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, 2010
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, 2010
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 1.3 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} @gol
168 -fdump-ada-spec@r{[}-slim@r{]}}
170 @item C Language Options
171 @xref{C Dialect Options,,Options Controlling C Dialect}.
172 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
173 -aux-info @var{filename} @gol
174 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
175 -fhosted -ffreestanding -fopenmp -fms-extensions @gol
176 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
177 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
178 -fsigned-bitfields -fsigned-char @gol
179 -funsigned-bitfields -funsigned-char}
181 @item C++ Language Options
182 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
183 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
184 -fconserve-space -ffriend-injection @gol
185 -fno-elide-constructors @gol
186 -fno-enforce-eh-specs @gol
187 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
188 -fno-implicit-templates @gol
189 -fno-implicit-inline-templates @gol
190 -fno-implement-inlines -fms-extensions @gol
191 -fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol
192 -fno-optional-diags -fpermissive @gol
193 -fno-pretty-templates @gol
194 -frepo -fno-rtti -fstats -ftemplate-depth=@var{n} @gol
195 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
196 -fno-default-inline -fvisibility-inlines-hidden @gol
197 -fvisibility-ms-compat @gol
198 -Wabi -Wconversion-null -Wctor-dtor-privacy @gol
199 -Wnoexcept -Wnon-virtual-dtor -Wreorder @gol
200 -Weffc++ -Wstrict-null-sentinel @gol
201 -Wno-non-template-friend -Wold-style-cast @gol
202 -Woverloaded-virtual -Wno-pmf-conversions @gol
205 @item Objective-C and Objective-C++ Language Options
206 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
207 Objective-C and Objective-C++ Dialects}.
208 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
209 -fgnu-runtime -fnext-runtime @gol
210 -fno-nil-receivers @gol
211 -fobjc-call-cxx-cdtors @gol
212 -fobjc-direct-dispatch @gol
213 -fobjc-exceptions @gol
215 -freplace-objc-classes @gol
218 -Wassign-intercept @gol
219 -Wno-protocol -Wselector @gol
220 -Wstrict-selector-match @gol
221 -Wundeclared-selector}
223 @item Language Independent Options
224 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
225 @gccoptlist{-fmessage-length=@var{n} @gol
226 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
227 -fdiagnostics-show-option}
229 @item Warning Options
230 @xref{Warning Options,,Options to Request or Suppress Warnings}.
231 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
232 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
233 -Wno-attributes -Wno-builtin-macro-redefined @gol
234 -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual @gol
235 -Wchar-subscripts -Wclobbered -Wcomment @gol
236 -Wconversion -Wcoverage-mismatch -Wcpp -Wno-deprecated @gol
237 -Wno-deprecated-declarations -Wdisabled-optimization @gol
238 -Wno-div-by-zero -Wempty-body -Wenum-compare -Wno-endif-labels @gol
239 -Werror -Werror=* @gol
240 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
241 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
242 -Wformat-security -Wformat-y2k @gol
243 -Wframe-larger-than=@var{len} -Wjump-misses-init -Wignored-qualifiers @gol
244 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
245 -Winit-self -Winline @gol
246 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
247 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
248 -Wlogical-op -Wlong-long @gol
249 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
250 -Wmissing-format-attribute -Wmissing-include-dirs @gol
252 -Wno-multichar -Wnonnull -Wno-overflow @gol
253 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
254 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
255 -Wpointer-arith -Wno-pointer-to-int-cast @gol
256 -Wredundant-decls @gol
257 -Wreturn-type -Wsequence-point -Wshadow @gol
258 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
259 -Wstrict-aliasing -Wstrict-aliasing=n @gol
260 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
261 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]} @gol
262 -Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand @gol
263 -Wsystem-headers -Wtrampolines -Wtrigraphs -Wtype-limits -Wundef @gol
264 -Wuninitialized -Wunknown-pragmas -Wno-pragmas @gol
265 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
266 -Wunused-label -Wunused-parameter -Wno-unused-result -Wunused-value -Wunused-variable @gol
267 -Wunused-but-set-parameter -Wunused-but-set-variable -Wvariadic-macros -Wvla @gol
268 -Wvolatile-register-var -Wwrite-strings}
270 @item C and Objective-C-only Warning Options
271 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
272 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
273 -Wold-style-declaration -Wold-style-definition @gol
274 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
275 -Wdeclaration-after-statement -Wpointer-sign}
277 @item Debugging Options
278 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
279 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
280 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
281 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
282 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
283 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
284 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
285 -fdump-statistics @gol
287 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
288 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
289 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
291 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
292 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
293 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
294 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
295 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
296 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
297 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
298 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
299 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
300 -fdump-tree-nrv -fdump-tree-vect @gol
301 -fdump-tree-sink @gol
302 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
303 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
304 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
305 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
306 -ftree-vectorizer-verbose=@var{n} @gol
307 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
308 -fdump-final-insns=@var{file} @gol
309 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
310 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
311 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
312 -fenable-icf-debug @gol
313 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
314 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
315 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
316 -ftest-coverage -ftime-report -fvar-tracking @gol
317 -fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol
318 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
319 -ggdb -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
320 -gvms -gxcoff -gxcoff+ @gol
321 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
322 -fdebug-prefix-map=@var{old}=@var{new} @gol
323 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
324 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
325 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
326 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
327 -print-prog-name=@var{program} -print-search-dirs -Q @gol
328 -print-sysroot -print-sysroot-headers-suffix @gol
329 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
331 @item Optimization Options
332 @xref{Optimize Options,,Options that Control Optimization}.
334 -falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
335 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
336 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
337 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
338 -fcheck-data-deps -fconserve-stack -fcprop-registers -fcrossjumping @gol
339 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules -fcx-limited-range @gol
340 -fdata-sections -fdce -fdce @gol
341 -fdelayed-branch -fdelete-null-pointer-checks -fdse -fdse @gol
342 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffast-math @gol
343 -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
344 -fforward-propagate -ffunction-sections @gol
345 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
346 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
347 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
348 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg -fipa-pta @gol
349 -fipa-profile -fipa-pure-const -fipa-reference -fipa-struct-reorg @gol
350 -fira-algorithm=@var{algorithm} @gol
351 -fira-region=@var{region} -fira-coalesce @gol
352 -fira-loop-pressure -fno-ira-share-save-slots @gol
353 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
354 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
355 -floop-block -floop-interchange -floop-strip-mine @gol
356 -floop-parallelize-all -flto -flto-compression-level -flto-report -fltrans @gol
357 -fltrans-output-list -fmerge-all-constants -fmerge-constants -fmodulo-sched @gol
358 -fmodulo-sched-allow-regmoves -fmove-loop-invariants -fmudflap @gol
359 -fmudflapir -fmudflapth -fno-branch-count-reg -fno-default-inline @gol
360 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
361 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
362 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
363 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
364 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
365 -fpartial-inlining -fpeel-loops -fpredictive-commoning -fprefetch-loop-arrays @gol
366 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
367 -fprofile-generate=@var{path} @gol
368 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
369 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
370 -freorder-blocks-and-partition -freorder-functions @gol
371 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
372 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
373 -fsched-spec-load -fsched-spec-load-dangerous @gol
374 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
375 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
376 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
377 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
378 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
379 -fselective-scheduling -fselective-scheduling2 @gol
380 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
381 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
382 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
383 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer @gol
384 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
385 -ftree-copyrename -ftree-dce -ftree-dominator-opts -ftree-dse @gol
386 -ftree-forwprop -ftree-fre -ftree-loop-if-convert -ftree-loop-im @gol
387 -ftree-phiprop -ftree-loop-distribution @gol
388 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
389 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-pta -ftree-reassoc @gol
390 -ftree-sink -ftree-sra -ftree-switch-conversion @gol
391 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
392 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
393 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
394 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
395 -fwhole-program -fwhopr[=@var{n}] -fwpa -fuse-linker-plugin @gol
396 --param @var{name}=@var{value}
397 -O -O0 -O1 -O2 -O3 -Os -Ofast}
399 @item Preprocessor Options
400 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
401 @gccoptlist{-A@var{question}=@var{answer} @gol
402 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
403 -C -dD -dI -dM -dN @gol
404 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
405 -idirafter @var{dir} @gol
406 -include @var{file} -imacros @var{file} @gol
407 -iprefix @var{file} -iwithprefix @var{dir} @gol
408 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
409 -imultilib @var{dir} -isysroot @var{dir} @gol
410 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
411 -P -fworking-directory -remap @gol
412 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
413 -Xpreprocessor @var{option}}
415 @item Assembler Option
416 @xref{Assembler Options,,Passing Options to the Assembler}.
417 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
420 @xref{Link Options,,Options for Linking}.
421 @gccoptlist{@var{object-file-name} -l@var{library} @gol
422 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
423 -s -static -static-libgcc -static-libstdc++ -shared @gol
424 -shared-libgcc -symbolic @gol
425 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
428 @item Directory Options
429 @xref{Directory Options,,Options for Directory Search}.
430 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir}}
431 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I-
434 @item Machine Dependent Options
435 @xref{Submodel Options,,Hardware Models and Configurations}.
436 @c This list is ordered alphanumerically by subsection name.
437 @c Try and put the significant identifier (CPU or system) first,
438 @c so users have a clue at guessing where the ones they want will be.
441 @gccoptlist{-EB -EL @gol
442 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
443 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
446 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
447 -mabi=@var{name} @gol
448 -mapcs-stack-check -mno-apcs-stack-check @gol
449 -mapcs-float -mno-apcs-float @gol
450 -mapcs-reentrant -mno-apcs-reentrant @gol
451 -msched-prolog -mno-sched-prolog @gol
452 -mlittle-endian -mbig-endian -mwords-little-endian @gol
453 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
454 -mfp16-format=@var{name}
455 -mthumb-interwork -mno-thumb-interwork @gol
456 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
457 -mstructure-size-boundary=@var{n} @gol
458 -mabort-on-noreturn @gol
459 -mlong-calls -mno-long-calls @gol
460 -msingle-pic-base -mno-single-pic-base @gol
461 -mpic-register=@var{reg} @gol
462 -mnop-fun-dllimport @gol
463 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
464 -mpoke-function-name @gol
466 -mtpcs-frame -mtpcs-leaf-frame @gol
467 -mcaller-super-interworking -mcallee-super-interworking @gol
469 -mword-relocations @gol
470 -mfix-cortex-m3-ldrd}
473 @gccoptlist{-mmcu=@var{mcu} -mno-interrupts @gol
474 -mcall-prologues -mtiny-stack -mint8}
476 @emph{Blackfin Options}
477 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
478 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
479 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
480 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
481 -mno-id-shared-library -mshared-library-id=@var{n} @gol
482 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
483 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
484 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
488 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
489 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
490 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
491 -mstack-align -mdata-align -mconst-align @gol
492 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
493 -melf -maout -melinux -mlinux -sim -sim2 @gol
494 -mmul-bug-workaround -mno-mul-bug-workaround}
497 @gccoptlist{-mmac -mpush-args}
499 @emph{Darwin Options}
500 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
501 -arch_only -bind_at_load -bundle -bundle_loader @gol
502 -client_name -compatibility_version -current_version @gol
504 -dependency-file -dylib_file -dylinker_install_name @gol
505 -dynamic -dynamiclib -exported_symbols_list @gol
506 -filelist -flat_namespace -force_cpusubtype_ALL @gol
507 -force_flat_namespace -headerpad_max_install_names @gol
509 -image_base -init -install_name -keep_private_externs @gol
510 -multi_module -multiply_defined -multiply_defined_unused @gol
511 -noall_load -no_dead_strip_inits_and_terms @gol
512 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
513 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
514 -private_bundle -read_only_relocs -sectalign @gol
515 -sectobjectsymbols -whyload -seg1addr @gol
516 -sectcreate -sectobjectsymbols -sectorder @gol
517 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
518 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
519 -segprot -segs_read_only_addr -segs_read_write_addr @gol
520 -single_module -static -sub_library -sub_umbrella @gol
521 -twolevel_namespace -umbrella -undefined @gol
522 -unexported_symbols_list -weak_reference_mismatches @gol
523 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
524 -mkernel -mone-byte-bool}
526 @emph{DEC Alpha Options}
527 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
528 -mieee -mieee-with-inexact -mieee-conformant @gol
529 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
530 -mtrap-precision=@var{mode} -mbuild-constants @gol
531 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
532 -mbwx -mmax -mfix -mcix @gol
533 -mfloat-vax -mfloat-ieee @gol
534 -mexplicit-relocs -msmall-data -mlarge-data @gol
535 -msmall-text -mlarge-text @gol
536 -mmemory-latency=@var{time}}
538 @emph{DEC Alpha/VMS Options}
539 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
542 @gccoptlist{-msmall-model -mno-lsim}
545 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
546 -mhard-float -msoft-float @gol
547 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
548 -mdouble -mno-double @gol
549 -mmedia -mno-media -mmuladd -mno-muladd @gol
550 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
551 -mlinked-fp -mlong-calls -malign-labels @gol
552 -mlibrary-pic -macc-4 -macc-8 @gol
553 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
554 -moptimize-membar -mno-optimize-membar @gol
555 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
556 -mvliw-branch -mno-vliw-branch @gol
557 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
558 -mno-nested-cond-exec -mtomcat-stats @gol
562 @emph{GNU/Linux Options}
563 @gccoptlist{-mglibc -muclibc -mbionic -mandroid @gol
564 -tno-android-cc -tno-android-ld}
566 @emph{H8/300 Options}
567 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
570 @gccoptlist{-march=@var{architecture-type} @gol
571 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
572 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
573 -mfixed-range=@var{register-range} @gol
574 -mjump-in-delay -mlinker-opt -mlong-calls @gol
575 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
576 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
577 -mno-jump-in-delay -mno-long-load-store @gol
578 -mno-portable-runtime -mno-soft-float @gol
579 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
580 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
581 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
582 -munix=@var{unix-std} -nolibdld -static -threads}
584 @emph{i386 and x86-64 Options}
585 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
586 -mfpmath=@var{unit} @gol
587 -masm=@var{dialect} -mno-fancy-math-387 @gol
588 -mno-fp-ret-in-387 -msoft-float @gol
589 -mno-wide-multiply -mrtd -malign-double @gol
590 -mpreferred-stack-boundary=@var{num}
591 -mincoming-stack-boundary=@var{num}
592 -mcld -mcx16 -msahf -mmovbe -mcrc32 -mrecip @gol
593 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
594 -maes -mpclmul -mfsgsbase -mrdrnd -mf16c -mfused-madd @gol
595 -msse4a -m3dnow -mpopcnt -mabm -mfma4 -mxop -mlwp @gol
596 -mthreads -mno-align-stringops -minline-all-stringops @gol
597 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
598 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
599 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
600 -mveclibabi=@var{type} -mpc32 -mpc64 -mpc80 -mstackrealign @gol
601 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
602 -mcmodel=@var{code-model} -mabi=@var{name} @gol
603 -m32 -m64 -mlarge-data-threshold=@var{num} @gol
607 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
608 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
609 -mconstant-gp -mauto-pic -mfused-madd @gol
610 -minline-float-divide-min-latency @gol
611 -minline-float-divide-max-throughput @gol
612 -mno-inline-float-divide @gol
613 -minline-int-divide-min-latency @gol
614 -minline-int-divide-max-throughput @gol
615 -mno-inline-int-divide @gol
616 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
617 -mno-inline-sqrt @gol
618 -mdwarf2-asm -mearly-stop-bits @gol
619 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
620 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
621 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
622 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
623 -msched-spec-ldc -msched-spec-control-ldc @gol
624 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
625 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
626 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
627 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
629 @emph{IA-64/VMS Options}
630 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
633 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
634 -msign-extend-enabled -muser-enabled}
636 @emph{M32R/D Options}
637 @gccoptlist{-m32r2 -m32rx -m32r @gol
639 -malign-loops -mno-align-loops @gol
640 -missue-rate=@var{number} @gol
641 -mbranch-cost=@var{number} @gol
642 -mmodel=@var{code-size-model-type} @gol
643 -msdata=@var{sdata-type} @gol
644 -mno-flush-func -mflush-func=@var{name} @gol
645 -mno-flush-trap -mflush-trap=@var{number} @gol
649 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
651 @emph{M680x0 Options}
652 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
653 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
654 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
655 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
656 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
657 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
658 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
659 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
662 @emph{M68hc1x Options}
663 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
664 -mauto-incdec -minmax -mlong-calls -mshort @gol
665 -msoft-reg-count=@var{count}}
668 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
669 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
670 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
671 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
672 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
675 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
676 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
677 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
678 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
682 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
683 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
684 -mips64 -mips64r2 @gol
685 -mips16 -mno-mips16 -mflip-mips16 @gol
686 -minterlink-mips16 -mno-interlink-mips16 @gol
687 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
688 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
689 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
690 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
691 -mfpu=@var{fpu-type} @gol
692 -msmartmips -mno-smartmips @gol
693 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
694 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
695 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
696 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
697 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
698 -membedded-data -mno-embedded-data @gol
699 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
700 -mcode-readable=@var{setting} @gol
701 -msplit-addresses -mno-split-addresses @gol
702 -mexplicit-relocs -mno-explicit-relocs @gol
703 -mcheck-zero-division -mno-check-zero-division @gol
704 -mdivide-traps -mdivide-breaks @gol
705 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
706 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
707 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
708 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
709 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
710 -mflush-func=@var{func} -mno-flush-func @gol
711 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
712 -mfp-exceptions -mno-fp-exceptions @gol
713 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
714 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
717 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
718 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
719 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
720 -mno-base-addresses -msingle-exit -mno-single-exit}
722 @emph{MN10300 Options}
723 @gccoptlist{-mmult-bug -mno-mult-bug @gol
724 -mam33 -mno-am33 @gol
725 -mam33-2 -mno-am33-2 @gol
726 -mreturn-pointer-on-d0 @gol
729 @emph{PDP-11 Options}
730 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
731 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
732 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
733 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
734 -mbranch-expensive -mbranch-cheap @gol
735 -msplit -mno-split -munix-asm -mdec-asm}
737 @emph{picoChip Options}
738 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N}
739 -msymbol-as-address -mno-inefficient-warnings}
741 @emph{PowerPC Options}
742 See RS/6000 and PowerPC Options.
744 @emph{RS/6000 and PowerPC Options}
745 @gccoptlist{-mcpu=@var{cpu-type} @gol
746 -mtune=@var{cpu-type} @gol
747 -mcmodel=@var{code-model} @gol
748 -mpower -mno-power -mpower2 -mno-power2 @gol
749 -mpowerpc -mpowerpc64 -mno-powerpc @gol
750 -maltivec -mno-altivec @gol
751 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
752 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
753 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
754 -mfprnd -mno-fprnd @gol
755 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
756 -mnew-mnemonics -mold-mnemonics @gol
757 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
758 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
759 -malign-power -malign-natural @gol
760 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
761 -msingle-float -mdouble-float -msimple-fpu @gol
762 -mstring -mno-string -mupdate -mno-update @gol
763 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
764 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
765 -mstrict-align -mno-strict-align -mrelocatable @gol
766 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
767 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
768 -mdynamic-no-pic -maltivec -mswdiv @gol
769 -mprioritize-restricted-insns=@var{priority} @gol
770 -msched-costly-dep=@var{dependence_type} @gol
771 -minsert-sched-nops=@var{scheme} @gol
772 -mcall-sysv -mcall-netbsd @gol
773 -maix-struct-return -msvr4-struct-return @gol
774 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
775 -mblock-move-inline-limit=@var{num} @gol
776 -misel -mno-isel @gol
777 -misel=yes -misel=no @gol
779 -mspe=yes -mspe=no @gol
781 -mgen-cell-microcode -mwarn-cell-microcode @gol
782 -mvrsave -mno-vrsave @gol
783 -mmulhw -mno-mulhw @gol
784 -mdlmzb -mno-dlmzb @gol
785 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
786 -mprototype -mno-prototype @gol
787 -msim -mmvme -mads -myellowknife -memb -msdata @gol
788 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
789 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision -mno-recip-precision}
792 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
794 -mbig-endian-data -mlittle-endian-data @gol
797 -mas100-syntax -mno-as100-syntax@gol
799 -mmax-constant-size=@gol
801 -msave-acc-in-interrupts}
803 @emph{S/390 and zSeries Options}
804 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
805 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
806 -mlong-double-64 -mlong-double-128 @gol
807 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
808 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
809 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
810 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
811 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
814 @gccoptlist{-meb -mel @gol
818 -mscore5 -mscore5u -mscore7 -mscore7d}
821 @gccoptlist{-m1 -m2 -m2e @gol
822 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
824 -m4-nofpu -m4-single-only -m4-single -m4 @gol
825 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
826 -m5-64media -m5-64media-nofpu @gol
827 -m5-32media -m5-32media-nofpu @gol
828 -m5-compact -m5-compact-nofpu @gol
829 -mb -ml -mdalign -mrelax @gol
830 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
831 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
832 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
833 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
834 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
835 -maccumulate-outgoing-args -minvalid-symbols}
838 @gccoptlist{-mcpu=@var{cpu-type} @gol
839 -mtune=@var{cpu-type} @gol
840 -mcmodel=@var{code-model} @gol
841 -m32 -m64 -mapp-regs -mno-app-regs @gol
842 -mfaster-structs -mno-faster-structs @gol
843 -mfpu -mno-fpu -mhard-float -msoft-float @gol
844 -mhard-quad-float -msoft-quad-float @gol
845 -mimpure-text -mno-impure-text -mlittle-endian @gol
846 -mstack-bias -mno-stack-bias @gol
847 -munaligned-doubles -mno-unaligned-doubles @gol
848 -mv8plus -mno-v8plus -mvis -mno-vis
849 -threads -pthreads -pthread}
852 @gccoptlist{-mwarn-reloc -merror-reloc @gol
853 -msafe-dma -munsafe-dma @gol
855 -msmall-mem -mlarge-mem -mstdmain @gol
856 -mfixed-range=@var{register-range} @gol
858 -maddress-space-conversion -mno-address-space-conversion @gol
859 -mcache-size=@var{cache-size} @gol
860 -matomic-updates -mno-atomic-updates}
862 @emph{System V Options}
863 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
866 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
867 -mprolog-function -mno-prolog-function -mspace @gol
868 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
869 -mapp-regs -mno-app-regs @gol
870 -mdisable-callt -mno-disable-callt @gol
876 @gccoptlist{-mg -mgnu -munix}
878 @emph{VxWorks Options}
879 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
880 -Xbind-lazy -Xbind-now}
882 @emph{x86-64 Options}
883 See i386 and x86-64 Options.
885 @emph{i386 and x86-64 Windows Options}
886 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll
887 -mnop-fun-dllimport -mthread -municode -mwin32 -mwindows
888 -fno-set-stack-executable}
890 @emph{Xstormy16 Options}
893 @emph{Xtensa Options}
894 @gccoptlist{-mconst16 -mno-const16 @gol
895 -mfused-madd -mno-fused-madd @gol
897 -mserialize-volatile -mno-serialize-volatile @gol
898 -mtext-section-literals -mno-text-section-literals @gol
899 -mtarget-align -mno-target-align @gol
900 -mlongcalls -mno-longcalls}
902 @emph{zSeries Options}
903 See S/390 and zSeries Options.
905 @item Code Generation Options
906 @xref{Code Gen Options,,Options for Code Generation Conventions}.
907 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
908 -ffixed-@var{reg} -fexceptions @gol
909 -fnon-call-exceptions -funwind-tables @gol
910 -fasynchronous-unwind-tables @gol
911 -finhibit-size-directive -finstrument-functions @gol
912 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
913 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
914 -fno-common -fno-ident @gol
915 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
916 -fno-jump-tables @gol
917 -frecord-gcc-switches @gol
918 -freg-struct-return -fshort-enums @gol
919 -fshort-double -fshort-wchar @gol
920 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
921 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
922 -fno-stack-limit @gol
923 -fleading-underscore -ftls-model=@var{model} @gol
924 -ftrapv -fwrapv -fbounds-check @gol
929 * Overall Options:: Controlling the kind of output:
930 an executable, object files, assembler files,
931 or preprocessed source.
932 * C Dialect Options:: Controlling the variant of C language compiled.
933 * C++ Dialect Options:: Variations on C++.
934 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
936 * Language Independent Options:: Controlling how diagnostics should be
938 * Warning Options:: How picky should the compiler be?
939 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
940 * Optimize Options:: How much optimization?
941 * Preprocessor Options:: Controlling header files and macro definitions.
942 Also, getting dependency information for Make.
943 * Assembler Options:: Passing options to the assembler.
944 * Link Options:: Specifying libraries and so on.
945 * Directory Options:: Where to find header files and libraries.
946 Where to find the compiler executable files.
947 * Spec Files:: How to pass switches to sub-processes.
948 * Target Options:: Running a cross-compiler, or an old version of GCC.
951 @node Overall Options
952 @section Options Controlling the Kind of Output
954 Compilation can involve up to four stages: preprocessing, compilation
955 proper, assembly and linking, always in that order. GCC is capable of
956 preprocessing and compiling several files either into several
957 assembler input files, or into one assembler input file; then each
958 assembler input file produces an object file, and linking combines all
959 the object files (those newly compiled, and those specified as input)
960 into an executable file.
962 @cindex file name suffix
963 For any given input file, the file name suffix determines what kind of
968 C source code which must be preprocessed.
971 C source code which should not be preprocessed.
974 C++ source code which should not be preprocessed.
977 Objective-C source code. Note that you must link with the @file{libobjc}
978 library to make an Objective-C program work.
981 Objective-C source code which should not be preprocessed.
985 Objective-C++ source code. Note that you must link with the @file{libobjc}
986 library to make an Objective-C++ program work. Note that @samp{.M} refers
987 to a literal capital M@.
990 Objective-C++ source code which should not be preprocessed.
993 C, C++, Objective-C or Objective-C++ header file to be turned into a
994 precompiled header (default), or C, C++ header file to be turned into an
995 Ada spec (via the @option{-fdump-ada-spec} switch).
999 @itemx @var{file}.cxx
1000 @itemx @var{file}.cpp
1001 @itemx @var{file}.CPP
1002 @itemx @var{file}.c++
1004 C++ source code which must be preprocessed. Note that in @samp{.cxx},
1005 the last two letters must both be literally @samp{x}. Likewise,
1006 @samp{.C} refers to a literal capital C@.
1010 Objective-C++ source code which must be preprocessed.
1012 @item @var{file}.mii
1013 Objective-C++ source code which should not be preprocessed.
1017 @itemx @var{file}.hp
1018 @itemx @var{file}.hxx
1019 @itemx @var{file}.hpp
1020 @itemx @var{file}.HPP
1021 @itemx @var{file}.h++
1022 @itemx @var{file}.tcc
1023 C++ header file to be turned into a precompiled header or Ada spec.
1026 @itemx @var{file}.for
1027 @itemx @var{file}.ftn
1028 Fixed form Fortran source code which should not be preprocessed.
1031 @itemx @var{file}.FOR
1032 @itemx @var{file}.fpp
1033 @itemx @var{file}.FPP
1034 @itemx @var{file}.FTN
1035 Fixed form Fortran source code which must be preprocessed (with the traditional
1038 @item @var{file}.f90
1039 @itemx @var{file}.f95
1040 @itemx @var{file}.f03
1041 @itemx @var{file}.f08
1042 Free form Fortran source code which should not be preprocessed.
1044 @item @var{file}.F90
1045 @itemx @var{file}.F95
1046 @itemx @var{file}.F03
1047 @itemx @var{file}.F08
1048 Free form Fortran source code which must be preprocessed (with the
1049 traditional preprocessor).
1051 @c FIXME: Descriptions of Java file types.
1057 @item @var{file}.ads
1058 Ada source code file which contains a library unit declaration (a
1059 declaration of a package, subprogram, or generic, or a generic
1060 instantiation), or a library unit renaming declaration (a package,
1061 generic, or subprogram renaming declaration). Such files are also
1064 @item @var{file}.adb
1065 Ada source code file containing a library unit body (a subprogram or
1066 package body). Such files are also called @dfn{bodies}.
1068 @c GCC also knows about some suffixes for languages not yet included:
1079 @itemx @var{file}.sx
1080 Assembler code which must be preprocessed.
1083 An object file to be fed straight into linking.
1084 Any file name with no recognized suffix is treated this way.
1088 You can specify the input language explicitly with the @option{-x} option:
1091 @item -x @var{language}
1092 Specify explicitly the @var{language} for the following input files
1093 (rather than letting the compiler choose a default based on the file
1094 name suffix). This option applies to all following input files until
1095 the next @option{-x} option. Possible values for @var{language} are:
1097 c c-header c-cpp-output
1098 c++ c++-header c++-cpp-output
1099 objective-c objective-c-header objective-c-cpp-output
1100 objective-c++ objective-c++-header objective-c++-cpp-output
1101 assembler assembler-with-cpp
1103 f77 f77-cpp-input f95 f95-cpp-input
1108 Turn off any specification of a language, so that subsequent files are
1109 handled according to their file name suffixes (as they are if @option{-x}
1110 has not been used at all).
1112 @item -pass-exit-codes
1113 @opindex pass-exit-codes
1114 Normally the @command{gcc} program will exit with the code of 1 if any
1115 phase of the compiler returns a non-success return code. If you specify
1116 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1117 numerically highest error produced by any phase that returned an error
1118 indication. The C, C++, and Fortran frontends return 4, if an internal
1119 compiler error is encountered.
1122 If you only want some of the stages of compilation, you can use
1123 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1124 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1125 @command{gcc} is to stop. Note that some combinations (for example,
1126 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1131 Compile or assemble the source files, but do not link. The linking
1132 stage simply is not done. The ultimate output is in the form of an
1133 object file for each source file.
1135 By default, the object file name for a source file is made by replacing
1136 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1138 Unrecognized input files, not requiring compilation or assembly, are
1143 Stop after the stage of compilation proper; do not assemble. The output
1144 is in the form of an assembler code file for each non-assembler input
1147 By default, the assembler file name for a source file is made by
1148 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1150 Input files that don't require compilation are ignored.
1154 Stop after the preprocessing stage; do not run the compiler proper. The
1155 output is in the form of preprocessed source code, which is sent to the
1158 Input files which don't require preprocessing are ignored.
1160 @cindex output file option
1163 Place output in file @var{file}. This applies regardless to whatever
1164 sort of output is being produced, whether it be an executable file,
1165 an object file, an assembler file or preprocessed C code.
1167 If @option{-o} is not specified, the default is to put an executable
1168 file in @file{a.out}, the object file for
1169 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1170 assembler file in @file{@var{source}.s}, a precompiled header file in
1171 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1176 Print (on standard error output) the commands executed to run the stages
1177 of compilation. Also print the version number of the compiler driver
1178 program and of the preprocessor and the compiler proper.
1182 Like @option{-v} except the commands are not executed and arguments
1183 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1184 This is useful for shell scripts to capture the driver-generated command lines.
1188 Use pipes rather than temporary files for communication between the
1189 various stages of compilation. This fails to work on some systems where
1190 the assembler is unable to read from a pipe; but the GNU assembler has
1195 If you are compiling multiple source files, this option tells the driver
1196 to pass all the source files to the compiler at once (for those
1197 languages for which the compiler can handle this). This will allow
1198 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1199 language for which this is supported is C@. If you pass source files for
1200 multiple languages to the driver, using this option, the driver will invoke
1201 the compiler(s) that support IMA once each, passing each compiler all the
1202 source files appropriate for it. For those languages that do not support
1203 IMA this option will be ignored, and the compiler will be invoked once for
1204 each source file in that language. If you use this option in conjunction
1205 with @option{-save-temps}, the compiler will generate multiple
1207 (one for each source file), but only one (combined) @file{.o} or
1212 Print (on the standard output) a description of the command line options
1213 understood by @command{gcc}. If the @option{-v} option is also specified
1214 then @option{--help} will also be passed on to the various processes
1215 invoked by @command{gcc}, so that they can display the command line options
1216 they accept. If the @option{-Wextra} option has also been specified
1217 (prior to the @option{--help} option), then command line options which
1218 have no documentation associated with them will also be displayed.
1221 @opindex target-help
1222 Print (on the standard output) a description of target-specific command
1223 line options for each tool. For some targets extra target-specific
1224 information may also be printed.
1226 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1227 Print (on the standard output) a description of the command line
1228 options understood by the compiler that fit into all specified classes
1229 and qualifiers. These are the supported classes:
1232 @item @samp{optimizers}
1233 This will display all of the optimization options supported by the
1236 @item @samp{warnings}
1237 This will display all of the options controlling warning messages
1238 produced by the compiler.
1241 This will display target-specific options. Unlike the
1242 @option{--target-help} option however, target-specific options of the
1243 linker and assembler will not be displayed. This is because those
1244 tools do not currently support the extended @option{--help=} syntax.
1247 This will display the values recognized by the @option{--param}
1250 @item @var{language}
1251 This will display the options supported for @var{language}, where
1252 @var{language} is the name of one of the languages supported in this
1256 This will display the options that are common to all languages.
1259 These are the supported qualifiers:
1262 @item @samp{undocumented}
1263 Display only those options which are undocumented.
1266 Display options which take an argument that appears after an equal
1267 sign in the same continuous piece of text, such as:
1268 @samp{--help=target}.
1270 @item @samp{separate}
1271 Display options which take an argument that appears as a separate word
1272 following the original option, such as: @samp{-o output-file}.
1275 Thus for example to display all the undocumented target-specific
1276 switches supported by the compiler the following can be used:
1279 --help=target,undocumented
1282 The sense of a qualifier can be inverted by prefixing it with the
1283 @samp{^} character, so for example to display all binary warning
1284 options (i.e., ones that are either on or off and that do not take an
1285 argument), which have a description the following can be used:
1288 --help=warnings,^joined,^undocumented
1291 The argument to @option{--help=} should not consist solely of inverted
1294 Combining several classes is possible, although this usually
1295 restricts the output by so much that there is nothing to display. One
1296 case where it does work however is when one of the classes is
1297 @var{target}. So for example to display all the target-specific
1298 optimization options the following can be used:
1301 --help=target,optimizers
1304 The @option{--help=} option can be repeated on the command line. Each
1305 successive use will display its requested class of options, skipping
1306 those that have already been displayed.
1308 If the @option{-Q} option appears on the command line before the
1309 @option{--help=} option, then the descriptive text displayed by
1310 @option{--help=} is changed. Instead of describing the displayed
1311 options, an indication is given as to whether the option is enabled,
1312 disabled or set to a specific value (assuming that the compiler
1313 knows this at the point where the @option{--help=} option is used).
1315 Here is a truncated example from the ARM port of @command{gcc}:
1318 % gcc -Q -mabi=2 --help=target -c
1319 The following options are target specific:
1321 -mabort-on-noreturn [disabled]
1325 The output is sensitive to the effects of previous command line
1326 options, so for example it is possible to find out which optimizations
1327 are enabled at @option{-O2} by using:
1330 -Q -O2 --help=optimizers
1333 Alternatively you can discover which binary optimizations are enabled
1334 by @option{-O3} by using:
1337 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1338 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1339 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1342 @item -no-canonical-prefixes
1343 @opindex no-canonical-prefixes
1344 Do not expand any symbolic links, resolve references to @samp{/../}
1345 or @samp{/./}, or make the path absolute when generating a relative
1350 Display the version number and copyrights of the invoked GCC@.
1354 Invoke all subcommands under a wrapper program. It takes a single
1355 comma separated list as an argument, which will be used to invoke
1359 gcc -c t.c -wrapper gdb,--args
1362 This will invoke all subprograms of gcc under "gdb --args",
1363 thus cc1 invocation will be "gdb --args cc1 ...".
1365 @item -fplugin=@var{name}.so
1366 Load the plugin code in file @var{name}.so, assumed to be a
1367 shared object to be dlopen'd by the compiler. The base name of
1368 the shared object file is used to identify the plugin for the
1369 purposes of argument parsing (See
1370 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1371 Each plugin should define the callback functions specified in the
1374 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1375 Define an argument called @var{key} with a value of @var{value}
1376 for the plugin called @var{name}.
1378 @item -fdump-ada-spec@r{[}-slim@r{]}
1379 For C and C++ source and include files, generate corresponding Ada
1380 specs. @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1381 GNAT User's Guide}, which provides detailed documentation on this feature.
1383 @include @value{srcdir}/../libiberty/at-file.texi
1387 @section Compiling C++ Programs
1389 @cindex suffixes for C++ source
1390 @cindex C++ source file suffixes
1391 C++ source files conventionally use one of the suffixes @samp{.C},
1392 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1393 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1394 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1395 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1396 files with these names and compiles them as C++ programs even if you
1397 call the compiler the same way as for compiling C programs (usually
1398 with the name @command{gcc}).
1402 However, the use of @command{gcc} does not add the C++ library.
1403 @command{g++} is a program that calls GCC and treats @samp{.c},
1404 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1405 files unless @option{-x} is used, and automatically specifies linking
1406 against the C++ library. This program is also useful when
1407 precompiling a C header file with a @samp{.h} extension for use in C++
1408 compilations. On many systems, @command{g++} is also installed with
1409 the name @command{c++}.
1411 @cindex invoking @command{g++}
1412 When you compile C++ programs, you may specify many of the same
1413 command-line options that you use for compiling programs in any
1414 language; or command-line options meaningful for C and related
1415 languages; or options that are meaningful only for C++ programs.
1416 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1417 explanations of options for languages related to C@.
1418 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1419 explanations of options that are meaningful only for C++ programs.
1421 @node C Dialect Options
1422 @section Options Controlling C Dialect
1423 @cindex dialect options
1424 @cindex language dialect options
1425 @cindex options, dialect
1427 The following options control the dialect of C (or languages derived
1428 from C, such as C++, Objective-C and Objective-C++) that the compiler
1432 @cindex ANSI support
1436 In C mode, this is equivalent to @samp{-std=c90}. In C++ mode, it is
1437 equivalent to @samp{-std=c++98}.
1439 This turns off certain features of GCC that are incompatible with ISO
1440 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1441 such as the @code{asm} and @code{typeof} keywords, and
1442 predefined macros such as @code{unix} and @code{vax} that identify the
1443 type of system you are using. It also enables the undesirable and
1444 rarely used ISO trigraph feature. For the C compiler,
1445 it disables recognition of C++ style @samp{//} comments as well as
1446 the @code{inline} keyword.
1448 The alternate keywords @code{__asm__}, @code{__extension__},
1449 @code{__inline__} and @code{__typeof__} continue to work despite
1450 @option{-ansi}. You would not want to use them in an ISO C program, of
1451 course, but it is useful to put them in header files that might be included
1452 in compilations done with @option{-ansi}. Alternate predefined macros
1453 such as @code{__unix__} and @code{__vax__} are also available, with or
1454 without @option{-ansi}.
1456 The @option{-ansi} option does not cause non-ISO programs to be
1457 rejected gratuitously. For that, @option{-pedantic} is required in
1458 addition to @option{-ansi}. @xref{Warning Options}.
1460 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1461 option is used. Some header files may notice this macro and refrain
1462 from declaring certain functions or defining certain macros that the
1463 ISO standard doesn't call for; this is to avoid interfering with any
1464 programs that might use these names for other things.
1466 Functions that would normally be built in but do not have semantics
1467 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1468 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1469 built-in functions provided by GCC}, for details of the functions
1474 Determine the language standard. @xref{Standards,,Language Standards
1475 Supported by GCC}, for details of these standard versions. This option
1476 is currently only supported when compiling C or C++.
1478 The compiler can accept several base standards, such as @samp{c90} or
1479 @samp{c++98}, and GNU dialects of those standards, such as
1480 @samp{gnu90} or @samp{gnu++98}. By specifying a base standard, the
1481 compiler will accept all programs following that standard and those
1482 using GNU extensions that do not contradict it. For example,
1483 @samp{-std=c90} turns off certain features of GCC that are
1484 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1485 keywords, but not other GNU extensions that do not have a meaning in
1486 ISO C90, such as omitting the middle term of a @code{?:}
1487 expression. On the other hand, by specifying a GNU dialect of a
1488 standard, all features the compiler support are enabled, even when
1489 those features change the meaning of the base standard and some
1490 strict-conforming programs may be rejected. The particular standard
1491 is used by @option{-pedantic} to identify which features are GNU
1492 extensions given that version of the standard. For example
1493 @samp{-std=gnu90 -pedantic} would warn about C++ style @samp{//}
1494 comments, while @samp{-std=gnu99 -pedantic} would not.
1496 A value for this option must be provided; possible values are
1502 Support all ISO C90 programs (certain GNU extensions that conflict
1503 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1505 @item iso9899:199409
1506 ISO C90 as modified in amendment 1.
1512 ISO C99. Note that this standard is not yet fully supported; see
1513 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1514 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1517 ISO C1X, the draft of the next revision of the ISO C standard.
1518 Support is limited and experimental and features enabled by this
1519 option may be changed or removed if changed in or removed from the
1524 GNU dialect of ISO C90 (including some C99 features). This
1525 is the default for C code.
1529 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1530 this will become the default. The name @samp{gnu9x} is deprecated.
1533 GNU dialect of ISO C1X. Support is limited and experimental and
1534 features enabled by this option may be changed or removed if changed
1535 in or removed from the standard draft.
1538 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1542 GNU dialect of @option{-std=c++98}. This is the default for
1546 The working draft of the upcoming ISO C++0x standard. This option
1547 enables experimental features that are likely to be included in
1548 C++0x. The working draft is constantly changing, and any feature that is
1549 enabled by this flag may be removed from future versions of GCC if it is
1550 not part of the C++0x standard.
1553 GNU dialect of @option{-std=c++0x}. This option enables
1554 experimental features that may be removed in future versions of GCC.
1557 @item -fgnu89-inline
1558 @opindex fgnu89-inline
1559 The option @option{-fgnu89-inline} tells GCC to use the traditional
1560 GNU semantics for @code{inline} functions when in C99 mode.
1561 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1562 is accepted and ignored by GCC versions 4.1.3 up to but not including
1563 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1564 C99 mode. Using this option is roughly equivalent to adding the
1565 @code{gnu_inline} function attribute to all inline functions
1566 (@pxref{Function Attributes}).
1568 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1569 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1570 specifies the default behavior). This option was first supported in
1571 GCC 4.3. This option is not supported in @option{-std=c90} or
1572 @option{-std=gnu90} mode.
1574 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1575 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1576 in effect for @code{inline} functions. @xref{Common Predefined
1577 Macros,,,cpp,The C Preprocessor}.
1579 @item -aux-info @var{filename}
1581 Output to the given filename prototyped declarations for all functions
1582 declared and/or defined in a translation unit, including those in header
1583 files. This option is silently ignored in any language other than C@.
1585 Besides declarations, the file indicates, in comments, the origin of
1586 each declaration (source file and line), whether the declaration was
1587 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1588 @samp{O} for old, respectively, in the first character after the line
1589 number and the colon), and whether it came from a declaration or a
1590 definition (@samp{C} or @samp{F}, respectively, in the following
1591 character). In the case of function definitions, a K&R-style list of
1592 arguments followed by their declarations is also provided, inside
1593 comments, after the declaration.
1597 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1598 keyword, so that code can use these words as identifiers. You can use
1599 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1600 instead. @option{-ansi} implies @option{-fno-asm}.
1602 In C++, this switch only affects the @code{typeof} keyword, since
1603 @code{asm} and @code{inline} are standard keywords. You may want to
1604 use the @option{-fno-gnu-keywords} flag instead, which has the same
1605 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1606 switch only affects the @code{asm} and @code{typeof} keywords, since
1607 @code{inline} is a standard keyword in ISO C99.
1610 @itemx -fno-builtin-@var{function}
1611 @opindex fno-builtin
1612 @cindex built-in functions
1613 Don't recognize built-in functions that do not begin with
1614 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1615 functions provided by GCC}, for details of the functions affected,
1616 including those which are not built-in functions when @option{-ansi} or
1617 @option{-std} options for strict ISO C conformance are used because they
1618 do not have an ISO standard meaning.
1620 GCC normally generates special code to handle certain built-in functions
1621 more efficiently; for instance, calls to @code{alloca} may become single
1622 instructions that adjust the stack directly, and calls to @code{memcpy}
1623 may become inline copy loops. The resulting code is often both smaller
1624 and faster, but since the function calls no longer appear as such, you
1625 cannot set a breakpoint on those calls, nor can you change the behavior
1626 of the functions by linking with a different library. In addition,
1627 when a function is recognized as a built-in function, GCC may use
1628 information about that function to warn about problems with calls to
1629 that function, or to generate more efficient code, even if the
1630 resulting code still contains calls to that function. For example,
1631 warnings are given with @option{-Wformat} for bad calls to
1632 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1633 known not to modify global memory.
1635 With the @option{-fno-builtin-@var{function}} option
1636 only the built-in function @var{function} is
1637 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1638 function is named that is not built-in in this version of GCC, this
1639 option is ignored. There is no corresponding
1640 @option{-fbuiltin-@var{function}} option; if you wish to enable
1641 built-in functions selectively when using @option{-fno-builtin} or
1642 @option{-ffreestanding}, you may define macros such as:
1645 #define abs(n) __builtin_abs ((n))
1646 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1651 @cindex hosted environment
1653 Assert that compilation takes place in a hosted environment. This implies
1654 @option{-fbuiltin}. A hosted environment is one in which the
1655 entire standard library is available, and in which @code{main} has a return
1656 type of @code{int}. Examples are nearly everything except a kernel.
1657 This is equivalent to @option{-fno-freestanding}.
1659 @item -ffreestanding
1660 @opindex ffreestanding
1661 @cindex hosted environment
1663 Assert that compilation takes place in a freestanding environment. This
1664 implies @option{-fno-builtin}. A freestanding environment
1665 is one in which the standard library may not exist, and program startup may
1666 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1667 This is equivalent to @option{-fno-hosted}.
1669 @xref{Standards,,Language Standards Supported by GCC}, for details of
1670 freestanding and hosted environments.
1674 @cindex openmp parallel
1675 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1676 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1677 compiler generates parallel code according to the OpenMP Application
1678 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1679 implies @option{-pthread}, and thus is only supported on targets that
1680 have support for @option{-pthread}.
1682 @item -fms-extensions
1683 @opindex fms-extensions
1684 Accept some non-standard constructs used in Microsoft header files.
1686 Some cases of unnamed fields in structures and unions are only
1687 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1688 fields within structs/unions}, for details.
1692 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1693 options for strict ISO C conformance) implies @option{-trigraphs}.
1695 @item -no-integrated-cpp
1696 @opindex no-integrated-cpp
1697 Performs a compilation in two passes: preprocessing and compiling. This
1698 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1699 @option{-B} option. The user supplied compilation step can then add in
1700 an additional preprocessing step after normal preprocessing but before
1701 compiling. The default is to use the integrated cpp (internal cpp)
1703 The semantics of this option will change if "cc1", "cc1plus", and
1704 "cc1obj" are merged.
1706 @cindex traditional C language
1707 @cindex C language, traditional
1709 @itemx -traditional-cpp
1710 @opindex traditional-cpp
1711 @opindex traditional
1712 Formerly, these options caused GCC to attempt to emulate a pre-standard
1713 C compiler. They are now only supported with the @option{-E} switch.
1714 The preprocessor continues to support a pre-standard mode. See the GNU
1715 CPP manual for details.
1717 @item -fcond-mismatch
1718 @opindex fcond-mismatch
1719 Allow conditional expressions with mismatched types in the second and
1720 third arguments. The value of such an expression is void. This option
1721 is not supported for C++.
1723 @item -flax-vector-conversions
1724 @opindex flax-vector-conversions
1725 Allow implicit conversions between vectors with differing numbers of
1726 elements and/or incompatible element types. This option should not be
1729 @item -funsigned-char
1730 @opindex funsigned-char
1731 Let the type @code{char} be unsigned, like @code{unsigned char}.
1733 Each kind of machine has a default for what @code{char} should
1734 be. It is either like @code{unsigned char} by default or like
1735 @code{signed char} by default.
1737 Ideally, a portable program should always use @code{signed char} or
1738 @code{unsigned char} when it depends on the signedness of an object.
1739 But many programs have been written to use plain @code{char} and
1740 expect it to be signed, or expect it to be unsigned, depending on the
1741 machines they were written for. This option, and its inverse, let you
1742 make such a program work with the opposite default.
1744 The type @code{char} is always a distinct type from each of
1745 @code{signed char} or @code{unsigned char}, even though its behavior
1746 is always just like one of those two.
1749 @opindex fsigned-char
1750 Let the type @code{char} be signed, like @code{signed char}.
1752 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1753 the negative form of @option{-funsigned-char}. Likewise, the option
1754 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1756 @item -fsigned-bitfields
1757 @itemx -funsigned-bitfields
1758 @itemx -fno-signed-bitfields
1759 @itemx -fno-unsigned-bitfields
1760 @opindex fsigned-bitfields
1761 @opindex funsigned-bitfields
1762 @opindex fno-signed-bitfields
1763 @opindex fno-unsigned-bitfields
1764 These options control whether a bit-field is signed or unsigned, when the
1765 declaration does not use either @code{signed} or @code{unsigned}. By
1766 default, such a bit-field is signed, because this is consistent: the
1767 basic integer types such as @code{int} are signed types.
