdump-rtl-barriers
4749 @opindex fdump-rtl-barriers
4750 Dump after cleaning up the barrier instructions.
4752 @item -fdump-rtl-bbpart
4753 @opindex fdump-rtl-bbpart
4754 Dump after partitioning hot and cold basic blocks.
4756 @item -fdump-rtl-bbro
4757 @opindex fdump-rtl-bbro
4758 Dump after block reordering.
4760 @item -fdump-rtl-btl1
4761 @itemx -fdump-rtl-btl2
4762 @opindex fdump-rtl-btl2
4763 @opindex fdump-rtl-btl2
4764 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
4765 after the two branch
4766 target load optimization passes.
4768 @item -fdump-rtl-bypass
4769 @opindex fdump-rtl-bypass
4770 Dump after jump bypassing and control flow optimizations.
4772 @item -fdump-rtl-combine
4773 @opindex fdump-rtl-combine
4774 Dump after the RTL instruction combination pass.
4776 @item -fdump-rtl-compgotos
4777 @opindex fdump-rtl-compgotos
4778 Dump after duplicating the computed gotos.
4780 @item -fdump-rtl-ce1
4781 @itemx -fdump-rtl-ce2
4782 @itemx -fdump-rtl-ce3
4783 @opindex fdump-rtl-ce1
4784 @opindex fdump-rtl-ce2
4785 @opindex fdump-rtl-ce3
4786 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
4787 @option{-fdump-rtl-ce3} enable dumping after the three
4788 if conversion passes.
4790 @itemx -fdump-rtl-cprop_hardreg
4791 @opindex fdump-rtl-cprop_hardreg
4792 Dump after hard register copy propagation.
4794 @itemx -fdump-rtl-csa
4795 @opindex fdump-rtl-csa
4796 Dump after combining stack adjustments.
4798 @item -fdump-rtl-cse1
4799 @itemx -fdump-rtl-cse2
4800 @opindex fdump-rtl-cse1
4801 @opindex fdump-rtl-cse2
4802 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
4803 the two common sub-expression elimination passes.
4805 @itemx -fdump-rtl-dce
4806 @opindex fdump-rtl-dce
4807 Dump after the standalone dead code elimination passes.
4809 @itemx -fdump-rtl-dbr
4810 @opindex fdump-rtl-dbr
4811 Dump after delayed branch scheduling.
4813 @item -fdump-rtl-dce1
4814 @itemx -fdump-rtl-dce2
4815 @opindex fdump-rtl-dce1
4816 @opindex fdump-rtl-dce2
4817 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
4818 the two dead store elimination passes.
4821 @opindex fdump-rtl-eh
4822 Dump after finalization of EH handling code.
4824 @item -fdump-rtl-eh_ranges
4825 @opindex fdump-rtl-eh_ranges
4826 Dump after conversion of EH handling range regions.
4828 @item -fdump-rtl-expand
4829 @opindex fdump-rtl-expand
4830 Dump after RTL generation.
4832 @item -fdump-rtl-fwprop1
4833 @itemx -fdump-rtl-fwprop2
4834 @opindex fdump-rtl-fwprop1
4835 @opindex fdump-rtl-fwprop2
4836 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
4837 dumping after the two forward propagation passes.
4839 @item -fdump-rtl-gcse1
4840 @itemx -fdump-rtl-gcse2
4841 @opindex fdump-rtl-gcse1
4842 @opindex fdump-rtl-gcse2
4843 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
4844 after global common subexpression elimination.
4846 @item -fdump-rtl-init-regs
4847 @opindex fdump-rtl-init-regs
4848 Dump after the initialization of the registers.
4850 @item -fdump-rtl-initvals
4851 @opindex fdump-rtl-initvals
4852 Dump after the computation of the initial value sets.
4854 @itemx -fdump-rtl-into_cfglayout
4855 @opindex fdump-rtl-into_cfglayout
4856 Dump after converting to cfglayout mode.
4858 @item -fdump-rtl-ira
4859 @opindex fdump-rtl-ira
4860 Dump after iterated register allocation.
4862 @item -fdump-rtl-jump
4863 @opindex fdump-rtl-jump
4864 Dump after the second jump optimization.
4866 @item -fdump-rtl-loop2
4867 @opindex fdump-rtl-loop2
4868 @option{-fdump-rtl-loop2} enables dumping after the rtl
4869 loop optimization passes.
4871 @item -fdump-rtl-mach
4872 @opindex fdump-rtl-mach
4873 Dump after performing the machine dependent reorganization pass, if that
4876 @item -fdump-rtl-mode_sw
4877 @opindex fdump-rtl-mode_sw
4878 Dump after removing redundant mode switches.
4880 @item -fdump-rtl-rnreg
4881 @opindex fdump-rtl-rnreg
4882 Dump after register renumbering.
4884 @itemx -fdump-rtl-outof_cfglayout
4885 @opindex fdump-rtl-outof_cfglayout
4886 Dump after converting from cfglayout mode.
4888 @item -fdump-rtl-peephole2
4889 @opindex fdump-rtl-peephole2
4890 Dump after the peephole pass.
4892 @item -fdump-rtl-postreload
4893 @opindex fdump-rtl-postreload
4894 Dump after post-reload optimizations.
4896 @itemx -fdump-rtl-pro_and_epilogue
4897 @opindex fdump-rtl-pro_and_epilogue
4898 Dump after generating the function pro and epilogues.
4900 @item -fdump-rtl-regmove
4901 @opindex fdump-rtl-regmove
4902 Dump after the register move pass.
4904 @item -fdump-rtl-sched1
4905 @itemx -fdump-rtl-sched2
4906 @opindex fdump-rtl-sched1
4907 @opindex fdump-rtl-sched2
4908 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
4909 after the basic block scheduling passes.
4911 @item -fdump-rtl-see
4912 @opindex fdump-rtl-see
4913 Dump after sign extension elimination.
4915 @item -fdump-rtl-seqabstr
4916 @opindex fdump-rtl-seqabstr
4917 Dump after common sequence discovery.
4919 @item -fdump-rtl-shorten
4920 @opindex fdump-rtl-shorten
4921 Dump after shortening branches.
4923 @item -fdump-rtl-sibling
4924 @opindex fdump-rtl-sibling
4925 Dump after sibling call optimizations.
4927 @item -fdump-rtl-split1
4928 @itemx -fdump-rtl-split2
4929 @itemx -fdump-rtl-split3
4930 @itemx -fdump-rtl-split4
4931 @itemx -fdump-rtl-split5
4932 @opindex fdump-rtl-split1
4933 @opindex fdump-rtl-split2
4934 @opindex fdump-rtl-split3
4935 @opindex fdump-rtl-split4
4936 @opindex fdump-rtl-split5
4937 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
4938 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
4939 @option{-fdump-rtl-split5} enable dumping after five rounds of
4940 instruction splitting.
4942 @item -fdump-rtl-sms
4943 @opindex fdump-rtl-sms
4944 Dump after modulo scheduling. This pass is only run on some
4947 @item -fdump-rtl-stack
4948 @opindex fdump-rtl-stack
4949 Dump after conversion from GCC's "flat register file" registers to the
4950 x87's stack-like registers. This pass is only run on x86 variants.
4952 @item -fdump-rtl-subreg1
4953 @itemx -fdump-rtl-subreg2
4954 @opindex fdump-rtl-subreg1
4955 @opindex fdump-rtl-subreg2
4956 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
4957 the two subreg expansion passes.
4959 @item -fdump-rtl-unshare
4960 @opindex fdump-rtl-unshare
4961 Dump after all rtl has been unshared.
4963 @item -fdump-rtl-vartrack
4964 @opindex fdump-rtl-vartrack
4965 Dump after variable tracking.
4967 @item -fdump-rtl-vregs
4968 @opindex fdump-rtl-vregs
4969 Dump after converting virtual registers to hard registers.
4971 @item -fdump-rtl-web
4972 @opindex fdump-rtl-web
4973 Dump after live range splitting.
4975 @item -fdump-rtl-regclass
4976 @itemx -fdump-rtl-subregs_of_mode_init
4977 @itemx -fdump-rtl-subregs_of_mode_finish
4978 @itemx -fdump-rtl-dfinit
4979 @itemx -fdump-rtl-dfinish
4980 @opindex fdump-rtl-regclass
4981 @opindex fdump-rtl-subregs_of_mode_init
4982 @opindex fdump-rtl-subregs_of_mode_finish
4983 @opindex fdump-rtl-dfinit
4984 @opindex fdump-rtl-dfinish
4985 These dumps are defined but always produce empty files.
4987 @item -fdump-rtl-all
4988 @opindex fdump-rtl-all
4989 Produce all the dumps listed above.
4993 Annotate the assembler output with miscellaneous debugging information.
4997 Dump all macro definitions, at the end of preprocessing, in addition to
5002 Produce a core dump whenever an error occurs.
5006 Print statistics on memory usage, at the end of the run, to
5011 Annotate the assembler output with a comment indicating which
5012 pattern and alternative was used. The length of each instruction is
5017 Dump the RTL in the assembler output as a comment before each instruction.
5018 Also turns on @option{-dp} annotation.
5022 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
5023 dump a representation of the control flow graph suitable for viewing with VCG
5024 to @file{@var{file}.@var{pass}.vcg}.
5028 Just generate RTL for a function instead of compiling it. Usually used
5029 with @option{-fdump-rtl-expand}.
5033 Dump debugging information during parsing, to standard error.
5037 @opindex fdump-noaddr
5038 When doing debugging dumps, suppress address output. This makes it more
5039 feasible to use diff on debugging dumps for compiler invocations with
5040 different compiler binaries and/or different
5041 text / bss / data / heap / stack / dso start locations.
5043 @item -fdump-unnumbered
5044 @opindex fdump-unnumbered
5045 When doing debugging dumps, suppress instruction numbers and address output.
5046 This makes it more feasible to use diff on debugging dumps for compiler
5047 invocations with different options, in particular with and without
5050 @item -fdump-unnumbered-links
5051 @opindex fdump-unnumbered-links
5052 When doing debugging dumps (see @option{-d} option above), suppress
5053 instruction numbers for the links to the previous and next instructions
5056 @item -fdump-translation-unit @r{(C++ only)}
5057 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5058 @opindex fdump-translation-unit
5059 Dump a representation of the tree structure for the entire translation
5060 unit to a file. The file name is made by appending @file{.tu} to the
5061 source file name, and the file is created in the same directory as the
5062 output file. If the @samp{-@var{options}} form is used, @var{options}
5063 controls the details of the dump as described for the
5064 @option{-fdump-tree} options.
5066 @item -fdump-class-hierarchy @r{(C++ only)}
5067 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5068 @opindex fdump-class-hierarchy
5069 Dump a representation of each class's hierarchy and virtual function
5070 table layout to a file. The file name is made by appending
5071 @file{.class} to the source file name, and the file is created in the
5072 same directory as the output file. If the @samp{-@var{options}} form
5073 is used, @var{options} controls the details of the dump as described
5074 for the @option{-fdump-tree} options.
5076 @item -fdump-ipa-@var{switch}
5078 Control the dumping at various stages of inter-procedural analysis
5079 language tree to a file. The file name is generated by appending a
5080 switch specific suffix to the source file name, and the file is created
5081 in the same directory as the output file. The following dumps are
5086 Enables all inter-procedural analysis dumps.
5089 Dumps information about call-graph optimization, unused function removal,
5090 and inlining decisions.
5093 Dump after function inlining.
5097 @item -fdump-statistics-@var{option}
5098 @opindex fdump-statistics
5099 Enable and control dumping of pass statistics in a separate file. The
5100 file name is generated by appending a suffix ending in
5101 @samp{.statistics} to the source file name, and the file is created in
5102 the same directory as the output file. If the @samp{-@var{option}}
5103 form is used, @samp{-stats} will cause counters to be summed over the
5104 whole compilation unit while @samp{-details} will dump every event as
5105 the passes generate them. The default with no option is to sum
5106 counters for each function compiled.
5108 @item -fdump-tree-@var{switch}
5109 @itemx -fdump-tree-@var{switch}-@var{options}
5111 Control the dumping at various stages of processing the intermediate
5112 language tree to a file. The file name is generated by appending a
5113 switch specific suffix to the source file name, and the file is
5114 created in the same directory as the output file. If the
5115 @samp{-@var{options}} form is used, @var{options} is a list of
5116 @samp{-} separated options that control the details of the dump. Not
5117 all options are applicable to all dumps, those which are not
5118 meaningful will be ignored. The following options are available
5122 Print the address of each node. Usually this is not meaningful as it
5123 changes according to the environment and source file. Its primary use
5124 is for tying up a dump file with a debug environment.
5126 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
5127 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
5128 use working backward from mangled names in the assembly file.
5130 Inhibit dumping of members of a scope or body of a function merely
5131 because that scope has been reached. Only dump such items when they
5132 are directly reachable by some other path. When dumping pretty-printed
5133 trees, this option inhibits dumping the bodies of control structures.
5135 Print a raw representation of the tree. By default, trees are
5136 pretty-printed into a C-like representation.
5138 Enable more detailed dumps (not honored by every dump option).
5140 Enable dumping various statistics about the pass (not honored by every dump
5143 Enable showing basic block boundaries (disabled in raw dumps).
5145 Enable showing virtual operands for every statement.
5147 Enable showing line numbers for statements.
5149 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5151 Enable showing the tree dump for each statement.
5153 Enable showing the EH region number holding each statement.
5155 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5156 and @option{lineno}.
5159 The following tree dumps are possible:
5163 @opindex fdump-tree-original
5164 Dump before any tree based optimization, to @file{@var{file}.original}.
5167 @opindex fdump-tree-optimized
5168 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5171 @opindex fdump-tree-gimple
5172 Dump each function before and after the gimplification pass to a file. The
5173 file name is made by appending @file{.gimple} to the source file name.
5176 @opindex fdump-tree-cfg
5177 Dump the control flow graph of each function to a file. The file name is
5178 made by appending @file{.cfg} to the source file name.
5181 @opindex fdump-tree-vcg
5182 Dump the control flow graph of each function to a file in VCG format. The
5183 file name is made by appending @file{.vcg} to the source file name. Note
5184 that if the file contains more than one function, the generated file cannot
5185 be used directly by VCG@. You will need to cut and paste each function's
5186 graph into its own separate file first.
5189 @opindex fdump-tree-ch
5190 Dump each function after copying loop headers. The file name is made by
5191 appending @file{.ch} to the source file name.
5194 @opindex fdump-tree-ssa
5195 Dump SSA related information to a file. The file name is made by appending
5196 @file{.ssa} to the source file name.
5199 @opindex fdump-tree-alias
5200 Dump aliasing information for each function. The file name is made by
5201 appending @file{.alias} to the source file name.
5204 @opindex fdump-tree-ccp
5205 Dump each function after CCP@. The file name is made by appending
5206 @file{.ccp} to the source file name.
5209 @opindex fdump-tree-storeccp
5210 Dump each function after STORE-CCP@. The file name is made by appending
5211 @file{.storeccp} to the source file name.
5214 @opindex fdump-tree-pre
5215 Dump trees after partial redundancy elimination. The file name is made
5216 by appending @file{.pre} to the source file name.
5219 @opindex fdump-tree-fre
5220 Dump trees after full redundancy elimination. The file name is made
5221 by appending @file{.fre} to the source file name.
5224 @opindex fdump-tree-copyprop
5225 Dump trees after copy propagation. The file name is made
5226 by appending @file{.copyprop} to the source file name.
5228 @item store_copyprop
5229 @opindex fdump-tree-store_copyprop
5230 Dump trees after store copy-propagation. The file name is made
5231 by appending @file{.store_copyprop} to the source file name.
5234 @opindex fdump-tree-dce
5235 Dump each function after dead code elimination. The file name is made by
5236 appending @file{.dce} to the source file name.
5239 @opindex fdump-tree-mudflap
5240 Dump each function after adding mudflap instrumentation. The file name is
5241 made by appending @file{.mudflap} to the source file name.
5244 @opindex fdump-tree-sra
5245 Dump each function after performing scalar replacement of aggregates. The
5246 file name is made by appending @file{.sra} to the source file name.
5249 @opindex fdump-tree-sink
5250 Dump each function after performing code sinking. The file name is made
5251 by appending @file{.sink} to the source file name.
5254 @opindex fdump-tree-dom
5255 Dump each function after applying dominator tree optimizations. The file
5256 name is made by appending @file{.dom} to the source file name.
5259 @opindex fdump-tree-dse
5260 Dump each function after applying dead store elimination. The file
5261 name is made by appending @file{.dse} to the source file name.
5264 @opindex fdump-tree-phiopt
5265 Dump each function after optimizing PHI nodes into straightline code. The file
5266 name is made by appending @file{.phiopt} to the source file name.
5269 @opindex fdump-tree-forwprop
5270 Dump each function after forward propagating single use variables. The file
5271 name is made by appending @file{.forwprop} to the source file name.
5274 @opindex fdump-tree-copyrename
5275 Dump each function after applying the copy rename optimization. The file
5276 name is made by appending @file{.copyrename} to the source file name.
5279 @opindex fdump-tree-nrv
5280 Dump each function after applying the named return value optimization on
5281 generic trees. The file name is made by appending @file{.nrv} to the source
5285 @opindex fdump-tree-vect
5286 Dump each function after applying vectorization of loops. The file name is
5287 made by appending @file{.vect} to the source file name.
5290 @opindex fdump-tree-vrp
5291 Dump each function after Value Range Propagation (VRP). The file name
5292 is made by appending @file{.vrp} to the source file name.
5295 @opindex fdump-tree-all
5296 Enable all the available tree dumps with the flags provided in this option.
5299 @item -ftree-vectorizer-verbose=@var{n}
5300 @opindex ftree-vectorizer-verbose
5301 This option controls the amount of debugging output the vectorizer prints.
5302 This information is written to standard error, unless
5303 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5304 in which case it is output to the usual dump listing file, @file{.vect}.
5305 For @var{n}=0 no diagnostic information is reported.
5306 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5307 and the total number of loops that got vectorized.
5308 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5309 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5310 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5311 level that @option{-fdump-tree-vect-stats} uses.
5312 Higher verbosity levels mean either more information dumped for each
5313 reported loop, or same amount of information reported for more loops:
5314 If @var{n}=3, alignment related information is added to the reports.
5315 If @var{n}=4, data-references related information (e.g.@: memory dependences,
5316 memory access-patterns) is added to the reports.
5317 If @var{n}=5, the vectorizer reports also non-vectorized inner-most loops
5318 that did not pass the first analysis phase (i.e., may not be countable, or
5319 may have complicated control-flow).
5320 If @var{n}=6, the vectorizer reports also non-vectorized nested loops.
5321 For @var{n}=7, all the information the vectorizer generates during its
5322 analysis and transformation is reported. This is the same verbosity level
5323 that @option{-fdump-tree-vect-details} uses.
5325 @item -frandom-seed=@var{string}
5326 @opindex frandom-seed
5327 This option provides a seed that GCC uses when it would otherwise use
5328 random numbers. It is used to generate certain symbol names
5329 that have to be different in every compiled file. It is also used to
5330 place unique stamps in coverage data files and the object files that
5331 produce them. You can use the @option{-frandom-seed} option to produce
5332 reproducibly identical object files.
5334 The @var{string} should be different for every file you compile.
5336 @item -fsched-verbose=@var{n}
5337 @opindex fsched-verbose
5338 On targets that use instruction scheduling, this option controls the
5339 amount of debugging output the scheduler prints. This information is
5340 written to standard error, unless @option{-fdump-rtl-sched1} or
5341 @option{-fdump-rtl-sched2} is specified, in which case it is output
5342 to the usual dump listing file, @file{.sched} or @file{.sched2}
5343 respectively. However for @var{n} greater than nine, the output is
5344 always printed to standard error.
5346 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5347 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5348 For @var{n} greater than one, it also output basic block probabilities,
5349 detailed ready list information and unit/insn info. For @var{n} greater
5350 than two, it includes RTL at abort point, control-flow and regions info.
5351 And for @var{n} over four, @option{-fsched-verbose} also includes
5355 @itemx -save-temps=cwd
5357 Store the usual ``temporary'' intermediate files permanently; place them
5358 in the current directory and name them based on the source file. Thus,
5359 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5360 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5361 preprocessed @file{foo.i} output file even though the compiler now
5362 normally uses an integrated preprocessor.
5364 When used in combination with the @option{-x} command line option,
5365 @option{-save-temps} is sensible enough to avoid over writing an
5366 input source file with the same extension as an intermediate file.
5367 The corresponding intermediate file may be obtained by renaming the
5368 source file before using @option{-save-temps}.
5370 If you invoke GCC in parallel, compiling several different source
5371 files that share a common base name in different subdirectories or the
5372 same source file compiled for multiple output destinations, it is
5373 likely that the different parallel compilers will interfere with each
5374 other, and overwrite the temporary files. For instance:
5377 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5378 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5381 may result in @file{foo.i} and @file{foo.o} being written to
5382 simultaneously by both compilers.
5384 @item -save-temps=obj
5385 @opindex save-temps=obj
5386 Store the usual ``temporary'' intermediate files permanently. If the
5387 @option{-o} option is used, the temporary files are based on the
5388 object file. If the @option{-o} option is not used, the
5389 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5394 gcc -save-temps=obj -c foo.c
5395 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5396 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5399 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5400 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5401 @file{dir2/yfoobar.o}.
5403 @item -time@r{[}=@var{file}@r{]}
5405 Report the CPU time taken by each subprocess in the compilation
5406 sequence. For C source files, this is the compiler proper and assembler
5407 (plus the linker if linking is done).
5409 Without the specification of an output file, the output looks like this:
5416 The first number on each line is the ``user time'', that is time spent
5417 executing the program itself. The second number is ``system time'',
5418 time spent executing operating system routines on behalf of the program.
5419 Both numbers are in seconds.
5421 With the specification of an output file, the output is appended to the
5422 named file, and it looks like this:
5425 0.12 0.01 cc1 @var{options}
5426 0.00 0.01 as @var{options}
5429 The ``user time'' and the ``system time'' are moved before the program
5430 name, and the options passed to the program are displayed, so that one
5431 can later tell what file was being compiled, and with which options.
5433 @item -fvar-tracking
5434 @opindex fvar-tracking
5435 Run variable tracking pass. It computes where variables are stored at each
5436 position in code. Better debugging information is then generated
5437 (if the debugging information format supports this information).
5439 It is enabled by default when compiling with optimization (@option{-Os},
5440 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5441 the debug info format supports it.
5443 @item -print-file-name=@var{library}
5444 @opindex print-file-name
5445 Print the full absolute name of the library file @var{library} that
5446 would be used when linking---and don't do anything else. With this
5447 option, GCC does not compile or link anything; it just prints the
5450 @item -print-multi-directory
5451 @opindex print-multi-directory
5452 Print the directory name corresponding to the multilib selected by any
5453 other switches present in the command line. This directory is supposed
5454 to exist in @env{GCC_EXEC_PREFIX}.
5456 @item -print-multi-lib
5457 @opindex print-multi-lib
5458 Print the mapping from multilib directory names to compiler switches
5459 that enable them. The directory name is separated from the switches by
5460 @samp{;}, and each switch starts with an @samp{@@} instead of the
5461 @samp{-}, without spaces between multiple switches. This is supposed to
5462 ease shell-processing.
5464 @item -print-prog-name=@var{program}
5465 @opindex print-prog-name
5466 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5468 @item -print-libgcc-file-name
5469 @opindex print-libgcc-file-name
5470 Same as @option{-print-file-name=libgcc.a}.
5472 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5473 but you do want to link with @file{libgcc.a}. You can do
5476 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5479 @item -print-search-dirs
5480 @opindex print-search-dirs
5481 Print the name of the configured installation directory and a list of
5482 program and library directories @command{gcc} will search---and don't do anything else.
5484 This is useful when @command{gcc} prints the error message
5485 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5486 To resolve this you either need to put @file{cpp0} and the other compiler
5487 components where @command{gcc} expects to find them, or you can set the environment
5488 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5489 Don't forget the trailing @samp{/}.
5490 @xref{Environment Variables}.
5492 @item -print-sysroot
5493 @opindex print-sysroot
5494 Print the target sysroot directory that will be used during
5495 compilation. This is the target sysroot specified either at configure
5496 time or using the @option{--sysroot} option, possibly with an extra
5497 suffix that depends on compilation options. If no target sysroot is
5498 specified, the option prints nothing.
5500 @item -print-sysroot-headers-suffix
5501 @opindex print-sysroot-headers-suffix
5502 Print the suffix added to the target sysroot when searching for
5503 headers, or give an error if the compiler is not configured with such
5504 a suffix---and don't do anything else.
5507 @opindex dumpmachine
5508 Print the compiler's target machine (for example,
5509 @samp{i686-pc-linux-gnu})---and don't do anything else.
5512 @opindex dumpversion
5513 Print the compiler version (for example, @samp{3.0})---and don't do
5518 Print the compiler's built-in specs---and don't do anything else. (This
5519 is used when GCC itself is being built.) @xref{Spec Files}.
5521 @item -feliminate-unused-debug-types
5522 @opindex feliminate-unused-debug-types
5523 Normally, when producing DWARF2 output, GCC will emit debugging
5524 information for all types declared in a compilation
5525 unit, regardless of whether or not they are actually used
5526 in that compilation unit. Sometimes this is useful, such as
5527 if, in the debugger, you want to cast a value to a type that is
5528 not actually used in your program (but is declared). More often,
5529 however, this results in a significant amount of wasted space.
5530 With this option, GCC will avoid producing debug symbol output
5531 for types that are nowhere used in the source file being compiled.
5534 @node Optimize Options
5535 @section Options That Control Optimization
5536 @cindex optimize options
5537 @cindex options, optimization
5539 These options control various sorts of optimizations.
5541 Without any optimization option, the compiler's goal is to reduce the
5542 cost of compilation and to make debugging produce the expected
5543 results. Statements are independent: if you stop the program with a
5544 breakpoint between statements, you can then assign a new value to any
5545 variable or change the program counter to any other statement in the
5546 function and get exactly the results you would expect from the source
5549 Turning on optimization flags makes the compiler attempt to improve
5550 the performance and/or code size at the expense of compilation time
5551 and possibly the ability to debug the program.
5553 The compiler performs optimization based on the knowledge it has of the
5554 program. Compiling multiple files at once to a single output file mode allows
5555 the compiler to use information gained from all of the files when compiling
5558 Not all optimizations are controlled directly by a flag. Only
5559 optimizations that have a flag are listed in this section.
5561 Depending on the target and how GCC was configured, a slightly different
5562 set of optimizations may be enabled at each @option{-O} level than
5563 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
5564 to find out the exact set of optimizations that are enabled at each level.
5565 @xref{Overall Options}, for examples.
5572 Optimize. Optimizing compilation takes somewhat more time, and a lot
5573 more memory for a large function.
5575 With @option{-O}, the compiler tries to reduce code size and execution
5576 time, without performing any optimizations that take a great deal of
5579 @option{-O} turns on the following optimization flags:
5582 -fcprop-registers @gol
5585 -fdelayed-branch @gol
5587 -fguess-branch-probability @gol
5588 -fif-conversion2 @gol
5589 -fif-conversion @gol
5590 -finline-small-functions @gol
5591 -fipa-pure-const @gol
5592 -fipa-reference @gol
5594 -fsplit-wide-types @gol
5595 -ftree-builtin-call-dce @gol
5598 -ftree-copyrename @gol
5600 -ftree-dominator-opts @gol
5602 -ftree-forwprop @gol
5610 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5611 where doing so does not interfere with debugging.
5615 Optimize even more. GCC performs nearly all supported optimizations
5616 that do not involve a space-speed tradeoff.
5617 As compared to @option{-O}, this option increases both compilation time
5618 and the performance of the generated code.
5620 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5621 also turns on the following optimization flags:
5622 @gccoptlist{-fthread-jumps @gol
5623 -falign-functions -falign-jumps @gol
5624 -falign-loops -falign-labels @gol
5627 -fcse-follow-jumps -fcse-skip-blocks @gol
5628 -fdelete-null-pointer-checks @gol
5629 -fexpensive-optimizations @gol
5630 -fgcse -fgcse-lm @gol
5631 -findirect-inlining @gol
5632 -foptimize-sibling-calls @gol
5635 -freorder-blocks -freorder-functions @gol
5636 -frerun-cse-after-loop @gol
5637 -fsched-interblock -fsched-spec @gol
5638 -fschedule-insns -fschedule-insns2 @gol
5639 -fstrict-aliasing -fstrict-overflow @gol
5640 -ftree-switch-conversion @gol
5644 Please note the warning under @option{-fgcse} about
5645 invoking @option{-O2} on programs that use computed gotos.
5649 Optimize yet more. @option{-O3} turns on all optimizations specified
5650 by @option{-O2} and also turns on the @option{-finline-functions},
5651 @option{-funswitch-loops}, @option{-fpredictive-commoning},
5652 @option{-fgcse-after-reload} and @option{-ftree-vectorize} options.
5656 Reduce compilation time and make debugging produce the expected
5657 results. This is the default.
5661 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5662 do not typically increase code size. It also performs further
5663 optimizations designed to reduce code size.
5665 @option{-Os} disables the following optimization flags:
5666 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5667 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5668 -fprefetch-loop-arrays -ftree-vect-loop-version}
5670 If you use multiple @option{-O} options, with or without level numbers,
5671 the last such option is the one that is effective.
5674 Options of the form @option{-f@var{flag}} specify machine-independent
5675 flags. Most flags have both positive and negative forms; the negative
5676 form of @option{-ffoo} would be @option{-fno-foo}. In the table
5677 below, only one of the forms is listed---the one you typically will
5678 use. You can figure out the other form by either removing @samp{no-}
5681 The following options control specific optimizations. They are either
5682 activated by @option{-O} options or are related to ones that are. You
5683 can use the following flags in the rare cases when ``fine-tuning'' of
5684 optimizations to be performed is desired.
5687 @item -fno-default-inline
5688 @opindex fno-default-inline
5689 Do not make member functions inline by default merely because they are
5690 defined inside the class scope (C++ only). Otherwise, when you specify
5691 @w{@option{-O}}, member functions defined inside class scope are compiled
5692 inline by default; i.e., you don't need to add @samp{inline} in front of
5693 the member function name.
5695 @item -fno-defer-pop
5696 @opindex fno-defer-pop
5697 Always pop the arguments to each function call as soon as that function
5698 returns. For machines which must pop arguments after a function call,
5699 the compiler normally lets arguments accumulate on the stack for several
5700 function calls and pops them all at once.
5702 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5704 @item -fforward-propagate
5705 @opindex fforward-propagate
5706 Perform a forward propagation pass on RTL@. The pass tries to combine two
5707 instructions and checks if the result can be simplified. If loop unrolling
5708 is active, two passes are performed and the second is scheduled after
5711 This option is enabled by default at optimization levels @option{-O},
5712 @option{-O2}, @option{-O3}, @option{-Os}.
5714 @item -fomit-frame-pointer
5715 @opindex fomit-frame-pointer
5716 Don't keep the frame pointer in a register for functions that
5717 don't need one. This avoids the instructions to save, set up and
5718 restore frame pointers; it also makes an extra register available
5719 in many functions. @strong{It also makes debugging impossible on
5722 On some machines, such as the VAX, this flag has no effect, because
5723 the standard calling sequence automatically handles the frame pointer
5724 and nothing is saved by pretending it doesn't exist. The
5725 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
5726 whether a target machine supports this flag. @xref{Registers,,Register
5727 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
5729 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5731 @item -foptimize-sibling-calls
5732 @opindex foptimize-sibling-calls
5733 Optimize sibling and tail recursive calls.
5735 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5739 Don't pay attention to the @code{inline} keyword. Normally this option
5740 is used to keep the compiler from expanding any functions inline.
5741 Note that if you are not optimizing, no functions can be expanded inline.
5743 @item -finline-small-functions
5744 @opindex finline-small-functions
5745 Integrate functions into their callers when their body is smaller than expected
5746 function call code (so overall size of program gets smaller). The compiler
5747 heuristically decides which functions are simple enough to be worth integrating
5750 Enabled at level @option{-O2}.
5752 @item -findirect-inlining
5753 @opindex findirect-inlining
5754 Inline also indirect calls that are discovered to be known at compile
5755 time thanks to previous inlining. This option has any effect only
5756 when inlining itself is turned on by the @option{-finline-functions}
5757 or @option{-finline-small-functions} options.
5759 Enabled at level @option{-O2}.
5761 @item -finline-functions
5762 @opindex finline-functions
5763 Integrate all simple functions into their callers. The compiler
5764 heuristically decides which functions are simple enough to be worth
5765 integrating in this way.
5767 If all calls to a given function are integrated, and the function is
5768 declared @code{static}, then the function is normally not output as
5769 assembler code in its own right.
5771 Enabled at level @option{-O3}.
5773 @item -finline-functions-called-once
5774 @opindex finline-functions-called-once
5775 Consider all @code{static} functions called once for inlining into their
5776 caller even if they are not marked @code{inline}. If a call to a given
5777 function is integrated, then the function is not output as assembler code
5780 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
5782 @item -fearly-inlining
5783 @opindex fearly-inlining
5784 Inline functions marked by @code{always_inline} and functions whose body seems
5785 smaller than the function call overhead early before doing
5786 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
5787 makes profiling significantly cheaper and usually inlining faster on programs
5788 having large chains of nested wrapper functions.
5792 @item -finline-limit=@var{n}
5793 @opindex finline-limit
5794 By default, GCC limits the size of functions that can be inlined. This flag
5795 allows coarse control of this limit. @var{n} is the size of functions that
5796 can be inlined in number of pseudo instructions.
5798 Inlining is actually controlled by a number of parameters, which may be
5799 specified individually by using @option{--param @var{name}=@var{value}}.
5800 The @option{-finline-limit=@var{n}} option sets some of these parameters
5804 @item max-inline-insns-single
5805 is set to @var{n}/2.
5806 @item max-inline-insns-auto
5807 is set to @var{n}/2.
5810 See below for a documentation of the individual
5811 parameters controlling inlining and for the defaults of these parameters.
5813 @emph{Note:} there may be no value to @option{-finline-limit} that results
5814 in default behavior.
5816 @emph{Note:} pseudo instruction represents, in this particular context, an
5817 abstract measurement of function's size. In no way does it represent a count
5818 of assembly instructions and as such its exact meaning might change from one
5819 release to an another.
5821 @item -fkeep-inline-functions
5822 @opindex fkeep-inline-functions
5823 In C, emit @code{static} functions that are declared @code{inline}
5824 into the object file, even if the function has been inlined into all
5825 of its callers. This switch does not affect functions using the
5826 @code{extern inline} extension in GNU C89@. In C++, emit any and all
5827 inline functions into the object file.
5829 @item -fkeep-static-consts
5830 @opindex fkeep-static-consts
5831 Emit variables declared @code{static const} when optimization isn't turned
5832 on, even if the variables aren't referenced.
5834 GCC enables this option by default. If you want to force the compiler to
5835 check if the variable was referenced, regardless of whether or not
5836 optimization is turned on, use the @option{-fno-keep-static-consts} option.
5838 @item -fmerge-constants
5839 @opindex fmerge-constants
5840 Attempt to merge identical constants (string constants and floating point
5841 constants) across compilation units.
5843 This option is the default for optimized compilation if the assembler and
5844 linker support it. Use @option{-fno-merge-constants} to inhibit this
5847 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5849 @item -fmerge-all-constants
5850 @opindex fmerge-all-constants
5851 Attempt to merge identical constants and identical variables.
5853 This option implies @option{-fmerge-constants}. In addition to
5854 @option{-fmerge-constants} this considers e.g.@: even constant initialized
5855 arrays or initialized constant variables with integral or floating point
5856 types. Languages like C or C++ require each variable, including multiple
5857 instances of the same variable in recursive calls, to have distinct locations,
5858 so using this option will result in non-conforming
5861 @item -fmodulo-sched
5862 @opindex fmodulo-sched
5863 Perform swing modulo scheduling immediately before the first scheduling
5864 pass. This pass looks at innermost loops and reorders their
5865 instructions by overlapping different iterations.
5867 @item -fmodulo-sched-allow-regmoves
5868 @opindex fmodulo-sched-allow-regmoves
5869 Perform more aggressive SMS based modulo scheduling with register moves
5870 allowed. By setting this flag certain anti-dependences edges will be
5871 deleted which will trigger the generation of reg-moves based on the
5872 life-range analysis. This option is effective only with
5873 @option{-fmodulo-sched} enabled.
5875 @item -fno-branch-count-reg
5876 @opindex fno-branch-count-reg
5877 Do not use ``decrement and branch'' instructions on a count register,
5878 but instead generate a sequence of instructions that decrement a
5879 register, compare it against zero, then branch based upon the result.
5880 This option is only meaningful on architectures that support such
5881 instructions, which include x86, PowerPC, IA-64 and S/390.
5883 The default is @option{-fbranch-count-reg}.
5885 @item -fno-function-cse
5886 @opindex fno-function-cse
5887 Do not put function addresses in registers; make each instruction that
5888 calls a constant function contain the function's address explicitly.
5890 This option results in less efficient code, but some strange hacks
5891 that alter the assembler output may be confused by the optimizations
5892 performed when this option is not used.
5894 The default is @option{-ffunction-cse}
5896 @item -fno-zero-initialized-in-bss
5897 @opindex fno-zero-initialized-in-bss
5898 If the target supports a BSS section, GCC by default puts variables that
5899 are initialized to zero into BSS@. This can save space in the resulting
5902 This option turns off this behavior because some programs explicitly
5903 rely on variables going to the data section. E.g., so that the
5904 resulting executable can find the beginning of that section and/or make
5905 assumptions based on that.
5907 The default is @option{-fzero-initialized-in-bss}.
5909 @item -fmudflap -fmudflapth -fmudflapir
5913 @cindex bounds checking
5915 For front-ends that support it (C and C++), instrument all risky
5916 pointer/array dereferencing operations, some standard library
5917 string/heap functions, and some other associated constructs with
5918 range/validity tests. Modules so instrumented should be immune to
5919 buffer overflows, invalid heap use, and some other classes of C/C++
5920 programming errors. The instrumentation relies on a separate runtime
5921 library (@file{libmudflap}), which will be linked into a program if
5922 @option{-fmudflap} is given at link time. Run-time behavior of the
5923 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
5924 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
5927 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
5928 link if your program is multi-threaded. Use @option{-fmudflapir}, in
5929 addition to @option{-fmudflap} or @option{-fmudflapth}, if
5930 instrumentation should ignore pointer reads. This produces less
5931 instrumentation (and therefore faster execution) and still provides
5932 some protection against outright memory corrupting writes, but allows
5933 erroneously read data to propagate within a program.
5935 @item -fthread-jumps
5936 @opindex fthread-jumps
5937 Perform optimizations where we check to see if a jump branches to a
5938 location where another comparison subsumed by the first is found. If
5939 so, the first branch is redirected to either the destination of the
5940 second branch or a point immediately following it, depending on whether
5941 the condition is known to be true or false.
5943 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5945 @item -fsplit-wide-types
5946 @opindex fsplit-wide-types
5947 When using a type that occupies multiple registers, such as @code{long
5948 long} on a 32-bit system, split the registers apart and allocate them
5949 independently. This normally generates better code for those types,
5950 but may make debugging more difficult.
5952 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
5955 @item -fcse-follow-jumps
5956 @opindex fcse-follow-jumps
5957 In common subexpression elimination (CSE), scan through jump instructions
5958 when the target of the jump is not reached by any other path. For
5959 example, when CSE encounters an @code{if} statement with an
5960 @code{else} clause, CSE will follow the jump when the condition
5963 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5965 @item -fcse-skip-blocks
5966 @opindex fcse-skip-blocks
5967 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
5968 follow jumps which conditionally skip over blocks. When CSE
5969 encounters a simple @code{if} statement with no else clause,
5970 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
5971 body of the @code{if}.
5973 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5975 @item -frerun-cse-after-loop
5976 @opindex frerun-cse-after-loop
5977 Re-run common subexpression elimination after loop optimizations has been
5980 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5984 Perform a global common subexpression elimination pass.
5985 This pass also performs global constant and copy propagation.
5987 @emph{Note:} When compiling a program using computed gotos, a GCC
5988 extension, you may get better runtime performance if you disable
5989 the global common subexpression elimination pass by adding
5990 @option{-fno-gcse} to the command line.
5992 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5996 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
5997 attempt to move loads which are only killed by stores into themselves. This
5998 allows a loop containing a load/store sequence to be changed to a load outside
5999 the loop, and a copy/store within the loop.
6001 Enabled by default when gcse is enabled.
6005 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6006 global common subexpression elimination. This pass will attempt to move
6007 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6008 loops containing a load/store sequence can be changed to a load before
6009 the loop and a store after the loop.
6011 Not enabled at any optimization level.
6015 When @option{-fgcse-las} is enabled, the global common subexpression
6016 elimination pass eliminates redundant loads that come after stores to the
6017 same memory location (both partial and full redundancies).
6019 Not enabled at any optimization level.
6021 @item -fgcse-after-reload
6022 @opindex fgcse-after-reload
6023 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6024 pass is performed after reload. The purpose of this pass is to cleanup
6027 @item -funsafe-loop-optimizations
6028 @opindex funsafe-loop-optimizations
6029 If given, the loop optimizer will assume that loop indices do not
6030 overflow, and that the loops with nontrivial exit condition are not
6031 infinite. This enables a wider range of loop optimizations even if
6032 the loop optimizer itself cannot prove that these assumptions are valid.
6033 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
6034 if it finds this kind of loop.
6036 @item -fcrossjumping
6037 @opindex fcrossjumping
6038 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6039 resulting code may or may not perform better than without cross-jumping.
6041 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6043 @item -fauto-inc-dec
6044 @opindex fauto-inc-dec
6045 Combine increments or decrements of addresses with memory accesses.
6046 This pass is always skipped on architectures that do not have
6047 instructions to support this. Enabled by default at @option{-O} and
6048 higher on architectures that support this.
6052 Perform dead code elimination (DCE) on RTL@.
6053 Enabled by default at @option{-O} and higher.
6057 Perform dead store elimination (DSE) on RTL@.
6058 Enabled by default at @option{-O} and higher.
6060 @item -fif-conversion
6061 @opindex fif-conversion
6062 Attempt to transform conditional jumps into branch-less equivalents. This
6063 include use of conditional moves, min, max, set flags and abs instructions, and
6064 some tricks doable by standard arithmetics. The use of conditional execution
6065 on chips where it is available is controlled by @code{if-conversion2}.
6067 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6069 @item -fif-conversion2
6070 @opindex fif-conversion2
6071 Use conditional execution (where available) to transform conditional jumps into
6072 branch-less equivalents.
6074 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6076 @item -fdelete-null-pointer-checks
6077 @opindex fdelete-null-pointer-checks
6078 Assume that programs cannot safely dereference null pointers, and that
6079 no code or data element resides there. This enables simple constant
6080 folding optimizations at all optimization levels. In addition, other
6081 optimization passes in GCC use this flag to control global dataflow
6082 analyses that eliminate useless checks for null pointers; these assume
6083 that if a pointer is checked after it has already been dereferenced,
6086 Note however that in some environments this assumption is not true.
6087 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6088 for programs which depend on that behavior.
6090 Some targets, especially embedded ones, disable this option at all levels.
6091 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6092 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6093 are enabled independently at different optimization levels.
6095 @item -fexpensive-optimizations
6096 @opindex fexpensive-optimizations
6097 Perform a number of minor optimizations that are relatively expensive.
6099 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6101 @item -foptimize-register-move
6103 @opindex foptimize-register-move
6105 Attempt to reassign register numbers in move instructions and as
6106 operands of other simple instructions in order to maximize the amount of
6107 register tying. This is especially helpful on machines with two-operand
6110 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6113 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6115 @item -fira-algorithm=@var{algorithm}
6116 Use specified coloring algorithm for the integrated register
6117 allocator. The @var{algorithm} argument should be @code{priority} or
6118 @code{CB}. The first algorithm specifies Chow's priority coloring,
6119 the second one specifies Chaitin-Briggs coloring. The second
6120 algorithm can be unimplemented for some architectures. If it is
6121 implemented, it is the default because Chaitin-Briggs coloring as a
6122 rule generates a better code.
6124 @item -fira-region=@var{region}
6125 Use specified regions for the integrated register allocator. The
6126 @var{region} argument should be one of @code{all}, @code{mixed}, or
6127 @code{one}. The first value means using all loops as register
6128 allocation regions, the second value which is the default means using
6129 all loops except for loops with small register pressure as the
6130 regions, and third one means using all function as a single region.
6131 The first value can give best result for machines with small size and
6132 irregular register set, the third one results in faster and generates
6133 decent code and the smallest size code, and the default value usually
6134 give the best results in most cases and for most architectures.
6136 @item -fira-coalesce
6137 @opindex fira-coalesce
6138 Do optimistic register coalescing. This option might be profitable for
6139 architectures with big regular register files.
6141 @item -fno-ira-share-save-slots
6142 @opindex fno-ira-share-save-slots
6143 Switch off sharing stack slots used for saving call used hard
6144 registers living through a call. Each hard register will get a
6145 separate stack slot and as a result function stack frame will be
6148 @item -fno-ira-share-spill-slots
6149 @opindex fno-ira-share-spill-slots
6150 Switch off sharing stack slots allocated for pseudo-registers. Each
6151 pseudo-register which did not get a hard register will get a separate
6152 stack slot and as a result function stack frame will be bigger.
6154 @item -fira-verbose=@var{n}
6155 @opindex fira-verbose
6156 Set up how verbose dump file for the integrated register allocator
6157 will be. Default value is 5. If the value is greater or equal to 10,
6158 the dump file will be stderr as if the value were @var{n} minus 10.
6160 @item -fdelayed-branch
6161 @opindex fdelayed-branch
6162 If supported for the target machine, attempt to reorder instructions
6163 to exploit instruction slots available after delayed branch
6166 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6168 @item -fschedule-insns
6169 @opindex fschedule-insns
6170 If supported for the target machine, attempt to reorder instructions to
6171 eliminate execution stalls due to required data being unavailable. This
6172 helps machines that have slow floating point or memory load instructions
6173 by allowing other instructions to be issued until the result of the load
6174 or floating point instruction is required.
6176 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6178 @item -fschedule-insns2
6179 @opindex fschedule-insns2
6180 Similar to @option{-fschedule-insns}, but requests an additional pass of
6181 instruction scheduling after register allocation has been done. This is
6182 especially useful on machines with a relatively small number of
6183 registers and where memory load instructions take more than one cycle.
6185 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6187 @item -fno-sched-interblock
6188 @opindex fno-sched-interblock
6189 Don't schedule instructions across basic blocks. This is normally
6190 enabled by default when scheduling before register allocation, i.e.@:
6191 with @option{-fschedule-insns} or at @option{-O2} or higher.
6193 @item -fno-sched-spec
6194 @opindex fno-sched-spec
6195 Don't allow speculative motion of non-load instructions. This is normally
6196 enabled by default when scheduling before register allocation, i.e.@:
6197 with @option{-fschedule-insns} or at @option{-O2} or higher.
6199 @item -fsched-spec-load
6200 @opindex fsched-spec-load
6201 Allow speculative motion of some load instructions. This only makes
6202 sense when scheduling before register allocation, i.e.@: with
6203 @option{-fschedule-insns} or at @option{-O2} or higher.
6205 @item -fsched-spec-load-dangerous
6206 @opindex fsched-spec-load-dangerous
6207 Allow speculative motion of more load instructions. This only makes
6208 sense when scheduling before register allocation, i.e.@: with
6209 @option{-fschedule-insns} or at @option{-O2} or higher.
6211 @item -fsched-stalled-insns
6212 @itemx -fsched-stalled-insns=@var{n}
6213 @opindex fsched-stalled-insns
6214 Define how many insns (if any) can be moved prematurely from the queue
6215 of stalled insns into the ready list, during the second scheduling pass.
6216 @option{-fno-sched-stalled-insns} means that no insns will be moved
6217 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6218 on how many queued insns can be moved prematurely.
6219 @option{-fsched-stalled-insns} without a value is equivalent to
6220 @option{-fsched-stalled-insns=1}.
6222 @item -fsched-stalled-insns-dep
6223 @itemx -fsched-stalled-insns-dep=@var{n}
6224 @opindex fsched-stalled-insns-dep
6225 Define how many insn groups (cycles) will be examined for a dependency
6226 on a stalled insn that is candidate for premature removal from the queue
6227 of stalled insns. This has an effect only during the second scheduling pass,
6228 and only if @option{-fsched-stalled-insns} is used.
6229 @option{-fno-sched-stalled-insns-dep} is equivalent to
6230 @option{-fsched-stalled-insns-dep=0}.