1770 @node C++ Dialect Options
1771 @section Options Controlling C++ Dialect
1773 @cindex compiler options, C++
1774 @cindex C++ options, command line
1775 @cindex options, C++
1776 This section describes the command-line options that are only meaningful
1777 for C++ programs; but you can also use most of the GNU compiler options
1778 regardless of what language your program is in. For example, you
1779 might compile a file @code{firstClass.C} like this:
1782 g++ -g -frepo -O -c firstClass.C
1786 In this example, only @option{-frepo} is an option meant
1787 only for C++ programs; you can use the other options with any
1788 language supported by GCC@.
1790 Here is a list of options that are @emph{only} for compiling C++ programs:
1794 @item -fabi-version=@var{n}
1795 @opindex fabi-version
1796 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1797 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1798 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1799 the version that conforms most closely to the C++ ABI specification.
1800 Therefore, the ABI obtained using version 0 will change as ABI bugs
1803 The default is version 2.
1805 Version 3 corrects an error in mangling a constant address as a
1808 Version 4 implements a standard mangling for vector types.
1810 See also @option{-Wabi}.
1812 @item -fno-access-control
1813 @opindex fno-access-control
1814 Turn off all access checking. This switch is mainly useful for working
1815 around bugs in the access control code.
1819 Check that the pointer returned by @code{operator new} is non-null
1820 before attempting to modify the storage allocated. This check is
1821 normally unnecessary because the C++ standard specifies that
1822 @code{operator new} will only return @code{0} if it is declared
1823 @samp{throw()}, in which case the compiler will always check the
1824 return value even without this option. In all other cases, when
1825 @code{operator new} has a non-empty exception specification, memory
1826 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1827 @samp{new (nothrow)}.
1829 @item -fconserve-space
1830 @opindex fconserve-space
1831 Put uninitialized or runtime-initialized global variables into the
1832 common segment, as C does. This saves space in the executable at the
1833 cost of not diagnosing duplicate definitions. If you compile with this
1834 flag and your program mysteriously crashes after @code{main()} has
1835 completed, you may have an object that is being destroyed twice because
1836 two definitions were merged.
1838 This option is no longer useful on most targets, now that support has
1839 been added for putting variables into BSS without making them common.
1841 @item -fno-deduce-init-list
1842 @opindex fno-deduce-init-list
1843 Disable deduction of a template type parameter as
1844 std::initializer_list from a brace-enclosed initializer list, i.e.
1847 template <class T> auto forward(T t) -> decltype (realfn (t))
1854 forward(@{1,2@}); // call forward<std::initializer_list<int>>
1858 This option is present because this deduction is an extension to the
1859 current specification in the C++0x working draft, and there was
1860 some concern about potential overload resolution problems.
1862 @item -ffriend-injection
1863 @opindex ffriend-injection
1864 Inject friend functions into the enclosing namespace, so that they are
1865 visible outside the scope of the class in which they are declared.
1866 Friend functions were documented to work this way in the old Annotated
1867 C++ Reference Manual, and versions of G++ before 4.1 always worked
1868 that way. However, in ISO C++ a friend function which is not declared
1869 in an enclosing scope can only be found using argument dependent
1870 lookup. This option causes friends to be injected as they were in
1873 This option is for compatibility, and may be removed in a future
1876 @item -fno-elide-constructors
1877 @opindex fno-elide-constructors
1878 The C++ standard allows an implementation to omit creating a temporary
1879 which is only used to initialize another object of the same type.
1880 Specifying this option disables that optimization, and forces G++ to
1881 call the copy constructor in all cases.
1883 @item -fno-enforce-eh-specs
1884 @opindex fno-enforce-eh-specs
1885 Don't generate code to check for violation of exception specifications
1886 at runtime. This option violates the C++ standard, but may be useful
1887 for reducing code size in production builds, much like defining
1888 @samp{NDEBUG}. This does not give user code permission to throw
1889 exceptions in violation of the exception specifications; the compiler
1890 will still optimize based on the specifications, so throwing an
1891 unexpected exception will result in undefined behavior.
1894 @itemx -fno-for-scope
1896 @opindex fno-for-scope
1897 If @option{-ffor-scope} is specified, the scope of variables declared in
1898 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1899 as specified by the C++ standard.
1900 If @option{-fno-for-scope} is specified, the scope of variables declared in
1901 a @i{for-init-statement} extends to the end of the enclosing scope,
1902 as was the case in old versions of G++, and other (traditional)
1903 implementations of C++.
1905 The default if neither flag is given to follow the standard,
1906 but to allow and give a warning for old-style code that would
1907 otherwise be invalid, or have different behavior.
1909 @item -fno-gnu-keywords
1910 @opindex fno-gnu-keywords
1911 Do not recognize @code{typeof} as a keyword, so that code can use this
1912 word as an identifier. You can use the keyword @code{__typeof__} instead.
1913 @option{-ansi} implies @option{-fno-gnu-keywords}.
1915 @item -fno-implicit-templates
1916 @opindex fno-implicit-templates
1917 Never emit code for non-inline templates which are instantiated
1918 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1919 @xref{Template Instantiation}, for more information.
1921 @item -fno-implicit-inline-templates
1922 @opindex fno-implicit-inline-templates
1923 Don't emit code for implicit instantiations of inline templates, either.
1924 The default is to handle inlines differently so that compiles with and
1925 without optimization will need the same set of explicit instantiations.
1927 @item -fno-implement-inlines
1928 @opindex fno-implement-inlines
1929 To save space, do not emit out-of-line copies of inline functions
1930 controlled by @samp{#pragma implementation}. This will cause linker
1931 errors if these functions are not inlined everywhere they are called.
1933 @item -fms-extensions
1934 @opindex fms-extensions
1935 Disable pedantic warnings about constructs used in MFC, such as implicit
1936 int and getting a pointer to member function via non-standard syntax.
1938 @item -fno-nonansi-builtins
1939 @opindex fno-nonansi-builtins
1940 Disable built-in declarations of functions that are not mandated by
1941 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1942 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1945 @opindex fnothrow-opt
1946 Treat a @code{throw()} exception specification as though it were a
1947 @code{noexcept} specification to reduce or eliminate the text size
1948 overhead relative to a function with no exception specification. If
1949 the function has local variables of types with non-trivial
1950 destructors, the exception specification will actually make the
1951 function smaller because the EH cleanups for those variables can be
1952 optimized away. The semantic effect is that an exception thrown out of
1953 a function with such an exception specification will result in a call
1954 to @code{terminate} rather than @code{unexpected}.
1956 @item -fno-operator-names
1957 @opindex fno-operator-names
1958 Do not treat the operator name keywords @code{and}, @code{bitand},
1959 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1960 synonyms as keywords.
1962 @item -fno-optional-diags
1963 @opindex fno-optional-diags
1964 Disable diagnostics that the standard says a compiler does not need to
1965 issue. Currently, the only such diagnostic issued by G++ is the one for
1966 a name having multiple meanings within a class.
1969 @opindex fpermissive
1970 Downgrade some diagnostics about nonconformant code from errors to
1971 warnings. Thus, using @option{-fpermissive} will allow some
1972 nonconforming code to compile.
1974 @item -fno-pretty-templates
1975 @opindex fno-pretty-templates
1976 When an error message refers to a specialization of a function
1977 template, the compiler will normally print the signature of the
1978 template followed by the template arguments and any typedefs or
1979 typenames in the signature (e.g. @code{void f(T) [with T = int]}
1980 rather than @code{void f(int)}) so that it's clear which template is
1981 involved. When an error message refers to a specialization of a class
1982 template, the compiler will omit any template arguments which match
1983 the default template arguments for that template. If either of these
1984 behaviors make it harder to understand the error message rather than
1985 easier, using @option{-fno-pretty-templates} will disable them.
1989 Enable automatic template instantiation at link time. This option also
1990 implies @option{-fno-implicit-templates}. @xref{Template
1991 Instantiation}, for more information.
1995 Disable generation of information about every class with virtual
1996 functions for use by the C++ runtime type identification features
1997 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1998 of the language, you can save some space by using this flag. Note that
1999 exception handling uses the same information, but it will generate it as
2000 needed. The @samp{dynamic_cast} operator can still be used for casts that
2001 do not require runtime type information, i.e.@: casts to @code{void *} or to
2002 unambiguous base classes.
2006 Emit statistics about front-end processing at the end of the compilation.
2007 This information is generally only useful to the G++ development team.
2009 @item -fstrict-enums
2010 @opindex fstrict-enums
2011 Allow the compiler to optimize using the assumption that a value of
2012 enumeration type can only be one of the values of the enumeration (as
2013 defined in the C++ standard; basically, a value which can be
2014 represented in the minimum number of bits needed to represent all the
2015 enumerators). This assumption may not be valid if the program uses a
2016 cast to convert an arbitrary integer value to the enumeration type.
2018 @item -ftemplate-depth=@var{n}
2019 @opindex ftemplate-depth
2020 Set the maximum instantiation depth for template classes to @var{n}.
2021 A limit on the template instantiation depth is needed to detect
2022 endless recursions during template class instantiation. ANSI/ISO C++
2023 conforming programs must not rely on a maximum depth greater than 17
2024 (changed to 1024 in C++0x).
2026 @item -fno-threadsafe-statics
2027 @opindex fno-threadsafe-statics
2028 Do not emit the extra code to use the routines specified in the C++
2029 ABI for thread-safe initialization of local statics. You can use this
2030 option to reduce code size slightly in code that doesn't need to be
2033 @item -fuse-cxa-atexit
2034 @opindex fuse-cxa-atexit
2035 Register destructors for objects with static storage duration with the
2036 @code{__cxa_atexit} function rather than the @code{atexit} function.
2037 This option is required for fully standards-compliant handling of static
2038 destructors, but will only work if your C library supports
2039 @code{__cxa_atexit}.
2041 @item -fno-use-cxa-get-exception-ptr
2042 @opindex fno-use-cxa-get-exception-ptr
2043 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2044 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
2045 if the runtime routine is not available.
2047 @item -fvisibility-inlines-hidden
2048 @opindex fvisibility-inlines-hidden
2049 This switch declares that the user does not attempt to compare
2050 pointers to inline methods where the addresses of the two functions
2051 were taken in different shared objects.
2053 The effect of this is that GCC may, effectively, mark inline methods with
2054 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2055 appear in the export table of a DSO and do not require a PLT indirection
2056 when used within the DSO@. Enabling this option can have a dramatic effect
2057 on load and link times of a DSO as it massively reduces the size of the
2058 dynamic export table when the library makes heavy use of templates.
2060 The behavior of this switch is not quite the same as marking the
2061 methods as hidden directly, because it does not affect static variables
2062 local to the function or cause the compiler to deduce that
2063 the function is defined in only one shared object.
2065 You may mark a method as having a visibility explicitly to negate the
2066 effect of the switch for that method. For example, if you do want to
2067 compare pointers to a particular inline method, you might mark it as
2068 having default visibility. Marking the enclosing class with explicit
2069 visibility will have no effect.
2071 Explicitly instantiated inline methods are unaffected by this option
2072 as their linkage might otherwise cross a shared library boundary.
2073 @xref{Template Instantiation}.
2075 @item -fvisibility-ms-compat
2076 @opindex fvisibility-ms-compat
2077 This flag attempts to use visibility settings to make GCC's C++
2078 linkage model compatible with that of Microsoft Visual Studio.
2080 The flag makes these changes to GCC's linkage model:
2084 It sets the default visibility to @code{hidden}, like
2085 @option{-fvisibility=hidden}.
2088 Types, but not their members, are not hidden by default.
2091 The One Definition Rule is relaxed for types without explicit
2092 visibility specifications which are defined in more than one different
2093 shared object: those declarations are permitted if they would have
2094 been permitted when this option was not used.
2097 In new code it is better to use @option{-fvisibility=hidden} and
2098 export those classes which are intended to be externally visible.
2099 Unfortunately it is possible for code to rely, perhaps accidentally,
2100 on the Visual Studio behavior.
2102 Among the consequences of these changes are that static data members
2103 of the same type with the same name but defined in different shared
2104 objects will be different, so changing one will not change the other;
2105 and that pointers to function members defined in different shared
2106 objects may not compare equal. When this flag is given, it is a
2107 violation of the ODR to define types with the same name differently.
2111 Do not use weak symbol support, even if it is provided by the linker.
2112 By default, G++ will use weak symbols if they are available. This
2113 option exists only for testing, and should not be used by end-users;
2114 it will result in inferior code and has no benefits. This option may
2115 be removed in a future release of G++.
2119 Do not search for header files in the standard directories specific to
2120 C++, but do still search the other standard directories. (This option
2121 is used when building the C++ library.)
2124 In addition, these optimization, warning, and code generation options
2125 have meanings only for C++ programs:
2128 @item -fno-default-inline
2129 @opindex fno-default-inline
2130 Do not assume @samp{inline} for functions defined inside a class scope.
2131 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2132 functions will have linkage like inline functions; they just won't be
2135 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2138 Warn when G++ generates code that is probably not compatible with the
2139 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2140 all such cases, there are probably some cases that are not warned about,
2141 even though G++ is generating incompatible code. There may also be
2142 cases where warnings are emitted even though the code that is generated
2145 You should rewrite your code to avoid these warnings if you are
2146 concerned about the fact that code generated by G++ may not be binary
2147 compatible with code generated by other compilers.
2149 The known incompatibilities in @option{-fabi-version=2} (the default) include:
2154 A template with a non-type template parameter of reference type is
2155 mangled incorrectly:
2158 template <int &> struct S @{@};
2162 This is fixed in @option{-fabi-version=3}.
2165 SIMD vector types declared using @code{__attribute ((vector_size))} are
2166 mangled in a non-standard way that does not allow for overloading of
2167 functions taking vectors of different sizes.
2169 The mangling is changed in @option{-fabi-version=4}.
2172 The known incompatibilities in @option{-fabi-version=1} include:
2177 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2178 pack data into the same byte as a base class. For example:
2181 struct A @{ virtual void f(); int f1 : 1; @};
2182 struct B : public A @{ int f2 : 1; @};
2186 In this case, G++ will place @code{B::f2} into the same byte
2187 as@code{A::f1}; other compilers will not. You can avoid this problem
2188 by explicitly padding @code{A} so that its size is a multiple of the
2189 byte size on your platform; that will cause G++ and other compilers to
2190 layout @code{B} identically.
2193 Incorrect handling of tail-padding for virtual bases. G++ does not use
2194 tail padding when laying out virtual bases. For example:
2197 struct A @{ virtual void f(); char c1; @};
2198 struct B @{ B(); char c2; @};
2199 struct C : public A, public virtual B @{@};
2203 In this case, G++ will not place @code{B} into the tail-padding for
2204 @code{A}; other compilers will. You can avoid this problem by
2205 explicitly padding @code{A} so that its size is a multiple of its
2206 alignment (ignoring virtual base classes); that will cause G++ and other
2207 compilers to layout @code{C} identically.
2210 Incorrect handling of bit-fields with declared widths greater than that
2211 of their underlying types, when the bit-fields appear in a union. For
2215 union U @{ int i : 4096; @};
2219 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2220 union too small by the number of bits in an @code{int}.
2223 Empty classes can be placed at incorrect offsets. For example:
2233 struct C : public B, public A @{@};
2237 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2238 it should be placed at offset zero. G++ mistakenly believes that the
2239 @code{A} data member of @code{B} is already at offset zero.
2242 Names of template functions whose types involve @code{typename} or
2243 template template parameters can be mangled incorrectly.
2246 template <typename Q>
2247 void f(typename Q::X) @{@}
2249 template <template <typename> class Q>
2250 void f(typename Q<int>::X) @{@}
2254 Instantiations of these templates may be mangled incorrectly.
2258 It also warns psABI related changes. The known psABI changes at this
2264 For SYSV/x86-64, when passing union with long double, it is changed to
2265 pass in memory as specified in psABI. For example:
2275 @code{union U} will always be passed in memory.
2279 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2280 @opindex Wctor-dtor-privacy
2281 @opindex Wno-ctor-dtor-privacy
2282 Warn when a class seems unusable because all the constructors or
2283 destructors in that class are private, and it has neither friends nor
2284 public static member functions.
2286 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2288 @opindex Wno-noexcept
2289 Warn when a noexcept-expression evaluates to false because of a call
2290 to a function that does not have a non-throwing exception
2291 specification (i.e. @samp{throw()} or @samp{noexcept}) but is known by
2292 the compiler to never throw an exception.
2294 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2295 @opindex Wnon-virtual-dtor
2296 @opindex Wno-non-virtual-dtor
2297 Warn when a class has virtual functions and accessible non-virtual
2298 destructor, in which case it would be possible but unsafe to delete
2299 an instance of a derived class through a pointer to the base class.
2300 This warning is also enabled if -Weffc++ is specified.
2302 @item -Wreorder @r{(C++ and Objective-C++ only)}
2304 @opindex Wno-reorder
2305 @cindex reordering, warning
2306 @cindex warning for reordering of member initializers
2307 Warn when the order of member initializers given in the code does not
2308 match the order in which they must be executed. For instance:
2314 A(): j (0), i (1) @{ @}
2318 The compiler will rearrange the member initializers for @samp{i}
2319 and @samp{j} to match the declaration order of the members, emitting
2320 a warning to that effect. This warning is enabled by @option{-Wall}.
2323 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2326 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2329 Warn about violations of the following style guidelines from Scott Meyers'
2330 @cite{Effective C++} book:
2334 Item 11: Define a copy constructor and an assignment operator for classes
2335 with dynamically allocated memory.
2338 Item 12: Prefer initialization to assignment in constructors.
2341 Item 14: Make destructors virtual in base classes.
2344 Item 15: Have @code{operator=} return a reference to @code{*this}.
2347 Item 23: Don't try to return a reference when you must return an object.
2351 Also warn about violations of the following style guidelines from
2352 Scott Meyers' @cite{More Effective C++} book:
2356 Item 6: Distinguish between prefix and postfix forms of increment and
2357 decrement operators.
2360 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2364 When selecting this option, be aware that the standard library
2365 headers do not obey all of these guidelines; use @samp{grep -v}
2366 to filter out those warnings.
2368 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2369 @opindex Wstrict-null-sentinel
2370 @opindex Wno-strict-null-sentinel
2371 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2372 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2373 to @code{__null}. Although it is a null pointer constant not a null pointer,
2374 it is guaranteed to be of the same size as a pointer. But this use is
2375 not portable across different compilers.
2377 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2378 @opindex Wno-non-template-friend
2379 @opindex Wnon-template-friend
2380 Disable warnings when non-templatized friend functions are declared
2381 within a template. Since the advent of explicit template specification
2382 support in G++, if the name of the friend is an unqualified-id (i.e.,
2383 @samp{friend foo(int)}), the C++ language specification demands that the
2384 friend declare or define an ordinary, nontemplate function. (Section
2385 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2386 could be interpreted as a particular specialization of a templatized
2387 function. Because this non-conforming behavior is no longer the default
2388 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2389 check existing code for potential trouble spots and is on by default.
2390 This new compiler behavior can be turned off with
2391 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2392 but disables the helpful warning.
2394 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2395 @opindex Wold-style-cast
2396 @opindex Wno-old-style-cast
2397 Warn if an old-style (C-style) cast to a non-void type is used within
2398 a C++ program. The new-style casts (@samp{dynamic_cast},
2399 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2400 less vulnerable to unintended effects and much easier to search for.
2402 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2403 @opindex Woverloaded-virtual
2404 @opindex Wno-overloaded-virtual
2405 @cindex overloaded virtual fn, warning
2406 @cindex warning for overloaded virtual fn
2407 Warn when a function declaration hides virtual functions from a
2408 base class. For example, in:
2415 struct B: public A @{
2420 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2428 will fail to compile.
2430 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2431 @opindex Wno-pmf-conversions
2432 @opindex Wpmf-conversions
2433 Disable the diagnostic for converting a bound pointer to member function
2436 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2437 @opindex Wsign-promo
2438 @opindex Wno-sign-promo
2439 Warn when overload resolution chooses a promotion from unsigned or
2440 enumerated type to a signed type, over a conversion to an unsigned type of
2441 the same size. Previous versions of G++ would try to preserve
2442 unsignedness, but the standard mandates the current behavior.
2447 A& operator = (int);
2457 In this example, G++ will synthesize a default @samp{A& operator =
2458 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2461 @node Objective-C and Objective-C++ Dialect Options
2462 @section Options Controlling Objective-C and Objective-C++ Dialects
2464 @cindex compiler options, Objective-C and Objective-C++
2465 @cindex Objective-C and Objective-C++ options, command line
2466 @cindex options, Objective-C and Objective-C++
2467 (NOTE: This manual does not describe the Objective-C and Objective-C++
2468 languages themselves. See @xref{Standards,,Language Standards
2469 Supported by GCC}, for references.)
2471 This section describes the command-line options that are only meaningful
2472 for Objective-C and Objective-C++ programs, but you can also use most of
2473 the language-independent GNU compiler options.
2474 For example, you might compile a file @code{some_class.m} like this:
2477 gcc -g -fgnu-runtime -O -c some_class.m
2481 In this example, @option{-fgnu-runtime} is an option meant only for
2482 Objective-C and Objective-C++ programs; you can use the other options with
2483 any language supported by GCC@.
2485 Note that since Objective-C is an extension of the C language, Objective-C
2486 compilations may also use options specific to the C front-end (e.g.,
2487 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2488 C++-specific options (e.g., @option{-Wabi}).
2490 Here is a list of options that are @emph{only} for compiling Objective-C
2491 and Objective-C++ programs:
2494 @item -fconstant-string-class=@var{class-name}
2495 @opindex fconstant-string-class
2496 Use @var{class-name} as the name of the class to instantiate for each
2497 literal string specified with the syntax @code{@@"@dots{}"}. The default
2498 class name is @code{NXConstantString} if the GNU runtime is being used, and
2499 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2500 @option{-fconstant-cfstrings} option, if also present, will override the
2501 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2502 to be laid out as constant CoreFoundation strings.
2505 @opindex fgnu-runtime
2506 Generate object code compatible with the standard GNU Objective-C
2507 runtime. This is the default for most types of systems.
2509 @item -fnext-runtime
2510 @opindex fnext-runtime
2511 Generate output compatible with the NeXT runtime. This is the default
2512 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2513 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2516 @item -fno-nil-receivers
2517 @opindex fno-nil-receivers
2518 Assume that all Objective-C message dispatches (e.g.,
2519 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
2520 is not @code{nil}. This allows for more efficient entry points in the runtime
2521 to be used. Currently, this option is only available in conjunction with
2522 the NeXT runtime on Mac OS X 10.3 and later.
2524 @item -fobjc-call-cxx-cdtors
2525 @opindex fobjc-call-cxx-cdtors
2526 For each Objective-C class, check if any of its instance variables is a
2527 C++ object with a non-trivial default constructor. If so, synthesize a
2528 special @code{- (id) .cxx_construct} instance method that will run
2529 non-trivial default constructors on any such instance variables, in order,
2530 and then return @code{self}. Similarly, check if any instance variable
2531 is a C++ object with a non-trivial destructor, and if so, synthesize a
2532 special @code{- (void) .cxx_destruct} method that will run
2533 all such default destructors, in reverse order.
2535 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
2536 thusly generated will only operate on instance variables declared in the
2537 current Objective-C class, and not those inherited from superclasses. It
2538 is the responsibility of the Objective-C runtime to invoke all such methods
2539 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
2540 will be invoked by the runtime immediately after a new object
2541 instance is allocated; the @code{- (void) .cxx_destruct} methods will
2542 be invoked immediately before the runtime deallocates an object instance.
2544 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2545 support for invoking the @code{- (id) .cxx_construct} and
2546 @code{- (void) .cxx_destruct} methods.
2548 @item -fobjc-direct-dispatch
2549 @opindex fobjc-direct-dispatch
2550 Allow fast jumps to the message dispatcher. On Darwin this is
2551 accomplished via the comm page.
2553 @item -fobjc-exceptions
2554 @opindex fobjc-exceptions
2555 Enable syntactic support for structured exception handling in Objective-C,
2556 similar to what is offered by C++ and Java. This option is
2557 unavailable in conjunction with the NeXT runtime on Mac OS X 10.2 and
2566 @@catch (AnObjCClass *exc) @{
2573 @@catch (AnotherClass *exc) @{
2576 @@catch (id allOthers) @{
2586 The @code{@@throw} statement may appear anywhere in an Objective-C or
2587 Objective-C++ program; when used inside of a @code{@@catch} block, the
2588 @code{@@throw} may appear without an argument (as shown above), in which case
2589 the object caught by the @code{@@catch} will be rethrown.
2591 Note that only (pointers to) Objective-C objects may be thrown and
2592 caught using this scheme. When an object is thrown, it will be caught
2593 by the nearest @code{@@catch} clause capable of handling objects of that type,
2594 analogously to how @code{catch} blocks work in C++ and Java. A
2595 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2596 any and all Objective-C exceptions not caught by previous @code{@@catch}
2599 The @code{@@finally} clause, if present, will be executed upon exit from the
2600 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2601 regardless of whether any exceptions are thrown, caught or rethrown
2602 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2603 of the @code{finally} clause in Java.
2605 There are several caveats to using the new exception mechanism:
2609 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2610 idioms provided by the @code{NSException} class, the new
2611 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2612 systems, due to additional functionality needed in the (NeXT) Objective-C
2616 As mentioned above, the new exceptions do not support handling
2617 types other than Objective-C objects. Furthermore, when used from
2618 Objective-C++, the Objective-C exception model does not interoperate with C++
2619 exceptions at this time. This means you cannot @code{@@throw} an exception
2620 from Objective-C and @code{catch} it in C++, or vice versa
2621 (i.e., @code{throw @dots{} @@catch}).
2624 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2625 blocks for thread-safe execution:
2628 @@synchronized (ObjCClass *guard) @{
2633 Upon entering the @code{@@synchronized} block, a thread of execution shall
2634 first check whether a lock has been placed on the corresponding @code{guard}
2635 object by another thread. If it has, the current thread shall wait until
2636 the other thread relinquishes its lock. Once @code{guard} becomes available,
2637 the current thread will place its own lock on it, execute the code contained in
2638 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2639 making @code{guard} available to other threads).
2641 Unlike Java, Objective-C does not allow for entire methods to be marked
2642 @code{@@synchronized}. Note that throwing exceptions out of
2643 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2644 to be unlocked properly.
2648 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2650 @item -freplace-objc-classes
2651 @opindex freplace-objc-classes
2652 Emit a special marker instructing @command{ld(1)} not to statically link in
2653 the resulting object file, and allow @command{dyld(1)} to load it in at
2654 run time instead. This is used in conjunction with the Fix-and-Continue
2655 debugging mode, where the object file in question may be recompiled and
2656 dynamically reloaded in the course of program execution, without the need
2657 to restart the program itself. Currently, Fix-and-Continue functionality
2658 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2663 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2664 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2665 compile time) with static class references that get initialized at load time,
2666 which improves run-time performance. Specifying the @option{-fzero-link} flag
2667 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2668 to be retained. This is useful in Zero-Link debugging mode, since it allows
2669 for individual class implementations to be modified during program execution.
2673 Dump interface declarations for all classes seen in the source file to a
2674 file named @file{@var{sourcename}.decl}.
2676 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2677 @opindex Wassign-intercept
2678 @opindex Wno-assign-intercept
2679 Warn whenever an Objective-C assignment is being intercepted by the
2682 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2683 @opindex Wno-protocol
2685 If a class is declared to implement a protocol, a warning is issued for
2686 every method in the protocol that is not implemented by the class. The
2687 default behavior is to issue a warning for every method not explicitly
2688 implemented in the class, even if a method implementation is inherited
2689 from the superclass. If you use the @option{-Wno-protocol} option, then
2690 methods inherited from the superclass are considered to be implemented,
2691 and no warning is issued for them.
2693 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2695 @opindex Wno-selector
2696 Warn if multiple methods of different types for the same selector are
2697 found during compilation. The check is performed on the list of methods
2698 in the final stage of compilation. Additionally, a check is performed
2699 for each selector appearing in a @code{@@selector(@dots{})}
2700 expression, and a corresponding method for that selector has been found
2701 during compilation. Because these checks scan the method table only at
2702 the end of compilation, these warnings are not produced if the final
2703 stage of compilation is not reached, for example because an error is
2704 found during compilation, or because the @option{-fsyntax-only} option is
2707 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2708 @opindex Wstrict-selector-match
2709 @opindex Wno-strict-selector-match
2710 Warn if multiple methods with differing argument and/or return types are
2711 found for a given selector when attempting to send a message using this
2712 selector to a receiver of type @code{id} or @code{Class}. When this flag
2713 is off (which is the default behavior), the compiler will omit such warnings
2714 if any differences found are confined to types which share the same size
2717 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2718 @opindex Wundeclared-selector
2719 @opindex Wno-undeclared-selector
2720 Warn if a @code{@@selector(@dots{})} expression referring to an
2721 undeclared selector is found. A selector is considered undeclared if no
2722 method with that name has been declared before the
2723 @code{@@selector(@dots{})} expression, either explicitly in an
2724 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2725 an @code{@@implementation} section. This option always performs its
2726 checks as soon as a @code{@@selector(@dots{})} expression is found,
2727 while @option{-Wselector} only performs its checks in the final stage of
2728 compilation. This also enforces the coding style convention
2729 that methods and selectors must be declared before being used.
2731 @item -print-objc-runtime-info
2732 @opindex print-objc-runtime-info
2733 Generate C header describing the largest structure that is passed by
2738 @node Language Independent Options
2739 @section Options to Control Diagnostic Messages Formatting
2740 @cindex options to control diagnostics formatting
2741 @cindex diagnostic messages
2742 @cindex message formatting
2744 Traditionally, diagnostic messages have been formatted irrespective of
2745 the output device's aspect (e.g.@: its width, @dots{}). The options described
2746 below can be used to control the diagnostic messages formatting
2747 algorithm, e.g.@: how many characters per line, how often source location
2748 information should be reported. Right now, only the C++ front end can
2749 honor these options. However it is expected, in the near future, that
2750 the remaining front ends would be able to digest them correctly.
2753 @item -fmessage-length=@var{n}
2754 @opindex fmessage-length
2755 Try to format error messages so that they fit on lines of about @var{n}
2756 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2757 the front ends supported by GCC@. If @var{n} is zero, then no
2758 line-wrapping will be done; each error message will appear on a single
2761 @opindex fdiagnostics-show-location
2762 @item -fdiagnostics-show-location=once
2763 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2764 reporter to emit @emph{once} source location information; that is, in
2765 case the message is too long to fit on a single physical line and has to
2766 be wrapped, the source location won't be emitted (as prefix) again,
2767 over and over, in subsequent continuation lines. This is the default
2770 @item -fdiagnostics-show-location=every-line
2771 Only meaningful in line-wrapping mode. Instructs the diagnostic
2772 messages reporter to emit the same source location information (as
2773 prefix) for physical lines that result from the process of breaking
2774 a message which is too long to fit on a single line.
2776 @item -fdiagnostics-show-option
2777 @opindex fdiagnostics-show-option
2778 This option instructs the diagnostic machinery to add text to each
2779 diagnostic emitted, which indicates which command line option directly
2780 controls that diagnostic, when such an option is known to the
2781 diagnostic machinery.
2783 @item -Wcoverage-mismatch
2784 @opindex Wcoverage-mismatch
2785 Warn if feedback profiles do not match when using the
2786 @option{-fprofile-use} option.
2787 If a source file was changed between @option{-fprofile-gen} and
2788 @option{-fprofile-use}, the files with the profile feedback can fail
2789 to match the source file and GCC can not use the profile feedback
2790 information. By default, this warning is enabled and is treated as an
2791 error. @option{-Wno-coverage-mismatch} can be used to disable the
2792 warning or @option{-Wno-error=coverage-mismatch} can be used to
2793 disable the error. Disable the error for this warning can result in
2794 poorly optimized code, so disabling the error is useful only in the
2795 case of very minor changes such as bug fixes to an existing code-base.
2796 Completely disabling the warning is not recommended.
2800 @node Warning Options
2801 @section Options to Request or Suppress Warnings
2802 @cindex options to control warnings
2803 @cindex warning messages
2804 @cindex messages, warning
2805 @cindex suppressing warnings
2807 Warnings are diagnostic messages that report constructions which
2808 are not inherently erroneous but which are risky or suggest there
2809 may have been an error.
2811 The following language-independent options do not enable specific
2812 warnings but control the kinds of diagnostics produced by GCC.
2815 @cindex syntax checking
2817 @opindex fsyntax-only
2818 Check the code for syntax errors, but don't do anything beyond that.
2822 Inhibit all warning messages.
2827 Make all warnings into errors.
2832 Make the specified warning into an error. The specifier for a warning
2833 is appended, for example @option{-Werror=switch} turns the warnings
2834 controlled by @option{-Wswitch} into errors. This switch takes a
2835 negative form, to be used to negate @option{-Werror} for specific
2836 warnings, for example @option{-Wno-error=switch} makes
2837 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2838 is in effect. You can use the @option{-fdiagnostics-show-option}
2839 option to have each controllable warning amended with the option which
2840 controls it, to determine what to use with this option.
2842 Note that specifying @option{-Werror=}@var{foo} automatically implies
2843 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2846 @item -Wfatal-errors
2847 @opindex Wfatal-errors
2848 @opindex Wno-fatal-errors
2849 This option causes the compiler to abort compilation on the first error
2850 occurred rather than trying to keep going and printing further error
2855 You can request many specific warnings with options beginning
2856 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2857 implicit declarations. Each of these specific warning options also
2858 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2859 example, @option{-Wno-implicit}. This manual lists only one of the
2860 two forms, whichever is not the default. For further,
2861 language-specific options also refer to @ref{C++ Dialect Options} and
2862 @ref{Objective-C and Objective-C++ Dialect Options}.
2864 When an unrecognized warning label is requested (e.g.,
2865 @option{-Wunknown-warning}), GCC will emit a diagnostic stating
2866 that the option is not recognized. However, if the @samp{-Wno-} form
2867 is used, the behavior is slightly different: No diagnostic will be
2868 produced for @option{-Wno-unknown-warning} unless other diagnostics
2869 are being produced. This allows the use of new @option{-Wno-} options
2870 with old compilers, but if something goes wrong, the compiler will
2871 warn that an unrecognized option was used.
2876 Issue all the warnings demanded by strict ISO C and ISO C++;
2877 reject all programs that use forbidden extensions, and some other
2878 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2879 version of the ISO C standard specified by any @option{-std} option used.
2881 Valid ISO C and ISO C++ programs should compile properly with or without
2882 this option (though a rare few will require @option{-ansi} or a
2883 @option{-std} option specifying the required version of ISO C)@. However,
2884 without this option, certain GNU extensions and traditional C and C++
2885 features are supported as well. With this option, they are rejected.
2887 @option{-pedantic} does not cause warning messages for use of the
2888 alternate keywords whose names begin and end with @samp{__}. Pedantic
2889 warnings are also disabled in the expression that follows
2890 @code{__extension__}. However, only system header files should use
2891 these escape routes; application programs should avoid them.
2892 @xref{Alternate Keywords}.
2894 Some users try to use @option{-pedantic} to check programs for strict ISO
2895 C conformance. They soon find that it does not do quite what they want:
2896 it finds some non-ISO practices, but not all---only those for which
2897 ISO C @emph{requires} a diagnostic, and some others for which
2898 diagnostics have been added.
2900 A feature to report any failure to conform to ISO C might be useful in
2901 some instances, but would require considerable additional work and would
2902 be quite different from @option{-pedantic}. We don't have plans to
2903 support such a feature in the near future.
2905 Where the standard specified with @option{-std} represents a GNU
2906 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
2907 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2908 extended dialect is based. Warnings from @option{-pedantic} are given
2909 where they are required by the base standard. (It would not make sense
2910 for such warnings to be given only for features not in the specified GNU
2911 C dialect, since by definition the GNU dialects of C include all
2912 features the compiler supports with the given option, and there would be
2913 nothing to warn about.)
2915 @item -pedantic-errors
2916 @opindex pedantic-errors
2917 Like @option{-pedantic}, except that errors are produced rather than
2923 This enables all the warnings about constructions that some users
2924 consider questionable, and that are easy to avoid (or modify to
2925 prevent the warning), even in conjunction with macros. This also
2926 enables some language-specific warnings described in @ref{C++ Dialect
2927 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2929 @option{-Wall} turns on the following warning flags:
2931 @gccoptlist{-Waddress @gol
2932 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2934 -Wchar-subscripts @gol
2935 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
2936 -Wimplicit-int @r{(C and Objective-C only)} @gol
2937 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
2940 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2941 -Wmissing-braces @gol
2947 -Wsequence-point @gol
2948 -Wsign-compare @r{(only in C++)} @gol
2949 -Wstrict-aliasing @gol
2950 -Wstrict-overflow=1 @gol
2953 -Wuninitialized @gol
2954 -Wunknown-pragmas @gol
2955 -Wunused-function @gol
2958 -Wunused-variable @gol
2959 -Wvolatile-register-var @gol
2962 Note that some warning flags are not implied by @option{-Wall}. Some of
2963 them warn about constructions that users generally do not consider
2964 questionable, but which occasionally you might wish to check for;
2965 others warn about constructions that are necessary or hard to avoid in
2966 some cases, and there is no simple way to modify the code to suppress
2967 the warning. Some of them are enabled by @option{-Wextra} but many of
2968 them must be enabled individually.
2974 This enables some extra warning flags that are not enabled by
2975 @option{-Wall}. (This option used to be called @option{-W}. The older
2976 name is still supported, but the newer name is more descriptive.)
2978 @gccoptlist{-Wclobbered @gol
2980 -Wignored-qualifiers @gol
2981 -Wmissing-field-initializers @gol
2982 -Wmissing-parameter-type @r{(C only)} @gol
2983 -Wold-style-declaration @r{(C only)} @gol
2984 -Woverride-init @gol
2987 -Wuninitialized @gol
2988 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2989 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2992 The option @option{-Wextra} also prints warning messages for the
2998 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2999 @samp{>}, or @samp{>=}.
3002 (C++ only) An enumerator and a non-enumerator both appear in a
3003 conditional expression.
3006 (C++ only) Ambiguous virtual bases.
3009 (C++ only) Subscripting an array which has been declared @samp{register}.
3012 (C++ only) Taking the address of a variable which has been declared
3016 (C++ only) A base class is not initialized in a derived class' copy
3021 @item -Wchar-subscripts
3022 @opindex Wchar-subscripts
3023 @opindex Wno-char-subscripts
3024 Warn if an array subscript has type @code{char}. This is a common cause
3025 of error, as programmers often forget that this type is signed on some
3027 This warning is enabled by @option{-Wall}.
3031 @opindex Wno-comment
3032 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3033 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3034 This warning is enabled by @option{-Wall}.
3037 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3039 Suppress warning messages emitted by @code{#warning} directives.
3044 @opindex ffreestanding
3045 @opindex fno-builtin
3046 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3047 the arguments supplied have types appropriate to the format string
3048 specified, and that the conversions specified in the format string make
3049 sense. This includes standard functions, and others specified by format
3050 attributes (@pxref{Function Attributes}), in the @code{printf},
3051 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3052 not in the C standard) families (or other target-specific families).
3053 Which functions are checked without format attributes having been
3054 specified depends on the standard version selected, and such checks of
3055 functions without the attribute specified are disabled by
3056 @option{-ffreestanding} or @option{-fno-builtin}.
3058 The formats are checked against the format features supported by GNU
3059 libc version 2.2. These include all ISO C90 and C99 features, as well
3060 as features from the Single Unix Specification and some BSD and GNU
3061 extensions. Other library implementations may not support all these
3062 features; GCC does not support warning about features that go beyond a
3063 particular library's limitations. However, if @option{-pedantic} is used
3064 with @option{-Wformat}, warnings will be given about format features not
3065 in the selected standard version (but not for @code{strfmon} formats,
3066 since those are not in any version of the C standard). @xref{C Dialect
3067 Options,,Options Controlling C Dialect}.
3069 Since @option{-Wformat} also checks for null format arguments for
3070 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
3072 @option{-Wformat} is included in @option{-Wall}. For more control over some
3073 aspects of format checking, the options @option{-Wformat-y2k},
3074 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
3075 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
3076 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
3079 @opindex Wformat-y2k
3080 @opindex Wno-format-y2k
3081 If @option{-Wformat} is specified, also warn about @code{strftime}
3082 formats which may yield only a two-digit year.
3084 @item -Wno-format-contains-nul
3085 @opindex Wno-format-contains-nul
3086 @opindex Wformat-contains-nul
3087 If @option{-Wformat} is specified, do not warn about format strings that
3090 @item -Wno-format-extra-args
3091 @opindex Wno-format-extra-args
3092 @opindex Wformat-extra-args
3093 If @option{-Wformat} is specified, do not warn about excess arguments to a
3094 @code{printf} or @code{scanf} format function. The C standard specifies
3095 that such arguments are ignored.
3097 Where the unused arguments lie between used arguments that are
3098 specified with @samp{$} operand number specifications, normally
3099 warnings are still given, since the implementation could not know what
3100 type to pass to @code{va_arg} to skip the unused arguments. However,
3101 in the case of @code{scanf} formats, this option will suppress the
3102 warning if the unused arguments are all pointers, since the Single
3103 Unix Specification says that such unused arguments are allowed.
3105 @item -Wno-format-zero-length @r{(C and Objective-C only)}
3106 @opindex Wno-format-zero-length
3107 @opindex Wformat-zero-length
3108 If @option{-Wformat} is specified, do not warn about zero-length formats.
3109 The C standard specifies that zero-length formats are allowed.
3111 @item -Wformat-nonliteral
3112 @opindex Wformat-nonliteral
3113 @opindex Wno-format-nonliteral
3114 If @option{-Wformat} is specified, also warn if the format string is not a
3115 string literal and so cannot be checked, unless the format function
3116 takes its format arguments as a @code{va_list}.
3118 @item -Wformat-security
3119 @opindex Wformat-security
3120 @opindex Wno-format-security
3121 If @option{-Wformat} is specified, also warn about uses of format
3122 functions that represent possible security problems. At present, this
3123 warns about calls to @code{printf} and @code{scanf} functions where the
3124 format string is not a string literal and there are no format arguments,
3125 as in @code{printf (foo);}. This may be a security hole if the format
3126 string came from untrusted input and contains @samp{%n}. (This is
3127 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3128 in future warnings may be added to @option{-Wformat-security} that are not
3129 included in @option{-Wformat-nonliteral}.)
3133 @opindex Wno-format=2
3134 Enable @option{-Wformat} plus format checks not included in
3135 @option{-Wformat}. Currently equivalent to @samp{-Wformat
3136 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
3138 @item -Wnonnull @r{(C and Objective-C only)}
3140 @opindex Wno-nonnull
3141 Warn about passing a null pointer for arguments marked as
3142 requiring a non-null value by the @code{nonnull} function attribute.
3144 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3145 can be disabled with the @option{-Wno-nonnull} option.
3147 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3149 @opindex Wno-init-self
3150 Warn about uninitialized variables which are initialized with themselves.
3151 Note this option can only be used with the @option{-Wuninitialized} option.
3153 For example, GCC will warn about @code{i} being uninitialized in the
3154 following snippet only when @option{-Winit-self} has been specified:
3165 @item -Wimplicit-int @r{(C and Objective-C only)}
3166 @opindex Wimplicit-int
3167 @opindex Wno-implicit-int
3168 Warn when a declaration does not specify a type.
3169 This warning is enabled by @option{-Wall}.
3171 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3172 @opindex Wimplicit-function-declaration
3173 @opindex Wno-implicit-function-declaration
3174 Give a warning whenever a function is used before being declared. In
3175 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3176 enabled by default and it is made into an error by
3177 @option{-pedantic-errors}. This warning is also enabled by
3180 @item -Wimplicit @r{(C and Objective-C only)}
3182 @opindex Wno-implicit
3183 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3184 This warning is enabled by @option{-Wall}.
3186 @item -Wignored-qualifiers @r{(C and C++ only)}
3187 @opindex Wignored-qualifiers
3188 @opindex Wno-ignored-qualifiers
3189 Warn if the return type of a function has a type qualifier
3190 such as @code{const}. For ISO C such a type qualifier has no effect,
3191 since the value returned by a function is not an lvalue.
3192 For C++, the warning is only emitted for scalar types or @code{void}.
3193 ISO C prohibits qualified @code{void} return types on function
3194 definitions, so such return types always receive a warning
3195 even without this option.
3197 This warning is also enabled by @option{-Wextra}.
3202 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3203 a function with external linkage, returning int, taking either zero
3204 arguments, two, or three arguments of appropriate types. This warning
3205 is enabled by default in C++ and is enabled by either @option{-Wall}
3206 or @option{-pedantic}.
3208 @item -Wmissing-braces
3209 @opindex Wmissing-braces
3210 @opindex Wno-missing-braces
3211 Warn if an aggregate or union initializer is not fully bracketed. In
3212 the following example, the initializer for @samp{a} is not fully
3213 bracketed, but that for @samp{b} is fully bracketed.
3216 int a[2][2] = @{ 0, 1, 2, 3 @};
3217 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3220 This warning is enabled by @option{-Wall}.
3222 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3223 @opindex Wmissing-include-dirs
3224 @opindex Wno-missing-include-dirs
3225 Warn if a user-supplied include directory does not exist.
3228 @opindex Wparentheses
3229 @opindex Wno-parentheses
3230 Warn if parentheses are omitted in certain contexts, such
3231 as when there is an assignment in a context where a truth value
3232 is expected, or when operators are nested whose precedence people
3233 often get confused about.
3235 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3236 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3237 interpretation from that of ordinary mathematical notation.
3239 Also warn about constructions where there may be confusion to which
3240 @code{if} statement an @code{else} branch belongs. Here is an example of
3255 In C/C++, every @code{else} branch belongs to the innermost possible
3256 @code{if} statement, which in this example is @code{if (b)}. This is
3257 often not what the programmer expected, as illustrated in the above
3258 example by indentation the programmer chose. When there is the
3259 potential for this confusion, GCC will issue a warning when this flag
3260 is specified. To eliminate the warning, add explicit braces around
3261 the innermost @code{if} statement so there is no way the @code{else}
3262 could belong to the enclosing @code{if}. The resulting code would
3279 Also warn for dangerous uses of the
3280 ?: with omitted middle operand GNU extension. When the condition
3281 in the ?: operator is a boolean expression the omitted value will
3282 be always 1. Often the user expects it to be a value computed
3283 inside the conditional expression instead.
3285 This warning is enabled by @option{-Wall}.
3287 @item -Wsequence-point
3288 @opindex Wsequence-point
3289 @opindex Wno-sequence-point
3290 Warn about code that may have undefined semantics because of violations
3291 of sequence point rules in the C and C++ standards.
3293 The C and C++ standards defines the order in which expressions in a C/C++
3294 program are evaluated in terms of @dfn{sequence points}, which represent
3295 a partial ordering between the execution of parts of the program: those
3296 executed before the sequence point, and those executed after it. These
3297 occur after the evaluation of a full expression (one which is not part
3298 of a larger expression), after the evaluation of the first operand of a
3299 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3300 function is called (but after the evaluation of its arguments and the
3301 expression denoting the called function), and in certain other places.
3302 Other than as expressed by the sequence point rules, the order of
3303 evaluation of subexpressions of an expression is not specified. All
3304 these rules describe only a partial order rather than a total order,
3305 since, for example, if two functions are called within one expression
3306 with no sequence point between them, the order in which the functions
3307 are called is not specified. However, the standards committee have
3308 ruled that function calls do not overlap.
3310 It is not specified when between sequence points modifications to the
3311 values of objects take effect. Programs whose behavior depends on this
3312 have undefined behavior; the C and C++ standards specify that ``Between
3313 the previous and next sequence point an object shall have its stored
3314 value modified at most once by the evaluation of an expression.
3315 Furthermore, the prior value shall be read only to determine the value
3316 to be stored.''. If a program breaks these rules, the results on any
3317 particular implementation are entirely unpredictable.
3319 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3320 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3321 diagnosed by this option, and it may give an occasional false positive
3322 result, but in general it has been found fairly effective at detecting
3323 this sort of problem in programs.
3325 The standard is worded confusingly, therefore there is some debate
3326 over the precise meaning of the sequence point rules in subtle cases.
3327 Links to discussions of the problem, including proposed formal
3328 definitions, may be found on the GCC readings page, at
3329 @w{@uref{http://gcc.gnu.org/readings.html}}.
3331 This warning is enabled by @option{-Wall} for C and C++.
3334 @opindex Wreturn-type
3335 @opindex Wno-return-type
3336 Warn whenever a function is defined with a return-type that defaults
3337 to @code{int}. Also warn about any @code{return} statement with no
3338 return-value in a function whose return-type is not @code{void}
3339 (falling off the end of the function body is considered returning
3340 without a value), and about a @code{return} statement with an
3341 expression in a function whose return-type is @code{void}.