6231 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6232 @option{-fsched-stalled-insns-dep=1}.
6234 @item -fsched2-use-superblocks
6235 @opindex fsched2-use-superblocks
6236 When scheduling after register allocation, do use superblock scheduling
6237 algorithm. Superblock scheduling allows motion across basic block boundaries
6238 resulting on faster schedules. This option is experimental, as not all machine
6239 descriptions used by GCC model the CPU closely enough to avoid unreliable
6240 results from the algorithm.
6242 This only makes sense when scheduling after register allocation, i.e.@: with
6243 @option{-fschedule-insns2} or at @option{-O2} or higher.
6245 @item -fsched-group-heuristic
6246 @opindex fsched-group-heuristic
6247 Enable the group heuristic in the scheduler. This heuristic favors
6248 the instruction that belongs to a schedule group. This is enabled
6249 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6250 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6252 @item -fsched-critical-path-heuristic
6253 @opindex fsched-critical-path-heuristic
6254 Enable the critical-path heuristic in the scheduler. This heuristic favors
6255 instructions on the critical path. This is enabled by default when
6256 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6257 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6259 @item -fsched-spec-insn-heuristic
6260 @opindex fsched-spec-insn-heuristic
6261 Enable the speculative instruction heuristic in the scheduler. This
6262 heuristic favors speculative instructions with greater dependency weakness.
6263 This is enabled by default when scheduling is enabled, i.e.@:
6264 with @option{-fschedule-insns} or @option{-fschedule-insns2}
6265 or at @option{-O2} or higher.
6267 @item -fsched-reg-pressure-heuristic
6268 @opindex fsched-reg-pressure-heuristic
6269 Enable the register pressure heuristic in the scheduler. This heuristic
6270 favors the instruction with smaller contribution to register pressure.
6271 This only makes sense when scheduling before register allocation, i.e.@:
6272 with @option{-fschedule-insns} or at @option{-O2} or higher.
6274 @item -fsched-rank-heuristic
6275 @opindex fsched-rank-heuristic
6276 Enable the rank heuristic in the scheduler. This heuristic favors
6277 the instruction belonging to a basic block with greater size or frequency.
6278 This is enabled by default when scheduling is enabled, i.e.@:
6279 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6280 at @option{-O2} or higher.
6282 @item -fsched-last-insn-heuristic
6283 @opindex fsched-last-insn-heuristic
6284 Enable the last-instruction heuristic in the scheduler. This heuristic
6285 favors the instruction that is less dependent on the last instruction
6286 scheduled. This is enabled by default when scheduling is enabled,
6287 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6288 at @option{-O2} or higher.
6290 @item -fsched-dep-count-heuristic
6291 @opindex fsched-dep-count-heuristic
6292 Enable the dependent-count heuristic in the scheduler. This heuristic
6293 favors the instruction that has more instructions depending on it.
6294 This is enabled by default when scheduling is enabled, i.e.@:
6295 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6296 at @option{-O2} or higher.
6298 @item -fsched2-use-traces
6299 @opindex fsched2-use-traces
6300 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
6301 allocation and additionally perform code duplication in order to increase the
6302 size of superblocks using tracer pass. See @option{-ftracer} for details on
6305 This mode should produce faster but significantly longer programs. Also
6306 without @option{-fbranch-probabilities} the traces constructed may not
6307 match the reality and hurt the performance. This only makes
6308 sense when scheduling after register allocation, i.e.@: with
6309 @option{-fschedule-insns2} or at @option{-O2} or higher.
6311 @item -freschedule-modulo-scheduled-loops
6312 @opindex freschedule-modulo-scheduled-loops
6313 The modulo scheduling comes before the traditional scheduling, if a loop
6314 was modulo scheduled we may want to prevent the later scheduling passes
6315 from changing its schedule, we use this option to control that.
6317 @item -fselective-scheduling
6318 @opindex fselective-scheduling
6319 Schedule instructions using selective scheduling algorithm. Selective
6320 scheduling runs instead of the first scheduler pass.
6322 @item -fselective-scheduling2
6323 @opindex fselective-scheduling2
6324 Schedule instructions using selective scheduling algorithm. Selective
6325 scheduling runs instead of the second scheduler pass.
6327 @item -fsel-sched-pipelining
6328 @opindex fsel-sched-pipelining
6329 Enable software pipelining of innermost loops during selective scheduling.
6330 This option has no effect until one of @option{-fselective-scheduling} or
6331 @option{-fselective-scheduling2} is turned on.
6333 @item -fsel-sched-pipelining-outer-loops
6334 @opindex fsel-sched-pipelining-outer-loops
6335 When pipelining loops during selective scheduling, also pipeline outer loops.
6336 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6338 @item -fcaller-saves
6339 @opindex fcaller-saves
6340 Enable values to be allocated in registers that will be clobbered by
6341 function calls, by emitting extra instructions to save and restore the
6342 registers around such calls. Such allocation is done only when it
6343 seems to result in better code than would otherwise be produced.
6345 This option is always enabled by default on certain machines, usually
6346 those which have no call-preserved registers to use instead.
6348 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6350 @item -fconserve-stack
6351 @opindex fconserve-stack
6352 Attempt to minimize stack usage. The compiler will attempt to use less
6353 stack space, even if that makes the program slower. This option
6354 implies setting the @option{large-stack-frame} parameter to 100
6355 and the @option{large-stack-frame-growth} parameter to 400.
6357 @item -ftree-reassoc
6358 @opindex ftree-reassoc
6359 Perform reassociation on trees. This flag is enabled by default
6360 at @option{-O} and higher.
6364 Perform partial redundancy elimination (PRE) on trees. This flag is
6365 enabled by default at @option{-O2} and @option{-O3}.
6367 @item -ftree-forwprop
6368 @opindex ftree-forwprop
6369 Perform forward propagation on trees. This flag is enabled by default
6370 at @option{-O} and higher.
6374 Perform full redundancy elimination (FRE) on trees. The difference
6375 between FRE and PRE is that FRE only considers expressions
6376 that are computed on all paths leading to the redundant computation.
6377 This analysis is faster than PRE, though it exposes fewer redundancies.
6378 This flag is enabled by default at @option{-O} and higher.
6380 @item -ftree-phiprop
6381 @opindex ftree-phiprop
6382 Perform hoisting of loads from conditional pointers on trees. This
6383 pass is enabled by default at @option{-O} and higher.
6385 @item -ftree-copy-prop
6386 @opindex ftree-copy-prop
6387 Perform copy propagation on trees. This pass eliminates unnecessary
6388 copy operations. This flag is enabled by default at @option{-O} and
6391 @item -fipa-pure-const
6392 @opindex fipa-pure-const
6393 Discover which functions are pure or constant.
6394 Enabled by default at @option{-O} and higher.
6396 @item -fipa-reference
6397 @opindex fipa-reference
6398 Discover which static variables do not escape cannot escape the
6400 Enabled by default at @option{-O} and higher.
6402 @item -fipa-struct-reorg
6403 @opindex fipa-struct-reorg
6404 Perform structure reorganization optimization, that change C-like structures
6405 layout in order to better utilize spatial locality. This transformation is
6406 affective for programs containing arrays of structures. Available in two
6407 compilation modes: profile-based (enabled with @option{-fprofile-generate})
6408 or static (which uses built-in heuristics). Require @option{-fipa-type-escape}
6409 to provide the safety of this transformation. It works only in whole program
6410 mode, so it requires @option{-fwhole-program} and @option{-combine} to be
6411 enabled. Structures considered @samp{cold} by this transformation are not
6412 affected (see @option{--param struct-reorg-cold-struct-ratio=@var{value}}).
6414 With this flag, the program debug info reflects a new structure layout.
6418 Perform interprocedural pointer analysis. This option is experimental
6419 and does not affect generated code.
6423 Perform interprocedural constant propagation.
6424 This optimization analyzes the program to determine when values passed
6425 to functions are constants and then optimizes accordingly.
6426 This optimization can substantially increase performance
6427 if the application has constants passed to functions.
6428 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
6430 @item -fipa-cp-clone
6431 @opindex fipa-cp-clone
6432 Perform function cloning to make interprocedural constant propagation stronger.
6433 When enabled, interprocedural constant propagation will perform function cloning
6434 when externally visible function can be called with constant arguments.
6435 Because this optimization can create multiple copies of functions,
6436 it may significantly increase code size
6437 (see @option{--param ipcp-unit-growth=@var{value}}).
6438 This flag is enabled by default at @option{-O3}.
6440 @item -fipa-matrix-reorg
6441 @opindex fipa-matrix-reorg
6442 Perform matrix flattening and transposing.
6443 Matrix flattening tries to replace an @math{m}-dimensional matrix
6444 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
6445 This reduces the level of indirection needed for accessing the elements
6446 of the matrix. The second optimization is matrix transposing that
6447 attempts to change the order of the matrix's dimensions in order to
6448 improve cache locality.
6449 Both optimizations need the @option{-fwhole-program} flag.
6450 Transposing is enabled only if profiling information is available.
6454 Perform forward store motion on trees. This flag is
6455 enabled by default at @option{-O} and higher.
6459 Perform sparse conditional constant propagation (CCP) on trees. This
6460 pass only operates on local scalar variables and is enabled by default
6461 at @option{-O} and higher.
6463 @item -ftree-switch-conversion
6464 Perform conversion of simple initializations in a switch to
6465 initializations from a scalar array. This flag is enabled by default
6466 at @option{-O2} and higher.
6470 Perform dead code elimination (DCE) on trees. This flag is enabled by
6471 default at @option{-O} and higher.
6473 @item -ftree-builtin-call-dce
6474 @opindex ftree-builtin-call-dce
6475 Perform conditional dead code elimination (DCE) for calls to builtin functions
6476 that may set @code{errno} but are otherwise side-effect free. This flag is
6477 enabled by default at @option{-O2} and higher if @option{-Os} is not also
6480 @item -ftree-dominator-opts
6481 @opindex ftree-dominator-opts
6482 Perform a variety of simple scalar cleanups (constant/copy
6483 propagation, redundancy elimination, range propagation and expression
6484 simplification) based on a dominator tree traversal. This also
6485 performs jump threading (to reduce jumps to jumps). This flag is
6486 enabled by default at @option{-O} and higher.
6490 Perform dead store elimination (DSE) on trees. A dead store is a store into
6491 a memory location which will later be overwritten by another store without
6492 any intervening loads. In this case the earlier store can be deleted. This
6493 flag is enabled by default at @option{-O} and higher.
6497 Perform loop header copying on trees. This is beneficial since it increases
6498 effectiveness of code motion optimizations. It also saves one jump. This flag
6499 is enabled by default at @option{-O} and higher. It is not enabled
6500 for @option{-Os}, since it usually increases code size.
6502 @item -ftree-loop-optimize
6503 @opindex ftree-loop-optimize
6504 Perform loop optimizations on trees. This flag is enabled by default
6505 at @option{-O} and higher.
6507 @item -ftree-loop-linear
6508 @opindex ftree-loop-linear
6509 Perform linear loop transformations on tree. This flag can improve cache
6510 performance and allow further loop optimizations to take place.
6512 @item -floop-interchange
6513 Perform loop interchange transformations on loops. Interchanging two
6514 nested loops switches the inner and outer loops. For example, given a
6519 A(J, I) = A(J, I) * C
6523 loop interchange will transform the loop as if the user had written:
6527 A(J, I) = A(J, I) * C
6531 which can be beneficial when @code{N} is larger than the caches,
6532 because in Fortran, the elements of an array are stored in memory
6533 contiguously by column, and the original loop iterates over rows,
6534 potentially creating at each access a cache miss. This optimization
6535 applies to all the languages supported by GCC and is not limited to
6536 Fortran. To use this code transformation, GCC has to be configured
6537 with @option{--with-ppl} and @option{--with-cloog} to enable the
6538 Graphite loop transformation infrastructure.
6540 @item -floop-strip-mine
6541 Perform loop strip mining transformations on loops. Strip mining
6542 splits a loop into two nested loops. The outer loop has strides
6543 equal to the strip size and the inner loop has strides of the
6544 original loop within a strip. For example, given a loop like:
6550 loop strip mining will transform the loop as if the user had written:
6553 DO I = II, min (II + 3, N)
6558 This optimization applies to all the languages supported by GCC and is
6559 not limited to Fortran. To use this code transformation, GCC has to
6560 be configured with @option{--with-ppl} and @option{--with-cloog} to
6561 enable the Graphite loop transformation infrastructure.
6564 Perform loop blocking transformations on loops. Blocking strip mines
6565 each loop in the loop nest such that the memory accesses of the
6566 element loops fit inside caches. For example, given a loop like:
6570 A(J, I) = B(I) + C(J)
6574 loop blocking will transform the loop as if the user had written:
6578 DO I = II, min (II + 63, N)
6579 DO J = JJ, min (JJ + 63, M)
6580 A(J, I) = B(I) + C(J)
6586 which can be beneficial when @code{M} is larger than the caches,
6587 because the innermost loop will iterate over a smaller amount of data
6588 that can be kept in the caches. This optimization applies to all the
6589 languages supported by GCC and is not limited to Fortran. To use this
6590 code transformation, GCC has to be configured with @option{--with-ppl}
6591 and @option{--with-cloog} to enable the Graphite loop transformation
6594 @item -fcheck-data-deps
6595 @opindex fcheck-data-deps
6596 Compare the results of several data dependence analyzers. This option
6597 is used for debugging the data dependence analyzers.
6599 @item -ftree-loop-distribution
6600 Perform loop distribution. This flag can improve cache performance on
6601 big loop bodies and allow further loop optimizations, like
6602 parallelization or vectorization, to take place. For example, the loop
6619 @item -ftree-loop-im
6620 @opindex ftree-loop-im
6621 Perform loop invariant motion on trees. This pass moves only invariants that
6622 would be hard to handle at RTL level (function calls, operations that expand to
6623 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
6624 operands of conditions that are invariant out of the loop, so that we can use
6625 just trivial invariantness analysis in loop unswitching. The pass also includes
6628 @item -ftree-loop-ivcanon
6629 @opindex ftree-loop-ivcanon
6630 Create a canonical counter for number of iterations in the loop for that
6631 determining number of iterations requires complicated analysis. Later
6632 optimizations then may determine the number easily. Useful especially
6633 in connection with unrolling.
6637 Perform induction variable optimizations (strength reduction, induction
6638 variable merging and induction variable elimination) on trees.
6640 @item -ftree-parallelize-loops=n
6641 @opindex ftree-parallelize-loops
6642 Parallelize loops, i.e., split their iteration space to run in n threads.
6643 This is only possible for loops whose iterations are independent
6644 and can be arbitrarily reordered. The optimization is only
6645 profitable on multiprocessor machines, for loops that are CPU-intensive,
6646 rather than constrained e.g.@: by memory bandwidth. This option
6647 implies @option{-pthread}, and thus is only supported on targets
6648 that have support for @option{-pthread}.
6652 Perform function-local points-to analysis on trees. This flag is
6653 enabled by default at @option{-O} and higher.
6657 Perform scalar replacement of aggregates. This pass replaces structure
6658 references with scalars to prevent committing structures to memory too
6659 early. This flag is enabled by default at @option{-O} and higher.
6661 @item -ftree-copyrename
6662 @opindex ftree-copyrename
6663 Perform copy renaming on trees. This pass attempts to rename compiler
6664 temporaries to other variables at copy locations, usually resulting in
6665 variable names which more closely resemble the original variables. This flag
6666 is enabled by default at @option{-O} and higher.
6670 Perform temporary expression replacement during the SSA->normal phase. Single
6671 use/single def temporaries are replaced at their use location with their
6672 defining expression. This results in non-GIMPLE code, but gives the expanders
6673 much more complex trees to work on resulting in better RTL generation. This is
6674 enabled by default at @option{-O} and higher.
6676 @item -ftree-vectorize
6677 @opindex ftree-vectorize
6678 Perform loop vectorization on trees. This flag is enabled by default at
6681 @item -ftree-vect-loop-version
6682 @opindex ftree-vect-loop-version
6683 Perform loop versioning when doing loop vectorization on trees. When a loop
6684 appears to be vectorizable except that data alignment or data dependence cannot
6685 be determined at compile time then vectorized and non-vectorized versions of
6686 the loop are generated along with runtime checks for alignment or dependence
6687 to control which version is executed. This option is enabled by default
6688 except at level @option{-Os} where it is disabled.
6690 @item -fvect-cost-model
6691 @opindex fvect-cost-model
6692 Enable cost model for vectorization.
6696 Perform Value Range Propagation on trees. This is similar to the
6697 constant propagation pass, but instead of values, ranges of values are
6698 propagated. This allows the optimizers to remove unnecessary range
6699 checks like array bound checks and null pointer checks. This is
6700 enabled by default at @option{-O2} and higher. Null pointer check
6701 elimination is only done if @option{-fdelete-null-pointer-checks} is
6706 Perform tail duplication to enlarge superblock size. This transformation
6707 simplifies the control flow of the function allowing other optimizations to do
6710 @item -funroll-loops
6711 @opindex funroll-loops
6712 Unroll loops whose number of iterations can be determined at compile
6713 time or upon entry to the loop. @option{-funroll-loops} implies
6714 @option{-frerun-cse-after-loop}. This option makes code larger,
6715 and may or may not make it run faster.
6717 @item -funroll-all-loops
6718 @opindex funroll-all-loops
6719 Unroll all loops, even if their number of iterations is uncertain when
6720 the loop is entered. This usually makes programs run more slowly.
6721 @option{-funroll-all-loops} implies the same options as
6722 @option{-funroll-loops},
6724 @item -fsplit-ivs-in-unroller
6725 @opindex fsplit-ivs-in-unroller
6726 Enables expressing of values of induction variables in later iterations
6727 of the unrolled loop using the value in the first iteration. This breaks
6728 long dependency chains, thus improving efficiency of the scheduling passes.
6730 Combination of @option{-fweb} and CSE is often sufficient to obtain the
6731 same effect. However in cases the loop body is more complicated than
6732 a single basic block, this is not reliable. It also does not work at all
6733 on some of the architectures due to restrictions in the CSE pass.
6735 This optimization is enabled by default.
6737 @item -fvariable-expansion-in-unroller
6738 @opindex fvariable-expansion-in-unroller
6739 With this option, the compiler will create multiple copies of some
6740 local variables when unrolling a loop which can result in superior code.
6742 @item -fpredictive-commoning
6743 @opindex fpredictive-commoning
6744 Perform predictive commoning optimization, i.e., reusing computations
6745 (especially memory loads and stores) performed in previous
6746 iterations of loops.
6748 This option is enabled at level @option{-O3}.
6750 @item -fprefetch-loop-arrays
6751 @opindex fprefetch-loop-arrays
6752 If supported by the target machine, generate instructions to prefetch
6753 memory to improve the performance of loops that access large arrays.
6755 This option may generate better or worse code; results are highly
6756 dependent on the structure of loops within the source code.
6758 Disabled at level @option{-Os}.
6761 @itemx -fno-peephole2
6762 @opindex fno-peephole
6763 @opindex fno-peephole2
6764 Disable any machine-specific peephole optimizations. The difference
6765 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
6766 are implemented in the compiler; some targets use one, some use the
6767 other, a few use both.
6769 @option{-fpeephole} is enabled by default.
6770 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6772 @item -fno-guess-branch-probability
6773 @opindex fno-guess-branch-probability
6774 Do not guess branch probabilities using heuristics.
6776 GCC will use heuristics to guess branch probabilities if they are
6777 not provided by profiling feedback (@option{-fprofile-arcs}). These
6778 heuristics are based on the control flow graph. If some branch probabilities
6779 are specified by @samp{__builtin_expect}, then the heuristics will be
6780 used to guess branch probabilities for the rest of the control flow graph,
6781 taking the @samp{__builtin_expect} info into account. The interactions
6782 between the heuristics and @samp{__builtin_expect} can be complex, and in
6783 some cases, it may be useful to disable the heuristics so that the effects
6784 of @samp{__builtin_expect} are easier to understand.
6786 The default is @option{-fguess-branch-probability} at levels
6787 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6789 @item -freorder-blocks
6790 @opindex freorder-blocks
6791 Reorder basic blocks in the compiled function in order to reduce number of
6792 taken branches and improve code locality.
6794 Enabled at levels @option{-O2}, @option{-O3}.
6796 @item -freorder-blocks-and-partition
6797 @opindex freorder-blocks-and-partition
6798 In addition to reordering basic blocks in the compiled function, in order
6799 to reduce number of taken branches, partitions hot and cold basic blocks
6800 into separate sections of the assembly and .o files, to improve
6801 paging and cache locality performance.
6803 This optimization is automatically turned off in the presence of
6804 exception handling, for linkonce sections, for functions with a user-defined
6805 section attribute and on any architecture that does not support named
6808 @item -freorder-functions
6809 @opindex freorder-functions
6810 Reorder functions in the object file in order to
6811 improve code locality. This is implemented by using special
6812 subsections @code{.text.hot} for most frequently executed functions and
6813 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
6814 the linker so object file format must support named sections and linker must
6815 place them in a reasonable way.
6817 Also profile feedback must be available in to make this option effective. See
6818 @option{-fprofile-arcs} for details.
6820 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6822 @item -fstrict-aliasing
6823 @opindex fstrict-aliasing
6824 Allows the compiler to assume the strictest aliasing rules applicable to
6825 the language being compiled. For C (and C++), this activates
6826 optimizations based on the type of expressions. In particular, an
6827 object of one type is assumed never to reside at the same address as an
6828 object of a different type, unless the types are almost the same. For
6829 example, an @code{unsigned int} can alias an @code{int}, but not a
6830 @code{void*} or a @code{double}. A character type may alias any other
6833 @anchor{Type-punning}Pay special attention to code like this:
6846 The practice of reading from a different union member than the one most
6847 recently written to (called ``type-punning'') is common. Even with
6848 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
6849 is accessed through the union type. So, the code above will work as
6850 expected. @xref{Structures unions enumerations and bit-fields
6851 implementation}. However, this code might not:
6862 Similarly, access by taking the address, casting the resulting pointer
6863 and dereferencing the result has undefined behavior, even if the cast
6864 uses a union type, e.g.:
6868 return ((union a_union *) &d)->i;
6872 The @option{-fstrict-aliasing} option is enabled at levels
6873 @option{-O2}, @option{-O3}, @option{-Os}.
6875 @item -fstrict-overflow
6876 @opindex fstrict-overflow
6877 Allow the compiler to assume strict signed overflow rules, depending
6878 on the language being compiled. For C (and C++) this means that
6879 overflow when doing arithmetic with signed numbers is undefined, which
6880 means that the compiler may assume that it will not happen. This
6881 permits various optimizations. For example, the compiler will assume
6882 that an expression like @code{i + 10 > i} will always be true for
6883 signed @code{i}. This assumption is only valid if signed overflow is
6884 undefined, as the expression is false if @code{i + 10} overflows when
6885 using twos complement arithmetic. When this option is in effect any
6886 attempt to determine whether an operation on signed numbers will
6887 overflow must be written carefully to not actually involve overflow.
6889 This option also allows the compiler to assume strict pointer
6890 semantics: given a pointer to an object, if adding an offset to that
6891 pointer does not produce a pointer to the same object, the addition is
6892 undefined. This permits the compiler to conclude that @code{p + u >
6893 p} is always true for a pointer @code{p} and unsigned integer
6894 @code{u}. This assumption is only valid because pointer wraparound is
6895 undefined, as the expression is false if @code{p + u} overflows using
6896 twos complement arithmetic.
6898 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
6899 that integer signed overflow is fully defined: it wraps. When
6900 @option{-fwrapv} is used, there is no difference between
6901 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
6902 integers. With @option{-fwrapv} certain types of overflow are
6903 permitted. For example, if the compiler gets an overflow when doing
6904 arithmetic on constants, the overflowed value can still be used with
6905 @option{-fwrapv}, but not otherwise.
6907 The @option{-fstrict-overflow} option is enabled at levels
6908 @option{-O2}, @option{-O3}, @option{-Os}.
6910 @item -falign-functions
6911 @itemx -falign-functions=@var{n}
6912 @opindex falign-functions
6913 Align the start of functions to the next power-of-two greater than
6914 @var{n}, skipping up to @var{n} bytes. For instance,
6915 @option{-falign-functions=32} aligns functions to the next 32-byte
6916 boundary, but @option{-falign-functions=24} would align to the next
6917 32-byte boundary only if this can be done by skipping 23 bytes or less.
6919 @option{-fno-align-functions} and @option{-falign-functions=1} are
6920 equivalent and mean that functions will not be aligned.
6922 Some assemblers only support this flag when @var{n} is a power of two;
6923 in that case, it is rounded up.
6925 If @var{n} is not specified or is zero, use a machine-dependent default.
6927 Enabled at levels @option{-O2}, @option{-O3}.
6929 @item -falign-labels
6930 @itemx -falign-labels=@var{n}
6931 @opindex falign-labels
6932 Align all branch targets to a power-of-two boundary, skipping up to
6933 @var{n} bytes like @option{-falign-functions}. This option can easily
6934 make code slower, because it must insert dummy operations for when the
6935 branch target is reached in the usual flow of the code.
6937 @option{-fno-align-labels} and @option{-falign-labels=1} are
6938 equivalent and mean that labels will not be aligned.
6940 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
6941 are greater than this value, then their values are used instead.
6943 If @var{n} is not specified or is zero, use a machine-dependent default
6944 which is very likely to be @samp{1}, meaning no alignment.
6946 Enabled at levels @option{-O2}, @option{-O3}.
6949 @itemx -falign-loops=@var{n}
6950 @opindex falign-loops
6951 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
6952 like @option{-falign-functions}. The hope is that the loop will be
6953 executed many times, which will make up for any execution of the dummy
6956 @option{-fno-align-loops} and @option{-falign-loops=1} are
6957 equivalent and mean that loops will not be aligned.
6959 If @var{n} is not specified or is zero, use a machine-dependent default.
6961 Enabled at levels @option{-O2}, @option{-O3}.
6964 @itemx -falign-jumps=@var{n}
6965 @opindex falign-jumps
6966 Align branch targets to a power-of-two boundary, for branch targets
6967 where the targets can only be reached by jumping, skipping up to @var{n}
6968 bytes like @option{-falign-functions}. In this case, no dummy operations
6971 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
6972 equivalent and mean that loops will not be aligned.
6974 If @var{n} is not specified or is zero, use a machine-dependent default.
6976 Enabled at levels @option{-O2}, @option{-O3}.
6978 @item -funit-at-a-time
6979 @opindex funit-at-a-time
6980 This option is left for compatibility reasons. @option{-funit-at-a-time}
6981 has no effect, while @option{-fno-unit-at-a-time} implies
6982 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
6986 @item -fno-toplevel-reorder
6987 @opindex fno-toplevel-reorder
6988 Do not reorder top-level functions, variables, and @code{asm}
6989 statements. Output them in the same order that they appear in the
6990 input file. When this option is used, unreferenced static variables
6991 will not be removed. This option is intended to support existing code
6992 which relies on a particular ordering. For new code, it is better to
6995 Enabled at level @option{-O0}. When disabled explicitly, it also imply
6996 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
7001 Constructs webs as commonly used for register allocation purposes and assign
7002 each web individual pseudo register. This allows the register allocation pass
7003 to operate on pseudos directly, but also strengthens several other optimization
7004 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7005 however, make debugging impossible, since variables will no longer stay in a
7008 Enabled by default with @option{-funroll-loops}.
7010 @item -fwhole-program
7011 @opindex fwhole-program
7012 Assume that the current compilation unit represents the whole program being
7013 compiled. All public functions and variables with the exception of @code{main}
7014 and those merged by attribute @code{externally_visible} become static functions
7015 and in effect are optimized more aggressively by interprocedural optimizers.
7016 While this option is equivalent to proper use of the @code{static} keyword for
7017 programs consisting of a single file, in combination with option
7018 @option{--combine} this flag can be used to compile many smaller scale C
7019 programs since the functions and variables become local for the whole combined
7020 compilation unit, not for the single source file itself.
7022 This option implies @option{-fwhole-file} for Fortran programs.
7024 @item -fcprop-registers
7025 @opindex fcprop-registers
7026 After register allocation and post-register allocation instruction splitting,
7027 we perform a copy-propagation pass to try to reduce scheduling dependencies
7028 and occasionally eliminate the copy.
7030 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7032 @item -fprofile-correction
7033 @opindex fprofile-correction
7034 Profiles collected using an instrumented binary for multi-threaded programs may
7035 be inconsistent due to missed counter updates. When this option is specified,
7036 GCC will use heuristics to correct or smooth out such inconsistencies. By
7037 default, GCC will emit an error message when an inconsistent profile is detected.
7039 @item -fprofile-dir=@var{path}
7040 @opindex fprofile-dir
7042 Set the directory to search the profile data files in to @var{path}.
7043 This option affects only the profile data generated by
7044 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
7045 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
7046 and its related options.
7047 By default, GCC will use the current directory as @var{path}
7048 thus the profile data file will appear in the same directory as the object file.
7050 @item -fprofile-generate
7051 @itemx -fprofile-generate=@var{path}
7052 @opindex fprofile-generate
7054 Enable options usually used for instrumenting application to produce
7055 profile useful for later recompilation with profile feedback based
7056 optimization. You must use @option{-fprofile-generate} both when
7057 compiling and when linking your program.
7059 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
7061 If @var{path} is specified, GCC will look at the @var{path} to find
7062 the profile feedback data files. See @option{-fprofile-dir}.
7065 @itemx -fprofile-use=@var{path}
7066 @opindex fprofile-use
7067 Enable profile feedback directed optimizations, and optimizations
7068 generally profitable only with profile feedback available.
7070 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
7071 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
7073 By default, GCC emits an error message if the feedback profiles do not
7074 match the source code. This error can be turned into a warning by using
7075 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
7078 If @var{path} is specified, GCC will look at the @var{path} to find
7079 the profile feedback data files. See @option{-fprofile-dir}.
7082 The following options control compiler behavior regarding floating
7083 point arithmetic. These options trade off between speed and
7084 correctness. All must be specifically enabled.
7088 @opindex ffloat-store
7089 Do not store floating point variables in registers, and inhibit other
7090 options that might change whether a floating point value is taken from a
7093 @cindex floating point precision
7094 This option prevents undesirable excess precision on machines such as
7095 the 68000 where the floating registers (of the 68881) keep more
7096 precision than a @code{double} is supposed to have. Similarly for the
7097 x86 architecture. For most programs, the excess precision does only
7098 good, but a few programs rely on the precise definition of IEEE floating
7099 point. Use @option{-ffloat-store} for such programs, after modifying
7100 them to store all pertinent intermediate computations into variables.
7102 @item -fexcess-precision=@var{style}
7103 @opindex fexcess-precision
7104 This option allows further control over excess precision on machines
7105 where floating-point registers have more precision than the IEEE
7106 @code{float} and @code{double} types and the processor does not
7107 support operations rounding to those types. By default,
7108 @option{-fexcess-precision=fast} is in effect; this means that
7109 operations are carried out in the precision of the registers and that
7110 it is unpredictable when rounding to the types specified in the source
7111 code takes place. When compiling C, if
7112 @option{-fexcess-precision=standard} is specified then excess
7113 precision will follow the rules specified in ISO C99; in particular,
7114 both casts and assignments cause values to be rounded to their
7115 semantic types (whereas @option{-ffloat-store} only affects
7116 assignments). This option is enabled by default for C if a strict
7117 conformance option such as @option{-std=c99} is used.
7120 @option{-fexcess-precision=standard} is not implemented for languages
7121 other than C, and has no effect if
7122 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
7123 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
7124 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
7125 semantics apply without excess precision, and in the latter, rounding
7130 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
7131 @option{-ffinite-math-only}, @option{-fno-rounding-math},
7132 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
7134 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
7136 This option is not turned on by any @option{-O} option since
7137 it can result in incorrect output for programs which depend on
7138 an exact implementation of IEEE or ISO rules/specifications for
7139 math functions. It may, however, yield faster code for programs
7140 that do not require the guarantees of these specifications.
7142 @item -fno-math-errno
7143 @opindex fno-math-errno
7144 Do not set ERRNO after calling math functions that are executed
7145 with a single instruction, e.g., sqrt. A program that relies on
7146 IEEE exceptions for math error handling may want to use this flag
7147 for speed while maintaining IEEE arithmetic compatibility.
7149 This option is not turned on by any @option{-O} option since
7150 it can result in incorrect output for programs which depend on
7151 an exact implementation of IEEE or ISO rules/specifications for
7152 math functions. It may, however, yield faster code for programs
7153 that do not require the guarantees of these specifications.
7155 The default is @option{-fmath-errno}.
7157 On Darwin systems, the math library never sets @code{errno}. There is
7158 therefore no reason for the compiler to consider the possibility that
7159 it might, and @option{-fno-math-errno} is the default.
7161 @item -funsafe-math-optimizations
7162 @opindex funsafe-math-optimizations
7164 Allow optimizations for floating-point arithmetic that (a) assume
7165 that arguments and results are valid and (b) may violate IEEE or
7166 ANSI standards. When used at link-time, it may include libraries
7167 or startup files that change the default FPU control word or other
7168 similar optimizations.
7170 This option is not turned on by any @option{-O} option since
7171 it can result in incorrect output for programs which depend on
7172 an exact implementation of IEEE or ISO rules/specifications for
7173 math functions. It may, however, yield faster code for programs
7174 that do not require the guarantees of these specifications.
7175 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
7176 @option{-fassociative-math} and @option{-freciprocal-math}.
7178 The default is @option{-fno-unsafe-math-optimizations}.
7180 @item -fassociative-math
7181 @opindex fassociative-math
7183 Allow re-association of operands in series of floating-point operations.
7184 This violates the ISO C and C++ language standard by possibly changing
7185 computation result. NOTE: re-ordering may change the sign of zero as
7186 well as ignore NaNs and inhibit or create underflow or overflow (and
7187 thus cannot be used on a code which relies on rounding behavior like
7188 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
7189 and thus may not be used when ordered comparisons are required.
7190 This option requires that both @option{-fno-signed-zeros} and
7191 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
7192 much sense with @option{-frounding-math}.
7194 The default is @option{-fno-associative-math}.
7196 @item -freciprocal-math
7197 @opindex freciprocal-math
7199 Allow the reciprocal of a value to be used instead of dividing by
7200 the value if this enables optimizations. For example @code{x / y}
7201 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
7202 is subject to common subexpression elimination. Note that this loses
7203 precision and increases the number of flops operating on the value.
7205 The default is @option{-fno-reciprocal-math}.
7207 @item -ffinite-math-only
7208 @opindex ffinite-math-only
7209 Allow optimizations for floating-point arithmetic that assume
7210 that arguments and results are not NaNs or +-Infs.
7212 This option is not turned on by any @option{-O} option since
7213 it can result in incorrect output for programs which depend on
7214 an exact implementation of IEEE or ISO rules/specifications for
7215 math functions. It may, however, yield faster code for programs
7216 that do not require the guarantees of these specifications.
7218 The default is @option{-fno-finite-math-only}.
7220 @item -fno-signed-zeros
7221 @opindex fno-signed-zeros
7222 Allow optimizations for floating point arithmetic that ignore the
7223 signedness of zero. IEEE arithmetic specifies the behavior of
7224 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
7225 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
7226 This option implies that the sign of a zero result isn't significant.
7228 The default is @option{-fsigned-zeros}.
7230 @item -fno-trapping-math
7231 @opindex fno-trapping-math
7232 Compile code assuming that floating-point operations cannot generate
7233 user-visible traps. These traps include division by zero, overflow,
7234 underflow, inexact result and invalid operation. This option requires
7235 that @option{-fno-signaling-nans} be in effect. Setting this option may
7236 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
7238 This option should never be turned on by any @option{-O} option since
7239 it can result in incorrect output for programs which depend on
7240 an exact implementation of IEEE or ISO rules/specifications for
7243 The default is @option{-ftrapping-math}.
7245 @item -frounding-math
7246 @opindex frounding-math
7247 Disable transformations and optimizations that assume default floating
7248 point rounding behavior. This is round-to-zero for all floating point
7249 to integer conversions, and round-to-nearest for all other arithmetic
7250 truncations. This option should be specified for programs that change
7251 the FP rounding mode dynamically, or that may be executed with a
7252 non-default rounding mode. This option disables constant folding of
7253 floating point expressions at compile-time (which may be affected by
7254 rounding mode) and arithmetic transformations that are unsafe in the
7255 presence of sign-dependent rounding modes.
7257 The default is @option{-fno-rounding-math}.
7259 This option is experimental and does not currently guarantee to
7260 disable all GCC optimizations that are affected by rounding mode.
7261 Future versions of GCC may provide finer control of this setting
7262 using C99's @code{FENV_ACCESS} pragma. This command line option
7263 will be used to specify the default state for @code{FENV_ACCESS}.
7265 @item -fsignaling-nans
7266 @opindex fsignaling-nans
7267 Compile code assuming that IEEE signaling NaNs may generate user-visible
7268 traps during floating-point operations. Setting this option disables
7269 optimizations that may change the number of exceptions visible with
7270 signaling NaNs. This option implies @option{-ftrapping-math}.
7272 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
7275 The default is @option{-fno-signaling-nans}.
7277 This option is experimental and does not currently guarantee to
7278 disable all GCC optimizations that affect signaling NaN behavior.
7280 @item -fsingle-precision-constant
7281 @opindex fsingle-precision-constant
7282 Treat floating point constant as single precision constant instead of
7283 implicitly converting it to double precision constant.
7285 @item -fcx-limited-range
7286 @opindex fcx-limited-range
7287 When enabled, this option states that a range reduction step is not
7288 needed when performing complex division. Also, there is no checking
7289 whether the result of a complex multiplication or division is @code{NaN
7290 + I*NaN}, with an attempt to rescue the situation in that case. The
7291 default is @option{-fno-cx-limited-range}, but is enabled by
7292 @option{-ffast-math}.
7294 This option controls the default setting of the ISO C99
7295 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
7298 @item -fcx-fortran-rules
7299 @opindex fcx-fortran-rules
7300 Complex multiplication and division follow Fortran rules. Range
7301 reduction is done as part of complex division, but there is no checking
7302 whether the result of a complex multiplication or division is @code{NaN
7303 + I*NaN}, with an attempt to rescue the situation in that case.
7305 The default is @option{-fno-cx-fortran-rules}.
7309 The following options control optimizations that may improve
7310 performance, but are not enabled by any @option{-O} options. This
7311 section includes experimental options that may produce broken code.
7314 @item -fbranch-probabilities
7315 @opindex fbranch-probabilities
7316 After running a program compiled with @option{-fprofile-arcs}
7317 (@pxref{Debugging Options,, Options for Debugging Your Program or
7318 @command{gcc}}), you can compile it a second time using
7319 @option{-fbranch-probabilities}, to improve optimizations based on
7320 the number of times each branch was taken. When the program
7321 compiled with @option{-fprofile-arcs} exits it saves arc execution
7322 counts to a file called @file{@var{sourcename}.gcda} for each source
7323 file. The information in this data file is very dependent on the
7324 structure of the generated code, so you must use the same source code
7325 and the same optimization options for both compilations.
7327 With @option{-fbranch-probabilities}, GCC puts a
7328 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
7329 These can be used to improve optimization. Currently, they are only
7330 used in one place: in @file{reorg.c}, instead of guessing which path a
7331 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
7332 exactly determine which path is taken more often.
7334 @item -fprofile-values
7335 @opindex fprofile-values
7336 If combined with @option{-fprofile-arcs}, it adds code so that some
7337 data about values of expressions in the program is gathered.
7339 With @option{-fbranch-probabilities}, it reads back the data gathered
7340 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
7341 notes to instructions for their later usage in optimizations.
7343 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
7347 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
7348 a code to gather information about values of expressions.
7350 With @option{-fbranch-probabilities}, it reads back the data gathered
7351 and actually performs the optimizations based on them.
7352 Currently the optimizations include specialization of division operation
7353 using the knowledge about the value of the denominator.
7355 @item -frename-registers
7356 @opindex frename-registers
7357 Attempt to avoid false dependencies in scheduled code by making use
7358 of registers left over after register allocation. This optimization
7359 will most benefit processors with lots of registers. Depending on the
7360 debug information format adopted by the target, however, it can
7361 make debugging impossible, since variables will no longer stay in
7362 a ``home register''.
7364 Enabled by default with @option{-funroll-loops}.
7368 Perform tail duplication to enlarge superblock size. This transformation
7369 simplifies the control flow of the function allowing other optimizations to do
7372 Enabled with @option{-fprofile-use}.
7374 @item -funroll-loops
7375 @opindex funroll-loops
7376 Unroll loops whose number of iterations can be determined at compile time or
7377 upon entry to the loop. @option{-funroll-loops} implies
7378 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
7379 It also turns on complete loop peeling (i.e.@: complete removal of loops with
7380 small constant number of iterations). This option makes code larger, and may
7381 or may not make it run faster.
7383 Enabled with @option{-fprofile-use}.
7385 @item -funroll-all-loops
7386 @opindex funroll-all-loops
7387 Unroll all loops, even if their number of iterations is uncertain when
7388 the loop is entered. This usually makes programs run more slowly.
7389 @option{-funroll-all-loops} implies the same options as
7390 @option{-funroll-loops}.
7393 @opindex fpeel-loops
7394 Peels the loops for that there is enough information that they do not
7395 roll much (from profile feedback). It also turns on complete loop peeling
7396 (i.e.@: complete removal of loops with small constant number of iterations).
7398 Enabled with @option{-fprofile-use}.
7400 @item -fmove-loop-invariants
7401 @opindex fmove-loop-invariants
7402 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
7403 at level @option{-O1}
7405 @item -funswitch-loops
7406 @opindex funswitch-loops
7407 Move branches with loop invariant conditions out of the loop, with duplicates
7408 of the loop on both branches (modified according to result of the condition).
7410 @item -ffunction-sections
7411 @itemx -fdata-sections
7412 @opindex ffunction-sections
7413 @opindex fdata-sections
7414 Place each function or data item into its own section in the output
7415 file if the target supports arbitrary sections. The name of the
7416 function or the name of the data item determines the section's name
7419 Use these options on systems where the linker can perform optimizations
7420 to improve locality of reference in the instruction space. Most systems
7421 using the ELF object format and SPARC processors running Solaris 2 have
7422 linkers with such optimizations. AIX may have these optimizations in
7425 Only use these options when there are significant benefits from doing
7426 so. When you specify these options, the assembler and linker will
7427 create larger object and executable files and will also be slower.
7428 You will not be able to use @code{gprof} on all systems if you
7429 specify this option and you may have problems with debugging if
7430 you specify both this option and @option{-g}.
7432 @item -fbranch-target-load-optimize
7433 @opindex fbranch-target-load-optimize
7434 Perform branch target register load optimization before prologue / epilogue
7436 The use of target registers can typically be exposed only during reload,
7437 thus hoisting loads out of loops and doing inter-block scheduling needs
7438 a separate optimization pass.
7440 @item -fbranch-target-load-optimize2
7441 @opindex fbranch-target-load-optimize2
7442 Perform branch target register load optimization after prologue / epilogue
7445 @item -fbtr-bb-exclusive
7446 @opindex fbtr-bb-exclusive
7447 When performing branch target register load optimization, don't reuse
7448 branch target registers in within any basic block.
7450 @item -fstack-protector
7451 @opindex fstack-protector
7452 Emit extra code to check for buffer overflows, such as stack smashing
7453 attacks. This is done by adding a guard variable to functions with
7454 vulnerable objects. This includes functions that call alloca, and
7455 functions with buffers larger than 8 bytes. The guards are initialized
7456 when a function is entered and then checked when the function exits.
7457 If a guard check fails, an error message is printed and the program exits.
7459 @item -fstack-protector-all
7460 @opindex fstack-protector-all
7461 Like @option{-fstack-protector} except that all functions are protected.
7463 @item -fsection-anchors
7464 @opindex fsection-anchors
7465 Try to reduce the number of symbolic address calculations by using
7466 shared ``anchor'' symbols to address nearby objects. This transformation
7467 can help to reduce the number of GOT entries and GOT accesses on some
7470 For example, the implementation of the following function @code{foo}:
7474 int foo (void) @{ return a + b + c; @}
7477 would usually calculate the addresses of all three variables, but if you
7478 compile it with @option{-fsection-anchors}, it will access the variables
7479 from a common anchor point instead. The effect is similar to the
7480 following pseudocode (which isn't valid C):
7485 register int *xr = &x;
7486 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
7490 Not all targets support this option.
7492 @item --param @var{name}=@var{value}
7494 In some places, GCC uses various constants to control the amount of
7495 optimization that is done. For example, GCC will not inline functions
7496 that contain more that a certain number of instructions. You can
7497 control some of these constants on the command-line using the
7498 @option{--param} option.
7500 The names of specific parameters, and the meaning of the values, are
7501 tied to the internals of the compiler, and are subject to change
7502 without notice in future releases.
7504 In each case, the @var{value} is an integer. The allowable choices for
7505 @var{name} are given in the following table:
7508 @item struct-reorg-cold-struct-ratio
7509 The threshold ratio (as a percentage) between a structure frequency
7510 and the frequency of the hottest structure in the program. This parameter
7511 is used by struct-reorg optimization enabled by @option{-fipa-struct-reorg}.
7512 We say that if the ratio of a structure frequency, calculated by profiling,
7513 to the hottest structure frequency in the program is less than this
7514 parameter, then structure reorganization is not applied to this structure.
7517 @item predictable-branch-cost-outcome
7518 When branch is predicted to be taken with probability lower than this threshold
7519 (in percent), then it is considered well predictable. The default is 10.
7521 @item max-crossjump-edges
7522 The maximum number of incoming edges to consider for crossjumping.
7523 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
7524 the number of edges incoming to each block. Increasing values mean
7525 more aggressive optimization, making the compile time increase with
7526 probably small improvement in executable size.
7528 @item min-crossjump-insns
7529 The minimum number of instructions which must be matched at the end
7530 of two blocks before crossjumping will be performed on them. This
7531 value is ignored in the case where all instructions in the block being
7532 crossjumped from are matched. The default value is 5.
7534 @item max-grow-copy-bb-insns
7535 The maximum code size expansion factor when copying basic blocks
7536 instead of jumping. The expansion is relative to a jump instruction.
7537 The default value is 8.
7539 @item max-goto-duplication-insns
7540 The maximum number of instructions to duplicate to a block that jumps
7541 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
7542 passes, GCC factors computed gotos early in the compilation process,
7543 and unfactors them as late as possible. Only computed jumps at the
7544 end of a basic blocks with no more than max-goto-duplication-insns are
7545 unfactored. The default value is 8.
7547 @item max-delay-slot-insn-search
7548 The maximum number of instructions to consider when looking for an
7549 instruction to fill a delay slot. If more than this arbitrary number of
7550 instructions is searched, the time savings from filling the delay slot
7551 will be minimal so stop searching. Increasing values mean more
7552 aggressive optimization, making the compile time increase with probably
7553 small improvement in executable run time.
7555 @item max-delay-slot-live-search
7556 When trying to fill delay slots, the maximum number of instructions to
7557 consider when searching for a block with valid live register
7558 information. Increasing this arbitrarily chosen value means more
7559 aggressive optimization, increasing the compile time. This parameter
7560 should be removed when the delay slot code is rewritten to maintain the
7563 @item max-gcse-memory
7564 The approximate maximum amount of memory that will be allocated in
7565 order to perform the global common subexpression elimination
7566 optimization. If more memory than specified is required, the
7567 optimization will not be done.
7569 @item max-pending-list-length
7570 The maximum number of pending dependencies scheduling will allow
7571 before flushing the current state and starting over. Large functions
7572 with few branches or calls can create excessively large lists which
7573 needlessly consume memory and resources.
7575 @item max-inline-insns-single
7576 Several parameters control the tree inliner used in gcc.
7577 This number sets the maximum number of instructions (counted in GCC's
7578 internal representation) in a single function that the tree inliner
7579 will consider for inlining. This only affects functions declared
7580 inline and methods implemented in a class declaration (C++).
7581 The default value is 300.
7583 @item max-inline-insns-auto
7584 When you use @option{-finline-functions} (included in @option{-O3}),
7585 a lot of functions that would otherwise not be considered for inlining
7586 by the compiler will be investigated. To those functions, a different
7587 (more restrictive) limit compared to functions declared inline can
7589 The default value is 60.
7591 @item large-function-insns
7592 The limit specifying really large functions. For functions larger than this
7593 limit after inlining, inlining is constrained by
7594 @option{--param large-function-growth}. This parameter is useful primarily
7595 to avoid extreme compilation time caused by non-linear algorithms used by the
7597 The default value is 2700.
7599 @item large-function-growth
7600 Specifies maximal growth of large function caused by inlining in percents.
7601 The default value is 100 which limits large function growth to 2.0 times
7604 @item large-unit-insns
7605 The limit specifying large translation unit. Growth caused by inlining of
7606 units larger than this limit is limited by @option{--param inline-unit-growth}.