3343 For C++, a function without return type always produces a diagnostic
3344 message, even when @option{-Wno-return-type} is specified. The only
3345 exceptions are @samp{main} and functions defined in system headers.
3347 This warning is enabled by @option{-Wall}.
3352 Warn whenever a @code{switch} statement has an index of enumerated type
3353 and lacks a @code{case} for one or more of the named codes of that
3354 enumeration. (The presence of a @code{default} label prevents this
3355 warning.) @code{case} labels outside the enumeration range also
3356 provoke warnings when this option is used (even if there is a
3357 @code{default} label).
3358 This warning is enabled by @option{-Wall}.
3360 @item -Wswitch-default
3361 @opindex Wswitch-default
3362 @opindex Wno-switch-default
3363 Warn whenever a @code{switch} statement does not have a @code{default}
3367 @opindex Wswitch-enum
3368 @opindex Wno-switch-enum
3369 Warn whenever a @code{switch} statement has an index of enumerated type
3370 and lacks a @code{case} for one or more of the named codes of that
3371 enumeration. @code{case} labels outside the enumeration range also
3372 provoke warnings when this option is used. The only difference
3373 between @option{-Wswitch} and this option is that this option gives a
3374 warning about an omitted enumeration code even if there is a
3375 @code{default} label.
3377 @item -Wsync-nand @r{(C and C++ only)}
3379 @opindex Wno-sync-nand
3380 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3381 built-in functions are used. These functions changed semantics in GCC 4.4.
3385 @opindex Wno-trigraphs
3386 Warn if any trigraphs are encountered that might change the meaning of
3387 the program (trigraphs within comments are not warned about).
3388 This warning is enabled by @option{-Wall}.
3390 @item -Wunused-but-set-parameter
3391 @opindex Wunused-but-set-parameter
3392 @opindex Wno-unused-but-set-parameter
3393 Warn whenever a function parameter is assigned to, but otherwise unused
3394 (aside from its declaration).
3396 To suppress this warning use the @samp{unused} attribute
3397 (@pxref{Variable Attributes}).
3399 This warning is also enabled by @option{-Wunused} together with
3402 @item -Wunused-but-set-variable
3403 @opindex Wunused-but-set-variable
3404 @opindex Wno-unused-but-set-variable
3405 Warn whenever a local variable is assigned to, but otherwise unused
3406 (aside from its declaration).
3407 This warning is enabled by @option{-Wall}.
3409 To suppress this warning use the @samp{unused} attribute
3410 (@pxref{Variable Attributes}).
3412 This warning is also enabled by @option{-Wunused}, which is enabled
3415 @item -Wunused-function
3416 @opindex Wunused-function
3417 @opindex Wno-unused-function
3418 Warn whenever a static function is declared but not defined or a
3419 non-inline static function is unused.
3420 This warning is enabled by @option{-Wall}.
3422 @item -Wunused-label
3423 @opindex Wunused-label
3424 @opindex Wno-unused-label
3425 Warn whenever a label is declared but not used.
3426 This warning is enabled by @option{-Wall}.
3428 To suppress this warning use the @samp{unused} attribute
3429 (@pxref{Variable Attributes}).
3431 @item -Wunused-parameter
3432 @opindex Wunused-parameter
3433 @opindex Wno-unused-parameter
3434 Warn whenever a function parameter is unused aside from its declaration.
3436 To suppress this warning use the @samp{unused} attribute
3437 (@pxref{Variable Attributes}).
3439 @item -Wno-unused-result
3440 @opindex Wunused-result
3441 @opindex Wno-unused-result
3442 Do not warn if a caller of a function marked with attribute
3443 @code{warn_unused_result} (@pxref{Variable Attributes}) does not use
3444 its return value. The default is @option{-Wunused-result}.
3446 @item -Wunused-variable
3447 @opindex Wunused-variable
3448 @opindex Wno-unused-variable
3449 Warn whenever a local variable or non-constant static variable is unused
3450 aside from its declaration.
3451 This warning is enabled by @option{-Wall}.
3453 To suppress this warning use the @samp{unused} attribute
3454 (@pxref{Variable Attributes}).
3456 @item -Wunused-value
3457 @opindex Wunused-value
3458 @opindex Wno-unused-value
3459 Warn whenever a statement computes a result that is explicitly not
3460 used. To suppress this warning cast the unused expression to
3461 @samp{void}. This includes an expression-statement or the left-hand
3462 side of a comma expression that contains no side effects. For example,
3463 an expression such as @samp{x[i,j]} will cause a warning, while
3464 @samp{x[(void)i,j]} will not.
3466 This warning is enabled by @option{-Wall}.
3471 All the above @option{-Wunused} options combined.
3473 In order to get a warning about an unused function parameter, you must
3474 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3475 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3477 @item -Wuninitialized
3478 @opindex Wuninitialized
3479 @opindex Wno-uninitialized
3480 Warn if an automatic variable is used without first being initialized
3481 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3482 warn if a non-static reference or non-static @samp{const} member
3483 appears in a class without constructors.
3485 If you want to warn about code which uses the uninitialized value of the
3486 variable in its own initializer, use the @option{-Winit-self} option.
3488 These warnings occur for individual uninitialized or clobbered
3489 elements of structure, union or array variables as well as for
3490 variables which are uninitialized or clobbered as a whole. They do
3491 not occur for variables or elements declared @code{volatile}. Because
3492 these warnings depend on optimization, the exact variables or elements
3493 for which there are warnings will depend on the precise optimization
3494 options and version of GCC used.
3496 Note that there may be no warning about a variable that is used only
3497 to compute a value that itself is never used, because such
3498 computations may be deleted by data flow analysis before the warnings
3501 These warnings are made optional because GCC is not smart
3502 enough to see all the reasons why the code might be correct
3503 despite appearing to have an error. Here is one example of how
3524 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3525 always initialized, but GCC doesn't know this. Here is
3526 another common case:
3531 if (change_y) save_y = y, y = new_y;
3533 if (change_y) y = save_y;
3538 This has no bug because @code{save_y} is used only if it is set.
3540 @cindex @code{longjmp} warnings
3541 This option also warns when a non-volatile automatic variable might be
3542 changed by a call to @code{longjmp}. These warnings as well are possible
3543 only in optimizing compilation.
3545 The compiler sees only the calls to @code{setjmp}. It cannot know
3546 where @code{longjmp} will be called; in fact, a signal handler could
3547 call it at any point in the code. As a result, you may get a warning
3548 even when there is in fact no problem because @code{longjmp} cannot
3549 in fact be called at the place which would cause a problem.
3551 Some spurious warnings can be avoided if you declare all the functions
3552 you use that never return as @code{noreturn}. @xref{Function
3555 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3557 @item -Wunknown-pragmas
3558 @opindex Wunknown-pragmas
3559 @opindex Wno-unknown-pragmas
3560 @cindex warning for unknown pragmas
3561 @cindex unknown pragmas, warning
3562 @cindex pragmas, warning of unknown
3563 Warn when a #pragma directive is encountered which is not understood by
3564 GCC@. If this command line option is used, warnings will even be issued
3565 for unknown pragmas in system header files. This is not the case if
3566 the warnings were only enabled by the @option{-Wall} command line option.
3569 @opindex Wno-pragmas
3571 Do not warn about misuses of pragmas, such as incorrect parameters,
3572 invalid syntax, or conflicts between pragmas. See also
3573 @samp{-Wunknown-pragmas}.
3575 @item -Wstrict-aliasing
3576 @opindex Wstrict-aliasing
3577 @opindex Wno-strict-aliasing
3578 This option is only active when @option{-fstrict-aliasing} is active.
3579 It warns about code which might break the strict aliasing rules that the
3580 compiler is using for optimization. The warning does not catch all
3581 cases, but does attempt to catch the more common pitfalls. It is
3582 included in @option{-Wall}.
3583 It is equivalent to @option{-Wstrict-aliasing=3}
3585 @item -Wstrict-aliasing=n
3586 @opindex Wstrict-aliasing=n
3587 @opindex Wno-strict-aliasing=n
3588 This option is only active when @option{-fstrict-aliasing} is active.
3589 It warns about code which might break the strict aliasing rules that the
3590 compiler is using for optimization.
3591 Higher levels correspond to higher accuracy (fewer false positives).
3592 Higher levels also correspond to more effort, similar to the way -O works.
3593 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3596 Level 1: Most aggressive, quick, least accurate.
3597 Possibly useful when higher levels
3598 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3599 false negatives. However, it has many false positives.
3600 Warns for all pointer conversions between possibly incompatible types,
3601 even if never dereferenced. Runs in the frontend only.
3603 Level 2: Aggressive, quick, not too precise.
3604 May still have many false positives (not as many as level 1 though),
3605 and few false negatives (but possibly more than level 1).
3606 Unlike level 1, it only warns when an address is taken. Warns about
3607 incomplete types. Runs in the frontend only.
3609 Level 3 (default for @option{-Wstrict-aliasing}):
3610 Should have very few false positives and few false
3611 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3612 Takes care of the common pun+dereference pattern in the frontend:
3613 @code{*(int*)&some_float}.
3614 If optimization is enabled, it also runs in the backend, where it deals
3615 with multiple statement cases using flow-sensitive points-to information.
3616 Only warns when the converted pointer is dereferenced.
3617 Does not warn about incomplete types.
3619 @item -Wstrict-overflow
3620 @itemx -Wstrict-overflow=@var{n}
3621 @opindex Wstrict-overflow
3622 @opindex Wno-strict-overflow
3623 This option is only active when @option{-fstrict-overflow} is active.
3624 It warns about cases where the compiler optimizes based on the
3625 assumption that signed overflow does not occur. Note that it does not
3626 warn about all cases where the code might overflow: it only warns
3627 about cases where the compiler implements some optimization. Thus
3628 this warning depends on the optimization level.
3630 An optimization which assumes that signed overflow does not occur is
3631 perfectly safe if the values of the variables involved are such that
3632 overflow never does, in fact, occur. Therefore this warning can
3633 easily give a false positive: a warning about code which is not
3634 actually a problem. To help focus on important issues, several
3635 warning levels are defined. No warnings are issued for the use of
3636 undefined signed overflow when estimating how many iterations a loop
3637 will require, in particular when determining whether a loop will be
3641 @item -Wstrict-overflow=1
3642 Warn about cases which are both questionable and easy to avoid. For
3643 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3644 compiler will simplify this to @code{1}. This level of
3645 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3646 are not, and must be explicitly requested.
3648 @item -Wstrict-overflow=2
3649 Also warn about other cases where a comparison is simplified to a
3650 constant. For example: @code{abs (x) >= 0}. This can only be
3651 simplified when @option{-fstrict-overflow} is in effect, because
3652 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3653 zero. @option{-Wstrict-overflow} (with no level) is the same as
3654 @option{-Wstrict-overflow=2}.
3656 @item -Wstrict-overflow=3
3657 Also warn about other cases where a comparison is simplified. For
3658 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3660 @item -Wstrict-overflow=4
3661 Also warn about other simplifications not covered by the above cases.
3662 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3664 @item -Wstrict-overflow=5
3665 Also warn about cases where the compiler reduces the magnitude of a
3666 constant involved in a comparison. For example: @code{x + 2 > y} will
3667 be simplified to @code{x + 1 >= y}. This is reported only at the
3668 highest warning level because this simplification applies to many
3669 comparisons, so this warning level will give a very large number of
3673 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]}
3674 @opindex Wsuggest-attribute=
3675 @opindex Wno-suggest-attribute=
3676 Warn for cases where adding an attribute may be beneficial. The
3677 attributes currently supported are listed below.
3680 @item -Wsuggest-attribute=pure
3681 @itemx -Wsuggest-attribute=const
3682 @itemx -Wsuggest-attribute=noreturn
3683 @opindex Wsuggest-attribute=pure
3684 @opindex Wno-suggest-attribute=pure
3685 @opindex Wsuggest-attribute=const
3686 @opindex Wno-suggest-attribute=const
3687 @opindex Wsuggest-attribute=noreturn
3688 @opindex Wno-suggest-attribute=noreturn
3690 Warn about functions which might be candidates for attributes
3691 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
3692 functions visible in other compilation units or (in the case of @code{pure} and
3693 @code{const}) if it cannot prove that the function returns normally. A function
3694 returns normally if it doesn't contain an infinite loop nor returns abnormally
3695 by throwing, calling @code{abort()} or trapping. This analysis requires option
3696 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
3697 higher. Higher optimization levels improve the accuracy of the analysis.
3700 @item -Warray-bounds
3701 @opindex Wno-array-bounds
3702 @opindex Warray-bounds
3703 This option is only active when @option{-ftree-vrp} is active
3704 (default for @option{-O2} and above). It warns about subscripts to arrays
3705 that are always out of bounds. This warning is enabled by @option{-Wall}.
3707 @item -Wno-div-by-zero
3708 @opindex Wno-div-by-zero
3709 @opindex Wdiv-by-zero
3710 Do not warn about compile-time integer division by zero. Floating point
3711 division by zero is not warned about, as it can be a legitimate way of
3712 obtaining infinities and NaNs.
3714 @item -Wsystem-headers
3715 @opindex Wsystem-headers
3716 @opindex Wno-system-headers
3717 @cindex warnings from system headers
3718 @cindex system headers, warnings from
3719 Print warning messages for constructs found in system header files.
3720 Warnings from system headers are normally suppressed, on the assumption
3721 that they usually do not indicate real problems and would only make the
3722 compiler output harder to read. Using this command line option tells
3723 GCC to emit warnings from system headers as if they occurred in user
3724 code. However, note that using @option{-Wall} in conjunction with this
3725 option will @emph{not} warn about unknown pragmas in system
3726 headers---for that, @option{-Wunknown-pragmas} must also be used.
3729 @opindex Wtrampolines
3730 @opindex Wno-trampolines
3731 Warn about trampolines generated for pointers to nested functions.
3733 A trampoline is a small piece of data or code that is created at run
3734 time on the stack when the address of a nested function is taken, and
3735 is used to call the nested function indirectly. For some targets, it
3736 is made up of data only and thus requires no special treatment. But,
3737 for most targets, it is made up of code and thus requires the stack
3738 to be made executable in order for the program to work properly.
3741 @opindex Wfloat-equal
3742 @opindex Wno-float-equal
3743 Warn if floating point values are used in equality comparisons.
3745 The idea behind this is that sometimes it is convenient (for the
3746 programmer) to consider floating-point values as approximations to
3747 infinitely precise real numbers. If you are doing this, then you need
3748 to compute (by analyzing the code, or in some other way) the maximum or
3749 likely maximum error that the computation introduces, and allow for it
3750 when performing comparisons (and when producing output, but that's a
3751 different problem). In particular, instead of testing for equality, you
3752 would check to see whether the two values have ranges that overlap; and
3753 this is done with the relational operators, so equality comparisons are
3756 @item -Wtraditional @r{(C and Objective-C only)}
3757 @opindex Wtraditional
3758 @opindex Wno-traditional
3759 Warn about certain constructs that behave differently in traditional and
3760 ISO C@. Also warn about ISO C constructs that have no traditional C
3761 equivalent, and/or problematic constructs which should be avoided.
3765 Macro parameters that appear within string literals in the macro body.
3766 In traditional C macro replacement takes place within string literals,
3767 but does not in ISO C@.
3770 In traditional C, some preprocessor directives did not exist.
3771 Traditional preprocessors would only consider a line to be a directive
3772 if the @samp{#} appeared in column 1 on the line. Therefore
3773 @option{-Wtraditional} warns about directives that traditional C
3774 understands but would ignore because the @samp{#} does not appear as the
3775 first character on the line. It also suggests you hide directives like
3776 @samp{#pragma} not understood by traditional C by indenting them. Some
3777 traditional implementations would not recognize @samp{#elif}, so it
3778 suggests avoiding it altogether.
3781 A function-like macro that appears without arguments.
3784 The unary plus operator.
3787 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3788 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3789 constants.) Note, these suffixes appear in macros defined in the system
3790 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3791 Use of these macros in user code might normally lead to spurious
3792 warnings, however GCC's integrated preprocessor has enough context to
3793 avoid warning in these cases.
3796 A function declared external in one block and then used after the end of
3800 A @code{switch} statement has an operand of type @code{long}.
3803 A non-@code{static} function declaration follows a @code{static} one.
3804 This construct is not accepted by some traditional C compilers.
3807 The ISO type of an integer constant has a different width or
3808 signedness from its traditional type. This warning is only issued if
3809 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3810 typically represent bit patterns, are not warned about.
3813 Usage of ISO string concatenation is detected.
3816 Initialization of automatic aggregates.
3819 Identifier conflicts with labels. Traditional C lacks a separate
3820 namespace for labels.
3823 Initialization of unions. If the initializer is zero, the warning is
3824 omitted. This is done under the assumption that the zero initializer in
3825 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3826 initializer warnings and relies on default initialization to zero in the
3830 Conversions by prototypes between fixed/floating point values and vice
3831 versa. The absence of these prototypes when compiling with traditional
3832 C would cause serious problems. This is a subset of the possible
3833 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3836 Use of ISO C style function definitions. This warning intentionally is
3837 @emph{not} issued for prototype declarations or variadic functions
3838 because these ISO C features will appear in your code when using
3839 libiberty's traditional C compatibility macros, @code{PARAMS} and
3840 @code{VPARAMS}. This warning is also bypassed for nested functions
3841 because that feature is already a GCC extension and thus not relevant to
3842 traditional C compatibility.
3845 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3846 @opindex Wtraditional-conversion
3847 @opindex Wno-traditional-conversion
3848 Warn if a prototype causes a type conversion that is different from what
3849 would happen to the same argument in the absence of a prototype. This
3850 includes conversions of fixed point to floating and vice versa, and
3851 conversions changing the width or signedness of a fixed point argument
3852 except when the same as the default promotion.
3854 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3855 @opindex Wdeclaration-after-statement
3856 @opindex Wno-declaration-after-statement
3857 Warn when a declaration is found after a statement in a block. This
3858 construct, known from C++, was introduced with ISO C99 and is by default
3859 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3860 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3865 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3867 @item -Wno-endif-labels
3868 @opindex Wno-endif-labels
3869 @opindex Wendif-labels
3870 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3875 Warn whenever a local variable or type declaration shadows another variable,
3876 parameter, type, or class member (in C++), or whenever a built-in function
3877 is shadowed. Note that in C++, the compiler will not warn if a local variable
3878 shadows a struct/class/enum, but will warn if it shadows an explicit typedef.
3880 @item -Wlarger-than=@var{len}
3881 @opindex Wlarger-than=@var{len}
3882 @opindex Wlarger-than-@var{len}
3883 Warn whenever an object of larger than @var{len} bytes is defined.
3885 @item -Wframe-larger-than=@var{len}
3886 @opindex Wframe-larger-than
3887 Warn if the size of a function frame is larger than @var{len} bytes.
3888 The computation done to determine the stack frame size is approximate
3889 and not conservative.
3890 The actual requirements may be somewhat greater than @var{len}
3891 even if you do not get a warning. In addition, any space allocated
3892 via @code{alloca}, variable-length arrays, or related constructs
3893 is not included by the compiler when determining
3894 whether or not to issue a warning.
3896 @item -Wunsafe-loop-optimizations
3897 @opindex Wunsafe-loop-optimizations
3898 @opindex Wno-unsafe-loop-optimizations
3899 Warn if the loop cannot be optimized because the compiler could not
3900 assume anything on the bounds of the loop indices. With
3901 @option{-funsafe-loop-optimizations} warn if the compiler made
3904 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
3905 @opindex Wno-pedantic-ms-format
3906 @opindex Wpedantic-ms-format
3907 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
3908 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
3909 depending on the MS runtime, when you are using the options @option{-Wformat}
3910 and @option{-pedantic} without gnu-extensions.
3912 @item -Wpointer-arith
3913 @opindex Wpointer-arith
3914 @opindex Wno-pointer-arith
3915 Warn about anything that depends on the ``size of'' a function type or
3916 of @code{void}. GNU C assigns these types a size of 1, for
3917 convenience in calculations with @code{void *} pointers and pointers
3918 to functions. In C++, warn also when an arithmetic operation involves
3919 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3922 @opindex Wtype-limits
3923 @opindex Wno-type-limits
3924 Warn if a comparison is always true or always false due to the limited
3925 range of the data type, but do not warn for constant expressions. For
3926 example, warn if an unsigned variable is compared against zero with
3927 @samp{<} or @samp{>=}. This warning is also enabled by
3930 @item -Wbad-function-cast @r{(C and Objective-C only)}
3931 @opindex Wbad-function-cast
3932 @opindex Wno-bad-function-cast
3933 Warn whenever a function call is cast to a non-matching type.
3934 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3936 @item -Wc++-compat @r{(C and Objective-C only)}
3937 Warn about ISO C constructs that are outside of the common subset of
3938 ISO C and ISO C++, e.g.@: request for implicit conversion from
3939 @code{void *} to a pointer to non-@code{void} type.
3941 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3942 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3943 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3944 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3948 @opindex Wno-cast-qual
3949 Warn whenever a pointer is cast so as to remove a type qualifier from
3950 the target type. For example, warn if a @code{const char *} is cast
3951 to an ordinary @code{char *}.
3953 Also warn when making a cast which introduces a type qualifier in an
3954 unsafe way. For example, casting @code{char **} to @code{const char **}
3955 is unsafe, as in this example:
3958 /* p is char ** value. */
3959 const char **q = (const char **) p;
3960 /* Assignment of readonly string to const char * is OK. */
3962 /* Now char** pointer points to read-only memory. */
3967 @opindex Wcast-align
3968 @opindex Wno-cast-align
3969 Warn whenever a pointer is cast such that the required alignment of the
3970 target is increased. For example, warn if a @code{char *} is cast to
3971 an @code{int *} on machines where integers can only be accessed at
3972 two- or four-byte boundaries.
3974 @item -Wwrite-strings
3975 @opindex Wwrite-strings
3976 @opindex Wno-write-strings
3977 When compiling C, give string constants the type @code{const
3978 char[@var{length}]} so that copying the address of one into a
3979 non-@code{const} @code{char *} pointer will get a warning. These
3980 warnings will help you find at compile time code that can try to write
3981 into a string constant, but only if you have been very careful about
3982 using @code{const} in declarations and prototypes. Otherwise, it will
3983 just be a nuisance. This is why we did not make @option{-Wall} request
3986 When compiling C++, warn about the deprecated conversion from string
3987 literals to @code{char *}. This warning is enabled by default for C++
3992 @opindex Wno-clobbered
3993 Warn for variables that might be changed by @samp{longjmp} or
3994 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
3997 @opindex Wconversion
3998 @opindex Wno-conversion
3999 Warn for implicit conversions that may alter a value. This includes
4000 conversions between real and integer, like @code{abs (x)} when
4001 @code{x} is @code{double}; conversions between signed and unsigned,
4002 like @code{unsigned ui = -1}; and conversions to smaller types, like
4003 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
4004 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
4005 changed by the conversion like in @code{abs (2.0)}. Warnings about
4006 conversions between signed and unsigned integers can be disabled by
4007 using @option{-Wno-sign-conversion}.
4009 For C++, also warn for confusing overload resolution for user-defined
4010 conversions; and conversions that will never use a type conversion
4011 operator: conversions to @code{void}, the same type, a base class or a
4012 reference to them. Warnings about conversions between signed and
4013 unsigned integers are disabled by default in C++ unless
4014 @option{-Wsign-conversion} is explicitly enabled.
4016 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
4017 @opindex Wconversion-null
4018 @opindex Wno-conversion-null
4019 Do not warn for conversions between @code{NULL} and non-pointer
4020 types. @option{-Wconversion-null} is enabled by default.
4023 @opindex Wempty-body
4024 @opindex Wno-empty-body
4025 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
4026 while} statement. This warning is also enabled by @option{-Wextra}.
4028 @item -Wenum-compare
4029 @opindex Wenum-compare
4030 @opindex Wno-enum-compare
4031 Warn about a comparison between values of different enum types. In C++
4032 this warning is enabled by default. In C this warning is enabled by
4035 @item -Wjump-misses-init @r{(C, Objective-C only)}
4036 @opindex Wjump-misses-init
4037 @opindex Wno-jump-misses-init
4038 Warn if a @code{goto} statement or a @code{switch} statement jumps
4039 forward across the initialization of a variable, or jumps backward to a
4040 label after the variable has been initialized. This only warns about
4041 variables which are initialized when they are declared. This warning is
4042 only supported for C and Objective C; in C++ this sort of branch is an
4045 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
4046 can be disabled with the @option{-Wno-jump-misses-init} option.
4048 @item -Wsign-compare
4049 @opindex Wsign-compare
4050 @opindex Wno-sign-compare
4051 @cindex warning for comparison of signed and unsigned values
4052 @cindex comparison of signed and unsigned values, warning
4053 @cindex signed and unsigned values, comparison warning
4054 Warn when a comparison between signed and unsigned values could produce
4055 an incorrect result when the signed value is converted to unsigned.
4056 This warning is also enabled by @option{-Wextra}; to get the other warnings
4057 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
4059 @item -Wsign-conversion
4060 @opindex Wsign-conversion
4061 @opindex Wno-sign-conversion
4062 Warn for implicit conversions that may change the sign of an integer
4063 value, like assigning a signed integer expression to an unsigned
4064 integer variable. An explicit cast silences the warning. In C, this
4065 option is enabled also by @option{-Wconversion}.
4069 @opindex Wno-address
4070 Warn about suspicious uses of memory addresses. These include using
4071 the address of a function in a conditional expression, such as
4072 @code{void func(void); if (func)}, and comparisons against the memory
4073 address of a string literal, such as @code{if (x == "abc")}. Such
4074 uses typically indicate a programmer error: the address of a function
4075 always evaluates to true, so their use in a conditional usually
4076 indicate that the programmer forgot the parentheses in a function
4077 call; and comparisons against string literals result in unspecified
4078 behavior and are not portable in C, so they usually indicate that the
4079 programmer intended to use @code{strcmp}. This warning is enabled by
4083 @opindex Wlogical-op
4084 @opindex Wno-logical-op
4085 Warn about suspicious uses of logical operators in expressions.
4086 This includes using logical operators in contexts where a
4087 bit-wise operator is likely to be expected.
4089 @item -Waggregate-return
4090 @opindex Waggregate-return
4091 @opindex Wno-aggregate-return
4092 Warn if any functions that return structures or unions are defined or
4093 called. (In languages where you can return an array, this also elicits
4096 @item -Wno-attributes
4097 @opindex Wno-attributes
4098 @opindex Wattributes
4099 Do not warn if an unexpected @code{__attribute__} is used, such as
4100 unrecognized attributes, function attributes applied to variables,
4101 etc. This will not stop errors for incorrect use of supported
4104 @item -Wno-builtin-macro-redefined
4105 @opindex Wno-builtin-macro-redefined
4106 @opindex Wbuiltin-macro-redefined
4107 Do not warn if certain built-in macros are redefined. This suppresses
4108 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
4109 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
4111 @item -Wstrict-prototypes @r{(C and Objective-C only)}
4112 @opindex Wstrict-prototypes
4113 @opindex Wno-strict-prototypes
4114 Warn if a function is declared or defined without specifying the
4115 argument types. (An old-style function definition is permitted without
4116 a warning if preceded by a declaration which specifies the argument
4119 @item -Wold-style-declaration @r{(C and Objective-C only)}
4120 @opindex Wold-style-declaration
4121 @opindex Wno-old-style-declaration
4122 Warn for obsolescent usages, according to the C Standard, in a
4123 declaration. For example, warn if storage-class specifiers like
4124 @code{static} are not the first things in a declaration. This warning
4125 is also enabled by @option{-Wextra}.
4127 @item -Wold-style-definition @r{(C and Objective-C only)}
4128 @opindex Wold-style-definition
4129 @opindex Wno-old-style-definition
4130 Warn if an old-style function definition is used. A warning is given
4131 even if there is a previous prototype.
4133 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
4134 @opindex Wmissing-parameter-type
4135 @opindex Wno-missing-parameter-type
4136 A function parameter is declared without a type specifier in K&R-style
4143 This warning is also enabled by @option{-Wextra}.
4145 @item -Wmissing-prototypes @r{(C and Objective-C only)}
4146 @opindex Wmissing-prototypes
4147 @opindex Wno-missing-prototypes
4148 Warn if a global function is defined without a previous prototype
4149 declaration. This warning is issued even if the definition itself
4150 provides a prototype. The aim is to detect global functions that fail
4151 to be declared in header files.
4153 @item -Wmissing-declarations
4154 @opindex Wmissing-declarations
4155 @opindex Wno-missing-declarations
4156 Warn if a global function is defined without a previous declaration.
4157 Do so even if the definition itself provides a prototype.
4158 Use this option to detect global functions that are not declared in
4159 header files. In C++, no warnings are issued for function templates,
4160 or for inline functions, or for functions in anonymous namespaces.
4162 @item -Wmissing-field-initializers
4163 @opindex Wmissing-field-initializers
4164 @opindex Wno-missing-field-initializers
4168 Warn if a structure's initializer has some fields missing. For
4169 example, the following code would cause such a warning, because
4170 @code{x.h} is implicitly zero:
4173 struct s @{ int f, g, h; @};
4174 struct s x = @{ 3, 4 @};
4177 This option does not warn about designated initializers, so the following
4178 modification would not trigger a warning:
4181 struct s @{ int f, g, h; @};
4182 struct s x = @{ .f = 3, .g = 4 @};
4185 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
4186 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
4188 @item -Wmissing-format-attribute
4189 @opindex Wmissing-format-attribute
4190 @opindex Wno-missing-format-attribute
4193 Warn about function pointers which might be candidates for @code{format}
4194 attributes. Note these are only possible candidates, not absolute ones.
4195 GCC will guess that function pointers with @code{format} attributes that
4196 are used in assignment, initialization, parameter passing or return
4197 statements should have a corresponding @code{format} attribute in the
4198 resulting type. I.e.@: the left-hand side of the assignment or
4199 initialization, the type of the parameter variable, or the return type
4200 of the containing function respectively should also have a @code{format}
4201 attribute to avoid the warning.
4203 GCC will also warn about function definitions which might be
4204 candidates for @code{format} attributes. Again, these are only
4205 possible candidates. GCC will guess that @code{format} attributes
4206 might be appropriate for any function that calls a function like
4207 @code{vprintf} or @code{vscanf}, but this might not always be the
4208 case, and some functions for which @code{format} attributes are
4209 appropriate may not be detected.
4211 @item -Wno-multichar
4212 @opindex Wno-multichar
4214 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4215 Usually they indicate a typo in the user's code, as they have
4216 implementation-defined values, and should not be used in portable code.
4218 @item -Wnormalized=<none|id|nfc|nfkc>
4219 @opindex Wnormalized=
4222 @cindex character set, input normalization
4223 In ISO C and ISO C++, two identifiers are different if they are
4224 different sequences of characters. However, sometimes when characters
4225 outside the basic ASCII character set are used, you can have two
4226 different character sequences that look the same. To avoid confusion,
4227 the ISO 10646 standard sets out some @dfn{normalization rules} which
4228 when applied ensure that two sequences that look the same are turned into
4229 the same sequence. GCC can warn you if you are using identifiers which
4230 have not been normalized; this option controls that warning.
4232 There are four levels of warning that GCC supports. The default is
4233 @option{-Wnormalized=nfc}, which warns about any identifier which is
4234 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4235 recommended form for most uses.
4237 Unfortunately, there are some characters which ISO C and ISO C++ allow
4238 in identifiers that when turned into NFC aren't allowable as
4239 identifiers. That is, there's no way to use these symbols in portable
4240 ISO C or C++ and have all your identifiers in NFC@.
4241 @option{-Wnormalized=id} suppresses the warning for these characters.
4242 It is hoped that future versions of the standards involved will correct
4243 this, which is why this option is not the default.
4245 You can switch the warning off for all characters by writing
4246 @option{-Wnormalized=none}. You would only want to do this if you
4247 were using some other normalization scheme (like ``D''), because
4248 otherwise you can easily create bugs that are literally impossible to see.
4250 Some characters in ISO 10646 have distinct meanings but look identical
4251 in some fonts or display methodologies, especially once formatting has
4252 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4253 LETTER N'', will display just like a regular @code{n} which has been
4254 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4255 normalization scheme to convert all these into a standard form as
4256 well, and GCC will warn if your code is not in NFKC if you use
4257 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4258 about every identifier that contains the letter O because it might be
4259 confused with the digit 0, and so is not the default, but may be
4260 useful as a local coding convention if the programming environment is
4261 unable to be fixed to display these characters distinctly.
4263 @item -Wno-deprecated
4264 @opindex Wno-deprecated
4265 @opindex Wdeprecated
4266 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4268 @item -Wno-deprecated-declarations
4269 @opindex Wno-deprecated-declarations
4270 @opindex Wdeprecated-declarations
4271 Do not warn about uses of functions (@pxref{Function Attributes}),
4272 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4273 Attributes}) marked as deprecated by using the @code{deprecated}
4277 @opindex Wno-overflow
4279 Do not warn about compile-time overflow in constant expressions.
4281 @item -Woverride-init @r{(C and Objective-C only)}
4282 @opindex Woverride-init
4283 @opindex Wno-override-init
4287 Warn if an initialized field without side effects is overridden when
4288 using designated initializers (@pxref{Designated Inits, , Designated
4291 This warning is included in @option{-Wextra}. To get other
4292 @option{-Wextra} warnings without this one, use @samp{-Wextra
4293 -Wno-override-init}.
4298 Warn if a structure is given the packed attribute, but the packed
4299 attribute has no effect on the layout or size of the structure.
4300 Such structures may be mis-aligned for little benefit. For
4301 instance, in this code, the variable @code{f.x} in @code{struct bar}
4302 will be misaligned even though @code{struct bar} does not itself
4303 have the packed attribute:
4310 @} __attribute__((packed));
4318 @item -Wpacked-bitfield-compat
4319 @opindex Wpacked-bitfield-compat
4320 @opindex Wno-packed-bitfield-compat
4321 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4322 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4323 the change can lead to differences in the structure layout. GCC
4324 informs you when the offset of such a field has changed in GCC 4.4.
4325 For example there is no longer a 4-bit padding between field @code{a}
4326 and @code{b} in this structure:
4333 @} __attribute__ ((packed));
4336 This warning is enabled by default. Use
4337 @option{-Wno-packed-bitfield-compat} to disable this warning.
4342 Warn if padding is included in a structure, either to align an element
4343 of the structure or to align the whole structure. Sometimes when this
4344 happens it is possible to rearrange the fields of the structure to
4345 reduce the padding and so make the structure smaller.
4347 @item -Wredundant-decls
4348 @opindex Wredundant-decls
4349 @opindex Wno-redundant-decls
4350 Warn if anything is declared more than once in the same scope, even in
4351 cases where multiple declaration is valid and changes nothing.
4353 @item -Wnested-externs @r{(C and Objective-C only)}
4354 @opindex Wnested-externs
4355 @opindex Wno-nested-externs
4356 Warn if an @code{extern} declaration is encountered within a function.
4361 Warn if a function can not be inlined and it was declared as inline.
4362 Even with this option, the compiler will not warn about failures to
4363 inline functions declared in system headers.
4365 The compiler uses a variety of heuristics to determine whether or not
4366 to inline a function. For example, the compiler takes into account
4367 the size of the function being inlined and the amount of inlining
4368 that has already been done in the current function. Therefore,
4369 seemingly insignificant changes in the source program can cause the
4370 warnings produced by @option{-Winline} to appear or disappear.
4372 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4373 @opindex Wno-invalid-offsetof
4374 @opindex Winvalid-offsetof
4375 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4376 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4377 to a non-POD type is undefined. In existing C++ implementations,
4378 however, @samp{offsetof} typically gives meaningful results even when
4379 applied to certain kinds of non-POD types. (Such as a simple
4380 @samp{struct} that fails to be a POD type only by virtue of having a
4381 constructor.) This flag is for users who are aware that they are
4382 writing nonportable code and who have deliberately chosen to ignore the
4385 The restrictions on @samp{offsetof} may be relaxed in a future version
4386 of the C++ standard.
4388 @item -Wno-int-to-pointer-cast
4389 @opindex Wno-int-to-pointer-cast
4390 @opindex Wint-to-pointer-cast
4391 Suppress warnings from casts to pointer type of an integer of a
4392 different size. In C++, casting to a pointer type of smaller size is
4393 an error. @option{Wint-to-pointer-cast} is enabled by default.
4396 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4397 @opindex Wno-pointer-to-int-cast
4398 @opindex Wpointer-to-int-cast
4399 Suppress warnings from casts from a pointer to an integer type of a
4403 @opindex Winvalid-pch
4404 @opindex Wno-invalid-pch
4405 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4406 the search path but can't be used.
4410 @opindex Wno-long-long
4411 Warn if @samp{long long} type is used. This is enabled by either
4412 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4413 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4415 @item -Wvariadic-macros
4416 @opindex Wvariadic-macros
4417 @opindex Wno-variadic-macros
4418 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4419 alternate syntax when in pedantic ISO C99 mode. This is default.
4420 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4425 Warn if variable length array is used in the code.
4426 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4427 the variable length array.
4429 @item -Wvolatile-register-var
4430 @opindex Wvolatile-register-var
4431 @opindex Wno-volatile-register-var
4432 Warn if a register variable is declared volatile. The volatile
4433 modifier does not inhibit all optimizations that may eliminate reads
4434 and/or writes to register variables. This warning is enabled by
4437 @item -Wdisabled-optimization
4438 @opindex Wdisabled-optimization
4439 @opindex Wno-disabled-optimization
4440 Warn if a requested optimization pass is disabled. This warning does
4441 not generally indicate that there is anything wrong with your code; it
4442 merely indicates that GCC's optimizers were unable to handle the code
4443 effectively. Often, the problem is that your code is too big or too
4444 complex; GCC will refuse to optimize programs when the optimization
4445 itself is likely to take inordinate amounts of time.
4447 @item -Wpointer-sign @r{(C and Objective-C only)}
4448 @opindex Wpointer-sign
4449 @opindex Wno-pointer-sign
4450 Warn for pointer argument passing or assignment with different signedness.
4451 This option is only supported for C and Objective-C@. It is implied by
4452 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4453 @option{-Wno-pointer-sign}.
4455 @item -Wstack-protector
4456 @opindex Wstack-protector
4457 @opindex Wno-stack-protector
4458 This option is only active when @option{-fstack-protector} is active. It
4459 warns about functions that will not be protected against stack smashing.
4462 @opindex Wno-mudflap
4463 Suppress warnings about constructs that cannot be instrumented by
4466 @item -Woverlength-strings
4467 @opindex Woverlength-strings
4468 @opindex Wno-overlength-strings
4469 Warn about string constants which are longer than the ``minimum
4470 maximum'' length specified in the C standard. Modern compilers
4471 generally allow string constants which are much longer than the
4472 standard's minimum limit, but very portable programs should avoid
4473 using longer strings.
4475 The limit applies @emph{after} string constant concatenation, and does
4476 not count the trailing NUL@. In C90, the limit was 509 characters; in
4477 C99, it was raised to 4095. C++98 does not specify a normative
4478 minimum maximum, so we do not diagnose overlength strings in C++@.
4480 This option is implied by @option{-pedantic}, and can be disabled with
4481 @option{-Wno-overlength-strings}.
4483 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
4484 @opindex Wunsuffixed-float-constants
4486 GCC will issue a warning for any floating constant that does not have
4487 a suffix. When used together with @option{-Wsystem-headers} it will
4488 warn about such constants in system header files. This can be useful
4489 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4490 from the decimal floating-point extension to C99.
4493 @node Debugging Options
4494 @section Options for Debugging Your Program or GCC
4495 @cindex options, debugging
4496 @cindex debugging information options
4498 GCC has various special options that are used for debugging
4499 either your program or GCC:
4504 Produce debugging information in the operating system's native format
4505 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4508 On most systems that use stabs format, @option{-g} enables use of extra
4509 debugging information that only GDB can use; this extra information
4510 makes debugging work better in GDB but will probably make other debuggers
4512 refuse to read the program. If you want to control for certain whether
4513 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4514 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4516 GCC allows you to use @option{-g} with
4517 @option{-O}. The shortcuts taken by optimized code may occasionally
4518 produce surprising results: some variables you declared may not exist
4519 at all; flow of control may briefly move where you did not expect it;
4520 some statements may not be executed because they compute constant
4521 results or their values were already at hand; some statements may
4522 execute in different places because they were moved out of loops.
4524 Nevertheless it proves possible to debug optimized output. This makes
4525 it reasonable to use the optimizer for programs that might have bugs.
4527 The following options are useful when GCC is generated with the
4528 capability for more than one debugging format.
4532 Produce debugging information for use by GDB@. This means to use the
4533 most expressive format available (DWARF 2, stabs, or the native format
4534 if neither of those are supported), including GDB extensions if at all
4539 Produce debugging information in stabs format (if that is supported),
4540 without GDB extensions. This is the format used by DBX on most BSD
4541 systems. On MIPS, Alpha and System V Release 4 systems this option
4542 produces stabs debugging output which is not understood by DBX or SDB@.
4543 On System V Release 4 systems this option requires the GNU assembler.
4545 @item -feliminate-unused-debug-symbols
4546 @opindex feliminate-unused-debug-symbols
4547 Produce debugging information in stabs format (if that is supported),
4548 for only symbols that are actually used.
4550 @item -femit-class-debug-always
4551 Instead of emitting debugging information for a C++ class in only one
4552 object file, emit it in all object files using the class. This option
4553 should be used only with debuggers that are unable to handle the way GCC
4554 normally emits debugging information for classes because using this
4555 option will increase the size of debugging information by as much as a
4560 Produce debugging information in stabs format (if that is supported),
4561 using GNU extensions understood only by the GNU debugger (GDB)@. The
4562 use of these extensions is likely to make other debuggers crash or
4563 refuse to read the program.
4567 Produce debugging information in COFF format (if that is supported).
4568 This is the format used by SDB on most System V systems prior to
4573 Produce debugging information in XCOFF format (if that is supported).
4574 This is the format used by the DBX debugger on IBM RS/6000 systems.
4578 Produce debugging information in XCOFF format (if that is supported),
4579 using GNU extensions understood only by the GNU debugger (GDB)@. The
4580 use of these extensions is likely to make other debuggers crash or
4581 refuse to read the program, and may cause assemblers other than the GNU
4582 assembler (GAS) to fail with an error.
4584 @item -gdwarf-@var{version}
4585 @opindex gdwarf-@var{version}
4586 Produce debugging information in DWARF format (if that is
4587 supported). This is the format used by DBX on IRIX 6. The value
4588 of @var{version} may be either 2, 3 or 4; the default version is 2.
4590 Note that with DWARF version 2 some ports require, and will always
4591 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4593 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
4594 for maximum benefit.
4596 @item -gstrict-dwarf
4597 @opindex gstrict-dwarf
4598 Disallow using extensions of later DWARF standard version than selected
4599 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
4600 DWARF extensions from later standard versions is allowed.
4602 @item -gno-strict-dwarf
4603 @opindex gno-strict-dwarf
4604 Allow using extensions of later DWARF standard version than selected with
4605 @option{-gdwarf-@var{version}}.
4609 Produce debugging information in VMS debug format (if that is
4610 supported). This is the format used by DEBUG on VMS systems.
4613 @itemx -ggdb@var{level}
4614 @itemx -gstabs@var{level}
4615 @itemx -gcoff@var{level}
4616 @itemx -gxcoff@var{level}
4617 @itemx -gvms@var{level}
4618 Request debugging information and also use @var{level} to specify how
4619 much information. The default level is 2.
4621 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4624 Level 1 produces minimal information, enough for making backtraces in
4625 parts of the program that you don't plan to debug. This includes
4626 descriptions of functions and external variables, but no information
4627 about local variables and no line numbers.
4629 Level 3 includes extra information, such as all the macro definitions
4630 present in the program. Some debuggers support macro expansion when
4631 you use @option{-g3}.
4633 @option{-gdwarf-2} does not accept a concatenated debug level, because
4634 GCC used to support an option @option{-gdwarf} that meant to generate
4635 debug information in version 1 of the DWARF format (which is very
4636 different from version 2), and it would have been too confusing. That
4637 debug format is long obsolete, but the option cannot be changed now.
4638 Instead use an additional @option{-g@var{level}} option to change the
4639 debug level for DWARF.
4643 Turn off generation of debug info, if leaving out this option would have
4644 generated it, or turn it on at level 2 otherwise. The position of this
4645 argument in the command line does not matter, it takes effect after all
4646 other options are processed, and it does so only once, no matter how
4647 many times it is given. This is mainly intended to be used with
4648 @option{-fcompare-debug}.
4650 @item -fdump-final-insns@r{[}=@var{file}@r{]}
4651 @opindex fdump-final-insns
4652 Dump the final internal representation (RTL) to @var{file}. If the
4653 optional argument is omitted (or if @var{file} is @code{.}), the name
4654 of the dump file will be determined by appending @code{.gkd} to the
4655 compilation output file name.
4657 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4658 @opindex fcompare-debug
4659 @opindex fno-compare-debug
4660 If no error occurs during compilation, run the compiler a second time,
4661 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4662 passed to the second compilation. Dump the final internal
4663 representation in both compilations, and print an error if they differ.
4665 If the equal sign is omitted, the default @option{-gtoggle} is used.
4667 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4668 and nonzero, implicitly enables @option{-fcompare-debug}. If
4669 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4670 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4673 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4674 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4675 of the final representation and the second compilation, preventing even
4676 @env{GCC_COMPARE_DEBUG} from taking effect.
4678 To verify full coverage during @option{-fcompare-debug} testing, set
4679 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4680 which GCC will reject as an invalid option in any actual compilation
4681 (rather than preprocessing, assembly or linking). To get just a
4682 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4683 not overridden} will do.
4685 @item -fcompare-debug-second
4686 @opindex fcompare-debug-second
4687 This option is implicitly passed to the compiler for the second
4688 compilation requested by @option{-fcompare-debug}, along with options to
4689 silence warnings, and omitting other options that would cause
4690 side-effect compiler outputs to files or to the standard output. Dump
4691 files and preserved temporary files are renamed so as to contain the
4692 @code{.gk} additional extension during the second compilation, to avoid
4693 overwriting those generated by the first.
4695 When this option is passed to the compiler driver, it causes the
4696 @emph{first} compilation to be skipped, which makes it useful for little
4697 other than debugging the compiler proper.
4699 @item -feliminate-dwarf2-dups
4700 @opindex feliminate-dwarf2-dups
4701 Compress DWARF2 debugging information by eliminating duplicated
4702 information about each symbol. This option only makes sense when
4703 generating DWARF2 debugging information with @option{-gdwarf-2}.
4705 @item -femit-struct-debug-baseonly
4706 Emit debug information for struct-like types
4707 only when the base name of the compilation source file
4708 matches the base name of file in which the struct was defined.
4710 This option substantially reduces the size of debugging information,
4711 but at significant potential loss in type information to the debugger.
4712 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4713 See @option{-femit-struct-debug-detailed} for more detailed control.
4715 This option works only with DWARF 2.
4717 @item -femit-struct-debug-reduced
4718 Emit debug information for struct-like types
4719 only when the base name of the compilation source file
4720 matches the base name of file in which the type was defined,
4721 unless the struct is a template or defined in a system header.
4723 This option significantly reduces the size of debugging information,
4724 with some potential loss in type information to the debugger.
4725 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4726 See @option{-femit-struct-debug-detailed} for more detailed control.
4728 This option works only with DWARF 2.
4730 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4731 Specify the struct-like types
4732 for which the compiler will generate debug information.
4733 The intent is to reduce duplicate struct debug information
4734 between different object files within the same program.
4736 This option is a detailed version of
4737 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4738 which will serve for most needs.
4740 A specification has the syntax
4741 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4743 The optional first word limits the specification to
4744 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4745 A struct type is used directly when it is the type of a variable, member.
4746 Indirect uses arise through pointers to structs.
4747 That is, when use of an incomplete struct would be legal, the use is indirect.
4749 @samp{struct one direct; struct two * indirect;}.
4751 The optional second word limits the specification to
4752 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4753 Generic structs are a bit complicated to explain.
4754 For C++, these are non-explicit specializations of template classes,
4755 or non-template classes within the above.
4756 Other programming languages have generics,
4757 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4759 The third word specifies the source files for those
4760 structs for which the compiler will emit debug information.
4761 The values @samp{none} and @samp{any} have the normal meaning.
4762 The value @samp{base} means that
4763 the base of name of the file in which the type declaration appears
4764 must match the base of the name of the main compilation file.
4765 In practice, this means that
4766 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4767 but types declared in other header will not.
4768 The value @samp{sys} means those types satisfying @samp{base}
4769 or declared in system or compiler headers.
4771 You may need to experiment to determine the best settings for your application.
4773 The default is @samp{-femit-struct-debug-detailed=all}.
4775 This option works only with DWARF 2.
4777 @item -fenable-icf-debug
4778 @opindex fenable-icf-debug
4779 Generate additional debug information to support identical code folding (ICF).
4780 This option only works with DWARF version 2 or higher.
4782 @item -fno-merge-debug-strings
4783 @opindex fmerge-debug-strings
4784 @opindex fno-merge-debug-strings
4785 Direct the linker to not merge together strings in the debugging
4786 information which are identical in different object files. Merging is
4787 not supported by all assemblers or linkers. Merging decreases the size
4788 of the debug information in the output file at the cost of increasing
4789 link processing time. Merging is enabled by default.