7607 For small units this might be too tight (consider unit consisting of function A
7608 that is inline and B that just calls A three time. If B is small relative to
7609 A, the growth of unit is 300\% and yet such inlining is very sane. For very
7610 large units consisting of small inlineable functions however the overall unit
7611 growth limit is needed to avoid exponential explosion of code size. Thus for
7612 smaller units, the size is increased to @option{--param large-unit-insns}
7613 before applying @option{--param inline-unit-growth}. The default is 10000
7615 @item inline-unit-growth
7616 Specifies maximal overall growth of the compilation unit caused by inlining.
7617 The default value is 30 which limits unit growth to 1.3 times the original
7620 @item ipcp-unit-growth
7621 Specifies maximal overall growth of the compilation unit caused by
7622 interprocedural constant propagation. The default value is 10 which limits
7623 unit growth to 1.1 times the original size.
7625 @item large-stack-frame
7626 The limit specifying large stack frames. While inlining the algorithm is trying
7627 to not grow past this limit too much. Default value is 256 bytes.
7629 @item large-stack-frame-growth
7630 Specifies maximal growth of large stack frames caused by inlining in percents.
7631 The default value is 1000 which limits large stack frame growth to 11 times
7634 @item max-inline-insns-recursive
7635 @itemx max-inline-insns-recursive-auto
7636 Specifies maximum number of instructions out-of-line copy of self recursive inline
7637 function can grow into by performing recursive inlining.
7639 For functions declared inline @option{--param max-inline-insns-recursive} is
7640 taken into account. For function not declared inline, recursive inlining
7641 happens only when @option{-finline-functions} (included in @option{-O3}) is
7642 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
7643 default value is 450.
7645 @item max-inline-recursive-depth
7646 @itemx max-inline-recursive-depth-auto
7647 Specifies maximum recursion depth used by the recursive inlining.
7649 For functions declared inline @option{--param max-inline-recursive-depth} is
7650 taken into account. For function not declared inline, recursive inlining
7651 happens only when @option{-finline-functions} (included in @option{-O3}) is
7652 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
7655 @item min-inline-recursive-probability
7656 Recursive inlining is profitable only for function having deep recursion
7657 in average and can hurt for function having little recursion depth by
7658 increasing the prologue size or complexity of function body to other
7661 When profile feedback is available (see @option{-fprofile-generate}) the actual
7662 recursion depth can be guessed from probability that function will recurse via
7663 given call expression. This parameter limits inlining only to call expression
7664 whose probability exceeds given threshold (in percents). The default value is
7667 @item early-inlining-insns
7668 Specify growth that early inliner can make. In effect it increases amount of
7669 inlining for code having large abstraction penalty. The default value is 12.
7671 @item max-early-inliner-iterations
7672 @itemx max-early-inliner-iterations
7673 Limit of iterations of early inliner. This basically bounds number of nested
7674 indirect calls early inliner can resolve. Deeper chains are still handled by
7677 @item min-vect-loop-bound
7678 The minimum number of iterations under which a loop will not get vectorized
7679 when @option{-ftree-vectorize} is used. The number of iterations after
7680 vectorization needs to be greater than the value specified by this option
7681 to allow vectorization. The default value is 0.
7683 @item max-unrolled-insns
7684 The maximum number of instructions that a loop should have if that loop
7685 is unrolled, and if the loop is unrolled, it determines how many times
7686 the loop code is unrolled.
7688 @item max-average-unrolled-insns
7689 The maximum number of instructions biased by probabilities of their execution
7690 that a loop should have if that loop is unrolled, and if the loop is unrolled,
7691 it determines how many times the loop code is unrolled.
7693 @item max-unroll-times
7694 The maximum number of unrollings of a single loop.
7696 @item max-peeled-insns
7697 The maximum number of instructions that a loop should have if that loop
7698 is peeled, and if the loop is peeled, it determines how many times
7699 the loop code is peeled.
7701 @item max-peel-times
7702 The maximum number of peelings of a single loop.
7704 @item max-completely-peeled-insns
7705 The maximum number of insns of a completely peeled loop.
7707 @item max-completely-peel-times
7708 The maximum number of iterations of a loop to be suitable for complete peeling.
7710 @item max-unswitch-insns
7711 The maximum number of insns of an unswitched loop.
7713 @item max-unswitch-level
7714 The maximum number of branches unswitched in a single loop.
7717 The minimum cost of an expensive expression in the loop invariant motion.
7719 @item iv-consider-all-candidates-bound
7720 Bound on number of candidates for induction variables below that
7721 all candidates are considered for each use in induction variable
7722 optimizations. Only the most relevant candidates are considered
7723 if there are more candidates, to avoid quadratic time complexity.
7725 @item iv-max-considered-uses
7726 The induction variable optimizations give up on loops that contain more
7727 induction variable uses.
7729 @item iv-always-prune-cand-set-bound
7730 If number of candidates in the set is smaller than this value,
7731 we always try to remove unnecessary ivs from the set during its
7732 optimization when a new iv is added to the set.
7734 @item scev-max-expr-size
7735 Bound on size of expressions used in the scalar evolutions analyzer.
7736 Large expressions slow the analyzer.
7738 @item omega-max-vars
7739 The maximum number of variables in an Omega constraint system.
7740 The default value is 128.
7742 @item omega-max-geqs
7743 The maximum number of inequalities in an Omega constraint system.
7744 The default value is 256.
7747 The maximum number of equalities in an Omega constraint system.
7748 The default value is 128.
7750 @item omega-max-wild-cards
7751 The maximum number of wildcard variables that the Omega solver will
7752 be able to insert. The default value is 18.
7754 @item omega-hash-table-size
7755 The size of the hash table in the Omega solver. The default value is
7758 @item omega-max-keys
7759 The maximal number of keys used by the Omega solver. The default
7762 @item omega-eliminate-redundant-constraints
7763 When set to 1, use expensive methods to eliminate all redundant
7764 constraints. The default value is 0.
7766 @item vect-max-version-for-alignment-checks
7767 The maximum number of runtime checks that can be performed when
7768 doing loop versioning for alignment in the vectorizer. See option
7769 ftree-vect-loop-version for more information.
7771 @item vect-max-version-for-alias-checks
7772 The maximum number of runtime checks that can be performed when
7773 doing loop versioning for alias in the vectorizer. See option
7774 ftree-vect-loop-version for more information.
7776 @item max-iterations-to-track
7778 The maximum number of iterations of a loop the brute force algorithm
7779 for analysis of # of iterations of the loop tries to evaluate.
7781 @item hot-bb-count-fraction
7782 Select fraction of the maximal count of repetitions of basic block in program
7783 given basic block needs to have to be considered hot.
7785 @item hot-bb-frequency-fraction
7786 Select fraction of the maximal frequency of executions of basic block in
7787 function given basic block needs to have to be considered hot
7789 @item max-predicted-iterations
7790 The maximum number of loop iterations we predict statically. This is useful
7791 in cases where function contain single loop with known bound and other loop
7792 with unknown. We predict the known number of iterations correctly, while
7793 the unknown number of iterations average to roughly 10. This means that the
7794 loop without bounds would appear artificially cold relative to the other one.
7796 @item align-threshold
7798 Select fraction of the maximal frequency of executions of basic block in
7799 function given basic block will get aligned.
7801 @item align-loop-iterations
7803 A loop expected to iterate at lest the selected number of iterations will get
7806 @item tracer-dynamic-coverage
7807 @itemx tracer-dynamic-coverage-feedback
7809 This value is used to limit superblock formation once the given percentage of
7810 executed instructions is covered. This limits unnecessary code size
7813 The @option{tracer-dynamic-coverage-feedback} is used only when profile
7814 feedback is available. The real profiles (as opposed to statically estimated
7815 ones) are much less balanced allowing the threshold to be larger value.
7817 @item tracer-max-code-growth
7818 Stop tail duplication once code growth has reached given percentage. This is
7819 rather hokey argument, as most of the duplicates will be eliminated later in
7820 cross jumping, so it may be set to much higher values than is the desired code
7823 @item tracer-min-branch-ratio
7825 Stop reverse growth when the reverse probability of best edge is less than this
7826 threshold (in percent).
7828 @item tracer-min-branch-ratio
7829 @itemx tracer-min-branch-ratio-feedback
7831 Stop forward growth if the best edge do have probability lower than this
7834 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
7835 compilation for profile feedback and one for compilation without. The value
7836 for compilation with profile feedback needs to be more conservative (higher) in
7837 order to make tracer effective.
7839 @item max-cse-path-length
7841 Maximum number of basic blocks on path that cse considers. The default is 10.
7844 The maximum instructions CSE process before flushing. The default is 1000.
7846 @item ggc-min-expand
7848 GCC uses a garbage collector to manage its own memory allocation. This
7849 parameter specifies the minimum percentage by which the garbage
7850 collector's heap should be allowed to expand between collections.
7851 Tuning this may improve compilation speed; it has no effect on code
7854 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
7855 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
7856 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
7857 GCC is not able to calculate RAM on a particular platform, the lower
7858 bound of 30% is used. Setting this parameter and
7859 @option{ggc-min-heapsize} to zero causes a full collection to occur at
7860 every opportunity. This is extremely slow, but can be useful for
7863 @item ggc-min-heapsize
7865 Minimum size of the garbage collector's heap before it begins bothering
7866 to collect garbage. The first collection occurs after the heap expands
7867 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
7868 tuning this may improve compilation speed, and has no effect on code
7871 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
7872 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
7873 with a lower bound of 4096 (four megabytes) and an upper bound of
7874 131072 (128 megabytes). If GCC is not able to calculate RAM on a
7875 particular platform, the lower bound is used. Setting this parameter
7876 very large effectively disables garbage collection. Setting this
7877 parameter and @option{ggc-min-expand} to zero causes a full collection
7878 to occur at every opportunity.
7880 @item max-reload-search-insns
7881 The maximum number of instruction reload should look backward for equivalent
7882 register. Increasing values mean more aggressive optimization, making the
7883 compile time increase with probably slightly better performance. The default
7886 @item max-cselib-memory-locations
7887 The maximum number of memory locations cselib should take into account.
7888 Increasing values mean more aggressive optimization, making the compile time
7889 increase with probably slightly better performance. The default value is 500.
7891 @item reorder-blocks-duplicate
7892 @itemx reorder-blocks-duplicate-feedback
7894 Used by basic block reordering pass to decide whether to use unconditional
7895 branch or duplicate the code on its destination. Code is duplicated when its
7896 estimated size is smaller than this value multiplied by the estimated size of
7897 unconditional jump in the hot spots of the program.
7899 The @option{reorder-block-duplicate-feedback} is used only when profile
7900 feedback is available and may be set to higher values than
7901 @option{reorder-block-duplicate} since information about the hot spots is more
7904 @item max-sched-ready-insns
7905 The maximum number of instructions ready to be issued the scheduler should
7906 consider at any given time during the first scheduling pass. Increasing
7907 values mean more thorough searches, making the compilation time increase
7908 with probably little benefit. The default value is 100.
7910 @item max-sched-region-blocks
7911 The maximum number of blocks in a region to be considered for
7912 interblock scheduling. The default value is 10.
7914 @item max-pipeline-region-blocks
7915 The maximum number of blocks in a region to be considered for
7916 pipelining in the selective scheduler. The default value is 15.
7918 @item max-sched-region-insns
7919 The maximum number of insns in a region to be considered for
7920 interblock scheduling. The default value is 100.
7922 @item max-pipeline-region-insns
7923 The maximum number of insns in a region to be considered for
7924 pipelining in the selective scheduler. The default value is 200.
7927 The minimum probability (in percents) of reaching a source block
7928 for interblock speculative scheduling. The default value is 40.
7930 @item max-sched-extend-regions-iters
7931 The maximum number of iterations through CFG to extend regions.
7932 0 - disable region extension,
7933 N - do at most N iterations.
7934 The default value is 0.
7936 @item max-sched-insn-conflict-delay
7937 The maximum conflict delay for an insn to be considered for speculative motion.
7938 The default value is 3.
7940 @item sched-spec-prob-cutoff
7941 The minimal probability of speculation success (in percents), so that
7942 speculative insn will be scheduled.
7943 The default value is 40.
7945 @item sched-mem-true-dep-cost
7946 Minimal distance (in CPU cycles) between store and load targeting same
7947 memory locations. The default value is 1.
7949 @item selsched-max-lookahead
7950 The maximum size of the lookahead window of selective scheduling. It is a
7951 depth of search for available instructions.
7952 The default value is 50.
7954 @item selsched-max-sched-times
7955 The maximum number of times that an instruction will be scheduled during
7956 selective scheduling. This is the limit on the number of iterations
7957 through which the instruction may be pipelined. The default value is 2.
7959 @item selsched-max-insns-to-rename
7960 The maximum number of best instructions in the ready list that are considered
7961 for renaming in the selective scheduler. The default value is 2.
7963 @item max-last-value-rtl
7964 The maximum size measured as number of RTLs that can be recorded in an expression
7965 in combiner for a pseudo register as last known value of that register. The default
7968 @item integer-share-limit
7969 Small integer constants can use a shared data structure, reducing the
7970 compiler's memory usage and increasing its speed. This sets the maximum
7971 value of a shared integer constant. The default value is 256.
7973 @item min-virtual-mappings
7974 Specifies the minimum number of virtual mappings in the incremental
7975 SSA updater that should be registered to trigger the virtual mappings
7976 heuristic defined by virtual-mappings-ratio. The default value is
7979 @item virtual-mappings-ratio
7980 If the number of virtual mappings is virtual-mappings-ratio bigger
7981 than the number of virtual symbols to be updated, then the incremental
7982 SSA updater switches to a full update for those symbols. The default
7985 @item ssp-buffer-size
7986 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
7987 protection when @option{-fstack-protection} is used.
7989 @item max-jump-thread-duplication-stmts
7990 Maximum number of statements allowed in a block that needs to be
7991 duplicated when threading jumps.
7993 @item max-fields-for-field-sensitive
7994 Maximum number of fields in a structure we will treat in
7995 a field sensitive manner during pointer analysis. The default is zero
7996 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
7998 @item prefetch-latency
7999 Estimate on average number of instructions that are executed before
8000 prefetch finishes. The distance we prefetch ahead is proportional
8001 to this constant. Increasing this number may also lead to less
8002 streams being prefetched (see @option{simultaneous-prefetches}).
8004 @item simultaneous-prefetches
8005 Maximum number of prefetches that can run at the same time.
8007 @item l1-cache-line-size
8008 The size of cache line in L1 cache, in bytes.
8011 The size of L1 cache, in kilobytes.
8014 The size of L2 cache, in kilobytes.
8016 @item min-insn-to-prefetch-ratio
8017 The minimum ratio between the number of instructions and the
8018 number of prefetches to enable prefetching in a loop with an
8021 @item prefetch-min-insn-to-mem-ratio
8022 The minimum ratio between the number of instructions and the
8023 number of memory references to enable prefetching in a loop.
8025 @item use-canonical-types
8026 Whether the compiler should use the ``canonical'' type system. By
8027 default, this should always be 1, which uses a more efficient internal
8028 mechanism for comparing types in C++ and Objective-C++. However, if
8029 bugs in the canonical type system are causing compilation failures,
8030 set this value to 0 to disable canonical types.
8032 @item switch-conversion-max-branch-ratio
8033 Switch initialization conversion will refuse to create arrays that are
8034 bigger than @option{switch-conversion-max-branch-ratio} times the number of
8035 branches in the switch.
8037 @item max-partial-antic-length
8038 Maximum length of the partial antic set computed during the tree
8039 partial redundancy elimination optimization (@option{-ftree-pre}) when
8040 optimizing at @option{-O3} and above. For some sorts of source code
8041 the enhanced partial redundancy elimination optimization can run away,
8042 consuming all of the memory available on the host machine. This
8043 parameter sets a limit on the length of the sets that are computed,
8044 which prevents the runaway behavior. Setting a value of 0 for
8045 this parameter will allow an unlimited set length.
8047 @item sccvn-max-scc-size
8048 Maximum size of a strongly connected component (SCC) during SCCVN
8049 processing. If this limit is hit, SCCVN processing for the whole
8050 function will not be done and optimizations depending on it will
8051 be disabled. The default maximum SCC size is 10000.
8053 @item ira-max-loops-num
8054 IRA uses a regional register allocation by default. If a function
8055 contains loops more than number given by the parameter, only at most
8056 given number of the most frequently executed loops will form regions
8057 for the regional register allocation. The default value of the
8060 @item ira-max-conflict-table-size
8061 Although IRA uses a sophisticated algorithm of compression conflict
8062 table, the table can be still big for huge functions. If the conflict
8063 table for a function could be more than size in MB given by the
8064 parameter, the conflict table is not built and faster, simpler, and
8065 lower quality register allocation algorithm will be used. The
8066 algorithm do not use pseudo-register conflicts. The default value of
8067 the parameter is 2000.
8069 @item loop-invariant-max-bbs-in-loop
8070 Loop invariant motion can be very expensive, both in compile time and
8071 in amount of needed compile time memory, with very large loops. Loops
8072 with more basic blocks than this parameter won't have loop invariant
8073 motion optimization performed on them. The default value of the
8074 parameter is 1000 for -O1 and 10000 for -O2 and above.
8079 @node Preprocessor Options
8080 @section Options Controlling the Preprocessor
8081 @cindex preprocessor options
8082 @cindex options, preprocessor
8084 These options control the C preprocessor, which is run on each C source
8085 file before actual compilation.
8087 If you use the @option{-E} option, nothing is done except preprocessing.
8088 Some of these options make sense only together with @option{-E} because
8089 they cause the preprocessor output to be unsuitable for actual
8093 @item -Wp,@var{option}
8095 You can use @option{-Wp,@var{option}} to bypass the compiler driver
8096 and pass @var{option} directly through to the preprocessor. If
8097 @var{option} contains commas, it is split into multiple options at the
8098 commas. However, many options are modified, translated or interpreted
8099 by the compiler driver before being passed to the preprocessor, and
8100 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
8101 interface is undocumented and subject to change, so whenever possible
8102 you should avoid using @option{-Wp} and let the driver handle the
8105 @item -Xpreprocessor @var{option}
8106 @opindex Xpreprocessor
8107 Pass @var{option} as an option to the preprocessor. You can use this to
8108 supply system-specific preprocessor options which GCC does not know how to
8111 If you want to pass an option that takes an argument, you must use
8112 @option{-Xpreprocessor} twice, once for the option and once for the argument.
8115 @include cppopts.texi
8117 @node Assembler Options
8118 @section Passing Options to the Assembler
8120 @c prevent bad page break with this line
8121 You can pass options to the assembler.
8124 @item -Wa,@var{option}
8126 Pass @var{option} as an option to the assembler. If @var{option}
8127 contains commas, it is split into multiple options at the commas.
8129 @item -Xassembler @var{option}
8131 Pass @var{option} as an option to the assembler. You can use this to
8132 supply system-specific assembler options which GCC does not know how to
8135 If you want to pass an option that takes an argument, you must use
8136 @option{-Xassembler} twice, once for the option and once for the argument.
8141 @section Options for Linking
8142 @cindex link options
8143 @cindex options, linking
8145 These options come into play when the compiler links object files into
8146 an executable output file. They are meaningless if the compiler is
8147 not doing a link step.
8151 @item @var{object-file-name}
8152 A file name that does not end in a special recognized suffix is
8153 considered to name an object file or library. (Object files are
8154 distinguished from libraries by the linker according to the file
8155 contents.) If linking is done, these object files are used as input
8164 If any of these options is used, then the linker is not run, and
8165 object file names should not be used as arguments. @xref{Overall
8169 @item -l@var{library}
8170 @itemx -l @var{library}
8172 Search the library named @var{library} when linking. (The second
8173 alternative with the library as a separate argument is only for
8174 POSIX compliance and is not recommended.)
8176 It makes a difference where in the command you write this option; the
8177 linker searches and processes libraries and object files in the order they
8178 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
8179 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
8180 to functions in @samp{z}, those functions may not be loaded.
8182 The linker searches a standard list of directories for the library,
8183 which is actually a file named @file{lib@var{library}.a}. The linker
8184 then uses this file as if it had been specified precisely by name.
8186 The directories searched include several standard system directories
8187 plus any that you specify with @option{-L}.
8189 Normally the files found this way are library files---archive files
8190 whose members are object files. The linker handles an archive file by
8191 scanning through it for members which define symbols that have so far
8192 been referenced but not defined. But if the file that is found is an
8193 ordinary object file, it is linked in the usual fashion. The only
8194 difference between using an @option{-l} option and specifying a file name
8195 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
8196 and searches several directories.
8200 You need this special case of the @option{-l} option in order to
8201 link an Objective-C or Objective-C++ program.
8204 @opindex nostartfiles
8205 Do not use the standard system startup files when linking.
8206 The standard system libraries are used normally, unless @option{-nostdlib}
8207 or @option{-nodefaultlibs} is used.
8209 @item -nodefaultlibs
8210 @opindex nodefaultlibs
8211 Do not use the standard system libraries when linking.
8212 Only the libraries you specify will be passed to the linker, options
8213 specifying linkage of the system libraries, such as @code{-static-libgcc}
8214 or @code{-shared-libgcc}, will be ignored.
8215 The standard startup files are used normally, unless @option{-nostartfiles}
8216 is used. The compiler may generate calls to @code{memcmp},
8217 @code{memset}, @code{memcpy} and @code{memmove}.
8218 These entries are usually resolved by entries in
8219 libc. These entry points should be supplied through some other
8220 mechanism when this option is specified.
8224 Do not use the standard system startup files or libraries when linking.
8225 No startup files and only the libraries you specify will be passed to
8226 the linker, options specifying linkage of the system libraries, such as
8227 @code{-static-libgcc} or @code{-shared-libgcc}, will be ignored.
8228 The compiler may generate calls to @code{memcmp}, @code{memset},
8229 @code{memcpy} and @code{memmove}.
8230 These entries are usually resolved by entries in
8231 libc. These entry points should be supplied through some other
8232 mechanism when this option is specified.
8234 @cindex @option{-lgcc}, use with @option{-nostdlib}
8235 @cindex @option{-nostdlib} and unresolved references
8236 @cindex unresolved references and @option{-nostdlib}
8237 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
8238 @cindex @option{-nodefaultlibs} and unresolved references
8239 @cindex unresolved references and @option{-nodefaultlibs}
8240 One of the standard libraries bypassed by @option{-nostdlib} and
8241 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
8242 that GCC uses to overcome shortcomings of particular machines, or special
8243 needs for some languages.
8244 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
8245 Collection (GCC) Internals},
8246 for more discussion of @file{libgcc.a}.)
8247 In most cases, you need @file{libgcc.a} even when you want to avoid
8248 other standard libraries. In other words, when you specify @option{-nostdlib}
8249 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
8250 This ensures that you have no unresolved references to internal GCC
8251 library subroutines. (For example, @samp{__main}, used to ensure C++
8252 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
8253 GNU Compiler Collection (GCC) Internals}.)
8257 Produce a position independent executable on targets which support it.
8258 For predictable results, you must also specify the same set of options
8259 that were used to generate code (@option{-fpie}, @option{-fPIE},
8260 or model suboptions) when you specify this option.
8264 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
8265 that support it. This instructs the linker to add all symbols, not
8266 only used ones, to the dynamic symbol table. This option is needed
8267 for some uses of @code{dlopen} or to allow obtaining backtraces
8268 from within a program.
8272 Remove all symbol table and relocation information from the executable.
8276 On systems that support dynamic linking, this prevents linking with the shared
8277 libraries. On other systems, this option has no effect.
8281 Produce a shared object which can then be linked with other objects to
8282 form an executable. Not all systems support this option. For predictable
8283 results, you must also specify the same set of options that were used to
8284 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
8285 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
8286 needs to build supplementary stub code for constructors to work. On
8287 multi-libbed systems, @samp{gcc -shared} must select the correct support
8288 libraries to link against. Failing to supply the correct flags may lead
8289 to subtle defects. Supplying them in cases where they are not necessary
8292 @item -shared-libgcc
8293 @itemx -static-libgcc
8294 @opindex shared-libgcc
8295 @opindex static-libgcc
8296 On systems that provide @file{libgcc} as a shared library, these options
8297 force the use of either the shared or static version respectively.
8298 If no shared version of @file{libgcc} was built when the compiler was
8299 configured, these options have no effect.
8301 There are several situations in which an application should use the
8302 shared @file{libgcc} instead of the static version. The most common
8303 of these is when the application wishes to throw and catch exceptions
8304 across different shared libraries. In that case, each of the libraries
8305 as well as the application itself should use the shared @file{libgcc}.
8307 Therefore, the G++ and GCJ drivers automatically add
8308 @option{-shared-libgcc} whenever you build a shared library or a main
8309 executable, because C++ and Java programs typically use exceptions, so
8310 this is the right thing to do.
8312 If, instead, you use the GCC driver to create shared libraries, you may
8313 find that they will not always be linked with the shared @file{libgcc}.
8314 If GCC finds, at its configuration time, that you have a non-GNU linker
8315 or a GNU linker that does not support option @option{--eh-frame-hdr},
8316 it will link the shared version of @file{libgcc} into shared libraries
8317 by default. Otherwise, it will take advantage of the linker and optimize
8318 away the linking with the shared version of @file{libgcc}, linking with
8319 the static version of libgcc by default. This allows exceptions to
8320 propagate through such shared libraries, without incurring relocation
8321 costs at library load time.
8323 However, if a library or main executable is supposed to throw or catch
8324 exceptions, you must link it using the G++ or GCJ driver, as appropriate
8325 for the languages used in the program, or using the option
8326 @option{-shared-libgcc}, such that it is linked with the shared
8329 @item -static-libstdc++
8330 When the @command{g++} program is used to link a C++ program, it will
8331 normally automatically link against @option{libstdc++}. If
8332 @file{libstdc++} is available as a shared library, and the
8333 @option{-static} option is not used, then this will link against the
8334 shared version of @file{libstdc++}. That is normally fine. However, it
8335 is sometimes useful to freeze the version of @file{libstdc++} used by
8336 the program without going all the way to a fully static link. The
8337 @option{-static-libstdc++} option directs the @command{g++} driver to
8338 link @file{libstdc++} statically, without necessarily linking other
8339 libraries statically.
8343 Bind references to global symbols when building a shared object. Warn
8344 about any unresolved references (unless overridden by the link editor
8345 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
8348 @item -T @var{script}
8350 @cindex linker script
8351 Use @var{script} as the linker script. This option is supported by most
8352 systems using the GNU linker. On some targets, such as bare-board
8353 targets without an operating system, the @option{-T} option may be required
8354 when linking to avoid references to undefined symbols.
8356 @item -Xlinker @var{option}
8358 Pass @var{option} as an option to the linker. You can use this to
8359 supply system-specific linker options which GCC does not know how to
8362 If you want to pass an option that takes a separate argument, you must use
8363 @option{-Xlinker} twice, once for the option and once for the argument.
8364 For example, to pass @option{-assert definitions}, you must write
8365 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
8366 @option{-Xlinker "-assert definitions"}, because this passes the entire
8367 string as a single argument, which is not what the linker expects.
8369 When using the GNU linker, it is usually more convenient to pass
8370 arguments to linker options using the @option{@var{option}=@var{value}}
8371 syntax than as separate arguments. For example, you can specify
8372 @samp{-Xlinker -Map=output.map} rather than
8373 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
8374 this syntax for command-line options.
8376 @item -Wl,@var{option}
8378 Pass @var{option} as an option to the linker. If @var{option} contains
8379 commas, it is split into multiple options at the commas. You can use this
8380 syntax to pass an argument to the option.
8381 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
8382 linker. When using the GNU linker, you can also get the same effect with
8383 @samp{-Wl,-Map=output.map}.
8385 @item -u @var{symbol}
8387 Pretend the symbol @var{symbol} is undefined, to force linking of
8388 library modules to define it. You can use @option{-u} multiple times with
8389 different symbols to force loading of additional library modules.
8392 @node Directory Options
8393 @section Options for Directory Search
8394 @cindex directory options
8395 @cindex options, directory search
8398 These options specify directories to search for header files, for
8399 libraries and for parts of the compiler:
8404 Add the directory @var{dir} to the head of the list of directories to be
8405 searched for header files. This can be used to override a system header
8406 file, substituting your own version, since these directories are
8407 searched before the system header file directories. However, you should
8408 not use this option to add directories that contain vendor-supplied
8409 system header files (use @option{-isystem} for that). If you use more than
8410 one @option{-I} option, the directories are scanned in left-to-right
8411 order; the standard system directories come after.
8413 If a standard system include directory, or a directory specified with
8414 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
8415 option will be ignored. The directory will still be searched but as a
8416 system directory at its normal position in the system include chain.
8417 This is to ensure that GCC's procedure to fix buggy system headers and
8418 the ordering for the include_next directive are not inadvertently changed.
8419 If you really need to change the search order for system directories,
8420 use the @option{-nostdinc} and/or @option{-isystem} options.
8422 @item -iquote@var{dir}
8424 Add the directory @var{dir} to the head of the list of directories to
8425 be searched for header files only for the case of @samp{#include
8426 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
8427 otherwise just like @option{-I}.
8431 Add directory @var{dir} to the list of directories to be searched
8434 @item -B@var{prefix}
8436 This option specifies where to find the executables, libraries,
8437 include files, and data files of the compiler itself.
8439 The compiler driver program runs one or more of the subprograms
8440 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
8441 @var{prefix} as a prefix for each program it tries to run, both with and
8442 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
8444 For each subprogram to be run, the compiler driver first tries the
8445 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
8446 was not specified, the driver tries two standard prefixes, which are
8447 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
8448 those results in a file name that is found, the unmodified program
8449 name is searched for using the directories specified in your
8450 @env{PATH} environment variable.
8452 The compiler will check to see if the path provided by the @option{-B}
8453 refers to a directory, and if necessary it will add a directory
8454 separator character at the end of the path.
8456 @option{-B} prefixes that effectively specify directory names also apply
8457 to libraries in the linker, because the compiler translates these
8458 options into @option{-L} options for the linker. They also apply to
8459 includes files in the preprocessor, because the compiler translates these
8460 options into @option{-isystem} options for the preprocessor. In this case,
8461 the compiler appends @samp{include} to the prefix.
8463 The run-time support file @file{libgcc.a} can also be searched for using
8464 the @option{-B} prefix, if needed. If it is not found there, the two
8465 standard prefixes above are tried, and that is all. The file is left
8466 out of the link if it is not found by those means.
8468 Another way to specify a prefix much like the @option{-B} prefix is to use
8469 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
8472 As a special kludge, if the path provided by @option{-B} is
8473 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
8474 9, then it will be replaced by @file{[dir/]include}. This is to help
8475 with boot-strapping the compiler.
8477 @item -specs=@var{file}
8479 Process @var{file} after the compiler reads in the standard @file{specs}
8480 file, in order to override the defaults that the @file{gcc} driver
8481 program uses when determining what switches to pass to @file{cc1},
8482 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
8483 @option{-specs=@var{file}} can be specified on the command line, and they
8484 are processed in order, from left to right.
8486 @item --sysroot=@var{dir}
8488 Use @var{dir} as the logical root directory for headers and libraries.
8489 For example, if the compiler would normally search for headers in
8490 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
8491 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
8493 If you use both this option and the @option{-isysroot} option, then
8494 the @option{--sysroot} option will apply to libraries, but the
8495 @option{-isysroot} option will apply to header files.
8497 The GNU linker (beginning with version 2.16) has the necessary support
8498 for this option. If your linker does not support this option, the
8499 header file aspect of @option{--sysroot} will still work, but the
8500 library aspect will not.
8504 This option has been deprecated. Please use @option{-iquote} instead for
8505 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
8506 Any directories you specify with @option{-I} options before the @option{-I-}
8507 option are searched only for the case of @samp{#include "@var{file}"};
8508 they are not searched for @samp{#include <@var{file}>}.
8510 If additional directories are specified with @option{-I} options after
8511 the @option{-I-}, these directories are searched for all @samp{#include}
8512 directives. (Ordinarily @emph{all} @option{-I} directories are used
8515 In addition, the @option{-I-} option inhibits the use of the current
8516 directory (where the current input file came from) as the first search
8517 directory for @samp{#include "@var{file}"}. There is no way to
8518 override this effect of @option{-I-}. With @option{-I.} you can specify
8519 searching the directory which was current when the compiler was
8520 invoked. That is not exactly the same as what the preprocessor does
8521 by default, but it is often satisfactory.
8523 @option{-I-} does not inhibit the use of the standard system directories
8524 for header files. Thus, @option{-I-} and @option{-nostdinc} are
8531 @section Specifying subprocesses and the switches to pass to them
8534 @command{gcc} is a driver program. It performs its job by invoking a
8535 sequence of other programs to do the work of compiling, assembling and
8536 linking. GCC interprets its command-line parameters and uses these to
8537 deduce which programs it should invoke, and which command-line options
8538 it ought to place on their command lines. This behavior is controlled
8539 by @dfn{spec strings}. In most cases there is one spec string for each
8540 program that GCC can invoke, but a few programs have multiple spec
8541 strings to control their behavior. The spec strings built into GCC can
8542 be overridden by using the @option{-specs=} command-line switch to specify
8545 @dfn{Spec files} are plaintext files that are used to construct spec
8546 strings. They consist of a sequence of directives separated by blank
8547 lines. The type of directive is determined by the first non-whitespace
8548 character on the line and it can be one of the following:
8551 @item %@var{command}
8552 Issues a @var{command} to the spec file processor. The commands that can
8556 @item %include <@var{file}>
8558 Search for @var{file} and insert its text at the current point in the
8561 @item %include_noerr <@var{file}>
8562 @cindex %include_noerr
8563 Just like @samp{%include}, but do not generate an error message if the include
8564 file cannot be found.
8566 @item %rename @var{old_name} @var{new_name}
8568 Rename the spec string @var{old_name} to @var{new_name}.
8572 @item *[@var{spec_name}]:
8573 This tells the compiler to create, override or delete the named spec
8574 string. All lines after this directive up to the next directive or
8575 blank line are considered to be the text for the spec string. If this
8576 results in an empty string then the spec will be deleted. (Or, if the
8577 spec did not exist, then nothing will happened.) Otherwise, if the spec
8578 does not currently exist a new spec will be created. If the spec does
8579 exist then its contents will be overridden by the text of this
8580 directive, unless the first character of that text is the @samp{+}
8581 character, in which case the text will be appended to the spec.
8583 @item [@var{suffix}]:
8584 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
8585 and up to the next directive or blank line are considered to make up the
8586 spec string for the indicated suffix. When the compiler encounters an
8587 input file with the named suffix, it will processes the spec string in
8588 order to work out how to compile that file. For example:
8595 This says that any input file whose name ends in @samp{.ZZ} should be
8596 passed to the program @samp{z-compile}, which should be invoked with the
8597 command-line switch @option{-input} and with the result of performing the
8598 @samp{%i} substitution. (See below.)
8600 As an alternative to providing a spec string, the text that follows a
8601 suffix directive can be one of the following:
8604 @item @@@var{language}
8605 This says that the suffix is an alias for a known @var{language}. This is
8606 similar to using the @option{-x} command-line switch to GCC to specify a
8607 language explicitly. For example:
8614 Says that .ZZ files are, in fact, C++ source files.
8617 This causes an error messages saying:
8620 @var{name} compiler not installed on this system.
8624 GCC already has an extensive list of suffixes built into it.
8625 This directive will add an entry to the end of the list of suffixes, but
8626 since the list is searched from the end backwards, it is effectively
8627 possible to override earlier entries using this technique.
8631 GCC has the following spec strings built into it. Spec files can
8632 override these strings or create their own. Note that individual
8633 targets can also add their own spec strings to this list.
8636 asm Options to pass to the assembler
8637 asm_final Options to pass to the assembler post-processor
8638 cpp Options to pass to the C preprocessor
8639 cc1 Options to pass to the C compiler
8640 cc1plus Options to pass to the C++ compiler
8641 endfile Object files to include at the end of the link
8642 link Options to pass to the linker
8643 lib Libraries to include on the command line to the linker
8644 libgcc Decides which GCC support library to pass to the linker
8645 linker Sets the name of the linker
8646 predefines Defines to be passed to the C preprocessor
8647 signed_char Defines to pass to CPP to say whether @code{char} is signed
8649 startfile Object files to include at the start of the link
8652 Here is a small example of a spec file:
8658 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
8661 This example renames the spec called @samp{lib} to @samp{old_lib} and
8662 then overrides the previous definition of @samp{lib} with a new one.
8663 The new definition adds in some extra command-line options before
8664 including the text of the old definition.
8666 @dfn{Spec strings} are a list of command-line options to be passed to their
8667 corresponding program. In addition, the spec strings can contain
8668 @samp{%}-prefixed sequences to substitute variable text or to
8669 conditionally insert text into the command line. Using these constructs
8670 it is possible to generate quite complex command lines.
8672 Here is a table of all defined @samp{%}-sequences for spec
8673 strings. Note that spaces are not generated automatically around the
8674 results of expanding these sequences. Therefore you can concatenate them
8675 together or combine them with constant text in a single argument.
8679 Substitute one @samp{%} into the program name or argument.
8682 Substitute the name of the input file being processed.
8685 Substitute the basename of the input file being processed.
8686 This is the substring up to (and not including) the last period
8687 and not including the directory.
8690 This is the same as @samp{%b}, but include the file suffix (text after
8694 Marks the argument containing or following the @samp{%d} as a
8695 temporary file name, so that that file will be deleted if GCC exits
8696 successfully. Unlike @samp{%g}, this contributes no text to the
8699 @item %g@var{suffix}
8700 Substitute a file name that has suffix @var{suffix} and is chosen
8701 once per compilation, and mark the argument in the same way as
8702 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
8703 name is now chosen in a way that is hard to predict even when previously
8704 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
8705 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
8706 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
8707 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
8708 was simply substituted with a file name chosen once per compilation,
8709 without regard to any appended suffix (which was therefore treated
8710 just like ordinary text), making such attacks more likely to succeed.
8712 @item %u@var{suffix}
8713 Like @samp{%g}, but generates a new temporary file name even if
8714 @samp{%u@var{suffix}} was already seen.
8716 @item %U@var{suffix}
8717 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
8718 new one if there is no such last file name. In the absence of any
8719 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
8720 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
8721 would involve the generation of two distinct file names, one
8722 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
8723 simply substituted with a file name chosen for the previous @samp{%u},
8724 without regard to any appended suffix.
8726 @item %j@var{suffix}
8727 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
8728 writable, and if save-temps is off; otherwise, substitute the name
8729 of a temporary file, just like @samp{%u}. This temporary file is not
8730 meant for communication between processes, but rather as a junk
8733 @item %|@var{suffix}
8734 @itemx %m@var{suffix}
8735 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
8736 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
8737 all. These are the two most common ways to instruct a program that it
8738 should read from standard input or write to standard output. If you
8739 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
8740 construct: see for example @file{f/lang-specs.h}.
8742 @item %.@var{SUFFIX}
8743 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
8744 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
8745 terminated by the next space or %.
8748 Marks the argument containing or following the @samp{%w} as the
8749 designated output file of this compilation. This puts the argument
8750 into the sequence of arguments that @samp{%o} will substitute later.
8753 Substitutes the names of all the output files, with spaces
8754 automatically placed around them. You should write spaces
8755 around the @samp{%o} as well or the results are undefined.
8756 @samp{%o} is for use in the specs for running the linker.
8757 Input files whose names have no recognized suffix are not compiled
8758 at all, but they are included among the output files, so they will
8762 Substitutes the suffix for object files. Note that this is
8763 handled specially when it immediately follows @samp{%g, %u, or %U},
8764 because of the need for those to form complete file names. The
8765 handling is such that @samp{%O} is treated exactly as if it had already
8766 been substituted, except that @samp{%g, %u, and %U} do not currently
8767 support additional @var{suffix} characters following @samp{%O} as they would
8768 following, for example, @samp{.o}.
8771 Substitutes the standard macro predefinitions for the
8772 current target machine. Use this when running @code{cpp}.
8775 Like @samp{%p}, but puts @samp{__} before and after the name of each
8776 predefined macro, except for macros that start with @samp{__} or with
8777 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
8781 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
8782 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
8783 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
8784 and @option{-imultilib} as necessary.
8787 Current argument is the name of a library or startup file of some sort.
8788 Search for that file in a standard list of directories and substitute
8789 the full name found.
8792 Print @var{str} as an error message. @var{str} is terminated by a newline.
8793 Use this when inconsistent options are detected.
8796 Substitute the contents of spec string @var{name} at this point.
8799 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
8801 @item %x@{@var{option}@}
8802 Accumulate an option for @samp{%X}.
8805 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
8809 Output the accumulated assembler options specified by @option{-Wa}.
8812 Output the accumulated preprocessor options specified by @option{-Wp}.
8815 Process the @code{asm} spec. This is used to compute the
8816 switches to be passed to the assembler.
8819 Process the @code{asm_final} spec. This is a spec string for
8820 passing switches to an assembler post-processor, if such a program is
8824 Process the @code{link} spec. This is the spec for computing the
8825 command line passed to the linker. Typically it will make use of the
8826 @samp{%L %G %S %D and %E} sequences.
8829 Dump out a @option{-L} option for each directory that GCC believes might
8830 contain startup files. If the target supports multilibs then the
8831 current multilib directory will be prepended to each of these paths.
8834 Process the @code{lib} spec. This is a spec string for deciding which
8835 libraries should be included on the command line to the linker.
8838 Process the @code{libgcc} spec. This is a spec string for deciding
8839 which GCC support library should be included on the command line to the linker.
8842 Process the @code{startfile} spec. This is a spec for deciding which
8843 object files should be the first ones passed to the linker. Typically
8844 this might be a file named @file{crt0.o}.
8847 Process the @code{endfile} spec. This is a spec string that specifies
8848 the last object files that will be passed to the linker.
8851 Process the @code{cpp} spec. This is used to construct the arguments
8852 to be passed to the C preprocessor.
8855 Process the @code{cc1} spec. This is used to construct the options to be
8856 passed to the actual C compiler (@samp{cc1}).
8859 Process the @code{cc1plus} spec. This is used to construct the options to be
8860 passed to the actual C++ compiler (@samp{cc1plus}).
8863 Substitute the variable part of a matched option. See below.
8864 Note that each comma in the substituted string is replaced by
8868 Remove all occurrences of @code{-S} from the command line. Note---this
8869 command is position dependent. @samp{%} commands in the spec string
8870 before this one will see @code{-S}, @samp{%} commands in the spec string
8871 after this one will not.
8873 @item %:@var{function}(@var{args})
8874 Call the named function @var{function}, passing it @var{args}.
8875 @var{args} is first processed as a nested spec string, then split
8876 into an argument vector in the usual fashion. The function returns
8877 a string which is processed as if it had appeared literally as part
8878 of the current spec.
8880 The following built-in spec functions are provided:
8884 The @code{getenv} spec function takes two arguments: an environment
8885 variable name and a string. If the environment variable is not
8886 defined, a fatal error is issued. Otherwise, the return value is the
8887 value of the environment variable concatenated with the string. For
8888 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
8891 %:getenv(TOPDIR /include)
8894 expands to @file{/path/to/top/include}.
8896 @item @code{if-exists}
8897 The @code{if-exists} spec function takes one argument, an absolute
8898 pathname to a file. If the file exists, @code{if-exists} returns the
8899 pathname. Here is a small example of its usage:
8903 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
8906 @item @code{if-exists-else}
8907 The @code{if-exists-else} spec function is similar to the @code{if-exists}
8908 spec function, except that it takes two arguments. The first argument is
8909 an absolute pathname to a file. If the file exists, @code{if-exists-else}
8910 returns the pathname. If it does not exist, it returns the second argument.
8911 This way, @code{if-exists-else} can be used to select one file or another,
8912 based on the existence of the first. Here is a small example of its usage:
8916 crt0%O%s %:if-exists(crti%O%s) \
8917 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
8920 @item @code{replace-outfile}
8921 The @code{replace-outfile} spec function takes two arguments. It looks for the
8922 first argument in the outfiles array and replaces it with the second argument. Here
8923 is a small example of its usage:
8926 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
8929 @item @code{print-asm-header}
8930 The @code{print-asm-header} function takes no arguments and simply
8931 prints a banner like:
8937 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
8940 It is used to separate compiler options from assembler options
8941 in the @option{--target-help} output.
8945 Substitutes the @code{-S} switch, if that switch was given to GCC@.
8946 If that switch was not specified, this substitutes nothing. Note that
8947 the leading dash is omitted when specifying this option, and it is
8948 automatically inserted if the substitution is performed. Thus the spec
8949 string @samp{%@{foo@}} would match the command-line option @option{-foo}
8950 and would output the command line option @option{-foo}.
8952 @item %W@{@code{S}@}
8953 Like %@{@code{S}@} but mark last argument supplied within as a file to be
8956 @item %@{@code{S}*@}
8957 Substitutes all the switches specified to GCC whose names start
8958 with @code{-S}, but which also take an argument. This is used for
8959 switches like @option{-o}, @option{-D}, @option{-I}, etc.
8960 GCC considers @option{-o foo} as being
8961 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
8962 text, including the space. Thus two arguments would be generated.
8964 @item %@{@code{S}*&@code{T}*@}
8965 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
8966 (the order of @code{S} and @code{T} in the spec is not significant).
8967 There can be any number of ampersand-separated variables; for each the
8968 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
8970 @item %@{@code{S}:@code{X}@}
8971 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
8973 @item %@{!@code{S}:@code{X}@}
8974 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
8976 @item %@{@code{S}*:@code{X}@}
8977 Substitutes @code{X} if one or more switches whose names start with
8978 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
8979 once, no matter how many such switches appeared. However, if @code{%*}
8980 appears somewhere in @code{X}, then @code{X} will be substituted once
8981 for each matching switch, with the @code{%*} replaced by the part of
8982 that switch that matched the @code{*}.
8984 @item %@{.@code{S}:@code{X}@}
8985 Substitutes @code{X}, if processing a file with suffix @code{S}.
8987 @item %@{!.@code{S}:@code{X}@}
8988 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
8990 @item %@{,@code{S}:@code{X}@}
8991 Substitutes @code{X}, if processing a file for language @code{S}.
8993 @item %@{!,@code{S}:@code{X}@}
8994 Substitutes @code{X}, if not processing a file for language @code{S}.
8996 @item %@{@code{S}|@code{P}:@code{X}@}
8997 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
8998 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
8999 @code{*} sequences as well, although they have a stronger binding than
9000 the @samp{|}. If @code{%*} appears in @code{X}, all of the
9001 alternatives must be starred, and only the first matching alternative
9004 For example, a spec string like this:
9007 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
9010 will output the following command-line options from the following input
9011 command-line options:
9016 -d fred.c -foo -baz -boggle
9017 -d jim.d -bar -baz -boggle
9020 @item %@{S:X; T:Y; :D@}
9022 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
9023 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
9024 be as many clauses as you need. This may be combined with @code{.},
9025 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
9030 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
9031 construct may contain other nested @samp{%} constructs or spaces, or
9032 even newlines. They are processed as usual, as described above.
9033 Trailing white space in @code{X} is ignored. White space may also
9034 appear anywhere on the left side of the colon in these constructs,
9035 except between @code{.} or @code{*} and the corresponding word.
9037 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
9038 handled specifically in these constructs. If another value of
9039 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
9040 @option{-W} switch is found later in the command line, the earlier
9041 switch value is ignored, except with @{@code{S}*@} where @code{S} is
9042 just one letter, which passes all matching options.
9044 The character @samp{|} at the beginning of the predicate text is used to
9045 indicate that a command should be piped to the following command, but
9046 only if @option{-pipe} is specified.
9048 It is built into GCC which switches take arguments and which do not.
9049 (You might think it would be useful to generalize this to allow each
9050 compiler's spec to say which switches take arguments. But this cannot
9051 be done in a consistent fashion. GCC cannot even decide which input
9052 files have been specified without knowing which switches take arguments,
9053 and it must know which input files to compile in order to tell which
9056 GCC also knows implicitly that arguments starting in @option{-l} are to be
9057 treated as compiler output files, and passed to the linker in their
9058 proper position among the other output files.
9060 @c man begin OPTIONS
9062 @node Target Options
9063 @section Specifying Target Machine and Compiler Version
9064 @cindex target options
9065 @cindex cross compiling
9066 @cindex specifying machine version
9067 @cindex specifying compiler version and target machine
9068 @cindex compiler version, specifying
9069 @cindex target machine, specifying
9071 The usual way to run GCC is to run the executable called @file{gcc}, or
9072 @file{<machine>-gcc} when cross-compiling, or
9073 @file{<machine>-gcc-<version>} to run a version other than the one that
9074 was installed last. Sometimes this is inconvenient, so GCC provides
9075 options that will switch to another cross-compiler or version.
9078 @item -b @var{machine}
9080 The argument @var{machine} specifies the target machine for compilation.
9082 The value to use for @var{machine} is the same as was specified as the
9083 machine type when configuring GCC as a cross-compiler. For
9084 example, if a cross-compiler was configured with @samp{configure
9085 arm-elf}, meaning to compile for an arm processor with elf binaries,
9086 then you would specify @option{-b arm-elf} to run that cross compiler.
9087 Because there are other options beginning with @option{-b}, the
9088 configuration must contain a hyphen, or @option{-b} alone should be one
9089 argument followed by the configuration in the next argument.
9091 @item -V @var{version}
9093 The argument @var{version} specifies which version of GCC to run.
9094 This is useful when multiple versions are installed. For example,
9095 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
9098 The @option{-V} and @option{-b} options work by running the
9099 @file{<machine>-gcc-<version>} executable, so there's no real reason to
9100 use them if you can just run that directly.
9102 @node Submodel Options
9103 @section Hardware Models and Configurations
9104 @cindex submodel options
9105 @cindex specifying hardware config
9106 @cindex hardware models and configurations, specifying
9107 @cindex machine dependent options
9109 Earlier we discussed the standard option @option{-b} which chooses among
9110 different installed compilers for completely different target
9111 machines, such as VAX vs.@: 68000 vs.@: 80386.
9113 In addition, each of these target machine types can have its own
9114 special options, starting with @samp{-m}, to choose among various
9115 hardware models or configurations---for example, 68010 vs 68020,
9116 floating coprocessor or none. A single installed version of the
9117 compiler can compile for any model or configuration, according to the
9120 Some configurations of the compiler also support additional special
9121 options, usually for compatibility with other compilers on the same
9124 @c This list is ordered alphanumerically by subsection name.
9125 @c It should be the same order and spelling as these options are listed
9126 @c in Machine Dependent Options
9132 * Blackfin Options::
9136 * DEC Alpha Options::
9137 * DEC Alpha/VMS Options::
9140 * GNU/Linux Options::
9143 * i386 and x86-64 Options::
9144 * i386 and x86-64 Windows Options::
9156 * picoChip Options::
9158 * RS/6000 and PowerPC Options::
9159 * S/390 and zSeries Options::
9164 * System V Options::
9169 * Xstormy16 Options::
9175 @subsection ARC Options
9178 These options are defined for ARC implementations:
9183 Compile code for little endian mode. This is the default.
9187 Compile code for big endian mode.
9190 @opindex mmangle-cpu
9191 Prepend the name of the cpu to all public symbol names.
9192 In multiple-processor systems, there are many ARC variants with different
9193 instruction and register set characteristics. This flag prevents code
9194 compiled for one cpu to be linked with code compiled for another.
9195 No facility exists for handling variants that are ``almost identical''.
9196 This is an all or nothing option.
9198 @item -mcpu=@var{cpu}
9200 Compile code for ARC variant @var{cpu}.
9201 Which variants are supported depend on the configuration.
9202 All variants support @option{-mcpu=base}, this is the default.
9204 @item -mtext=@var{text-section}
9205 @itemx -mdata=@var{data-section}
9206 @itemx -mrodata=@var{readonly-data-section}
9210 Put functions, data, and readonly data in @var{text-section},
9211 @var{data-section}, and @var{readonly-data-section} respectively
9212 by default. This can be overridden with the @code{section} attribute.
9213 @xref{Variable Attributes}.
9215 @item -mfix-cortex-m3-ldrd
9216 @opindex mfix-cortex-m3-ldrd
9217 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
9218 with overlapping destination and base registers are used. This option avoids
9219 generating these instructions. This option is enabled by default when
9220 @option{-mcpu=cortex-m3} is specified.
9225 @subsection ARM Options
9228 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
9232 @item -mabi=@var{name}
9234 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
9235 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
9238 @opindex mapcs-frame
9239 Generate a stack frame that is compliant with the ARM Procedure Call
9240 Standard for all functions, even if this is not strictly necessary for
9241 correct execution of the code. Specifying @option{-fomit-frame-pointer}
9242 with this option will cause the stack frames not to be generated for
9243 leaf functions. The default is @option{-mno-apcs-frame}.
9247 This is a synonym for @option{-mapcs-frame}.
9250 @c not currently implemented
9251 @item -mapcs-stack-check
9252 @opindex mapcs-stack-check
9253 Generate code to check the amount of stack space available upon entry to
9254 every function (that actually uses some stack space). If there is
9255 insufficient space available then either the function
9256 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
9257 called, depending upon the amount of stack space required. The run time
9258 system is required to provide these functions. The default is
9259 @option{-mno-apcs-stack-check}, since this produces smaller code.
9261 @c not currently implemented
9263 @opindex mapcs-float
9264 Pass floating point arguments using the float point registers. This is
9265 one of the variants of the APCS@. This option is recommended if the
9266 target hardware has a floating point unit or if a lot of floating point
9267 arithmetic is going to be performed by the code. The default is
9268 @option{-mno-apcs-float}, since integer only code is slightly increased in
9269 size if @option{-mapcs-float} is used.
9271 @c not currently implemented
9272 @item -mapcs-reentrant
9273 @opindex mapcs-reentrant
9274 Generate reentrant, position independent code. The default is
9275 @option{-mno-apcs-reentrant}.
9278 @item -mthumb-interwork
9279 @opindex mthumb-interwork
9280 Generate code which supports calling between the ARM and Thumb
9281 instruction sets. Without this option the two instruction sets cannot
9282 be reliably used inside one program. The default is
9283 @option{-mno-thumb-interwork}, since slightly larger code is generated
9284 when @option{-mthumb-interwork} is specified.
9286 @item -mno-sched-prolog
9287 @opindex mno-sched-prolog
9288 Prevent the reordering of instructions in the function prolog, or the
9289 merging of those instruction with the instructions in the function's
9290 body. This means that all functions will start with a recognizable set
9291 of instructions (or in fact one of a choice from a small set of
9292 different function prologues), and this information can be used to
9293 locate the start if functions inside an executable piece of code. The
9294 default is @option{-msched-prolog}.
9296 @item -mfloat-abi=@var{name}
9298 Specifies which floating-point ABI to use. Permissible values
9299 are: @samp{soft}, @samp{softfp} and @samp{hard}.
9301 Specifying @samp{soft} causes GCC to generate output containing
9302 library calls for floating-point operations.
9303 @samp{softfp} allows the generation of code using hardware floating-point
9304 instructions, but still uses the soft-float calling conventions.
9305 @samp{hard} allows generation of floating-point instructions
9306 and uses FPU-specific calling conventions.
9308 Using @option{-mfloat-abi=hard} with VFP coprocessors is not supported.
9309 Use @option{-mfloat-abi=softfp} with the appropriate @option{-mfpu} option
9310 to allow the compiler to generate code that makes use of the hardware
9311 floating-point capabilities for these CPUs.
9313 The default depends on the specific target configuration. Note that
9314 the hard-float and soft-float ABIs are not link-compatible; you must
9315 compile your entire program with the same ABI, and link with a
9316 compatible set of libraries.
9319 @opindex mhard-float
9320 Equivalent to @option{-mfloat-abi=hard}.
9323 @opindex msoft-float
9324 Equivalent to @option{-mfloat-abi=soft}.
9326 @item -mlittle-endian
9327 @opindex mlittle-endian
9328 Generate code for a processor running in little-endian mode. This is
9329 the default for all standard configurations.
9332 @opindex mbig-endian
9333 Generate code for a processor running in big-endian mode; the default is
9334 to compile code for a little-endian processor.
9336 @item -mwords-little-endian
9337 @opindex mwords-little-endian
9338 This option only applies when generating code for big-endian processors.
9339 Generate code for a little-endian word order but a big-endian byte
9340 order. That is, a byte order of the form @samp{32107654}. Note: this
9341 option should only be used if you require compatibility with code for
9342 big-endian ARM processors generated by versions of the compiler prior to
9345 @item -mcpu=@var{name}
9347 This specifies the name of the target ARM processor. GCC uses this name
9348 to determine what kind of instructions it can emit when generating
9349 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
9350 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
9351 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
9352 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
9353 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
9355 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
9356 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
9357 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
9358 @samp{strongarm1110},
9359 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
9360 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
9361 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
9362 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
9363 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
9364 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
9365 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
9366 @samp{cortex-a8}, @samp{cortex-a9},
9367 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-m3},
9370 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9372 @item -mtune=@var{name}
9374 This option is very similar to the @option{-mcpu=} option, except that
9375 instead of specifying the actual target processor type, and hence
9376 restricting which instructions can be used, it specifies that GCC should
9377 tune the performance of the code as if the target were of the type
9378 specified in this option, but still choosing the instructions that it
9379 will generate based on the cpu specified by a @option{-mcpu=} option.
9380 For some ARM implementations better performance can be obtained by using
9383 @item -march=@var{name}
9385 This specifies the name of the target ARM architecture. GCC uses this
9386 name to determine what kind of instructions it can emit when generating
9387 assembly code. This option can be used in conjunction with or instead
9388 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
9389 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
9390 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
9391 @samp{armv6}, @samp{armv6j},
9392 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
9393 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
9394 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9396 @item -mfpu=@var{name}
9397 @itemx -mfpe=@var{number}
9398 @itemx -mfp=@var{number}
9402 This specifies what floating point hardware (or hardware emulation) is
9403 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
9404 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-d16},
9405 @samp{neon}, and @samp{neon-fp16}. @option{-mfp} and @option{-mfpe}
9406 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
9407 with older versions of GCC@.
9409 If @option{-msoft-float} is specified this specifies the format of
9410 floating point values.
9412 @item -mfp16-format=@var{name}
9413 @opindex mfp16-format
9414 Specify the format of the @code{__fp16} half-precision floating-point type.
9415 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
9416 the default is @samp{none}, in which case the @code{__fp16} type is not
9417 defined. @xref{Half-Precision}, for more information.
9419 @item -mstructure-size-boundary=@var{n}
9420 @opindex mstructure-size-boundary
9421 The size of all structures and unions will be rounded up to a multiple
9422 of the number of bits set by this option. Permissible values are 8, 32
9423 and 64. The default value varies for different toolchains. For the COFF
9424 targeted toolchain the default value is 8. A value of 64 is only allowed
9425 if the underlying ABI supports it.
9427 Specifying the larger number can produce faster, more efficient code, but
9428 can also increase the size of the program. Different values are potentially
9429 incompatible. Code compiled with one value cannot necessarily expect to
9430 work with code or libraries compiled with another value, if they exchange
9431 information using structures or unions.
9433 @item -mabort-on-noreturn
9434 @opindex mabort-on-noreturn
9435 Generate a call to the function @code{abort} at the end of a
9436 @code{noreturn} function. It will be executed if the function tries to
9440 @itemx -mno-long-calls
9441 @opindex mlong-calls
9442 @opindex mno-long-calls
9443 Tells the compiler to perform function calls by first loading the
9444 address of the function into a register and then performing a subroutine
9445 call on this register. This switch is needed if the target function
9446 will lie outside of the 64 megabyte addressing range of the offset based
9447 version of subroutine call instruction.
9449 Even if this switch is enabled, not all function calls will be turned
9450 into long calls. The heuristic is that static functions, functions
9451 which have the @samp{short-call} attribute, functions that are inside
9452 the scope of a @samp{#pragma no_long_calls} directive and functions whose
9453 definitions have already been compiled within the current compilation
9454 unit, will not be turned into long calls. The exception to this rule is
9455 that weak function definitions, functions with the @samp{long-call}
9456 attribute or the @samp{section} attribute, and functions that are within
9457 the scope of a @samp{#pragma long_calls} directive, will always be
9458 turned into long calls.
9460 This feature is not enabled by default. Specifying
9461 @option{-mno-long-calls} will restore the default behavior, as will
9462 placing the function calls within the scope of a @samp{#pragma
9463 long_calls_off} directive. Note these switches have no effect on how
9464 the compiler generates code to handle function calls via function
9467 @item -msingle-pic-base
9468 @opindex msingle-pic-base
9469 Treat the register used for PIC addressing as read-only, rather than
9470 loading it in the prologue for each function. The run-time system is
9471 responsible for initializing this register with an appropriate value
9472 before execution begins.
9474 @item -mpic-register=@var{reg}
9475 @opindex mpic-register
9476 Specify the register to be used for PIC addressing. The default is R10
9477 unless stack-checking is enabled, when R9 is used.
9479 @item -mcirrus-fix-invalid-insns
9480 @opindex mcirrus-fix-invalid-insns
9481 @opindex mno-cirrus-fix-invalid-insns
9482 Insert NOPs into the instruction stream to in order to work around
9483 problems with invalid Maverick instruction combinations. This option
9484 is only valid if the @option{-mcpu=ep9312} option has been used to
9485 enable generation of instructions for the Cirrus Maverick floating
9486 point co-processor. This option is not enabled by default, since the
9487 problem is only present in older Maverick implementations. The default
9488 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
9491 @item -mpoke-function-name
9492 @opindex mpoke-function-name
9493 Write the name of each function into the text section, directly
9494 preceding the function prologue. The generated code is similar to this:
9498 .ascii "arm_poke_function_name", 0
9501 .word 0xff000000 + (t1 - t0)
9502 arm_poke_function_name
9504 stmfd sp!, @{fp, ip, lr, pc@}
9508 When performing a stack backtrace, code can inspect the value of
9509 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
9510 location @code{pc - 12} and the top 8 bits are set, then we know that
9511 there is a function name embedded immediately preceding this location
9512 and has length @code{((pc[-3]) & 0xff000000)}.
9516 Generate code for the Thumb instruction set. The default is to
9517 use the 32-bit ARM instruction set.
9518 This option automatically enables either 16-bit Thumb-1 or
9519 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
9520 and @option{-march=@var{name}} options. This option is not passed to the
9521 assembler. If you want to force assembler files to be interpreted as Thumb code,
9522 either add a @samp{.thumb} directive to the source or pass the @option{-mthumb}
9523 option directly to the assembler by prefixing it with @option{-Wa}.
9526 @opindex mtpcs-frame
9527 Generate a stack frame that is compliant with the Thumb Procedure Call
9528 Standard for all non-leaf functions. (A leaf function is one that does
9529 not call any other functions.) The default is @option{-mno-tpcs-frame}.
9531 @item -mtpcs-leaf-frame
9532 @opindex mtpcs-leaf-frame
9533 Generate a stack frame that is compliant with the Thumb Procedure Call
9534 Standard for all leaf functions. (A leaf function is one that does
9535 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
9537 @item -mcallee-super-interworking
9538 @opindex mcallee-super-interworking
9539 Gives all externally visible functions in the file being compiled an ARM
9540 instruction set header which switches to Thumb mode before executing the
9541 rest of the function. This allows these functions to be called from
9542 non-interworking code. This option is not valid in AAPCS configurations
9543 because interworking is enabled by default.
9545 @item -mcaller-super-interworking
9546 @opindex mcaller-super-interworking
9547 Allows calls via function pointers (including virtual functions) to
9548 execute correctly regardless of whether the target code has been
9549 compiled for interworking or not. There is a small overhead in the cost
9550 of executing a function pointer if this option is enabled. This option
9551 is not valid in AAPCS configurations because interworking is enabled
9554 @item -mtp=@var{name}
9556 Specify the access model for the thread local storage pointer. The valid
9557 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
9558 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
9559 (supported in the arm6k architecture), and @option{auto}, which uses the
9560 best available method for the selected processor. The default setting is
9563 @item -mword-relocations
9564 @opindex mword-relocations
9565 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
9566 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
9567 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
9573 @subsection AVR Options
9576 These options are defined for AVR implementations:
9579 @item -mmcu=@var{mcu}
9581 Specify ATMEL AVR instruction set or MCU type.
9583 Instruction set avr1 is for the minimal AVR core, not supported by the C
9584 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
9585 attiny11, attiny12, attiny15, attiny28).
9587 Instruction set avr2 (default) is for the classic AVR core with up to
9588 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
9589 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
9590 at90c8534, at90s8535).
9592 Instruction set avr3 is for the classic AVR core with up to 128K program
9593 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
9595 Instruction set avr4 is for the enhanced AVR core with up to 8K program
9596 memory space (MCU types: atmega8, atmega83, atmega85).
9598 Instruction set avr5 is for the enhanced AVR core with up to 128K program
9599 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
9600 atmega64, atmega128, at43usb355, at94k).
9604 Output instruction sizes to the asm file.
9606 @item -minit-stack=@var{N}
9607 @opindex minit-stack
9608 Specify the initial stack address, which may be a symbol or numeric value,
9609 @samp{__stack} is the default.
9611 @item -mno-interrupts
9612 @opindex mno-interrupts
9613 Generated code is not compatible with hardware interrupts.
9614 Code size will be smaller.
9616 @item -mcall-prologues
9617 @opindex mcall-prologues
9618 Functions prologues/epilogues expanded as call to appropriate
9619 subroutines. Code size will be smaller.
9622 @opindex mtiny-stack
9623 Change only the low 8 bits of the stack pointer.
9627 Assume int to be 8 bit integer. This affects the sizes of all types: A
9628 char will be 1 byte, an int will be 1 byte, a long will be 2 bytes
9629 and long long will be 4 bytes. Please note that this option does not
9630 comply to the C standards, but it will provide you with smaller code
9634 @node Blackfin Options
9635 @subsection Blackfin Options
9636 @cindex Blackfin Options
9639 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
9641 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
9642 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
9643 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
9644 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
9645 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
9646 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
9648 The optional @var{sirevision} specifies the silicon revision of the target
9649 Blackfin processor. Any workarounds available for the targeted silicon revision
9650 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
9651 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
9652 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
9653 hexadecimal digits representing the major and minor numbers in the silicon
9654 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
9655 is not defined. If @var{sirevision} is @samp{any}, the
9656 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
9657 If this optional @var{sirevision} is not used, GCC assumes the latest known
9658 silicon revision of the targeted Blackfin processor.
9660 Support for @samp{bf561} is incomplete. For @samp{bf561},
9661 Only the processor macro is defined.
9662 Without this option, @samp{bf532} is used as the processor by default.
9663 The corresponding predefined processor macros for @var{cpu} is to
9664 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
9665 provided by libgloss to be linked in if @option{-msim} is not given.
9669 Specifies that the program will be run on the simulator. This causes
9670 the simulator BSP provided by libgloss to be linked in. This option
9671 has effect only for @samp{bfin-elf} toolchain.
9672 Certain other options, such as @option{-mid-shared-library} and
9673 @option{-mfdpic}, imply @option{-msim}.
9675 @item -momit-leaf-frame-pointer
9676 @opindex momit-leaf-frame-pointer
9677 Don't keep the frame pointer in a register for leaf functions. This
9678 avoids the instructions to save, set up and restore frame pointers and
9679 makes an extra register available in leaf functions. The option
9680 @option{-fomit-frame-pointer} removes the frame pointer for all functions
9681 which might make debugging harder.
9683 @item -mspecld-anomaly
9684 @opindex mspecld-anomaly
9685 When enabled, the compiler will ensure that the generated code does not
9686 contain speculative loads after jump instructions. If this option is used,
9687 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
9689 @item -mno-specld-anomaly
9690 @opindex mno-specld-anomaly
9691 Don't generate extra code to prevent speculative loads from occurring.
9693 @item -mcsync-anomaly
9694 @opindex mcsync-anomaly
9695 When enabled, the compiler will ensure that the generated code does not
9696 contain CSYNC or SSYNC instructions too soon after conditional branches.
9697 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
9699 @item -mno-csync-anomaly
9700 @opindex mno-csync-anomaly
9701 Don't generate extra code to prevent CSYNC or SSYNC instructions from
9702 occurring too soon after a conditional branch.
9706 When enabled, the compiler is free to take advantage of the knowledge that
9707 the entire program fits into the low 64k of memory.
9710 @opindex mno-low-64k
9711 Assume that the program is arbitrarily large. This is the default.
9713 @item -mstack-check-l1
9714 @opindex mstack-check-l1
9715 Do stack checking using information placed into L1 scratchpad memory by the
9718 @item -mid-shared-library
9719 @opindex mid-shared-library
9720 Generate code that supports shared libraries via the library ID method.
9721 This allows for execute in place and shared libraries in an environment
9722 without virtual memory management. This option implies @option{-fPIC}.
9723 With a @samp{bfin-elf} target, this option implies @option{-msim}.
9725 @item -mno-id-shared-library
9726 @opindex mno-id-shared-library
9727 Generate code that doesn't assume ID based shared libraries are being used.
9728 This is the default.
9730 @item -mleaf-id-shared-library
9731 @opindex mleaf-id-shared-library
9732 Generate code that supports shared libraries via the library ID method,
9733 but assumes that this library or executable won't link against any other
9734 ID shared libraries. That allows the compiler to use faster code for jumps
9737 @item -mno-leaf-id-shared-library
9738 @opindex mno-leaf-id-shared-library
9739 Do not assume that the code being compiled won't link against any ID shared
9740 libraries. Slower code will be generated for jump and call insns.
9742 @item -mshared-library-id=n
9743 @opindex mshared-library-id
9744 Specified the identification number of the ID based shared library being
9745 compiled. Specifying a value of 0 will generate more compact code, specifying
9746 other values will force the allocation of that number to the current
9747 library but is no more space or time efficient than omitting this option.
9751 Generate code that allows the data segment to be located in a different
9752 area of memory from the text segment. This allows for execute in place in
9753 an environment without virtual memory management by eliminating relocations
9754 against the text section.
9757 @opindex mno-sep-data
9758 Generate code that assumes that the data segment follows the text segment.
9759 This is the default.
9762 @itemx -mno-long-calls
9763 @opindex mlong-calls
9764 @opindex mno-long-calls
9765 Tells the compiler to perform function calls by first loading the
9766 address of the function into a register and then performing a subroutine
9767 call on this register. This switch is needed if the target function
9768 will lie outside of the 24 bit addressing range of the offset based
9769 version of subroutine call instruction.
9771 This feature is not enabled by default. Specifying
9772 @option{-mno-long-calls} will restore the default behavior. Note these
9773 switches have no effect on how the compiler generates code to handle
9774 function calls via function pointers.
9778 Link with the fast floating-point library. This library relaxes some of
9779 the IEEE floating-point standard's rules for checking inputs against
9780 Not-a-Number (NAN), in the interest of performance.
9783 @opindex minline-plt
9784 Enable inlining of PLT entries in function calls to functions that are
9785 not known to bind locally. It has no effect without @option{-mfdpic}.
9789 Build standalone application for multicore Blackfin processor. Proper
9790 start files and link scripts will be used to support multicore.
9791 This option defines @code{__BFIN_MULTICORE}. It can only be used with
9792 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
9793 @option{-mcorea} or @option{-mcoreb}. If it's used without
9794 @option{-mcorea} or @option{-mcoreb}, single application/dual core
9795 programming model is used. In this model, the main function of Core B
9796 should be named as coreb_main. If it's used with @option{-mcorea} or
9797 @option{-mcoreb}, one application per core programming model is used.
9798 If this option is not used, single core application programming
9803 Build standalone application for Core A of BF561 when using
9804 one application per core programming model. Proper start files
9805 and link scripts will be used to support Core A. This option
9806 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
9810 Build standalone application for Core B of BF561 when using
9811 one application per core programming model. Proper start files
9812 and link scripts will be used to support Core B. This option
9813 defines @code{__BFIN_COREB}. When this option is used, coreb_main
9814 should be used instead of main. It must be used with
9815 @option{-mmulticore}.
9819 Build standalone application for SDRAM. Proper start files and
9820 link scripts will be used to put the application into SDRAM.
9821 Loader should initialize SDRAM before loading the application
9822 into SDRAM. This option defines @code{__BFIN_SDRAM}.
9826 Assume that ICPLBs are enabled at runtime. This has an effect on certain
9827 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
9828 are enabled; for standalone applications the default is off.
9832 @subsection CRIS Options
9833 @cindex CRIS Options
9835 These options are defined specifically for the CRIS ports.
9838 @item -march=@var{architecture-type}
9839 @itemx -mcpu=@var{architecture-type}
9842 Generate code for the specified architecture. The choices for
9843 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
9844 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
9845 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
9848 @item -mtune=@var{architecture-type}
9850 Tune to @var{architecture-type} everything applicable about the generated
9851 code, except for the ABI and the set of available instructions. The
9852 choices for @var{architecture-type} are the same as for
9853 @option{-march=@var{architecture-type}}.
9855 @item -mmax-stack-frame=@var{n}
9856 @opindex mmax-stack-frame
9857 Warn when the stack frame of a function exceeds @var{n} bytes.
9863 The options @option{-metrax4} and @option{-metrax100} are synonyms for
9864 @option{-march=v3} and @option{-march=v8} respectively.
9866 @item -mmul-bug-workaround
9867 @itemx -mno-mul-bug-workaround
9868 @opindex mmul-bug-workaround
9869 @opindex mno-mul-bug-workaround
9870 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
9871 models where it applies. This option is active by default.
9875 Enable CRIS-specific verbose debug-related information in the assembly
9876 code. This option also has the effect to turn off the @samp{#NO_APP}
9877 formatted-code indicator to the assembler at the beginning of the
9882 Do not use condition-code results from previous instruction; always emit
9883 compare and test instructions before use of condition codes.
9885 @item -mno-side-effects
9886 @opindex mno-side-effects
9887 Do not emit instructions with side-effects in addressing modes other than
9891 @itemx -mno-stack-align
9893 @itemx -mno-data-align
9894 @itemx -mconst-align
9895 @itemx -mno-const-align
9896 @opindex mstack-align
9897 @opindex mno-stack-align
9898 @opindex mdata-align
9899 @opindex mno-data-align
9900 @opindex mconst-align
9901 @opindex mno-const-align
9902 These options (no-options) arranges (eliminate arrangements) for the
9903 stack-frame, individual data and constants to be aligned for the maximum
9904 single data access size for the chosen CPU model. The default is to
9905 arrange for 32-bit alignment. ABI details such as structure layout are
9906 not affected by these options.
9914 Similar to the stack- data- and const-align options above, these options
9915 arrange for stack-frame, writable data and constants to all be 32-bit,
9916 16-bit or 8-bit aligned. The default is 32-bit alignment.
9918 @item -mno-prologue-epilogue
9919 @itemx -mprologue-epilogue
9920 @opindex mno-prologue-epilogue
9921 @opindex mprologue-epilogue
9922 With @option{-mno-prologue-epilogue}, the normal function prologue and
9923 epilogue that sets up the stack-frame are omitted and no return
9924 instructions or return sequences are generated in the code. Use this
9925 option only together with visual inspection of the compiled code: no
9926 warnings or errors are generated when call-saved registers must be saved,
9927 or storage for local variable needs to be allocated.
9933 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
9934 instruction sequences that load addresses for functions from the PLT part
9935 of the GOT rather than (traditional on other architectures) calls to the
9936 PLT@. The default is @option{-mgotplt}.
9940 Legacy no-op option only recognized with the cris-axis-elf and
9941 cris-axis-linux-gnu targets.
9945 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
9949 This option, recognized for the cris-axis-elf arranges
9950 to link with input-output functions from a simulator library. Code,
9951 initialized data and zero-initialized data are allocated consecutively.
9955 Like @option{-sim}, but pass linker options to locate initialized data at
9956 0x40000000 and zero-initialized data at 0x80000000.
9960 @subsection CRX Options
9963 These options are defined specifically for the CRX ports.
9969 Enable the use of multiply-accumulate instructions. Disabled by default.
9973 Push instructions will be used to pass outgoing arguments when functions
9974 are called. Enabled by default.
9977 @node Darwin Options
9978 @subsection Darwin Options
9979 @cindex Darwin options
9981 These options are defined for all architectures running the Darwin operating
9984 FSF GCC on Darwin does not create ``fat'' object files; it will create
9985 an object file for the single architecture that it was built to
9986 target. Apple's GCC on Darwin does create ``fat'' files if multiple
9987 @option{-arch} options are used; it does so by running the compiler or
9988 linker multiple times and joining the results together with
9991 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
9992 @samp{i686}) is determined by the flags that specify the ISA
9993 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
9994 @option{-force_cpusubtype_ALL} option can be used to override this.
9996 The Darwin tools vary in their behavior when presented with an ISA
9997 mismatch. The assembler, @file{as}, will only permit instructions to
9998 be used that are valid for the subtype of the file it is generating,
9999 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
10000 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
10001 and print an error if asked to create a shared library with a less
10002 restrictive subtype than its input files (for instance, trying to put
10003 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
10004 for executables, @file{ld}, will quietly give the executable the most
10005 restrictive subtype of any of its input files.
10010 Add the framework directory @var{dir} to the head of the list of
10011 directories to be searched for header files. These directories are
10012 interleaved with those specified by @option{-I} options and are
10013 scanned in a left-to-right order.
10015 A framework directory is a directory with frameworks in it. A
10016 framework is a directory with a @samp{"Headers"} and/or
10017 @samp{"PrivateHeaders"} directory contained directly in it that ends
10018 in @samp{".framework"}. The name of a framework is the name of this
10019 directory excluding the @samp{".framework"}. Headers associated with
10020 the framework are found in one of those two directories, with
10021 @samp{"Headers"} being searched first. A subframework is a framework
10022 directory that is in a framework's @samp{"Frameworks"} directory.
10023 Includes of subframework headers can only appear in a header of a
10024 framework that contains the subframework, or in a sibling subframework
10025 header. Two subframeworks are siblings if they occur in the same
10026 framework. A subframework should not have the same name as a
10027 framework, a warning will be issued if this is violated. Currently a
10028 subframework cannot have subframeworks, in the future, the mechanism
10029 may be extended to support this. The standard frameworks can be found
10030 in @samp{"/System/Library/Frameworks"} and
10031 @samp{"/Library/Frameworks"}. An example include looks like
10032 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
10033 the name of the framework and header.h is found in the
10034 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
10036 @item -iframework@var{dir}
10037 @opindex iframework
10038 Like @option{-F} except the directory is a treated as a system
10039 directory. The main difference between this @option{-iframework} and
10040 @option{-F} is that with @option{-iframework} the compiler does not
10041 warn about constructs contained within header files found via
10042 @var{dir}. This option is valid only for the C family of languages.
10046 Emit debugging information for symbols that are used. For STABS
10047 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
10048 This is by default ON@.
10052 Emit debugging information for all symbols and types.
10054 @item -mmacosx-version-min=@var{version}
10055 The earliest version of MacOS X that this executable will run on
10056 is @var{version}. Typical values of @var{version} include @code{10.1},
10057 @code{10.2}, and @code{10.3.9}.
10059 If the compiler was built to use the system's headers by default,
10060 then the default for this option is the system version on which the
10061 compiler is running, otherwise the default is to make choices which
10062 are compatible with as many systems and code bases as possible.
10066 Enable kernel development mode. The @option{-mkernel} option sets
10067 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
10068 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
10069 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
10070 applicable. This mode also sets @option{-mno-altivec},
10071 @option{-msoft-float}, @option{-fno-builtin} and
10072 @option{-mlong-branch} for PowerPC targets.
10074 @item -mone-byte-bool
10075 @opindex mone-byte-bool
10076 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
10077 By default @samp{sizeof(bool)} is @samp{4} when compiling for
10078 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
10079 option has no effect on x86.
10081 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
10082 to generate code that is not binary compatible with code generated
10083 without that switch. Using this switch may require recompiling all
10084 other modules in a program, including system libraries. Use this
10085 switch to conform to a non-default data model.
10087 @item -mfix-and-continue
10088 @itemx -ffix-and-continue
10089 @itemx -findirect-data
10090 @opindex mfix-and-continue
10091 @opindex ffix-and-continue
10092 @opindex findirect-data
10093 Generate code suitable for fast turn around development. Needed to
10094 enable gdb to dynamically load @code{.o} files into already running
10095 programs. @option{-findirect-data} and @option{-ffix-and-continue}
10096 are provided for backwards compatibility.
10100 Loads all members of static archive libraries.
10101 See man ld(1) for more information.
10103 @item -arch_errors_fatal
10104 @opindex arch_errors_fatal
10105 Cause the errors having to do with files that have the wrong architecture
10108 @item -bind_at_load
10109 @opindex bind_at_load
10110 Causes the output file to be marked such that the dynamic linker will
10111 bind all undefined references when the file is loaded or launched.
10115 Produce a Mach-o bundle format file.
10116 See man ld(1) for more information.
10118 @item -bundle_loader @var{executable}
10119 @opindex bundle_loader
10120 This option specifies the @var{executable} that will be loading the build
10121 output file being linked. See man ld(1) for more information.
10124 @opindex dynamiclib
10125 When passed this option, GCC will produce a dynamic library instead of
10126 an executable when linking, using the Darwin @file{libtool} command.
10128 @item -force_cpusubtype_ALL
10129 @opindex force_cpusubtype_ALL
10130 This causes GCC's output file to have the @var{ALL} subtype, instead of
10131 one controlled by the @option{-mcpu} or @option{-march} option.
10133 @item -allowable_client @var{client_name}
10134 @itemx -client_name
10135 @itemx -compatibility_version
10136 @itemx -current_version
10138 @itemx -dependency-file
10140 @itemx -dylinker_install_name
10142 @itemx -exported_symbols_list
10144 @itemx -flat_namespace
10145 @itemx -force_flat_namespace
10146 @itemx -headerpad_max_install_names
10149 @itemx -install_name
10150 @itemx -keep_private_externs
10151 @itemx -multi_module
10152 @itemx -multiply_defined
10153 @itemx -multiply_defined_unused
10155 @itemx -no_dead_strip_inits_and_terms
10156 @itemx -nofixprebinding
10157 @itemx -nomultidefs
10159 @itemx -noseglinkedit
10160 @itemx -pagezero_size
10162 @itemx -prebind_all_twolevel_modules
10163 @itemx -private_bundle
10164 @itemx -read_only_relocs
10166 @itemx -sectobjectsymbols
10170 @itemx -sectobjectsymbols
10173 @itemx -segs_read_only_addr
10174 @itemx -segs_read_write_addr
10175 @itemx -seg_addr_table
10176 @itemx -seg_addr_table_filename
10177 @itemx -seglinkedit
10179 @itemx -segs_read_only_addr
10180 @itemx -segs_read_write_addr
10181 @itemx -single_module
10183 @itemx -sub_library
10184 @itemx -sub_umbrella
10185 @itemx -twolevel_namespace
10188 @itemx -unexported_symbols_list
10189 @itemx -weak_reference_mismatches
10190 @itemx -whatsloaded
10191 @opindex allowable_client
10192 @opindex client_name
10193 @opindex compatibility_version
10194 @opindex current_version
10195 @opindex dead_strip
10196 @opindex dependency-file
10197 @opindex dylib_file
10198 @opindex dylinker_install_name
10200 @opindex exported_symbols_list
10202 @opindex flat_namespace
10203 @opindex force_flat_namespace
10204 @opindex headerpad_max_install_names
10205 @opindex image_base
10207 @opindex install_name
10208 @opindex keep_private_externs
10209 @opindex multi_module
10210 @opindex multiply_defined
10211 @opindex multiply_defined_unused
10212 @opindex noall_load
10213 @opindex no_dead_strip_inits_and_terms
10214 @opindex nofixprebinding
10215 @opindex nomultidefs
10217 @opindex noseglinkedit
10218 @opindex pagezero_size
10220 @opindex prebind_all_twolevel_modules
10221 @opindex private_bundle
10222 @opindex read_only_relocs
10224 @opindex sectobjectsymbols
10227 @opindex sectcreate
10228 @opindex sectobjectsymbols
10231 @opindex segs_read_only_addr
10232 @opindex segs_read_write_addr
10233 @opindex seg_addr_table
10234 @opindex seg_addr_table_filename
10235 @opindex seglinkedit
10237 @opindex segs_read_only_addr
10238 @opindex segs_read_write_addr
10239 @opindex single_module
10241 @opindex sub_library
10242 @opindex sub_umbrella
10243 @opindex twolevel_namespace
10246 @opindex unexported_symbols_list
10247 @opindex weak_reference_mismatches
10248 @opindex whatsloaded
10249 These options are passed to the Darwin linker. The Darwin linker man page
10250 describes them in detail.
10253 @node DEC Alpha Options
10254 @subsection DEC Alpha Options
10256 These @samp{-m} options are defined for the DEC Alpha implementations:
10259 @item -mno-soft-float
10260 @itemx -msoft-float
10261 @opindex mno-soft-float
10262 @opindex msoft-float
10263 Use (do not use) the hardware floating-point instructions for
10264 floating-point operations. When @option{-msoft-float} is specified,
10265 functions in @file{libgcc.a} will be used to perform floating-point
10266 operations. Unless they are replaced by routines that emulate the
10267 floating-point operations, or compiled in such a way as to call such
10268 emulations routines, these routines will issue floating-point
10269 operations. If you are compiling for an Alpha without floating-point
10270 operations, you must ensure that the library is built so as not to call
10273 Note that Alpha implementations without floating-point operations are
10274 required to have floating-point registers.
10277 @itemx -mno-fp-regs
10279 @opindex mno-fp-regs
10280 Generate code that uses (does not use) the floating-point register set.
10281 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
10282 register set is not used, floating point operands are passed in integer
10283 registers as if they were integers and floating-point results are passed
10284 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
10285 so any function with a floating-point argument or return value called by code
10286 compiled with @option{-mno-fp-regs} must also be compiled with that
10289 A typical use of this option is building a kernel that does not use,
10290 and hence need not save and restore, any floating-point registers.
10294 The Alpha architecture implements floating-point hardware optimized for
10295 maximum performance. It is mostly compliant with the IEEE floating
10296 point standard. However, for full compliance, software assistance is
10297 required. This option generates code fully IEEE compliant code
10298 @emph{except} that the @var{inexact-flag} is not maintained (see below).
10299 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
10300 defined during compilation. The resulting code is less efficient but is
10301 able to correctly support denormalized numbers and exceptional IEEE
10302 values such as not-a-number and plus/minus infinity. Other Alpha
10303 compilers call this option @option{-ieee_with_no_inexact}.
10305 @item -mieee-with-inexact
10306 @opindex mieee-with-inexact
10307 This is like @option{-mieee} except the generated code also maintains
10308 the IEEE @var{inexact-flag}. Turning on this option causes the
10309 generated code to implement fully-compliant IEEE math. In addition to
10310 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
10311 macro. On some Alpha implementations the resulting code may execute
10312 significantly slower than the code generated by default. Since there is
10313 very little code that depends on the @var{inexact-flag}, you should
10314 normally not specify this option. Other Alpha compilers call this
10315 option @option{-ieee_with_inexact}.
10317 @item -mfp-trap-mode=@var{trap-mode}
10318 @opindex mfp-trap-mode
10319 This option controls what floating-point related traps are enabled.
10320 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
10321 The trap mode can be set to one of four values:
10325 This is the default (normal) setting. The only traps that are enabled
10326 are the ones that cannot be disabled in software (e.g., division by zero
10330 In addition to the traps enabled by @samp{n}, underflow traps are enabled
10334 Like @samp{u}, but the instructions are marked to be safe for software
10335 completion (see Alpha architecture manual for details).
10338 Like @samp{su}, but inexact traps are enabled as well.
10341 @item -mfp-rounding-mode=@var{rounding-mode}
10342 @opindex mfp-rounding-mode
10343 Selects the IEEE rounding mode. Other Alpha compilers call this option
10344 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
10349 Normal IEEE rounding mode. Floating point numbers are rounded towards
10350 the nearest machine number or towards the even machine number in case
10354 Round towards minus infinity.
10357 Chopped rounding mode. Floating point numbers are rounded towards zero.
10360 Dynamic rounding mode. A field in the floating point control register
10361 (@var{fpcr}, see Alpha architecture reference manual) controls the
10362 rounding mode in effect. The C library initializes this register for
10363 rounding towards plus infinity. Thus, unless your program modifies the
10364 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
10367 @item -mtrap-precision=@var{trap-precision}
10368 @opindex mtrap-precision
10369 In the Alpha architecture, floating point traps are imprecise. This
10370 means without software assistance it is impossible to recover from a
10371 floating trap and program execution normally needs to be terminated.
10372 GCC can generate code that can assist operating system trap handlers
10373 in determining the exact location that caused a floating point trap.
10374 Depending on the requirements of an application, different levels of
10375 precisions can be selected:
10379 Program precision. This option is the default and means a trap handler
10380 can only identify which program caused a floating point exception.
10383 Function precision. The trap handler can determine the function that
10384 caused a floating point exception.
10387 Instruction precision. The trap handler can determine the exact
10388 instruction that caused a floating point exception.
10391 Other Alpha compilers provide the equivalent options called
10392 @option{-scope_safe} and @option{-resumption_safe}.
10394 @item -mieee-conformant
10395 @opindex mieee-conformant
10396 This option marks the generated code as IEEE conformant. You must not
10397 use this option unless you also specify @option{-mtrap-precision=i} and either
10398 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
10399 is to emit the line @samp{.eflag 48} in the function prologue of the
10400 generated assembly file. Under DEC Unix, this has the effect that
10401 IEEE-conformant math library routines will be linked in.
10403 @item -mbuild-constants
10404 @opindex mbuild-constants
10405 Normally GCC examines a 32- or 64-bit integer constant to
10406 see if it can construct it from smaller constants in two or three
10407 instructions. If it cannot, it will output the constant as a literal and
10408 generate code to load it from the data segment at runtime.
10410 Use this option to require GCC to construct @emph{all} integer constants
10411 using code, even if it takes more instructions (the maximum is six).
10413 You would typically use this option to build a shared library dynamic
10414 loader. Itself a shared library, it must relocate itself in memory
10415 before it can find the variables and constants in its own data segment.
10421 Select whether to generate code to be assembled by the vendor-supplied
10422 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
10440 Indicate whether GCC should generate code to use the optional BWX,
10441 CIX, FIX and MAX instruction sets. The default is to use the instruction
10442 sets supported by the CPU type specified via @option{-mcpu=} option or that
10443 of the CPU on which GCC was built if none was specified.
10446 @itemx -mfloat-ieee
10447 @opindex mfloat-vax
10448 @opindex mfloat-ieee
10449 Generate code that uses (does not use) VAX F and G floating point
10450 arithmetic instead of IEEE single and double precision.
10452 @item -mexplicit-relocs
10453 @itemx -mno-explicit-relocs
10454 @opindex mexplicit-relocs
10455 @opindex mno-explicit-relocs
10456 Older Alpha assemblers provided no way to generate symbol relocations
10457 except via assembler macros. Use of these macros does not allow
10458 optimal instruction scheduling. GNU binutils as of version 2.12
10459 supports a new syntax that allows the compiler to explicitly mark
10460 which relocations should apply to which instructions. This option
10461 is mostly useful for debugging, as GCC detects the capabilities of
10462 the assembler when it is built and sets the default accordingly.
10465 @itemx -mlarge-data
10466 @opindex msmall-data
10467 @opindex mlarge-data
10468 When @option{-mexplicit-relocs} is in effect, static data is
10469 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
10470 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
10471 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
10472 16-bit relocations off of the @code{$gp} register. This limits the
10473 size of the small data area to 64KB, but allows the variables to be
10474 directly accessed via a single instruction.
10476 The default is @option{-mlarge-data}. With this option the data area
10477 is limited to just below 2GB@. Programs that require more than 2GB of
10478 data must use @code{malloc} or @code{mmap} to allocate the data in the
10479 heap instead of in the program's data segment.
10481 When generating code for shared libraries, @option{-fpic} implies
10482 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
10485 @itemx -mlarge-text
10486 @opindex msmall-text
10487 @opindex mlarge-text
10488 When @option{-msmall-text} is used, the compiler assumes that the
10489 code of the entire program (or shared library) fits in 4MB, and is
10490 thus reachable with a branch instruction. When @option{-msmall-data}
10491 is used, the compiler can assume that all local symbols share the
10492 same @code{$gp} value, and thus reduce the number of instructions
10493 required for a function call from 4 to 1.
10495 The default is @option{-mlarge-text}.
10497 @item -mcpu=@var{cpu_type}
10499 Set the instruction set and instruction scheduling parameters for
10500 machine type @var{cpu_type}. You can specify either the @samp{EV}
10501 style name or the corresponding chip number. GCC supports scheduling
10502 parameters for the EV4, EV5 and EV6 family of processors and will
10503 choose the default values for the instruction set from the processor
10504 you specify. If you do not specify a processor type, GCC will default
10505 to the processor on which the compiler was built.
10507 Supported values for @var{cpu_type} are
10513 Schedules as an EV4 and has no instruction set extensions.
10517 Schedules as an EV5 and has no instruction set extensions.
10521 Schedules as an EV5 and supports the BWX extension.
10526 Schedules as an EV5 and supports the BWX and MAX extensions.
10530 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
10534 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
10537 Native Linux/GNU toolchains also support the value @samp{native},
10538 which selects the best architecture option for the host processor.
10539 @option{-mcpu=native} has no effect if GCC does not recognize
10542 @item -mtune=@var{cpu_type}
10544 Set only the instruction scheduling parameters for machine type
10545 @var{cpu_type}. The instruction set is not changed.
10547 Native Linux/GNU toolchains also support the value @samp{native},
10548 which selects the best architecture option for the host processor.
10549 @option{-mtune=native} has no effect if GCC does not recognize
10552 @item -mmemory-latency=@var{time}
10553 @opindex mmemory-latency
10554 Sets the latency the scheduler should assume for typical memory
10555 references as seen by the application. This number is highly
10556 dependent on the memory access patterns used by the application
10557 and the size of the external cache on the machine.
10559 Valid options for @var{time} are
10563 A decimal number representing clock cycles.
10569 The compiler contains estimates of the number of clock cycles for
10570 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
10571 (also called Dcache, Scache, and Bcache), as well as to main memory.
10572 Note that L3 is only valid for EV5.
10577 @node DEC Alpha/VMS Options
10578 @subsection DEC Alpha/VMS Options
10580 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
10583 @item -mvms-return-codes
10584 @opindex mvms-return-codes
10585 Return VMS condition codes from main. The default is to return POSIX
10586 style condition (e.g.@: error) codes.
10588 @item -mdebug-main=@var{prefix}
10589 @opindex mdebug-main=@var{prefix}
10590 Flag the first routine whose name starts with @var{prefix} as the main
10591 routine for the debugger.
10595 @subsection FR30 Options
10596 @cindex FR30 Options
10598 These options are defined specifically for the FR30 port.
10602 @item -msmall-model
10603 @opindex msmall-model
10604 Use the small address space model. This can produce smaller code, but
10605 it does assume that all symbolic values and addresses will fit into a
10610 Assume that run-time support has been provided and so there is no need
10611 to include the simulator library (@file{libsim.a}) on the linker
10617 @subsection FRV Options
10618 @cindex FRV Options
10624 Only use the first 32 general purpose registers.
10629 Use all 64 general purpose registers.
10634 Use only the first 32 floating point registers.
10639 Use all 64 floating point registers
10642 @opindex mhard-float
10644 Use hardware instructions for floating point operations.
10647 @opindex msoft-float
10649 Use library routines for floating point operations.
10654 Dynamically allocate condition code registers.
10659 Do not try to dynamically allocate condition code registers, only
10660 use @code{icc0} and @code{fcc0}.
10665 Change ABI to use double word insns.
10670 Do not use double word instructions.
10675 Use floating point double instructions.
10678 @opindex mno-double
10680 Do not use floating point double instructions.
10685 Use media instructions.
10690 Do not use media instructions.
10695 Use multiply and add/subtract instructions.
10698 @opindex mno-muladd
10700 Do not use multiply and add/subtract instructions.
10705 Select the FDPIC ABI, that uses function descriptors to represent
10706 pointers to functions. Without any PIC/PIE-related options, it
10707 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
10708 assumes GOT entries and small data are within a 12-bit range from the
10709 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
10710 are computed with 32 bits.
10711 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10714 @opindex minline-plt
10716 Enable inlining of PLT entries in function calls to functions that are
10717 not known to bind locally. It has no effect without @option{-mfdpic}.
10718 It's enabled by default if optimizing for speed and compiling for
10719 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
10720 optimization option such as @option{-O3} or above is present in the
10726 Assume a large TLS segment when generating thread-local code.
10731 Do not assume a large TLS segment when generating thread-local code.
10736 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
10737 that is known to be in read-only sections. It's enabled by default,
10738 except for @option{-fpic} or @option{-fpie}: even though it may help
10739 make the global offset table smaller, it trades 1 instruction for 4.