4791 @item -fdebug-prefix-map=@var{old}=@var{new}
4792 @opindex fdebug-prefix-map
4793 When compiling files in directory @file{@var{old}}, record debugging
4794 information describing them as in @file{@var{new}} instead.
4796 @item -fno-dwarf2-cfi-asm
4797 @opindex fdwarf2-cfi-asm
4798 @opindex fno-dwarf2-cfi-asm
4799 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4800 instead of using GAS @code{.cfi_*} directives.
4802 @cindex @command{prof}
4805 Generate extra code to write profile information suitable for the
4806 analysis program @command{prof}. You must use this option when compiling
4807 the source files you want data about, and you must also use it when
4810 @cindex @command{gprof}
4813 Generate extra code to write profile information suitable for the
4814 analysis program @command{gprof}. You must use this option when compiling
4815 the source files you want data about, and you must also use it when
4820 Makes the compiler print out each function name as it is compiled, and
4821 print some statistics about each pass when it finishes.
4824 @opindex ftime-report
4825 Makes the compiler print some statistics about the time consumed by each
4826 pass when it finishes.
4829 @opindex fmem-report
4830 Makes the compiler print some statistics about permanent memory
4831 allocation when it finishes.
4833 @item -fpre-ipa-mem-report
4834 @opindex fpre-ipa-mem-report
4835 @item -fpost-ipa-mem-report
4836 @opindex fpost-ipa-mem-report
4837 Makes the compiler print some statistics about permanent memory
4838 allocation before or after interprocedural optimization.
4840 @item -fprofile-arcs
4841 @opindex fprofile-arcs
4842 Add code so that program flow @dfn{arcs} are instrumented. During
4843 execution the program records how many times each branch and call is
4844 executed and how many times it is taken or returns. When the compiled
4845 program exits it saves this data to a file called
4846 @file{@var{auxname}.gcda} for each source file. The data may be used for
4847 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4848 test coverage analysis (@option{-ftest-coverage}). Each object file's
4849 @var{auxname} is generated from the name of the output file, if
4850 explicitly specified and it is not the final executable, otherwise it is
4851 the basename of the source file. In both cases any suffix is removed
4852 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4853 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4854 @xref{Cross-profiling}.
4856 @cindex @command{gcov}
4860 This option is used to compile and link code instrumented for coverage
4861 analysis. The option is a synonym for @option{-fprofile-arcs}
4862 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4863 linking). See the documentation for those options for more details.
4868 Compile the source files with @option{-fprofile-arcs} plus optimization
4869 and code generation options. For test coverage analysis, use the
4870 additional @option{-ftest-coverage} option. You do not need to profile
4871 every source file in a program.
4874 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4875 (the latter implies the former).
4878 Run the program on a representative workload to generate the arc profile
4879 information. This may be repeated any number of times. You can run
4880 concurrent instances of your program, and provided that the file system
4881 supports locking, the data files will be correctly updated. Also
4882 @code{fork} calls are detected and correctly handled (double counting
4886 For profile-directed optimizations, compile the source files again with
4887 the same optimization and code generation options plus
4888 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4889 Control Optimization}).
4892 For test coverage analysis, use @command{gcov} to produce human readable
4893 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4894 @command{gcov} documentation for further information.
4898 With @option{-fprofile-arcs}, for each function of your program GCC
4899 creates a program flow graph, then finds a spanning tree for the graph.
4900 Only arcs that are not on the spanning tree have to be instrumented: the
4901 compiler adds code to count the number of times that these arcs are
4902 executed. When an arc is the only exit or only entrance to a block, the
4903 instrumentation code can be added to the block; otherwise, a new basic
4904 block must be created to hold the instrumentation code.
4907 @item -ftest-coverage
4908 @opindex ftest-coverage
4909 Produce a notes file that the @command{gcov} code-coverage utility
4910 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4911 show program coverage. Each source file's note file is called
4912 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4913 above for a description of @var{auxname} and instructions on how to
4914 generate test coverage data. Coverage data will match the source files
4915 more closely, if you do not optimize.
4917 @item -fdbg-cnt-list
4918 @opindex fdbg-cnt-list
4919 Print the name and the counter upperbound for all debug counters.
4921 @item -fdbg-cnt=@var{counter-value-list}
4923 Set the internal debug counter upperbound. @var{counter-value-list}
4924 is a comma-separated list of @var{name}:@var{value} pairs
4925 which sets the upperbound of each debug counter @var{name} to @var{value}.
4926 All debug counters have the initial upperbound of @var{UINT_MAX},
4927 thus dbg_cnt() returns true always unless the upperbound is set by this option.
4928 e.g. With -fdbg-cnt=dce:10,tail_call:0
4929 dbg_cnt(dce) will return true only for first 10 invocations
4930 and dbg_cnt(tail_call) will return false always.
4932 @item -d@var{letters}
4933 @itemx -fdump-rtl-@var{pass}
4935 Says to make debugging dumps during compilation at times specified by
4936 @var{letters}. This is used for debugging the RTL-based passes of the
4937 compiler. The file names for most of the dumps are made by appending
4938 a pass number and a word to the @var{dumpname}, and the files are
4939 created in the directory of the output file. @var{dumpname} is
4940 generated from the name of the output file, if explicitly specified
4941 and it is not an executable, otherwise it is the basename of the
4942 source file. These switches may have different effects when
4943 @option{-E} is used for preprocessing.
4945 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
4946 @option{-d} option @var{letters}. Here are the possible
4947 letters for use in @var{pass} and @var{letters}, and their meanings:
4951 @item -fdump-rtl-alignments
4952 @opindex fdump-rtl-alignments
4953 Dump after branch alignments have been computed.
4955 @item -fdump-rtl-asmcons
4956 @opindex fdump-rtl-asmcons
4957 Dump after fixing rtl statements that have unsatisfied in/out constraints.
4959 @item -fdump-rtl-auto_inc_dec
4960 @opindex fdump-rtl-auto_inc_dec
4961 Dump after auto-inc-dec discovery. This pass is only run on
4962 architectures that have auto inc or auto dec instructions.
4964 @item -fdump-rtl-barriers
4965 @opindex fdump-rtl-barriers
4966 Dump after cleaning up the barrier instructions.
4968 @item -fdump-rtl-bbpart
4969 @opindex fdump-rtl-bbpart
4970 Dump after partitioning hot and cold basic blocks.
4972 @item -fdump-rtl-bbro
4973 @opindex fdump-rtl-bbro
4974 Dump after block reordering.
4976 @item -fdump-rtl-btl1
4977 @itemx -fdump-rtl-btl2
4978 @opindex fdump-rtl-btl2
4979 @opindex fdump-rtl-btl2
4980 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
4981 after the two branch
4982 target load optimization passes.
4984 @item -fdump-rtl-bypass
4985 @opindex fdump-rtl-bypass
4986 Dump after jump bypassing and control flow optimizations.
4988 @item -fdump-rtl-combine
4989 @opindex fdump-rtl-combine
4990 Dump after the RTL instruction combination pass.
4992 @item -fdump-rtl-compgotos
4993 @opindex fdump-rtl-compgotos
4994 Dump after duplicating the computed gotos.
4996 @item -fdump-rtl-ce1
4997 @itemx -fdump-rtl-ce2
4998 @itemx -fdump-rtl-ce3
4999 @opindex fdump-rtl-ce1
5000 @opindex fdump-rtl-ce2
5001 @opindex fdump-rtl-ce3
5002 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
5003 @option{-fdump-rtl-ce3} enable dumping after the three
5004 if conversion passes.
5006 @itemx -fdump-rtl-cprop_hardreg
5007 @opindex fdump-rtl-cprop_hardreg
5008 Dump after hard register copy propagation.
5010 @itemx -fdump-rtl-csa
5011 @opindex fdump-rtl-csa
5012 Dump after combining stack adjustments.
5014 @item -fdump-rtl-cse1
5015 @itemx -fdump-rtl-cse2
5016 @opindex fdump-rtl-cse1
5017 @opindex fdump-rtl-cse2
5018 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
5019 the two common sub-expression elimination passes.
5021 @itemx -fdump-rtl-dce
5022 @opindex fdump-rtl-dce
5023 Dump after the standalone dead code elimination passes.
5025 @itemx -fdump-rtl-dbr
5026 @opindex fdump-rtl-dbr
5027 Dump after delayed branch scheduling.
5029 @item -fdump-rtl-dce1
5030 @itemx -fdump-rtl-dce2
5031 @opindex fdump-rtl-dce1
5032 @opindex fdump-rtl-dce2
5033 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
5034 the two dead store elimination passes.
5037 @opindex fdump-rtl-eh
5038 Dump after finalization of EH handling code.
5040 @item -fdump-rtl-eh_ranges
5041 @opindex fdump-rtl-eh_ranges
5042 Dump after conversion of EH handling range regions.
5044 @item -fdump-rtl-expand
5045 @opindex fdump-rtl-expand
5046 Dump after RTL generation.
5048 @item -fdump-rtl-fwprop1
5049 @itemx -fdump-rtl-fwprop2
5050 @opindex fdump-rtl-fwprop1
5051 @opindex fdump-rtl-fwprop2
5052 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
5053 dumping after the two forward propagation passes.
5055 @item -fdump-rtl-gcse1
5056 @itemx -fdump-rtl-gcse2
5057 @opindex fdump-rtl-gcse1
5058 @opindex fdump-rtl-gcse2
5059 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
5060 after global common subexpression elimination.
5062 @item -fdump-rtl-init-regs
5063 @opindex fdump-rtl-init-regs
5064 Dump after the initialization of the registers.
5066 @item -fdump-rtl-initvals
5067 @opindex fdump-rtl-initvals
5068 Dump after the computation of the initial value sets.
5070 @itemx -fdump-rtl-into_cfglayout
5071 @opindex fdump-rtl-into_cfglayout
5072 Dump after converting to cfglayout mode.
5074 @item -fdump-rtl-ira
5075 @opindex fdump-rtl-ira
5076 Dump after iterated register allocation.
5078 @item -fdump-rtl-jump
5079 @opindex fdump-rtl-jump
5080 Dump after the second jump optimization.
5082 @item -fdump-rtl-loop2
5083 @opindex fdump-rtl-loop2
5084 @option{-fdump-rtl-loop2} enables dumping after the rtl
5085 loop optimization passes.
5087 @item -fdump-rtl-mach
5088 @opindex fdump-rtl-mach
5089 Dump after performing the machine dependent reorganization pass, if that
5092 @item -fdump-rtl-mode_sw
5093 @opindex fdump-rtl-mode_sw
5094 Dump after removing redundant mode switches.
5096 @item -fdump-rtl-rnreg
5097 @opindex fdump-rtl-rnreg
5098 Dump after register renumbering.
5100 @itemx -fdump-rtl-outof_cfglayout
5101 @opindex fdump-rtl-outof_cfglayout
5102 Dump after converting from cfglayout mode.
5104 @item -fdump-rtl-peephole2
5105 @opindex fdump-rtl-peephole2
5106 Dump after the peephole pass.
5108 @item -fdump-rtl-postreload
5109 @opindex fdump-rtl-postreload
5110 Dump after post-reload optimizations.
5112 @itemx -fdump-rtl-pro_and_epilogue
5113 @opindex fdump-rtl-pro_and_epilogue
5114 Dump after generating the function pro and epilogues.
5116 @item -fdump-rtl-regmove
5117 @opindex fdump-rtl-regmove
5118 Dump after the register move pass.
5120 @item -fdump-rtl-sched1
5121 @itemx -fdump-rtl-sched2
5122 @opindex fdump-rtl-sched1
5123 @opindex fdump-rtl-sched2
5124 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
5125 after the basic block scheduling passes.
5127 @item -fdump-rtl-see
5128 @opindex fdump-rtl-see
5129 Dump after sign extension elimination.
5131 @item -fdump-rtl-seqabstr
5132 @opindex fdump-rtl-seqabstr
5133 Dump after common sequence discovery.
5135 @item -fdump-rtl-shorten
5136 @opindex fdump-rtl-shorten
5137 Dump after shortening branches.
5139 @item -fdump-rtl-sibling
5140 @opindex fdump-rtl-sibling
5141 Dump after sibling call optimizations.
5143 @item -fdump-rtl-split1
5144 @itemx -fdump-rtl-split2
5145 @itemx -fdump-rtl-split3
5146 @itemx -fdump-rtl-split4
5147 @itemx -fdump-rtl-split5
5148 @opindex fdump-rtl-split1
5149 @opindex fdump-rtl-split2
5150 @opindex fdump-rtl-split3
5151 @opindex fdump-rtl-split4
5152 @opindex fdump-rtl-split5
5153 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
5154 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
5155 @option{-fdump-rtl-split5} enable dumping after five rounds of
5156 instruction splitting.
5158 @item -fdump-rtl-sms
5159 @opindex fdump-rtl-sms
5160 Dump after modulo scheduling. This pass is only run on some
5163 @item -fdump-rtl-stack
5164 @opindex fdump-rtl-stack
5165 Dump after conversion from GCC's "flat register file" registers to the
5166 x87's stack-like registers. This pass is only run on x86 variants.
5168 @item -fdump-rtl-subreg1
5169 @itemx -fdump-rtl-subreg2
5170 @opindex fdump-rtl-subreg1
5171 @opindex fdump-rtl-subreg2
5172 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
5173 the two subreg expansion passes.
5175 @item -fdump-rtl-unshare
5176 @opindex fdump-rtl-unshare
5177 Dump after all rtl has been unshared.
5179 @item -fdump-rtl-vartrack
5180 @opindex fdump-rtl-vartrack
5181 Dump after variable tracking.
5183 @item -fdump-rtl-vregs
5184 @opindex fdump-rtl-vregs
5185 Dump after converting virtual registers to hard registers.
5187 @item -fdump-rtl-web
5188 @opindex fdump-rtl-web
5189 Dump after live range splitting.
5191 @item -fdump-rtl-regclass
5192 @itemx -fdump-rtl-subregs_of_mode_init
5193 @itemx -fdump-rtl-subregs_of_mode_finish
5194 @itemx -fdump-rtl-dfinit
5195 @itemx -fdump-rtl-dfinish
5196 @opindex fdump-rtl-regclass
5197 @opindex fdump-rtl-subregs_of_mode_init
5198 @opindex fdump-rtl-subregs_of_mode_finish
5199 @opindex fdump-rtl-dfinit
5200 @opindex fdump-rtl-dfinish
5201 These dumps are defined but always produce empty files.
5203 @item -fdump-rtl-all
5204 @opindex fdump-rtl-all
5205 Produce all the dumps listed above.
5209 Annotate the assembler output with miscellaneous debugging information.
5213 Dump all macro definitions, at the end of preprocessing, in addition to
5218 Produce a core dump whenever an error occurs.
5222 Print statistics on memory usage, at the end of the run, to
5227 Annotate the assembler output with a comment indicating which
5228 pattern and alternative was used. The length of each instruction is
5233 Dump the RTL in the assembler output as a comment before each instruction.
5234 Also turns on @option{-dp} annotation.
5238 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
5239 dump a representation of the control flow graph suitable for viewing with VCG
5240 to @file{@var{file}.@var{pass}.vcg}.
5244 Just generate RTL for a function instead of compiling it. Usually used
5245 with @option{-fdump-rtl-expand}.
5249 Dump debugging information during parsing, to standard error.
5253 @opindex fdump-noaddr
5254 When doing debugging dumps, suppress address output. This makes it more
5255 feasible to use diff on debugging dumps for compiler invocations with
5256 different compiler binaries and/or different
5257 text / bss / data / heap / stack / dso start locations.
5259 @item -fdump-unnumbered
5260 @opindex fdump-unnumbered
5261 When doing debugging dumps, suppress instruction numbers and address output.
5262 This makes it more feasible to use diff on debugging dumps for compiler
5263 invocations with different options, in particular with and without
5266 @item -fdump-unnumbered-links
5267 @opindex fdump-unnumbered-links
5268 When doing debugging dumps (see @option{-d} option above), suppress
5269 instruction numbers for the links to the previous and next instructions
5272 @item -fdump-translation-unit @r{(C++ only)}
5273 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5274 @opindex fdump-translation-unit
5275 Dump a representation of the tree structure for the entire translation
5276 unit to a file. The file name is made by appending @file{.tu} to the
5277 source file name, and the file is created in the same directory as the
5278 output file. If the @samp{-@var{options}} form is used, @var{options}
5279 controls the details of the dump as described for the
5280 @option{-fdump-tree} options.
5282 @item -fdump-class-hierarchy @r{(C++ only)}
5283 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5284 @opindex fdump-class-hierarchy
5285 Dump a representation of each class's hierarchy and virtual function
5286 table layout to a file. The file name is made by appending
5287 @file{.class} to the source file name, and the file is created in the
5288 same directory as the output file. If the @samp{-@var{options}} form
5289 is used, @var{options} controls the details of the dump as described
5290 for the @option{-fdump-tree} options.
5292 @item -fdump-ipa-@var{switch}
5294 Control the dumping at various stages of inter-procedural analysis
5295 language tree to a file. The file name is generated by appending a
5296 switch specific suffix to the source file name, and the file is created
5297 in the same directory as the output file. The following dumps are
5302 Enables all inter-procedural analysis dumps.
5305 Dumps information about call-graph optimization, unused function removal,
5306 and inlining decisions.
5309 Dump after function inlining.
5313 @item -fdump-statistics-@var{option}
5314 @opindex fdump-statistics
5315 Enable and control dumping of pass statistics in a separate file. The
5316 file name is generated by appending a suffix ending in
5317 @samp{.statistics} to the source file name, and the file is created in
5318 the same directory as the output file. If the @samp{-@var{option}}
5319 form is used, @samp{-stats} will cause counters to be summed over the
5320 whole compilation unit while @samp{-details} will dump every event as
5321 the passes generate them. The default with no option is to sum
5322 counters for each function compiled.
5324 @item -fdump-tree-@var{switch}
5325 @itemx -fdump-tree-@var{switch}-@var{options}
5327 Control the dumping at various stages of processing the intermediate
5328 language tree to a file. The file name is generated by appending a
5329 switch specific suffix to the source file name, and the file is
5330 created in the same directory as the output file. If the
5331 @samp{-@var{options}} form is used, @var{options} is a list of
5332 @samp{-} separated options that control the details of the dump. Not
5333 all options are applicable to all dumps, those which are not
5334 meaningful will be ignored. The following options are available
5338 Print the address of each node. Usually this is not meaningful as it
5339 changes according to the environment and source file. Its primary use
5340 is for tying up a dump file with a debug environment.
5342 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
5343 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
5344 use working backward from mangled names in the assembly file.
5346 Inhibit dumping of members of a scope or body of a function merely
5347 because that scope has been reached. Only dump such items when they
5348 are directly reachable by some other path. When dumping pretty-printed
5349 trees, this option inhibits dumping the bodies of control structures.
5351 Print a raw representation of the tree. By default, trees are
5352 pretty-printed into a C-like representation.
5354 Enable more detailed dumps (not honored by every dump option).
5356 Enable dumping various statistics about the pass (not honored by every dump
5359 Enable showing basic block boundaries (disabled in raw dumps).
5361 Enable showing virtual operands for every statement.
5363 Enable showing line numbers for statements.
5365 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5367 Enable showing the tree dump for each statement.
5369 Enable showing the EH region number holding each statement.
5371 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5372 and @option{lineno}.
5375 The following tree dumps are possible:
5379 @opindex fdump-tree-original
5380 Dump before any tree based optimization, to @file{@var{file}.original}.
5383 @opindex fdump-tree-optimized
5384 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5387 @opindex fdump-tree-gimple
5388 Dump each function before and after the gimplification pass to a file. The
5389 file name is made by appending @file{.gimple} to the source file name.
5392 @opindex fdump-tree-cfg
5393 Dump the control flow graph of each function to a file. The file name is
5394 made by appending @file{.cfg} to the source file name.
5397 @opindex fdump-tree-vcg
5398 Dump the control flow graph of each function to a file in VCG format. The
5399 file name is made by appending @file{.vcg} to the source file name. Note
5400 that if the file contains more than one function, the generated file cannot
5401 be used directly by VCG@. You will need to cut and paste each function's
5402 graph into its own separate file first.
5405 @opindex fdump-tree-ch
5406 Dump each function after copying loop headers. The file name is made by
5407 appending @file{.ch} to the source file name.
5410 @opindex fdump-tree-ssa
5411 Dump SSA related information to a file. The file name is made by appending
5412 @file{.ssa} to the source file name.
5415 @opindex fdump-tree-alias
5416 Dump aliasing information for each function. The file name is made by
5417 appending @file{.alias} to the source file name.
5420 @opindex fdump-tree-ccp
5421 Dump each function after CCP@. The file name is made by appending
5422 @file{.ccp} to the source file name.
5425 @opindex fdump-tree-storeccp
5426 Dump each function after STORE-CCP@. The file name is made by appending
5427 @file{.storeccp} to the source file name.
5430 @opindex fdump-tree-pre
5431 Dump trees after partial redundancy elimination. The file name is made
5432 by appending @file{.pre} to the source file name.
5435 @opindex fdump-tree-fre
5436 Dump trees after full redundancy elimination. The file name is made
5437 by appending @file{.fre} to the source file name.
5440 @opindex fdump-tree-copyprop
5441 Dump trees after copy propagation. The file name is made
5442 by appending @file{.copyprop} to the source file name.
5444 @item store_copyprop
5445 @opindex fdump-tree-store_copyprop
5446 Dump trees after store copy-propagation. The file name is made
5447 by appending @file{.store_copyprop} to the source file name.
5450 @opindex fdump-tree-dce
5451 Dump each function after dead code elimination. The file name is made by
5452 appending @file{.dce} to the source file name.
5455 @opindex fdump-tree-mudflap
5456 Dump each function after adding mudflap instrumentation. The file name is
5457 made by appending @file{.mudflap} to the source file name.
5460 @opindex fdump-tree-sra
5461 Dump each function after performing scalar replacement of aggregates. The
5462 file name is made by appending @file{.sra} to the source file name.
5465 @opindex fdump-tree-sink
5466 Dump each function after performing code sinking. The file name is made
5467 by appending @file{.sink} to the source file name.
5470 @opindex fdump-tree-dom
5471 Dump each function after applying dominator tree optimizations. The file
5472 name is made by appending @file{.dom} to the source file name.
5475 @opindex fdump-tree-dse
5476 Dump each function after applying dead store elimination. The file
5477 name is made by appending @file{.dse} to the source file name.
5480 @opindex fdump-tree-phiopt
5481 Dump each function after optimizing PHI nodes into straightline code. The file
5482 name is made by appending @file{.phiopt} to the source file name.
5485 @opindex fdump-tree-forwprop
5486 Dump each function after forward propagating single use variables. The file
5487 name is made by appending @file{.forwprop} to the source file name.
5490 @opindex fdump-tree-copyrename
5491 Dump each function after applying the copy rename optimization. The file
5492 name is made by appending @file{.copyrename} to the source file name.
5495 @opindex fdump-tree-nrv
5496 Dump each function after applying the named return value optimization on
5497 generic trees. The file name is made by appending @file{.nrv} to the source
5501 @opindex fdump-tree-vect
5502 Dump each function after applying vectorization of loops. The file name is
5503 made by appending @file{.vect} to the source file name.
5506 @opindex fdump-tree-slp
5507 Dump each function after applying vectorization of basic blocks. The file name
5508 is made by appending @file{.slp} to the source file name.
5511 @opindex fdump-tree-vrp
5512 Dump each function after Value Range Propagation (VRP). The file name
5513 is made by appending @file{.vrp} to the source file name.
5516 @opindex fdump-tree-all
5517 Enable all the available tree dumps with the flags provided in this option.
5520 @item -ftree-vectorizer-verbose=@var{n}
5521 @opindex ftree-vectorizer-verbose
5522 This option controls the amount of debugging output the vectorizer prints.
5523 This information is written to standard error, unless
5524 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5525 in which case it is output to the usual dump listing file, @file{.vect}.
5526 For @var{n}=0 no diagnostic information is reported.
5527 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5528 and the total number of loops that got vectorized.
5529 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5530 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5531 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5532 level that @option{-fdump-tree-vect-stats} uses.
5533 Higher verbosity levels mean either more information dumped for each
5534 reported loop, or same amount of information reported for more loops:
5535 if @var{n}=3, vectorizer cost model information is reported.
5536 If @var{n}=4, alignment related information is added to the reports.
5537 If @var{n}=5, data-references related information (e.g.@: memory dependences,
5538 memory access-patterns) is added to the reports.
5539 If @var{n}=6, the vectorizer reports also non-vectorized inner-most loops
5540 that did not pass the first analysis phase (i.e., may not be countable, or
5541 may have complicated control-flow).
5542 If @var{n}=7, the vectorizer reports also non-vectorized nested loops.
5543 If @var{n}=8, SLP related information is added to the reports.
5544 For @var{n}=9, all the information the vectorizer generates during its
5545 analysis and transformation is reported. This is the same verbosity level
5546 that @option{-fdump-tree-vect-details} uses.
5548 @item -frandom-seed=@var{string}
5549 @opindex frandom-seed
5550 This option provides a seed that GCC uses when it would otherwise use
5551 random numbers. It is used to generate certain symbol names
5552 that have to be different in every compiled file. It is also used to
5553 place unique stamps in coverage data files and the object files that
5554 produce them. You can use the @option{-frandom-seed} option to produce
5555 reproducibly identical object files.
5557 The @var{string} should be different for every file you compile.
5559 @item -fsched-verbose=@var{n}
5560 @opindex fsched-verbose
5561 On targets that use instruction scheduling, this option controls the
5562 amount of debugging output the scheduler prints. This information is
5563 written to standard error, unless @option{-fdump-rtl-sched1} or
5564 @option{-fdump-rtl-sched2} is specified, in which case it is output
5565 to the usual dump listing file, @file{.sched1} or @file{.sched2}
5566 respectively. However for @var{n} greater than nine, the output is
5567 always printed to standard error.
5569 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5570 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5571 For @var{n} greater than one, it also output basic block probabilities,
5572 detailed ready list information and unit/insn info. For @var{n} greater
5573 than two, it includes RTL at abort point, control-flow and regions info.
5574 And for @var{n} over four, @option{-fsched-verbose} also includes
5578 @itemx -save-temps=cwd
5580 Store the usual ``temporary'' intermediate files permanently; place them
5581 in the current directory and name them based on the source file. Thus,
5582 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5583 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5584 preprocessed @file{foo.i} output file even though the compiler now
5585 normally uses an integrated preprocessor.
5587 When used in combination with the @option{-x} command line option,
5588 @option{-save-temps} is sensible enough to avoid over writing an
5589 input source file with the same extension as an intermediate file.
5590 The corresponding intermediate file may be obtained by renaming the
5591 source file before using @option{-save-temps}.
5593 If you invoke GCC in parallel, compiling several different source
5594 files that share a common base name in different subdirectories or the
5595 same source file compiled for multiple output destinations, it is
5596 likely that the different parallel compilers will interfere with each
5597 other, and overwrite the temporary files. For instance:
5600 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5601 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5604 may result in @file{foo.i} and @file{foo.o} being written to
5605 simultaneously by both compilers.
5607 @item -save-temps=obj
5608 @opindex save-temps=obj
5609 Store the usual ``temporary'' intermediate files permanently. If the
5610 @option{-o} option is used, the temporary files are based on the
5611 object file. If the @option{-o} option is not used, the
5612 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5617 gcc -save-temps=obj -c foo.c
5618 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5619 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5622 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5623 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5624 @file{dir2/yfoobar.o}.
5626 @item -time@r{[}=@var{file}@r{]}
5628 Report the CPU time taken by each subprocess in the compilation
5629 sequence. For C source files, this is the compiler proper and assembler
5630 (plus the linker if linking is done).
5632 Without the specification of an output file, the output looks like this:
5639 The first number on each line is the ``user time'', that is time spent
5640 executing the program itself. The second number is ``system time'',
5641 time spent executing operating system routines on behalf of the program.
5642 Both numbers are in seconds.
5644 With the specification of an output file, the output is appended to the
5645 named file, and it looks like this:
5648 0.12 0.01 cc1 @var{options}
5649 0.00 0.01 as @var{options}
5652 The ``user time'' and the ``system time'' are moved before the program
5653 name, and the options passed to the program are displayed, so that one
5654 can later tell what file was being compiled, and with which options.
5656 @item -fvar-tracking
5657 @opindex fvar-tracking
5658 Run variable tracking pass. It computes where variables are stored at each
5659 position in code. Better debugging information is then generated
5660 (if the debugging information format supports this information).
5662 It is enabled by default when compiling with optimization (@option{-Os},
5663 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5664 the debug info format supports it.
5666 @item -fvar-tracking-assignments
5667 @opindex fvar-tracking-assignments
5668 @opindex fno-var-tracking-assignments
5669 Annotate assignments to user variables early in the compilation and
5670 attempt to carry the annotations over throughout the compilation all the
5671 way to the end, in an attempt to improve debug information while
5672 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
5674 It can be enabled even if var-tracking is disabled, in which case
5675 annotations will be created and maintained, but discarded at the end.
5677 @item -fvar-tracking-assignments-toggle
5678 @opindex fvar-tracking-assignments-toggle
5679 @opindex fno-var-tracking-assignments-toggle
5680 Toggle @option{-fvar-tracking-assignments}, in the same way that
5681 @option{-gtoggle} toggles @option{-g}.
5683 @item -print-file-name=@var{library}
5684 @opindex print-file-name
5685 Print the full absolute name of the library file @var{library} that
5686 would be used when linking---and don't do anything else. With this
5687 option, GCC does not compile or link anything; it just prints the
5690 @item -print-multi-directory
5691 @opindex print-multi-directory
5692 Print the directory name corresponding to the multilib selected by any
5693 other switches present in the command line. This directory is supposed
5694 to exist in @env{GCC_EXEC_PREFIX}.
5696 @item -print-multi-lib
5697 @opindex print-multi-lib
5698 Print the mapping from multilib directory names to compiler switches
5699 that enable them. The directory name is separated from the switches by
5700 @samp{;}, and each switch starts with an @samp{@@} instead of the
5701 @samp{-}, without spaces between multiple switches. This is supposed to
5702 ease shell-processing.
5704 @item -print-multi-os-directory
5705 @opindex print-multi-os-directory
5706 Print the path to OS libraries for the selected
5707 multilib, relative to some @file{lib} subdirectory. If OS libraries are
5708 present in the @file{lib} subdirectory and no multilibs are used, this is
5709 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
5710 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
5711 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
5712 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
5714 @item -print-prog-name=@var{program}
5715 @opindex print-prog-name
5716 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5718 @item -print-libgcc-file-name
5719 @opindex print-libgcc-file-name
5720 Same as @option{-print-file-name=libgcc.a}.
5722 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5723 but you do want to link with @file{libgcc.a}. You can do
5726 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5729 @item -print-search-dirs
5730 @opindex print-search-dirs
5731 Print the name of the configured installation directory and a list of
5732 program and library directories @command{gcc} will search---and don't do anything else.
5734 This is useful when @command{gcc} prints the error message
5735 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5736 To resolve this you either need to put @file{cpp0} and the other compiler
5737 components where @command{gcc} expects to find them, or you can set the environment
5738 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5739 Don't forget the trailing @samp{/}.
5740 @xref{Environment Variables}.
5742 @item -print-sysroot
5743 @opindex print-sysroot
5744 Print the target sysroot directory that will be used during
5745 compilation. This is the target sysroot specified either at configure
5746 time or using the @option{--sysroot} option, possibly with an extra
5747 suffix that depends on compilation options. If no target sysroot is
5748 specified, the option prints nothing.
5750 @item -print-sysroot-headers-suffix
5751 @opindex print-sysroot-headers-suffix
5752 Print the suffix added to the target sysroot when searching for
5753 headers, or give an error if the compiler is not configured with such
5754 a suffix---and don't do anything else.
5757 @opindex dumpmachine
5758 Print the compiler's target machine (for example,
5759 @samp{i686-pc-linux-gnu})---and don't do anything else.
5762 @opindex dumpversion
5763 Print the compiler version (for example, @samp{3.0})---and don't do
5768 Print the compiler's built-in specs---and don't do anything else. (This
5769 is used when GCC itself is being built.) @xref{Spec Files}.
5771 @item -feliminate-unused-debug-types
5772 @opindex feliminate-unused-debug-types
5773 Normally, when producing DWARF2 output, GCC will emit debugging
5774 information for all types declared in a compilation
5775 unit, regardless of whether or not they are actually used
5776 in that compilation unit. Sometimes this is useful, such as
5777 if, in the debugger, you want to cast a value to a type that is
5778 not actually used in your program (but is declared). More often,
5779 however, this results in a significant amount of wasted space.
5780 With this option, GCC will avoid producing debug symbol output
5781 for types that are nowhere used in the source file being compiled.
5784 @node Optimize Options
5785 @section Options That Control Optimization
5786 @cindex optimize options
5787 @cindex options, optimization
5789 These options control various sorts of optimizations.
5791 Without any optimization option, the compiler's goal is to reduce the
5792 cost of compilation and to make debugging produce the expected
5793 results. Statements are independent: if you stop the program with a
5794 breakpoint between statements, you can then assign a new value to any
5795 variable or change the program counter to any other statement in the
5796 function and get exactly the results you would expect from the source
5799 Turning on optimization flags makes the compiler attempt to improve
5800 the performance and/or code size at the expense of compilation time
5801 and possibly the ability to debug the program.
5803 The compiler performs optimization based on the knowledge it has of the
5804 program. Compiling multiple files at once to a single output file mode allows
5805 the compiler to use information gained from all of the files when compiling
5808 Not all optimizations are controlled directly by a flag. Only
5809 optimizations that have a flag are listed in this section.
5811 Most optimizations are only enabled if an @option{-O} level is set on
5812 the command line. Otherwise they are disabled, even if individual
5813 optimization flags are specified.
5815 Depending on the target and how GCC was configured, a slightly different
5816 set of optimizations may be enabled at each @option{-O} level than
5817 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
5818 to find out the exact set of optimizations that are enabled at each level.
5819 @xref{Overall Options}, for examples.
5826 Optimize. Optimizing compilation takes somewhat more time, and a lot
5827 more memory for a large function.
5829 With @option{-O}, the compiler tries to reduce code size and execution
5830 time, without performing any optimizations that take a great deal of
5833 @option{-O} turns on the following optimization flags:
5836 -fcprop-registers @gol
5839 -fdelayed-branch @gol
5841 -fguess-branch-probability @gol
5842 -fif-conversion2 @gol
5843 -fif-conversion @gol
5844 -fipa-pure-const @gol
5846 -fipa-reference @gol
5848 -fsplit-wide-types @gol
5849 -ftree-builtin-call-dce @gol
5852 -ftree-copyrename @gol
5854 -ftree-dominator-opts @gol
5856 -ftree-forwprop @gol
5864 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5865 where doing so does not interfere with debugging.
5869 Optimize even more. GCC performs nearly all supported optimizations
5870 that do not involve a space-speed tradeoff.
5871 As compared to @option{-O}, this option increases both compilation time
5872 and the performance of the generated code.
5874 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5875 also turns on the following optimization flags:
5876 @gccoptlist{-fthread-jumps @gol
5877 -falign-functions -falign-jumps @gol
5878 -falign-loops -falign-labels @gol
5881 -fcse-follow-jumps -fcse-skip-blocks @gol
5882 -fdelete-null-pointer-checks @gol
5883 -fexpensive-optimizations @gol
5884 -fgcse -fgcse-lm @gol
5885 -finline-small-functions @gol
5886 -findirect-inlining @gol
5888 -foptimize-sibling-calls @gol
5889 -fpartial-inlining @gol
5892 -freorder-blocks -freorder-functions @gol
5893 -frerun-cse-after-loop @gol
5894 -fsched-interblock -fsched-spec @gol
5895 -fschedule-insns -fschedule-insns2 @gol
5896 -fstrict-aliasing -fstrict-overflow @gol
5897 -ftree-switch-conversion @gol
5901 Please note the warning under @option{-fgcse} about
5902 invoking @option{-O2} on programs that use computed gotos.
5906 Optimize yet more. @option{-O3} turns on all optimizations specified
5907 by @option{-O2} and also turns on the @option{-finline-functions},
5908 @option{-funswitch-loops}, @option{-fpredictive-commoning},
5909 @option{-fgcse-after-reload} and @option{-ftree-vectorize} options.
5913 Reduce compilation time and make debugging produce the expected
5914 results. This is the default.
5918 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5919 do not typically increase code size. It also performs further
5920 optimizations designed to reduce code size.
5922 @option{-Os} disables the following optimization flags:
5923 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5924 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5925 -fprefetch-loop-arrays -ftree-vect-loop-version}
5929 Disregard strict standards compliance. @option{-Ofast} enables all
5930 @option{-O3} optimizations. It also enables optimizations that are not
5931 valid for all standard compliant programs.
5932 It turns on @option{-ffast-math}.
5934 If you use multiple @option{-O} options, with or without level numbers,
5935 the last such option is the one that is effective.
5938 Options of the form @option{-f@var{flag}} specify machine-independent
5939 flags. Most flags have both positive and negative forms; the negative
5940 form of @option{-ffoo} would be @option{-fno-foo}. In the table
5941 below, only one of the forms is listed---the one you typically will
5942 use. You can figure out the other form by either removing @samp{no-}
5945 The following options control specific optimizations. They are either
5946 activated by @option{-O} options or are related to ones that are. You
5947 can use the following flags in the rare cases when ``fine-tuning'' of
5948 optimizations to be performed is desired.
5951 @item -fno-default-inline
5952 @opindex fno-default-inline
5953 Do not make member functions inline by default merely because they are
5954 defined inside the class scope (C++ only). Otherwise, when you specify
5955 @w{@option{-O}}, member functions defined inside class scope are compiled
5956 inline by default; i.e., you don't need to add @samp{inline} in front of
5957 the member function name.
5959 @item -fno-defer-pop
5960 @opindex fno-defer-pop
5961 Always pop the arguments to each function call as soon as that function
5962 returns. For machines which must pop arguments after a function call,
5963 the compiler normally lets arguments accumulate on the stack for several
5964 function calls and pops them all at once.
5966 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5968 @item -fforward-propagate
5969 @opindex fforward-propagate
5970 Perform a forward propagation pass on RTL@. The pass tries to combine two
5971 instructions and checks if the result can be simplified. If loop unrolling
5972 is active, two passes are performed and the second is scheduled after
5975 This option is enabled by default at optimization levels @option{-O},
5976 @option{-O2}, @option{-O3}, @option{-Os}.
5978 @item -fomit-frame-pointer
5979 @opindex fomit-frame-pointer
5980 Don't keep the frame pointer in a register for functions that
5981 don't need one. This avoids the instructions to save, set up and
5982 restore frame pointers; it also makes an extra register available
5983 in many functions. @strong{It also makes debugging impossible on
5986 On some machines, such as the VAX, this flag has no effect, because
5987 the standard calling sequence automatically handles the frame pointer
5988 and nothing is saved by pretending it doesn't exist. The
5989 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
5990 whether a target machine supports this flag. @xref{Registers,,Register
5991 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
5993 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5995 @item -foptimize-sibling-calls
5996 @opindex foptimize-sibling-calls
5997 Optimize sibling and tail recursive calls.
5999 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6003 Don't pay attention to the @code{inline} keyword. Normally this option
6004 is used to keep the compiler from expanding any functions inline.
6005 Note that if you are not optimizing, no functions can be expanded inline.
6007 @item -finline-small-functions
6008 @opindex finline-small-functions
6009 Integrate functions into their callers when their body is smaller than expected
6010 function call code (so overall size of program gets smaller). The compiler
6011 heuristically decides which functions are simple enough to be worth integrating
6014 Enabled at level @option{-O2}.
6016 @item -findirect-inlining
6017 @opindex findirect-inlining
6018 Inline also indirect calls that are discovered to be known at compile
6019 time thanks to previous inlining. This option has any effect only
6020 when inlining itself is turned on by the @option{-finline-functions}
6021 or @option{-finline-small-functions} options.
6023 Enabled at level @option{-O2}.
6025 @item -finline-functions
6026 @opindex finline-functions
6027 Integrate all simple functions into their callers. The compiler
6028 heuristically decides which functions are simple enough to be worth
6029 integrating in this way.
6031 If all calls to a given function are integrated, and the function is
6032 declared @code{static}, then the function is normally not output as
6033 assembler code in its own right.
6035 Enabled at level @option{-O3}.
6037 @item -finline-functions-called-once
6038 @opindex finline-functions-called-once
6039 Consider all @code{static} functions called once for inlining into their
6040 caller even if they are not marked @code{inline}. If a call to a given
6041 function is integrated, then the function is not output as assembler code
6044 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
6046 @item -fearly-inlining
6047 @opindex fearly-inlining
6048 Inline functions marked by @code{always_inline} and functions whose body seems
6049 smaller than the function call overhead early before doing
6050 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
6051 makes profiling significantly cheaper and usually inlining faster on programs
6052 having large chains of nested wrapper functions.
6058 Perform interprocedural scalar replacement of aggregates, removal of
6059 unused parameters and replacement of parameters passed by reference
6060 by parameters passed by value.
6062 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
6064 @item -finline-limit=@var{n}
6065 @opindex finline-limit
6066 By default, GCC limits the size of functions that can be inlined. This flag
6067 allows coarse control of this limit. @var{n} is the size of functions that
6068 can be inlined in number of pseudo instructions.
6070 Inlining is actually controlled by a number of parameters, which may be
6071 specified individually by using @option{--param @var{name}=@var{value}}.
6072 The @option{-finline-limit=@var{n}} option sets some of these parameters
6076 @item max-inline-insns-single
6077 is set to @var{n}/2.
6078 @item max-inline-insns-auto
6079 is set to @var{n}/2.
6082 See below for a documentation of the individual
6083 parameters controlling inlining and for the defaults of these parameters.
6085 @emph{Note:} there may be no value to @option{-finline-limit} that results
6086 in default behavior.
6088 @emph{Note:} pseudo instruction represents, in this particular context, an
6089 abstract measurement of function's size. In no way does it represent a count
6090 of assembly instructions and as such its exact meaning might change from one
6091 release to an another.
6093 @item -fkeep-inline-functions
6094 @opindex fkeep-inline-functions
6095 In C, emit @code{static} functions that are declared @code{inline}
6096 into the object file, even if the function has been inlined into all
6097 of its callers. This switch does not affect functions using the
6098 @code{extern inline} extension in GNU C90@. In C++, emit any and all
6099 inline functions into the object file.
6101 @item -fkeep-static-consts
6102 @opindex fkeep-static-consts
6103 Emit variables declared @code{static const} when optimization isn't turned
6104 on, even if the variables aren't referenced.
6106 GCC enables this option by default. If you want to force the compiler to
6107 check if the variable was referenced, regardless of whether or not
6108 optimization is turned on, use the @option{-fno-keep-static-consts} option.
6110 @item -fmerge-constants
6111 @opindex fmerge-constants
6112 Attempt to merge identical constants (string constants and floating point
6113 constants) across compilation units.
6115 This option is the default for optimized compilation if the assembler and
6116 linker support it. Use @option{-fno-merge-constants} to inhibit this
6119 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6121 @item -fmerge-all-constants
6122 @opindex fmerge-all-constants
6123 Attempt to merge identical constants and identical variables.
6125 This option implies @option{-fmerge-constants}. In addition to
6126 @option{-fmerge-constants} this considers e.g.@: even constant initialized
6127 arrays or initialized constant variables with integral or floating point
6128 types. Languages like C or C++ require each variable, including multiple
6129 instances of the same variable in recursive calls, to have distinct locations,
6130 so using this option will result in non-conforming
6133 @item -fmodulo-sched
6134 @opindex fmodulo-sched
6135 Perform swing modulo scheduling immediately before the first scheduling
6136 pass. This pass looks at innermost loops and reorders their
6137 instructions by overlapping different iterations.
6139 @item -fmodulo-sched-allow-regmoves
6140 @opindex fmodulo-sched-allow-regmoves
6141 Perform more aggressive SMS based modulo scheduling with register moves
6142 allowed. By setting this flag certain anti-dependences edges will be
6143 deleted which will trigger the generation of reg-moves based on the
6144 life-range analysis. This option is effective only with
6145 @option{-fmodulo-sched} enabled.
6147 @item -fno-branch-count-reg
6148 @opindex fno-branch-count-reg
6149 Do not use ``decrement and branch'' instructions on a count register,
6150 but instead generate a sequence of instructions that decrement a
6151 register, compare it against zero, then branch based upon the result.
6152 This option is only meaningful on architectures that support such
6153 instructions, which include x86, PowerPC, IA-64 and S/390.
6155 The default is @option{-fbranch-count-reg}.
6157 @item -fno-function-cse
6158 @opindex fno-function-cse
6159 Do not put function addresses in registers; make each instruction that
6160 calls a constant function contain the function's address explicitly.
6162 This option results in less efficient code, but some strange hacks
6163 that alter the assembler output may be confused by the optimizations
6164 performed when this option is not used.
6166 The default is @option{-ffunction-cse}
6168 @item -fno-zero-initialized-in-bss
6169 @opindex fno-zero-initialized-in-bss
6170 If the target supports a BSS section, GCC by default puts variables that
6171 are initialized to zero into BSS@. This can save space in the resulting
6174 This option turns off this behavior because some programs explicitly
6175 rely on variables going to the data section. E.g., so that the
6176 resulting executable can find the beginning of that section and/or make
6177 assumptions based on that.
6179 The default is @option{-fzero-initialized-in-bss}.
6181 @item -fmudflap -fmudflapth -fmudflapir
6185 @cindex bounds checking
6187 For front-ends that support it (C and C++), instrument all risky
6188 pointer/array dereferencing operations, some standard library
6189 string/heap functions, and some other associated constructs with
6190 range/validity tests. Modules so instrumented should be immune to
6191 buffer overflows, invalid heap use, and some other classes of C/C++
6192 programming errors. The instrumentation relies on a separate runtime
6193 library (@file{libmudflap}), which will be linked into a program if
6194 @option{-fmudflap} is given at link time. Run-time behavior of the
6195 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
6196 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
6199 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
6200 link if your program is multi-threaded. Use @option{-fmudflapir}, in
6201 addition to @option{-fmudflap} or @option{-fmudflapth}, if
6202 instrumentation should ignore pointer reads. This produces less
6203 instrumentation (and therefore faster execution) and still provides
6204 some protection against outright memory corrupting writes, but allows
6205 erroneously read data to propagate within a program.
6207 @item -fthread-jumps
6208 @opindex fthread-jumps
6209 Perform optimizations where we check to see if a jump branches to a
6210 location where another comparison subsumed by the first is found. If
6211 so, the first branch is redirected to either the destination of the
6212 second branch or a point immediately following it, depending on whether
6213 the condition is known to be true or false.
6215 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6217 @item -fsplit-wide-types
6218 @opindex fsplit-wide-types
6219 When using a type that occupies multiple registers, such as @code{long
6220 long} on a 32-bit system, split the registers apart and allocate them
6221 independently. This normally generates better code for those types,
6222 but may make debugging more difficult.
6224 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
6227 @item -fcse-follow-jumps
6228 @opindex fcse-follow-jumps
6229 In common subexpression elimination (CSE), scan through jump instructions
6230 when the target of the jump is not reached by any other path. For
6231 example, when CSE encounters an @code{if} statement with an
6232 @code{else} clause, CSE will follow the jump when the condition
6235 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6237 @item -fcse-skip-blocks
6238 @opindex fcse-skip-blocks
6239 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
6240 follow jumps which conditionally skip over blocks. When CSE
6241 encounters a simple @code{if} statement with no else clause,
6242 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6243 body of the @code{if}.
6245 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6247 @item -frerun-cse-after-loop
6248 @opindex frerun-cse-after-loop
6249 Re-run common subexpression elimination after loop optimizations has been
6252 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6256 Perform a global common subexpression elimination pass.
6257 This pass also performs global constant and copy propagation.
6259 @emph{Note:} When compiling a program using computed gotos, a GCC
6260 extension, you may get better runtime performance if you disable
6261 the global common subexpression elimination pass by adding
6262 @option{-fno-gcse} to the command line.
6264 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6268 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
6269 attempt to move loads which are only killed by stores into themselves. This
6270 allows a loop containing a load/store sequence to be changed to a load outside
6271 the loop, and a copy/store within the loop.
6273 Enabled by default when gcse is enabled.
6277 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6278 global common subexpression elimination. This pass will attempt to move
6279 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6280 loops containing a load/store sequence can be changed to a load before
6281 the loop and a store after the loop.
6283 Not enabled at any optimization level.
6287 When @option{-fgcse-las} is enabled, the global common subexpression
6288 elimination pass eliminates redundant loads that come after stores to the
6289 same memory location (both partial and full redundancies).
6291 Not enabled at any optimization level.
6293 @item -fgcse-after-reload
6294 @opindex fgcse-after-reload
6295 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6296 pass is performed after reload. The purpose of this pass is to cleanup
6299 @item -funsafe-loop-optimizations
6300 @opindex funsafe-loop-optimizations
6301 If given, the loop optimizer will assume that loop indices do not
6302 overflow, and that the loops with nontrivial exit condition are not
6303 infinite. This enables a wider range of loop optimizations even if
6304 the loop optimizer itself cannot prove that these assumptions are valid.
6305 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
6306 if it finds this kind of loop.
6308 @item -fcrossjumping
6309 @opindex fcrossjumping
6310 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6311 resulting code may or may not perform better than without cross-jumping.
6313 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6315 @item -fauto-inc-dec
6316 @opindex fauto-inc-dec
6317 Combine increments or decrements of addresses with memory accesses.
6318 This pass is always skipped on architectures that do not have
6319 instructions to support this. Enabled by default at @option{-O} and
6320 higher on architectures that support this.
6324 Perform dead code elimination (DCE) on RTL@.
6325 Enabled by default at @option{-O} and higher.
6329 Perform dead store elimination (DSE) on RTL@.
6330 Enabled by default at @option{-O} and higher.
6332 @item -fif-conversion
6333 @opindex fif-conversion
6334 Attempt to transform conditional jumps into branch-less equivalents. This
6335 include use of conditional moves, min, max, set flags and abs instructions, and
6336 some tricks doable by standard arithmetics. The use of conditional execution
6337 on chips where it is available is controlled by @code{if-conversion2}.
6339 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6341 @item -fif-conversion2
6342 @opindex fif-conversion2
6343 Use conditional execution (where available) to transform conditional jumps into
6344 branch-less equivalents.
6346 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6348 @item -fdelete-null-pointer-checks
6349 @opindex fdelete-null-pointer-checks
6350 Assume that programs cannot safely dereference null pointers, and that
6351 no code or data element resides there. This enables simple constant
6352 folding optimizations at all optimization levels. In addition, other
6353 optimization passes in GCC use this flag to control global dataflow
6354 analyses that eliminate useless checks for null pointers; these assume
6355 that if a pointer is checked after it has already been dereferenced,
6358 Note however that in some environments this assumption is not true.
6359 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6360 for programs which depend on that behavior.
6362 Some targets, especially embedded ones, disable this option at all levels.
6363 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6364 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6365 are enabled independently at different optimization levels.
6367 @item -fexpensive-optimizations
6368 @opindex fexpensive-optimizations
6369 Perform a number of minor optimizations that are relatively expensive.
6371 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6373 @item -foptimize-register-move
6375 @opindex foptimize-register-move
6377 Attempt to reassign register numbers in move instructions and as
6378 operands of other simple instructions in order to maximize the amount of
6379 register tying. This is especially helpful on machines with two-operand
6382 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6385 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6387 @item -fira-algorithm=@var{algorithm}
6388 Use specified coloring algorithm for the integrated register
6389 allocator. The @var{algorithm} argument should be @code{priority} or
6390 @code{CB}. The first algorithm specifies Chow's priority coloring,
6391 the second one specifies Chaitin-Briggs coloring. The second
6392 algorithm can be unimplemented for some architectures. If it is
6393 implemented, it is the default because Chaitin-Briggs coloring as a
6394 rule generates a better code.
6396 @item -fira-region=@var{region}
6397 Use specified regions for the integrated register allocator. The
6398 @var{region} argument should be one of @code{all}, @code{mixed}, or
6399 @code{one}. The first value means using all loops as register
6400 allocation regions, the second value which is the default means using
6401 all loops except for loops with small register pressure as the
6402 regions, and third one means using all function as a single region.
6403 The first value can give best result for machines with small size and
6404 irregular register set, the third one results in faster and generates
6405 decent code and the smallest size code, and the default value usually
6406 give the best results in most cases and for most architectures.
6408 @item -fira-coalesce
6409 @opindex fira-coalesce
6410 Do optimistic register coalescing. This option might be profitable for
6411 architectures with big regular register files.
6413 @item -fira-loop-pressure
6414 @opindex fira-loop-pressure
6415 Use IRA to evaluate register pressure in loops for decision to move
6416 loop invariants. Usage of this option usually results in generation
6417 of faster and smaller code on machines with big register files (>= 32
6418 registers) but it can slow compiler down.
6420 This option is enabled at level @option{-O3} for some targets.
6422 @item -fno-ira-share-save-slots
6423 @opindex fno-ira-share-save-slots
6424 Switch off sharing stack slots used for saving call used hard
6425 registers living through a call. Each hard register will get a
6426 separate stack slot and as a result function stack frame will be
6429 @item -fno-ira-share-spill-slots
6430 @opindex fno-ira-share-spill-slots
6431 Switch off sharing stack slots allocated for pseudo-registers. Each
6432 pseudo-register which did not get a hard register will get a separate
6433 stack slot and as a result function stack frame will be bigger.
6435 @item -fira-verbose=@var{n}
6436 @opindex fira-verbose
6437 Set up how verbose dump file for the integrated register allocator
6438 will be. Default value is 5. If the value is greater or equal to 10,
6439 the dump file will be stderr as if the value were @var{n} minus 10.
6441 @item -fdelayed-branch
6442 @opindex fdelayed-branch
6443 If supported for the target machine, attempt to reorder instructions
6444 to exploit instruction slots available after delayed branch
6447 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6449 @item -fschedule-insns
6450 @opindex fschedule-insns
6451 If supported for the target machine, attempt to reorder instructions to
6452 eliminate execution stalls due to required data being unavailable. This
6453 helps machines that have slow floating point or memory load instructions
6454 by allowing other instructions to be issued until the result of the load
6455 or floating point instruction is required.
6457 Enabled at levels @option{-O2}, @option{-O3}.
6459 @item -fschedule-insns2
6460 @opindex fschedule-insns2
6461 Similar to @option{-fschedule-insns}, but requests an additional pass of
6462 instruction scheduling after register allocation has been done. This is
6463 especially useful on machines with a relatively small number of
6464 registers and where memory load instructions take more than one cycle.
6466 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6468 @item -fno-sched-interblock
6469 @opindex fno-sched-interblock
6470 Don't schedule instructions across basic blocks. This is normally
6471 enabled by default when scheduling before register allocation, i.e.@:
6472 with @option{-fschedule-insns} or at @option{-O2} or higher.
6474 @item -fno-sched-spec
6475 @opindex fno-sched-spec
6476 Don't allow speculative motion of non-load instructions. This is normally
6477 enabled by default when scheduling before register allocation, i.e.@:
6478 with @option{-fschedule-insns} or at @option{-O2} or higher.
6480 @item -fsched-pressure
6481 @opindex fsched-pressure
6482 Enable register pressure sensitive insn scheduling before the register
6483 allocation. This only makes sense when scheduling before register
6484 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
6485 @option{-O2} or higher. Usage of this option can improve the
6486 generated code and decrease its size by preventing register pressure
6487 increase above the number of available hard registers and as a
6488 consequence register spills in the register allocation.
6490 @item -fsched-spec-load
6491 @opindex fsched-spec-load
6492 Allow speculative motion of some load instructions. This only makes
6493 sense when scheduling before register allocation, i.e.@: with
6494 @option{-fschedule-insns} or at @option{-O2} or higher.
6496 @item -fsched-spec-load-dangerous
6497 @opindex fsched-spec-load-dangerous
6498 Allow speculative motion of more load instructions. This only makes
6499 sense when scheduling before register allocation, i.e.@: with
6500 @option{-fschedule-insns} or at @option{-O2} or higher.
6502 @item -fsched-stalled-insns
6503 @itemx -fsched-stalled-insns=@var{n}
6504 @opindex fsched-stalled-insns
6505 Define how many insns (if any) can be moved prematurely from the queue
6506 of stalled insns into the ready list, during the second scheduling pass.
6507 @option{-fno-sched-stalled-insns} means that no insns will be moved
6508 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6509 on how many queued insns can be moved prematurely.
6510 @option{-fsched-stalled-insns} without a value is equivalent to
6511 @option{-fsched-stalled-insns=1}.
6513 @item -fsched-stalled-insns-dep
6514 @itemx -fsched-stalled-insns-dep=@var{n}
6515 @opindex fsched-stalled-insns-dep
6516 Define how many insn groups (cycles) will be examined for a dependency
6517 on a stalled insn that is candidate for premature removal from the queue
6518 of stalled insns. This has an effect only during the second scheduling pass,
6519 and only if @option{-fsched-stalled-insns} is used.
6520 @option{-fno-sched-stalled-insns-dep} is equivalent to
6521 @option{-fsched-stalled-insns-dep=0}.
6522 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6523 @option{-fsched-stalled-insns-dep=1}.
6525 @item -fsched2-use-superblocks
6526 @opindex fsched2-use-superblocks
6527 When scheduling after register allocation, do use superblock scheduling
6528 algorithm. Superblock scheduling allows motion across basic block boundaries
6529 resulting on faster schedules. This option is experimental, as not all machine
6530 descriptions used by GCC model the CPU closely enough to avoid unreliable
6531 results from the algorithm.
6533 This only makes sense when scheduling after register allocation, i.e.@: with
6534 @option{-fschedule-insns2} or at @option{-O2} or higher.
6536 @item -fsched-group-heuristic
6537 @opindex fsched-group-heuristic
6538 Enable the group heuristic in the scheduler. This heuristic favors
6539 the instruction that belongs to a schedule group. This is enabled
6540 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6541 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6543 @item -fsched-critical-path-heuristic
6544 @opindex fsched-critical-path-heuristic
6545 Enable the critical-path heuristic in the scheduler. This heuristic favors
6546 instructions on the critical path. This is enabled by default when
6547 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6548 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6550 @item -fsched-spec-insn-heuristic
6551 @opindex fsched-spec-insn-heuristic
6552 Enable the speculative instruction heuristic in the scheduler. This
6553 heuristic favors speculative instructions with greater dependency weakness.
6554 This is enabled by default when scheduling is enabled, i.e.@:
6555 with @option{-fschedule-insns} or @option{-fschedule-insns2}
6556 or at @option{-O2} or higher.
6558 @item -fsched-rank-heuristic
6559 @opindex fsched-rank-heuristic
6560 Enable the rank heuristic in the scheduler. This heuristic favors
6561 the instruction belonging to a basic block with greater size or frequency.
6562 This is enabled by default when scheduling is enabled, i.e.@:
6563 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6564 at @option{-O2} or higher.
6566 @item -fsched-last-insn-heuristic
6567 @opindex fsched-last-insn-heuristic
6568 Enable the last-instruction heuristic in the scheduler. This heuristic
6569 favors the instruction that is less dependent on the last instruction
6570 scheduled. This is enabled by default when scheduling is enabled,
6571 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6572 at @option{-O2} or higher.
6574 @item -fsched-dep-count-heuristic
6575 @opindex fsched-dep-count-heuristic
6576 Enable the dependent-count heuristic in the scheduler. This heuristic
6577 favors the instruction that has more instructions depending on it.
6578 This is enabled by default when scheduling is enabled, i.e.@:
6579 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6580 at @option{-O2} or higher.
6582 @item -freschedule-modulo-scheduled-loops
6583 @opindex freschedule-modulo-scheduled-loops
6584 The modulo scheduling comes before the traditional scheduling, if a loop
6585 was modulo scheduled we may want to prevent the later scheduling passes
6586 from changing its schedule, we use this option to control that.
6588 @item -fselective-scheduling
6589 @opindex fselective-scheduling
6590 Schedule instructions using selective scheduling algorithm. Selective
6591 scheduling runs instead of the first scheduler pass.
6593 @item -fselective-scheduling2
6594 @opindex fselective-scheduling2
6595 Schedule instructions using selective scheduling algorithm. Selective
6596 scheduling runs instead of the second scheduler pass.
6598 @item -fsel-sched-pipelining
6599 @opindex fsel-sched-pipelining
6600 Enable software pipelining of innermost loops during selective scheduling.
6601 This option has no effect until one of @option{-fselective-scheduling} or
6602 @option{-fselective-scheduling2} is turned on.
6604 @item -fsel-sched-pipelining-outer-loops
6605 @opindex fsel-sched-pipelining-outer-loops
6606 When pipelining loops during selective scheduling, also pipeline outer loops.
6607 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6609 @item -fcaller-saves
6610 @opindex fcaller-saves
6611 Enable values to be allocated in registers that will be clobbered by
6612 function calls, by emitting extra instructions to save and restore the
6613 registers around such calls. Such allocation is done only when it
6614 seems to result in better code than would otherwise be produced.
6616 This option is always enabled by default on certain machines, usually
6617 those which have no call-preserved registers to use instead.
6619 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6621 @item -fconserve-stack
6622 @opindex fconserve-stack
6623 Attempt to minimize stack usage. The compiler will attempt to use less
6624 stack space, even if that makes the program slower. This option
6625 implies setting the @option{large-stack-frame} parameter to 100
6626 and the @option{large-stack-frame-growth} parameter to 400.
6628 @item -ftree-reassoc
6629 @opindex ftree-reassoc
6630 Perform reassociation on trees. This flag is enabled by default
6631 at @option{-O} and higher.
6635 Perform partial redundancy elimination (PRE) on trees. This flag is
6636 enabled by default at @option{-O2} and @option{-O3}.
6638 @item -ftree-forwprop
6639 @opindex ftree-forwprop
6640 Perform forward propagation on trees. This flag is enabled by default
6641 at @option{-O} and higher.
6645 Perform full redundancy elimination (FRE) on trees. The difference
6646 between FRE and PRE is that FRE only considers expressions
6647 that are computed on all paths leading to the redundant computation.
6648 This analysis is faster than PRE, though it exposes fewer redundancies.
6649 This flag is enabled by default at @option{-O} and higher.
6651 @item -ftree-phiprop
6652 @opindex ftree-phiprop
6653 Perform hoisting of loads from conditional pointers on trees. This
6654 pass is enabled by default at @option{-O} and higher.
6656 @item -ftree-copy-prop
6657 @opindex ftree-copy-prop
6658 Perform copy propagation on trees. This pass eliminates unnecessary
6659 copy operations. This flag is enabled by default at @option{-O} and
6662 @item -fipa-pure-const
6663 @opindex fipa-pure-const
6664 Discover which functions are pure or constant.
6665 Enabled by default at @option{-O} and higher.
6667 @item -fipa-reference
6668 @opindex fipa-reference
6669 Discover which static variables do not escape cannot escape the
6671 Enabled by default at @option{-O} and higher.
6673 @item -fipa-struct-reorg
6674 @opindex fipa-struct-reorg
6675 Perform structure reorganization optimization, that change C-like structures
6676 layout in order to better utilize spatial locality. This transformation is
6677 affective for programs containing arrays of structures. Available in two
6678 compilation modes: profile-based (enabled with @option{-fprofile-generate})
6679 or static (which uses built-in heuristics). It works only in whole program
6680 mode, so it requires @option{-fwhole-program} and @option{-combine} to be
6681 enabled. Structures considered @samp{cold} by this transformation are not
6682 affected (see @option{--param struct-reorg-cold-struct-ratio=@var{value}}).
6684 With this flag, the program debug info reflects a new structure layout.
6688 Perform interprocedural pointer analysis and interprocedural modification
6689 and reference analysis. This option can cause excessive memory and
6690 compile-time usage on large compilation units. It is not enabled by
6691 default at any optimization level.
6694 @opindex fipa-profile
6695 Perform interprocedural profile propagation. The functions called only from
6696 cold functions are marked as cold. Also functions executed once (such as
6697 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
6698 functions and loop less parts of functions executed once are then optimized for
6700 Enabled by default at @option{-O} and higher.
6704 Perform interprocedural constant propagation.
6705 This optimization analyzes the program to determine when values passed
6706 to functions are constants and then optimizes accordingly.
6707 This optimization can substantially increase performance
6708 if the application has constants passed to functions.
6709 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
6711 @item -fipa-cp-clone
6712 @opindex fipa-cp-clone
6713 Perform function cloning to make interprocedural constant propagation stronger.
6714 When enabled, interprocedural constant propagation will perform function cloning
6715 when externally visible function can be called with constant arguments.
6716 Because this optimization can create multiple copies of functions,
6717 it may significantly increase code size
6718 (see @option{--param ipcp-unit-growth=@var{value}}).
6719 This flag is enabled by default at @option{-O3}.
6721 @item -fipa-matrix-reorg
6722 @opindex fipa-matrix-reorg
6723 Perform matrix flattening and transposing.
6724 Matrix flattening tries to replace an @math{m}-dimensional matrix
6725 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
6726 This reduces the level of indirection needed for accessing the elements
6727 of the matrix. The second optimization is matrix transposing that
6728 attempts to change the order of the matrix's dimensions in order to
6729 improve cache locality.
6730 Both optimizations need the @option{-fwhole-program} flag.
6731 Transposing is enabled only if profiling information is available.
6735 Perform forward store motion on trees. This flag is
6736 enabled by default at @option{-O} and higher.
6740 Perform sparse conditional constant propagation (CCP) on trees. This
6741 pass only operates on local scalar variables and is enabled by default
6742 at @option{-O} and higher.
6744 @item -ftree-switch-conversion
6745 Perform conversion of simple initializations in a switch to
6746 initializations from a scalar array. This flag is enabled by default
6747 at @option{-O2} and higher.
6751 Perform dead code elimination (DCE) on trees. This flag is enabled by
6752 default at @option{-O} and higher.
6754 @item -ftree-builtin-call-dce
6755 @opindex ftree-builtin-call-dce
6756 Perform conditional dead code elimination (DCE) for calls to builtin functions
6757 that may set @code{errno} but are otherwise side-effect free. This flag is
6758 enabled by default at @option{-O2} and higher if @option{-Os} is not also
6761 @item -ftree-dominator-opts
6762 @opindex ftree-dominator-opts
6763 Perform a variety of simple scalar cleanups (constant/copy
6764 propagation, redundancy elimination, range propagation and expression
6765 simplification) based on a dominator tree traversal. This also
6766 performs jump threading (to reduce jumps to jumps). This flag is
6767 enabled by default at @option{-O} and higher.
6771 Perform dead store elimination (DSE) on trees. A dead store is a store into
6772 a memory location which will later be overwritten by another store without
6773 any intervening loads. In this case the earlier store can be deleted. This
6774 flag is enabled by default at @option{-O} and higher.
6778 Perform loop header copying on trees. This is beneficial since it increases
6779 effectiveness of code motion optimizations. It also saves one jump. This flag
6780 is enabled by default at @option{-O} and higher. It is not enabled
6781 for @option{-Os}, since it usually increases code size.
6783 @item -ftree-loop-optimize
6784 @opindex ftree-loop-optimize
6785 Perform loop optimizations on trees. This flag is enabled by default
6786 at @option{-O} and higher.
6788 @item -ftree-loop-linear
6789 @opindex ftree-loop-linear
6790 Perform linear loop transformations on tree. This flag can improve cache
6791 performance and allow further loop optimizations to take place.
6793 @item -floop-interchange
6794 Perform loop interchange transformations on loops. Interchanging two
6795 nested loops switches the inner and outer loops. For example, given a
6800 A(J, I) = A(J, I) * C
6804 loop interchange will transform the loop as if the user had written:
6808 A(J, I) = A(J, I) * C
6812 which can be beneficial when @code{N} is larger than the caches,
6813 because in Fortran, the elements of an array are stored in memory
6814 contiguously by column, and the original loop iterates over rows,
6815 potentially creating at each access a cache miss. This optimization
6816 applies to all the languages supported by GCC and is not limited to
6817 Fortran. To use this code transformation, GCC has to be configured
6818 with @option{--with-ppl} and @option{--with-cloog} to enable the
6819 Graphite loop transformation infrastructure.
6821 @item -floop-strip-mine
6822 Perform loop strip mining transformations on loops. Strip mining
6823 splits a loop into two nested loops. The outer loop has strides
6824 equal to the strip size and the inner loop has strides of the
6825 original loop within a strip. The strip length can be changed
6826 using the @option{loop-block-tile-size} parameter. For example,
6833 loop strip mining will transform the loop as if the user had written:
6836 DO I = II, min (II + 50, N)
6841 This optimization applies to all the languages supported by GCC and is
6842 not limited to Fortran. To use this code transformation, GCC has to
6843 be configured with @option{--with-ppl} and @option{--with-cloog} to
6844 enable the Graphite loop transformation infrastructure.
6847 Perform loop blocking transformations on loops. Blocking strip mines
6848 each loop in the loop nest such that the memory accesses of the
6849 element loops fit inside caches. The strip length can be changed
6850 using the @option{loop-block-tile-size} parameter. For example, given
6855 A(J, I) = B(I) + C(J)
6859 loop blocking will transform the loop as if the user had written:
6863 DO I = II, min (II + 50, N)
6864 DO J = JJ, min (JJ + 50, M)
6865 A(J, I) = B(I) + C(J)
6871 which can be beneficial when @code{M} is larger than the caches,
6872 because the innermost loop will iterate over a smaller amount of data
6873 that can be kept in the caches. This optimization applies to all the
6874 languages supported by GCC and is not limited to Fortran. To use this
6875 code transformation, GCC has to be configured with @option{--with-ppl}
6876 and @option{--with-cloog} to enable the Graphite loop transformation
6879 @item -fgraphite-identity
6880 @opindex fgraphite-identity
6881 Enable the identity transformation for graphite. For every SCoP we generate
6882 the polyhedral representation and transform it back to gimple. Using
6883 @option{-fgraphite-identity} we can check the costs or benefits of the
6884 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
6885 are also performed by the code generator CLooG, like index splitting and
6886 dead code elimination in loops.
6888 @item -floop-parallelize-all
6889 Use the Graphite data dependence analysis to identify loops that can
6890 be parallelized. Parallelize all the loops that can be analyzed to
6891 not contain loop carried dependences without checking that it is
6892 profitable to parallelize the loops.
6894 @item -fcheck-data-deps
6895 @opindex fcheck-data-deps
6896 Compare the results of several data dependence analyzers. This option
6897 is used for debugging the data dependence analyzers.
6899 @item -ftree-loop-if-convert
6900 Attempt to transform conditional jumps in the innermost loops to
6901 branch-less equivalents. The intent is to remove control-flow from
6902 the innermost loops in order to improve the ability of the
6903 vectorization pass to handle these loops. This is enabled by default
6904 if vectorization is enabled.
6906 @item -ftree-loop-distribution
6907 Perform loop distribution. This flag can improve cache performance on
6908 big loop bodies and allow further loop optimizations, like
6909 parallelization or vectorization, to take place. For example, the loop
6926 @item -ftree-loop-im
6927 @opindex ftree-loop-im
6928 Perform loop invariant motion on trees. This pass moves only invariants that
6929 would be hard to handle at RTL level (function calls, operations that expand to
6930 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
6931 operands of conditions that are invariant out of the loop, so that we can use
6932 just trivial invariantness analysis in loop unswitching. The pass also includes
6935 @item -ftree-loop-ivcanon
6936 @opindex ftree-loop-ivcanon
6937 Create a canonical counter for number of iterations in the loop for that
6938 determining number of iterations requires complicated analysis. Later
6939 optimizations then may determine the number easily. Useful especially
6940 in connection with unrolling.
6944 Perform induction variable optimizations (strength reduction, induction
6945 variable merging and induction variable elimination) on trees.
6947 @item -ftree-parallelize-loops=n
6948 @opindex ftree-parallelize-loops
6949 Parallelize loops, i.e., split their iteration space to run in n threads.
6950 This is only possible for loops whose iterations are independent
6951 and can be arbitrarily reordered. The optimization is only
6952 profitable on multiprocessor machines, for loops that are CPU-intensive,
6953 rather than constrained e.g.@: by memory bandwidth. This option
6954 implies @option{-pthread}, and thus is only supported on targets
6955 that have support for @option{-pthread}.
6959 Perform function-local points-to analysis on trees. This flag is
6960 enabled by default at @option{-O} and higher.
6964 Perform scalar replacement of aggregates. This pass replaces structure
6965 references with scalars to prevent committing structures to memory too
6966 early. This flag is enabled by default at @option{-O} and higher.
6968 @item -ftree-copyrename
6969 @opindex ftree-copyrename
6970 Perform copy renaming on trees. This pass attempts to rename compiler
6971 temporaries to other variables at copy locations, usually resulting in
6972 variable names which more closely resemble the original variables. This flag
6973 is enabled by default at @option{-O} and higher.
6977 Perform temporary expression replacement during the SSA->normal phase. Single
6978 use/single def temporaries are replaced at their use location with their
6979 defining expression. This results in non-GIMPLE code, but gives the expanders
6980 much more complex trees to work on resulting in better RTL generation. This is
6981 enabled by default at @option{-O} and higher.
6983 @item -ftree-vectorize
6984 @opindex ftree-vectorize
6985 Perform loop vectorization on trees. This flag is enabled by default at
6988 @item -ftree-slp-vectorize
6989 @opindex ftree-slp-vectorize
6990 Perform basic block vectorization on trees. This flag is enabled by default at
6991 @option{-O3} and when @option{-ftree-vectorize} is enabled.
6993 @item -ftree-vect-loop-version
6994 @opindex ftree-vect-loop-version
6995 Perform loop versioning when doing loop vectorization on trees. When a loop
6996 appears to be vectorizable except that data alignment or data dependence cannot
6997 be determined at compile time then vectorized and non-vectorized versions of
6998 the loop are generated along with runtime checks for alignment or dependence
6999 to control which version is executed. This option is enabled by default
7000 except at level @option{-Os} where it is disabled.
7002 @item -fvect-cost-model
7003 @opindex fvect-cost-model
7004 Enable cost model for vectorization.
7008 Perform Value Range Propagation on trees. This is similar to the
7009 constant propagation pass, but instead of values, ranges of values are
7010 propagated. This allows the optimizers to remove unnecessary range
7011 checks like array bound checks and null pointer checks. This is
7012 enabled by default at @option{-O2} and higher. Null pointer check
7013 elimination is only done if @option{-fdelete-null-pointer-checks} is
7018 Perform tail duplication to enlarge superblock size. This transformation
7019 simplifies the control flow of the function allowing other optimizations to do
7022 @item -funroll-loops
7023 @opindex funroll-loops
7024 Unroll loops whose number of iterations can be determined at compile
7025 time or upon entry to the loop. @option{-funroll-loops} implies
7026 @option{-frerun-cse-after-loop}. This option makes code larger,
7027 and may or may not make it run faster.
7029 @item -funroll-all-loops
7030 @opindex funroll-all-loops
7031 Unroll all loops, even if their number of iterations is uncertain when
7032 the loop is entered. This usually makes programs run more slowly.
7033 @option{-funroll-all-loops} implies the same options as
7034 @option{-funroll-loops},
7036 @item -fsplit-ivs-in-unroller
7037 @opindex fsplit-ivs-in-unroller
7038 Enables expressing of values of induction variables in later iterations
7039 of the unrolled loop using the value in the first iteration. This breaks
7040 long dependency chains, thus improving efficiency of the scheduling passes.
7042 Combination of @option{-fweb} and CSE is often sufficient to obtain the
7043 same effect. However in cases the loop body is more complicated than
7044 a single basic block, this is not reliable. It also does not work at all
7045 on some of the architectures due to restrictions in the CSE pass.
7047 This optimization is enabled by default.
7049 @item -fvariable-expansion-in-unroller
7050 @opindex fvariable-expansion-in-unroller
7051 With this option, the compiler will create multiple copies of some
7052 local variables when unrolling a loop which can result in superior code.
7054 @item -fpartial-inlining
7055 @opindex fpartial-inlining
7056 Inline parts of functions. This option has any effect only
7057 when inlining itself is turned on by the @option{-finline-functions}
7058 or @option{-finline-small-functions} options.
7060 Enabled at level @option{-O2}.
7062 @item -fpredictive-commoning
7063 @opindex fpredictive-commoning
7064 Perform predictive commoning optimization, i.e., reusing computations
7065 (especially memory loads and stores) performed in previous
7066 iterations of loops.
7068 This option is enabled at level @option{-O3}.
7070 @item -fprefetch-loop-arrays
7071 @opindex fprefetch-loop-arrays
7072 If supported by the target machine, generate instructions to prefetch
7073 memory to improve the performance of loops that access large arrays.
7075 This option may generate better or worse code; results are highly
7076 dependent on the structure of loops within the source code.
7078 Disabled at level @option{-Os}.
7081 @itemx -fno-peephole2
7082 @opindex fno-peephole
7083 @opindex fno-peephole2
7084 Disable any machine-specific peephole optimizations. The difference
7085 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
7086 are implemented in the compiler; some targets use one, some use the
7087 other, a few use both.
7089 @option{-fpeephole} is enabled by default.
7090 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7092 @item -fno-guess-branch-probability
7093 @opindex fno-guess-branch-probability
7094 Do not guess branch probabilities using heuristics.
7096 GCC will use heuristics to guess branch probabilities if they are
7097 not provided by profiling feedback (@option{-fprofile-arcs}). These
7098 heuristics are based on the control flow graph. If some branch probabilities
7099 are specified by @samp{__builtin_expect}, then the heuristics will be
7100 used to guess branch probabilities for the rest of the control flow graph,
7101 taking the @samp{__builtin_expect} info into account. The interactions
7102 between the heuristics and @samp{__builtin_expect} can be complex, and in
7103 some cases, it may be useful to disable the heuristics so that the effects
7104 of @samp{__builtin_expect} are easier to understand.
7106 The default is @option{-fguess-branch-probability} at levels
7107 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7109 @item -freorder-blocks
7110 @opindex freorder-blocks
7111 Reorder basic blocks in the compiled function in order to reduce number of
7112 taken branches and improve code locality.
7114 Enabled at levels @option{-O2}, @option{-O3}.
7116 @item -freorder-blocks-and-partition
7117 @opindex freorder-blocks-and-partition
7118 In addition to reordering basic blocks in the compiled function, in order
7119 to reduce number of taken branches, partitions hot and cold basic blocks
7120 into separate sections of the assembly and .o files, to improve
7121 paging and cache locality performance.
7123 This optimization is automatically turned off in the presence of
7124 exception handling, for linkonce sections, for functions with a user-defined
7125 section attribute and on any architecture that does not support named
7128 @item -freorder-functions
7129 @opindex freorder-functions
7130 Reorder functions in the object file in order to
7131 improve code locality. This is implemented by using special
7132 subsections @code{.text.hot} for most frequently executed functions and
7133 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
7134 the linker so object file format must support named sections and linker must
7135 place them in a reasonable way.
7137 Also profile feedback must be available in to make this option effective. See
7138 @option{-fprofile-arcs} for details.
7140 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7142 @item -fstrict-aliasing
7143 @opindex fstrict-aliasing
7144 Allow the compiler to assume the strictest aliasing rules applicable to
7145 the language being compiled. For C (and C++), this activates
7146 optimizations based on the type of expressions. In particular, an
7147 object of one type is assumed never to reside at the same address as an
7148 object of a different type, unless the types are almost the same. For
7149 example, an @code{unsigned int} can alias an @code{int}, but not a
7150 @code{void*} or a @code{double}. A character type may alias any other
7153 @anchor{Type-punning}Pay special attention to code like this:
7166 The practice of reading from a different union member than the one most
7167 recently written to (called ``type-punning'') is common. Even with
7168 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
7169 is accessed through the union type. So, the code above will work as
7170 expected. @xref{Structures unions enumerations and bit-fields
7171 implementation}. However, this code might not:
7182 Similarly, access by taking the address, casting the resulting pointer
7183 and dereferencing the result has undefined behavior, even if the cast
7184 uses a union type, e.g.:
7188 return ((union a_union *) &d)->i;
7192 The @option{-fstrict-aliasing} option is enabled at levels
7193 @option{-O2}, @option{-O3}, @option{-Os}.
7195 @item -fstrict-overflow
7196 @opindex fstrict-overflow
7197 Allow the compiler to assume strict signed overflow rules, depending
7198 on the language being compiled. For C (and C++) this means that
7199 overflow when doing arithmetic with signed numbers is undefined, which
7200 means that the compiler may assume that it will not happen. This
7201 permits various optimizations. For example, the compiler will assume
7202 that an expression like @code{i + 10 > i} will always be true for
7203 signed @code{i}. This assumption is only valid if signed overflow is
7204 undefined, as the expression is false if @code{i + 10} overflows when
7205 using twos complement arithmetic. When this option is in effect any
7206 attempt to determine whether an operation on signed numbers will
7207 overflow must be written carefully to not actually involve overflow.
7209 This option also allows the compiler to assume strict pointer
7210 semantics: given a pointer to an object, if adding an offset to that
7211 pointer does not produce a pointer to the same object, the addition is
7212 undefined. This permits the compiler to conclude that @code{p + u >
7213 p} is always true for a pointer @code{p} and unsigned integer
7214 @code{u}. This assumption is only valid because pointer wraparound is
7215 undefined, as the expression is false if @code{p + u} overflows using
7216 twos complement arithmetic.
7218 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
7219 that integer signed overflow is fully defined: it wraps. When
7220 @option{-fwrapv} is used, there is no difference between
7221 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
7222 integers. With @option{-fwrapv} certain types of overflow are
7223 permitted. For example, if the compiler gets an overflow when doing
7224 arithmetic on constants, the overflowed value can still be used with
7225 @option{-fwrapv}, but not otherwise.
7227 The @option{-fstrict-overflow} option is enabled at levels
7228 @option{-O2}, @option{-O3}, @option{-Os}.
7230 @item -falign-functions
7231 @itemx -falign-functions=@var{n}
7232 @opindex falign-functions
7233 Align the start of functions to the next power-of-two greater than
7234 @var{n}, skipping up to @var{n} bytes. For instance,
7235 @option{-falign-functions=32} aligns functions to the next 32-byte
7236 boundary, but @option{-falign-functions=24} would align to the next
7237 32-byte boundary only if this can be done by skipping 23 bytes or less.
7239 @option{-fno-align-functions} and @option{-falign-functions=1} are
7240 equivalent and mean that functions will not be aligned.
7242 Some assemblers only support this flag when @var{n} is a power of two;
7243 in that case, it is rounded up.
7245 If @var{n} is not specified or is zero, use a machine-dependent default.
7247 Enabled at levels @option{-O2}, @option{-O3}.
7249 @item -falign-labels
7250 @itemx -falign-labels=@var{n}
7251 @opindex falign-labels
7252 Align all branch targets to a power-of-two boundary, skipping up to
7253 @var{n} bytes like @option{-falign-functions}. This option can easily
7254 make code slower, because it must insert dummy operations for when the
7255 branch target is reached in the usual flow of the code.
7257 @option{-fno-align-labels} and @option{-falign-labels=1} are
7258 equivalent and mean that labels will not be aligned.
7260 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
7261 are greater than this value, then their values are used instead.
7263 If @var{n} is not specified or is zero, use a machine-dependent default
7264 which is very likely to be @samp{1}, meaning no alignment.
7266 Enabled at levels @option{-O2}, @option{-O3}.
7269 @itemx -falign-loops=@var{n}
7270 @opindex falign-loops
7271 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7272 like @option{-falign-functions}. The hope is that the loop will be
7273 executed many times, which will make up for any execution of the dummy
7276 @option{-fno-align-loops} and @option{-falign-loops=1} are
7277 equivalent and mean that loops will not be aligned.
7279 If @var{n} is not specified or is zero, use a machine-dependent default.
7281 Enabled at levels @option{-O2}, @option{-O3}.
7284 @itemx -falign-jumps=@var{n}
7285 @opindex falign-jumps
7286 Align branch targets to a power-of-two boundary, for branch targets
7287 where the targets can only be reached by jumping, skipping up to @var{n}
7288 bytes like @option{-falign-functions}. In this case, no dummy operations
7291 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7292 equivalent and mean that loops will not be aligned.
7294 If @var{n} is not specified or is zero, use a machine-dependent default.
7296 Enabled at levels @option{-O2}, @option{-O3}.
7298 @item -funit-at-a-time
7299 @opindex funit-at-a-time
7300 This option is left for compatibility reasons. @option{-funit-at-a-time}
7301 has no effect, while @option{-fno-unit-at-a-time} implies
7302 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7306 @item -fno-toplevel-reorder
7307 @opindex fno-toplevel-reorder
7308 Do not reorder top-level functions, variables, and @code{asm}
7309 statements. Output them in the same order that they appear in the
7310 input file. When this option is used, unreferenced static variables
7311 will not be removed. This option is intended to support existing code
7312 which relies on a particular ordering. For new code, it is better to
7315 Enabled at level @option{-O0}. When disabled explicitly, it also imply
7316 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
7321 Constructs webs as commonly used for register allocation purposes and assign
7322 each web individual pseudo register. This allows the register allocation pass
7323 to operate on pseudos directly, but also strengthens several other optimization
7324 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7325 however, make debugging impossible, since variables will no longer stay in a
7328 Enabled by default with @option{-funroll-loops}.
7330 @item -fwhole-program
7331 @opindex fwhole-program
7332 Assume that the current compilation unit represents the whole program being
7333 compiled. All public functions and variables with the exception of @code{main}
7334 and those merged by attribute @code{externally_visible} become static functions
7335 and in effect are optimized more aggressively by interprocedural optimizers. If @command{gold} is used as the linker plugin, @code{externally_visible} attributes are automatically added to functions (not variable yet due to a current @command{gold} issue) that are accessed outside of LTO objects according to resolution file produced by @command{gold}. For other linkers that cannot generate resolution file, explicit @code{externally_visible} attributes are still necessary.
7336 While this option is equivalent to proper use of the @code{static} keyword for
7337 programs consisting of a single file, in combination with option
7338 @option{-combine}, @option{-flto} or @option{-fwhopr} this flag can be used to
7339 compile many smaller scale programs since the functions and variables become
7340 local for the whole combined compilation unit, not for the single source file
7343 This option implies @option{-fwhole-file} for Fortran programs.
7347 This option runs the standard link-time optimizer. When invoked
7348 with source code, it generates GIMPLE (one of GCC's internal
7349 representations) and writes it to special ELF sections in the object
7350 file. When the object files are linked together, all the function
7351 bodies are read from these ELF sections and instantiated as if they
7352 had been part of the same translation unit.
7354 To use the link-timer optimizer, @option{-flto} needs to be specified at
7355 compile time and during the final link. For example,
7358 gcc -c -O2 -flto foo.c
7359 gcc -c -O2 -flto bar.c
7360 gcc -o myprog -flto -O2 foo.o bar.o
7363 The first two invocations to GCC will save a bytecode representation
7364 of GIMPLE into special ELF sections inside @file{foo.o} and
7365 @file{bar.o}. The final invocation will read the GIMPLE bytecode from
7366 @file{foo.o} and @file{bar.o}, merge the two files into a single
7367 internal image, and compile the result as usual. Since both
7368 @file{foo.o} and @file{bar.o} are merged into a single image, this
7369 causes all the inter-procedural analyses and optimizations in GCC to
7370 work across the two files as if they were a single one. This means,
7371 for example, that the inliner will be able to inline functions in
7372 @file{bar.o} into functions in @file{foo.o} and vice-versa.
7374 Another (simpler) way to enable link-time optimization is,
7377 gcc -o myprog -flto -O2 foo.c bar.c
7380 The above will generate bytecode for @file{foo.c} and @file{bar.c},
7381 merge them together into a single GIMPLE representation and optimize
7382 them as usual to produce @file{myprog}.
7384 The only important thing to keep in mind is that to enable link-time
7385 optimizations the @option{-flto} flag needs to be passed to both the
7386 compile and the link commands.
7388 Note that when a file is compiled with @option{-flto}, the generated
7389 object file will be larger than a regular object file because it will
7390 contain GIMPLE bytecodes and the usual final code. This means that
7391 object files with LTO information can be linked as a normal object
7392 file. So, in the previous example, if the final link is done with
7395 gcc -o myprog foo.o bar.o
7398 The only difference will be that no inter-procedural optimizations
7399 will be applied to produce @file{myprog}. The two object files
7400 @file{foo.o} and @file{bar.o} will be simply sent to the regular
7403 Additionally, the optimization flags used to compile individual files
7404 are not necessarily related to those used at link-time. For instance,
7407 gcc -c -O0 -flto foo.c
7408 gcc -c -O0 -flto bar.c
7409 gcc -o myprog -flto -O3 foo.o bar.o
7412 This will produce individual object files with unoptimized assembler
7413 code, but the resulting binary @file{myprog} will be optimized at
7414 @option{-O3}. Now, if the final binary is generated without
7415 @option{-flto}, then @file{myprog} will not be optimized.
7417 When producing the final binary with @option{-flto}, GCC will only
7418 apply link-time optimizations to those files that contain bytecode.
7419 Therefore, you can mix and match object files and libraries with
7420 GIMPLE bytecodes and final object code. GCC will automatically select
7421 which files to optimize in LTO mode and which files to link without
7424 There are some code generation flags that GCC will preserve when
7425 generating bytecodes, as they need to be used during the final link
7426 stage. Currently, the following options are saved into the GIMPLE
7427 bytecode files: @option{-fPIC}, @option{-fcommon} and all the
7428 @option{-m} target flags.
7430 At link time, these options are read-in and reapplied. Note that the
7431 current implementation makes no attempt at recognizing conflicting
7432 values for these options. If two or more files have a conflicting
7433 value (e.g., one file is compiled with @option{-fPIC} and another
7434 isn't), the compiler will simply use the last value read from the
7435 bytecode files. It is recommended, then, that all the files
7436 participating in the same link be compiled with the same options.
7438 Another feature of LTO is that it is possible to apply interprocedural
7439 optimizations on files written in different languages. This requires
7440 some support in the language front end. Currently, the C, C++ and
7441 Fortran front ends are capable of emitting GIMPLE bytecodes, so
7442 something like this should work
7447 gfortran -c -flto baz.f90
7448 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
7451 Notice that the final link is done with @command{g++} to get the C++
7452 runtime libraries and @option{-lgfortran} is added to get the Fortran
7453 runtime libraries. In general, when mixing languages in LTO mode, you
7454 should use the same link command used when mixing languages in a
7455 regular (non-LTO) compilation. This means that if your build process
7456 was mixing languages before, all you need to add is @option{-flto} to
7457 all the compile and link commands.
7459 If LTO encounters objects with C linkage declared with incompatible
7460 types in separate translation units to be linked together (undefined
7461 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
7462 issued. The behavior is still undefined at runtime.
7464 If object files containing GIMPLE bytecode are stored in a library
7465 archive, say @file{libfoo.a}, it is possible to extract and use them
7466 in an LTO link if you are using @command{gold} as the linker (which,
7467 in turn requires GCC to be configured with @option{--enable-gold}).
7468 To enable this feature, use the flag @option{-fuse-linker-plugin} at
7472 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
7475 With the linker plugin enabled, @command{gold} will extract the needed
7476 GIMPLE files from @file{libfoo.a} and pass them on to the running GCC
7477 to make them part of the aggregated GIMPLE image to be optimized.
7479 If you are not using @command{gold} and/or do not specify
7480 @option{-fuse-linker-plugin} then the objects inside @file{libfoo.a}
7481 will be extracted and linked as usual, but they will not participate
7482 in the LTO optimization process.
7484 Link time optimizations do not require the presence of the whole
7485 program to operate. If the program does not require any symbols to
7486 be exported, it is possible to combine @option{-flto} and
7487 @option{-fwhopr} with @option{-fwhole-program} to allow the
7488 interprocedural optimizers to use more aggressive assumptions which
7489 may lead to improved optimization opportunities.
7491 Regarding portability: the current implementation of LTO makes no
7492 attempt at generating bytecode that can be ported between different
7493 types of hosts. The bytecode files are versioned and there is a
7494 strict version check, so bytecode files generated in one version of
7495 GCC will not work with an older/newer version of GCC.
7497 Link time optimization does not play well with generating debugging
7498 information. Combining @option{-flto} or @option{-fwhopr} with
7499 @option{-g} is experimental.
7501 This option is disabled by default.
7503 @item -fwhopr[=@var{n}]
7505 This option is identical in functionality to @option{-flto} but it
7506 differs in how the final link stage is executed. Instead of loading
7507 all the function bodies in memory, the callgraph is analyzed and
7508 optimization decisions are made (whole program analysis or WPA). Once
7509 optimization decisions are made, the callgraph is partitioned and the
7510 different sections are compiled separately (local transformations or
7511 LTRANS)@. This process allows optimizations on very large programs
7512 that otherwise would not fit in memory. This option enables
7513 @option{-fwpa} and @option{-fltrans} automatically.
7515 If you specify the optional @var{n} the link stage is executed in
7516 parallel using @var{n} parallel jobs by utilizing an installed
7517 @command{make} program. The environment variable @env{MAKE} may be
7518 used to override the program used.
7520 Disabled by default.
7524 This is an internal option used by GCC when compiling with
7525 @option{-fwhopr}. You should never need to use it.
7527 This option runs the link-time optimizer in the whole-program-analysis
7528 (WPA) mode, which reads in summary information from all inputs and
7529 performs a whole-program analysis based on summary information only.
7530 It generates object files for subsequent runs of the link-time
7531 optimizer where individual object files are optimized using both
7532 summary information from the WPA mode and the actual function bodies.
7533 It then drives the LTRANS phase.
7535 Disabled by default.
7539 This is an internal option used by GCC when compiling with
7540 @option{-fwhopr}. You should never need to use it.
7542 This option runs the link-time optimizer in the local-transformation (LTRANS)
7543 mode, which reads in output from a previous run of the LTO in WPA mode.
7544 In the LTRANS mode, LTO optimizes an object and produces the final assembly.
7546 Disabled by default.
7548 @item -fltrans-output-list=@var{file}
7549 @opindex fltrans-output-list
7550 This is an internal option used by GCC when compiling with
7551 @option{-fwhopr}. You should never need to use it.
7553 This option specifies a file to which the names of LTRANS output files are
7554 written. This option is only meaningful in conjunction with @option{-fwpa}.
7556 Disabled by default.
7558 @item -flto-compression-level=@var{n}
7559 This option specifies the level of compression used for intermediate
7560 language written to LTO object files, and is only meaningful in
7561 conjunction with LTO mode (@option{-fwhopr}, @option{-flto}). Valid
7562 values are 0 (no compression) to 9 (maximum compression). Values
7563 outside this range are clamped to either 0 or 9. If the option is not
7564 given, a default balanced compression setting is used.
7567 Prints a report with internal details on the workings of the link-time
7568 optimizer. The contents of this report vary from version to version,
7569 it is meant to be useful to GCC developers when processing object
7570 files in LTO mode (via @option{-fwhopr} or @option{-flto}).
7572 Disabled by default.
7574 @item -fuse-linker-plugin
7575 Enables the extraction of objects with GIMPLE bytecode information
7576 from library archives. This option relies on features available only
7577 in @command{gold}, so to use this you must configure GCC with
7578 @option{--enable-gold}. See @option{-flto} for a description on the
7579 effect of this flag and how to use it.
7581 Disabled by default.
7583 @item -fcprop-registers
7584 @opindex fcprop-registers
7585 After register allocation and post-register allocation instruction splitting,
7586 we perform a copy-propagation pass to try to reduce scheduling dependencies
7587 and occasionally eliminate the copy.
7589 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7591 @item -fprofile-correction
7592 @opindex fprofile-correction
7593 Profiles collected using an instrumented binary for multi-threaded programs may
7594 be inconsistent due to missed counter updates. When this option is specified,
7595 GCC will use heuristics to correct or smooth out such inconsistencies. By
7596 default, GCC will emit an error message when an inconsistent profile is detected.
7598 @item -fprofile-dir=@var{path}
7599 @opindex fprofile-dir
7601 Set the directory to search the profile data files in to @var{path}.
7602 This option affects only the profile data generated by
7603 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
7604 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
7605 and its related options.
7606 By default, GCC will use the current directory as @var{path}
7607 thus the profile data file will appear in the same directory as the object file.
7609 @item -fprofile-generate
7610 @itemx -fprofile-generate=@var{path}
7611 @opindex fprofile-generate
7613 Enable options usually used for instrumenting application to produce
7614 profile useful for later recompilation with profile feedback based
7615 optimization. You must use @option{-fprofile-generate} both when
7616 compiling and when linking your program.
7618 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
7620 If @var{path} is specified, GCC will look at the @var{path} to find
7621 the profile feedback data files. See @option{-fprofile-dir}.
7624 @itemx -fprofile-use=@var{path}
7625 @opindex fprofile-use
7626 Enable profile feedback directed optimizations, and optimizations
7627 generally profitable only with profile feedback available.
7629 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
7630 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
7632 By default, GCC emits an error message if the feedback profiles do not
7633 match the source code. This error can be turned into a warning by using
7634 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
7637 If @var{path} is specified, GCC will look at the @var{path} to find
7638 the profile feedback data files. See @option{-fprofile-dir}.
7641 The following options control compiler behavior regarding floating
7642 point arithmetic. These options trade off between speed and
7643 correctness. All must be specifically enabled.