10740 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
10741 one of which may be shared by multiple symbols, and it avoids the need
10742 for a GOT entry for the referenced symbol, so it's more likely to be a
10743 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
10745 @item -multilib-library-pic
10746 @opindex multilib-library-pic
10748 Link with the (library, not FD) pic libraries. It's implied by
10749 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
10750 @option{-fpic} without @option{-mfdpic}. You should never have to use
10754 @opindex mlinked-fp
10756 Follow the EABI requirement of always creating a frame pointer whenever
10757 a stack frame is allocated. This option is enabled by default and can
10758 be disabled with @option{-mno-linked-fp}.
10761 @opindex mlong-calls
10763 Use indirect addressing to call functions outside the current
10764 compilation unit. This allows the functions to be placed anywhere
10765 within the 32-bit address space.
10767 @item -malign-labels
10768 @opindex malign-labels
10770 Try to align labels to an 8-byte boundary by inserting nops into the
10771 previous packet. This option only has an effect when VLIW packing
10772 is enabled. It doesn't create new packets; it merely adds nops to
10775 @item -mlibrary-pic
10776 @opindex mlibrary-pic
10778 Generate position-independent EABI code.
10783 Use only the first four media accumulator registers.
10788 Use all eight media accumulator registers.
10793 Pack VLIW instructions.
10798 Do not pack VLIW instructions.
10801 @opindex mno-eflags
10803 Do not mark ABI switches in e_flags.
10806 @opindex mcond-move
10808 Enable the use of conditional-move instructions (default).
10810 This switch is mainly for debugging the compiler and will likely be removed
10811 in a future version.
10813 @item -mno-cond-move
10814 @opindex mno-cond-move
10816 Disable the use of conditional-move instructions.
10818 This switch is mainly for debugging the compiler and will likely be removed
10819 in a future version.
10824 Enable the use of conditional set instructions (default).
10826 This switch is mainly for debugging the compiler and will likely be removed
10827 in a future version.
10832 Disable the use of conditional set instructions.
10834 This switch is mainly for debugging the compiler and will likely be removed
10835 in a future version.
10838 @opindex mcond-exec
10840 Enable the use of conditional execution (default).
10842 This switch is mainly for debugging the compiler and will likely be removed
10843 in a future version.
10845 @item -mno-cond-exec
10846 @opindex mno-cond-exec
10848 Disable the use of conditional execution.
10850 This switch is mainly for debugging the compiler and will likely be removed
10851 in a future version.
10853 @item -mvliw-branch
10854 @opindex mvliw-branch
10856 Run a pass to pack branches into VLIW instructions (default).
10858 This switch is mainly for debugging the compiler and will likely be removed
10859 in a future version.
10861 @item -mno-vliw-branch
10862 @opindex mno-vliw-branch
10864 Do not run a pass to pack branches into VLIW instructions.
10866 This switch is mainly for debugging the compiler and will likely be removed
10867 in a future version.
10869 @item -mmulti-cond-exec
10870 @opindex mmulti-cond-exec
10872 Enable optimization of @code{&&} and @code{||} in conditional execution
10875 This switch is mainly for debugging the compiler and will likely be removed
10876 in a future version.
10878 @item -mno-multi-cond-exec
10879 @opindex mno-multi-cond-exec
10881 Disable optimization of @code{&&} and @code{||} in conditional execution.
10883 This switch is mainly for debugging the compiler and will likely be removed
10884 in a future version.
10886 @item -mnested-cond-exec
10887 @opindex mnested-cond-exec
10889 Enable nested conditional execution optimizations (default).
10891 This switch is mainly for debugging the compiler and will likely be removed
10892 in a future version.
10894 @item -mno-nested-cond-exec
10895 @opindex mno-nested-cond-exec
10897 Disable nested conditional execution optimizations.
10899 This switch is mainly for debugging the compiler and will likely be removed
10900 in a future version.
10902 @item -moptimize-membar
10903 @opindex moptimize-membar
10905 This switch removes redundant @code{membar} instructions from the
10906 compiler generated code. It is enabled by default.
10908 @item -mno-optimize-membar
10909 @opindex mno-optimize-membar
10911 This switch disables the automatic removal of redundant @code{membar}
10912 instructions from the generated code.
10914 @item -mtomcat-stats
10915 @opindex mtomcat-stats
10917 Cause gas to print out tomcat statistics.
10919 @item -mcpu=@var{cpu}
10922 Select the processor type for which to generate code. Possible values are
10923 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
10924 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
10928 @node GNU/Linux Options
10929 @subsection GNU/Linux Options
10931 These @samp{-m} options are defined for GNU/Linux targets:
10936 Use the GNU C library instead of uClibc. This is the default except
10937 on @samp{*-*-linux-*uclibc*} targets.
10941 Use uClibc instead of the GNU C library. This is the default on
10942 @samp{*-*-linux-*uclibc*} targets.
10945 @node H8/300 Options
10946 @subsection H8/300 Options
10948 These @samp{-m} options are defined for the H8/300 implementations:
10953 Shorten some address references at link time, when possible; uses the
10954 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
10955 ld, Using ld}, for a fuller description.
10959 Generate code for the H8/300H@.
10963 Generate code for the H8S@.
10967 Generate code for the H8S and H8/300H in the normal mode. This switch
10968 must be used either with @option{-mh} or @option{-ms}.
10972 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
10976 Make @code{int} data 32 bits by default.
10979 @opindex malign-300
10980 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
10981 The default for the H8/300H and H8S is to align longs and floats on 4
10983 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
10984 This option has no effect on the H8/300.
10988 @subsection HPPA Options
10989 @cindex HPPA Options
10991 These @samp{-m} options are defined for the HPPA family of computers:
10994 @item -march=@var{architecture-type}
10996 Generate code for the specified architecture. The choices for
10997 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
10998 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
10999 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
11000 architecture option for your machine. Code compiled for lower numbered
11001 architectures will run on higher numbered architectures, but not the
11004 @item -mpa-risc-1-0
11005 @itemx -mpa-risc-1-1
11006 @itemx -mpa-risc-2-0
11007 @opindex mpa-risc-1-0
11008 @opindex mpa-risc-1-1
11009 @opindex mpa-risc-2-0
11010 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
11013 @opindex mbig-switch
11014 Generate code suitable for big switch tables. Use this option only if
11015 the assembler/linker complain about out of range branches within a switch
11018 @item -mjump-in-delay
11019 @opindex mjump-in-delay
11020 Fill delay slots of function calls with unconditional jump instructions
11021 by modifying the return pointer for the function call to be the target
11022 of the conditional jump.
11024 @item -mdisable-fpregs
11025 @opindex mdisable-fpregs
11026 Prevent floating point registers from being used in any manner. This is
11027 necessary for compiling kernels which perform lazy context switching of
11028 floating point registers. If you use this option and attempt to perform
11029 floating point operations, the compiler will abort.
11031 @item -mdisable-indexing
11032 @opindex mdisable-indexing
11033 Prevent the compiler from using indexing address modes. This avoids some
11034 rather obscure problems when compiling MIG generated code under MACH@.
11036 @item -mno-space-regs
11037 @opindex mno-space-regs
11038 Generate code that assumes the target has no space registers. This allows
11039 GCC to generate faster indirect calls and use unscaled index address modes.
11041 Such code is suitable for level 0 PA systems and kernels.
11043 @item -mfast-indirect-calls
11044 @opindex mfast-indirect-calls
11045 Generate code that assumes calls never cross space boundaries. This
11046 allows GCC to emit code which performs faster indirect calls.
11048 This option will not work in the presence of shared libraries or nested
11051 @item -mfixed-range=@var{register-range}
11052 @opindex mfixed-range
11053 Generate code treating the given register range as fixed registers.
11054 A fixed register is one that the register allocator can not use. This is
11055 useful when compiling kernel code. A register range is specified as
11056 two registers separated by a dash. Multiple register ranges can be
11057 specified separated by a comma.
11059 @item -mlong-load-store
11060 @opindex mlong-load-store
11061 Generate 3-instruction load and store sequences as sometimes required by
11062 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
11065 @item -mportable-runtime
11066 @opindex mportable-runtime
11067 Use the portable calling conventions proposed by HP for ELF systems.
11071 Enable the use of assembler directives only GAS understands.
11073 @item -mschedule=@var{cpu-type}
11075 Schedule code according to the constraints for the machine type
11076 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
11077 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
11078 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
11079 proper scheduling option for your machine. The default scheduling is
11083 @opindex mlinker-opt
11084 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
11085 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
11086 linkers in which they give bogus error messages when linking some programs.
11089 @opindex msoft-float
11090 Generate output containing library calls for floating point.
11091 @strong{Warning:} the requisite libraries are not available for all HPPA
11092 targets. Normally the facilities of the machine's usual C compiler are
11093 used, but this cannot be done directly in cross-compilation. You must make
11094 your own arrangements to provide suitable library functions for
11097 @option{-msoft-float} changes the calling convention in the output file;
11098 therefore, it is only useful if you compile @emph{all} of a program with
11099 this option. In particular, you need to compile @file{libgcc.a}, the
11100 library that comes with GCC, with @option{-msoft-float} in order for
11105 Generate the predefine, @code{_SIO}, for server IO@. The default is
11106 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
11107 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
11108 options are available under HP-UX and HI-UX@.
11112 Use GNU ld specific options. This passes @option{-shared} to ld when
11113 building a shared library. It is the default when GCC is configured,
11114 explicitly or implicitly, with the GNU linker. This option does not
11115 have any affect on which ld is called, it only changes what parameters
11116 are passed to that ld. The ld that is called is determined by the
11117 @option{--with-ld} configure option, GCC's program search path, and
11118 finally by the user's @env{PATH}. The linker used by GCC can be printed
11119 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
11120 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11124 Use HP ld specific options. This passes @option{-b} to ld when building
11125 a shared library and passes @option{+Accept TypeMismatch} to ld on all
11126 links. It is the default when GCC is configured, explicitly or
11127 implicitly, with the HP linker. This option does not have any affect on
11128 which ld is called, it only changes what parameters are passed to that
11129 ld. The ld that is called is determined by the @option{--with-ld}
11130 configure option, GCC's program search path, and finally by the user's
11131 @env{PATH}. The linker used by GCC can be printed using @samp{which
11132 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
11133 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11136 @opindex mno-long-calls
11137 Generate code that uses long call sequences. This ensures that a call
11138 is always able to reach linker generated stubs. The default is to generate
11139 long calls only when the distance from the call site to the beginning
11140 of the function or translation unit, as the case may be, exceeds a
11141 predefined limit set by the branch type being used. The limits for
11142 normal calls are 7,600,000 and 240,000 bytes, respectively for the
11143 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
11146 Distances are measured from the beginning of functions when using the
11147 @option{-ffunction-sections} option, or when using the @option{-mgas}
11148 and @option{-mno-portable-runtime} options together under HP-UX with
11151 It is normally not desirable to use this option as it will degrade
11152 performance. However, it may be useful in large applications,
11153 particularly when partial linking is used to build the application.
11155 The types of long calls used depends on the capabilities of the
11156 assembler and linker, and the type of code being generated. The
11157 impact on systems that support long absolute calls, and long pic
11158 symbol-difference or pc-relative calls should be relatively small.
11159 However, an indirect call is used on 32-bit ELF systems in pic code
11160 and it is quite long.
11162 @item -munix=@var{unix-std}
11164 Generate compiler predefines and select a startfile for the specified
11165 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
11166 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
11167 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
11168 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
11169 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
11172 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
11173 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
11174 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
11175 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
11176 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
11177 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
11179 It is @emph{important} to note that this option changes the interfaces
11180 for various library routines. It also affects the operational behavior
11181 of the C library. Thus, @emph{extreme} care is needed in using this
11184 Library code that is intended to operate with more than one UNIX
11185 standard must test, set and restore the variable @var{__xpg4_extended_mask}
11186 as appropriate. Most GNU software doesn't provide this capability.
11190 Suppress the generation of link options to search libdld.sl when the
11191 @option{-static} option is specified on HP-UX 10 and later.
11195 The HP-UX implementation of setlocale in libc has a dependency on
11196 libdld.sl. There isn't an archive version of libdld.sl. Thus,
11197 when the @option{-static} option is specified, special link options
11198 are needed to resolve this dependency.
11200 On HP-UX 10 and later, the GCC driver adds the necessary options to
11201 link with libdld.sl when the @option{-static} option is specified.
11202 This causes the resulting binary to be dynamic. On the 64-bit port,
11203 the linkers generate dynamic binaries by default in any case. The
11204 @option{-nolibdld} option can be used to prevent the GCC driver from
11205 adding these link options.
11209 Add support for multithreading with the @dfn{dce thread} library
11210 under HP-UX@. This option sets flags for both the preprocessor and
11214 @node i386 and x86-64 Options
11215 @subsection Intel 386 and AMD x86-64 Options
11216 @cindex i386 Options
11217 @cindex x86-64 Options
11218 @cindex Intel 386 Options
11219 @cindex AMD x86-64 Options
11221 These @samp{-m} options are defined for the i386 and x86-64 family of
11225 @item -mtune=@var{cpu-type}
11227 Tune to @var{cpu-type} everything applicable about the generated code, except
11228 for the ABI and the set of available instructions. The choices for
11229 @var{cpu-type} are:
11232 Produce code optimized for the most common IA32/AMD64/EM64T processors.
11233 If you know the CPU on which your code will run, then you should use
11234 the corresponding @option{-mtune} option instead of
11235 @option{-mtune=generic}. But, if you do not know exactly what CPU users
11236 of your application will have, then you should use this option.
11238 As new processors are deployed in the marketplace, the behavior of this
11239 option will change. Therefore, if you upgrade to a newer version of
11240 GCC, the code generated option will change to reflect the processors
11241 that were most common when that version of GCC was released.
11243 There is no @option{-march=generic} option because @option{-march}
11244 indicates the instruction set the compiler can use, and there is no
11245 generic instruction set applicable to all processors. In contrast,
11246 @option{-mtune} indicates the processor (or, in this case, collection of
11247 processors) for which the code is optimized.
11249 This selects the CPU to tune for at compilation time by determining
11250 the processor type of the compiling machine. Using @option{-mtune=native}
11251 will produce code optimized for the local machine under the constraints
11252 of the selected instruction set. Using @option{-march=native} will
11253 enable all instruction subsets supported by the local machine (hence
11254 the result might not run on different machines).
11256 Original Intel's i386 CPU@.
11258 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
11259 @item i586, pentium
11260 Intel Pentium CPU with no MMX support.
11262 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
11264 Intel PentiumPro CPU@.
11266 Same as @code{generic}, but when used as @code{march} option, PentiumPro
11267 instruction set will be used, so the code will run on all i686 family chips.
11269 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
11270 @item pentium3, pentium3m
11271 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
11274 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
11275 support. Used by Centrino notebooks.
11276 @item pentium4, pentium4m
11277 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
11279 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
11282 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
11283 SSE2 and SSE3 instruction set support.
11285 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
11286 instruction set support.
11288 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
11289 instruction set support.
11291 AMD K6 CPU with MMX instruction set support.
11293 Improved versions of AMD K6 CPU with MMX and 3dNOW!@: instruction set support.
11294 @item athlon, athlon-tbird
11295 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and SSE prefetch instructions
11297 @item athlon-4, athlon-xp, athlon-mp
11298 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and full SSE
11299 instruction set support.
11300 @item k8, opteron, athlon64, athlon-fx
11301 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
11302 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW!@: and 64-bit instruction set extensions.)
11303 @item k8-sse3, opteron-sse3, athlon64-sse3
11304 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
11305 @item amdfam10, barcelona
11306 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
11307 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3dNOW!, enhanced 3dNOW!, ABM and 64-bit
11308 instruction set extensions.)
11310 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
11313 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!@:
11314 instruction set support.
11316 Via C3 CPU with MMX and 3dNOW!@: instruction set support. (No scheduling is
11317 implemented for this chip.)
11319 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
11320 implemented for this chip.)
11322 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
11325 While picking a specific @var{cpu-type} will schedule things appropriately
11326 for that particular chip, the compiler will not generate any code that
11327 does not run on the i386 without the @option{-march=@var{cpu-type}} option
11330 @item -march=@var{cpu-type}
11332 Generate instructions for the machine type @var{cpu-type}. The choices
11333 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
11334 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
11336 @item -mcpu=@var{cpu-type}
11338 A deprecated synonym for @option{-mtune}.
11340 @item -mfpmath=@var{unit}
11342 Generate floating point arithmetics for selected unit @var{unit}. The choices
11343 for @var{unit} are:
11347 Use the standard 387 floating point coprocessor present majority of chips and
11348 emulated otherwise. Code compiled with this option will run almost everywhere.
11349 The temporary results are computed in 80bit precision instead of precision
11350 specified by the type resulting in slightly different results compared to most
11351 of other chips. See @option{-ffloat-store} for more detailed description.
11353 This is the default choice for i386 compiler.
11356 Use scalar floating point instructions present in the SSE instruction set.
11357 This instruction set is supported by Pentium3 and newer chips, in the AMD line
11358 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
11359 instruction set supports only single precision arithmetics, thus the double and
11360 extended precision arithmetics is still done using 387. Later version, present
11361 only in Pentium4 and the future AMD x86-64 chips supports double precision
11364 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
11365 or @option{-msse2} switches to enable SSE extensions and make this option
11366 effective. For the x86-64 compiler, these extensions are enabled by default.
11368 The resulting code should be considerably faster in the majority of cases and avoid
11369 the numerical instability problems of 387 code, but may break some existing
11370 code that expects temporaries to be 80bit.
11372 This is the default choice for the x86-64 compiler.
11377 Attempt to utilize both instruction sets at once. This effectively double the
11378 amount of available registers and on chips with separate execution units for
11379 387 and SSE the execution resources too. Use this option with care, as it is
11380 still experimental, because the GCC register allocator does not model separate
11381 functional units well resulting in instable performance.
11384 @item -masm=@var{dialect}
11385 @opindex masm=@var{dialect}
11386 Output asm instructions using selected @var{dialect}. Supported
11387 choices are @samp{intel} or @samp{att} (the default one). Darwin does
11388 not support @samp{intel}.
11391 @itemx -mno-ieee-fp
11393 @opindex mno-ieee-fp
11394 Control whether or not the compiler uses IEEE floating point
11395 comparisons. These handle correctly the case where the result of a
11396 comparison is unordered.
11399 @opindex msoft-float
11400 Generate output containing library calls for floating point.
11401 @strong{Warning:} the requisite libraries are not part of GCC@.
11402 Normally the facilities of the machine's usual C compiler are used, but
11403 this can't be done directly in cross-compilation. You must make your
11404 own arrangements to provide suitable library functions for
11407 On machines where a function returns floating point results in the 80387
11408 register stack, some floating point opcodes may be emitted even if
11409 @option{-msoft-float} is used.
11411 @item -mno-fp-ret-in-387
11412 @opindex mno-fp-ret-in-387
11413 Do not use the FPU registers for return values of functions.
11415 The usual calling convention has functions return values of types
11416 @code{float} and @code{double} in an FPU register, even if there
11417 is no FPU@. The idea is that the operating system should emulate
11420 The option @option{-mno-fp-ret-in-387} causes such values to be returned
11421 in ordinary CPU registers instead.
11423 @item -mno-fancy-math-387
11424 @opindex mno-fancy-math-387
11425 Some 387 emulators do not support the @code{sin}, @code{cos} and
11426 @code{sqrt} instructions for the 387. Specify this option to avoid
11427 generating those instructions. This option is the default on FreeBSD,
11428 OpenBSD and NetBSD@. This option is overridden when @option{-march}
11429 indicates that the target cpu will always have an FPU and so the
11430 instruction will not need emulation. As of revision 2.6.1, these
11431 instructions are not generated unless you also use the
11432 @option{-funsafe-math-optimizations} switch.
11434 @item -malign-double
11435 @itemx -mno-align-double
11436 @opindex malign-double
11437 @opindex mno-align-double
11438 Control whether GCC aligns @code{double}, @code{long double}, and
11439 @code{long long} variables on a two word boundary or a one word
11440 boundary. Aligning @code{double} variables on a two word boundary will
11441 produce code that runs somewhat faster on a @samp{Pentium} at the
11442 expense of more memory.
11444 On x86-64, @option{-malign-double} is enabled by default.
11446 @strong{Warning:} if you use the @option{-malign-double} switch,
11447 structures containing the above types will be aligned differently than
11448 the published application binary interface specifications for the 386
11449 and will not be binary compatible with structures in code compiled
11450 without that switch.
11452 @item -m96bit-long-double
11453 @itemx -m128bit-long-double
11454 @opindex m96bit-long-double
11455 @opindex m128bit-long-double
11456 These switches control the size of @code{long double} type. The i386
11457 application binary interface specifies the size to be 96 bits,
11458 so @option{-m96bit-long-double} is the default in 32 bit mode.
11460 Modern architectures (Pentium and newer) would prefer @code{long double}
11461 to be aligned to an 8 or 16 byte boundary. In arrays or structures
11462 conforming to the ABI, this would not be possible. So specifying a
11463 @option{-m128bit-long-double} will align @code{long double}
11464 to a 16 byte boundary by padding the @code{long double} with an additional
11467 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
11468 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
11470 Notice that neither of these options enable any extra precision over the x87
11471 standard of 80 bits for a @code{long double}.
11473 @strong{Warning:} if you override the default value for your target ABI, the
11474 structures and arrays containing @code{long double} variables will change
11475 their size as well as function calling convention for function taking
11476 @code{long double} will be modified. Hence they will not be binary
11477 compatible with arrays or structures in code compiled without that switch.
11479 @item -mlarge-data-threshold=@var{number}
11480 @opindex mlarge-data-threshold=@var{number}
11481 When @option{-mcmodel=medium} is specified, the data greater than
11482 @var{threshold} are placed in large data section. This value must be the
11483 same across all object linked into the binary and defaults to 65535.
11487 Use a different function-calling convention, in which functions that
11488 take a fixed number of arguments return with the @code{ret} @var{num}
11489 instruction, which pops their arguments while returning. This saves one
11490 instruction in the caller since there is no need to pop the arguments
11493 You can specify that an individual function is called with this calling
11494 sequence with the function attribute @samp{stdcall}. You can also
11495 override the @option{-mrtd} option by using the function attribute
11496 @samp{cdecl}. @xref{Function Attributes}.
11498 @strong{Warning:} this calling convention is incompatible with the one
11499 normally used on Unix, so you cannot use it if you need to call
11500 libraries compiled with the Unix compiler.
11502 Also, you must provide function prototypes for all functions that
11503 take variable numbers of arguments (including @code{printf});
11504 otherwise incorrect code will be generated for calls to those
11507 In addition, seriously incorrect code will result if you call a
11508 function with too many arguments. (Normally, extra arguments are
11509 harmlessly ignored.)
11511 @item -mregparm=@var{num}
11513 Control how many registers are used to pass integer arguments. By
11514 default, no registers are used to pass arguments, and at most 3
11515 registers can be used. You can control this behavior for a specific
11516 function by using the function attribute @samp{regparm}.
11517 @xref{Function Attributes}.
11519 @strong{Warning:} if you use this switch, and
11520 @var{num} is nonzero, then you must build all modules with the same
11521 value, including any libraries. This includes the system libraries and
11525 @opindex msseregparm
11526 Use SSE register passing conventions for float and double arguments
11527 and return values. You can control this behavior for a specific
11528 function by using the function attribute @samp{sseregparm}.
11529 @xref{Function Attributes}.
11531 @strong{Warning:} if you use this switch then you must build all
11532 modules with the same value, including any libraries. This includes
11533 the system libraries and startup modules.
11542 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
11543 is specified, the significands of results of floating-point operations are
11544 rounded to 24 bits (single precision); @option{-mpc64} rounds the
11545 significands of results of floating-point operations to 53 bits (double
11546 precision) and @option{-mpc80} rounds the significands of results of
11547 floating-point operations to 64 bits (extended double precision), which is
11548 the default. When this option is used, floating-point operations in higher
11549 precisions are not available to the programmer without setting the FPU
11550 control word explicitly.
11552 Setting the rounding of floating-point operations to less than the default
11553 80 bits can speed some programs by 2% or more. Note that some mathematical
11554 libraries assume that extended precision (80 bit) floating-point operations
11555 are enabled by default; routines in such libraries could suffer significant
11556 loss of accuracy, typically through so-called "catastrophic cancellation",
11557 when this option is used to set the precision to less than extended precision.
11559 @item -mstackrealign
11560 @opindex mstackrealign
11561 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
11562 option will generate an alternate prologue and epilogue that realigns the
11563 runtime stack if necessary. This supports mixing legacy codes that keep
11564 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
11565 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
11566 applicable to individual functions.
11568 @item -mpreferred-stack-boundary=@var{num}
11569 @opindex mpreferred-stack-boundary
11570 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
11571 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
11572 the default is 4 (16 bytes or 128 bits).
11574 @item -mincoming-stack-boundary=@var{num}
11575 @opindex mincoming-stack-boundary
11576 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
11577 boundary. If @option{-mincoming-stack-boundary} is not specified,
11578 the one specified by @option{-mpreferred-stack-boundary} will be used.
11580 On Pentium and PentiumPro, @code{double} and @code{long double} values
11581 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
11582 suffer significant run time performance penalties. On Pentium III, the
11583 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
11584 properly if it is not 16 byte aligned.
11586 To ensure proper alignment of this values on the stack, the stack boundary
11587 must be as aligned as that required by any value stored on the stack.
11588 Further, every function must be generated such that it keeps the stack
11589 aligned. Thus calling a function compiled with a higher preferred
11590 stack boundary from a function compiled with a lower preferred stack
11591 boundary will most likely misalign the stack. It is recommended that
11592 libraries that use callbacks always use the default setting.
11594 This extra alignment does consume extra stack space, and generally
11595 increases code size. Code that is sensitive to stack space usage, such
11596 as embedded systems and operating system kernels, may want to reduce the
11597 preferred alignment to @option{-mpreferred-stack-boundary=2}.
11637 These switches enable or disable the use of instructions in the MMX,
11638 SSE, SSE2, SSE3, SSSE3, SSE4.1, AVX, AES, PCLMUL, SSE4A, SSE5, ABM or
11639 3DNow!@: extended instruction sets.
11640 These extensions are also available as built-in functions: see
11641 @ref{X86 Built-in Functions}, for details of the functions enabled and
11642 disabled by these switches.
11644 To have SSE/SSE2 instructions generated automatically from floating-point
11645 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
11647 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
11648 generates new AVX instructions or AVX equivalence for all SSEx instructions
11651 These options will enable GCC to use these extended instructions in
11652 generated code, even without @option{-mfpmath=sse}. Applications which
11653 perform runtime CPU detection must compile separate files for each
11654 supported architecture, using the appropriate flags. In particular,
11655 the file containing the CPU detection code should be compiled without
11660 This option instructs GCC to emit a @code{cld} instruction in the prologue
11661 of functions that use string instructions. String instructions depend on
11662 the DF flag to select between autoincrement or autodecrement mode. While the
11663 ABI specifies the DF flag to be cleared on function entry, some operating
11664 systems violate this specification by not clearing the DF flag in their
11665 exception dispatchers. The exception handler can be invoked with the DF flag
11666 set which leads to wrong direction mode, when string instructions are used.
11667 This option can be enabled by default on 32-bit x86 targets by configuring
11668 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
11669 instructions can be suppressed with the @option{-mno-cld} compiler option
11674 This option will enable GCC to use CMPXCHG16B instruction in generated code.
11675 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
11676 data types. This is useful for high resolution counters that could be updated
11677 by multiple processors (or cores). This instruction is generated as part of
11678 atomic built-in functions: see @ref{Atomic Builtins} for details.
11682 This option will enable GCC to use SAHF instruction in generated 64-bit code.
11683 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
11684 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
11685 SAHF are load and store instructions, respectively, for certain status flags.
11686 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
11687 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
11691 This option will enable GCC to use movbe instruction to implement
11692 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
11696 This option will enable built-in functions, @code{__builtin_ia32_crc32qi},
11697 @code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and
11698 @code{__builtin_ia32_crc32di} to generate the crc32 machine instruction.
11702 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
11703 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
11704 to increase precision instead of DIVSS and SQRTSS (and their vectorized
11705 variants) for single precision floating point arguments. These instructions
11706 are generated only when @option{-funsafe-math-optimizations} is enabled
11707 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
11708 Note that while the throughput of the sequence is higher than the throughput
11709 of the non-reciprocal instruction, the precision of the sequence can be
11710 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
11712 @item -mveclibabi=@var{type}
11713 @opindex mveclibabi
11714 Specifies the ABI type to use for vectorizing intrinsics using an
11715 external library. Supported types are @code{svml} for the Intel short
11716 vector math library and @code{acml} for the AMD math core library style
11717 of interfacing. GCC will currently emit calls to @code{vmldExp2},
11718 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
11719 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
11720 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
11721 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
11722 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
11723 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
11724 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
11725 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
11726 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
11727 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
11728 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
11729 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
11730 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
11731 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
11732 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
11733 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
11734 compatible library will have to be specified at link time.
11736 @item -mabi=@var{name}
11738 Generate code for the specified calling convention. Permissible values
11739 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
11740 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
11741 ABI when targeting Windows. On all other systems, the default is the
11742 SYSV ABI. You can control this behavior for a specific function by
11743 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
11744 @xref{Function Attributes}.
11747 @itemx -mno-push-args
11748 @opindex mpush-args
11749 @opindex mno-push-args
11750 Use PUSH operations to store outgoing parameters. This method is shorter
11751 and usually equally fast as method using SUB/MOV operations and is enabled
11752 by default. In some cases disabling it may improve performance because of
11753 improved scheduling and reduced dependencies.
11755 @item -maccumulate-outgoing-args
11756 @opindex maccumulate-outgoing-args
11757 If enabled, the maximum amount of space required for outgoing arguments will be
11758 computed in the function prologue. This is faster on most modern CPUs
11759 because of reduced dependencies, improved scheduling and reduced stack usage
11760 when preferred stack boundary is not equal to 2. The drawback is a notable
11761 increase in code size. This switch implies @option{-mno-push-args}.
11765 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
11766 on thread-safe exception handling must compile and link all code with the
11767 @option{-mthreads} option. When compiling, @option{-mthreads} defines
11768 @option{-D_MT}; when linking, it links in a special thread helper library
11769 @option{-lmingwthrd} which cleans up per thread exception handling data.
11771 @item -mno-align-stringops
11772 @opindex mno-align-stringops
11773 Do not align destination of inlined string operations. This switch reduces
11774 code size and improves performance in case the destination is already aligned,
11775 but GCC doesn't know about it.
11777 @item -minline-all-stringops
11778 @opindex minline-all-stringops
11779 By default GCC inlines string operations only when destination is known to be
11780 aligned at least to 4 byte boundary. This enables more inlining, increase code
11781 size, but may improve performance of code that depends on fast memcpy, strlen
11782 and memset for short lengths.
11784 @item -minline-stringops-dynamically
11785 @opindex minline-stringops-dynamically
11786 For string operation of unknown size, inline runtime checks so for small
11787 blocks inline code is used, while for large blocks library call is used.
11789 @item -mstringop-strategy=@var{alg}
11790 @opindex mstringop-strategy=@var{alg}
11791 Overwrite internal decision heuristic about particular algorithm to inline
11792 string operation with. The allowed values are @code{rep_byte},
11793 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
11794 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
11795 expanding inline loop, @code{libcall} for always expanding library call.
11797 @item -momit-leaf-frame-pointer
11798 @opindex momit-leaf-frame-pointer
11799 Don't keep the frame pointer in a register for leaf functions. This
11800 avoids the instructions to save, set up and restore frame pointers and
11801 makes an extra register available in leaf functions. The option
11802 @option{-fomit-frame-pointer} removes the frame pointer for all functions
11803 which might make debugging harder.
11805 @item -mtls-direct-seg-refs
11806 @itemx -mno-tls-direct-seg-refs
11807 @opindex mtls-direct-seg-refs
11808 Controls whether TLS variables may be accessed with offsets from the
11809 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
11810 or whether the thread base pointer must be added. Whether or not this
11811 is legal depends on the operating system, and whether it maps the
11812 segment to cover the entire TLS area.
11814 For systems that use GNU libc, the default is on.
11817 @itemx -mno-fused-madd
11818 @opindex mfused-madd
11819 Enable automatic generation of fused floating point multiply-add instructions
11820 if the ISA supports such instructions. The -mfused-madd option is on by
11821 default. The fused multiply-add instructions have a different
11822 rounding behavior compared to executing a multiply followed by an add.
11825 @itemx -mno-sse2avx
11827 Specify that the assembler should encode SSE instructions with VEX
11828 prefix. The option @option{-mavx} turns this on by default.
11831 These @samp{-m} switches are supported in addition to the above
11832 on AMD x86-64 processors in 64-bit environments.
11839 Generate code for a 32-bit or 64-bit environment.
11840 The 32-bit environment sets int, long and pointer to 32 bits and
11841 generates code that runs on any i386 system.
11842 The 64-bit environment sets int to 32 bits and long and pointer
11843 to 64 bits and generates code for AMD's x86-64 architecture. For
11844 darwin only the -m64 option turns off the @option{-fno-pic} and
11845 @option{-mdynamic-no-pic} options.
11847 @item -mno-red-zone
11848 @opindex mno-red-zone
11849 Do not use a so called red zone for x86-64 code. The red zone is mandated
11850 by the x86-64 ABI, it is a 128-byte area beyond the location of the
11851 stack pointer that will not be modified by signal or interrupt handlers
11852 and therefore can be used for temporary data without adjusting the stack
11853 pointer. The flag @option{-mno-red-zone} disables this red zone.
11855 @item -mcmodel=small
11856 @opindex mcmodel=small
11857 Generate code for the small code model: the program and its symbols must
11858 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
11859 Programs can be statically or dynamically linked. This is the default
11862 @item -mcmodel=kernel
11863 @opindex mcmodel=kernel
11864 Generate code for the kernel code model. The kernel runs in the
11865 negative 2 GB of the address space.
11866 This model has to be used for Linux kernel code.
11868 @item -mcmodel=medium
11869 @opindex mcmodel=medium
11870 Generate code for the medium model: The program is linked in the lower 2
11871 GB of the address space. Small symbols are also placed there. Symbols
11872 with sizes larger than @option{-mlarge-data-threshold} are put into
11873 large data or bss sections and can be located above 2GB. Programs can
11874 be statically or dynamically linked.
11876 @item -mcmodel=large
11877 @opindex mcmodel=large
11878 Generate code for the large model: This model makes no assumptions
11879 about addresses and sizes of sections.
11882 @node IA-64 Options
11883 @subsection IA-64 Options
11884 @cindex IA-64 Options
11886 These are the @samp{-m} options defined for the Intel IA-64 architecture.
11890 @opindex mbig-endian
11891 Generate code for a big endian target. This is the default for HP-UX@.
11893 @item -mlittle-endian
11894 @opindex mlittle-endian
11895 Generate code for a little endian target. This is the default for AIX5
11901 @opindex mno-gnu-as
11902 Generate (or don't) code for the GNU assembler. This is the default.
11903 @c Also, this is the default if the configure option @option{--with-gnu-as}
11909 @opindex mno-gnu-ld
11910 Generate (or don't) code for the GNU linker. This is the default.
11911 @c Also, this is the default if the configure option @option{--with-gnu-ld}
11916 Generate code that does not use a global pointer register. The result
11917 is not position independent code, and violates the IA-64 ABI@.
11919 @item -mvolatile-asm-stop
11920 @itemx -mno-volatile-asm-stop
11921 @opindex mvolatile-asm-stop
11922 @opindex mno-volatile-asm-stop
11923 Generate (or don't) a stop bit immediately before and after volatile asm
11926 @item -mregister-names
11927 @itemx -mno-register-names
11928 @opindex mregister-names
11929 @opindex mno-register-names
11930 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
11931 the stacked registers. This may make assembler output more readable.
11937 Disable (or enable) optimizations that use the small data section. This may
11938 be useful for working around optimizer bugs.
11940 @item -mconstant-gp
11941 @opindex mconstant-gp
11942 Generate code that uses a single constant global pointer value. This is
11943 useful when compiling kernel code.
11947 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
11948 This is useful when compiling firmware code.
11950 @item -minline-float-divide-min-latency
11951 @opindex minline-float-divide-min-latency
11952 Generate code for inline divides of floating point values
11953 using the minimum latency algorithm.
11955 @item -minline-float-divide-max-throughput
11956 @opindex minline-float-divide-max-throughput
11957 Generate code for inline divides of floating point values
11958 using the maximum throughput algorithm.
11960 @item -mno-inline-float-divide
11961 @opindex mno-inline-float-divide
11962 Do not generate inline code for divides of floating point values.
11964 @item -minline-int-divide-min-latency
11965 @opindex minline-int-divide-min-latency
11966 Generate code for inline divides of integer values
11967 using the minimum latency algorithm.
11969 @item -minline-int-divide-max-throughput
11970 @opindex minline-int-divide-max-throughput
11971 Generate code for inline divides of integer values
11972 using the maximum throughput algorithm.
11974 @item -mno-inline-int-divide
11975 @opindex mno-inline-int-divide
11976 Do not generate inline code for divides of integer values.
11978 @item -minline-sqrt-min-latency
11979 @opindex minline-sqrt-min-latency
11980 Generate code for inline square roots
11981 using the minimum latency algorithm.
11983 @item -minline-sqrt-max-throughput
11984 @opindex minline-sqrt-max-throughput
11985 Generate code for inline square roots
11986 using the maximum throughput algorithm.
11988 @item -mno-inline-sqrt
11989 @opindex mno-inline-sqrt
11990 Do not generate inline code for sqrt.
11993 @itemx -mno-fused-madd
11994 @opindex mfused-madd
11995 @opindex mno-fused-madd
11996 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
11997 instructions. The default is to use these instructions.
11999 @item -mno-dwarf2-asm
12000 @itemx -mdwarf2-asm
12001 @opindex mno-dwarf2-asm
12002 @opindex mdwarf2-asm
12003 Don't (or do) generate assembler code for the DWARF2 line number debugging
12004 info. This may be useful when not using the GNU assembler.
12006 @item -mearly-stop-bits
12007 @itemx -mno-early-stop-bits
12008 @opindex mearly-stop-bits
12009 @opindex mno-early-stop-bits
12010 Allow stop bits to be placed earlier than immediately preceding the
12011 instruction that triggered the stop bit. This can improve instruction
12012 scheduling, but does not always do so.
12014 @item -mfixed-range=@var{register-range}
12015 @opindex mfixed-range
12016 Generate code treating the given register range as fixed registers.
12017 A fixed register is one that the register allocator can not use. This is
12018 useful when compiling kernel code. A register range is specified as
12019 two registers separated by a dash. Multiple register ranges can be
12020 specified separated by a comma.
12022 @item -mtls-size=@var{tls-size}
12024 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
12027 @item -mtune=@var{cpu-type}
12029 Tune the instruction scheduling for a particular CPU, Valid values are
12030 itanium, itanium1, merced, itanium2, and mckinley.
12036 Generate code for a 32-bit or 64-bit environment.
12037 The 32-bit environment sets int, long and pointer to 32 bits.
12038 The 64-bit environment sets int to 32 bits and long and pointer
12039 to 64 bits. These are HP-UX specific flags.
12041 @item -mno-sched-br-data-spec
12042 @itemx -msched-br-data-spec
12043 @opindex mno-sched-br-data-spec
12044 @opindex msched-br-data-spec
12045 (Dis/En)able data speculative scheduling before reload.
12046 This will result in generation of the ld.a instructions and
12047 the corresponding check instructions (ld.c / chk.a).
12048 The default is 'disable'.
12050 @item -msched-ar-data-spec
12051 @itemx -mno-sched-ar-data-spec
12052 @opindex msched-ar-data-spec
12053 @opindex mno-sched-ar-data-spec
12054 (En/Dis)able data speculative scheduling after reload.
12055 This will result in generation of the ld.a instructions and
12056 the corresponding check instructions (ld.c / chk.a).
12057 The default is 'enable'.
12059 @item -mno-sched-control-spec
12060 @itemx -msched-control-spec
12061 @opindex mno-sched-control-spec
12062 @opindex msched-control-spec
12063 (Dis/En)able control speculative scheduling. This feature is
12064 available only during region scheduling (i.e.@: before reload).
12065 This will result in generation of the ld.s instructions and
12066 the corresponding check instructions chk.s .
12067 The default is 'disable'.
12069 @item -msched-br-in-data-spec
12070 @itemx -mno-sched-br-in-data-spec
12071 @opindex msched-br-in-data-spec
12072 @opindex mno-sched-br-in-data-spec
12073 (En/Dis)able speculative scheduling of the instructions that
12074 are dependent on the data speculative loads before reload.
12075 This is effective only with @option{-msched-br-data-spec} enabled.
12076 The default is 'enable'.
12078 @item -msched-ar-in-data-spec
12079 @itemx -mno-sched-ar-in-data-spec
12080 @opindex msched-ar-in-data-spec
12081 @opindex mno-sched-ar-in-data-spec
12082 (En/Dis)able speculative scheduling of the instructions that
12083 are dependent on the data speculative loads after reload.
12084 This is effective only with @option{-msched-ar-data-spec} enabled.
12085 The default is 'enable'.
12087 @item -msched-in-control-spec
12088 @itemx -mno-sched-in-control-spec
12089 @opindex msched-in-control-spec
12090 @opindex mno-sched-in-control-spec
12091 (En/Dis)able speculative scheduling of the instructions that
12092 are dependent on the control speculative loads.
12093 This is effective only with @option{-msched-control-spec} enabled.
12094 The default is 'enable'.
12096 @item -mno-sched-prefer-non-data-spec-insns
12097 @itemx -msched-prefer-non-data-spec-insns
12098 @opindex mno-sched-prefer-non-data-spec-insns
12099 @opindex msched-prefer-non-data-spec-insns
12100 If enabled, data speculative instructions will be chosen for schedule
12101 only if there are no other choices at the moment. This will make
12102 the use of the data speculation much more conservative.
12103 The default is 'disable'.
12105 @item -mno-sched-prefer-non-control-spec-insns
12106 @itemx -msched-prefer-non-control-spec-insns
12107 @opindex mno-sched-prefer-non-control-spec-insns
12108 @opindex msched-prefer-non-control-spec-insns
12109 If enabled, control speculative instructions will be chosen for schedule
12110 only if there are no other choices at the moment. This will make
12111 the use of the control speculation much more conservative.
12112 The default is 'disable'.
12114 @item -mno-sched-count-spec-in-critical-path
12115 @itemx -msched-count-spec-in-critical-path
12116 @opindex mno-sched-count-spec-in-critical-path
12117 @opindex msched-count-spec-in-critical-path
12118 If enabled, speculative dependencies will be considered during
12119 computation of the instructions priorities. This will make the use of the
12120 speculation a bit more conservative.
12121 The default is 'disable'.
12123 @item -msched-spec-ldc
12124 @opindex msched-spec-ldc
12125 Use a simple data speculation check. This option is on by default.
12127 @item -msched-control-spec-ldc
12128 @opindex msched-spec-ldc
12129 Use a simple check for control speculation. This option is on by default.
12131 @item -msched-stop-bits-after-every-cycle
12132 @opindex msched-stop-bits-after-every-cycle
12133 Place a stop bit after every cycle when scheduling. This option is on
12136 @item -msched-fp-mem-deps-zero-cost
12137 @opindex msched-fp-mem-deps-zero-cost
12138 Assume that floating-point stores and loads are not likely to cause a conflict
12139 when placed into the same instruction group. This option is disabled by
12142 @item -msel-sched-dont-check-control-spec
12143 @opindex msel-sched-dont-check-control-spec
12144 Generate checks for control speculation in selective scheduling.
12145 This flag is disabled by default.
12147 @item -msched-max-memory-insns=@var{max-insns}
12148 @opindex msched-max-memory-insns
12149 Limit on the number of memory insns per instruction group, giving lower
12150 priority to subsequent memory insns attempting to schedule in the same
12151 instruction group. Frequently useful to prevent cache bank conflicts.
12152 The default value is 1.
12154 @item -msched-max-memory-insns-hard-limit
12155 @opindex msched-max-memory-insns-hard-limit
12156 Disallow more than `msched-max-memory-insns' in instruction group.
12157 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
12158 when limit is reached but may still schedule memory operations.
12163 @subsection M32C Options
12164 @cindex M32C options
12167 @item -mcpu=@var{name}
12169 Select the CPU for which code is generated. @var{name} may be one of
12170 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
12171 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
12172 the M32C/80 series.
12176 Specifies that the program will be run on the simulator. This causes
12177 an alternate runtime library to be linked in which supports, for
12178 example, file I/O@. You must not use this option when generating
12179 programs that will run on real hardware; you must provide your own
12180 runtime library for whatever I/O functions are needed.
12182 @item -memregs=@var{number}
12184 Specifies the number of memory-based pseudo-registers GCC will use
12185 during code generation. These pseudo-registers will be used like real
12186 registers, so there is a tradeoff between GCC's ability to fit the
12187 code into available registers, and the performance penalty of using
12188 memory instead of registers. Note that all modules in a program must
12189 be compiled with the same value for this option. Because of that, you
12190 must not use this option with the default runtime libraries gcc
12195 @node M32R/D Options
12196 @subsection M32R/D Options
12197 @cindex M32R/D options
12199 These @option{-m} options are defined for Renesas M32R/D architectures:
12204 Generate code for the M32R/2@.
12208 Generate code for the M32R/X@.
12212 Generate code for the M32R@. This is the default.
12214 @item -mmodel=small
12215 @opindex mmodel=small
12216 Assume all objects live in the lower 16MB of memory (so that their addresses
12217 can be loaded with the @code{ld24} instruction), and assume all subroutines
12218 are reachable with the @code{bl} instruction.
12219 This is the default.
12221 The addressability of a particular object can be set with the
12222 @code{model} attribute.
12224 @item -mmodel=medium
12225 @opindex mmodel=medium
12226 Assume objects may be anywhere in the 32-bit address space (the compiler
12227 will generate @code{seth/add3} instructions to load their addresses), and
12228 assume all subroutines are reachable with the @code{bl} instruction.
12230 @item -mmodel=large
12231 @opindex mmodel=large
12232 Assume objects may be anywhere in the 32-bit address space (the compiler
12233 will generate @code{seth/add3} instructions to load their addresses), and
12234 assume subroutines may not be reachable with the @code{bl} instruction
12235 (the compiler will generate the much slower @code{seth/add3/jl}
12236 instruction sequence).
12239 @opindex msdata=none
12240 Disable use of the small data area. Variables will be put into
12241 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
12242 @code{section} attribute has been specified).
12243 This is the default.
12245 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
12246 Objects may be explicitly put in the small data area with the
12247 @code{section} attribute using one of these sections.
12249 @item -msdata=sdata
12250 @opindex msdata=sdata
12251 Put small global and static data in the small data area, but do not
12252 generate special code to reference them.
12255 @opindex msdata=use
12256 Put small global and static data in the small data area, and generate
12257 special instructions to reference them.
12261 @cindex smaller data references
12262 Put global and static objects less than or equal to @var{num} bytes
12263 into the small data or bss sections instead of the normal data or bss
12264 sections. The default value of @var{num} is 8.
12265 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
12266 for this option to have any effect.
12268 All modules should be compiled with the same @option{-G @var{num}} value.
12269 Compiling with different values of @var{num} may or may not work; if it
12270 doesn't the linker will give an error message---incorrect code will not be
12275 Makes the M32R specific code in the compiler display some statistics
12276 that might help in debugging programs.
12278 @item -malign-loops
12279 @opindex malign-loops
12280 Align all loops to a 32-byte boundary.
12282 @item -mno-align-loops
12283 @opindex mno-align-loops
12284 Do not enforce a 32-byte alignment for loops. This is the default.
12286 @item -missue-rate=@var{number}
12287 @opindex missue-rate=@var{number}
12288 Issue @var{number} instructions per cycle. @var{number} can only be 1
12291 @item -mbranch-cost=@var{number}
12292 @opindex mbranch-cost=@var{number}
12293 @var{number} can only be 1 or 2. If it is 1 then branches will be
12294 preferred over conditional code, if it is 2, then the opposite will
12297 @item -mflush-trap=@var{number}
12298 @opindex mflush-trap=@var{number}
12299 Specifies the trap number to use to flush the cache. The default is
12300 12. Valid numbers are between 0 and 15 inclusive.
12302 @item -mno-flush-trap
12303 @opindex mno-flush-trap
12304 Specifies that the cache cannot be flushed by using a trap.
12306 @item -mflush-func=@var{name}
12307 @opindex mflush-func=@var{name}
12308 Specifies the name of the operating system function to call to flush
12309 the cache. The default is @emph{_flush_cache}, but a function call
12310 will only be used if a trap is not available.
12312 @item -mno-flush-func
12313 @opindex mno-flush-func
12314 Indicates that there is no OS function for flushing the cache.
12318 @node M680x0 Options
12319 @subsection M680x0 Options
12320 @cindex M680x0 options
12322 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
12323 The default settings depend on which architecture was selected when
12324 the compiler was configured; the defaults for the most common choices
12328 @item -march=@var{arch}
12330 Generate code for a specific M680x0 or ColdFire instruction set
12331 architecture. Permissible values of @var{arch} for M680x0
12332 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
12333 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
12334 architectures are selected according to Freescale's ISA classification
12335 and the permissible values are: @samp{isaa}, @samp{isaaplus},
12336 @samp{isab} and @samp{isac}.
12338 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
12339 code for a ColdFire target. The @var{arch} in this macro is one of the
12340 @option{-march} arguments given above.
12342 When used together, @option{-march} and @option{-mtune} select code
12343 that runs on a family of similar processors but that is optimized
12344 for a particular microarchitecture.
12346 @item -mcpu=@var{cpu}
12348 Generate code for a specific M680x0 or ColdFire processor.
12349 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
12350 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
12351 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
12352 below, which also classifies the CPUs into families:
12354 @multitable @columnfractions 0.20 0.80
12355 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
12356 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
12357 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
12358 @item @samp{5206e} @tab @samp{5206e}
12359 @item @samp{5208} @tab @samp{5207} @samp{5208}
12360 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
12361 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
12362 @item @samp{5216} @tab @samp{5214} @samp{5216}
12363 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
12364 @item @samp{5225} @tab @samp{5224} @samp{5225}
12365 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
12366 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
12367 @item @samp{5249} @tab @samp{5249}
12368 @item @samp{5250} @tab @samp{5250}
12369 @item @samp{5271} @tab @samp{5270} @samp{5271}
12370 @item @samp{5272} @tab @samp{5272}
12371 @item @samp{5275} @tab @samp{5274} @samp{5275}
12372 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
12373 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
12374 @item @samp{5307} @tab @samp{5307}
12375 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
12376 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
12377 @item @samp{5407} @tab @samp{5407}
12378 @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}
12381 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
12382 @var{arch} is compatible with @var{cpu}. Other combinations of
12383 @option{-mcpu} and @option{-march} are rejected.
12385 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
12386 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
12387 where the value of @var{family} is given by the table above.
12389 @item -mtune=@var{tune}
12391 Tune the code for a particular microarchitecture, within the
12392 constraints set by @option{-march} and @option{-mcpu}.
12393 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
12394 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
12395 and @samp{cpu32}. The ColdFire microarchitectures
12396 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
12398 You can also use @option{-mtune=68020-40} for code that needs
12399 to run relatively well on 68020, 68030 and 68040 targets.
12400 @option{-mtune=68020-60} is similar but includes 68060 targets
12401 as well. These two options select the same tuning decisions as
12402 @option{-m68020-40} and @option{-m68020-60} respectively.
12404 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
12405 when tuning for 680x0 architecture @var{arch}. It also defines
12406 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
12407 option is used. If gcc is tuning for a range of architectures,
12408 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
12409 it defines the macros for every architecture in the range.
12411 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
12412 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
12413 of the arguments given above.
12419 Generate output for a 68000. This is the default
12420 when the compiler is configured for 68000-based systems.
12421 It is equivalent to @option{-march=68000}.
12423 Use this option for microcontrollers with a 68000 or EC000 core,
12424 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
12428 Generate output for a 68010. This is the default
12429 when the compiler is configured for 68010-based systems.
12430 It is equivalent to @option{-march=68010}.
12436 Generate output for a 68020. This is the default
12437 when the compiler is configured for 68020-based systems.
12438 It is equivalent to @option{-march=68020}.
12442 Generate output for a 68030. This is the default when the compiler is
12443 configured for 68030-based systems. It is equivalent to
12444 @option{-march=68030}.
12448 Generate output for a 68040. This is the default when the compiler is
12449 configured for 68040-based systems. It is equivalent to
12450 @option{-march=68040}.
12452 This option inhibits the use of 68881/68882 instructions that have to be
12453 emulated by software on the 68040. Use this option if your 68040 does not
12454 have code to emulate those instructions.
12458 Generate output for a 68060. This is the default when the compiler is
12459 configured for 68060-based systems. It is equivalent to
12460 @option{-march=68060}.
12462 This option inhibits the use of 68020 and 68881/68882 instructions that
12463 have to be emulated by software on the 68060. Use this option if your 68060
12464 does not have code to emulate those instructions.
12468 Generate output for a CPU32. This is the default
12469 when the compiler is configured for CPU32-based systems.
12470 It is equivalent to @option{-march=cpu32}.
12472 Use this option for microcontrollers with a
12473 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
12474 68336, 68340, 68341, 68349 and 68360.
12478 Generate output for a 520X ColdFire CPU@. This is the default
12479 when the compiler is configured for 520X-based systems.
12480 It is equivalent to @option{-mcpu=5206}, and is now deprecated
12481 in favor of that option.
12483 Use this option for microcontroller with a 5200 core, including
12484 the MCF5202, MCF5203, MCF5204 and MCF5206.
12488 Generate output for a 5206e ColdFire CPU@. The option is now
12489 deprecated in favor of the equivalent @option{-mcpu=5206e}.
12493 Generate output for a member of the ColdFire 528X family.
12494 The option is now deprecated in favor of the equivalent
12495 @option{-mcpu=528x}.
12499 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
12500 in favor of the equivalent @option{-mcpu=5307}.
12504 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
12505 in favor of the equivalent @option{-mcpu=5407}.
12509 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
12510 This includes use of hardware floating point instructions.
12511 The option is equivalent to @option{-mcpu=547x}, and is now
12512 deprecated in favor of that option.
12516 Generate output for a 68040, without using any of the new instructions.
12517 This results in code which can run relatively efficiently on either a
12518 68020/68881 or a 68030 or a 68040. The generated code does use the
12519 68881 instructions that are emulated on the 68040.
12521 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
12525 Generate output for a 68060, without using any of the new instructions.
12526 This results in code which can run relatively efficiently on either a
12527 68020/68881 or a 68030 or a 68040. The generated code does use the
12528 68881 instructions that are emulated on the 68060.
12530 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
12534 @opindex mhard-float
12536 Generate floating-point instructions. This is the default for 68020
12537 and above, and for ColdFire devices that have an FPU@. It defines the
12538 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
12539 on ColdFire targets.
12542 @opindex msoft-float
12543 Do not generate floating-point instructions; use library calls instead.
12544 This is the default for 68000, 68010, and 68832 targets. It is also
12545 the default for ColdFire devices that have no FPU.
12551 Generate (do not generate) ColdFire hardware divide and remainder
12552 instructions. If @option{-march} is used without @option{-mcpu},
12553 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
12554 architectures. Otherwise, the default is taken from the target CPU
12555 (either the default CPU, or the one specified by @option{-mcpu}). For
12556 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
12557 @option{-mcpu=5206e}.
12559 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
12563 Consider type @code{int} to be 16 bits wide, like @code{short int}.
12564 Additionally, parameters passed on the stack are also aligned to a
12565 16-bit boundary even on targets whose API mandates promotion to 32-bit.
12569 Do not consider type @code{int} to be 16 bits wide. This is the default.
12572 @itemx -mno-bitfield
12573 @opindex mnobitfield
12574 @opindex mno-bitfield
12575 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
12576 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
12580 Do use the bit-field instructions. The @option{-m68020} option implies
12581 @option{-mbitfield}. This is the default if you use a configuration
12582 designed for a 68020.
12586 Use a different function-calling convention, in which functions
12587 that take a fixed number of arguments return with the @code{rtd}
12588 instruction, which pops their arguments while returning. This
12589 saves one instruction in the caller since there is no need to pop
12590 the arguments there.
12592 This calling convention is incompatible with the one normally
12593 used on Unix, so you cannot use it if you need to call libraries
12594 compiled with the Unix compiler.
12596 Also, you must provide function prototypes for all functions that
12597 take variable numbers of arguments (including @code{printf});
12598 otherwise incorrect code will be generated for calls to those
12601 In addition, seriously incorrect code will result if you call a
12602 function with too many arguments. (Normally, extra arguments are
12603 harmlessly ignored.)
12605 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
12606 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
12610 Do not use the calling conventions selected by @option{-mrtd}.
12611 This is the default.
12614 @itemx -mno-align-int
12615 @opindex malign-int
12616 @opindex mno-align-int
12617 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
12618 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
12619 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
12620 Aligning variables on 32-bit boundaries produces code that runs somewhat
12621 faster on processors with 32-bit busses at the expense of more memory.
12623 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
12624 align structures containing the above types differently than
12625 most published application binary interface specifications for the m68k.
12629 Use the pc-relative addressing mode of the 68000 directly, instead of
12630 using a global offset table. At present, this option implies @option{-fpic},
12631 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
12632 not presently supported with @option{-mpcrel}, though this could be supported for
12633 68020 and higher processors.
12635 @item -mno-strict-align
12636 @itemx -mstrict-align
12637 @opindex mno-strict-align
12638 @opindex mstrict-align
12639 Do not (do) assume that unaligned memory references will be handled by
12643 Generate code that allows the data segment to be located in a different
12644 area of memory from the text segment. This allows for execute in place in
12645 an environment without virtual memory management. This option implies
12648 @item -mno-sep-data
12649 Generate code that assumes that the data segment follows the text segment.
12650 This is the default.
12652 @item -mid-shared-library
12653 Generate code that supports shared libraries via the library ID method.
12654 This allows for execute in place and shared libraries in an environment
12655 without virtual memory management. This option implies @option{-fPIC}.
12657 @item -mno-id-shared-library
12658 Generate code that doesn't assume ID based shared libraries are being used.
12659 This is the default.
12661 @item -mshared-library-id=n
12662 Specified the identification number of the ID based shared library being
12663 compiled. Specifying a value of 0 will generate more compact code, specifying
12664 other values will force the allocation of that number to the current
12665 library but is no more space or time efficient than omitting this option.
12671 When generating position-independent code for ColdFire, generate code
12672 that works if the GOT has more than 8192 entries. This code is
12673 larger and slower than code generated without this option. On M680x0
12674 processors, this option is not needed; @option{-fPIC} suffices.
12676 GCC normally uses a single instruction to load values from the GOT@.
12677 While this is relatively efficient, it only works if the GOT
12678 is smaller than about 64k. Anything larger causes the linker
12679 to report an error such as:
12681 @cindex relocation truncated to fit (ColdFire)
12683 relocation truncated to fit: R_68K_GOT16O foobar
12686 If this happens, you should recompile your code with @option{-mxgot}.
12687 It should then work with very large GOTs. However, code generated with
12688 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
12689 the value of a global symbol.
12691 Note that some linkers, including newer versions of the GNU linker,
12692 can create multiple GOTs and sort GOT entries. If you have such a linker,
12693 you should only need to use @option{-mxgot} when compiling a single
12694 object file that accesses more than 8192 GOT entries. Very few do.
12696 These options have no effect unless GCC is generating
12697 position-independent code.
12701 @node M68hc1x Options
12702 @subsection M68hc1x Options
12703 @cindex M68hc1x options
12705 These are the @samp{-m} options defined for the 68hc11 and 68hc12
12706 microcontrollers. The default values for these options depends on
12707 which style of microcontroller was selected when the compiler was configured;
12708 the defaults for the most common choices are given below.
12715 Generate output for a 68HC11. This is the default
12716 when the compiler is configured for 68HC11-based systems.
12722 Generate output for a 68HC12. This is the default
12723 when the compiler is configured for 68HC12-based systems.
12729 Generate output for a 68HCS12.
12731 @item -mauto-incdec
12732 @opindex mauto-incdec
12733 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
12740 Enable the use of 68HC12 min and max instructions.
12743 @itemx -mno-long-calls
12744 @opindex mlong-calls
12745 @opindex mno-long-calls
12746 Treat all calls as being far away (near). If calls are assumed to be
12747 far away, the compiler will use the @code{call} instruction to
12748 call a function and the @code{rtc} instruction for returning.
12752 Consider type @code{int} to be 16 bits wide, like @code{short int}.
12754 @item -msoft-reg-count=@var{count}
12755 @opindex msoft-reg-count
12756 Specify the number of pseudo-soft registers which are used for the
12757 code generation. The maximum number is 32. Using more pseudo-soft
12758 register may or may not result in better code depending on the program.
12759 The default is 4 for 68HC11 and 2 for 68HC12.
12763 @node MCore Options
12764 @subsection MCore Options
12765 @cindex MCore options
12767 These are the @samp{-m} options defined for the Motorola M*Core
12773 @itemx -mno-hardlit
12775 @opindex mno-hardlit
12776 Inline constants into the code stream if it can be done in two
12777 instructions or less.
12783 Use the divide instruction. (Enabled by default).
12785 @item -mrelax-immediate
12786 @itemx -mno-relax-immediate
12787 @opindex mrelax-immediate
12788 @opindex mno-relax-immediate
12789 Allow arbitrary sized immediates in bit operations.
12791 @item -mwide-bitfields
12792 @itemx -mno-wide-bitfields
12793 @opindex mwide-bitfields
12794 @opindex mno-wide-bitfields
12795 Always treat bit-fields as int-sized.
12797 @item -m4byte-functions
12798 @itemx -mno-4byte-functions
12799 @opindex m4byte-functions
12800 @opindex mno-4byte-functions
12801 Force all functions to be aligned to a four byte boundary.
12803 @item -mcallgraph-data
12804 @itemx -mno-callgraph-data
12805 @opindex mcallgraph-data
12806 @opindex mno-callgraph-data
12807 Emit callgraph information.
12810 @itemx -mno-slow-bytes
12811 @opindex mslow-bytes
12812 @opindex mno-slow-bytes
12813 Prefer word access when reading byte quantities.
12815 @item -mlittle-endian
12816 @itemx -mbig-endian
12817 @opindex mlittle-endian
12818 @opindex mbig-endian
12819 Generate code for a little endian target.
12825 Generate code for the 210 processor.
12829 Assume that run-time support has been provided and so omit the
12830 simulator library (@file{libsim.a)} from the linker command line.
12832 @item -mstack-increment=@var{size}
12833 @opindex mstack-increment
12834 Set the maximum amount for a single stack increment operation. Large
12835 values can increase the speed of programs which contain functions
12836 that need a large amount of stack space, but they can also trigger a
12837 segmentation fault if the stack is extended too much. The default
12843 @subsection MeP Options
12844 @cindex MeP options
12850 Enables the @code{abs} instruction, which is the absolute difference
12851 between two registers.
12855 Enables all the optional instructions - average, multiply, divide, bit
12856 operations, leading zero, absolute difference, min/max, clip, and
12862 Enables the @code{ave} instruction, which computes the average of two
12865 @item -mbased=@var{n}
12867 Variables of size @var{n} bytes or smaller will be placed in the
12868 @code{.based} section by default. Based variables use the @code{$tp}
12869 register as a base register, and there is a 128 byte limit to the
12870 @code{.based} section.
12874 Enables the bit operation instructions - bit test (@code{btstm}), set
12875 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
12876 test-and-set (@code{tas}).
12878 @item -mc=@var{name}
12880 Selects which section constant data will be placed in. @var{name} may
12881 be @code{tiny}, @code{near}, or @code{far}.
12885 Enables the @code{clip} instruction. Note that @code{-mclip} is not
12886 useful unless you also provide @code{-mminmax}.
12888 @item -mconfig=@var{name}
12890 Selects one of the build-in core configurations. Each MeP chip has
12891 one or more modules in it; each module has a core CPU and a variety of
12892 coprocessors, optional instructions, and peripherals. The
12893 @code{MeP-Integrator} tool, not part of GCC, provides these
12894 configurations through this option; using this option is the same as
12895 using all the corresponding command line options. The default
12896 configuration is @code{default}.
12900 Enables the coprocessor instructions. By default, this is a 32-bit
12901 coprocessor. Note that the coprocessor is normally enabled via the
12902 @code{-mconfig=} option.
12906 Enables the 32-bit coprocessor's instructions.
12910 Enables the 64-bit coprocessor's instructions.
12914 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
12918 Causes constant variables to be placed in the @code{.near} section.
12922 Enables the @code{div} and @code{divu} instructions.
12926 Generate big-endian code.
12930 Generate little-endian code.
12932 @item -mio-volatile
12933 @opindex mio-volatile
12934 Tells the compiler that any variable marked with the @code{io}
12935 attribute is to be considered volatile.
12939 Causes variables to be assigned to the @code{.far} section by default.
12943 Enables the @code{leadz} (leading zero) instruction.
12947 Causes variables to be assigned to the @code{.near} section by default.
12951 Enables the @code{min} and @code{max} instructions.
12955 Enables the multiplication and multiply-accumulate instructions.
12959 Disables all the optional instructions enabled by @code{-mall-opts}.
12963 Enables the @code{repeat} and @code{erepeat} instructions, used for
12964 low-overhead looping.
12968 Causes all variables to default to the @code{.tiny} section. Note
12969 that there is a 65536 byte limit to this section. Accesses to these
12970 variables use the @code{%gp} base register.
12974 Enables the saturation instructions. Note that the compiler does not
12975 currently generate these itself, but this option is included for
12976 compatibility with other tools, like @code{as}.
12980 Link the SDRAM-based runtime instead of the default ROM-based runtime.
12984 Link the simulator runtime libraries.
12988 Link the simulator runtime libraries, excluding built-in support
12989 for reset and exception vectors and tables.
12993 Causes all functions to default to the @code{.far} section. Without
12994 this option, functions default to the @code{.near} section.
12996 @item -mtiny=@var{n}
12998 Variables that are @var{n} bytes or smaller will be allocated to the
12999 @code{.tiny} section. These variables use the @code{$gp} base
13000 register. The default for this option is 4, but note that there's a
13001 65536 byte limit to the @code{.tiny} section.
13006 @subsection MIPS Options
13007 @cindex MIPS options
13013 Generate big-endian code.
13017 Generate little-endian code. This is the default for @samp{mips*el-*-*}
13020 @item -march=@var{arch}
13022 Generate code that will run on @var{arch}, which can be the name of a
13023 generic MIPS ISA, or the name of a particular processor.
13025 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
13026 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
13027 The processor names are:
13028 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
13029 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
13030 @samp{5kc}, @samp{5kf},
13032 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
13033 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
13034 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
13035 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
13036 @samp{loongson2e}, @samp{loongson2f},
13040 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
13041 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
13042 @samp{rm7000}, @samp{rm9000},
13043 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
13046 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
13047 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
13049 The special value @samp{from-abi} selects the
13050 most compatible architecture for the selected ABI (that is,
13051 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
13053 Native Linux/GNU toolchains also support the value @samp{native},
13054 which selects the best architecture option for the host processor.
13055 @option{-march=native} has no effect if GCC does not recognize
13058 In processor names, a final @samp{000} can be abbreviated as @samp{k}
13059 (for example, @samp{-march=r2k}). Prefixes are optional, and
13060 @samp{vr} may be written @samp{r}.
13062 Names of the form @samp{@var{n}f2_1} refer to processors with
13063 FPUs clocked at half the rate of the core, names of the form
13064 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
13065 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
13066 processors with FPUs clocked a ratio of 3:2 with respect to the core.
13067 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
13068 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
13069 accepted as synonyms for @samp{@var{n}f1_1}.
13071 GCC defines two macros based on the value of this option. The first
13072 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
13073 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
13074 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
13075 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
13076 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
13078 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
13079 above. In other words, it will have the full prefix and will not
13080 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
13081 the macro names the resolved architecture (either @samp{"mips1"} or
13082 @samp{"mips3"}). It names the default architecture when no
13083 @option{-march} option is given.
13085 @item -mtune=@var{arch}
13087 Optimize for @var{arch}. Among other things, this option controls
13088 the way instructions are scheduled, and the perceived cost of arithmetic
13089 operations. The list of @var{arch} values is the same as for
13092 When this option is not used, GCC will optimize for the processor
13093 specified by @option{-march}. By using @option{-march} and
13094 @option{-mtune} together, it is possible to generate code that will
13095 run on a family of processors, but optimize the code for one
13096 particular member of that family.
13098 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
13099 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
13100 @samp{-march} ones described above.
13104 Equivalent to @samp{-march=mips1}.
13108 Equivalent to @samp{-march=mips2}.
13112 Equivalent to @samp{-march=mips3}.
13116 Equivalent to @samp{-march=mips4}.
13120 Equivalent to @samp{-march=mips32}.
13124 Equivalent to @samp{-march=mips32r2}.
13128 Equivalent to @samp{-march=mips64}.
13132 Equivalent to @samp{-march=mips64r2}.
13137 @opindex mno-mips16
13138 Generate (do not generate) MIPS16 code. If GCC is targetting a
13139 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
13141 MIPS16 code generation can also be controlled on a per-function basis
13142 by means of @code{mips16} and @code{nomips16} attributes.
13143 @xref{Function Attributes}, for more information.
13145 @item -mflip-mips16
13146 @opindex mflip-mips16
13147 Generate MIPS16 code on alternating functions. This option is provided
13148 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
13149 not intended for ordinary use in compiling user code.
13151 @item -minterlink-mips16
13152 @itemx -mno-interlink-mips16
13153 @opindex minterlink-mips16
13154 @opindex mno-interlink-mips16
13155 Require (do not require) that non-MIPS16 code be link-compatible with
13158 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
13159 it must either use a call or an indirect jump. @option{-minterlink-mips16}
13160 therefore disables direct jumps unless GCC knows that the target of the
13161 jump is not MIPS16.
13173 Generate code for the given ABI@.
13175 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
13176 generates 64-bit code when you select a 64-bit architecture, but you
13177 can use @option{-mgp32} to get 32-bit code instead.
13179 For information about the O64 ABI, see
13180 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
13182 GCC supports a variant of the o32 ABI in which floating-point registers
13183 are 64 rather than 32 bits wide. You can select this combination with
13184 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
13185 and @samp{mfhc1} instructions and is therefore only supported for
13186 MIPS32R2 processors.
13188 The register assignments for arguments and return values remain the
13189 same, but each scalar value is passed in a single 64-bit register
13190 rather than a pair of 32-bit registers. For example, scalar
13191 floating-point values are returned in @samp{$f0} only, not a
13192 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
13193 remains the same, but all 64 bits are saved.
13196 @itemx -mno-abicalls
13198 @opindex mno-abicalls
13199 Generate (do not generate) code that is suitable for SVR4-style
13200 dynamic objects. @option{-mabicalls} is the default for SVR4-based
13205 Generate (do not generate) code that is fully position-independent,
13206 and that can therefore be linked into shared libraries. This option
13207 only affects @option{-mabicalls}.
13209 All @option{-mabicalls} code has traditionally been position-independent,
13210 regardless of options like @option{-fPIC} and @option{-fpic}. However,
13211 as an extension, the GNU toolchain allows executables to use absolute
13212 accesses for locally-binding symbols. It can also use shorter GP
13213 initialization sequences and generate direct calls to locally-defined
13214 functions. This mode is selected by @option{-mno-shared}.
13216 @option{-mno-shared} depends on binutils 2.16 or higher and generates
13217 objects that can only be linked by the GNU linker. However, the option
13218 does not affect the ABI of the final executable; it only affects the ABI
13219 of relocatable objects. Using @option{-mno-shared} will generally make
13220 executables both smaller and quicker.
13222 @option{-mshared} is the default.
13228 Assume (do not assume) that the static and dynamic linkers
13229 support PLTs and copy relocations. This option only affects
13230 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
13231 has no effect without @samp{-msym32}.
13233 You can make @option{-mplt} the default by configuring
13234 GCC with @option{--with-mips-plt}. The default is
13235 @option{-mno-plt} otherwise.
13241 Lift (do not lift) the usual restrictions on the size of the global
13244 GCC normally uses a single instruction to load values from the GOT@.
13245 While this is relatively efficient, it will only work if the GOT
13246 is smaller than about 64k. Anything larger will cause the linker
13247 to report an error such as:
13249 @cindex relocation truncated to fit (MIPS)
13251 relocation truncated to fit: R_MIPS_GOT16 foobar
13254 If this happens, you should recompile your code with @option{-mxgot}.
13255 It should then work with very large GOTs, although it will also be
13256 less efficient, since it will take three instructions to fetch the
13257 value of a global symbol.
13259 Note that some linkers can create multiple GOTs. If you have such a
13260 linker, you should only need to use @option{-mxgot} when a single object
13261 file accesses more than 64k's worth of GOT entries. Very few do.
13263 These options have no effect unless GCC is generating position
13268 Assume that general-purpose registers are 32 bits wide.
13272 Assume that general-purpose registers are 64 bits wide.
13276 Assume that floating-point registers are 32 bits wide.
13280 Assume that floating-point registers are 64 bits wide.
13283 @opindex mhard-float
13284 Use floating-point coprocessor instructions.
13287 @opindex msoft-float
13288 Do not use floating-point coprocessor instructions. Implement
13289 floating-point calculations using library calls instead.
13291 @item -msingle-float
13292 @opindex msingle-float
13293 Assume that the floating-point coprocessor only supports single-precision
13296 @item -mdouble-float
13297 @opindex mdouble-float
13298 Assume that the floating-point coprocessor supports double-precision
13299 operations. This is the default.
13305 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
13306 implement atomic memory built-in functions. When neither option is
13307 specified, GCC will use the instructions if the target architecture
13310 @option{-mllsc} is useful if the runtime environment can emulate the
13311 instructions and @option{-mno-llsc} can be useful when compiling for
13312 nonstandard ISAs. You can make either option the default by
13313 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
13314 respectively. @option{--with-llsc} is the default for some
13315 configurations; see the installation documentation for details.
13321 Use (do not use) revision 1 of the MIPS DSP ASE@.
13322 @xref{MIPS DSP Built-in Functions}. This option defines the
13323 preprocessor macro @samp{__mips_dsp}. It also defines
13324 @samp{__mips_dsp_rev} to 1.
13330 Use (do not use) revision 2 of the MIPS DSP ASE@.
13331 @xref{MIPS DSP Built-in Functions}. This option defines the
13332 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
13333 It also defines @samp{__mips_dsp_rev} to 2.
13336 @itemx -mno-smartmips
13337 @opindex msmartmips
13338 @opindex mno-smartmips
13339 Use (do not use) the MIPS SmartMIPS ASE.
13341 @item -mpaired-single
13342 @itemx -mno-paired-single
13343 @opindex mpaired-single
13344 @opindex mno-paired-single
13345 Use (do not use) paired-single floating-point instructions.
13346 @xref{MIPS Paired-Single Support}. This option requires
13347 hardware floating-point support to be enabled.
13353 Use (do not use) MIPS Digital Media Extension instructions.
13354 This option can only be used when generating 64-bit code and requires
13355 hardware floating-point support to be enabled.
13360 @opindex mno-mips3d
13361 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
13362 The option @option{-mips3d} implies @option{-mpaired-single}.
13368 Use (do not use) MT Multithreading instructions.
13372 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
13373 an explanation of the default and the way that the pointer size is
13378 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
13380 The default size of @code{int}s, @code{long}s and pointers depends on
13381 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
13382 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
13383 32-bit @code{long}s. Pointers are the same size as @code{long}s,
13384 or the same size as integer registers, whichever is smaller.
13390 Assume (do not assume) that all symbols have 32-bit values, regardless
13391 of the selected ABI@. This option is useful in combination with
13392 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
13393 to generate shorter and faster references to symbolic addresses.
13397 Put definitions of externally-visible data in a small data section
13398 if that data is no bigger than @var{num} bytes. GCC can then access
13399 the data more efficiently; see @option{-mgpopt} for details.
13401 The default @option{-G} option depends on the configuration.
13403 @item -mlocal-sdata
13404 @itemx -mno-local-sdata
13405 @opindex mlocal-sdata
13406 @opindex mno-local-sdata
13407 Extend (do not extend) the @option{-G} behavior to local data too,
13408 such as to static variables in C@. @option{-mlocal-sdata} is the
13409 default for all configurations.
13411 If the linker complains that an application is using too much small data,
13412 you might want to try rebuilding the less performance-critical parts with
13413 @option{-mno-local-sdata}. You might also want to build large
13414 libraries with @option{-mno-local-sdata}, so that the libraries leave
13415 more room for the main program.
13417 @item -mextern-sdata
13418 @itemx -mno-extern-sdata
13419 @opindex mextern-sdata
13420 @opindex mno-extern-sdata
13421 Assume (do not assume) that externally-defined data will be in
13422 a small data section if that data is within the @option{-G} limit.
13423 @option{-mextern-sdata} is the default for all configurations.
13425 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
13426 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
13427 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
13428 is placed in a small data section. If @var{Var} is defined by another
13429 module, you must either compile that module with a high-enough
13430 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
13431 definition. If @var{Var} is common, you must link the application
13432 with a high-enough @option{-G} setting.
13434 The easiest way of satisfying these restrictions is to compile
13435 and link every module with the same @option{-G} option. However,
13436 you may wish to build a library that supports several different
13437 small data limits. You can do this by compiling the library with
13438 the highest supported @option{-G} setting and additionally using
13439 @option{-mno-extern-sdata} to stop the library from making assumptions
13440 about externally-defined data.
13446 Use (do not use) GP-relative accesses for symbols that are known to be
13447 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
13448 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
13451 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
13452 might not hold the value of @code{_gp}. For example, if the code is
13453 part of a library that might be used in a boot monitor, programs that
13454 call boot monitor routines will pass an unknown value in @code{$gp}.
13455 (In such situations, the boot monitor itself would usually be compiled
13456 with @option{-G0}.)
13458 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
13459 @option{-mno-extern-sdata}.
13461 @item -membedded-data
13462 @itemx -mno-embedded-data
13463 @opindex membedded-data
13464 @opindex mno-embedded-data
13465 Allocate variables to the read-only data section first if possible, then
13466 next in the small data section if possible, otherwise in data. This gives
13467 slightly slower code than the default, but reduces the amount of RAM required
13468 when executing, and thus may be preferred for some embedded systems.
13470 @item -muninit-const-in-rodata
13471 @itemx -mno-uninit-const-in-rodata
13472 @opindex muninit-const-in-rodata
13473 @opindex mno-uninit-const-in-rodata
13474 Put uninitialized @code{const} variables in the read-only data section.
13475 This option is only meaningful in conjunction with @option{-membedded-data}.
13477 @item -mcode-readable=@var{setting}
13478 @opindex mcode-readable
13479 Specify whether GCC may generate code that reads from executable sections.
13480 There are three possible settings:
13483 @item -mcode-readable=yes
13484 Instructions may freely access executable sections. This is the
13487 @item -mcode-readable=pcrel
13488 MIPS16 PC-relative load instructions can access executable sections,
13489 but other instructions must not do so. This option is useful on 4KSc
13490 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
13491 It is also useful on processors that can be configured to have a dual
13492 instruction/data SRAM interface and that, like the M4K, automatically
13493 redirect PC-relative loads to the instruction RAM.
13495 @item -mcode-readable=no
13496 Instructions must not access executable sections. This option can be
13497 useful on targets that are configured to have a dual instruction/data
13498 SRAM interface but that (unlike the M4K) do not automatically redirect
13499 PC-relative loads to the instruction RAM.
13502 @item -msplit-addresses
13503 @itemx -mno-split-addresses
13504 @opindex msplit-addresses
13505 @opindex mno-split-addresses
13506 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
13507 relocation operators. This option has been superseded by
13508 @option{-mexplicit-relocs} but is retained for backwards compatibility.
13510 @item -mexplicit-relocs
13511 @itemx -mno-explicit-relocs
13512 @opindex mexplicit-relocs
13513 @opindex mno-explicit-relocs
13514 Use (do not use) assembler relocation operators when dealing with symbolic
13515 addresses. The alternative, selected by @option{-mno-explicit-relocs},
13516 is to use assembler macros instead.
13518 @option{-mexplicit-relocs} is the default if GCC was configured
13519 to use an assembler that supports relocation operators.
13521 @item -mcheck-zero-division
13522 @itemx -mno-check-zero-division
13523 @opindex mcheck-zero-division
13524 @opindex mno-check-zero-division
13525 Trap (do not trap) on integer division by zero.
13527 The default is @option{-mcheck-zero-division}.
13529 @item -mdivide-traps
13530 @itemx -mdivide-breaks
13531 @opindex mdivide-traps
13532 @opindex mdivide-breaks
13533 MIPS systems check for division by zero by generating either a
13534 conditional trap or a break instruction. Using traps results in
13535 smaller code, but is only supported on MIPS II and later. Also, some
13536 versions of the Linux kernel have a bug that prevents trap from
13537 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
13538 allow conditional traps on architectures that support them and
13539 @option{-mdivide-breaks} to force the use of breaks.
13541 The default is usually @option{-mdivide-traps}, but this can be
13542 overridden at configure time using @option{--with-divide=breaks}.
13543 Divide-by-zero checks can be completely disabled using
13544 @option{-mno-check-zero-division}.
13549 @opindex mno-memcpy
13550 Force (do not force) the use of @code{memcpy()} for non-trivial block
13551 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
13552 most constant-sized copies.
13555 @itemx -mno-long-calls
13556 @opindex mlong-calls
13557 @opindex mno-long-calls
13558 Disable (do not disable) use of the @code{jal} instruction. Calling
13559 functions using @code{jal} is more efficient but requires the caller
13560 and callee to be in the same 256 megabyte segment.
13562 This option has no effect on abicalls code. The default is
13563 @option{-mno-long-calls}.
13569 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
13570 instructions, as provided by the R4650 ISA@.
13573 @itemx -mno-fused-madd
13574 @opindex mfused-madd
13575 @opindex mno-fused-madd
13576 Enable (disable) use of the floating point multiply-accumulate
13577 instructions, when they are available. The default is
13578 @option{-mfused-madd}.
13580 When multiply-accumulate instructions are used, the intermediate
13581 product is calculated to infinite precision and is not subject to
13582 the FCSR Flush to Zero bit. This may be undesirable in some
13587 Tell the MIPS assembler to not run its preprocessor over user
13588 assembler files (with a @samp{.s} suffix) when assembling them.
13591 @itemx -mno-fix-r4000
13592 @opindex mfix-r4000
13593 @opindex mno-fix-r4000
13594 Work around certain R4000 CPU errata:
13597 A double-word or a variable shift may give an incorrect result if executed
13598 immediately after starting an integer division.
13600 A double-word or a variable shift may give an incorrect result if executed
13601 while an integer multiplication is in progress.
13603 An integer division may give an incorrect result if started in a delay slot
13604 of a taken branch or a jump.
13608 @itemx -mno-fix-r4400
13609 @opindex mfix-r4400
13610 @opindex mno-fix-r4400
13611 Work around certain R4400 CPU errata:
13614 A double-word or a variable shift may give an incorrect result if executed
13615 immediately after starting an integer division.
13619 @itemx -mno-fix-r10000
13620 @opindex mfix-r10000
13621 @opindex mno-fix-r10000
13622 Work around certain R10000 errata:
13625 @code{ll}/@code{sc} sequences may not behave atomically on revisions
13626 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
13629 This option can only be used if the target architecture supports
13630 branch-likely instructions. @option{-mfix-r10000} is the default when
13631 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
13635 @itemx -mno-fix-vr4120
13636 @opindex mfix-vr4120
13637 Work around certain VR4120 errata:
13640 @code{dmultu} does not always produce the correct result.
13642 @code{div} and @code{ddiv} do not always produce the correct result if one
13643 of the operands is negative.
13645 The workarounds for the division errata rely on special functions in
13646 @file{libgcc.a}. At present, these functions are only provided by
13647 the @code{mips64vr*-elf} configurations.
13649 Other VR4120 errata require a nop to be inserted between certain pairs of
13650 instructions. These errata are handled by the assembler, not by GCC itself.
13653 @opindex mfix-vr4130
13654 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
13655 workarounds are implemented by the assembler rather than by GCC,
13656 although GCC will avoid using @code{mflo} and @code{mfhi} if the
13657 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
13658 instructions are available instead.
13661 @itemx -mno-fix-sb1
13663 Work around certain SB-1 CPU core errata.
13664 (This flag currently works around the SB-1 revision 2
13665 ``F1'' and ``F2'' floating point errata.)
13667 @item -mr10k-cache-barrier=@var{setting}
13668 @opindex mr10k-cache-barrier
13669 Specify whether GCC should insert cache barriers to avoid the
13670 side-effects of speculation on R10K processors.
13672 In common with many processors, the R10K tries to predict the outcome
13673 of a conditional branch and speculatively executes instructions from
13674 the ``taken'' branch. It later aborts these instructions if the
13675 predicted outcome was wrong. However, on the R10K, even aborted
13676 instructions can have side effects.
13678 This problem only affects kernel stores and, depending on the system,
13679 kernel loads. As an example, a speculatively-executed store may load
13680 the target memory into cache and mark the cache line as dirty, even if
13681 the store itself is later aborted. If a DMA operation writes to the
13682 same area of memory before the ``dirty'' line is flushed, the cached
13683 data will overwrite the DMA-ed data. See the R10K processor manual
13684 for a full description, including other potential problems.
13686 One workaround is to insert cache barrier instructions before every memory
13687 access that might be speculatively executed and that might have side
13688 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
13689 controls GCC's implementation of this workaround. It assumes that
13690 aborted accesses to any byte in the following regions will not have
13695 the memory occupied by the current function's stack frame;
13698 the memory occupied by an incoming stack argument;
13701 the memory occupied by an object with a link-time-constant address.
13704 It is the kernel's responsibility to ensure that speculative
13705 accesses to these regions are indeed safe.
13707 If the input program contains a function declaration such as:
13713 then the implementation of @code{foo} must allow @code{j foo} and
13714 @code{jal foo} to be executed speculatively. GCC honors this
13715 restriction for functions it compiles itself. It expects non-GCC
13716 functions (such as hand-written assembly code) to do the same.
13718 The option has three forms:
13721 @item -mr10k-cache-barrier=load-store
13722 Insert a cache barrier before a load or store that might be
13723 speculatively executed and that might have side effects even
13726 @item -mr10k-cache-barrier=store
13727 Insert a cache barrier before a store that might be speculatively
13728 executed and that might have side effects even if aborted.
13730 @item -mr10k-cache-barrier=none
13731 Disable the insertion of cache barriers. This is the default setting.
13734 @item -mflush-func=@var{func}
13735 @itemx -mno-flush-func
13736 @opindex mflush-func
13737 Specifies the function to call to flush the I and D caches, or to not
13738 call any such function. If called, the function must take the same
13739 arguments as the common @code{_flush_func()}, that is, the address of the
13740 memory range for which the cache is being flushed, the size of the
13741 memory range, and the number 3 (to flush both caches). The default
13742 depends on the target GCC was configured for, but commonly is either
13743 @samp{_flush_func} or @samp{__cpu_flush}.
13745 @item mbranch-cost=@var{num}
13746 @opindex mbranch-cost
13747 Set the cost of branches to roughly @var{num} ``simple'' instructions.
13748 This cost is only a heuristic and is not guaranteed to produce
13749 consistent results across releases. A zero cost redundantly selects
13750 the default, which is based on the @option{-mtune} setting.
13752 @item -mbranch-likely
13753 @itemx -mno-branch-likely
13754 @opindex mbranch-likely
13755 @opindex mno-branch-likely
13756 Enable or disable use of Branch Likely instructions, regardless of the
13757 default for the selected architecture. By default, Branch Likely
13758 instructions may be generated if they are supported by the selected
13759 architecture. An exception is for the MIPS32 and MIPS64 architectures
13760 and processors which implement those architectures; for those, Branch
13761 Likely instructions will not be generated by default because the MIPS32
13762 and MIPS64 architectures specifically deprecate their use.
13764 @item -mfp-exceptions
13765 @itemx -mno-fp-exceptions
13766 @opindex mfp-exceptions
13767 Specifies whether FP exceptions are enabled. This affects how we schedule
13768 FP instructions for some processors. The default is that FP exceptions are
13771 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
13772 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
13775 @item -mvr4130-align
13776 @itemx -mno-vr4130-align
13777 @opindex mvr4130-align
13778 The VR4130 pipeline is two-way superscalar, but can only issue two
13779 instructions together if the first one is 8-byte aligned. When this
13780 option is enabled, GCC will align pairs of instructions that it
13781 thinks should execute in parallel.
13783 This option only has an effect when optimizing for the VR4130.
13784 It normally makes code faster, but at the expense of making it bigger.
13785 It is enabled by default at optimization level @option{-O3}.
13790 Enable (disable) generation of @code{synci} instructions on
13791 architectures that support it. The @code{synci} instructions (if
13792 enabled) will be generated when @code{__builtin___clear_cache()} is
13795 This option defaults to @code{-mno-synci}, but the default can be
13796 overridden by configuring with @code{--with-synci}.
13798 When compiling code for single processor systems, it is generally safe
13799 to use @code{synci}. However, on many multi-core (SMP) systems, it
13800 will not invalidate the instruction caches on all cores and may lead
13801 to undefined behavior.
13805 @subsection MMIX Options
13806 @cindex MMIX Options
13808 These options are defined for the MMIX:
13812 @itemx -mno-libfuncs
13814 @opindex mno-libfuncs
13815 Specify that intrinsic library functions are being compiled, passing all
13816 values in registers, no matter the size.
13819 @itemx -mno-epsilon
13821 @opindex mno-epsilon
13822 Generate floating-point comparison instructions that compare with respect
13823 to the @code{rE} epsilon register.
13825 @item -mabi=mmixware
13827 @opindex mabi=mmixware
13829 Generate code that passes function parameters and return values that (in
13830 the called function) are seen as registers @code{$0} and up, as opposed to
13831 the GNU ABI which uses global registers @code{$231} and up.
13833 @item -mzero-extend
13834 @itemx -mno-zero-extend
13835 @opindex mzero-extend
13836 @opindex mno-zero-extend
13837 When reading data from memory in sizes shorter than 64 bits, use (do not
13838 use) zero-extending load instructions by default, rather than
13839 sign-extending ones.
13842 @itemx -mno-knuthdiv
13844 @opindex mno-knuthdiv
13845 Make the result of a division yielding a remainder have the same sign as
13846 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
13847 remainder follows the sign of the dividend. Both methods are
13848 arithmetically valid, the latter being almost exclusively used.
13850 @item -mtoplevel-symbols
13851 @itemx -mno-toplevel-symbols
13852 @opindex mtoplevel-symbols
13853 @opindex mno-toplevel-symbols
13854 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
13855 code can be used with the @code{PREFIX} assembly directive.
13859 Generate an executable in the ELF format, rather than the default
13860 @samp{mmo} format used by the @command{mmix} simulator.
13862 @item -mbranch-predict
13863 @itemx -mno-branch-predict
13864 @opindex mbranch-predict
13865 @opindex mno-branch-predict
13866 Use (do not use) the probable-branch instructions, when static branch
13867 prediction indicates a probable branch.
13869 @item -mbase-addresses
13870 @itemx -mno-base-addresses
13871 @opindex mbase-addresses
13872 @opindex mno-base-addresses
13873 Generate (do not generate) code that uses @emph{base addresses}. Using a
13874 base address automatically generates a request (handled by the assembler
13875 and the linker) for a constant to be set up in a global register. The
13876 register is used for one or more base address requests within the range 0
13877 to 255 from the value held in the register. The generally leads to short
13878 and fast code, but the number of different data items that can be
13879 addressed is limited. This means that a program that uses lots of static
13880 data may require @option{-mno-base-addresses}.
13882 @item -msingle-exit
13883 @itemx -mno-single-exit
13884 @opindex msingle-exit
13885 @opindex mno-single-exit
13886 Force (do not force) generated code to have a single exit point in each
13890 @node MN10300 Options
13891 @subsection MN10300 Options
13892 @cindex MN10300 options
13894 These @option{-m} options are defined for Matsushita MN10300 architectures:
13899 Generate code to avoid bugs in the multiply instructions for the MN10300
13900 processors. This is the default.
13902 @item -mno-mult-bug
13903 @opindex mno-mult-bug
13904 Do not generate code to avoid bugs in the multiply instructions for the
13905 MN10300 processors.
13909 Generate code which uses features specific to the AM33 processor.
13913 Do not generate code which uses features specific to the AM33 processor. This
13916 @item -mreturn-pointer-on-d0
13917 @opindex mreturn-pointer-on-d0
13918 When generating a function which returns a pointer, return the pointer
13919 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
13920 only in a0, and attempts to call such functions without a prototype
13921 would result in errors. Note that this option is on by default; use
13922 @option{-mno-return-pointer-on-d0} to disable it.
13926 Do not link in the C run-time initialization object file.
13930 Indicate to the linker that it should perform a relaxation optimization pass
13931 to shorten branches, calls and absolute memory addresses. This option only
13932 has an effect when used on the command line for the final link step.
13934 This option makes symbolic debugging impossible.
13937 @node PDP-11 Options
13938 @subsection PDP-11 Options
13939 @cindex PDP-11 Options
13941 These options are defined for the PDP-11:
13946 Use hardware FPP floating point. This is the default. (FIS floating
13947 point on the PDP-11/40 is not supported.)
13950 @opindex msoft-float
13951 Do not use hardware floating point.
13955 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
13959 Return floating-point results in memory. This is the default.
13963 Generate code for a PDP-11/40.
13967 Generate code for a PDP-11/45. This is the default.
13971 Generate code for a PDP-11/10.
13973 @item -mbcopy-builtin
13974 @opindex mbcopy-builtin
13975 Use inline @code{movmemhi} patterns for copying memory. This is the
13980 Do not use inline @code{movmemhi} patterns for copying memory.
13986 Use 16-bit @code{int}. This is the default.
13992 Use 32-bit @code{int}.
13995 @itemx -mno-float32
13997 @opindex mno-float32
13998 Use 64-bit @code{float}. This is the default.
14001 @itemx -mno-float64
14003 @opindex mno-float64
14004 Use 32-bit @code{float}.
14008 Use @code{abshi2} pattern. This is the default.
14012 Do not use @code{abshi2} pattern.
14014 @item -mbranch-expensive
14015 @opindex mbranch-expensive
14016 Pretend that branches are expensive. This is for experimenting with
14017 code generation only.
14019 @item -mbranch-cheap
14020 @opindex mbranch-cheap
14021 Do not pretend that branches are expensive. This is the default.
14025 Generate code for a system with split I&D@.
14029 Generate code for a system without split I&D@. This is the default.
14033 Use Unix assembler syntax. This is the default when configured for
14034 @samp{pdp11-*-bsd}.
14038 Use DEC assembler syntax. This is the default when configured for any
14039 PDP-11 target other than @samp{pdp11-*-bsd}.
14042 @node picoChip Options
14043 @subsection picoChip Options
14044 @cindex picoChip options
14046 These @samp{-m} options are defined for picoChip implementations:
14050 @item -mae=@var{ae_type}
14052 Set the instruction set, register set, and instruction scheduling
14053 parameters for array element type @var{ae_type}. Supported values
14054 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
14056 @option{-mae=ANY} selects a completely generic AE type. Code
14057 generated with this option will run on any of the other AE types. The
14058 code will not be as efficient as it would be if compiled for a specific
14059 AE type, and some types of operation (e.g., multiplication) will not
14060 work properly on all types of AE.
14062 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
14063 for compiled code, and is the default.
14065 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
14066 option may suffer from poor performance of byte (char) manipulation,
14067 since the DSP AE does not provide hardware support for byte load/stores.
14069 @item -msymbol-as-address
14070 Enable the compiler to directly use a symbol name as an address in a
14071 load/store instruction, without first loading it into a
14072 register. Typically, the use of this option will generate larger
14073 programs, which run faster than when the option isn't used. However, the
14074 results vary from program to program, so it is left as a user option,
14075 rather than being permanently enabled.
14077 @item -mno-inefficient-warnings
14078 Disables warnings about the generation of inefficient code. These
14079 warnings can be generated, for example, when compiling code which
14080 performs byte-level memory operations on the MAC AE type. The MAC AE has
14081 no hardware support for byte-level memory operations, so all byte
14082 load/stores must be synthesized from word load/store operations. This is
14083 inefficient and a warning will be generated indicating to the programmer
14084 that they should rewrite the code to avoid byte operations, or to target
14085 an AE type which has the necessary hardware support. This option enables
14086 the warning to be turned off.