7647 @opindex ffloat-store
7648 Do not store floating point variables in registers, and inhibit other
7649 options that might change whether a floating point value is taken from a
7652 @cindex floating point precision
7653 This option prevents undesirable excess precision on machines such as
7654 the 68000 where the floating registers (of the 68881) keep more
7655 precision than a @code{double} is supposed to have. Similarly for the
7656 x86 architecture. For most programs, the excess precision does only
7657 good, but a few programs rely on the precise definition of IEEE floating
7658 point. Use @option{-ffloat-store} for such programs, after modifying
7659 them to store all pertinent intermediate computations into variables.
7661 @item -fexcess-precision=@var{style}
7662 @opindex fexcess-precision
7663 This option allows further control over excess precision on machines
7664 where floating-point registers have more precision than the IEEE
7665 @code{float} and @code{double} types and the processor does not
7666 support operations rounding to those types. By default,
7667 @option{-fexcess-precision=fast} is in effect; this means that
7668 operations are carried out in the precision of the registers and that
7669 it is unpredictable when rounding to the types specified in the source
7670 code takes place. When compiling C, if
7671 @option{-fexcess-precision=standard} is specified then excess
7672 precision will follow the rules specified in ISO C99; in particular,
7673 both casts and assignments cause values to be rounded to their
7674 semantic types (whereas @option{-ffloat-store} only affects
7675 assignments). This option is enabled by default for C if a strict
7676 conformance option such as @option{-std=c99} is used.
7679 @option{-fexcess-precision=standard} is not implemented for languages
7680 other than C, and has no effect if
7681 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
7682 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
7683 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
7684 semantics apply without excess precision, and in the latter, rounding
7689 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
7690 @option{-ffinite-math-only}, @option{-fno-rounding-math},
7691 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
7693 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
7695 This option is not turned on by any @option{-O} option since
7696 it can result in incorrect output for programs which depend on
7697 an exact implementation of IEEE or ISO rules/specifications for
7698 math functions. It may, however, yield faster code for programs
7699 that do not require the guarantees of these specifications.
7701 @item -fno-math-errno
7702 @opindex fno-math-errno
7703 Do not set ERRNO after calling math functions that are executed
7704 with a single instruction, e.g., sqrt. A program that relies on
7705 IEEE exceptions for math error handling may want to use this flag
7706 for speed while maintaining IEEE arithmetic compatibility.
7708 This option is not turned on by any @option{-O} option since
7709 it can result in incorrect output for programs which depend on
7710 an exact implementation of IEEE or ISO rules/specifications for
7711 math functions. It may, however, yield faster code for programs
7712 that do not require the guarantees of these specifications.
7714 The default is @option{-fmath-errno}.
7716 On Darwin systems, the math library never sets @code{errno}. There is
7717 therefore no reason for the compiler to consider the possibility that
7718 it might, and @option{-fno-math-errno} is the default.
7720 @item -funsafe-math-optimizations
7721 @opindex funsafe-math-optimizations
7723 Allow optimizations for floating-point arithmetic that (a) assume
7724 that arguments and results are valid and (b) may violate IEEE or
7725 ANSI standards. When used at link-time, it may include libraries
7726 or startup files that change the default FPU control word or other
7727 similar optimizations.
7729 This option is not turned on by any @option{-O} option since
7730 it can result in incorrect output for programs which depend on
7731 an exact implementation of IEEE or ISO rules/specifications for
7732 math functions. It may, however, yield faster code for programs
7733 that do not require the guarantees of these specifications.
7734 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
7735 @option{-fassociative-math} and @option{-freciprocal-math}.
7737 The default is @option{-fno-unsafe-math-optimizations}.
7739 @item -fassociative-math
7740 @opindex fassociative-math
7742 Allow re-association of operands in series of floating-point operations.
7743 This violates the ISO C and C++ language standard by possibly changing
7744 computation result. NOTE: re-ordering may change the sign of zero as
7745 well as ignore NaNs and inhibit or create underflow or overflow (and
7746 thus cannot be used on a code which relies on rounding behavior like
7747 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
7748 and thus may not be used when ordered comparisons are required.
7749 This option requires that both @option{-fno-signed-zeros} and
7750 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
7751 much sense with @option{-frounding-math}. For Fortran the option
7752 is automatically enabled when both @option{-fno-signed-zeros} and
7753 @option{-fno-trapping-math} are in effect.
7755 The default is @option{-fno-associative-math}.
7757 @item -freciprocal-math
7758 @opindex freciprocal-math
7760 Allow the reciprocal of a value to be used instead of dividing by
7761 the value if this enables optimizations. For example @code{x / y}
7762 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
7763 is subject to common subexpression elimination. Note that this loses
7764 precision and increases the number of flops operating on the value.
7766 The default is @option{-fno-reciprocal-math}.
7768 @item -ffinite-math-only
7769 @opindex ffinite-math-only
7770 Allow optimizations for floating-point arithmetic that assume
7771 that arguments and results are not NaNs or +-Infs.
7773 This option is not turned on by any @option{-O} option since
7774 it can result in incorrect output for programs which depend on
7775 an exact implementation of IEEE or ISO rules/specifications for
7776 math functions. It may, however, yield faster code for programs
7777 that do not require the guarantees of these specifications.
7779 The default is @option{-fno-finite-math-only}.
7781 @item -fno-signed-zeros
7782 @opindex fno-signed-zeros
7783 Allow optimizations for floating point arithmetic that ignore the
7784 signedness of zero. IEEE arithmetic specifies the behavior of
7785 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
7786 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
7787 This option implies that the sign of a zero result isn't significant.
7789 The default is @option{-fsigned-zeros}.
7791 @item -fno-trapping-math
7792 @opindex fno-trapping-math
7793 Compile code assuming that floating-point operations cannot generate
7794 user-visible traps. These traps include division by zero, overflow,
7795 underflow, inexact result and invalid operation. This option requires
7796 that @option{-fno-signaling-nans} be in effect. Setting this option may
7797 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
7799 This option should never be turned on by any @option{-O} option since
7800 it can result in incorrect output for programs which depend on
7801 an exact implementation of IEEE or ISO rules/specifications for
7804 The default is @option{-ftrapping-math}.
7806 @item -frounding-math
7807 @opindex frounding-math
7808 Disable transformations and optimizations that assume default floating
7809 point rounding behavior. This is round-to-zero for all floating point
7810 to integer conversions, and round-to-nearest for all other arithmetic
7811 truncations. This option should be specified for programs that change
7812 the FP rounding mode dynamically, or that may be executed with a
7813 non-default rounding mode. This option disables constant folding of
7814 floating point expressions at compile-time (which may be affected by
7815 rounding mode) and arithmetic transformations that are unsafe in the
7816 presence of sign-dependent rounding modes.
7818 The default is @option{-fno-rounding-math}.
7820 This option is experimental and does not currently guarantee to
7821 disable all GCC optimizations that are affected by rounding mode.
7822 Future versions of GCC may provide finer control of this setting
7823 using C99's @code{FENV_ACCESS} pragma. This command line option
7824 will be used to specify the default state for @code{FENV_ACCESS}.
7826 @item -fsignaling-nans
7827 @opindex fsignaling-nans
7828 Compile code assuming that IEEE signaling NaNs may generate user-visible
7829 traps during floating-point operations. Setting this option disables
7830 optimizations that may change the number of exceptions visible with
7831 signaling NaNs. This option implies @option{-ftrapping-math}.
7833 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
7836 The default is @option{-fno-signaling-nans}.
7838 This option is experimental and does not currently guarantee to
7839 disable all GCC optimizations that affect signaling NaN behavior.
7841 @item -fsingle-precision-constant
7842 @opindex fsingle-precision-constant
7843 Treat floating point constant as single precision constant instead of
7844 implicitly converting it to double precision constant.
7846 @item -fcx-limited-range
7847 @opindex fcx-limited-range
7848 When enabled, this option states that a range reduction step is not
7849 needed when performing complex division. Also, there is no checking
7850 whether the result of a complex multiplication or division is @code{NaN
7851 + I*NaN}, with an attempt to rescue the situation in that case. The
7852 default is @option{-fno-cx-limited-range}, but is enabled by
7853 @option{-ffast-math}.
7855 This option controls the default setting of the ISO C99
7856 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
7859 @item -fcx-fortran-rules
7860 @opindex fcx-fortran-rules
7861 Complex multiplication and division follow Fortran rules. Range
7862 reduction is done as part of complex division, but there is no checking
7863 whether the result of a complex multiplication or division is @code{NaN
7864 + I*NaN}, with an attempt to rescue the situation in that case.
7866 The default is @option{-fno-cx-fortran-rules}.
7870 The following options control optimizations that may improve
7871 performance, but are not enabled by any @option{-O} options. This
7872 section includes experimental options that may produce broken code.
7875 @item -fbranch-probabilities
7876 @opindex fbranch-probabilities
7877 After running a program compiled with @option{-fprofile-arcs}
7878 (@pxref{Debugging Options,, Options for Debugging Your Program or
7879 @command{gcc}}), you can compile it a second time using
7880 @option{-fbranch-probabilities}, to improve optimizations based on
7881 the number of times each branch was taken. When the program
7882 compiled with @option{-fprofile-arcs} exits it saves arc execution
7883 counts to a file called @file{@var{sourcename}.gcda} for each source
7884 file. The information in this data file is very dependent on the
7885 structure of the generated code, so you must use the same source code
7886 and the same optimization options for both compilations.
7888 With @option{-fbranch-probabilities}, GCC puts a
7889 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
7890 These can be used to improve optimization. Currently, they are only
7891 used in one place: in @file{reorg.c}, instead of guessing which path a
7892 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
7893 exactly determine which path is taken more often.
7895 @item -fprofile-values
7896 @opindex fprofile-values
7897 If combined with @option{-fprofile-arcs}, it adds code so that some
7898 data about values of expressions in the program is gathered.
7900 With @option{-fbranch-probabilities}, it reads back the data gathered
7901 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
7902 notes to instructions for their later usage in optimizations.
7904 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
7908 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
7909 a code to gather information about values of expressions.
7911 With @option{-fbranch-probabilities}, it reads back the data gathered
7912 and actually performs the optimizations based on them.
7913 Currently the optimizations include specialization of division operation
7914 using the knowledge about the value of the denominator.
7916 @item -frename-registers
7917 @opindex frename-registers
7918 Attempt to avoid false dependencies in scheduled code by making use
7919 of registers left over after register allocation. This optimization
7920 will most benefit processors with lots of registers. Depending on the
7921 debug information format adopted by the target, however, it can
7922 make debugging impossible, since variables will no longer stay in
7923 a ``home register''.
7925 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
7929 Perform tail duplication to enlarge superblock size. This transformation
7930 simplifies the control flow of the function allowing other optimizations to do
7933 Enabled with @option{-fprofile-use}.
7935 @item -funroll-loops
7936 @opindex funroll-loops
7937 Unroll loops whose number of iterations can be determined at compile time or
7938 upon entry to the loop. @option{-funroll-loops} implies
7939 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
7940 It also turns on complete loop peeling (i.e.@: complete removal of loops with
7941 small constant number of iterations). This option makes code larger, and may
7942 or may not make it run faster.
7944 Enabled with @option{-fprofile-use}.
7946 @item -funroll-all-loops
7947 @opindex funroll-all-loops
7948 Unroll all loops, even if their number of iterations is uncertain when
7949 the loop is entered. This usually makes programs run more slowly.
7950 @option{-funroll-all-loops} implies the same options as
7951 @option{-funroll-loops}.
7954 @opindex fpeel-loops
7955 Peels the loops for that there is enough information that they do not
7956 roll much (from profile feedback). It also turns on complete loop peeling
7957 (i.e.@: complete removal of loops with small constant number of iterations).
7959 Enabled with @option{-fprofile-use}.
7961 @item -fmove-loop-invariants
7962 @opindex fmove-loop-invariants
7963 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
7964 at level @option{-O1}
7966 @item -funswitch-loops
7967 @opindex funswitch-loops
7968 Move branches with loop invariant conditions out of the loop, with duplicates
7969 of the loop on both branches (modified according to result of the condition).
7971 @item -ffunction-sections
7972 @itemx -fdata-sections
7973 @opindex ffunction-sections
7974 @opindex fdata-sections
7975 Place each function or data item into its own section in the output
7976 file if the target supports arbitrary sections. The name of the
7977 function or the name of the data item determines the section's name
7980 Use these options on systems where the linker can perform optimizations
7981 to improve locality of reference in the instruction space. Most systems
7982 using the ELF object format and SPARC processors running Solaris 2 have
7983 linkers with such optimizations. AIX may have these optimizations in
7986 Only use these options when there are significant benefits from doing
7987 so. When you specify these options, the assembler and linker will
7988 create larger object and executable files and will also be slower.
7989 You will not be able to use @code{gprof} on all systems if you
7990 specify this option and you may have problems with debugging if
7991 you specify both this option and @option{-g}.
7993 @item -fbranch-target-load-optimize
7994 @opindex fbranch-target-load-optimize
7995 Perform branch target register load optimization before prologue / epilogue
7997 The use of target registers can typically be exposed only during reload,
7998 thus hoisting loads out of loops and doing inter-block scheduling needs
7999 a separate optimization pass.
8001 @item -fbranch-target-load-optimize2
8002 @opindex fbranch-target-load-optimize2
8003 Perform branch target register load optimization after prologue / epilogue
8006 @item -fbtr-bb-exclusive
8007 @opindex fbtr-bb-exclusive
8008 When performing branch target register load optimization, don't reuse
8009 branch target registers in within any basic block.
8011 @item -fstack-protector
8012 @opindex fstack-protector
8013 Emit extra code to check for buffer overflows, such as stack smashing
8014 attacks. This is done by adding a guard variable to functions with
8015 vulnerable objects. This includes functions that call alloca, and
8016 functions with buffers larger than 8 bytes. The guards are initialized
8017 when a function is entered and then checked when the function exits.
8018 If a guard check fails, an error message is printed and the program exits.
8020 @item -fstack-protector-all
8021 @opindex fstack-protector-all
8022 Like @option{-fstack-protector} except that all functions are protected.
8024 @item -fsection-anchors
8025 @opindex fsection-anchors
8026 Try to reduce the number of symbolic address calculations by using
8027 shared ``anchor'' symbols to address nearby objects. This transformation
8028 can help to reduce the number of GOT entries and GOT accesses on some
8031 For example, the implementation of the following function @code{foo}:
8035 int foo (void) @{ return a + b + c; @}
8038 would usually calculate the addresses of all three variables, but if you
8039 compile it with @option{-fsection-anchors}, it will access the variables
8040 from a common anchor point instead. The effect is similar to the
8041 following pseudocode (which isn't valid C):
8046 register int *xr = &x;
8047 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
8051 Not all targets support this option.
8053 @item --param @var{name}=@var{value}
8055 In some places, GCC uses various constants to control the amount of
8056 optimization that is done. For example, GCC will not inline functions
8057 that contain more that a certain number of instructions. You can
8058 control some of these constants on the command-line using the
8059 @option{--param} option.
8061 The names of specific parameters, and the meaning of the values, are
8062 tied to the internals of the compiler, and are subject to change
8063 without notice in future releases.
8065 In each case, the @var{value} is an integer. The allowable choices for
8066 @var{name} are given in the following table:
8069 @item struct-reorg-cold-struct-ratio
8070 The threshold ratio (as a percentage) between a structure frequency
8071 and the frequency of the hottest structure in the program. This parameter
8072 is used by struct-reorg optimization enabled by @option{-fipa-struct-reorg}.
8073 We say that if the ratio of a structure frequency, calculated by profiling,
8074 to the hottest structure frequency in the program is less than this
8075 parameter, then structure reorganization is not applied to this structure.
8078 @item predictable-branch-outcome
8079 When branch is predicted to be taken with probability lower than this threshold
8080 (in percent), then it is considered well predictable. The default is 10.
8082 @item max-crossjump-edges
8083 The maximum number of incoming edges to consider for crossjumping.
8084 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
8085 the number of edges incoming to each block. Increasing values mean
8086 more aggressive optimization, making the compile time increase with
8087 probably small improvement in executable size.
8089 @item min-crossjump-insns
8090 The minimum number of instructions which must be matched at the end
8091 of two blocks before crossjumping will be performed on them. This
8092 value is ignored in the case where all instructions in the block being
8093 crossjumped from are matched. The default value is 5.
8095 @item max-grow-copy-bb-insns
8096 The maximum code size expansion factor when copying basic blocks
8097 instead of jumping. The expansion is relative to a jump instruction.
8098 The default value is 8.
8100 @item max-goto-duplication-insns
8101 The maximum number of instructions to duplicate to a block that jumps
8102 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
8103 passes, GCC factors computed gotos early in the compilation process,
8104 and unfactors them as late as possible. Only computed jumps at the
8105 end of a basic blocks with no more than max-goto-duplication-insns are
8106 unfactored. The default value is 8.
8108 @item max-delay-slot-insn-search
8109 The maximum number of instructions to consider when looking for an
8110 instruction to fill a delay slot. If more than this arbitrary number of
8111 instructions is searched, the time savings from filling the delay slot
8112 will be minimal so stop searching. Increasing values mean more
8113 aggressive optimization, making the compile time increase with probably
8114 small improvement in executable run time.
8116 @item max-delay-slot-live-search
8117 When trying to fill delay slots, the maximum number of instructions to
8118 consider when searching for a block with valid live register
8119 information. Increasing this arbitrarily chosen value means more
8120 aggressive optimization, increasing the compile time. This parameter
8121 should be removed when the delay slot code is rewritten to maintain the
8124 @item max-gcse-memory
8125 The approximate maximum amount of memory that will be allocated in
8126 order to perform the global common subexpression elimination
8127 optimization. If more memory than specified is required, the
8128 optimization will not be done.
8130 @item max-pending-list-length
8131 The maximum number of pending dependencies scheduling will allow
8132 before flushing the current state and starting over. Large functions
8133 with few branches or calls can create excessively large lists which
8134 needlessly consume memory and resources.
8136 @item max-inline-insns-single
8137 Several parameters control the tree inliner used in gcc.
8138 This number sets the maximum number of instructions (counted in GCC's
8139 internal representation) in a single function that the tree inliner
8140 will consider for inlining. This only affects functions declared
8141 inline and methods implemented in a class declaration (C++).
8142 The default value is 300.
8144 @item max-inline-insns-auto
8145 When you use @option{-finline-functions} (included in @option{-O3}),
8146 a lot of functions that would otherwise not be considered for inlining
8147 by the compiler will be investigated. To those functions, a different
8148 (more restrictive) limit compared to functions declared inline can
8150 The default value is 40.
8152 @item large-function-insns
8153 The limit specifying really large functions. For functions larger than this
8154 limit after inlining, inlining is constrained by
8155 @option{--param large-function-growth}. This parameter is useful primarily
8156 to avoid extreme compilation time caused by non-linear algorithms used by the
8158 The default value is 2700.
8160 @item large-function-growth
8161 Specifies maximal growth of large function caused by inlining in percents.
8162 The default value is 100 which limits large function growth to 2.0 times
8165 @item large-unit-insns
8166 The limit specifying large translation unit. Growth caused by inlining of
8167 units larger than this limit is limited by @option{--param inline-unit-growth}.
8168 For small units this might be too tight (consider unit consisting of function A
8169 that is inline and B that just calls A three time. If B is small relative to
8170 A, the growth of unit is 300\% and yet such inlining is very sane. For very
8171 large units consisting of small inlineable functions however the overall unit
8172 growth limit is needed to avoid exponential explosion of code size. Thus for
8173 smaller units, the size is increased to @option{--param large-unit-insns}
8174 before applying @option{--param inline-unit-growth}. The default is 10000
8176 @item inline-unit-growth
8177 Specifies maximal overall growth of the compilation unit caused by inlining.
8178 The default value is 30 which limits unit growth to 1.3 times the original
8181 @item ipcp-unit-growth
8182 Specifies maximal overall growth of the compilation unit caused by
8183 interprocedural constant propagation. The default value is 10 which limits
8184 unit growth to 1.1 times the original size.
8186 @item large-stack-frame
8187 The limit specifying large stack frames. While inlining the algorithm is trying
8188 to not grow past this limit too much. Default value is 256 bytes.
8190 @item large-stack-frame-growth
8191 Specifies maximal growth of large stack frames caused by inlining in percents.
8192 The default value is 1000 which limits large stack frame growth to 11 times
8195 @item max-inline-insns-recursive
8196 @itemx max-inline-insns-recursive-auto
8197 Specifies maximum number of instructions out-of-line copy of self recursive inline
8198 function can grow into by performing recursive inlining.
8200 For functions declared inline @option{--param max-inline-insns-recursive} is
8201 taken into account. For function not declared inline, recursive inlining
8202 happens only when @option{-finline-functions} (included in @option{-O3}) is
8203 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
8204 default value is 450.
8206 @item max-inline-recursive-depth
8207 @itemx max-inline-recursive-depth-auto
8208 Specifies maximum recursion depth used by the recursive inlining.
8210 For functions declared inline @option{--param max-inline-recursive-depth} is
8211 taken into account. For function not declared inline, recursive inlining
8212 happens only when @option{-finline-functions} (included in @option{-O3}) is
8213 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
8216 @item min-inline-recursive-probability
8217 Recursive inlining is profitable only for function having deep recursion
8218 in average and can hurt for function having little recursion depth by
8219 increasing the prologue size or complexity of function body to other
8222 When profile feedback is available (see @option{-fprofile-generate}) the actual
8223 recursion depth can be guessed from probability that function will recurse via
8224 given call expression. This parameter limits inlining only to call expression
8225 whose probability exceeds given threshold (in percents). The default value is
8228 @item early-inlining-insns
8229 Specify growth that early inliner can make. In effect it increases amount of
8230 inlining for code having large abstraction penalty. The default value is 8.
8232 @item max-early-inliner-iterations
8233 @itemx max-early-inliner-iterations
8234 Limit of iterations of early inliner. This basically bounds number of nested
8235 indirect calls early inliner can resolve. Deeper chains are still handled by
8238 @item min-vect-loop-bound
8239 The minimum number of iterations under which a loop will not get vectorized
8240 when @option{-ftree-vectorize} is used. The number of iterations after
8241 vectorization needs to be greater than the value specified by this option
8242 to allow vectorization. The default value is 0.
8244 @item max-unrolled-insns
8245 The maximum number of instructions that a loop should have if that loop
8246 is unrolled, and if the loop is unrolled, it determines how many times
8247 the loop code is unrolled.
8249 @item max-average-unrolled-insns
8250 The maximum number of instructions biased by probabilities of their execution
8251 that a loop should have if that loop is unrolled, and if the loop is unrolled,
8252 it determines how many times the loop code is unrolled.
8254 @item max-unroll-times
8255 The maximum number of unrollings of a single loop.
8257 @item max-peeled-insns
8258 The maximum number of instructions that a loop should have if that loop
8259 is peeled, and if the loop is peeled, it determines how many times
8260 the loop code is peeled.
8262 @item max-peel-times
8263 The maximum number of peelings of a single loop.
8265 @item max-completely-peeled-insns
8266 The maximum number of insns of a completely peeled loop.
8268 @item max-completely-peel-times
8269 The maximum number of iterations of a loop to be suitable for complete peeling.
8271 @item max-completely-peel-loop-nest-depth
8272 The maximum depth of a loop nest suitable for complete peeling.
8274 @item max-unswitch-insns
8275 The maximum number of insns of an unswitched loop.
8277 @item max-unswitch-level
8278 The maximum number of branches unswitched in a single loop.
8281 The minimum cost of an expensive expression in the loop invariant motion.
8283 @item iv-consider-all-candidates-bound
8284 Bound on number of candidates for induction variables below that
8285 all candidates are considered for each use in induction variable
8286 optimizations. Only the most relevant candidates are considered
8287 if there are more candidates, to avoid quadratic time complexity.
8289 @item iv-max-considered-uses
8290 The induction variable optimizations give up on loops that contain more
8291 induction variable uses.
8293 @item iv-always-prune-cand-set-bound
8294 If number of candidates in the set is smaller than this value,
8295 we always try to remove unnecessary ivs from the set during its
8296 optimization when a new iv is added to the set.
8298 @item scev-max-expr-size
8299 Bound on size of expressions used in the scalar evolutions analyzer.
8300 Large expressions slow the analyzer.
8302 @item omega-max-vars
8303 The maximum number of variables in an Omega constraint system.
8304 The default value is 128.
8306 @item omega-max-geqs
8307 The maximum number of inequalities in an Omega constraint system.
8308 The default value is 256.
8311 The maximum number of equalities in an Omega constraint system.
8312 The default value is 128.
8314 @item omega-max-wild-cards
8315 The maximum number of wildcard variables that the Omega solver will
8316 be able to insert. The default value is 18.
8318 @item omega-hash-table-size
8319 The size of the hash table in the Omega solver. The default value is
8322 @item omega-max-keys
8323 The maximal number of keys used by the Omega solver. The default
8326 @item omega-eliminate-redundant-constraints
8327 When set to 1, use expensive methods to eliminate all redundant
8328 constraints. The default value is 0.
8330 @item vect-max-version-for-alignment-checks
8331 The maximum number of runtime checks that can be performed when
8332 doing loop versioning for alignment in the vectorizer. See option
8333 ftree-vect-loop-version for more information.
8335 @item vect-max-version-for-alias-checks
8336 The maximum number of runtime checks that can be performed when
8337 doing loop versioning for alias in the vectorizer. See option
8338 ftree-vect-loop-version for more information.
8340 @item max-iterations-to-track
8342 The maximum number of iterations of a loop the brute force algorithm
8343 for analysis of # of iterations of the loop tries to evaluate.
8345 @item hot-bb-count-fraction
8346 Select fraction of the maximal count of repetitions of basic block in program
8347 given basic block needs to have to be considered hot.
8349 @item hot-bb-frequency-fraction
8350 Select fraction of the maximal frequency of executions of basic block in
8351 function given basic block needs to have to be considered hot
8353 @item max-predicted-iterations
8354 The maximum number of loop iterations we predict statically. This is useful
8355 in cases where function contain single loop with known bound and other loop
8356 with unknown. We predict the known number of iterations correctly, while
8357 the unknown number of iterations average to roughly 10. This means that the
8358 loop without bounds would appear artificially cold relative to the other one.
8360 @item align-threshold
8362 Select fraction of the maximal frequency of executions of basic block in
8363 function given basic block will get aligned.
8365 @item align-loop-iterations
8367 A loop expected to iterate at lest the selected number of iterations will get
8370 @item tracer-dynamic-coverage
8371 @itemx tracer-dynamic-coverage-feedback
8373 This value is used to limit superblock formation once the given percentage of
8374 executed instructions is covered. This limits unnecessary code size
8377 The @option{tracer-dynamic-coverage-feedback} is used only when profile
8378 feedback is available. The real profiles (as opposed to statically estimated
8379 ones) are much less balanced allowing the threshold to be larger value.
8381 @item tracer-max-code-growth
8382 Stop tail duplication once code growth has reached given percentage. This is
8383 rather hokey argument, as most of the duplicates will be eliminated later in
8384 cross jumping, so it may be set to much higher values than is the desired code
8387 @item tracer-min-branch-ratio
8389 Stop reverse growth when the reverse probability of best edge is less than this
8390 threshold (in percent).
8392 @item tracer-min-branch-ratio
8393 @itemx tracer-min-branch-ratio-feedback
8395 Stop forward growth if the best edge do have probability lower than this
8398 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
8399 compilation for profile feedback and one for compilation without. The value
8400 for compilation with profile feedback needs to be more conservative (higher) in
8401 order to make tracer effective.
8403 @item max-cse-path-length
8405 Maximum number of basic blocks on path that cse considers. The default is 10.
8408 The maximum instructions CSE process before flushing. The default is 1000.
8410 @item ggc-min-expand
8412 GCC uses a garbage collector to manage its own memory allocation. This
8413 parameter specifies the minimum percentage by which the garbage
8414 collector's heap should be allowed to expand between collections.
8415 Tuning this may improve compilation speed; it has no effect on code
8418 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
8419 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
8420 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
8421 GCC is not able to calculate RAM on a particular platform, the lower
8422 bound of 30% is used. Setting this parameter and
8423 @option{ggc-min-heapsize} to zero causes a full collection to occur at
8424 every opportunity. This is extremely slow, but can be useful for
8427 @item ggc-min-heapsize
8429 Minimum size of the garbage collector's heap before it begins bothering
8430 to collect garbage. The first collection occurs after the heap expands
8431 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
8432 tuning this may improve compilation speed, and has no effect on code
8435 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
8436 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
8437 with a lower bound of 4096 (four megabytes) and an upper bound of
8438 131072 (128 megabytes). If GCC is not able to calculate RAM on a
8439 particular platform, the lower bound is used. Setting this parameter
8440 very large effectively disables garbage collection. Setting this
8441 parameter and @option{ggc-min-expand} to zero causes a full collection
8442 to occur at every opportunity.
8444 @item max-reload-search-insns
8445 The maximum number of instruction reload should look backward for equivalent
8446 register. Increasing values mean more aggressive optimization, making the
8447 compile time increase with probably slightly better performance. The default
8450 @item max-cselib-memory-locations
8451 The maximum number of memory locations cselib should take into account.
8452 Increasing values mean more aggressive optimization, making the compile time
8453 increase with probably slightly better performance. The default value is 500.
8455 @item reorder-blocks-duplicate
8456 @itemx reorder-blocks-duplicate-feedback
8458 Used by basic block reordering pass to decide whether to use unconditional
8459 branch or duplicate the code on its destination. Code is duplicated when its
8460 estimated size is smaller than this value multiplied by the estimated size of
8461 unconditional jump in the hot spots of the program.
8463 The @option{reorder-block-duplicate-feedback} is used only when profile
8464 feedback is available and may be set to higher values than
8465 @option{reorder-block-duplicate} since information about the hot spots is more
8468 @item max-sched-ready-insns
8469 The maximum number of instructions ready to be issued the scheduler should
8470 consider at any given time during the first scheduling pass. Increasing
8471 values mean more thorough searches, making the compilation time increase
8472 with probably little benefit. The default value is 100.
8474 @item max-sched-region-blocks
8475 The maximum number of blocks in a region to be considered for
8476 interblock scheduling. The default value is 10.
8478 @item max-pipeline-region-blocks
8479 The maximum number of blocks in a region to be considered for
8480 pipelining in the selective scheduler. The default value is 15.
8482 @item max-sched-region-insns
8483 The maximum number of insns in a region to be considered for
8484 interblock scheduling. The default value is 100.
8486 @item max-pipeline-region-insns
8487 The maximum number of insns in a region to be considered for
8488 pipelining in the selective scheduler. The default value is 200.
8491 The minimum probability (in percents) of reaching a source block
8492 for interblock speculative scheduling. The default value is 40.
8494 @item max-sched-extend-regions-iters
8495 The maximum number of iterations through CFG to extend regions.
8496 0 - disable region extension,
8497 N - do at most N iterations.
8498 The default value is 0.
8500 @item max-sched-insn-conflict-delay
8501 The maximum conflict delay for an insn to be considered for speculative motion.
8502 The default value is 3.
8504 @item sched-spec-prob-cutoff
8505 The minimal probability of speculation success (in percents), so that
8506 speculative insn will be scheduled.
8507 The default value is 40.
8509 @item sched-mem-true-dep-cost
8510 Minimal distance (in CPU cycles) between store and load targeting same
8511 memory locations. The default value is 1.
8513 @item selsched-max-lookahead
8514 The maximum size of the lookahead window of selective scheduling. It is a
8515 depth of search for available instructions.
8516 The default value is 50.
8518 @item selsched-max-sched-times
8519 The maximum number of times that an instruction will be scheduled during
8520 selective scheduling. This is the limit on the number of iterations
8521 through which the instruction may be pipelined. The default value is 2.
8523 @item selsched-max-insns-to-rename
8524 The maximum number of best instructions in the ready list that are considered
8525 for renaming in the selective scheduler. The default value is 2.
8527 @item max-last-value-rtl
8528 The maximum size measured as number of RTLs that can be recorded in an expression
8529 in combiner for a pseudo register as last known value of that register. The default
8532 @item integer-share-limit
8533 Small integer constants can use a shared data structure, reducing the
8534 compiler's memory usage and increasing its speed. This sets the maximum
8535 value of a shared integer constant. The default value is 256.
8537 @item min-virtual-mappings
8538 Specifies the minimum number of virtual mappings in the incremental
8539 SSA updater that should be registered to trigger the virtual mappings
8540 heuristic defined by virtual-mappings-ratio. The default value is
8543 @item virtual-mappings-ratio
8544 If the number of virtual mappings is virtual-mappings-ratio bigger
8545 than the number of virtual symbols to be updated, then the incremental
8546 SSA updater switches to a full update for those symbols. The default
8549 @item ssp-buffer-size
8550 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
8551 protection when @option{-fstack-protection} is used.
8553 @item max-jump-thread-duplication-stmts
8554 Maximum number of statements allowed in a block that needs to be
8555 duplicated when threading jumps.
8557 @item max-fields-for-field-sensitive
8558 Maximum number of fields in a structure we will treat in
8559 a field sensitive manner during pointer analysis. The default is zero
8560 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
8562 @item prefetch-latency
8563 Estimate on average number of instructions that are executed before
8564 prefetch finishes. The distance we prefetch ahead is proportional
8565 to this constant. Increasing this number may also lead to less
8566 streams being prefetched (see @option{simultaneous-prefetches}).
8568 @item simultaneous-prefetches
8569 Maximum number of prefetches that can run at the same time.
8571 @item l1-cache-line-size
8572 The size of cache line in L1 cache, in bytes.
8575 The size of L1 cache, in kilobytes.
8578 The size of L2 cache, in kilobytes.
8580 @item min-insn-to-prefetch-ratio
8581 The minimum ratio between the number of instructions and the
8582 number of prefetches to enable prefetching in a loop.
8584 @item prefetch-min-insn-to-mem-ratio
8585 The minimum ratio between the number of instructions and the
8586 number of memory references to enable prefetching in a loop.
8588 @item use-canonical-types
8589 Whether the compiler should use the ``canonical'' type system. By
8590 default, this should always be 1, which uses a more efficient internal
8591 mechanism for comparing types in C++ and Objective-C++. However, if
8592 bugs in the canonical type system are causing compilation failures,
8593 set this value to 0 to disable canonical types.
8595 @item switch-conversion-max-branch-ratio
8596 Switch initialization conversion will refuse to create arrays that are
8597 bigger than @option{switch-conversion-max-branch-ratio} times the number of
8598 branches in the switch.
8600 @item max-partial-antic-length
8601 Maximum length of the partial antic set computed during the tree
8602 partial redundancy elimination optimization (@option{-ftree-pre}) when
8603 optimizing at @option{-O3} and above. For some sorts of source code
8604 the enhanced partial redundancy elimination optimization can run away,
8605 consuming all of the memory available on the host machine. This
8606 parameter sets a limit on the length of the sets that are computed,
8607 which prevents the runaway behavior. Setting a value of 0 for
8608 this parameter will allow an unlimited set length.
8610 @item sccvn-max-scc-size
8611 Maximum size of a strongly connected component (SCC) during SCCVN
8612 processing. If this limit is hit, SCCVN processing for the whole
8613 function will not be done and optimizations depending on it will
8614 be disabled. The default maximum SCC size is 10000.
8616 @item ira-max-loops-num
8617 IRA uses a regional register allocation by default. If a function
8618 contains loops more than number given by the parameter, only at most
8619 given number of the most frequently executed loops will form regions
8620 for the regional register allocation. The default value of the
8623 @item ira-max-conflict-table-size
8624 Although IRA uses a sophisticated algorithm of compression conflict
8625 table, the table can be still big for huge functions. If the conflict
8626 table for a function could be more than size in MB given by the
8627 parameter, the conflict table is not built and faster, simpler, and
8628 lower quality register allocation algorithm will be used. The
8629 algorithm do not use pseudo-register conflicts. The default value of
8630 the parameter is 2000.
8632 @item ira-loop-reserved-regs
8633 IRA can be used to evaluate more accurate register pressure in loops
8634 for decision to move loop invariants (see @option{-O3}). The number
8635 of available registers reserved for some other purposes is described
8636 by this parameter. The default value of the parameter is 2 which is
8637 minimal number of registers needed for execution of typical
8638 instruction. This value is the best found from numerous experiments.
8640 @item loop-invariant-max-bbs-in-loop
8641 Loop invariant motion can be very expensive, both in compile time and
8642 in amount of needed compile time memory, with very large loops. Loops
8643 with more basic blocks than this parameter won't have loop invariant
8644 motion optimization performed on them. The default value of the
8645 parameter is 1000 for -O1 and 10000 for -O2 and above.
8647 @item max-vartrack-size
8648 Sets a maximum number of hash table slots to use during variable
8649 tracking dataflow analysis of any function. If this limit is exceeded
8650 with variable tracking at assignments enabled, analysis for that
8651 function is retried without it, after removing all debug insns from
8652 the function. If the limit is exceeded even without debug insns, var
8653 tracking analysis is completely disabled for the function. Setting
8654 the parameter to zero makes it unlimited.
8656 @item min-nondebug-insn-uid
8657 Use uids starting at this parameter for nondebug insns. The range below
8658 the parameter is reserved exclusively for debug insns created by
8659 @option{-fvar-tracking-assignments}, but debug insns may get
8660 (non-overlapping) uids above it if the reserved range is exhausted.
8662 @item ipa-sra-ptr-growth-factor
8663 IPA-SRA will replace a pointer to an aggregate with one or more new
8664 parameters only when their cumulative size is less or equal to
8665 @option{ipa-sra-ptr-growth-factor} times the size of the original
8668 @item graphite-max-nb-scop-params
8669 To avoid exponential effects in the Graphite loop transforms, the
8670 number of parameters in a Static Control Part (SCoP) is bounded. The
8671 default value is 10 parameters. A variable whose value is unknown at
8672 compile time and defined outside a SCoP is a parameter of the SCoP.
8674 @item graphite-max-bbs-per-function
8675 To avoid exponential effects in the detection of SCoPs, the size of
8676 the functions analyzed by Graphite is bounded. The default value is
8679 @item loop-block-tile-size
8680 Loop blocking or strip mining transforms, enabled with
8681 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
8682 loop in the loop nest by a given number of iterations. The strip
8683 length can be changed using the @option{loop-block-tile-size}
8684 parameter. The default value is 51 iterations.
8689 @node Preprocessor Options
8690 @section Options Controlling the Preprocessor
8691 @cindex preprocessor options
8692 @cindex options, preprocessor
8694 These options control the C preprocessor, which is run on each C source
8695 file before actual compilation.
8697 If you use the @option{-E} option, nothing is done except preprocessing.
8698 Some of these options make sense only together with @option{-E} because
8699 they cause the preprocessor output to be unsuitable for actual
8703 @item -Wp,@var{option}
8705 You can use @option{-Wp,@var{option}} to bypass the compiler driver
8706 and pass @var{option} directly through to the preprocessor. If
8707 @var{option} contains commas, it is split into multiple options at the
8708 commas. However, many options are modified, translated or interpreted
8709 by the compiler driver before being passed to the preprocessor, and
8710 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
8711 interface is undocumented and subject to change, so whenever possible
8712 you should avoid using @option{-Wp} and let the driver handle the
8715 @item -Xpreprocessor @var{option}
8716 @opindex Xpreprocessor
8717 Pass @var{option} as an option to the preprocessor. You can use this to
8718 supply system-specific preprocessor options which GCC does not know how to
8721 If you want to pass an option that takes an argument, you must use
8722 @option{-Xpreprocessor} twice, once for the option and once for the argument.
8725 @include cppopts.texi
8727 @node Assembler Options
8728 @section Passing Options to the Assembler
8730 @c prevent bad page break with this line
8731 You can pass options to the assembler.
8734 @item -Wa,@var{option}
8736 Pass @var{option} as an option to the assembler. If @var{option}
8737 contains commas, it is split into multiple options at the commas.
8739 @item -Xassembler @var{option}
8741 Pass @var{option} as an option to the assembler. You can use this to
8742 supply system-specific assembler options which GCC does not know how to
8745 If you want to pass an option that takes an argument, you must use
8746 @option{-Xassembler} twice, once for the option and once for the argument.
8751 @section Options for Linking
8752 @cindex link options
8753 @cindex options, linking
8755 These options come into play when the compiler links object files into
8756 an executable output file. They are meaningless if the compiler is
8757 not doing a link step.
8761 @item @var{object-file-name}
8762 A file name that does not end in a special recognized suffix is
8763 considered to name an object file or library. (Object files are
8764 distinguished from libraries by the linker according to the file
8765 contents.) If linking is done, these object files are used as input
8774 If any of these options is used, then the linker is not run, and
8775 object file names should not be used as arguments. @xref{Overall
8779 @item -l@var{library}
8780 @itemx -l @var{library}
8782 Search the library named @var{library} when linking. (The second
8783 alternative with the library as a separate argument is only for
8784 POSIX compliance and is not recommended.)
8786 It makes a difference where in the command you write this option; the
8787 linker searches and processes libraries and object files in the order they
8788 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
8789 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
8790 to functions in @samp{z}, those functions may not be loaded.
8792 The linker searches a standard list of directories for the library,
8793 which is actually a file named @file{lib@var{library}.a}. The linker
8794 then uses this file as if it had been specified precisely by name.
8796 The directories searched include several standard system directories
8797 plus any that you specify with @option{-L}.
8799 Normally the files found this way are library files---archive files
8800 whose members are object files. The linker handles an archive file by
8801 scanning through it for members which define symbols that have so far
8802 been referenced but not defined. But if the file that is found is an
8803 ordinary object file, it is linked in the usual fashion. The only
8804 difference between using an @option{-l} option and specifying a file name
8805 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
8806 and searches several directories.
8810 You need this special case of the @option{-l} option in order to
8811 link an Objective-C or Objective-C++ program.
8814 @opindex nostartfiles
8815 Do not use the standard system startup files when linking.
8816 The standard system libraries are used normally, unless @option{-nostdlib}
8817 or @option{-nodefaultlibs} is used.
8819 @item -nodefaultlibs
8820 @opindex nodefaultlibs
8821 Do not use the standard system libraries when linking.
8822 Only the libraries you specify will be passed to the linker, options
8823 specifying linkage of the system libraries, such as @code{-static-libgcc}
8824 or @code{-shared-libgcc}, will be ignored.
8825 The standard startup files are used normally, unless @option{-nostartfiles}
8826 is used. The compiler may generate calls to @code{memcmp},
8827 @code{memset}, @code{memcpy} and @code{memmove}.
8828 These entries are usually resolved by entries in
8829 libc. These entry points should be supplied through some other
8830 mechanism when this option is specified.
8834 Do not use the standard system startup files or libraries when linking.
8835 No startup files and only the libraries you specify will be passed to
8836 the linker, options specifying linkage of the system libraries, such as
8837 @code{-static-libgcc} or @code{-shared-libgcc}, will be ignored.
8838 The compiler may generate calls to @code{memcmp}, @code{memset},
8839 @code{memcpy} and @code{memmove}.
8840 These entries are usually resolved by entries in
8841 libc. These entry points should be supplied through some other
8842 mechanism when this option is specified.
8844 @cindex @option{-lgcc}, use with @option{-nostdlib}
8845 @cindex @option{-nostdlib} and unresolved references
8846 @cindex unresolved references and @option{-nostdlib}
8847 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
8848 @cindex @option{-nodefaultlibs} and unresolved references
8849 @cindex unresolved references and @option{-nodefaultlibs}
8850 One of the standard libraries bypassed by @option{-nostdlib} and
8851 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
8852 that GCC uses to overcome shortcomings of particular machines, or special
8853 needs for some languages.
8854 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
8855 Collection (GCC) Internals},
8856 for more discussion of @file{libgcc.a}.)
8857 In most cases, you need @file{libgcc.a} even when you want to avoid
8858 other standard libraries. In other words, when you specify @option{-nostdlib}
8859 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
8860 This ensures that you have no unresolved references to internal GCC
8861 library subroutines. (For example, @samp{__main}, used to ensure C++
8862 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
8863 GNU Compiler Collection (GCC) Internals}.)
8867 Produce a position independent executable on targets which support it.
8868 For predictable results, you must also specify the same set of options
8869 that were used to generate code (@option{-fpie}, @option{-fPIE},
8870 or model suboptions) when you specify this option.
8874 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
8875 that support it. This instructs the linker to add all symbols, not
8876 only used ones, to the dynamic symbol table. This option is needed
8877 for some uses of @code{dlopen} or to allow obtaining backtraces
8878 from within a program.
8882 Remove all symbol table and relocation information from the executable.
8886 On systems that support dynamic linking, this prevents linking with the shared
8887 libraries. On other systems, this option has no effect.
8891 Produce a shared object which can then be linked with other objects to
8892 form an executable. Not all systems support this option. For predictable
8893 results, you must also specify the same set of options that were used to
8894 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
8895 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
8896 needs to build supplementary stub code for constructors to work. On
8897 multi-libbed systems, @samp{gcc -shared} must select the correct support
8898 libraries to link against. Failing to supply the correct flags may lead
8899 to subtle defects. Supplying them in cases where they are not necessary
8902 @item -shared-libgcc
8903 @itemx -static-libgcc
8904 @opindex shared-libgcc
8905 @opindex static-libgcc
8906 On systems that provide @file{libgcc} as a shared library, these options
8907 force the use of either the shared or static version respectively.
8908 If no shared version of @file{libgcc} was built when the compiler was
8909 configured, these options have no effect.
8911 There are several situations in which an application should use the
8912 shared @file{libgcc} instead of the static version. The most common
8913 of these is when the application wishes to throw and catch exceptions
8914 across different shared libraries. In that case, each of the libraries
8915 as well as the application itself should use the shared @file{libgcc}.
8917 Therefore, the G++ and GCJ drivers automatically add
8918 @option{-shared-libgcc} whenever you build a shared library or a main
8919 executable, because C++ and Java programs typically use exceptions, so
8920 this is the right thing to do.
8922 If, instead, you use the GCC driver to create shared libraries, you may
8923 find that they will not always be linked with the shared @file{libgcc}.
8924 If GCC finds, at its configuration time, that you have a non-GNU linker
8925 or a GNU linker that does not support option @option{--eh-frame-hdr},
8926 it will link the shared version of @file{libgcc} into shared libraries
8927 by default. Otherwise, it will take advantage of the linker and optimize
8928 away the linking with the shared version of @file{libgcc}, linking with
8929 the static version of libgcc by default. This allows exceptions to
8930 propagate through such shared libraries, without incurring relocation
8931 costs at library load time.
8933 However, if a library or main executable is supposed to throw or catch
8934 exceptions, you must link it using the G++ or GCJ driver, as appropriate
8935 for the languages used in the program, or using the option
8936 @option{-shared-libgcc}, such that it is linked with the shared
8939 @item -static-libstdc++
8940 When the @command{g++} program is used to link a C++ program, it will
8941 normally automatically link against @option{libstdc++}. If
8942 @file{libstdc++} is available as a shared library, and the
8943 @option{-static} option is not used, then this will link against the
8944 shared version of @file{libstdc++}. That is normally fine. However, it
8945 is sometimes useful to freeze the version of @file{libstdc++} used by
8946 the program without going all the way to a fully static link. The
8947 @option{-static-libstdc++} option directs the @command{g++} driver to
8948 link @file{libstdc++} statically, without necessarily linking other
8949 libraries statically.
8953 Bind references to global symbols when building a shared object. Warn
8954 about any unresolved references (unless overridden by the link editor
8955 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
8958 @item -T @var{script}
8960 @cindex linker script
8961 Use @var{script} as the linker script. This option is supported by most
8962 systems using the GNU linker. On some targets, such as bare-board
8963 targets without an operating system, the @option{-T} option may be required
8964 when linking to avoid references to undefined symbols.
8966 @item -Xlinker @var{option}
8968 Pass @var{option} as an option to the linker. You can use this to
8969 supply system-specific linker options which GCC does not know how to
8972 If you want to pass an option that takes a separate argument, you must use
8973 @option{-Xlinker} twice, once for the option and once for the argument.
8974 For example, to pass @option{-assert definitions}, you must write
8975 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
8976 @option{-Xlinker "-assert definitions"}, because this passes the entire
8977 string as a single argument, which is not what the linker expects.
8979 When using the GNU linker, it is usually more convenient to pass
8980 arguments to linker options using the @option{@var{option}=@var{value}}
8981 syntax than as separate arguments. For example, you can specify
8982 @samp{-Xlinker -Map=output.map} rather than
8983 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
8984 this syntax for command-line options.
8986 @item -Wl,@var{option}
8988 Pass @var{option} as an option to the linker. If @var{option} contains
8989 commas, it is split into multiple options at the commas. You can use this
8990 syntax to pass an argument to the option.
8991 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
8992 linker. When using the GNU linker, you can also get the same effect with
8993 @samp{-Wl,-Map=output.map}.
8995 @item -u @var{symbol}
8997 Pretend the symbol @var{symbol} is undefined, to force linking of
8998 library modules to define it. You can use @option{-u} multiple times with
8999 different symbols to force loading of additional library modules.