14090 @node PowerPC Options
14091 @subsection PowerPC Options
14092 @cindex PowerPC options
14094 These are listed under @xref{RS/6000 and PowerPC Options}.
14096 @node RS/6000 and PowerPC Options
14097 @subsection IBM RS/6000 and PowerPC Options
14098 @cindex RS/6000 and PowerPC Options
14099 @cindex IBM RS/6000 and PowerPC Options
14101 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
14108 @itemx -mno-powerpc
14109 @itemx -mpowerpc-gpopt
14110 @itemx -mno-powerpc-gpopt
14111 @itemx -mpowerpc-gfxopt
14112 @itemx -mno-powerpc-gfxopt
14114 @itemx -mno-powerpc64
14118 @itemx -mno-popcntb
14126 @itemx -mno-hard-dfp
14130 @opindex mno-power2
14132 @opindex mno-powerpc
14133 @opindex mpowerpc-gpopt
14134 @opindex mno-powerpc-gpopt
14135 @opindex mpowerpc-gfxopt
14136 @opindex mno-powerpc-gfxopt
14137 @opindex mpowerpc64
14138 @opindex mno-powerpc64
14142 @opindex mno-popcntb
14148 @opindex mno-mfpgpr
14150 @opindex mno-hard-dfp
14151 GCC supports two related instruction set architectures for the
14152 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
14153 instructions supported by the @samp{rios} chip set used in the original
14154 RS/6000 systems and the @dfn{PowerPC} instruction set is the
14155 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
14156 the IBM 4xx, 6xx, and follow-on microprocessors.
14158 Neither architecture is a subset of the other. However there is a
14159 large common subset of instructions supported by both. An MQ
14160 register is included in processors supporting the POWER architecture.
14162 You use these options to specify which instructions are available on the
14163 processor you are using. The default value of these options is
14164 determined when configuring GCC@. Specifying the
14165 @option{-mcpu=@var{cpu_type}} overrides the specification of these
14166 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
14167 rather than the options listed above.
14169 The @option{-mpower} option allows GCC to generate instructions that
14170 are found only in the POWER architecture and to use the MQ register.
14171 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
14172 to generate instructions that are present in the POWER2 architecture but
14173 not the original POWER architecture.
14175 The @option{-mpowerpc} option allows GCC to generate instructions that
14176 are found only in the 32-bit subset of the PowerPC architecture.
14177 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
14178 GCC to use the optional PowerPC architecture instructions in the
14179 General Purpose group, including floating-point square root. Specifying
14180 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
14181 use the optional PowerPC architecture instructions in the Graphics
14182 group, including floating-point select.
14184 The @option{-mmfcrf} option allows GCC to generate the move from
14185 condition register field instruction implemented on the POWER4
14186 processor and other processors that support the PowerPC V2.01
14188 The @option{-mpopcntb} option allows GCC to generate the popcount and
14189 double precision FP reciprocal estimate instruction implemented on the
14190 POWER5 processor and other processors that support the PowerPC V2.02
14192 The @option{-mfprnd} option allows GCC to generate the FP round to
14193 integer instructions implemented on the POWER5+ processor and other
14194 processors that support the PowerPC V2.03 architecture.
14195 The @option{-mcmpb} option allows GCC to generate the compare bytes
14196 instruction implemented on the POWER6 processor and other processors
14197 that support the PowerPC V2.05 architecture.
14198 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
14199 general purpose register instructions implemented on the POWER6X
14200 processor and other processors that support the extended PowerPC V2.05
14202 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
14203 point instructions implemented on some POWER processors.
14205 The @option{-mpowerpc64} option allows GCC to generate the additional
14206 64-bit instructions that are found in the full PowerPC64 architecture
14207 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
14208 @option{-mno-powerpc64}.
14210 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
14211 will use only the instructions in the common subset of both
14212 architectures plus some special AIX common-mode calls, and will not use
14213 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
14214 permits GCC to use any instruction from either architecture and to
14215 allow use of the MQ register; specify this for the Motorola MPC601.
14217 @item -mnew-mnemonics
14218 @itemx -mold-mnemonics
14219 @opindex mnew-mnemonics
14220 @opindex mold-mnemonics
14221 Select which mnemonics to use in the generated assembler code. With
14222 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
14223 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
14224 assembler mnemonics defined for the POWER architecture. Instructions
14225 defined in only one architecture have only one mnemonic; GCC uses that
14226 mnemonic irrespective of which of these options is specified.
14228 GCC defaults to the mnemonics appropriate for the architecture in
14229 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
14230 value of these option. Unless you are building a cross-compiler, you
14231 should normally not specify either @option{-mnew-mnemonics} or
14232 @option{-mold-mnemonics}, but should instead accept the default.
14234 @item -mcpu=@var{cpu_type}
14236 Set architecture type, register usage, choice of mnemonics, and
14237 instruction scheduling parameters for machine type @var{cpu_type}.
14238 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
14239 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
14240 @samp{505}, @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
14241 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
14242 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
14243 @samp{860}, @samp{970}, @samp{8540}, @samp{e300c2}, @samp{e300c3},
14244 @samp{e500mc}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
14245 @samp{power}, @samp{power2}, @samp{power3}, @samp{power4},
14246 @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x}, @samp{power7}
14247 @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
14248 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
14250 @option{-mcpu=common} selects a completely generic processor. Code
14251 generated under this option will run on any POWER or PowerPC processor.
14252 GCC will use only the instructions in the common subset of both
14253 architectures, and will not use the MQ register. GCC assumes a generic
14254 processor model for scheduling purposes.
14256 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
14257 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
14258 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
14259 types, with an appropriate, generic processor model assumed for
14260 scheduling purposes.
14262 The other options specify a specific processor. Code generated under
14263 those options will run best on that processor, and may not run at all on
14266 The @option{-mcpu} options automatically enable or disable the
14269 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
14270 -mnew-mnemonics -mpopcntb -mpower -mpower2 -mpowerpc64 @gol
14271 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
14272 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr}
14274 The particular options set for any particular CPU will vary between
14275 compiler versions, depending on what setting seems to produce optimal
14276 code for that CPU; it doesn't necessarily reflect the actual hardware's
14277 capabilities. If you wish to set an individual option to a particular
14278 value, you may specify it after the @option{-mcpu} option, like
14279 @samp{-mcpu=970 -mno-altivec}.
14281 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
14282 not enabled or disabled by the @option{-mcpu} option at present because
14283 AIX does not have full support for these options. You may still
14284 enable or disable them individually if you're sure it'll work in your
14287 @item -mtune=@var{cpu_type}
14289 Set the instruction scheduling parameters for machine type
14290 @var{cpu_type}, but do not set the architecture type, register usage, or
14291 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
14292 values for @var{cpu_type} are used for @option{-mtune} as for
14293 @option{-mcpu}. If both are specified, the code generated will use the
14294 architecture, registers, and mnemonics set by @option{-mcpu}, but the
14295 scheduling parameters set by @option{-mtune}.
14301 Generate code to compute division as reciprocal estimate and iterative
14302 refinement, creating opportunities for increased throughput. This
14303 feature requires: optional PowerPC Graphics instruction set for single
14304 precision and FRE instruction for double precision, assuming divides
14305 cannot generate user-visible traps, and the domain values not include
14306 Infinities, denormals or zero denominator.
14309 @itemx -mno-altivec
14311 @opindex mno-altivec
14312 Generate code that uses (does not use) AltiVec instructions, and also
14313 enable the use of built-in functions that allow more direct access to
14314 the AltiVec instruction set. You may also need to set
14315 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
14321 @opindex mno-vrsave
14322 Generate VRSAVE instructions when generating AltiVec code.
14324 @item -mgen-cell-microcode
14325 @opindex mgen-cell-microcode
14326 Generate Cell microcode instructions
14328 @item -mwarn-cell-microcode
14329 @opindex mwarn-cell-microcode
14330 Warning when a Cell microcode instruction is going to emitted. An example
14331 of a Cell microcode instruction is a variable shift.
14334 @opindex msecure-plt
14335 Generate code that allows ld and ld.so to build executables and shared
14336 libraries with non-exec .plt and .got sections. This is a PowerPC
14337 32-bit SYSV ABI option.
14341 Generate code that uses a BSS .plt section that ld.so fills in, and
14342 requires .plt and .got sections that are both writable and executable.
14343 This is a PowerPC 32-bit SYSV ABI option.
14349 This switch enables or disables the generation of ISEL instructions.
14351 @item -misel=@var{yes/no}
14352 This switch has been deprecated. Use @option{-misel} and
14353 @option{-mno-isel} instead.
14359 This switch enables or disables the generation of SPE simd
14365 @opindex mno-paired
14366 This switch enables or disables the generation of PAIRED simd
14369 @item -mspe=@var{yes/no}
14370 This option has been deprecated. Use @option{-mspe} and
14371 @option{-mno-spe} instead.
14373 @item -mfloat-gprs=@var{yes/single/double/no}
14374 @itemx -mfloat-gprs
14375 @opindex mfloat-gprs
14376 This switch enables or disables the generation of floating point
14377 operations on the general purpose registers for architectures that
14380 The argument @var{yes} or @var{single} enables the use of
14381 single-precision floating point operations.
14383 The argument @var{double} enables the use of single and
14384 double-precision floating point operations.
14386 The argument @var{no} disables floating point operations on the
14387 general purpose registers.
14389 This option is currently only available on the MPC854x.
14395 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
14396 targets (including GNU/Linux). The 32-bit environment sets int, long
14397 and pointer to 32 bits and generates code that runs on any PowerPC
14398 variant. The 64-bit environment sets int to 32 bits and long and
14399 pointer to 64 bits, and generates code for PowerPC64, as for
14400 @option{-mpowerpc64}.
14403 @itemx -mno-fp-in-toc
14404 @itemx -mno-sum-in-toc
14405 @itemx -mminimal-toc
14407 @opindex mno-fp-in-toc
14408 @opindex mno-sum-in-toc
14409 @opindex mminimal-toc
14410 Modify generation of the TOC (Table Of Contents), which is created for
14411 every executable file. The @option{-mfull-toc} option is selected by
14412 default. In that case, GCC will allocate at least one TOC entry for
14413 each unique non-automatic variable reference in your program. GCC
14414 will also place floating-point constants in the TOC@. However, only
14415 16,384 entries are available in the TOC@.
14417 If you receive a linker error message that saying you have overflowed
14418 the available TOC space, you can reduce the amount of TOC space used
14419 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
14420 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
14421 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
14422 generate code to calculate the sum of an address and a constant at
14423 run-time instead of putting that sum into the TOC@. You may specify one
14424 or both of these options. Each causes GCC to produce very slightly
14425 slower and larger code at the expense of conserving TOC space.
14427 If you still run out of space in the TOC even when you specify both of
14428 these options, specify @option{-mminimal-toc} instead. This option causes
14429 GCC to make only one TOC entry for every file. When you specify this
14430 option, GCC will produce code that is slower and larger but which
14431 uses extremely little TOC space. You may wish to use this option
14432 only on files that contain less frequently executed code.
14438 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
14439 @code{long} type, and the infrastructure needed to support them.
14440 Specifying @option{-maix64} implies @option{-mpowerpc64} and
14441 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
14442 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
14445 @itemx -mno-xl-compat
14446 @opindex mxl-compat
14447 @opindex mno-xl-compat
14448 Produce code that conforms more closely to IBM XL compiler semantics
14449 when using AIX-compatible ABI@. Pass floating-point arguments to
14450 prototyped functions beyond the register save area (RSA) on the stack
14451 in addition to argument FPRs. Do not assume that most significant
14452 double in 128-bit long double value is properly rounded when comparing
14453 values and converting to double. Use XL symbol names for long double
14456 The AIX calling convention was extended but not initially documented to
14457 handle an obscure K&R C case of calling a function that takes the
14458 address of its arguments with fewer arguments than declared. IBM XL
14459 compilers access floating point arguments which do not fit in the
14460 RSA from the stack when a subroutine is compiled without
14461 optimization. Because always storing floating-point arguments on the
14462 stack is inefficient and rarely needed, this option is not enabled by
14463 default and only is necessary when calling subroutines compiled by IBM
14464 XL compilers without optimization.
14468 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
14469 application written to use message passing with special startup code to
14470 enable the application to run. The system must have PE installed in the
14471 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
14472 must be overridden with the @option{-specs=} option to specify the
14473 appropriate directory location. The Parallel Environment does not
14474 support threads, so the @option{-mpe} option and the @option{-pthread}
14475 option are incompatible.
14477 @item -malign-natural
14478 @itemx -malign-power
14479 @opindex malign-natural
14480 @opindex malign-power
14481 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
14482 @option{-malign-natural} overrides the ABI-defined alignment of larger
14483 types, such as floating-point doubles, on their natural size-based boundary.
14484 The option @option{-malign-power} instructs GCC to follow the ABI-specified
14485 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
14487 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
14491 @itemx -mhard-float
14492 @opindex msoft-float
14493 @opindex mhard-float
14494 Generate code that does not use (uses) the floating-point register set.
14495 Software floating point emulation is provided if you use the
14496 @option{-msoft-float} option, and pass the option to GCC when linking.
14498 @item -msingle-float
14499 @itemx -mdouble-float
14500 @opindex msingle-float
14501 @opindex mdouble-float
14502 Generate code for single or double-precision floating point operations.
14503 @option{-mdouble-float} implies @option{-msingle-float}.
14506 @opindex msimple-fpu
14507 Do not generate sqrt and div instructions for hardware floating point unit.
14511 Specify type of floating point unit. Valid values are @var{sp_lite}
14512 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
14513 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
14514 and @var{dp_full} (equivalent to -mdouble-float).
14517 @opindex mxilinx-fpu
14518 Perform optimizations for floating point unit on Xilinx PPC 405/440.
14521 @itemx -mno-multiple
14523 @opindex mno-multiple
14524 Generate code that uses (does not use) the load multiple word
14525 instructions and the store multiple word instructions. These
14526 instructions are generated by default on POWER systems, and not
14527 generated on PowerPC systems. Do not use @option{-mmultiple} on little
14528 endian PowerPC systems, since those instructions do not work when the
14529 processor is in little endian mode. The exceptions are PPC740 and
14530 PPC750 which permit the instructions usage in little endian mode.
14535 @opindex mno-string
14536 Generate code that uses (does not use) the load string instructions
14537 and the store string word instructions to save multiple registers and
14538 do small block moves. These instructions are generated by default on
14539 POWER systems, and not generated on PowerPC systems. Do not use
14540 @option{-mstring} on little endian PowerPC systems, since those
14541 instructions do not work when the processor is in little endian mode.
14542 The exceptions are PPC740 and PPC750 which permit the instructions
14543 usage in little endian mode.
14548 @opindex mno-update
14549 Generate code that uses (does not use) the load or store instructions
14550 that update the base register to the address of the calculated memory
14551 location. These instructions are generated by default. If you use
14552 @option{-mno-update}, there is a small window between the time that the
14553 stack pointer is updated and the address of the previous frame is
14554 stored, which means code that walks the stack frame across interrupts or
14555 signals may get corrupted data.
14557 @item -mavoid-indexed-addresses
14558 @item -mno-avoid-indexed-addresses
14559 @opindex mavoid-indexed-addresses
14560 @opindex mno-avoid-indexed-addresses
14561 Generate code that tries to avoid (not avoid) the use of indexed load
14562 or store instructions. These instructions can incur a performance
14563 penalty on Power6 processors in certain situations, such as when
14564 stepping through large arrays that cross a 16M boundary. This option
14565 is enabled by default when targetting Power6 and disabled otherwise.
14568 @itemx -mno-fused-madd
14569 @opindex mfused-madd
14570 @opindex mno-fused-madd
14571 Generate code that uses (does not use) the floating point multiply and
14572 accumulate instructions. These instructions are generated by default if
14573 hardware floating is used.
14579 Generate code that uses (does not use) the half-word multiply and
14580 multiply-accumulate instructions on the IBM 405, 440 and 464 processors.
14581 These instructions are generated by default when targetting those
14588 Generate code that uses (does not use) the string-search @samp{dlmzb}
14589 instruction on the IBM 405, 440 and 464 processors. This instruction is
14590 generated by default when targetting those processors.
14592 @item -mno-bit-align
14594 @opindex mno-bit-align
14595 @opindex mbit-align
14596 On System V.4 and embedded PowerPC systems do not (do) force structures
14597 and unions that contain bit-fields to be aligned to the base type of the
14600 For example, by default a structure containing nothing but 8
14601 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
14602 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
14603 the structure would be aligned to a 1 byte boundary and be one byte in
14606 @item -mno-strict-align
14607 @itemx -mstrict-align
14608 @opindex mno-strict-align
14609 @opindex mstrict-align
14610 On System V.4 and embedded PowerPC systems do not (do) assume that
14611 unaligned memory references will be handled by the system.
14613 @item -mrelocatable
14614 @itemx -mno-relocatable
14615 @opindex mrelocatable
14616 @opindex mno-relocatable
14617 On embedded PowerPC systems generate code that allows (does not allow)
14618 the program to be relocated to a different address at runtime. If you
14619 use @option{-mrelocatable} on any module, all objects linked together must
14620 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
14622 @item -mrelocatable-lib
14623 @itemx -mno-relocatable-lib
14624 @opindex mrelocatable-lib
14625 @opindex mno-relocatable-lib
14626 On embedded PowerPC systems generate code that allows (does not allow)
14627 the program to be relocated to a different address at runtime. Modules
14628 compiled with @option{-mrelocatable-lib} can be linked with either modules
14629 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
14630 with modules compiled with the @option{-mrelocatable} options.
14636 On System V.4 and embedded PowerPC systems do not (do) assume that
14637 register 2 contains a pointer to a global area pointing to the addresses
14638 used in the program.
14641 @itemx -mlittle-endian
14643 @opindex mlittle-endian
14644 On System V.4 and embedded PowerPC systems compile code for the
14645 processor in little endian mode. The @option{-mlittle-endian} option is
14646 the same as @option{-mlittle}.
14649 @itemx -mbig-endian
14651 @opindex mbig-endian
14652 On System V.4 and embedded PowerPC systems compile code for the
14653 processor in big endian mode. The @option{-mbig-endian} option is
14654 the same as @option{-mbig}.
14656 @item -mdynamic-no-pic
14657 @opindex mdynamic-no-pic
14658 On Darwin and Mac OS X systems, compile code so that it is not
14659 relocatable, but that its external references are relocatable. The
14660 resulting code is suitable for applications, but not shared
14663 @item -mprioritize-restricted-insns=@var{priority}
14664 @opindex mprioritize-restricted-insns
14665 This option controls the priority that is assigned to
14666 dispatch-slot restricted instructions during the second scheduling
14667 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
14668 @var{no/highest/second-highest} priority to dispatch slot restricted
14671 @item -msched-costly-dep=@var{dependence_type}
14672 @opindex msched-costly-dep
14673 This option controls which dependences are considered costly
14674 by the target during instruction scheduling. The argument
14675 @var{dependence_type} takes one of the following values:
14676 @var{no}: no dependence is costly,
14677 @var{all}: all dependences are costly,
14678 @var{true_store_to_load}: a true dependence from store to load is costly,
14679 @var{store_to_load}: any dependence from store to load is costly,
14680 @var{number}: any dependence which latency >= @var{number} is costly.
14682 @item -minsert-sched-nops=@var{scheme}
14683 @opindex minsert-sched-nops
14684 This option controls which nop insertion scheme will be used during
14685 the second scheduling pass. The argument @var{scheme} takes one of the
14687 @var{no}: Don't insert nops.
14688 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
14689 according to the scheduler's grouping.
14690 @var{regroup_exact}: Insert nops to force costly dependent insns into
14691 separate groups. Insert exactly as many nops as needed to force an insn
14692 to a new group, according to the estimated processor grouping.
14693 @var{number}: Insert nops to force costly dependent insns into
14694 separate groups. Insert @var{number} nops to force an insn to a new group.
14697 @opindex mcall-sysv
14698 On System V.4 and embedded PowerPC systems compile code using calling
14699 conventions that adheres to the March 1995 draft of the System V
14700 Application Binary Interface, PowerPC processor supplement. This is the
14701 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
14703 @item -mcall-sysv-eabi
14705 @opindex mcall-sysv-eabi
14706 @opindex mcall-eabi
14707 Specify both @option{-mcall-sysv} and @option{-meabi} options.
14709 @item -mcall-sysv-noeabi
14710 @opindex mcall-sysv-noeabi
14711 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
14713 @item -mcall-aixdesc
14715 On System V.4 and embedded PowerPC systems compile code for the AIX
14719 @opindex mcall-linux
14720 On System V.4 and embedded PowerPC systems compile code for the
14721 Linux-based GNU system.
14725 On System V.4 and embedded PowerPC systems compile code for the
14726 Hurd-based GNU system.
14728 @item -mcall-freebsd
14729 @opindex mcall-freebsd
14730 On System V.4 and embedded PowerPC systems compile code for the
14731 FreeBSD operating system.
14733 @item -mcall-netbsd
14734 @opindex mcall-netbsd
14735 On System V.4 and embedded PowerPC systems compile code for the
14736 NetBSD operating system.
14738 @item -mcall-openbsd
14739 @opindex mcall-netbsd
14740 On System V.4 and embedded PowerPC systems compile code for the
14741 OpenBSD operating system.
14743 @item -maix-struct-return
14744 @opindex maix-struct-return
14745 Return all structures in memory (as specified by the AIX ABI)@.
14747 @item -msvr4-struct-return
14748 @opindex msvr4-struct-return
14749 Return structures smaller than 8 bytes in registers (as specified by the
14752 @item -mabi=@var{abi-type}
14754 Extend the current ABI with a particular extension, or remove such extension.
14755 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
14756 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
14760 Extend the current ABI with SPE ABI extensions. This does not change
14761 the default ABI, instead it adds the SPE ABI extensions to the current
14765 @opindex mabi=no-spe
14766 Disable Booke SPE ABI extensions for the current ABI@.
14768 @item -mabi=ibmlongdouble
14769 @opindex mabi=ibmlongdouble
14770 Change the current ABI to use IBM extended precision long double.
14771 This is a PowerPC 32-bit SYSV ABI option.
14773 @item -mabi=ieeelongdouble
14774 @opindex mabi=ieeelongdouble
14775 Change the current ABI to use IEEE extended precision long double.
14776 This is a PowerPC 32-bit Linux ABI option.
14779 @itemx -mno-prototype
14780 @opindex mprototype
14781 @opindex mno-prototype
14782 On System V.4 and embedded PowerPC systems assume that all calls to
14783 variable argument functions are properly prototyped. Otherwise, the
14784 compiler must insert an instruction before every non prototyped call to
14785 set or clear bit 6 of the condition code register (@var{CR}) to
14786 indicate whether floating point values were passed in the floating point
14787 registers in case the function takes a variable arguments. With
14788 @option{-mprototype}, only calls to prototyped variable argument functions
14789 will set or clear the bit.
14793 On embedded PowerPC systems, assume that the startup module is called
14794 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
14795 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
14800 On embedded PowerPC systems, assume that the startup module is called
14801 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
14806 On embedded PowerPC systems, assume that the startup module is called
14807 @file{crt0.o} and the standard C libraries are @file{libads.a} and
14810 @item -myellowknife
14811 @opindex myellowknife
14812 On embedded PowerPC systems, assume that the startup module is called
14813 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
14818 On System V.4 and embedded PowerPC systems, specify that you are
14819 compiling for a VxWorks system.
14823 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
14824 header to indicate that @samp{eabi} extended relocations are used.
14830 On System V.4 and embedded PowerPC systems do (do not) adhere to the
14831 Embedded Applications Binary Interface (eabi) which is a set of
14832 modifications to the System V.4 specifications. Selecting @option{-meabi}
14833 means that the stack is aligned to an 8 byte boundary, a function
14834 @code{__eabi} is called to from @code{main} to set up the eabi
14835 environment, and the @option{-msdata} option can use both @code{r2} and
14836 @code{r13} to point to two separate small data areas. Selecting
14837 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
14838 do not call an initialization function from @code{main}, and the
14839 @option{-msdata} option will only use @code{r13} to point to a single
14840 small data area. The @option{-meabi} option is on by default if you
14841 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
14844 @opindex msdata=eabi
14845 On System V.4 and embedded PowerPC systems, put small initialized
14846 @code{const} global and static data in the @samp{.sdata2} section, which
14847 is pointed to by register @code{r2}. Put small initialized
14848 non-@code{const} global and static data in the @samp{.sdata} section,
14849 which is pointed to by register @code{r13}. Put small uninitialized
14850 global and static data in the @samp{.sbss} section, which is adjacent to
14851 the @samp{.sdata} section. The @option{-msdata=eabi} option is
14852 incompatible with the @option{-mrelocatable} option. The
14853 @option{-msdata=eabi} option also sets the @option{-memb} option.
14856 @opindex msdata=sysv
14857 On System V.4 and embedded PowerPC systems, put small global and static
14858 data in the @samp{.sdata} section, which is pointed to by register
14859 @code{r13}. Put small uninitialized global and static data in the
14860 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
14861 The @option{-msdata=sysv} option is incompatible with the
14862 @option{-mrelocatable} option.
14864 @item -msdata=default
14866 @opindex msdata=default
14868 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
14869 compile code the same as @option{-msdata=eabi}, otherwise compile code the
14870 same as @option{-msdata=sysv}.
14873 @opindex msdata=data
14874 On System V.4 and embedded PowerPC systems, put small global
14875 data in the @samp{.sdata} section. Put small uninitialized global
14876 data in the @samp{.sbss} section. Do not use register @code{r13}
14877 to address small data however. This is the default behavior unless
14878 other @option{-msdata} options are used.
14882 @opindex msdata=none
14884 On embedded PowerPC systems, put all initialized global and static data
14885 in the @samp{.data} section, and all uninitialized data in the
14886 @samp{.bss} section.
14890 @cindex smaller data references (PowerPC)
14891 @cindex .sdata/.sdata2 references (PowerPC)
14892 On embedded PowerPC systems, put global and static items less than or
14893 equal to @var{num} bytes into the small data or bss sections instead of
14894 the normal data or bss section. By default, @var{num} is 8. The
14895 @option{-G @var{num}} switch is also passed to the linker.
14896 All modules should be compiled with the same @option{-G @var{num}} value.
14899 @itemx -mno-regnames
14901 @opindex mno-regnames
14902 On System V.4 and embedded PowerPC systems do (do not) emit register
14903 names in the assembly language output using symbolic forms.
14906 @itemx -mno-longcall
14908 @opindex mno-longcall
14909 By default assume that all calls are far away so that a longer more
14910 expensive calling sequence is required. This is required for calls
14911 further than 32 megabytes (33,554,432 bytes) from the current location.
14912 A short call will be generated if the compiler knows
14913 the call cannot be that far away. This setting can be overridden by
14914 the @code{shortcall} function attribute, or by @code{#pragma
14917 Some linkers are capable of detecting out-of-range calls and generating
14918 glue code on the fly. On these systems, long calls are unnecessary and
14919 generate slower code. As of this writing, the AIX linker can do this,
14920 as can the GNU linker for PowerPC/64. It is planned to add this feature
14921 to the GNU linker for 32-bit PowerPC systems as well.
14923 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
14924 callee, L42'', plus a ``branch island'' (glue code). The two target
14925 addresses represent the callee and the ``branch island''. The
14926 Darwin/PPC linker will prefer the first address and generate a ``bl
14927 callee'' if the PPC ``bl'' instruction will reach the callee directly;
14928 otherwise, the linker will generate ``bl L42'' to call the ``branch
14929 island''. The ``branch island'' is appended to the body of the
14930 calling function; it computes the full 32-bit address of the callee
14933 On Mach-O (Darwin) systems, this option directs the compiler emit to
14934 the glue for every direct call, and the Darwin linker decides whether
14935 to use or discard it.
14937 In the future, we may cause GCC to ignore all longcall specifications
14938 when the linker is known to generate glue.
14940 @item -mtls-markers
14941 @itemx -mno-tls-markers
14942 @opindex mtls-markers
14943 @opindex mno-tls-markers
14944 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
14945 specifying the function argument. The relocation allows ld to
14946 reliably associate function call with argument setup instructions for
14947 TLS optimization, which in turn allows gcc to better schedule the
14952 Adds support for multithreading with the @dfn{pthreads} library.
14953 This option sets flags for both the preprocessor and linker.
14957 @node S/390 and zSeries Options
14958 @subsection S/390 and zSeries Options
14959 @cindex S/390 and zSeries Options
14961 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
14965 @itemx -msoft-float
14966 @opindex mhard-float
14967 @opindex msoft-float
14968 Use (do not use) the hardware floating-point instructions and registers
14969 for floating-point operations. When @option{-msoft-float} is specified,
14970 functions in @file{libgcc.a} will be used to perform floating-point
14971 operations. When @option{-mhard-float} is specified, the compiler
14972 generates IEEE floating-point instructions. This is the default.
14975 @itemx -mno-hard-dfp
14977 @opindex mno-hard-dfp
14978 Use (do not use) the hardware decimal-floating-point instructions for
14979 decimal-floating-point operations. When @option{-mno-hard-dfp} is
14980 specified, functions in @file{libgcc.a} will be used to perform
14981 decimal-floating-point operations. When @option{-mhard-dfp} is
14982 specified, the compiler generates decimal-floating-point hardware
14983 instructions. This is the default for @option{-march=z9-ec} or higher.
14985 @item -mlong-double-64
14986 @itemx -mlong-double-128
14987 @opindex mlong-double-64
14988 @opindex mlong-double-128
14989 These switches control the size of @code{long double} type. A size
14990 of 64bit makes the @code{long double} type equivalent to the @code{double}
14991 type. This is the default.
14994 @itemx -mno-backchain
14995 @opindex mbackchain
14996 @opindex mno-backchain
14997 Store (do not store) the address of the caller's frame as backchain pointer
14998 into the callee's stack frame.
14999 A backchain may be needed to allow debugging using tools that do not understand
15000 DWARF-2 call frame information.
15001 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
15002 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
15003 the backchain is placed into the topmost word of the 96/160 byte register
15006 In general, code compiled with @option{-mbackchain} is call-compatible with
15007 code compiled with @option{-mmo-backchain}; however, use of the backchain
15008 for debugging purposes usually requires that the whole binary is built with
15009 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
15010 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
15011 to build a linux kernel use @option{-msoft-float}.
15013 The default is to not maintain the backchain.
15015 @item -mpacked-stack
15016 @itemx -mno-packed-stack
15017 @opindex mpacked-stack
15018 @opindex mno-packed-stack
15019 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
15020 specified, the compiler uses the all fields of the 96/160 byte register save
15021 area only for their default purpose; unused fields still take up stack space.
15022 When @option{-mpacked-stack} is specified, register save slots are densely
15023 packed at the top of the register save area; unused space is reused for other
15024 purposes, allowing for more efficient use of the available stack space.
15025 However, when @option{-mbackchain} is also in effect, the topmost word of
15026 the save area is always used to store the backchain, and the return address
15027 register is always saved two words below the backchain.
15029 As long as the stack frame backchain is not used, code generated with
15030 @option{-mpacked-stack} is call-compatible with code generated with
15031 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
15032 S/390 or zSeries generated code that uses the stack frame backchain at run
15033 time, not just for debugging purposes. Such code is not call-compatible
15034 with code compiled with @option{-mpacked-stack}. Also, note that the
15035 combination of @option{-mbackchain},
15036 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
15037 to build a linux kernel use @option{-msoft-float}.
15039 The default is to not use the packed stack layout.
15042 @itemx -mno-small-exec
15043 @opindex msmall-exec
15044 @opindex mno-small-exec
15045 Generate (or do not generate) code using the @code{bras} instruction
15046 to do subroutine calls.
15047 This only works reliably if the total executable size does not
15048 exceed 64k. The default is to use the @code{basr} instruction instead,
15049 which does not have this limitation.
15055 When @option{-m31} is specified, generate code compliant to the
15056 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
15057 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
15058 particular to generate 64-bit instructions. For the @samp{s390}
15059 targets, the default is @option{-m31}, while the @samp{s390x}
15060 targets default to @option{-m64}.
15066 When @option{-mzarch} is specified, generate code using the
15067 instructions available on z/Architecture.
15068 When @option{-mesa} is specified, generate code using the
15069 instructions available on ESA/390. Note that @option{-mesa} is
15070 not possible with @option{-m64}.
15071 When generating code compliant to the GNU/Linux for S/390 ABI,
15072 the default is @option{-mesa}. When generating code compliant
15073 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
15079 Generate (or do not generate) code using the @code{mvcle} instruction
15080 to perform block moves. When @option{-mno-mvcle} is specified,
15081 use a @code{mvc} loop instead. This is the default unless optimizing for
15088 Print (or do not print) additional debug information when compiling.
15089 The default is to not print debug information.
15091 @item -march=@var{cpu-type}
15093 Generate code that will run on @var{cpu-type}, which is the name of a system
15094 representing a certain processor type. Possible values for
15095 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
15096 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
15097 When generating code using the instructions available on z/Architecture,
15098 the default is @option{-march=z900}. Otherwise, the default is
15099 @option{-march=g5}.
15101 @item -mtune=@var{cpu-type}
15103 Tune to @var{cpu-type} everything applicable about the generated code,
15104 except for the ABI and the set of available instructions.
15105 The list of @var{cpu-type} values is the same as for @option{-march}.
15106 The default is the value used for @option{-march}.
15109 @itemx -mno-tpf-trace
15110 @opindex mtpf-trace
15111 @opindex mno-tpf-trace
15112 Generate code that adds (does not add) in TPF OS specific branches to trace
15113 routines in the operating system. This option is off by default, even
15114 when compiling for the TPF OS@.
15117 @itemx -mno-fused-madd
15118 @opindex mfused-madd
15119 @opindex mno-fused-madd
15120 Generate code that uses (does not use) the floating point multiply and
15121 accumulate instructions. These instructions are generated by default if
15122 hardware floating point is used.
15124 @item -mwarn-framesize=@var{framesize}
15125 @opindex mwarn-framesize
15126 Emit a warning if the current function exceeds the given frame size. Because
15127 this is a compile time check it doesn't need to be a real problem when the program
15128 runs. It is intended to identify functions which most probably cause
15129 a stack overflow. It is useful to be used in an environment with limited stack
15130 size e.g.@: the linux kernel.
15132 @item -mwarn-dynamicstack
15133 @opindex mwarn-dynamicstack
15134 Emit a warning if the function calls alloca or uses dynamically
15135 sized arrays. This is generally a bad idea with a limited stack size.
15137 @item -mstack-guard=@var{stack-guard}
15138 @itemx -mstack-size=@var{stack-size}
15139 @opindex mstack-guard
15140 @opindex mstack-size
15141 If these options are provided the s390 back end emits additional instructions in
15142 the function prologue which trigger a trap if the stack size is @var{stack-guard}
15143 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
15144 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
15145 the frame size of the compiled function is chosen.
15146 These options are intended to be used to help debugging stack overflow problems.
15147 The additionally emitted code causes only little overhead and hence can also be
15148 used in production like systems without greater performance degradation. The given
15149 values have to be exact powers of 2 and @var{stack-size} has to be greater than
15150 @var{stack-guard} without exceeding 64k.
15151 In order to be efficient the extra code makes the assumption that the stack starts
15152 at an address aligned to the value given by @var{stack-size}.
15153 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
15156 @node Score Options
15157 @subsection Score Options
15158 @cindex Score Options
15160 These options are defined for Score implementations:
15165 Compile code for big endian mode. This is the default.
15169 Compile code for little endian mode.
15173 Disable generate bcnz instruction.
15177 Enable generate unaligned load and store instruction.
15181 Enable the use of multiply-accumulate instructions. Disabled by default.
15185 Specify the SCORE5 as the target architecture.
15189 Specify the SCORE5U of the target architecture.
15193 Specify the SCORE7 as the target architecture. This is the default.
15197 Specify the SCORE7D as the target architecture.
15201 @subsection SH Options
15203 These @samp{-m} options are defined for the SH implementations:
15208 Generate code for the SH1.
15212 Generate code for the SH2.
15215 Generate code for the SH2e.
15219 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
15220 that the floating-point unit is not used.
15222 @item -m2a-single-only
15223 @opindex m2a-single-only
15224 Generate code for the SH2a-FPU, in such a way that no double-precision
15225 floating point operations are used.
15228 @opindex m2a-single
15229 Generate code for the SH2a-FPU assuming the floating-point unit is in
15230 single-precision mode by default.
15234 Generate code for the SH2a-FPU assuming the floating-point unit is in
15235 double-precision mode by default.
15239 Generate code for the SH3.
15243 Generate code for the SH3e.
15247 Generate code for the SH4 without a floating-point unit.
15249 @item -m4-single-only
15250 @opindex m4-single-only
15251 Generate code for the SH4 with a floating-point unit that only
15252 supports single-precision arithmetic.
15256 Generate code for the SH4 assuming the floating-point unit is in
15257 single-precision mode by default.
15261 Generate code for the SH4.
15265 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
15266 floating-point unit is not used.
15268 @item -m4a-single-only
15269 @opindex m4a-single-only
15270 Generate code for the SH4a, in such a way that no double-precision
15271 floating point operations are used.
15274 @opindex m4a-single
15275 Generate code for the SH4a assuming the floating-point unit is in
15276 single-precision mode by default.
15280 Generate code for the SH4a.
15284 Same as @option{-m4a-nofpu}, except that it implicitly passes
15285 @option{-dsp} to the assembler. GCC doesn't generate any DSP
15286 instructions at the moment.
15290 Compile code for the processor in big endian mode.
15294 Compile code for the processor in little endian mode.
15298 Align doubles at 64-bit boundaries. Note that this changes the calling
15299 conventions, and thus some functions from the standard C library will
15300 not work unless you recompile it first with @option{-mdalign}.
15304 Shorten some address references at link time, when possible; uses the
15305 linker option @option{-relax}.
15309 Use 32-bit offsets in @code{switch} tables. The default is to use
15314 Enable the use of bit manipulation instructions on SH2A.
15318 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
15319 alignment constraints.
15323 Comply with the calling conventions defined by Renesas.
15327 Comply with the calling conventions defined by Renesas.
15331 Comply with the calling conventions defined for GCC before the Renesas
15332 conventions were available. This option is the default for all
15333 targets of the SH toolchain except for @samp{sh-symbianelf}.
15336 @opindex mnomacsave
15337 Mark the @code{MAC} register as call-clobbered, even if
15338 @option{-mhitachi} is given.
15342 Increase IEEE-compliance of floating-point code.
15343 At the moment, this is equivalent to @option{-fno-finite-math-only}.
15344 When generating 16 bit SH opcodes, getting IEEE-conforming results for
15345 comparisons of NANs / infinities incurs extra overhead in every
15346 floating point comparison, therefore the default is set to
15347 @option{-ffinite-math-only}.
15349 @item -minline-ic_invalidate
15350 @opindex minline-ic_invalidate
15351 Inline code to invalidate instruction cache entries after setting up
15352 nested function trampolines.
15353 This option has no effect if -musermode is in effect and the selected
15354 code generation option (e.g. -m4) does not allow the use of the icbi
15356 If the selected code generation option does not allow the use of the icbi
15357 instruction, and -musermode is not in effect, the inlined code will
15358 manipulate the instruction cache address array directly with an associative
15359 write. This not only requires privileged mode, but it will also
15360 fail if the cache line had been mapped via the TLB and has become unmapped.
15364 Dump instruction size and location in the assembly code.
15367 @opindex mpadstruct
15368 This option is deprecated. It pads structures to multiple of 4 bytes,
15369 which is incompatible with the SH ABI@.
15373 Optimize for space instead of speed. Implied by @option{-Os}.
15376 @opindex mprefergot
15377 When generating position-independent code, emit function calls using
15378 the Global Offset Table instead of the Procedure Linkage Table.
15382 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
15383 if the inlined code would not work in user mode.
15384 This is the default when the target is @code{sh-*-linux*}.
15386 @item -multcost=@var{number}
15387 @opindex multcost=@var{number}
15388 Set the cost to assume for a multiply insn.
15390 @item -mdiv=@var{strategy}
15391 @opindex mdiv=@var{strategy}
15392 Set the division strategy to use for SHmedia code. @var{strategy} must be
15393 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
15394 inv:call2, inv:fp .
15395 "fp" performs the operation in floating point. This has a very high latency,
15396 but needs only a few instructions, so it might be a good choice if
15397 your code has enough easily exploitable ILP to allow the compiler to
15398 schedule the floating point instructions together with other instructions.
15399 Division by zero causes a floating point exception.
15400 "inv" uses integer operations to calculate the inverse of the divisor,
15401 and then multiplies the dividend with the inverse. This strategy allows
15402 cse and hoisting of the inverse calculation. Division by zero calculates
15403 an unspecified result, but does not trap.
15404 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
15405 have been found, or if the entire operation has been hoisted to the same
15406 place, the last stages of the inverse calculation are intertwined with the
15407 final multiply to reduce the overall latency, at the expense of using a few
15408 more instructions, and thus offering fewer scheduling opportunities with
15410 "call" calls a library function that usually implements the inv:minlat
15412 This gives high code density for m5-*media-nofpu compilations.
15413 "call2" uses a different entry point of the same library function, where it
15414 assumes that a pointer to a lookup table has already been set up, which
15415 exposes the pointer load to cse / code hoisting optimizations.
15416 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
15417 code generation, but if the code stays unoptimized, revert to the "call",
15418 "call2", or "fp" strategies, respectively. Note that the
15419 potentially-trapping side effect of division by zero is carried by a
15420 separate instruction, so it is possible that all the integer instructions
15421 are hoisted out, but the marker for the side effect stays where it is.
15422 A recombination to fp operations or a call is not possible in that case.
15423 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
15424 that the inverse calculation was nor separated from the multiply, they speed
15425 up division where the dividend fits into 20 bits (plus sign where applicable),
15426 by inserting a test to skip a number of operations in this case; this test
15427 slows down the case of larger dividends. inv20u assumes the case of a such
15428 a small dividend to be unlikely, and inv20l assumes it to be likely.
15430 @item -mdivsi3_libfunc=@var{name}
15431 @opindex mdivsi3_libfunc=@var{name}
15432 Set the name of the library function used for 32 bit signed division to
15433 @var{name}. This only affect the name used in the call and inv:call
15434 division strategies, and the compiler will still expect the same
15435 sets of input/output/clobbered registers as if this option was not present.
15437 @item -mfixed-range=@var{register-range}
15438 @opindex mfixed-range
15439 Generate code treating the given register range as fixed registers.
15440 A fixed register is one that the register allocator can not use. This is
15441 useful when compiling kernel code. A register range is specified as
15442 two registers separated by a dash. Multiple register ranges can be
15443 specified separated by a comma.
15445 @item -madjust-unroll
15446 @opindex madjust-unroll
15447 Throttle unrolling to avoid thrashing target registers.
15448 This option only has an effect if the gcc code base supports the
15449 TARGET_ADJUST_UNROLL_MAX target hook.
15451 @item -mindexed-addressing
15452 @opindex mindexed-addressing
15453 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
15454 This is only safe if the hardware and/or OS implement 32 bit wrap-around
15455 semantics for the indexed addressing mode. The architecture allows the
15456 implementation of processors with 64 bit MMU, which the OS could use to
15457 get 32 bit addressing, but since no current hardware implementation supports
15458 this or any other way to make the indexed addressing mode safe to use in
15459 the 32 bit ABI, the default is -mno-indexed-addressing.
15461 @item -mgettrcost=@var{number}
15462 @opindex mgettrcost=@var{number}
15463 Set the cost assumed for the gettr instruction to @var{number}.
15464 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
15468 Assume pt* instructions won't trap. This will generally generate better
15469 scheduled code, but is unsafe on current hardware. The current architecture
15470 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
15471 This has the unintentional effect of making it unsafe to schedule ptabs /
15472 ptrel before a branch, or hoist it out of a loop. For example,
15473 __do_global_ctors, a part of libgcc that runs constructors at program
15474 startup, calls functions in a list which is delimited by @minus{}1. With the
15475 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
15476 That means that all the constructors will be run a bit quicker, but when
15477 the loop comes to the end of the list, the program crashes because ptabs
15478 loads @minus{}1 into a target register. Since this option is unsafe for any
15479 hardware implementing the current architecture specification, the default
15480 is -mno-pt-fixed. Unless the user specifies a specific cost with
15481 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
15482 this deters register allocation using target registers for storing
15485 @item -minvalid-symbols
15486 @opindex minvalid-symbols
15487 Assume symbols might be invalid. Ordinary function symbols generated by
15488 the compiler will always be valid to load with movi/shori/ptabs or
15489 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
15490 to generate symbols that will cause ptabs / ptrel to trap.
15491 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
15492 It will then prevent cross-basic-block cse, hoisting and most scheduling
15493 of symbol loads. The default is @option{-mno-invalid-symbols}.
15496 @node SPARC Options
15497 @subsection SPARC Options
15498 @cindex SPARC options
15500 These @samp{-m} options are supported on the SPARC:
15503 @item -mno-app-regs
15505 @opindex mno-app-regs
15507 Specify @option{-mapp-regs} to generate output using the global registers
15508 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
15511 To be fully SVR4 ABI compliant at the cost of some performance loss,
15512 specify @option{-mno-app-regs}. You should compile libraries and system
15513 software with this option.
15516 @itemx -mhard-float
15518 @opindex mhard-float
15519 Generate output containing floating point instructions. This is the
15523 @itemx -msoft-float
15525 @opindex msoft-float
15526 Generate output containing library calls for floating point.
15527 @strong{Warning:} the requisite libraries are not available for all SPARC
15528 targets. Normally the facilities of the machine's usual C compiler are
15529 used, but this cannot be done directly in cross-compilation. You must make
15530 your own arrangements to provide suitable library functions for
15531 cross-compilation. The embedded targets @samp{sparc-*-aout} and
15532 @samp{sparclite-*-*} do provide software floating point support.
15534 @option{-msoft-float} changes the calling convention in the output file;
15535 therefore, it is only useful if you compile @emph{all} of a program with
15536 this option. In particular, you need to compile @file{libgcc.a}, the
15537 library that comes with GCC, with @option{-msoft-float} in order for
15540 @item -mhard-quad-float
15541 @opindex mhard-quad-float
15542 Generate output containing quad-word (long double) floating point
15545 @item -msoft-quad-float
15546 @opindex msoft-quad-float
15547 Generate output containing library calls for quad-word (long double)
15548 floating point instructions. The functions called are those specified
15549 in the SPARC ABI@. This is the default.
15551 As of this writing, there are no SPARC implementations that have hardware
15552 support for the quad-word floating point instructions. They all invoke
15553 a trap handler for one of these instructions, and then the trap handler
15554 emulates the effect of the instruction. Because of the trap handler overhead,
15555 this is much slower than calling the ABI library routines. Thus the
15556 @option{-msoft-quad-float} option is the default.
15558 @item -mno-unaligned-doubles
15559 @itemx -munaligned-doubles
15560 @opindex mno-unaligned-doubles
15561 @opindex munaligned-doubles
15562 Assume that doubles have 8 byte alignment. This is the default.
15564 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
15565 alignment only if they are contained in another type, or if they have an
15566 absolute address. Otherwise, it assumes they have 4 byte alignment.
15567 Specifying this option avoids some rare compatibility problems with code
15568 generated by other compilers. It is not the default because it results
15569 in a performance loss, especially for floating point code.
15571 @item -mno-faster-structs
15572 @itemx -mfaster-structs
15573 @opindex mno-faster-structs
15574 @opindex mfaster-structs
15575 With @option{-mfaster-structs}, the compiler assumes that structures
15576 should have 8 byte alignment. This enables the use of pairs of
15577 @code{ldd} and @code{std} instructions for copies in structure
15578 assignment, in place of twice as many @code{ld} and @code{st} pairs.
15579 However, the use of this changed alignment directly violates the SPARC
15580 ABI@. Thus, it's intended only for use on targets where the developer
15581 acknowledges that their resulting code will not be directly in line with
15582 the rules of the ABI@.
15584 @item -mimpure-text
15585 @opindex mimpure-text
15586 @option{-mimpure-text}, used in addition to @option{-shared}, tells
15587 the compiler to not pass @option{-z text} to the linker when linking a
15588 shared object. Using this option, you can link position-dependent
15589 code into a shared object.
15591 @option{-mimpure-text} suppresses the ``relocations remain against
15592 allocatable but non-writable sections'' linker error message.
15593 However, the necessary relocations will trigger copy-on-write, and the
15594 shared object is not actually shared across processes. Instead of
15595 using @option{-mimpure-text}, you should compile all source code with
15596 @option{-fpic} or @option{-fPIC}.
15598 This option is only available on SunOS and Solaris.
15600 @item -mcpu=@var{cpu_type}
15602 Set the instruction set, register set, and instruction scheduling parameters
15603 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
15604 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
15605 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
15606 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
15607 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
15609 Default instruction scheduling parameters are used for values that select
15610 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
15611 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
15613 Here is a list of each supported architecture and their supported
15618 v8: supersparc, hypersparc
15619 sparclite: f930, f934, sparclite86x
15621 v9: ultrasparc, ultrasparc3, niagara, niagara2
15624 By default (unless configured otherwise), GCC generates code for the V7
15625 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
15626 additionally optimizes it for the Cypress CY7C602 chip, as used in the
15627 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
15628 SPARCStation 1, 2, IPX etc.
15630 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
15631 architecture. The only difference from V7 code is that the compiler emits
15632 the integer multiply and integer divide instructions which exist in SPARC-V8
15633 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
15634 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
15637 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
15638 the SPARC architecture. This adds the integer multiply, integer divide step
15639 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
15640 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
15641 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
15642 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
15643 MB86934 chip, which is the more recent SPARClite with FPU@.
15645 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
15646 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
15647 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
15648 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
15649 optimizes it for the TEMIC SPARClet chip.
15651 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
15652 architecture. This adds 64-bit integer and floating-point move instructions,
15653 3 additional floating-point condition code registers and conditional move
15654 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
15655 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
15656 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
15657 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
15658 @option{-mcpu=niagara}, the compiler additionally optimizes it for
15659 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
15660 additionally optimizes it for Sun UltraSPARC T2 chips.
15662 @item -mtune=@var{cpu_type}
15664 Set the instruction scheduling parameters for machine type
15665 @var{cpu_type}, but do not set the instruction set or register set that the
15666 option @option{-mcpu=@var{cpu_type}} would.
15668 The same values for @option{-mcpu=@var{cpu_type}} can be used for
15669 @option{-mtune=@var{cpu_type}}, but the only useful values are those
15670 that select a particular cpu implementation. Those are @samp{cypress},
15671 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
15672 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
15673 @samp{ultrasparc3}, @samp{niagara}, and @samp{niagara2}.
15678 @opindex mno-v8plus
15679 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
15680 difference from the V8 ABI is that the global and out registers are
15681 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
15682 mode for all SPARC-V9 processors.
15688 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
15689 Visual Instruction Set extensions. The default is @option{-mno-vis}.
15692 These @samp{-m} options are supported in addition to the above
15693 on SPARC-V9 processors in 64-bit environments:
15696 @item -mlittle-endian
15697 @opindex mlittle-endian
15698 Generate code for a processor running in little-endian mode. It is only
15699 available for a few configurations and most notably not on Solaris and Linux.
15705 Generate code for a 32-bit or 64-bit environment.
15706 The 32-bit environment sets int, long and pointer to 32 bits.
15707 The 64-bit environment sets int to 32 bits and long and pointer
15710 @item -mcmodel=medlow
15711 @opindex mcmodel=medlow
15712 Generate code for the Medium/Low code model: 64-bit addresses, programs
15713 must be linked in the low 32 bits of memory. Programs can be statically
15714 or dynamically linked.
15716 @item -mcmodel=medmid
15717 @opindex mcmodel=medmid
15718 Generate code for the Medium/Middle code model: 64-bit addresses, programs
15719 must be linked in the low 44 bits of memory, the text and data segments must
15720 be less than 2GB in size and the data segment must be located within 2GB of
15723 @item -mcmodel=medany
15724 @opindex mcmodel=medany
15725 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
15726 may be linked anywhere in memory, the text and data segments must be less
15727 than 2GB in size and the data segment must be located within 2GB of the
15730 @item -mcmodel=embmedany
15731 @opindex mcmodel=embmedany
15732 Generate code for the Medium/Anywhere code model for embedded systems:
15733 64-bit addresses, the text and data segments must be less than 2GB in
15734 size, both starting anywhere in memory (determined at link time). The
15735 global register %g4 points to the base of the data segment. Programs
15736 are statically linked and PIC is not supported.
15739 @itemx -mno-stack-bias
15740 @opindex mstack-bias
15741 @opindex mno-stack-bias
15742 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
15743 frame pointer if present, are offset by @minus{}2047 which must be added back
15744 when making stack frame references. This is the default in 64-bit mode.
15745 Otherwise, assume no such offset is present.
15748 These switches are supported in addition to the above on Solaris:
15753 Add support for multithreading using the Solaris threads library. This
15754 option sets flags for both the preprocessor and linker. This option does
15755 not affect the thread safety of object code produced by the compiler or
15756 that of libraries supplied with it.
15760 Add support for multithreading using the POSIX threads library. This
15761 option sets flags for both the preprocessor and linker. This option does
15762 not affect the thread safety of object code produced by the compiler or
15763 that of libraries supplied with it.
15767 This is a synonym for @option{-pthreads}.
15771 @subsection SPU Options
15772 @cindex SPU options
15774 These @samp{-m} options are supported on the SPU:
15778 @itemx -merror-reloc
15779 @opindex mwarn-reloc
15780 @opindex merror-reloc
15782 The loader for SPU does not handle dynamic relocations. By default, GCC
15783 will give an error when it generates code that requires a dynamic
15784 relocation. @option{-mno-error-reloc} disables the error,
15785 @option{-mwarn-reloc} will generate a warning instead.
15788 @itemx -munsafe-dma
15790 @opindex munsafe-dma
15792 Instructions which initiate or test completion of DMA must not be
15793 reordered with respect to loads and stores of the memory which is being
15794 accessed. Users typically address this problem using the volatile
15795 keyword, but that can lead to inefficient code in places where the
15796 memory is known to not change. Rather than mark the memory as volatile
15797 we treat the DMA instructions as potentially effecting all memory. With
15798 @option{-munsafe-dma} users must use the volatile keyword to protect
15801 @item -mbranch-hints
15802 @opindex mbranch-hints
15804 By default, GCC will generate a branch hint instruction to avoid
15805 pipeline stalls for always taken or probably taken branches. A hint
15806 will not be generated closer than 8 instructions away from its branch.
15807 There is little reason to disable them, except for debugging purposes,
15808 or to make an object a little bit smaller.
15812 @opindex msmall-mem
15813 @opindex mlarge-mem
15815 By default, GCC generates code assuming that addresses are never larger
15816 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
15817 a full 32 bit address.
15822 By default, GCC links against startup code that assumes the SPU-style
15823 main function interface (which has an unconventional parameter list).
15824 With @option{-mstdmain}, GCC will link your program against startup
15825 code that assumes a C99-style interface to @code{main}, including a
15826 local copy of @code{argv} strings.
15828 @item -mfixed-range=@var{register-range}
15829 @opindex mfixed-range
15830 Generate code treating the given register range as fixed registers.
15831 A fixed register is one that the register allocator can not use. This is
15832 useful when compiling kernel code. A register range is specified as
15833 two registers separated by a dash. Multiple register ranges can be
15834 specified separated by a comma.
15837 @itemx -mdual-nops=@var{n}
15838 @opindex mdual-nops
15839 By default, GCC will insert nops to increase dual issue when it expects
15840 it to increase performance. @var{n} can be a value from 0 to 10. A
15841 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
15842 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
15844 @item -mhint-max-nops=@var{n}
15845 @opindex mhint-max-nops
15846 Maximum number of nops to insert for a branch hint. A branch hint must
15847 be at least 8 instructions away from the branch it is effecting. GCC
15848 will insert up to @var{n} nops to enforce this, otherwise it will not
15849 generate the branch hint.
15851 @item -mhint-max-distance=@var{n}
15852 @opindex mhint-max-distance
15853 The encoding of the branch hint instruction limits the hint to be within
15854 256 instructions of the branch it is effecting. By default, GCC makes
15855 sure it is within 125.
15858 @opindex msafe-hints
15859 Work around a hardware bug which causes the SPU to stall indefinitely.
15860 By default, GCC will insert the @code{hbrp} instruction to make sure
15861 this stall won't happen.
15865 @node System V Options
15866 @subsection Options for System V
15868 These additional options are available on System V Release 4 for
15869 compatibility with other compilers on those systems:
15874 Create a shared object.
15875 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
15879 Identify the versions of each tool used by the compiler, in a
15880 @code{.ident} assembler directive in the output.
15884 Refrain from adding @code{.ident} directives to the output file (this is
15887 @item -YP,@var{dirs}
15889 Search the directories @var{dirs}, and no others, for libraries
15890 specified with @option{-l}.
15892 @item -Ym,@var{dir}
15894 Look in the directory @var{dir} to find the M4 preprocessor.
15895 The assembler uses this option.
15896 @c This is supposed to go with a -Yd for predefined M4 macro files, but
15897 @c the generic assembler that comes with Solaris takes just -Ym.
15901 @subsection V850 Options
15902 @cindex V850 Options
15904 These @samp{-m} options are defined for V850 implementations:
15908 @itemx -mno-long-calls
15909 @opindex mlong-calls
15910 @opindex mno-long-calls
15911 Treat all calls as being far away (near). If calls are assumed to be
15912 far away, the compiler will always load the functions address up into a
15913 register, and call indirect through the pointer.
15919 Do not optimize (do optimize) basic blocks that use the same index
15920 pointer 4 or more times to copy pointer into the @code{ep} register, and
15921 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
15922 option is on by default if you optimize.
15924 @item -mno-prolog-function
15925 @itemx -mprolog-function
15926 @opindex mno-prolog-function
15927 @opindex mprolog-function
15928 Do not use (do use) external functions to save and restore registers
15929 at the prologue and epilogue of a function. The external functions
15930 are slower, but use less code space if more than one function saves
15931 the same number of registers. The @option{-mprolog-function} option
15932 is on by default if you optimize.
15936 Try to make the code as small as possible. At present, this just turns
15937 on the @option{-mep} and @option{-mprolog-function} options.
15939 @item -mtda=@var{n}
15941 Put static or global variables whose size is @var{n} bytes or less into
15942 the tiny data area that register @code{ep} points to. The tiny data
15943 area can hold up to 256 bytes in total (128 bytes for byte references).
15945 @item -msda=@var{n}
15947 Put static or global variables whose size is @var{n} bytes or less into
15948 the small data area that register @code{gp} points to. The small data
15949 area can hold up to 64 kilobytes.
15951 @item -mzda=@var{n}
15953 Put static or global variables whose size is @var{n} bytes or less into
15954 the first 32 kilobytes of memory.
15958 Specify that the target processor is the V850.
15961 @opindex mbig-switch
15962 Generate code suitable for big switch tables. Use this option only if
15963 the assembler/linker complain about out of range branches within a switch
15968 This option will cause r2 and r5 to be used in the code generated by
15969 the compiler. This setting is the default.
15971 @item -mno-app-regs
15972 @opindex mno-app-regs
15973 This option will cause r2 and r5 to be treated as fixed registers.
15977 Specify that the target processor is the V850E1. The preprocessor
15978 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
15979 this option is used.
15983 Specify that the target processor is the V850E@. The preprocessor
15984 constant @samp{__v850e__} will be defined if this option is used.
15986 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
15987 are defined then a default target processor will be chosen and the
15988 relevant @samp{__v850*__} preprocessor constant will be defined.
15990 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
15991 defined, regardless of which processor variant is the target.
15993 @item -mdisable-callt
15994 @opindex mdisable-callt
15995 This option will suppress generation of the CALLT instruction for the
15996 v850e and v850e1 flavors of the v850 architecture. The default is
15997 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
16002 @subsection VAX Options
16003 @cindex VAX options
16005 These @samp{-m} options are defined for the VAX:
16010 Do not output certain jump instructions (@code{aobleq} and so on)
16011 that the Unix assembler for the VAX cannot handle across long
16016 Do output those jump instructions, on the assumption that you
16017 will assemble with the GNU assembler.
16021 Output code for g-format floating point numbers instead of d-format.
16024 @node VxWorks Options
16025 @subsection VxWorks Options
16026 @cindex VxWorks Options
16028 The options in this section are defined for all VxWorks targets.
16029 Options specific to the target hardware are listed with the other
16030 options for that target.
16035 GCC can generate code for both VxWorks kernels and real time processes
16036 (RTPs). This option switches from the former to the latter. It also
16037 defines the preprocessor macro @code{__RTP__}.
16040 @opindex non-static
16041 Link an RTP executable against shared libraries rather than static
16042 libraries. The options @option{-static} and @option{-shared} can
16043 also be used for RTPs (@pxref{Link Options}); @option{-static}
16050 These options are passed down to the linker. They are defined for
16051 compatibility with Diab.
16054 @opindex Xbind-lazy
16055 Enable lazy binding of function calls. This option is equivalent to
16056 @option{-Wl,-z,now} and is defined for compatibility with Diab.
16060 Disable lazy binding of function calls. This option is the default and
16061 is defined for compatibility with Diab.
16064 @node x86-64 Options
16065 @subsection x86-64 Options
16066 @cindex x86-64 options
16068 These are listed under @xref{i386 and x86-64 Options}.
16070 @node i386 and x86-64 Windows Options
16071 @subsection i386 and x86-64 Windows Options
16072 @cindex i386 and x86-64 Windows Options
16074 These additional options are available for Windows targets:
16079 This option is available for Cygwin and MinGW targets. It
16080 specifies that a console application is to be generated, by
16081 instructing the linker to set the PE header subsystem type
16082 required for console applications.
16083 This is the default behavior for Cygwin and MinGW targets.
16087 This option is available for Cygwin targets. It specifies that
16088 the Cygwin internal interface is to be used for predefined
16089 preprocessor macros, C runtime libraries and related linker
16090 paths and options. For Cygwin targets this is the default behavior.
16091 This option is deprecated and will be removed in a future release.
16094 @opindex mno-cygwin
16095 This option is available for Cygwin targets. It specifies that
16096 the MinGW internal interface is to be used instead of Cygwin's, by
16097 setting MinGW-related predefined macros and linker paths and default
16099 This option is deprecated and will be removed in a future release.
16103 This option is available for Cygwin and MinGW targets. It
16104 specifies that a DLL - a dynamic link library - is to be
16105 generated, enabling the selection of the required runtime
16106 startup object and entry point.
16108 @item -mnop-fun-dllimport
16109 @opindex mnop-fun-dllimport
16110 This option is available for Cygwin and MinGW targets. It
16111 specifies that the dllimport attribute should be ignored.
16115 This option is available for MinGW targets. It specifies
16116 that MinGW-specific thread support is to be used.
16120 This option is available for mingw-w64 targets. It specifies
16121 that the UNICODE macro is getting pre-defined and that the
16122 unicode capable runtime startup code is choosen.
16126 This option is available for Cygwin and MinGW targets. It
16127 specifies that the typical Windows pre-defined macros are to
16128 be set in the pre-processor, but does not influence the choice
16129 of runtime library/startup code.
16133 This option is available for Cygwin and MinGW targets. It
16134 specifies that a GUI application is to be generated by
16135 instructing the linker to set the PE header subsystem type
16138 @item -mpe-aligned-commons
16139 @opindex mpe-aligned-commons
16140 This option is available for Cygwin and MinGW targets. It
16141 specifies that the GNU extension to the PE file format that
16142 permits the correct alignment of COMMON variables should be
16143 used when generating code. It will be enabled by default if
16144 GCC detects that the target assembler found during configuration
16145 supports the feature.
16148 See also under @ref{i386 and x86-64 Options} for standard options.
16150 @node Xstormy16 Options
16151 @subsection Xstormy16 Options
16152 @cindex Xstormy16 Options
16154 These options are defined for Xstormy16:
16159 Choose startup files and linker script suitable for the simulator.
16162 @node Xtensa Options
16163 @subsection Xtensa Options
16164 @cindex Xtensa Options
16166 These options are supported for Xtensa targets:
16170 @itemx -mno-const16
16172 @opindex mno-const16
16173 Enable or disable use of @code{CONST16} instructions for loading
16174 constant values. The @code{CONST16} instruction is currently not a
16175 standard option from Tensilica. When enabled, @code{CONST16}
16176 instructions are always used in place of the standard @code{L32R}
16177 instructions. The use of @code{CONST16} is enabled by default only if
16178 the @code{L32R} instruction is not available.
16181 @itemx -mno-fused-madd
16182 @opindex mfused-madd
16183 @opindex mno-fused-madd
16184 Enable or disable use of fused multiply/add and multiply/subtract
16185 instructions in the floating-point option. This has no effect if the
16186 floating-point option is not also enabled. Disabling fused multiply/add
16187 and multiply/subtract instructions forces the compiler to use separate
16188 instructions for the multiply and add/subtract operations. This may be
16189 desirable in some cases where strict IEEE 754-compliant results are
16190 required: the fused multiply add/subtract instructions do not round the
16191 intermediate result, thereby producing results with @emph{more} bits of
16192 precision than specified by the IEEE standard. Disabling fused multiply
16193 add/subtract instructions also ensures that the program output is not
16194 sensitive to the compiler's ability to combine multiply and add/subtract
16197 @item -mserialize-volatile
16198 @itemx -mno-serialize-volatile
16199 @opindex mserialize-volatile
16200 @opindex mno-serialize-volatile
16201 When this option is enabled, GCC inserts @code{MEMW} instructions before
16202 @code{volatile} memory references to guarantee sequential consistency.
16203 The default is @option{-mserialize-volatile}. Use
16204 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
16206 @item -mtext-section-literals
16207 @itemx -mno-text-section-literals
16208 @opindex mtext-section-literals
16209 @opindex mno-text-section-literals
16210 Control the treatment of literal pools. The default is
16211 @option{-mno-text-section-literals}, which places literals in a separate
16212 section in the output file. This allows the literal pool to be placed
16213 in a data RAM/ROM, and it also allows the linker to combine literal
16214 pools from separate object files to remove redundant literals and
16215 improve code size. With @option{-mtext-section-literals}, the literals
16216 are interspersed in the text section in order to keep them as close as
16217 possible to their references. This may be necessary for large assembly
16220 @item -mtarget-align
16221 @itemx -mno-target-align
16222 @opindex mtarget-align
16223 @opindex mno-target-align
16224 When this option is enabled, GCC instructs the assembler to
16225 automatically align instructions to reduce branch penalties at the
16226 expense of some code density. The assembler attempts to widen density
16227 instructions to align branch targets and the instructions following call
16228 instructions. If there are not enough preceding safe density
16229 instructions to align a target, no widening will be performed. The
16230 default is @option{-mtarget-align}. These options do not affect the
16231 treatment of auto-aligned instructions like @code{LOOP}, which the
16232 assembler will always align, either by widening density instructions or
16233 by inserting no-op instructions.
16236 @itemx -mno-longcalls
16237 @opindex mlongcalls
16238 @opindex mno-longcalls
16239 When this option is enabled, GCC instructs the assembler to translate
16240 direct calls to indirect calls unless it can determine that the target
16241 of a direct call is in the range allowed by the call instruction. This
16242 translation typically occurs for calls to functions in other source
16243 files. Specifically, the assembler translates a direct @code{CALL}
16244 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
16245 The default is @option{-mno-longcalls}. This option should be used in
16246 programs where the call target can potentially be out of range. This
16247 option is implemented in the assembler, not the compiler, so the
16248 assembly code generated by GCC will still show direct call
16249 instructions---look at the disassembled object code to see the actual
16250 instructions. Note that the assembler will use an indirect call for
16251 every cross-file call, not just those that really will be out of range.
16254 @node zSeries Options
16255 @subsection zSeries Options
16256 @cindex zSeries options
16258 These are listed under @xref{S/390 and zSeries Options}.
16260 @node Code Gen Options
16261 @section Options for Code Generation Conventions
16262 @cindex code generation conventions
16263 @cindex options, code generation
16264 @cindex run-time options
16266 These machine-independent options control the interface conventions
16267 used in code generation.
16269 Most of them have both positive and negative forms; the negative form
16270 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
16271 one of the forms is listed---the one which is not the default. You
16272 can figure out the other form by either removing @samp{no-} or adding
16276 @item -fbounds-check
16277 @opindex fbounds-check
16278 For front-ends that support it, generate additional code to check that
16279 indices used to access arrays are within the declared range. This is
16280 currently only supported by the Java and Fortran front-ends, where
16281 this option defaults to true and false respectively.
16285 This option generates traps for signed overflow on addition, subtraction,
16286 multiplication operations.
16290 This option instructs the compiler to assume that signed arithmetic
16291 overflow of addition, subtraction and multiplication wraps around
16292 using twos-complement representation. This flag enables some optimizations
16293 and disables others. This option is enabled by default for the Java
16294 front-end, as required by the Java language specification.
16297 @opindex fexceptions
16298 Enable exception handling. Generates extra code needed to propagate
16299 exceptions. For some targets, this implies GCC will generate frame
16300 unwind information for all functions, which can produce significant data
16301 size overhead, although it does not affect execution. If you do not
16302 specify this option, GCC will enable it by default for languages like
16303 C++ which normally require exception handling, and disable it for
16304 languages like C that do not normally require it. However, you may need
16305 to enable this option when compiling C code that needs to interoperate
16306 properly with exception handlers written in C++. You may also wish to
16307 disable this option if you are compiling older C++ programs that don't
16308 use exception handling.
16310 @item -fnon-call-exceptions
16311 @opindex fnon-call-exceptions
16312 Generate code that allows trapping instructions to throw exceptions.
16313 Note that this requires platform-specific runtime support that does
16314 not exist everywhere. Moreover, it only allows @emph{trapping}
16315 instructions to throw exceptions, i.e.@: memory references or floating
16316 point instructions. It does not allow exceptions to be thrown from
16317 arbitrary signal handlers such as @code{SIGALRM}.
16319 @item -funwind-tables
16320 @opindex funwind-tables
16321 Similar to @option{-fexceptions}, except that it will just generate any needed
16322 static data, but will not affect the generated code in any other way.
16323 You will normally not enable this option; instead, a language processor
16324 that needs this handling would enable it on your behalf.
16326 @item -fasynchronous-unwind-tables
16327 @opindex fasynchronous-unwind-tables
16328 Generate unwind table in dwarf2 format, if supported by target machine. The
16329 table is exact at each instruction boundary, so it can be used for stack
16330 unwinding from asynchronous events (such as debugger or garbage collector).
16332 @item -fpcc-struct-return
16333 @opindex fpcc-struct-return
16334 Return ``short'' @code{struct} and @code{union} values in memory like
16335 longer ones, rather than in registers. This convention is less
16336 efficient, but it has the advantage of allowing intercallability between
16337 GCC-compiled files and files compiled with other compilers, particularly
16338 the Portable C Compiler (pcc).
16340 The precise convention for returning structures in memory depends
16341 on the target configuration macros.
16343 Short structures and unions are those whose size and alignment match
16344 that of some integer type.
16346 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
16347 switch is not binary compatible with code compiled with the
16348 @option{-freg-struct-return} switch.
16349 Use it to conform to a non-default application binary interface.
16351 @item -freg-struct-return
16352 @opindex freg-struct-return
16353 Return @code{struct} and @code{union} values in registers when possible.
16354 This is more efficient for small structures than
16355 @option{-fpcc-struct-return}.
16357 If you specify neither @option{-fpcc-struct-return} nor
16358 @option{-freg-struct-return}, GCC defaults to whichever convention is
16359 standard for the target. If there is no standard convention, GCC
16360 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
16361 the principal compiler. In those cases, we can choose the standard, and
16362 we chose the more efficient register return alternative.
16364 @strong{Warning:} code compiled with the @option{-freg-struct-return}
16365 switch is not binary compatible with code compiled with the
16366 @option{-fpcc-struct-return} switch.
16367 Use it to conform to a non-default application binary interface.
16369 @item -fshort-enums
16370 @opindex fshort-enums
16371 Allocate to an @code{enum} type only as many bytes as it needs for the
16372 declared range of possible values. Specifically, the @code{enum} type
16373 will be equivalent to the smallest integer type which has enough room.
16375 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
16376 code that is not binary compatible with code generated without that switch.
16377 Use it to conform to a non-default application binary interface.
16379 @item -fshort-double
16380 @opindex fshort-double
16381 Use the same size for @code{double} as for @code{float}.
16383 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
16384 code that is not binary compatible with code generated without that switch.
16385 Use it to conform to a non-default application binary interface.
16387 @item -fshort-wchar
16388 @opindex fshort-wchar
16389 Override the underlying type for @samp{wchar_t} to be @samp{short
16390 unsigned int} instead of the default for the target. This option is
16391 useful for building programs to run under WINE@.
16393 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
16394 code that is not binary compatible with code generated without that switch.
16395 Use it to conform to a non-default application binary interface.
16398 @opindex fno-common
16399 In C code, controls the placement of uninitialized global variables.
16400 Unix C compilers have traditionally permitted multiple definitions of
16401 such variables in different compilation units by placing the variables
16403 This is the behavior specified by @option{-fcommon}, and is the default
16404 for GCC on most targets.
16405 On the other hand, this behavior is not required by ISO C, and on some
16406 targets may carry a speed or code size penalty on variable references.
16407 The @option{-fno-common} option specifies that the compiler should place
16408 uninitialized global variables in the data section of the object file,
16409 rather than generating them as common blocks.
16410 This has the effect that if the same variable is declared
16411 (without @code{extern}) in two different compilations,
16412 you will get a multiple-definition error when you link them.
16413 In this case, you must compile with @option{-fcommon} instead.
16414 Compiling with @option{-fno-common} is useful on targets for which
16415 it provides better performance, or if you wish to verify that the
16416 program will work on other systems which always treat uninitialized
16417 variable declarations this way.
16421 Ignore the @samp{#ident} directive.
16423 @item -finhibit-size-directive
16424 @opindex finhibit-size-directive
16425 Don't output a @code{.size} assembler directive, or anything else that
16426 would cause trouble if the function is split in the middle, and the
16427 two halves are placed at locations far apart in memory. This option is
16428 used when compiling @file{crtstuff.c}; you should not need to use it
16431 @item -fverbose-asm
16432 @opindex fverbose-asm
16433 Put extra commentary information in the generated assembly code to
16434 make it more readable. This option is generally only of use to those
16435 who actually need to read the generated assembly code (perhaps while
16436 debugging the compiler itself).
16438 @option{-fno-verbose-asm}, the default, causes the
16439 extra information to be omitted and is useful when comparing two assembler
16442 @item -frecord-gcc-switches
16443 @opindex frecord-gcc-switches
16444 This switch causes the command line that was used to invoke the
16445 compiler to be recorded into the object file that is being created.
16446 This switch is only implemented on some targets and the exact format
16447 of the recording is target and binary file format dependent, but it
16448 usually takes the form of a section containing ASCII text. This
16449 switch is related to the @option{-fverbose-asm} switch, but that
16450 switch only records information in the assembler output file as
16451 comments, so it never reaches the object file.
16455 @cindex global offset table
16457 Generate position-independent code (PIC) suitable for use in a shared
16458 library, if supported for the target machine. Such code accesses all
16459 constant addresses through a global offset table (GOT)@. The dynamic
16460 loader resolves the GOT entries when the program starts (the dynamic
16461 loader is not part of GCC; it is part of the operating system). If
16462 the GOT size for the linked executable exceeds a machine-specific
16463 maximum size, you get an error message from the linker indicating that
16464 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
16465 instead. (These maximums are 8k on the SPARC and 32k
16466 on the m68k and RS/6000. The 386 has no such limit.)
16468 Position-independent code requires special support, and therefore works
16469 only on certain machines. For the 386, GCC supports PIC for System V
16470 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
16471 position-independent.
16473 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
16478 If supported for the target machine, emit position-independent code,
16479 suitable for dynamic linking and avoiding any limit on the size of the
16480 global offset table. This option makes a difference on the m68k,
16481 PowerPC and SPARC@.
16483 Position-independent code requires special support, and therefore works
16484 only on certain machines.
16486 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
16493 These options are similar to @option{-fpic} and @option{-fPIC}, but
16494 generated position independent code can be only linked into executables.
16495 Usually these options are used when @option{-pie} GCC option will be
16496 used during linking.
16498 @option{-fpie} and @option{-fPIE} both define the macros
16499 @code{__pie__} and @code{__PIE__}. The macros have the value 1
16500 for @option{-fpie} and 2 for @option{-fPIE}.
16502 @item -fno-jump-tables
16503 @opindex fno-jump-tables
16504 Do not use jump tables for switch statements even where it would be
16505 more efficient than other code generation strategies. This option is
16506 of use in conjunction with @option{-fpic} or @option{-fPIC} for
16507 building code which forms part of a dynamic linker and cannot
16508 reference the address of a jump table. On some targets, jump tables
16509 do not require a GOT and this option is not needed.
16511 @item -ffixed-@var{reg}
16513 Treat the register named @var{reg} as a fixed register; generated code
16514 should never refer to it (except perhaps as a stack pointer, frame
16515 pointer or in some other fixed role).
16517 @var{reg} must be the name of a register. The register names accepted
16518 are machine-specific and are defined in the @code{REGISTER_NAMES}
16519 macro in the machine description macro file.
16521 This flag does not have a negative form, because it specifies a
16524 @item -fcall-used-@var{reg}
16525 @opindex fcall-used
16526 Treat the register named @var{reg} as an allocable register that is
16527 clobbered by function calls. It may be allocated for temporaries or
16528 variables that do not live across a call. Functions compiled this way
16529 will not save and restore the register @var{reg}.
16531 It is an error to used this flag with the frame pointer or stack pointer.
16532 Use of this flag for other registers that have fixed pervasive roles in
16533 the machine's execution model will produce disastrous results.
16535 This flag does not have a negative form, because it specifies a
16538 @item -fcall-saved-@var{reg}
16539 @opindex fcall-saved
16540 Treat the register named @var{reg} as an allocable register saved by
16541 functions. It may be allocated even for temporaries or variables that
16542 live across a call. Functions compiled this way will save and restore
16543 the register @var{reg} if they use it.
16545 It is an error to used this flag with the frame pointer or stack pointer.
16546 Use of this flag for other registers that have fixed pervasive roles in
16547 the machine's execution model will produce disastrous results.
16549 A different sort of disaster will result from the use of this flag for
16550 a register in which function values may be returned.
16552 This flag does not have a negative form, because it specifies a
16555 @item -fpack-struct[=@var{n}]
16556 @opindex fpack-struct
16557 Without a value specified, pack all structure members together without
16558 holes. When a value is specified (which must be a small power of two), pack
16559 structure members according to this value, representing the maximum
16560 alignment (that is, objects with default alignment requirements larger than
16561 this will be output potentially unaligned at the next fitting location.
16563 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
16564 code that is not binary compatible with code generated without that switch.
16565 Additionally, it makes the code suboptimal.
16566 Use it to conform to a non-default application binary interface.
16568 @item -finstrument-functions
16569 @opindex finstrument-functions
16570 Generate instrumentation calls for entry and exit to functions. Just
16571 after function entry and just before function exit, the following
16572 profiling functions will be called with the address of the current
16573 function and its call site. (On some platforms,
16574 @code{__builtin_return_address} does not work beyond the current
16575 function, so the call site information may not be available to the
16576 profiling functions otherwise.)
16579 void __cyg_profile_func_enter (void *this_fn,
16581 void __cyg_profile_func_exit (void *this_fn,
16585 The first argument is the address of the start of the current function,
16586 which may be looked up exactly in the symbol table.
16588 This instrumentation is also done for functions expanded inline in other
16589 functions. The profiling calls will indicate where, conceptually, the
16590 inline function is entered and exited. This means that addressable
16591 versions of such functions must be available. If all your uses of a
16592 function are expanded inline, this may mean an additional expansion of
16593 code size. If you use @samp{extern inline} in your C code, an
16594 addressable version of such functions must be provided. (This is
16595 normally the case anyways, but if you get lucky and the optimizer always
16596 expands the functions inline, you might have gotten away without
16597 providing static copies.)
16599 A function may be given the attribute @code{no_instrument_function}, in
16600 which case this instrumentation will not be done. This can be used, for
16601 example, for the profiling functions listed above, high-priority
16602 interrupt routines, and any functions from which the profiling functions
16603 cannot safely be called (perhaps signal handlers, if the profiling
16604 routines generate output or allocate memory).
16606 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
16607 @opindex finstrument-functions-exclude-file-list
16609 Set the list of functions that are excluded from instrumentation (see
16610 the description of @code{-finstrument-functions}). If the file that
16611 contains a function definition matches with one of @var{file}, then
16612 that function is not instrumented. The match is done on substrings:
16613 if the @var{file} parameter is a substring of the file name, it is
16614 considered to be a match.
16617 @code{-finstrument-functions-exclude-file-list=/bits/stl,include/sys}
16618 will exclude any inline function defined in files whose pathnames
16619 contain @code{/bits/stl} or @code{include/sys}.
16621 If, for some reason, you want to include letter @code{','} in one of
16622 @var{sym}, write @code{'\,'}. For example,
16623 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
16624 (note the single quote surrounding the option).
16626 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
16627 @opindex finstrument-functions-exclude-function-list
16629 This is similar to @code{-finstrument-functions-exclude-file-list},
16630 but this option sets the list of function names to be excluded from
16631 instrumentation. The function name to be matched is its user-visible
16632 name, such as @code{vector<int> blah(const vector<int> &)}, not the
16633 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
16634 match is done on substrings: if the @var{sym} parameter is a substring
16635 of the function name, it is considered to be a match. For C99 and C++
16636 extended identifiers, the function name must be given in UTF-8, not
16637 using universal character names.
16639 @item -fstack-check
16640 @opindex fstack-check
16641 Generate code to verify that you do not go beyond the boundary of the
16642 stack. You should specify this flag if you are running in an
16643 environment with multiple threads, but only rarely need to specify it in
16644 a single-threaded environment since stack overflow is automatically
16645 detected on nearly all systems if there is only one stack.
16647 Note that this switch does not actually cause checking to be done; the
16648 operating system or the language runtime must do that. The switch causes
16649 generation of code to ensure that they see the stack being extended.
16651 You can additionally specify a string parameter: @code{no} means no
16652 checking, @code{generic} means force the use of old-style checking,
16653 @code{specific} means use the best checking method and is equivalent
16654 to bare @option{-fstack-check}.
16656 Old-style checking is a generic mechanism that requires no specific
16657 target support in the compiler but comes with the following drawbacks:
16661 Modified allocation strategy for large objects: they will always be
16662 allocated dynamically if their size exceeds a fixed threshold.
16665 Fixed limit on the size of the static frame of functions: when it is
16666 topped by a particular function, stack checking is not reliable and
16667 a warning is issued by the compiler.
16670 Inefficiency: because of both the modified allocation strategy and the
16671 generic implementation, the performances of the code are hampered.
16674 Note that old-style stack checking is also the fallback method for
16675 @code{specific} if no target support has been added in the compiler.
16677 @item -fstack-limit-register=@var{reg}
16678 @itemx -fstack-limit-symbol=@var{sym}
16679 @itemx -fno-stack-limit
16680 @opindex fstack-limit-register
16681 @opindex fstack-limit-symbol
16682 @opindex fno-stack-limit
16683 Generate code to ensure that the stack does not grow beyond a certain value,
16684 either the value of a register or the address of a symbol. If the stack
16685 would grow beyond the value, a signal is raised. For most targets,
16686 the signal is raised before the stack overruns the boundary, so
16687 it is possible to catch the signal without taking special precautions.
16689 For instance, if the stack starts at absolute address @samp{0x80000000}
16690 and grows downwards, you can use the flags
16691 @option{-fstack-limit-symbol=__stack_limit} and
16692 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
16693 of 128KB@. Note that this may only work with the GNU linker.
16695 @cindex aliasing of parameters
16696 @cindex parameters, aliased
16697 @item -fargument-alias
16698 @itemx -fargument-noalias
16699 @itemx -fargument-noalias-global
16700 @itemx -fargument-noalias-anything
16701 @opindex fargument-alias
16702 @opindex fargument-noalias
16703 @opindex fargument-noalias-global
16704 @opindex fargument-noalias-anything
16705 Specify the possible relationships among parameters and between
16706 parameters and global data.
16708 @option{-fargument-alias} specifies that arguments (parameters) may
16709 alias each other and may alias global storage.@*
16710 @option{-fargument-noalias} specifies that arguments do not alias
16711 each other, but may alias global storage.@*
16712 @option{-fargument-noalias-global} specifies that arguments do not
16713 alias each other and do not alias global storage.
16714 @option{-fargument-noalias-anything} specifies that arguments do not
16715 alias any other storage.
16717 Each language will automatically use whatever option is required by
16718 the language standard. You should not need to use these options yourself.
16720 @item -fleading-underscore
16721 @opindex fleading-underscore
16722 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
16723 change the way C symbols are represented in the object file. One use
16724 is to help link with legacy assembly code.
16726 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
16727 generate code that is not binary compatible with code generated without that
16728 switch. Use it to conform to a non-default application binary interface.
16729 Not all targets provide complete support for this switch.
16731 @item -ftls-model=@var{model}
16732 @opindex ftls-model
16733 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
16734 The @var{model} argument should be one of @code{global-dynamic},
16735 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
16737 The default without @option{-fpic} is @code{initial-exec}; with
16738 @option{-fpic} the default is @code{global-dynamic}.
16740 @item -fvisibility=@var{default|internal|hidden|protected}
16741 @opindex fvisibility
16742 Set the default ELF image symbol visibility to the specified option---all
16743 symbols will be marked with this unless overridden within the code.
16744 Using this feature can very substantially improve linking and
16745 load times of shared object libraries, produce more optimized
16746 code, provide near-perfect API export and prevent symbol clashes.
16747 It is @strong{strongly} recommended that you use this in any shared objects
16750 Despite the nomenclature, @code{default} always means public ie;
16751 available to be linked against from outside the shared object.
16752 @code{protected} and @code{internal} are pretty useless in real-world
16753 usage so the only other commonly used option will be @code{hidden}.
16754 The default if @option{-fvisibility} isn't specified is
16755 @code{default}, i.e., make every
16756 symbol public---this causes the same behavior as previous versions of
16759 A good explanation of the benefits offered by ensuring ELF
16760 symbols have the correct visibility is given by ``How To Write
16761 Shared Libraries'' by Ulrich Drepper (which can be found at
16762 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
16763 solution made possible by this option to marking things hidden when
16764 the default is public is to make the default hidden and mark things
16765 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
16766 and @code{__attribute__ ((visibility("default")))} instead of
16767 @code{__declspec(dllexport)} you get almost identical semantics with
16768 identical syntax. This is a great boon to those working with
16769 cross-platform projects.
16771 For those adding visibility support to existing code, you may find
16772 @samp{#pragma GCC visibility} of use. This works by you enclosing
16773 the declarations you wish to set visibility for with (for example)
16774 @samp{#pragma GCC visibility push(hidden)} and
16775 @samp{#pragma GCC visibility pop}.
16776 Bear in mind that symbol visibility should be viewed @strong{as
16777 part of the API interface contract} and thus all new code should
16778 always specify visibility when it is not the default ie; declarations
16779 only for use within the local DSO should @strong{always} be marked explicitly
16780 as hidden as so to avoid PLT indirection overheads---making this
16781 abundantly clear also aids readability and self-documentation of the code.
16782 Note that due to ISO C++ specification requirements, operator new and
16783 operator delete must always be of default visibility.
16785 Be aware that headers from outside your project, in particular system
16786 headers and headers from any other library you use, may not be
16787 expecting to be compiled with visibility other than the default. You
16788 may need to explicitly say @samp{#pragma GCC visibility push(default)}
16789 before including any such headers.
16791 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
16792 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
16793 no modifications. However, this means that calls to @samp{extern}
16794 functions with no explicit visibility will use the PLT, so it is more
16795 effective to use @samp{__attribute ((visibility))} and/or
16796 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
16797 declarations should be treated as hidden.
16799 Note that @samp{-fvisibility} does affect C++ vague linkage
16800 entities. This means that, for instance, an exception class that will
16801 be thrown between DSOs must be explicitly marked with default
16802 visibility so that the @samp{type_info} nodes will be unified between
16805 An overview of these techniques, their benefits and how to use them
16806 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
16812 @node Environment Variables
16813 @section Environment Variables Affecting GCC
16814 @cindex environment variables
16816 @c man begin ENVIRONMENT
16817 This section describes several environment variables that affect how GCC
16818 operates. Some of them work by specifying directories or prefixes to use
16819 when searching for various kinds of files. Some are used to specify other
16820 aspects of the compilation environment.
16822 Note that you can also specify places to search using options such as
16823 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
16824 take precedence over places specified using environment variables, which
16825 in turn take precedence over those specified by the configuration of GCC@.
16826 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
16827 GNU Compiler Collection (GCC) Internals}.
16832 @c @itemx LC_COLLATE
16834 @c @itemx LC_MONETARY
16835 @c @itemx LC_NUMERIC
16840 @c @findex LC_COLLATE
16841 @findex LC_MESSAGES
16842 @c @findex LC_MONETARY
16843 @c @findex LC_NUMERIC
16847 These environment variables control the way that GCC uses
16848 localization information that allow GCC to work with different
16849 national conventions. GCC inspects the locale categories
16850 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
16851 so. These locale categories can be set to any value supported by your
16852 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
16853 Kingdom encoded in UTF-8.
16855 The @env{LC_CTYPE} environment variable specifies character
16856 classification. GCC uses it to determine the character boundaries in
16857 a string; this is needed for some multibyte encodings that contain quote
16858 and escape characters that would otherwise be interpreted as a string
16861 The @env{LC_MESSAGES} environment variable specifies the language to
16862 use in diagnostic messages.
16864 If the @env{LC_ALL} environment variable is set, it overrides the value
16865 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
16866 and @env{LC_MESSAGES} default to the value of the @env{LANG}
16867 environment variable. If none of these variables are set, GCC
16868 defaults to traditional C English behavior.
16872 If @env{TMPDIR} is set, it specifies the directory to use for temporary
16873 files. GCC uses temporary files to hold the output of one stage of
16874 compilation which is to be used as input to the next stage: for example,
16875 the output of the preprocessor, which is the input to the compiler
16878 @item GCC_EXEC_PREFIX
16879 @findex GCC_EXEC_PREFIX
16880 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
16881 names of the subprograms executed by the compiler. No slash is added
16882 when this prefix is combined with the name of a subprogram, but you can
16883 specify a prefix that ends with a slash if you wish.
16885 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
16886 an appropriate prefix to use based on the pathname it was invoked with.
16888 If GCC cannot find the subprogram using the specified prefix, it
16889 tries looking in the usual places for the subprogram.
16891 The default value of @env{GCC_EXEC_PREFIX} is
16892 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
16893 the installed compiler. In many cases @var{prefix} is the value
16894 of @code{prefix} when you ran the @file{configure} script.
16896 Other prefixes specified with @option{-B} take precedence over this prefix.
16898 This prefix is also used for finding files such as @file{crt0.o} that are
16901 In addition, the prefix is used in an unusual way in finding the
16902 directories to search for header files. For each of the standard
16903 directories whose name normally begins with @samp{/usr/local/lib/gcc}
16904 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
16905 replacing that beginning with the specified prefix to produce an
16906 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
16907 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
16908 These alternate directories are searched first; the standard directories
16909 come next. If a standard directory begins with the configured
16910 @var{prefix} then the value of @var{prefix} is replaced by
16911 @env{GCC_EXEC_PREFIX} when looking for header files.
16913 @item COMPILER_PATH
16914 @findex COMPILER_PATH
16915 The value of @env{COMPILER_PATH} is a colon-separated list of
16916 directories, much like @env{PATH}. GCC tries the directories thus
16917 specified when searching for subprograms, if it can't find the
16918 subprograms using @env{GCC_EXEC_PREFIX}.
16921 @findex LIBRARY_PATH
16922 The value of @env{LIBRARY_PATH} is a colon-separated list of
16923 directories, much like @env{PATH}. When configured as a native compiler,
16924 GCC tries the directories thus specified when searching for special
16925 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
16926 using GCC also uses these directories when searching for ordinary
16927 libraries for the @option{-l} option (but directories specified with
16928 @option{-L} come first).
16932 @cindex locale definition
16933 This variable is used to pass locale information to the compiler. One way in
16934 which this information is used is to determine the character set to be used
16935 when character literals, string literals and comments are parsed in C and C++.
16936 When the compiler is configured to allow multibyte characters,
16937 the following values for @env{LANG} are recognized:
16941 Recognize JIS characters.
16943 Recognize SJIS characters.
16945 Recognize EUCJP characters.
16948 If @env{LANG} is not defined, or if it has some other value, then the
16949 compiler will use mblen and mbtowc as defined by the default locale to
16950 recognize and translate multibyte characters.
16954 Some additional environments variables affect the behavior of the
16957 @include cppenv.texi
16961 @node Precompiled Headers
16962 @section Using Precompiled Headers
16963 @cindex precompiled headers
16964 @cindex speed of compilation
16966 Often large projects have many header files that are included in every
16967 source file. The time the compiler takes to process these header files
16968 over and over again can account for nearly all of the time required to
16969 build the project. To make builds faster, GCC allows users to
16970 `precompile' a header file; then, if builds can use the precompiled
16971 header file they will be much faster.
16973 To create a precompiled header file, simply compile it as you would any
16974 other file, if necessary using the @option{-x} option to make the driver
16975 treat it as a C or C++ header file. You will probably want to use a
16976 tool like @command{make} to keep the precompiled header up-to-date when
16977 the headers it contains change.
16979 A precompiled header file will be searched for when @code{#include} is
16980 seen in the compilation. As it searches for the included file
16981 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
16982 compiler looks for a precompiled header in each directory just before it
16983 looks for the include file in that directory. The name searched for is
16984 the name specified in the @code{#include} with @samp{.gch} appended. If
16985 the precompiled header file can't be used, it is ignored.
16987 For instance, if you have @code{#include "all.h"}, and you have
16988 @file{all.h.gch} in the same directory as @file{all.h}, then the
16989 precompiled header file will be used if possible, and the original
16990 header will be used otherwise.
16992 Alternatively, you might decide to put the precompiled header file in a
16993 directory and use @option{-I} to ensure that directory is searched
16994 before (or instead of) the directory containing the original header.
16995 Then, if you want to check that the precompiled header file is always
16996 used, you can put a file of the same name as the original header in this
16997 directory containing an @code{#error} command.
16999 This also works with @option{-include}. So yet another way to use
17000 precompiled headers, good for projects not designed with precompiled
17001 header files in mind, is to simply take most of the header files used by
17002 a project, include them from another header file, precompile that header
17003 file, and @option{-include} the precompiled header. If the header files
17004 have guards against multiple inclusion, they will be skipped because
17005 they've already been included (in the precompiled header).
17007 If you need to precompile the same header file for different
17008 languages, targets, or compiler options, you can instead make a
17009 @emph{directory} named like @file{all.h.gch}, and put each precompiled
17010 header in the directory, perhaps using @option{-o}. It doesn't matter
17011 what you call the files in the directory, every precompiled header in
17012 the directory will be considered. The first precompiled header
17013 encountered in the directory that is valid for this compilation will
17014 be used; they're searched in no particular order.
17016 There are many other possibilities, limited only by your imagination,
17017 good sense, and the constraints of your build system.
17019 A precompiled header file can be used only when these conditions apply:
17023 Only one precompiled header can be used in a particular compilation.
17026 A precompiled header can't be used once the first C token is seen. You
17027 can have preprocessor directives before a precompiled header; you can
17028 even include a precompiled header from inside another header, so long as
17029 there are no C tokens before the @code{#include}.
17032 The precompiled header file must be produced for the same language as
17033 the current compilation. You can't use a C precompiled header for a C++
17037 The precompiled header file must have been produced by the same compiler
17038 binary as the current compilation is using.
17041 Any macros defined before the precompiled header is included must
17042 either be defined in the same way as when the precompiled header was
17043 generated, or must not affect the precompiled header, which usually
17044 means that they don't appear in the precompiled header at all.
17046 The @option{-D} option is one way to define a macro before a
17047 precompiled header is included; using a @code{#define} can also do it.
17048 There are also some options that define macros implicitly, like
17049 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
17052 @item If debugging information is output when using the precompiled
17053 header, using @option{-g} or similar, the same kind of debugging information
17054 must have been output when building the precompiled header. However,
17055 a precompiled header built using @option{-g} can be used in a compilation
17056 when no debugging information is being output.
17058 @item The same @option{-m} options must generally be used when building
17059 and using the precompiled header. @xref{Submodel Options},
17060 for any cases where this rule is relaxed.
17062 @item Each of the following options must be the same when building and using
17063 the precompiled header:
17065 @gccoptlist{-fexceptions}
17068 Some other command-line options starting with @option{-f},
17069 @option{-p}, or @option{-O} must be defined in the same way as when
17070 the precompiled header was generated. At present, it's not clear
17071 which options are safe to change and which are not; the safest choice
17072 is to use exactly the same options when generating and using the
17073 precompiled header. The following are known to be safe:
17075 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
17076 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
17077 -fsched-verbose=<number> -fschedule-insns -fvisibility= @gol
17082 For all of these except the last, the compiler will automatically
17083 ignore the precompiled header if the conditions aren't met. If you
17084 find an option combination that doesn't work and doesn't cause the
17085 precompiled header to be ignored, please consider filing a bug report,
17088 If you do use differing options when generating and using the
17089 precompiled header, the actual behavior will be a mixture of the
17090 behavior for the options. For instance, if you use @option{-g} to
17091 generate the precompiled header but not when using it, you may or may
17092 not get debugging information for routines in the precompiled header.