9002 @node Directory Options
9003 @section Options for Directory Search
9004 @cindex directory options
9005 @cindex options, directory search
9008 These options specify directories to search for header files, for
9009 libraries and for parts of the compiler:
9014 Add the directory @var{dir} to the head of the list of directories to be
9015 searched for header files. This can be used to override a system header
9016 file, substituting your own version, since these directories are
9017 searched before the system header file directories. However, you should
9018 not use this option to add directories that contain vendor-supplied
9019 system header files (use @option{-isystem} for that). If you use more than
9020 one @option{-I} option, the directories are scanned in left-to-right
9021 order; the standard system directories come after.
9023 If a standard system include directory, or a directory specified with
9024 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
9025 option will be ignored. The directory will still be searched but as a
9026 system directory at its normal position in the system include chain.
9027 This is to ensure that GCC's procedure to fix buggy system headers and
9028 the ordering for the include_next directive are not inadvertently changed.
9029 If you really need to change the search order for system directories,
9030 use the @option{-nostdinc} and/or @option{-isystem} options.
9032 @item -iplugindir=@var{dir}
9033 Set the directory to search for plugins which are passed
9034 by @option{-fplugin=@var{name}} instead of
9035 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
9036 to be used by the user, but only passed by the driver.
9038 @item -iquote@var{dir}
9040 Add the directory @var{dir} to the head of the list of directories to
9041 be searched for header files only for the case of @samp{#include
9042 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
9043 otherwise just like @option{-I}.
9047 Add directory @var{dir} to the list of directories to be searched
9050 @item -B@var{prefix}
9052 This option specifies where to find the executables, libraries,
9053 include files, and data files of the compiler itself.
9055 The compiler driver program runs one or more of the subprograms
9056 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
9057 @var{prefix} as a prefix for each program it tries to run, both with and
9058 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
9060 For each subprogram to be run, the compiler driver first tries the
9061 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
9062 was not specified, the driver tries two standard prefixes, which are
9063 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
9064 those results in a file name that is found, the unmodified program
9065 name is searched for using the directories specified in your
9066 @env{PATH} environment variable.
9068 The compiler will check to see if the path provided by the @option{-B}
9069 refers to a directory, and if necessary it will add a directory
9070 separator character at the end of the path.
9072 @option{-B} prefixes that effectively specify directory names also apply
9073 to libraries in the linker, because the compiler translates these
9074 options into @option{-L} options for the linker. They also apply to
9075 includes files in the preprocessor, because the compiler translates these
9076 options into @option{-isystem} options for the preprocessor. In this case,
9077 the compiler appends @samp{include} to the prefix.
9079 The run-time support file @file{libgcc.a} can also be searched for using
9080 the @option{-B} prefix, if needed. If it is not found there, the two
9081 standard prefixes above are tried, and that is all. The file is left
9082 out of the link if it is not found by those means.
9084 Another way to specify a prefix much like the @option{-B} prefix is to use
9085 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
9088 As a special kludge, if the path provided by @option{-B} is
9089 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
9090 9, then it will be replaced by @file{[dir/]include}. This is to help
9091 with boot-strapping the compiler.
9093 @item -specs=@var{file}
9095 Process @var{file} after the compiler reads in the standard @file{specs}
9096 file, in order to override the defaults that the @file{gcc} driver
9097 program uses when determining what switches to pass to @file{cc1},
9098 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
9099 @option{-specs=@var{file}} can be specified on the command line, and they
9100 are processed in order, from left to right.
9102 @item --sysroot=@var{dir}
9104 Use @var{dir} as the logical root directory for headers and libraries.
9105 For example, if the compiler would normally search for headers in
9106 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
9107 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
9109 If you use both this option and the @option{-isysroot} option, then
9110 the @option{--sysroot} option will apply to libraries, but the
9111 @option{-isysroot} option will apply to header files.
9113 The GNU linker (beginning with version 2.16) has the necessary support
9114 for this option. If your linker does not support this option, the
9115 header file aspect of @option{--sysroot} will still work, but the
9116 library aspect will not.
9120 This option has been deprecated. Please use @option{-iquote} instead for
9121 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
9122 Any directories you specify with @option{-I} options before the @option{-I-}
9123 option are searched only for the case of @samp{#include "@var{file}"};
9124 they are not searched for @samp{#include <@var{file}>}.
9126 If additional directories are specified with @option{-I} options after
9127 the @option{-I-}, these directories are searched for all @samp{#include}
9128 directives. (Ordinarily @emph{all} @option{-I} directories are used
9131 In addition, the @option{-I-} option inhibits the use of the current
9132 directory (where the current input file came from) as the first search
9133 directory for @samp{#include "@var{file}"}. There is no way to
9134 override this effect of @option{-I-}. With @option{-I.} you can specify
9135 searching the directory which was current when the compiler was
9136 invoked. That is not exactly the same as what the preprocessor does
9137 by default, but it is often satisfactory.
9139 @option{-I-} does not inhibit the use of the standard system directories
9140 for header files. Thus, @option{-I-} and @option{-nostdinc} are
9147 @section Specifying subprocesses and the switches to pass to them
9150 @command{gcc} is a driver program. It performs its job by invoking a
9151 sequence of other programs to do the work of compiling, assembling and
9152 linking. GCC interprets its command-line parameters and uses these to
9153 deduce which programs it should invoke, and which command-line options
9154 it ought to place on their command lines. This behavior is controlled
9155 by @dfn{spec strings}. In most cases there is one spec string for each
9156 program that GCC can invoke, but a few programs have multiple spec
9157 strings to control their behavior. The spec strings built into GCC can
9158 be overridden by using the @option{-specs=} command-line switch to specify
9161 @dfn{Spec files} are plaintext files that are used to construct spec
9162 strings. They consist of a sequence of directives separated by blank
9163 lines. The type of directive is determined by the first non-whitespace
9164 character on the line and it can be one of the following:
9167 @item %@var{command}
9168 Issues a @var{command} to the spec file processor. The commands that can
9172 @item %include <@var{file}>
9174 Search for @var{file} and insert its text at the current point in the
9177 @item %include_noerr <@var{file}>
9178 @cindex %include_noerr
9179 Just like @samp{%include}, but do not generate an error message if the include
9180 file cannot be found.
9182 @item %rename @var{old_name} @var{new_name}
9184 Rename the spec string @var{old_name} to @var{new_name}.
9188 @item *[@var{spec_name}]:
9189 This tells the compiler to create, override or delete the named spec
9190 string. All lines after this directive up to the next directive or
9191 blank line are considered to be the text for the spec string. If this
9192 results in an empty string then the spec will be deleted. (Or, if the
9193 spec did not exist, then nothing will happened.) Otherwise, if the spec
9194 does not currently exist a new spec will be created. If the spec does
9195 exist then its contents will be overridden by the text of this
9196 directive, unless the first character of that text is the @samp{+}
9197 character, in which case the text will be appended to the spec.
9199 @item [@var{suffix}]:
9200 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
9201 and up to the next directive or blank line are considered to make up the
9202 spec string for the indicated suffix. When the compiler encounters an
9203 input file with the named suffix, it will processes the spec string in
9204 order to work out how to compile that file. For example:
9211 This says that any input file whose name ends in @samp{.ZZ} should be
9212 passed to the program @samp{z-compile}, which should be invoked with the
9213 command-line switch @option{-input} and with the result of performing the
9214 @samp{%i} substitution. (See below.)
9216 As an alternative to providing a spec string, the text that follows a
9217 suffix directive can be one of the following:
9220 @item @@@var{language}
9221 This says that the suffix is an alias for a known @var{language}. This is
9222 similar to using the @option{-x} command-line switch to GCC to specify a
9223 language explicitly. For example:
9230 Says that .ZZ files are, in fact, C++ source files.
9233 This causes an error messages saying:
9236 @var{name} compiler not installed on this system.
9240 GCC already has an extensive list of suffixes built into it.
9241 This directive will add an entry to the end of the list of suffixes, but
9242 since the list is searched from the end backwards, it is effectively
9243 possible to override earlier entries using this technique.
9247 GCC has the following spec strings built into it. Spec files can
9248 override these strings or create their own. Note that individual
9249 targets can also add their own spec strings to this list.
9252 asm Options to pass to the assembler
9253 asm_final Options to pass to the assembler post-processor
9254 cpp Options to pass to the C preprocessor
9255 cc1 Options to pass to the C compiler
9256 cc1plus Options to pass to the C++ compiler
9257 endfile Object files to include at the end of the link
9258 link Options to pass to the linker
9259 lib Libraries to include on the command line to the linker
9260 libgcc Decides which GCC support library to pass to the linker
9261 linker Sets the name of the linker
9262 predefines Defines to be passed to the C preprocessor
9263 signed_char Defines to pass to CPP to say whether @code{char} is signed
9265 startfile Object files to include at the start of the link
9268 Here is a small example of a spec file:
9274 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
9277 This example renames the spec called @samp{lib} to @samp{old_lib} and
9278 then overrides the previous definition of @samp{lib} with a new one.
9279 The new definition adds in some extra command-line options before
9280 including the text of the old definition.
9282 @dfn{Spec strings} are a list of command-line options to be passed to their
9283 corresponding program. In addition, the spec strings can contain
9284 @samp{%}-prefixed sequences to substitute variable text or to
9285 conditionally insert text into the command line. Using these constructs
9286 it is possible to generate quite complex command lines.
9288 Here is a table of all defined @samp{%}-sequences for spec
9289 strings. Note that spaces are not generated automatically around the
9290 results of expanding these sequences. Therefore you can concatenate them
9291 together or combine them with constant text in a single argument.
9295 Substitute one @samp{%} into the program name or argument.
9298 Substitute the name of the input file being processed.
9301 Substitute the basename of the input file being processed.
9302 This is the substring up to (and not including) the last period
9303 and not including the directory.
9306 This is the same as @samp{%b}, but include the file suffix (text after
9310 Marks the argument containing or following the @samp{%d} as a
9311 temporary file name, so that that file will be deleted if GCC exits
9312 successfully. Unlike @samp{%g}, this contributes no text to the
9315 @item %g@var{suffix}
9316 Substitute a file name that has suffix @var{suffix} and is chosen
9317 once per compilation, and mark the argument in the same way as
9318 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
9319 name is now chosen in a way that is hard to predict even when previously
9320 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
9321 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
9322 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
9323 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
9324 was simply substituted with a file name chosen once per compilation,
9325 without regard to any appended suffix (which was therefore treated
9326 just like ordinary text), making such attacks more likely to succeed.
9328 @item %u@var{suffix}
9329 Like @samp{%g}, but generates a new temporary file name even if
9330 @samp{%u@var{suffix}} was already seen.
9332 @item %U@var{suffix}
9333 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
9334 new one if there is no such last file name. In the absence of any
9335 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
9336 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
9337 would involve the generation of two distinct file names, one
9338 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
9339 simply substituted with a file name chosen for the previous @samp{%u},
9340 without regard to any appended suffix.
9342 @item %j@var{suffix}
9343 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
9344 writable, and if save-temps is off; otherwise, substitute the name
9345 of a temporary file, just like @samp{%u}. This temporary file is not
9346 meant for communication between processes, but rather as a junk
9349 @item %|@var{suffix}
9350 @itemx %m@var{suffix}
9351 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
9352 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
9353 all. These are the two most common ways to instruct a program that it
9354 should read from standard input or write to standard output. If you
9355 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
9356 construct: see for example @file{f/lang-specs.h}.
9358 @item %.@var{SUFFIX}
9359 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
9360 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
9361 terminated by the next space or %.
9364 Marks the argument containing or following the @samp{%w} as the
9365 designated output file of this compilation. This puts the argument
9366 into the sequence of arguments that @samp{%o} will substitute later.
9369 Substitutes the names of all the output files, with spaces
9370 automatically placed around them. You should write spaces
9371 around the @samp{%o} as well or the results are undefined.
9372 @samp{%o} is for use in the specs for running the linker.
9373 Input files whose names have no recognized suffix are not compiled
9374 at all, but they are included among the output files, so they will
9378 Substitutes the suffix for object files. Note that this is
9379 handled specially when it immediately follows @samp{%g, %u, or %U},
9380 because of the need for those to form complete file names. The
9381 handling is such that @samp{%O} is treated exactly as if it had already
9382 been substituted, except that @samp{%g, %u, and %U} do not currently
9383 support additional @var{suffix} characters following @samp{%O} as they would
9384 following, for example, @samp{.o}.
9387 Substitutes the standard macro predefinitions for the
9388 current target machine. Use this when running @code{cpp}.
9391 Like @samp{%p}, but puts @samp{__} before and after the name of each
9392 predefined macro, except for macros that start with @samp{__} or with
9393 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
9397 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
9398 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
9399 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
9400 and @option{-imultilib} as necessary.
9403 Current argument is the name of a library or startup file of some sort.
9404 Search for that file in a standard list of directories and substitute
9405 the full name found. The current working directory is included in the
9406 list of directories scanned.
9409 Current argument is the name of a linker script. Search for that file
9410 in the current list of directories to scan for libraries. If the file
9411 is located insert a @option{--script} option into the command line
9412 followed by the full path name found. If the file is not found then
9413 generate an error message. Note: the current working directory is not
9417 Print @var{str} as an error message. @var{str} is terminated by a newline.
9418 Use this when inconsistent options are detected.
9421 Substitute the contents of spec string @var{name} at this point.
9424 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
9426 @item %x@{@var{option}@}
9427 Accumulate an option for @samp{%X}.
9430 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
9434 Output the accumulated assembler options specified by @option{-Wa}.
9437 Output the accumulated preprocessor options specified by @option{-Wp}.
9440 Process the @code{asm} spec. This is used to compute the
9441 switches to be passed to the assembler.
9444 Process the @code{asm_final} spec. This is a spec string for
9445 passing switches to an assembler post-processor, if such a program is
9449 Process the @code{link} spec. This is the spec for computing the
9450 command line passed to the linker. Typically it will make use of the
9451 @samp{%L %G %S %D and %E} sequences.
9454 Dump out a @option{-L} option for each directory that GCC believes might
9455 contain startup files. If the target supports multilibs then the
9456 current multilib directory will be prepended to each of these paths.
9459 Process the @code{lib} spec. This is a spec string for deciding which
9460 libraries should be included on the command line to the linker.
9463 Process the @code{libgcc} spec. This is a spec string for deciding
9464 which GCC support library should be included on the command line to the linker.
9467 Process the @code{startfile} spec. This is a spec for deciding which
9468 object files should be the first ones passed to the linker. Typically
9469 this might be a file named @file{crt0.o}.
9472 Process the @code{endfile} spec. This is a spec string that specifies
9473 the last object files that will be passed to the linker.
9476 Process the @code{cpp} spec. This is used to construct the arguments
9477 to be passed to the C preprocessor.
9480 Process the @code{cc1} spec. This is used to construct the options to be
9481 passed to the actual C compiler (@samp{cc1}).
9484 Process the @code{cc1plus} spec. This is used to construct the options to be
9485 passed to the actual C++ compiler (@samp{cc1plus}).
9488 Substitute the variable part of a matched option. See below.
9489 Note that each comma in the substituted string is replaced by
9493 Remove all occurrences of @code{-S} from the command line. Note---this
9494 command is position dependent. @samp{%} commands in the spec string
9495 before this one will see @code{-S}, @samp{%} commands in the spec string
9496 after this one will not.
9498 @item %:@var{function}(@var{args})
9499 Call the named function @var{function}, passing it @var{args}.
9500 @var{args} is first processed as a nested spec string, then split
9501 into an argument vector in the usual fashion. The function returns
9502 a string which is processed as if it had appeared literally as part
9503 of the current spec.
9505 The following built-in spec functions are provided:
9509 The @code{getenv} spec function takes two arguments: an environment
9510 variable name and a string. If the environment variable is not
9511 defined, a fatal error is issued. Otherwise, the return value is the
9512 value of the environment variable concatenated with the string. For
9513 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
9516 %:getenv(TOPDIR /include)
9519 expands to @file{/path/to/top/include}.
9521 @item @code{if-exists}
9522 The @code{if-exists} spec function takes one argument, an absolute
9523 pathname to a file. If the file exists, @code{if-exists} returns the
9524 pathname. Here is a small example of its usage:
9528 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
9531 @item @code{if-exists-else}
9532 The @code{if-exists-else} spec function is similar to the @code{if-exists}
9533 spec function, except that it takes two arguments. The first argument is
9534 an absolute pathname to a file. If the file exists, @code{if-exists-else}
9535 returns the pathname. If it does not exist, it returns the second argument.
9536 This way, @code{if-exists-else} can be used to select one file or another,
9537 based on the existence of the first. Here is a small example of its usage:
9541 crt0%O%s %:if-exists(crti%O%s) \
9542 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
9545 @item @code{replace-outfile}
9546 The @code{replace-outfile} spec function takes two arguments. It looks for the
9547 first argument in the outfiles array and replaces it with the second argument. Here
9548 is a small example of its usage:
9551 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
9554 @item @code{print-asm-header}
9555 The @code{print-asm-header} function takes no arguments and simply
9556 prints a banner like:
9562 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
9565 It is used to separate compiler options from assembler options
9566 in the @option{--target-help} output.
9570 Substitutes the @code{-S} switch, if that switch was given to GCC@.
9571 If that switch was not specified, this substitutes nothing. Note that
9572 the leading dash is omitted when specifying this option, and it is
9573 automatically inserted if the substitution is performed. Thus the spec
9574 string @samp{%@{foo@}} would match the command-line option @option{-foo}
9575 and would output the command line option @option{-foo}.
9577 @item %W@{@code{S}@}
9578 Like %@{@code{S}@} but mark last argument supplied within as a file to be
9581 @item %@{@code{S}*@}
9582 Substitutes all the switches specified to GCC whose names start
9583 with @code{-S}, but which also take an argument. This is used for
9584 switches like @option{-o}, @option{-D}, @option{-I}, etc.
9585 GCC considers @option{-o foo} as being
9586 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
9587 text, including the space. Thus two arguments would be generated.
9589 @item %@{@code{S}*&@code{T}*@}
9590 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
9591 (the order of @code{S} and @code{T} in the spec is not significant).
9592 There can be any number of ampersand-separated variables; for each the
9593 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
9595 @item %@{@code{S}:@code{X}@}
9596 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
9598 @item %@{!@code{S}:@code{X}@}
9599 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
9601 @item %@{@code{S}*:@code{X}@}
9602 Substitutes @code{X} if one or more switches whose names start with
9603 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
9604 once, no matter how many such switches appeared. However, if @code{%*}
9605 appears somewhere in @code{X}, then @code{X} will be substituted once
9606 for each matching switch, with the @code{%*} replaced by the part of
9607 that switch that matched the @code{*}.
9609 @item %@{.@code{S}:@code{X}@}
9610 Substitutes @code{X}, if processing a file with suffix @code{S}.
9612 @item %@{!.@code{S}:@code{X}@}
9613 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
9615 @item %@{,@code{S}:@code{X}@}
9616 Substitutes @code{X}, if processing a file for language @code{S}.
9618 @item %@{!,@code{S}:@code{X}@}
9619 Substitutes @code{X}, if not processing a file for language @code{S}.
9621 @item %@{@code{S}|@code{P}:@code{X}@}
9622 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
9623 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
9624 @code{*} sequences as well, although they have a stronger binding than
9625 the @samp{|}. If @code{%*} appears in @code{X}, all of the
9626 alternatives must be starred, and only the first matching alternative
9629 For example, a spec string like this:
9632 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
9635 will output the following command-line options from the following input
9636 command-line options:
9641 -d fred.c -foo -baz -boggle
9642 -d jim.d -bar -baz -boggle
9645 @item %@{S:X; T:Y; :D@}
9647 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
9648 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
9649 be as many clauses as you need. This may be combined with @code{.},
9650 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
9655 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
9656 construct may contain other nested @samp{%} constructs or spaces, or
9657 even newlines. They are processed as usual, as described above.
9658 Trailing white space in @code{X} is ignored. White space may also
9659 appear anywhere on the left side of the colon in these constructs,
9660 except between @code{.} or @code{*} and the corresponding word.
9662 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
9663 handled specifically in these constructs. If another value of
9664 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
9665 @option{-W} switch is found later in the command line, the earlier
9666 switch value is ignored, except with @{@code{S}*@} where @code{S} is
9667 just one letter, which passes all matching options.
9669 The character @samp{|} at the beginning of the predicate text is used to
9670 indicate that a command should be piped to the following command, but
9671 only if @option{-pipe} is specified.
9673 It is built into GCC which switches take arguments and which do not.
9674 (You might think it would be useful to generalize this to allow each
9675 compiler's spec to say which switches take arguments. But this cannot
9676 be done in a consistent fashion. GCC cannot even decide which input
9677 files have been specified without knowing which switches take arguments,
9678 and it must know which input files to compile in order to tell which
9681 GCC also knows implicitly that arguments starting in @option{-l} are to be
9682 treated as compiler output files, and passed to the linker in their
9683 proper position among the other output files.
9685 @c man begin OPTIONS
9687 @node Target Options
9688 @section Specifying Target Machine and Compiler Version
9689 @cindex target options
9690 @cindex cross compiling
9691 @cindex specifying machine version
9692 @cindex specifying compiler version and target machine
9693 @cindex compiler version, specifying
9694 @cindex target machine, specifying
9696 The usual way to run GCC is to run the executable called @file{gcc}, or
9697 @file{<machine>-gcc} when cross-compiling, or
9698 @file{<machine>-gcc-<version>} to run a version other than the one that
9701 @node Submodel Options
9702 @section Hardware Models and Configurations
9703 @cindex submodel options
9704 @cindex specifying hardware config
9705 @cindex hardware models and configurations, specifying
9706 @cindex machine dependent options
9708 Each target machine types can have its own
9709 special options, starting with @samp{-m}, to choose among various
9710 hardware models or configurations---for example, 68010 vs 68020,
9711 floating coprocessor or none. A single installed version of the
9712 compiler can compile for any model or configuration, according to the
9715 Some configurations of the compiler also support additional special
9716 options, usually for compatibility with other compilers on the same
9719 @c This list is ordered alphanumerically by subsection name.
9720 @c It should be the same order and spelling as these options are listed
9721 @c in Machine Dependent Options
9727 * Blackfin Options::
9731 * DEC Alpha Options::
9732 * DEC Alpha/VMS Options::
9735 * GNU/Linux Options::
9738 * i386 and x86-64 Options::
9739 * i386 and x86-64 Windows Options::
9741 * IA-64/VMS Options::
9753 * picoChip Options::
9755 * RS/6000 and PowerPC Options::
9757 * S/390 and zSeries Options::
9762 * System V Options::
9767 * Xstormy16 Options::
9773 @subsection ARC Options
9776 These options are defined for ARC implementations:
9781 Compile code for little endian mode. This is the default.
9785 Compile code for big endian mode.
9788 @opindex mmangle-cpu
9789 Prepend the name of the cpu to all public symbol names.
9790 In multiple-processor systems, there are many ARC variants with different
9791 instruction and register set characteristics. This flag prevents code
9792 compiled for one cpu to be linked with code compiled for another.
9793 No facility exists for handling variants that are ``almost identical''.
9794 This is an all or nothing option.
9796 @item -mcpu=@var{cpu}
9798 Compile code for ARC variant @var{cpu}.
9799 Which variants are supported depend on the configuration.
9800 All variants support @option{-mcpu=base}, this is the default.
9802 @item -mtext=@var{text-section}
9803 @itemx -mdata=@var{data-section}
9804 @itemx -mrodata=@var{readonly-data-section}
9808 Put functions, data, and readonly data in @var{text-section},
9809 @var{data-section}, and @var{readonly-data-section} respectively
9810 by default. This can be overridden with the @code{section} attribute.
9811 @xref{Variable Attributes}.
9816 @subsection ARM Options
9819 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
9823 @item -mabi=@var{name}
9825 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
9826 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
9829 @opindex mapcs-frame
9830 Generate a stack frame that is compliant with the ARM Procedure Call
9831 Standard for all functions, even if this is not strictly necessary for
9832 correct execution of the code. Specifying @option{-fomit-frame-pointer}
9833 with this option will cause the stack frames not to be generated for
9834 leaf functions. The default is @option{-mno-apcs-frame}.
9838 This is a synonym for @option{-mapcs-frame}.
9841 @c not currently implemented
9842 @item -mapcs-stack-check
9843 @opindex mapcs-stack-check
9844 Generate code to check the amount of stack space available upon entry to
9845 every function (that actually uses some stack space). If there is
9846 insufficient space available then either the function
9847 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
9848 called, depending upon the amount of stack space required. The run time
9849 system is required to provide these functions. The default is
9850 @option{-mno-apcs-stack-check}, since this produces smaller code.
9852 @c not currently implemented
9854 @opindex mapcs-float
9855 Pass floating point arguments using the float point registers. This is
9856 one of the variants of the APCS@. This option is recommended if the
9857 target hardware has a floating point unit or if a lot of floating point
9858 arithmetic is going to be performed by the code. The default is
9859 @option{-mno-apcs-float}, since integer only code is slightly increased in
9860 size if @option{-mapcs-float} is used.
9862 @c not currently implemented
9863 @item -mapcs-reentrant
9864 @opindex mapcs-reentrant
9865 Generate reentrant, position independent code. The default is
9866 @option{-mno-apcs-reentrant}.
9869 @item -mthumb-interwork
9870 @opindex mthumb-interwork
9871 Generate code which supports calling between the ARM and Thumb
9872 instruction sets. Without this option the two instruction sets cannot
9873 be reliably used inside one program. The default is
9874 @option{-mno-thumb-interwork}, since slightly larger code is generated
9875 when @option{-mthumb-interwork} is specified.
9877 @item -mno-sched-prolog
9878 @opindex mno-sched-prolog
9879 Prevent the reordering of instructions in the function prolog, or the
9880 merging of those instruction with the instructions in the function's
9881 body. This means that all functions will start with a recognizable set
9882 of instructions (or in fact one of a choice from a small set of
9883 different function prologues), and this information can be used to
9884 locate the start if functions inside an executable piece of code. The
9885 default is @option{-msched-prolog}.
9887 @item -mfloat-abi=@var{name}
9889 Specifies which floating-point ABI to use. Permissible values
9890 are: @samp{soft}, @samp{softfp} and @samp{hard}.
9892 Specifying @samp{soft} causes GCC to generate output containing
9893 library calls for floating-point operations.
9894 @samp{softfp} allows the generation of code using hardware floating-point
9895 instructions, but still uses the soft-float calling conventions.
9896 @samp{hard} allows generation of floating-point instructions
9897 and uses FPU-specific calling conventions.
9899 The default depends on the specific target configuration. Note that
9900 the hard-float and soft-float ABIs are not link-compatible; you must
9901 compile your entire program with the same ABI, and link with a
9902 compatible set of libraries.
9905 @opindex mhard-float
9906 Equivalent to @option{-mfloat-abi=hard}.
9909 @opindex msoft-float
9910 Equivalent to @option{-mfloat-abi=soft}.
9912 @item -mlittle-endian
9913 @opindex mlittle-endian
9914 Generate code for a processor running in little-endian mode. This is
9915 the default for all standard configurations.
9918 @opindex mbig-endian
9919 Generate code for a processor running in big-endian mode; the default is
9920 to compile code for a little-endian processor.
9922 @item -mwords-little-endian
9923 @opindex mwords-little-endian
9924 This option only applies when generating code for big-endian processors.
9925 Generate code for a little-endian word order but a big-endian byte
9926 order. That is, a byte order of the form @samp{32107654}. Note: this
9927 option should only be used if you require compatibility with code for
9928 big-endian ARM processors generated by versions of the compiler prior to
9931 @item -mcpu=@var{name}
9933 This specifies the name of the target ARM processor. GCC uses this name
9934 to determine what kind of instructions it can emit when generating
9935 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
9936 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
9937 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
9938 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
9939 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
9941 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
9942 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
9943 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
9944 @samp{strongarm1110},
9945 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
9946 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
9947 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
9948 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
9949 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
9950 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
9951 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
9952 @samp{cortex-a5}, @samp{cortex-a8}, @samp{cortex-a9},
9953 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-m4}, @samp{cortex-m3},
9956 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9958 @item -mtune=@var{name}
9960 This option is very similar to the @option{-mcpu=} option, except that
9961 instead of specifying the actual target processor type, and hence
9962 restricting which instructions can be used, it specifies that GCC should
9963 tune the performance of the code as if the target were of the type
9964 specified in this option, but still choosing the instructions that it
9965 will generate based on the cpu specified by a @option{-mcpu=} option.
9966 For some ARM implementations better performance can be obtained by using
9969 @item -march=@var{name}
9971 This specifies the name of the target ARM architecture. GCC uses this
9972 name to determine what kind of instructions it can emit when generating
9973 assembly code. This option can be used in conjunction with or instead
9974 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
9975 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
9976 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
9977 @samp{armv6}, @samp{armv6j},
9978 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
9979 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
9980 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9982 @item -mfpu=@var{name}
9983 @itemx -mfpe=@var{number}
9984 @itemx -mfp=@var{number}
9988 This specifies what floating point hardware (or hardware emulation) is
9989 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
9990 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-fp16},
9991 @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd}, @samp{vfpv3xd-fp16},
9992 @samp{neon}, @samp{neon-fp16}, @samp{vfpv4}, @samp{vfpv4-d16},
9993 @samp{fpv4-sp-d16} and @samp{neon-vfpv4}.
9994 @option{-mfp} and @option{-mfpe} are synonyms for
9995 @option{-mfpu}=@samp{fpe}@var{number}, for compatibility with older versions
9998 If @option{-msoft-float} is specified this specifies the format of
9999 floating point values.
10001 If the selected floating-point hardware includes the NEON extension
10002 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
10003 operations will not be used by GCC's auto-vectorization pass unless
10004 @option{-funsafe-math-optimizations} is also specified. This is
10005 because NEON hardware does not fully implement the IEEE 754 standard for
10006 floating-point arithmetic (in particular denormal values are treated as
10007 zero), so the use of NEON instructions may lead to a loss of precision.
10009 @item -mfp16-format=@var{name}
10010 @opindex mfp16-format
10011 Specify the format of the @code{__fp16} half-precision floating-point type.
10012 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
10013 the default is @samp{none}, in which case the @code{__fp16} type is not
10014 defined. @xref{Half-Precision}, for more information.
10016 @item -mstructure-size-boundary=@var{n}
10017 @opindex mstructure-size-boundary
10018 The size of all structures and unions will be rounded up to a multiple
10019 of the number of bits set by this option. Permissible values are 8, 32
10020 and 64. The default value varies for different toolchains. For the COFF
10021 targeted toolchain the default value is 8. A value of 64 is only allowed
10022 if the underlying ABI supports it.
10024 Specifying the larger number can produce faster, more efficient code, but
10025 can also increase the size of the program. Different values are potentially
10026 incompatible. Code compiled with one value cannot necessarily expect to
10027 work with code or libraries compiled with another value, if they exchange
10028 information using structures or unions.
10030 @item -mabort-on-noreturn
10031 @opindex mabort-on-noreturn
10032 Generate a call to the function @code{abort} at the end of a
10033 @code{noreturn} function. It will be executed if the function tries to
10037 @itemx -mno-long-calls
10038 @opindex mlong-calls
10039 @opindex mno-long-calls
10040 Tells the compiler to perform function calls by first loading the
10041 address of the function into a register and then performing a subroutine
10042 call on this register. This switch is needed if the target function
10043 will lie outside of the 64 megabyte addressing range of the offset based
10044 version of subroutine call instruction.
10046 Even if this switch is enabled, not all function calls will be turned
10047 into long calls. The heuristic is that static functions, functions
10048 which have the @samp{short-call} attribute, functions that are inside
10049 the scope of a @samp{#pragma no_long_calls} directive and functions whose
10050 definitions have already been compiled within the current compilation
10051 unit, will not be turned into long calls. The exception to this rule is
10052 that weak function definitions, functions with the @samp{long-call}
10053 attribute or the @samp{section} attribute, and functions that are within
10054 the scope of a @samp{#pragma long_calls} directive, will always be
10055 turned into long calls.
10057 This feature is not enabled by default. Specifying
10058 @option{-mno-long-calls} will restore the default behavior, as will
10059 placing the function calls within the scope of a @samp{#pragma
10060 long_calls_off} directive. Note these switches have no effect on how
10061 the compiler generates code to handle function calls via function
10064 @item -msingle-pic-base
10065 @opindex msingle-pic-base
10066 Treat the register used for PIC addressing as read-only, rather than
10067 loading it in the prologue for each function. The run-time system is
10068 responsible for initializing this register with an appropriate value
10069 before execution begins.
10071 @item -mpic-register=@var{reg}
10072 @opindex mpic-register
10073 Specify the register to be used for PIC addressing. The default is R10
10074 unless stack-checking is enabled, when R9 is used.
10076 @item -mcirrus-fix-invalid-insns
10077 @opindex mcirrus-fix-invalid-insns
10078 @opindex mno-cirrus-fix-invalid-insns
10079 Insert NOPs into the instruction stream to in order to work around
10080 problems with invalid Maverick instruction combinations. This option
10081 is only valid if the @option{-mcpu=ep9312} option has been used to
10082 enable generation of instructions for the Cirrus Maverick floating
10083 point co-processor. This option is not enabled by default, since the
10084 problem is only present in older Maverick implementations. The default
10085 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
10088 @item -mpoke-function-name
10089 @opindex mpoke-function-name
10090 Write the name of each function into the text section, directly
10091 preceding the function prologue. The generated code is similar to this:
10095 .ascii "arm_poke_function_name", 0
10098 .word 0xff000000 + (t1 - t0)
10099 arm_poke_function_name
10101 stmfd sp!, @{fp, ip, lr, pc@}
10105 When performing a stack backtrace, code can inspect the value of
10106 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
10107 location @code{pc - 12} and the top 8 bits are set, then we know that
10108 there is a function name embedded immediately preceding this location
10109 and has length @code{((pc[-3]) & 0xff000000)}.
10113 Generate code for the Thumb instruction set. The default is to
10114 use the 32-bit ARM instruction set.
10115 This option automatically enables either 16-bit Thumb-1 or
10116 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
10117 and @option{-march=@var{name}} options. This option is not passed to the
10118 assembler. If you want to force assembler files to be interpreted as Thumb code,
10119 either add a @samp{.thumb} directive to the source or pass the @option{-mthumb}
10120 option directly to the assembler by prefixing it with @option{-Wa}.
10123 @opindex mtpcs-frame
10124 Generate a stack frame that is compliant with the Thumb Procedure Call
10125 Standard for all non-leaf functions. (A leaf function is one that does
10126 not call any other functions.) The default is @option{-mno-tpcs-frame}.
10128 @item -mtpcs-leaf-frame
10129 @opindex mtpcs-leaf-frame
10130 Generate a stack frame that is compliant with the Thumb Procedure Call
10131 Standard for all leaf functions. (A leaf function is one that does
10132 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
10134 @item -mcallee-super-interworking
10135 @opindex mcallee-super-interworking
10136 Gives all externally visible functions in the file being compiled an ARM
10137 instruction set header which switches to Thumb mode before executing the
10138 rest of the function. This allows these functions to be called from
10139 non-interworking code. This option is not valid in AAPCS configurations
10140 because interworking is enabled by default.
10142 @item -mcaller-super-interworking
10143 @opindex mcaller-super-interworking
10144 Allows calls via function pointers (including virtual functions) to
10145 execute correctly regardless of whether the target code has been
10146 compiled for interworking or not. There is a small overhead in the cost
10147 of executing a function pointer if this option is enabled. This option
10148 is not valid in AAPCS configurations because interworking is enabled
10151 @item -mtp=@var{name}
10153 Specify the access model for the thread local storage pointer. The valid
10154 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
10155 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
10156 (supported in the arm6k architecture), and @option{auto}, which uses the
10157 best available method for the selected processor. The default setting is
10160 @item -mword-relocations
10161 @opindex mword-relocations
10162 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
10163 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
10164 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
10167 @item -mfix-cortex-m3-ldrd
10168 @opindex mfix-cortex-m3-ldrd
10169 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
10170 with overlapping destination and base registers are used. This option avoids
10171 generating these instructions. This option is enabled by default when
10172 @option{-mcpu=cortex-m3} is specified.
10177 @subsection AVR Options
10178 @cindex AVR Options
10180 These options are defined for AVR implementations:
10183 @item -mmcu=@var{mcu}
10185 Specify ATMEL AVR instruction set or MCU type.
10187 Instruction set avr1 is for the minimal AVR core, not supported by the C
10188 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
10189 attiny11, attiny12, attiny15, attiny28).
10191 Instruction set avr2 (default) is for the classic AVR core with up to
10192 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
10193 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
10194 at90c8534, at90s8535).
10196 Instruction set avr3 is for the classic AVR core with up to 128K program
10197 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
10199 Instruction set avr4 is for the enhanced AVR core with up to 8K program
10200 memory space (MCU types: atmega8, atmega83, atmega85).
10202 Instruction set avr5 is for the enhanced AVR core with up to 128K program
10203 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
10204 atmega64, atmega128, at43usb355, at94k).
10206 @item -mno-interrupts
10207 @opindex mno-interrupts
10208 Generated code is not compatible with hardware interrupts.
10209 Code size will be smaller.
10211 @item -mcall-prologues
10212 @opindex mcall-prologues
10213 Functions prologues/epilogues expanded as call to appropriate
10214 subroutines. Code size will be smaller.
10217 @opindex mtiny-stack
10218 Change only the low 8 bits of the stack pointer.
10222 Assume int to be 8 bit integer. This affects the sizes of all types: A
10223 char will be 1 byte, an int will be 1 byte, a long will be 2 bytes
10224 and long long will be 4 bytes. Please note that this option does not
10225 comply to the C standards, but it will provide you with smaller code
10229 @node Blackfin Options
10230 @subsection Blackfin Options
10231 @cindex Blackfin Options
10234 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
10236 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
10237 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
10238 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
10239 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
10240 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
10241 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
10242 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
10244 The optional @var{sirevision} specifies the silicon revision of the target
10245 Blackfin processor. Any workarounds available for the targeted silicon revision
10246 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
10247 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
10248 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
10249 hexadecimal digits representing the major and minor numbers in the silicon
10250 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
10251 is not defined. If @var{sirevision} is @samp{any}, the
10252 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
10253 If this optional @var{sirevision} is not used, GCC assumes the latest known
10254 silicon revision of the targeted Blackfin processor.
10256 Support for @samp{bf561} is incomplete. For @samp{bf561},
10257 Only the processor macro is defined.
10258 Without this option, @samp{bf532} is used as the processor by default.
10259 The corresponding predefined processor macros for @var{cpu} is to
10260 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
10261 provided by libgloss to be linked in if @option{-msim} is not given.
10265 Specifies that the program will be run on the simulator. This causes
10266 the simulator BSP provided by libgloss to be linked in. This option
10267 has effect only for @samp{bfin-elf} toolchain.
10268 Certain other options, such as @option{-mid-shared-library} and
10269 @option{-mfdpic}, imply @option{-msim}.
10271 @item -momit-leaf-frame-pointer
10272 @opindex momit-leaf-frame-pointer
10273 Don't keep the frame pointer in a register for leaf functions. This
10274 avoids the instructions to save, set up and restore frame pointers and
10275 makes an extra register available in leaf functions. The option
10276 @option{-fomit-frame-pointer} removes the frame pointer for all functions
10277 which might make debugging harder.
10279 @item -mspecld-anomaly
10280 @opindex mspecld-anomaly
10281 When enabled, the compiler will ensure that the generated code does not
10282 contain speculative loads after jump instructions. If this option is used,
10283 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
10285 @item -mno-specld-anomaly
10286 @opindex mno-specld-anomaly
10287 Don't generate extra code to prevent speculative loads from occurring.
10289 @item -mcsync-anomaly
10290 @opindex mcsync-anomaly
10291 When enabled, the compiler will ensure that the generated code does not
10292 contain CSYNC or SSYNC instructions too soon after conditional branches.
10293 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
10295 @item -mno-csync-anomaly
10296 @opindex mno-csync-anomaly
10297 Don't generate extra code to prevent CSYNC or SSYNC instructions from
10298 occurring too soon after a conditional branch.
10302 When enabled, the compiler is free to take advantage of the knowledge that
10303 the entire program fits into the low 64k of memory.
10306 @opindex mno-low-64k
10307 Assume that the program is arbitrarily large. This is the default.
10309 @item -mstack-check-l1
10310 @opindex mstack-check-l1
10311 Do stack checking using information placed into L1 scratchpad memory by the
10314 @item -mid-shared-library
10315 @opindex mid-shared-library
10316 Generate code that supports shared libraries via the library ID method.
10317 This allows for execute in place and shared libraries in an environment
10318 without virtual memory management. This option implies @option{-fPIC}.
10319 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10321 @item -mno-id-shared-library
10322 @opindex mno-id-shared-library
10323 Generate code that doesn't assume ID based shared libraries are being used.
10324 This is the default.
10326 @item -mleaf-id-shared-library
10327 @opindex mleaf-id-shared-library
10328 Generate code that supports shared libraries via the library ID method,
10329 but assumes that this library or executable won't link against any other
10330 ID shared libraries. That allows the compiler to use faster code for jumps
10333 @item -mno-leaf-id-shared-library
10334 @opindex mno-leaf-id-shared-library
10335 Do not assume that the code being compiled won't link against any ID shared
10336 libraries. Slower code will be generated for jump and call insns.
10338 @item -mshared-library-id=n
10339 @opindex mshared-library-id
10340 Specified the identification number of the ID based shared library being
10341 compiled. Specifying a value of 0 will generate more compact code, specifying
10342 other values will force the allocation of that number to the current
10343 library but is no more space or time efficient than omitting this option.
10347 Generate code that allows the data segment to be located in a different
10348 area of memory from the text segment. This allows for execute in place in
10349 an environment without virtual memory management by eliminating relocations
10350 against the text section.
10352 @item -mno-sep-data
10353 @opindex mno-sep-data
10354 Generate code that assumes that the data segment follows the text segment.
10355 This is the default.
10358 @itemx -mno-long-calls
10359 @opindex mlong-calls
10360 @opindex mno-long-calls
10361 Tells the compiler to perform function calls by first loading the
10362 address of the function into a register and then performing a subroutine
10363 call on this register. This switch is needed if the target function
10364 will lie outside of the 24 bit addressing range of the offset based
10365 version of subroutine call instruction.
10367 This feature is not enabled by default. Specifying
10368 @option{-mno-long-calls} will restore the default behavior. Note these
10369 switches have no effect on how the compiler generates code to handle
10370 function calls via function pointers.
10374 Link with the fast floating-point library. This library relaxes some of
10375 the IEEE floating-point standard's rules for checking inputs against
10376 Not-a-Number (NAN), in the interest of performance.
10379 @opindex minline-plt
10380 Enable inlining of PLT entries in function calls to functions that are
10381 not known to bind locally. It has no effect without @option{-mfdpic}.
10384 @opindex mmulticore
10385 Build standalone application for multicore Blackfin processor. Proper
10386 start files and link scripts will be used to support multicore.
10387 This option defines @code{__BFIN_MULTICORE}. It can only be used with
10388 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
10389 @option{-mcorea} or @option{-mcoreb}. If it's used without
10390 @option{-mcorea} or @option{-mcoreb}, single application/dual core
10391 programming model is used. In this model, the main function of Core B
10392 should be named as coreb_main. If it's used with @option{-mcorea} or
10393 @option{-mcoreb}, one application per core programming model is used.
10394 If this option is not used, single core application programming
10399 Build standalone application for Core A of BF561 when using
10400 one application per core programming model. Proper start files
10401 and link scripts will be used to support Core A. This option
10402 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
10406 Build standalone application for Core B of BF561 when using
10407 one application per core programming model. Proper start files
10408 and link scripts will be used to support Core B. This option
10409 defines @code{__BFIN_COREB}. When this option is used, coreb_main
10410 should be used instead of main. It must be used with
10411 @option{-mmulticore}.
10415 Build standalone application for SDRAM. Proper start files and
10416 link scripts will be used to put the application into SDRAM.
10417 Loader should initialize SDRAM before loading the application
10418 into SDRAM. This option defines @code{__BFIN_SDRAM}.
10422 Assume that ICPLBs are enabled at runtime. This has an effect on certain
10423 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
10424 are enabled; for standalone applications the default is off.
10428 @subsection CRIS Options
10429 @cindex CRIS Options
10431 These options are defined specifically for the CRIS ports.
10434 @item -march=@var{architecture-type}
10435 @itemx -mcpu=@var{architecture-type}
10438 Generate code for the specified architecture. The choices for
10439 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
10440 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
10441 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
10444 @item -mtune=@var{architecture-type}
10446 Tune to @var{architecture-type} everything applicable about the generated
10447 code, except for the ABI and the set of available instructions. The
10448 choices for @var{architecture-type} are the same as for
10449 @option{-march=@var{architecture-type}}.
10451 @item -mmax-stack-frame=@var{n}
10452 @opindex mmax-stack-frame
10453 Warn when the stack frame of a function exceeds @var{n} bytes.
10459 The options @option{-metrax4} and @option{-metrax100} are synonyms for
10460 @option{-march=v3} and @option{-march=v8} respectively.
10462 @item -mmul-bug-workaround
10463 @itemx -mno-mul-bug-workaround
10464 @opindex mmul-bug-workaround
10465 @opindex mno-mul-bug-workaround
10466 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
10467 models where it applies. This option is active by default.
10471 Enable CRIS-specific verbose debug-related information in the assembly
10472 code. This option also has the effect to turn off the @samp{#NO_APP}
10473 formatted-code indicator to the assembler at the beginning of the
10478 Do not use condition-code results from previous instruction; always emit
10479 compare and test instructions before use of condition codes.
10481 @item -mno-side-effects
10482 @opindex mno-side-effects
10483 Do not emit instructions with side-effects in addressing modes other than
10486 @item -mstack-align
10487 @itemx -mno-stack-align
10488 @itemx -mdata-align
10489 @itemx -mno-data-align
10490 @itemx -mconst-align
10491 @itemx -mno-const-align
10492 @opindex mstack-align
10493 @opindex mno-stack-align
10494 @opindex mdata-align
10495 @opindex mno-data-align
10496 @opindex mconst-align
10497 @opindex mno-const-align
10498 These options (no-options) arranges (eliminate arrangements) for the
10499 stack-frame, individual data and constants to be aligned for the maximum
10500 single data access size for the chosen CPU model. The default is to
10501 arrange for 32-bit alignment. ABI details such as structure layout are
10502 not affected by these options.
10510 Similar to the stack- data- and const-align options above, these options
10511 arrange for stack-frame, writable data and constants to all be 32-bit,
10512 16-bit or 8-bit aligned. The default is 32-bit alignment.
10514 @item -mno-prologue-epilogue
10515 @itemx -mprologue-epilogue
10516 @opindex mno-prologue-epilogue
10517 @opindex mprologue-epilogue
10518 With @option{-mno-prologue-epilogue}, the normal function prologue and
10519 epilogue that sets up the stack-frame are omitted and no return
10520 instructions or return sequences are generated in the code. Use this
10521 option only together with visual inspection of the compiled code: no
10522 warnings or errors are generated when call-saved registers must be saved,
10523 or storage for local variable needs to be allocated.
10527 @opindex mno-gotplt
10529 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
10530 instruction sequences that load addresses for functions from the PLT part
10531 of the GOT rather than (traditional on other architectures) calls to the
10532 PLT@. The default is @option{-mgotplt}.
10536 Legacy no-op option only recognized with the cris-axis-elf and
10537 cris-axis-linux-gnu targets.
10541 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
10545 This option, recognized for the cris-axis-elf arranges
10546 to link with input-output functions from a simulator library. Code,
10547 initialized data and zero-initialized data are allocated consecutively.
10551 Like @option{-sim}, but pass linker options to locate initialized data at
10552 0x40000000 and zero-initialized data at 0x80000000.
10556 @subsection CRX Options
10557 @cindex CRX Options
10559 These options are defined specifically for the CRX ports.
10565 Enable the use of multiply-accumulate instructions. Disabled by default.
10568 @opindex mpush-args
10569 Push instructions will be used to pass outgoing arguments when functions
10570 are called. Enabled by default.
10573 @node Darwin Options
10574 @subsection Darwin Options
10575 @cindex Darwin options
10577 These options are defined for all architectures running the Darwin operating
10580 FSF GCC on Darwin does not create ``fat'' object files; it will create
10581 an object file for the single architecture that it was built to
10582 target. Apple's GCC on Darwin does create ``fat'' files if multiple
10583 @option{-arch} options are used; it does so by running the compiler or
10584 linker multiple times and joining the results together with
10587 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
10588 @samp{i686}) is determined by the flags that specify the ISA
10589 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
10590 @option{-force_cpusubtype_ALL} option can be used to override this.
10592 The Darwin tools vary in their behavior when presented with an ISA
10593 mismatch. The assembler, @file{as}, will only permit instructions to
10594 be used that are valid for the subtype of the file it is generating,
10595 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
10596 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
10597 and print an error if asked to create a shared library with a less
10598 restrictive subtype than its input files (for instance, trying to put
10599 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
10600 for executables, @file{ld}, will quietly give the executable the most
10601 restrictive subtype of any of its input files.
10606 Add the framework directory @var{dir} to the head of the list of
10607 directories to be searched for header files. These directories are
10608 interleaved with those specified by @option{-I} options and are
10609 scanned in a left-to-right order.
10611 A framework directory is a directory with frameworks in it. A
10612 framework is a directory with a @samp{"Headers"} and/or
10613 @samp{"PrivateHeaders"} directory contained directly in it that ends
10614 in @samp{".framework"}. The name of a framework is the name of this
10615 directory excluding the @samp{".framework"}. Headers associated with
10616 the framework are found in one of those two directories, with
10617 @samp{"Headers"} being searched first. A subframework is a framework
10618 directory that is in a framework's @samp{"Frameworks"} directory.
10619 Includes of subframework headers can only appear in a header of a
10620 framework that contains the subframework, or in a sibling subframework
10621 header. Two subframeworks are siblings if they occur in the same
10622 framework. A subframework should not have the same name as a
10623 framework, a warning will be issued if this is violated. Currently a
10624 subframework cannot have subframeworks, in the future, the mechanism
10625 may be extended to support this. The standard frameworks can be found
10626 in @samp{"/System/Library/Frameworks"} and
10627 @samp{"/Library/Frameworks"}. An example include looks like
10628 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
10629 the name of the framework and header.h is found in the
10630 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
10632 @item -iframework@var{dir}
10633 @opindex iframework
10634 Like @option{-F} except the directory is a treated as a system
10635 directory. The main difference between this @option{-iframework} and
10636 @option{-F} is that with @option{-iframework} the compiler does not
10637 warn about constructs contained within header files found via
10638 @var{dir}. This option is valid only for the C family of languages.
10642 Emit debugging information for symbols that are used. For STABS
10643 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
10644 This is by default ON@.
10648 Emit debugging information for all symbols and types.
10650 @item -mmacosx-version-min=@var{version}
10651 The earliest version of MacOS X that this executable will run on
10652 is @var{version}. Typical values of @var{version} include @code{10.1},
10653 @code{10.2}, and @code{10.3.9}.
10655 If the compiler was built to use the system's headers by default,
10656 then the default for this option is the system version on which the
10657 compiler is running, otherwise the default is to make choices which
10658 are compatible with as many systems and code bases as possible.
10662 Enable kernel development mode. The @option{-mkernel} option sets
10663 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
10664 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
10665 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
10666 applicable. This mode also sets @option{-mno-altivec},
10667 @option{-msoft-float}, @option{-fno-builtin} and
10668 @option{-mlong-branch} for PowerPC targets.
10670 @item -mone-byte-bool
10671 @opindex mone-byte-bool
10672 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
10673 By default @samp{sizeof(bool)} is @samp{4} when compiling for
10674 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
10675 option has no effect on x86.
10677 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
10678 to generate code that is not binary compatible with code generated
10679 without that switch. Using this switch may require recompiling all
10680 other modules in a program, including system libraries. Use this
10681 switch to conform to a non-default data model.
10683 @item -mfix-and-continue
10684 @itemx -ffix-and-continue
10685 @itemx -findirect-data
10686 @opindex mfix-and-continue
10687 @opindex ffix-and-continue
10688 @opindex findirect-data
10689 Generate code suitable for fast turn around development. Needed to
10690 enable gdb to dynamically load @code{.o} files into already running
10691 programs. @option{-findirect-data} and @option{-ffix-and-continue}
10692 are provided for backwards compatibility.
10696 Loads all members of static archive libraries.
10697 See man ld(1) for more information.
10699 @item -arch_errors_fatal
10700 @opindex arch_errors_fatal
10701 Cause the errors having to do with files that have the wrong architecture
10704 @item -bind_at_load
10705 @opindex bind_at_load
10706 Causes the output file to be marked such that the dynamic linker will
10707 bind all undefined references when the file is loaded or launched.
10711 Produce a Mach-o bundle format file.
10712 See man ld(1) for more information.
10714 @item -bundle_loader @var{executable}
10715 @opindex bundle_loader
10716 This option specifies the @var{executable} that will be loading the build
10717 output file being linked. See man ld(1) for more information.
10720 @opindex dynamiclib
10721 When passed this option, GCC will produce a dynamic library instead of
10722 an executable when linking, using the Darwin @file{libtool} command.
10724 @item -force_cpusubtype_ALL
10725 @opindex force_cpusubtype_ALL
10726 This causes GCC's output file to have the @var{ALL} subtype, instead of
10727 one controlled by the @option{-mcpu} or @option{-march} option.
10729 @item -allowable_client @var{client_name}
10730 @itemx -client_name
10731 @itemx -compatibility_version
10732 @itemx -current_version
10734 @itemx -dependency-file
10736 @itemx -dylinker_install_name
10738 @itemx -exported_symbols_list
10740 @itemx -flat_namespace
10741 @itemx -force_flat_namespace
10742 @itemx -headerpad_max_install_names
10745 @itemx -install_name
10746 @itemx -keep_private_externs
10747 @itemx -multi_module
10748 @itemx -multiply_defined
10749 @itemx -multiply_defined_unused
10751 @itemx -no_dead_strip_inits_and_terms
10752 @itemx -nofixprebinding
10753 @itemx -nomultidefs
10755 @itemx -noseglinkedit
10756 @itemx -pagezero_size
10758 @itemx -prebind_all_twolevel_modules
10759 @itemx -private_bundle
10760 @itemx -read_only_relocs
10762 @itemx -sectobjectsymbols
10766 @itemx -sectobjectsymbols
10769 @itemx -segs_read_only_addr
10770 @itemx -segs_read_write_addr
10771 @itemx -seg_addr_table
10772 @itemx -seg_addr_table_filename
10773 @itemx -seglinkedit
10775 @itemx -segs_read_only_addr
10776 @itemx -segs_read_write_addr
10777 @itemx -single_module
10779 @itemx -sub_library
10780 @itemx -sub_umbrella
10781 @itemx -twolevel_namespace
10784 @itemx -unexported_symbols_list
10785 @itemx -weak_reference_mismatches
10786 @itemx -whatsloaded
10787 @opindex allowable_client
10788 @opindex client_name
10789 @opindex compatibility_version
10790 @opindex current_version
10791 @opindex dead_strip
10792 @opindex dependency-file
10793 @opindex dylib_file
10794 @opindex dylinker_install_name
10796 @opindex exported_symbols_list
10798 @opindex flat_namespace
10799 @opindex force_flat_namespace
10800 @opindex headerpad_max_install_names
10801 @opindex image_base
10803 @opindex install_name
10804 @opindex keep_private_externs
10805 @opindex multi_module
10806 @opindex multiply_defined
10807 @opindex multiply_defined_unused
10808 @opindex noall_load
10809 @opindex no_dead_strip_inits_and_terms
10810 @opindex nofixprebinding
10811 @opindex nomultidefs
10813 @opindex noseglinkedit
10814 @opindex pagezero_size
10816 @opindex prebind_all_twolevel_modules
10817 @opindex private_bundle
10818 @opindex read_only_relocs
10820 @opindex sectobjectsymbols
10823 @opindex sectcreate
10824 @opindex sectobjectsymbols
10827 @opindex segs_read_only_addr
10828 @opindex segs_read_write_addr
10829 @opindex seg_addr_table
10830 @opindex seg_addr_table_filename
10831 @opindex seglinkedit
10833 @opindex segs_read_only_addr
10834 @opindex segs_read_write_addr
10835 @opindex single_module
10837 @opindex sub_library
10838 @opindex sub_umbrella
10839 @opindex twolevel_namespace
10842 @opindex unexported_symbols_list
10843 @opindex weak_reference_mismatches
10844 @opindex whatsloaded
10845 These options are passed to the Darwin linker. The Darwin linker man page
10846 describes them in detail.
10849 @node DEC Alpha Options
10850 @subsection DEC Alpha Options
10852 These @samp{-m} options are defined for the DEC Alpha implementations:
10855 @item -mno-soft-float
10856 @itemx -msoft-float
10857 @opindex mno-soft-float
10858 @opindex msoft-float
10859 Use (do not use) the hardware floating-point instructions for
10860 floating-point operations. When @option{-msoft-float} is specified,
10861 functions in @file{libgcc.a} will be used to perform floating-point
10862 operations. Unless they are replaced by routines that emulate the
10863 floating-point operations, or compiled in such a way as to call such
10864 emulations routines, these routines will issue floating-point
10865 operations. If you are compiling for an Alpha without floating-point
10866 operations, you must ensure that the library is built so as not to call
10869 Note that Alpha implementations without floating-point operations are
10870 required to have floating-point registers.
10873 @itemx -mno-fp-regs
10875 @opindex mno-fp-regs
10876 Generate code that uses (does not use) the floating-point register set.
10877 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
10878 register set is not used, floating point operands are passed in integer
10879 registers as if they were integers and floating-point results are passed
10880 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
10881 so any function with a floating-point argument or return value called by code
10882 compiled with @option{-mno-fp-regs} must also be compiled with that
10885 A typical use of this option is building a kernel that does not use,
10886 and hence need not save and restore, any floating-point registers.
10890 The Alpha architecture implements floating-point hardware optimized for
10891 maximum performance. It is mostly compliant with the IEEE floating
10892 point standard. However, for full compliance, software assistance is
10893 required. This option generates code fully IEEE compliant code
10894 @emph{except} that the @var{inexact-flag} is not maintained (see below).
10895 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
10896 defined during compilation. The resulting code is less efficient but is
10897 able to correctly support denormalized numbers and exceptional IEEE
10898 values such as not-a-number and plus/minus infinity. Other Alpha
10899 compilers call this option @option{-ieee_with_no_inexact}.
10901 @item -mieee-with-inexact
10902 @opindex mieee-with-inexact
10903 This is like @option{-mieee} except the generated code also maintains
10904 the IEEE @var{inexact-flag}. Turning on this option causes the
10905 generated code to implement fully-compliant IEEE math. In addition to
10906 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
10907 macro. On some Alpha implementations the resulting code may execute
10908 significantly slower than the code generated by default. Since there is
10909 very little code that depends on the @var{inexact-flag}, you should
10910 normally not specify this option. Other Alpha compilers call this
10911 option @option{-ieee_with_inexact}.
10913 @item -mfp-trap-mode=@var{trap-mode}
10914 @opindex mfp-trap-mode
10915 This option controls what floating-point related traps are enabled.
10916 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
10917 The trap mode can be set to one of four values:
10921 This is the default (normal) setting. The only traps that are enabled
10922 are the ones that cannot be disabled in software (e.g., division by zero
10926 In addition to the traps enabled by @samp{n}, underflow traps are enabled
10930 Like @samp{u}, but the instructions are marked to be safe for software
10931 completion (see Alpha architecture manual for details).
10934 Like @samp{su}, but inexact traps are enabled as well.
10937 @item -mfp-rounding-mode=@var{rounding-mode}
10938 @opindex mfp-rounding-mode
10939 Selects the IEEE rounding mode. Other Alpha compilers call this option
10940 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
10945 Normal IEEE rounding mode. Floating point numbers are rounded towards
10946 the nearest machine number or towards the even machine number in case
10950 Round towards minus infinity.
10953 Chopped rounding mode. Floating point numbers are rounded towards zero.
10956 Dynamic rounding mode. A field in the floating point control register
10957 (@var{fpcr}, see Alpha architecture reference manual) controls the
10958 rounding mode in effect. The C library initializes this register for
10959 rounding towards plus infinity. Thus, unless your program modifies the
10960 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
10963 @item -mtrap-precision=@var{trap-precision}
10964 @opindex mtrap-precision
10965 In the Alpha architecture, floating point traps are imprecise. This
10966 means without software assistance it is impossible to recover from a
10967 floating trap and program execution normally needs to be terminated.
10968 GCC can generate code that can assist operating system trap handlers
10969 in determining the exact location that caused a floating point trap.
10970 Depending on the requirements of an application, different levels of
10971 precisions can be selected:
10975 Program precision. This option is the default and means a trap handler
10976 can only identify which program caused a floating point exception.
10979 Function precision. The trap handler can determine the function that
10980 caused a floating point exception.
10983 Instruction precision. The trap handler can determine the exact
10984 instruction that caused a floating point exception.
10987 Other Alpha compilers provide the equivalent options called
10988 @option{-scope_safe} and @option{-resumption_safe}.
10990 @item -mieee-conformant
10991 @opindex mieee-conformant
10992 This option marks the generated code as IEEE conformant. You must not
10993 use this option unless you also specify @option{-mtrap-precision=i} and either
10994 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
10995 is to emit the line @samp{.eflag 48} in the function prologue of the
10996 generated assembly file. Under DEC Unix, this has the effect that
10997 IEEE-conformant math library routines will be linked in.
10999 @item -mbuild-constants
11000 @opindex mbuild-constants
11001 Normally GCC examines a 32- or 64-bit integer constant to
11002 see if it can construct it from smaller constants in two or three
11003 instructions. If it cannot, it will output the constant as a literal and
11004 generate code to load it from the data segment at runtime.
11006 Use this option to require GCC to construct @emph{all} integer constants
11007 using code, even if it takes more instructions (the maximum is six).
11009 You would typically use this option to build a shared library dynamic
11010 loader. Itself a shared library, it must relocate itself in memory
11011 before it can find the variables and constants in its own data segment.
11017 Select whether to generate code to be assembled by the vendor-supplied
11018 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
11036 Indicate whether GCC should generate code to use the optional BWX,
11037 CIX, FIX and MAX instruction sets. The default is to use the instruction
11038 sets supported by the CPU type specified via @option{-mcpu=} option or that
11039 of the CPU on which GCC was built if none was specified.
11042 @itemx -mfloat-ieee
11043 @opindex mfloat-vax
11044 @opindex mfloat-ieee
11045 Generate code that uses (does not use) VAX F and G floating point
11046 arithmetic instead of IEEE single and double precision.
11048 @item -mexplicit-relocs
11049 @itemx -mno-explicit-relocs
11050 @opindex mexplicit-relocs
11051 @opindex mno-explicit-relocs
11052 Older Alpha assemblers provided no way to generate symbol relocations
11053 except via assembler macros. Use of these macros does not allow
11054 optimal instruction scheduling. GNU binutils as of version 2.12
11055 supports a new syntax that allows the compiler to explicitly mark
11056 which relocations should apply to which instructions. This option
11057 is mostly useful for debugging, as GCC detects the capabilities of
11058 the assembler when it is built and sets the default accordingly.
11061 @itemx -mlarge-data
11062 @opindex msmall-data
11063 @opindex mlarge-data
11064 When @option{-mexplicit-relocs} is in effect, static data is
11065 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
11066 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
11067 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
11068 16-bit relocations off of the @code{$gp} register. This limits the
11069 size of the small data area to 64KB, but allows the variables to be
11070 directly accessed via a single instruction.
11072 The default is @option{-mlarge-data}. With this option the data area
11073 is limited to just below 2GB@. Programs that require more than 2GB of
11074 data must use @code{malloc} or @code{mmap} to allocate the data in the
11075 heap instead of in the program's data segment.
11077 When generating code for shared libraries, @option{-fpic} implies
11078 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
11081 @itemx -mlarge-text
11082 @opindex msmall-text
11083 @opindex mlarge-text
11084 When @option{-msmall-text} is used, the compiler assumes that the
11085 code of the entire program (or shared library) fits in 4MB, and is
11086 thus reachable with a branch instruction. When @option{-msmall-data}
11087 is used, the compiler can assume that all local symbols share the
11088 same @code{$gp} value, and thus reduce the number of instructions
11089 required for a function call from 4 to 1.
11091 The default is @option{-mlarge-text}.
11093 @item -mcpu=@var{cpu_type}
11095 Set the instruction set and instruction scheduling parameters for
11096 machine type @var{cpu_type}. You can specify either the @samp{EV}
11097 style name or the corresponding chip number. GCC supports scheduling
11098 parameters for the EV4, EV5 and EV6 family of processors and will
11099 choose the default values for the instruction set from the processor
11100 you specify. If you do not specify a processor type, GCC will default
11101 to the processor on which the compiler was built.
11103 Supported values for @var{cpu_type} are
11109 Schedules as an EV4 and has no instruction set extensions.
11113 Schedules as an EV5 and has no instruction set extensions.
11117 Schedules as an EV5 and supports the BWX extension.
11122 Schedules as an EV5 and supports the BWX and MAX extensions.
11126 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
11130 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
11133 Native Linux/GNU toolchains also support the value @samp{native},
11134 which selects the best architecture option for the host processor.
11135 @option{-mcpu=native} has no effect if GCC does not recognize
11138 @item -mtune=@var{cpu_type}
11140 Set only the instruction scheduling parameters for machine type
11141 @var{cpu_type}. The instruction set is not changed.
11143 Native Linux/GNU toolchains also support the value @samp{native},
11144 which selects the best architecture option for the host processor.
11145 @option{-mtune=native} has no effect if GCC does not recognize
11148 @item -mmemory-latency=@var{time}
11149 @opindex mmemory-latency
11150 Sets the latency the scheduler should assume for typical memory
11151 references as seen by the application. This number is highly
11152 dependent on the memory access patterns used by the application
11153 and the size of the external cache on the machine.
11155 Valid options for @var{time} are
11159 A decimal number representing clock cycles.
11165 The compiler contains estimates of the number of clock cycles for
11166 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
11167 (also called Dcache, Scache, and Bcache), as well as to main memory.
11168 Note that L3 is only valid for EV5.
11173 @node DEC Alpha/VMS Options
11174 @subsection DEC Alpha/VMS Options
11176 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
11179 @item -mvms-return-codes
11180 @opindex mvms-return-codes
11181 Return VMS condition codes from main. The default is to return POSIX
11182 style condition (e.g.@: error) codes.
11184 @item -mdebug-main=@var{prefix}
11185 @opindex mdebug-main=@var{prefix}
11186 Flag the first routine whose name starts with @var{prefix} as the main
11187 routine for the debugger.
11191 Default to 64bit memory allocation routines.
11195 @subsection FR30 Options
11196 @cindex FR30 Options
11198 These options are defined specifically for the FR30 port.
11202 @item -msmall-model
11203 @opindex msmall-model
11204 Use the small address space model. This can produce smaller code, but
11205 it does assume that all symbolic values and addresses will fit into a
11210 Assume that run-time support has been provided and so there is no need
11211 to include the simulator library (@file{libsim.a}) on the linker
11217 @subsection FRV Options
11218 @cindex FRV Options
11224 Only use the first 32 general purpose registers.
11229 Use all 64 general purpose registers.
11234 Use only the first 32 floating point registers.
11239 Use all 64 floating point registers
11242 @opindex mhard-float
11244 Use hardware instructions for floating point operations.
11247 @opindex msoft-float
11249 Use library routines for floating point operations.
11254 Dynamically allocate condition code registers.
11259 Do not try to dynamically allocate condition code registers, only
11260 use @code{icc0} and @code{fcc0}.
11265 Change ABI to use double word insns.
11270 Do not use double word instructions.
11275 Use floating point double instructions.
11278 @opindex mno-double
11280 Do not use floating point double instructions.
11285 Use media instructions.
11290 Do not use media instructions.
11295 Use multiply and add/subtract instructions.
11298 @opindex mno-muladd
11300 Do not use multiply and add/subtract instructions.
11305 Select the FDPIC ABI, that uses function descriptors to represent
11306 pointers to functions. Without any PIC/PIE-related options, it
11307 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
11308 assumes GOT entries and small data are within a 12-bit range from the
11309 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
11310 are computed with 32 bits.
11311 With a @samp{bfin-elf} target, this option implies @option{-msim}.
11314 @opindex minline-plt
11316 Enable inlining of PLT entries in function calls to functions that are
11317 not known to bind locally. It has no effect without @option{-mfdpic}.
11318 It's enabled by default if optimizing for speed and compiling for
11319 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
11320 optimization option such as @option{-O3} or above is present in the
11326 Assume a large TLS segment when generating thread-local code.
11331 Do not assume a large TLS segment when generating thread-local code.
11336 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
11337 that is known to be in read-only sections. It's enabled by default,
11338 except for @option{-fpic} or @option{-fpie}: even though it may help
11339 make the global offset table smaller, it trades 1 instruction for 4.
11340 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
11341 one of which may be shared by multiple symbols, and it avoids the need
11342 for a GOT entry for the referenced symbol, so it's more likely to be a
11343 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
11345 @item -multilib-library-pic
11346 @opindex multilib-library-pic
11348 Link with the (library, not FD) pic libraries. It's implied by
11349 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
11350 @option{-fpic} without @option{-mfdpic}. You should never have to use
11354 @opindex mlinked-fp
11356 Follow the EABI requirement of always creating a frame pointer whenever
11357 a stack frame is allocated. This option is enabled by default and can
11358 be disabled with @option{-mno-linked-fp}.
11361 @opindex mlong-calls
11363 Use indirect addressing to call functions outside the current
11364 compilation unit. This allows the functions to be placed anywhere
11365 within the 32-bit address space.
11367 @item -malign-labels
11368 @opindex malign-labels
11370 Try to align labels to an 8-byte boundary by inserting nops into the
11371 previous packet. This option only has an effect when VLIW packing
11372 is enabled. It doesn't create new packets; it merely adds nops to
11375 @item -mlibrary-pic
11376 @opindex mlibrary-pic
11378 Generate position-independent EABI code.
11383 Use only the first four media accumulator registers.
11388 Use all eight media accumulator registers.
11393 Pack VLIW instructions.
11398 Do not pack VLIW instructions.
11401 @opindex mno-eflags
11403 Do not mark ABI switches in e_flags.
11406 @opindex mcond-move
11408 Enable the use of conditional-move instructions (default).
11410 This switch is mainly for debugging the compiler and will likely be removed
11411 in a future version.
11413 @item -mno-cond-move
11414 @opindex mno-cond-move
11416 Disable the use of conditional-move instructions.
11418 This switch is mainly for debugging the compiler and will likely be removed
11419 in a future version.
11424 Enable the use of conditional set instructions (default).
11426 This switch is mainly for debugging the compiler and will likely be removed
11427 in a future version.
11432 Disable the use of conditional set instructions.
11434 This switch is mainly for debugging the compiler and will likely be removed
11435 in a future version.
11438 @opindex mcond-exec
11440 Enable the use of conditional execution (default).
11442 This switch is mainly for debugging the compiler and will likely be removed
11443 in a future version.
11445 @item -mno-cond-exec
11446 @opindex mno-cond-exec
11448 Disable the use of conditional execution.
11450 This switch is mainly for debugging the compiler and will likely be removed
11451 in a future version.
11453 @item -mvliw-branch
11454 @opindex mvliw-branch
11456 Run a pass to pack branches into VLIW instructions (default).
11458 This switch is mainly for debugging the compiler and will likely be removed
11459 in a future version.
11461 @item -mno-vliw-branch
11462 @opindex mno-vliw-branch
11464 Do not run a pass to pack branches into VLIW instructions.
11466 This switch is mainly for debugging the compiler and will likely be removed
11467 in a future version.
11469 @item -mmulti-cond-exec
11470 @opindex mmulti-cond-exec
11472 Enable optimization of @code{&&} and @code{||} in conditional execution
11475 This switch is mainly for debugging the compiler and will likely be removed
11476 in a future version.
11478 @item -mno-multi-cond-exec
11479 @opindex mno-multi-cond-exec
11481 Disable optimization of @code{&&} and @code{||} in conditional execution.
11483 This switch is mainly for debugging the compiler and will likely be removed
11484 in a future version.
11486 @item -mnested-cond-exec
11487 @opindex mnested-cond-exec
11489 Enable nested conditional execution optimizations (default).
11491 This switch is mainly for debugging the compiler and will likely be removed
11492 in a future version.
11494 @item -mno-nested-cond-exec
11495 @opindex mno-nested-cond-exec
11497 Disable nested conditional execution optimizations.
11499 This switch is mainly for debugging the compiler and will likely be removed
11500 in a future version.
11502 @item -moptimize-membar
11503 @opindex moptimize-membar
11505 This switch removes redundant @code{membar} instructions from the
11506 compiler generated code. It is enabled by default.
11508 @item -mno-optimize-membar
11509 @opindex mno-optimize-membar
11511 This switch disables the automatic removal of redundant @code{membar}
11512 instructions from the generated code.
11514 @item -mtomcat-stats
11515 @opindex mtomcat-stats
11517 Cause gas to print out tomcat statistics.
11519 @item -mcpu=@var{cpu}
11522 Select the processor type for which to generate code. Possible values are
11523 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
11524 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
11528 @node GNU/Linux Options
11529 @subsection GNU/Linux Options
11531 These @samp{-m} options are defined for GNU/Linux targets:
11536 Use the GNU C library. This is the default except
11537 on @samp{*-*-linux-*uclibc*} and @samp{*-*-linux-*android*} targets.
11541 Use uClibc C library. This is the default on
11542 @samp{*-*-linux-*uclibc*} targets.
11546 Use Bionic C library. This is the default on
11547 @samp{*-*-linux-*android*} targets.
11551 Compile code compatible with Android platform. This is the default on
11552 @samp{*-*-linux-*android*} targets.
11554 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
11555 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
11556 this option makes the GCC driver pass Android-specific options to the linker.
11557 Finally, this option causes the preprocessor macro @code{__ANDROID__}
11560 @item -tno-android-cc
11561 @opindex tno-android-cc
11562 Disable compilation effects of @option{-mandroid}, i.e., do not enable
11563 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
11564 @option{-fno-rtti} by default.
11566 @item -tno-android-ld
11567 @opindex tno-android-ld
11568 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
11569 linking options to the linker.
11573 @node H8/300 Options
11574 @subsection H8/300 Options
11576 These @samp{-m} options are defined for the H8/300 implementations:
11581 Shorten some address references at link time, when possible; uses the
11582 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
11583 ld, Using ld}, for a fuller description.
11587 Generate code for the H8/300H@.
11591 Generate code for the H8S@.
11595 Generate code for the H8S and H8/300H in the normal mode. This switch
11596 must be used either with @option{-mh} or @option{-ms}.
11600 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
11604 Make @code{int} data 32 bits by default.
11607 @opindex malign-300
11608 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
11609 The default for the H8/300H and H8S is to align longs and floats on 4
11611 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
11612 This option has no effect on the H8/300.
11616 @subsection HPPA Options
11617 @cindex HPPA Options
11619 These @samp{-m} options are defined for the HPPA family of computers:
11622 @item -march=@var{architecture-type}
11624 Generate code for the specified architecture. The choices for
11625 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
11626 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
11627 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
11628 architecture option for your machine. Code compiled for lower numbered
11629 architectures will run on higher numbered architectures, but not the
11632 @item -mpa-risc-1-0
11633 @itemx -mpa-risc-1-1
11634 @itemx -mpa-risc-2-0
11635 @opindex mpa-risc-1-0
11636 @opindex mpa-risc-1-1
11637 @opindex mpa-risc-2-0
11638 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
11641 @opindex mbig-switch
11642 Generate code suitable for big switch tables. Use this option only if
11643 the assembler/linker complain about out of range branches within a switch
11646 @item -mjump-in-delay
11647 @opindex mjump-in-delay
11648 Fill delay slots of function calls with unconditional jump instructions
11649 by modifying the return pointer for the function call to be the target
11650 of the conditional jump.
11652 @item -mdisable-fpregs
11653 @opindex mdisable-fpregs
11654 Prevent floating point registers from being used in any manner. This is
11655 necessary for compiling kernels which perform lazy context switching of
11656 floating point registers. If you use this option and attempt to perform
11657 floating point operations, the compiler will abort.
11659 @item -mdisable-indexing
11660 @opindex mdisable-indexing
11661 Prevent the compiler from using indexing address modes. This avoids some
11662 rather obscure problems when compiling MIG generated code under MACH@.
11664 @item -mno-space-regs
11665 @opindex mno-space-regs
11666 Generate code that assumes the target has no space registers. This allows
11667 GCC to generate faster indirect calls and use unscaled index address modes.
11669 Such code is suitable for level 0 PA systems and kernels.
11671 @item -mfast-indirect-calls
11672 @opindex mfast-indirect-calls
11673 Generate code that assumes calls never cross space boundaries. This
11674 allows GCC to emit code which performs faster indirect calls.
11676 This option will not work in the presence of shared libraries or nested
11679 @item -mfixed-range=@var{register-range}
11680 @opindex mfixed-range
11681 Generate code treating the given register range as fixed registers.
11682 A fixed register is one that the register allocator can not use. This is
11683 useful when compiling kernel code. A register range is specified as
11684 two registers separated by a dash. Multiple register ranges can be
11685 specified separated by a comma.
11687 @item -mlong-load-store
11688 @opindex mlong-load-store
11689 Generate 3-instruction load and store sequences as sometimes required by
11690 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
11693 @item -mportable-runtime
11694 @opindex mportable-runtime
11695 Use the portable calling conventions proposed by HP for ELF systems.
11699 Enable the use of assembler directives only GAS understands.
11701 @item -mschedule=@var{cpu-type}
11703 Schedule code according to the constraints for the machine type
11704 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
11705 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
11706 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
11707 proper scheduling option for your machine. The default scheduling is
11711 @opindex mlinker-opt
11712 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
11713 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
11714 linkers in which they give bogus error messages when linking some programs.
11717 @opindex msoft-float
11718 Generate output containing library calls for floating point.
11719 @strong{Warning:} the requisite libraries are not available for all HPPA
11720 targets. Normally the facilities of the machine's usual C compiler are
11721 used, but this cannot be done directly in cross-compilation. You must make
11722 your own arrangements to provide suitable library functions for
11725 @option{-msoft-float} changes the calling convention in the output file;
11726 therefore, it is only useful if you compile @emph{all} of a program with
11727 this option. In particular, you need to compile @file{libgcc.a}, the
11728 library that comes with GCC, with @option{-msoft-float} in order for
11733 Generate the predefine, @code{_SIO}, for server IO@. The default is
11734 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
11735 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
11736 options are available under HP-UX and HI-UX@.
11740 Use GNU ld specific options. This passes @option{-shared} to ld when
11741 building a shared library. It is the default when GCC is configured,
11742 explicitly or implicitly, with the GNU linker. This option does not
11743 have any affect on which ld is called, it only changes what parameters
11744 are passed to that ld. The ld that is called is determined by the
11745 @option{--with-ld} configure option, GCC's program search path, and
11746 finally by the user's @env{PATH}. The linker used by GCC can be printed
11747 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
11748 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11752 Use HP ld specific options. This passes @option{-b} to ld when building
11753 a shared library and passes @option{+Accept TypeMismatch} to ld on all
11754 links. It is the default when GCC is configured, explicitly or
11755 implicitly, with the HP linker. This option does not have any affect on
11756 which ld is called, it only changes what parameters are passed to that
11757 ld. The ld that is called is determined by the @option{--with-ld}
11758 configure option, GCC's program search path, and finally by the user's
11759 @env{PATH}. The linker used by GCC can be printed using @samp{which
11760 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
11761 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11764 @opindex mno-long-calls
11765 Generate code that uses long call sequences. This ensures that a call
11766 is always able to reach linker generated stubs. The default is to generate
11767 long calls only when the distance from the call site to the beginning
11768 of the function or translation unit, as the case may be, exceeds a
11769 predefined limit set by the branch type being used. The limits for
11770 normal calls are 7,600,000 and 240,000 bytes, respectively for the
11771 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
11774 Distances are measured from the beginning of functions when using the
11775 @option{-ffunction-sections} option, or when using the @option{-mgas}
11776 and @option{-mno-portable-runtime} options together under HP-UX with
11779 It is normally not desirable to use this option as it will degrade
11780 performance. However, it may be useful in large applications,
11781 particularly when partial linking is used to build the application.
11783 The types of long calls used depends on the capabilities of the
11784 assembler and linker, and the type of code being generated. The
11785 impact on systems that support long absolute calls, and long pic
11786 symbol-difference or pc-relative calls should be relatively small.
11787 However, an indirect call is used on 32-bit ELF systems in pic code
11788 and it is quite long.
11790 @item -munix=@var{unix-std}
11792 Generate compiler predefines and select a startfile for the specified
11793 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
11794 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
11795 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
11796 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
11797 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
11800 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
11801 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
11802 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
11803 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
11804 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
11805 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
11807 It is @emph{important} to note that this option changes the interfaces
11808 for various library routines. It also affects the operational behavior
11809 of the C library. Thus, @emph{extreme} care is needed in using this
11812 Library code that is intended to operate with more than one UNIX
11813 standard must test, set and restore the variable @var{__xpg4_extended_mask}
11814 as appropriate. Most GNU software doesn't provide this capability.
11818 Suppress the generation of link options to search libdld.sl when the
11819 @option{-static} option is specified on HP-UX 10 and later.
11823 The HP-UX implementation of setlocale in libc has a dependency on
11824 libdld.sl. There isn't an archive version of libdld.sl. Thus,
11825 when the @option{-static} option is specified, special link options
11826 are needed to resolve this dependency.
11828 On HP-UX 10 and later, the GCC driver adds the necessary options to
11829 link with libdld.sl when the @option{-static} option is specified.
11830 This causes the resulting binary to be dynamic. On the 64-bit port,
11831 the linkers generate dynamic binaries by default in any case. The
11832 @option{-nolibdld} option can be used to prevent the GCC driver from
11833 adding these link options.
11837 Add support for multithreading with the @dfn{dce thread} library
11838 under HP-UX@. This option sets flags for both the preprocessor and
11842 @node i386 and x86-64 Options
11843 @subsection Intel 386 and AMD x86-64 Options
11844 @cindex i386 Options
11845 @cindex x86-64 Options
11846 @cindex Intel 386 Options
11847 @cindex AMD x86-64 Options
11849 These @samp{-m} options are defined for the i386 and x86-64 family of
11853 @item -mtune=@var{cpu-type}
11855 Tune to @var{cpu-type} everything applicable about the generated code, except
11856 for the ABI and the set of available instructions. The choices for
11857 @var{cpu-type} are:
11860 Produce code optimized for the most common IA32/AMD64/EM64T processors.
11861 If you know the CPU on which your code will run, then you should use
11862 the corresponding @option{-mtune} option instead of
11863 @option{-mtune=generic}. But, if you do not know exactly what CPU users
11864 of your application will have, then you should use this option.
11866 As new processors are deployed in the marketplace, the behavior of this
11867 option will change. Therefore, if you upgrade to a newer version of
11868 GCC, the code generated option will change to reflect the processors
11869 that were most common when that version of GCC was released.
11871 There is no @option{-march=generic} option because @option{-march}
11872 indicates the instruction set the compiler can use, and there is no
11873 generic instruction set applicable to all processors. In contrast,
11874 @option{-mtune} indicates the processor (or, in this case, collection of
11875 processors) for which the code is optimized.
11877 This selects the CPU to tune for at compilation time by determining
11878 the processor type of the compiling machine. Using @option{-mtune=native}
11879 will produce code optimized for the local machine under the constraints
11880 of the selected instruction set. Using @option{-march=native} will
11881 enable all instruction subsets supported by the local machine (hence
11882 the result might not run on different machines).
11884 Original Intel's i386 CPU@.
11886 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
11887 @item i586, pentium
11888 Intel Pentium CPU with no MMX support.
11890 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
11892 Intel PentiumPro CPU@.
11894 Same as @code{generic}, but when used as @code{march} option, PentiumPro
11895 instruction set will be used, so the code will run on all i686 family chips.
11897 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
11898 @item pentium3, pentium3m
11899 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
11902 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
11903 support. Used by Centrino notebooks.
11904 @item pentium4, pentium4m
11905 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
11907 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
11910 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
11911 SSE2 and SSE3 instruction set support.
11913 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
11914 instruction set support.
11916 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
11917 instruction set support.
11919 AMD K6 CPU with MMX instruction set support.
11921 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
11922 @item athlon, athlon-tbird
11923 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
11925 @item athlon-4, athlon-xp, athlon-mp
11926 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
11927 instruction set support.
11928 @item k8, opteron, athlon64, athlon-fx
11929 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
11930 MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit instruction set extensions.)
11931 @item k8-sse3, opteron-sse3, athlon64-sse3
11932 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
11933 @item amdfam10, barcelona
11934 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
11935 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
11936 instruction set extensions.)
11938 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
11941 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
11942 instruction set support.
11944 Via C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
11945 implemented for this chip.)
11947 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
11948 implemented for this chip.)
11950 Embedded AMD CPU with MMX and 3DNow!@: instruction set support.
11953 While picking a specific @var{cpu-type} will schedule things appropriately
11954 for that particular chip, the compiler will not generate any code that
11955 does not run on the i386 without the @option{-march=@var{cpu-type}} option
11958 @item -march=@var{cpu-type}
11960 Generate instructions for the machine type @var{cpu-type}. The choices
11961 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
11962 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
11964 @item -mcpu=@var{cpu-type}
11966 A deprecated synonym for @option{-mtune}.
11968 @item -mfpmath=@var{unit}
11970 Generate floating point arithmetics for selected unit @var{unit}. The choices
11971 for @var{unit} are:
11975 Use the standard 387 floating point coprocessor present majority of chips and
11976 emulated otherwise. Code compiled with this option will run almost everywhere.
11977 The temporary results are computed in 80bit precision instead of precision
11978 specified by the type resulting in slightly different results compared to most
11979 of other chips. See @option{-ffloat-store} for more detailed description.
11981 This is the default choice for i386 compiler.
11984 Use scalar floating point instructions present in the SSE instruction set.
11985 This instruction set is supported by Pentium3 and newer chips, in the AMD line
11986 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
11987 instruction set supports only single precision arithmetics, thus the double and
11988 extended precision arithmetics is still done using 387. Later version, present
11989 only in Pentium4 and the future AMD x86-64 chips supports double precision
11992 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
11993 or @option{-msse2} switches to enable SSE extensions and make this option
11994 effective. For the x86-64 compiler, these extensions are enabled by default.
11996 The resulting code should be considerably faster in the majority of cases and avoid
11997 the numerical instability problems of 387 code, but may break some existing
11998 code that expects temporaries to be 80bit.
12000 This is the default choice for the x86-64 compiler.
12005 Attempt to utilize both instruction sets at once. This effectively double the
12006 amount of available registers and on chips with separate execution units for
12007 387 and SSE the execution resources too. Use this option with care, as it is
12008 still experimental, because the GCC register allocator does not model separate
12009 functional units well resulting in instable performance.
12012 @item -masm=@var{dialect}
12013 @opindex masm=@var{dialect}
12014 Output asm instructions using selected @var{dialect}. Supported
12015 choices are @samp{intel} or @samp{att} (the default one). Darwin does
12016 not support @samp{intel}.
12019 @itemx -mno-ieee-fp
12021 @opindex mno-ieee-fp
12022 Control whether or not the compiler uses IEEE floating point
12023 comparisons. These handle correctly the case where the result of a
12024 comparison is unordered.
12027 @opindex msoft-float
12028 Generate output containing library calls for floating point.
12029 @strong{Warning:} the requisite libraries are not part of GCC@.
12030 Normally the facilities of the machine's usual C compiler are used, but
12031 this can't be done directly in cross-compilation. You must make your
12032 own arrangements to provide suitable library functions for
12035 On machines where a function returns floating point results in the 80387
12036 register stack, some floating point opcodes may be emitted even if
12037 @option{-msoft-float} is used.
12039 @item -mno-fp-ret-in-387
12040 @opindex mno-fp-ret-in-387
12041 Do not use the FPU registers for return values of functions.
12043 The usual calling convention has functions return values of types
12044 @code{float} and @code{double} in an FPU register, even if there
12045 is no FPU@. The idea is that the operating system should emulate
12048 The option @option{-mno-fp-ret-in-387} causes such values to be returned
12049 in ordinary CPU registers instead.
12051 @item -mno-fancy-math-387
12052 @opindex mno-fancy-math-387
12053 Some 387 emulators do not support the @code{sin}, @code{cos} and
12054 @code{sqrt} instructions for the 387. Specify this option to avoid
12055 generating those instructions. This option is the default on FreeBSD,
12056 OpenBSD and NetBSD@. This option is overridden when @option{-march}
12057 indicates that the target cpu will always have an FPU and so the
12058 instruction will not need emulation. As of revision 2.6.1, these
12059 instructions are not generated unless you also use the
12060 @option{-funsafe-math-optimizations} switch.
12062 @item -malign-double
12063 @itemx -mno-align-double
12064 @opindex malign-double
12065 @opindex mno-align-double
12066 Control whether GCC aligns @code{double}, @code{long double}, and
12067 @code{long long} variables on a two word boundary or a one word
12068 boundary. Aligning @code{double} variables on a two word boundary will
12069 produce code that runs somewhat faster on a @samp{Pentium} at the
12070 expense of more memory.
12072 On x86-64, @option{-malign-double} is enabled by default.
12074 @strong{Warning:} if you use the @option{-malign-double} switch,
12075 structures containing the above types will be aligned differently than
12076 the published application binary interface specifications for the 386
12077 and will not be binary compatible with structures in code compiled
12078 without that switch.
12080 @item -m96bit-long-double
12081 @itemx -m128bit-long-double
12082 @opindex m96bit-long-double
12083 @opindex m128bit-long-double
12084 These switches control the size of @code{long double} type. The i386
12085 application binary interface specifies the size to be 96 bits,
12086 so @option{-m96bit-long-double} is the default in 32 bit mode.
12088 Modern architectures (Pentium and newer) would prefer @code{long double}
12089 to be aligned to an 8 or 16 byte boundary. In arrays or structures
12090 conforming to the ABI, this would not be possible. So specifying a
12091 @option{-m128bit-long-double} will align @code{long double}
12092 to a 16 byte boundary by padding the @code{long double} with an additional
12095 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
12096 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
12098 Notice that neither of these options enable any extra precision over the x87
12099 standard of 80 bits for a @code{long double}.
12101 @strong{Warning:} if you override the default value for your target ABI, the
12102 structures and arrays containing @code{long double} variables will change
12103 their size as well as function calling convention for function taking
12104 @code{long double} will be modified. Hence they will not be binary
12105 compatible with arrays or structures in code compiled without that switch.
12107 @item -mlarge-data-threshold=@var{number}
12108 @opindex mlarge-data-threshold=@var{number}
12109 When @option{-mcmodel=medium} is specified, the data greater than
12110 @var{threshold} are placed in large data section. This value must be the
12111 same across all object linked into the binary and defaults to 65535.
12115 Use a different function-calling convention, in which functions that
12116 take a fixed number of arguments return with the @code{ret} @var{num}
12117 instruction, which pops their arguments while returning. This saves one
12118 instruction in the caller since there is no need to pop the arguments
12121 You can specify that an individual function is called with this calling
12122 sequence with the function attribute @samp{stdcall}. You can also
12123 override the @option{-mrtd} option by using the function attribute
12124 @samp{cdecl}. @xref{Function Attributes}.
12126 @strong{Warning:} this calling convention is incompatible with the one
12127 normally used on Unix, so you cannot use it if you need to call
12128 libraries compiled with the Unix compiler.
12130 Also, you must provide function prototypes for all functions that
12131 take variable numbers of arguments (including @code{printf});
12132 otherwise incorrect code will be generated for calls to those
12135 In addition, seriously incorrect code will result if you call a
12136 function with too many arguments. (Normally, extra arguments are
12137 harmlessly ignored.)
12139 @item -mregparm=@var{num}
12141 Control how many registers are used to pass integer arguments. By
12142 default, no registers are used to pass arguments, and at most 3
12143 registers can be used. You can control this behavior for a specific
12144 function by using the function attribute @samp{regparm}.
12145 @xref{Function Attributes}.
12147 @strong{Warning:} if you use this switch, and
12148 @var{num} is nonzero, then you must build all modules with the same
12149 value, including any libraries. This includes the system libraries and
12153 @opindex msseregparm
12154 Use SSE register passing conventions for float and double arguments
12155 and return values. You can control this behavior for a specific
12156 function by using the function attribute @samp{sseregparm}.
12157 @xref{Function Attributes}.
12159 @strong{Warning:} if you use this switch then you must build all
12160 modules with the same value, including any libraries. This includes
12161 the system libraries and startup modules.
12170 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
12171 is specified, the significands of results of floating-point operations are
12172 rounded to 24 bits (single precision); @option{-mpc64} rounds the
12173 significands of results of floating-point operations to 53 bits (double
12174 precision) and @option{-mpc80} rounds the significands of results of
12175 floating-point operations to 64 bits (extended double precision), which is
12176 the default. When this option is used, floating-point operations in higher
12177 precisions are not available to the programmer without setting the FPU
12178 control word explicitly.
12180 Setting the rounding of floating-point operations to less than the default
12181 80 bits can speed some programs by 2% or more. Note that some mathematical
12182 libraries assume that extended precision (80 bit) floating-point operations
12183 are enabled by default; routines in such libraries could suffer significant
12184 loss of accuracy, typically through so-called "catastrophic cancellation",
12185 when this option is used to set the precision to less than extended precision.
12187 @item -mstackrealign
12188 @opindex mstackrealign
12189 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
12190 option will generate an alternate prologue and epilogue that realigns the
12191 runtime stack if necessary. This supports mixing legacy codes that keep
12192 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
12193 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
12194 applicable to individual functions.
12196 @item -mpreferred-stack-boundary=@var{num}
12197 @opindex mpreferred-stack-boundary
12198 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
12199 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
12200 the default is 4 (16 bytes or 128 bits).
12202 @item -mincoming-stack-boundary=@var{num}
12203 @opindex mincoming-stack-boundary
12204 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
12205 boundary. If @option{-mincoming-stack-boundary} is not specified,
12206 the one specified by @option{-mpreferred-stack-boundary} will be used.
12208 On Pentium and PentiumPro, @code{double} and @code{long double} values
12209 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
12210 suffer significant run time performance penalties. On Pentium III, the
12211 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
12212 properly if it is not 16 byte aligned.
12214 To ensure proper alignment of this values on the stack, the stack boundary
12215 must be as aligned as that required by any value stored on the stack.
12216 Further, every function must be generated such that it keeps the stack
12217 aligned. Thus calling a function compiled with a higher preferred
12218 stack boundary from a function compiled with a lower preferred stack
12219 boundary will most likely misalign the stack. It is recommended that
12220 libraries that use callbacks always use the default setting.
12222 This extra alignment does consume extra stack space, and generally
12223 increases code size. Code that is sensitive to stack space usage, such
12224 as embedded systems and operating system kernels, may want to reduce the
12225 preferred alignment to @option{-mpreferred-stack-boundary=2}.
12250 @itemx -mno-fsgsbase
12275 These switches enable or disable the use of instructions in the MMX,
12276 SSE, SSE2, SSE3, SSSE3, SSE4.1, AVX, AES, PCLMUL, FSGSBASE, RDRND,
12277 F16C, SSE4A, FMA4, XOP, LWP, ABM or 3DNow!@: extended instruction sets.
12278 These extensions are also available as built-in functions: see
12279 @ref{X86 Built-in Functions}, for details of the functions enabled and
12280 disabled by these switches.
12282 To have SSE/SSE2 instructions generated automatically from floating-point
12283 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
12285 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
12286 generates new AVX instructions or AVX equivalence for all SSEx instructions
12289 These options will enable GCC to use these extended instructions in
12290 generated code, even without @option{-mfpmath=sse}. Applications which
12291 perform runtime CPU detection must compile separate files for each
12292 supported architecture, using the appropriate flags. In particular,
12293 the file containing the CPU detection code should be compiled without
12297 @itemx -mno-fused-madd
12298 @opindex mfused-madd
12299 @opindex mno-fused-madd
12300 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
12301 instructions. The default is to use these instructions.
12305 This option instructs GCC to emit a @code{cld} instruction in the prologue
12306 of functions that use string instructions. String instructions depend on
12307 the DF flag to select between autoincrement or autodecrement mode. While the
12308 ABI specifies the DF flag to be cleared on function entry, some operating
12309 systems violate this specification by not clearing the DF flag in their
12310 exception dispatchers. The exception handler can be invoked with the DF flag
12311 set which leads to wrong direction mode, when string instructions are used.
12312 This option can be enabled by default on 32-bit x86 targets by configuring
12313 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
12314 instructions can be suppressed with the @option{-mno-cld} compiler option
12319 This option will enable GCC to use CMPXCHG16B instruction in generated code.
12320 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
12321 data types. This is useful for high resolution counters that could be updated
12322 by multiple processors (or cores). This instruction is generated as part of
12323 atomic built-in functions: see @ref{Atomic Builtins} for details.
12327 This option will enable GCC to use SAHF instruction in generated 64-bit code.
12328 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
12329 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
12330 SAHF are load and store instructions, respectively, for certain status flags.
12331 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
12332 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
12336 This option will enable GCC to use movbe instruction to implement
12337 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
12341 This option will enable built-in functions, @code{__builtin_ia32_crc32qi},
12342 @code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and
12343 @code{__builtin_ia32_crc32di} to generate the crc32 machine instruction.
12347 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
12348 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
12349 to increase precision instead of DIVSS and SQRTSS (and their vectorized
12350 variants) for single precision floating point arguments. These instructions
12351 are generated only when @option{-funsafe-math-optimizations} is enabled
12352 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
12353 Note that while the throughput of the sequence is higher than the throughput
12354 of the non-reciprocal instruction, the precision of the sequence can be
12355 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
12357 Note that GCC implements 1.0f/sqrtf(x) in terms of RSQRTSS (or RSQRTPS)
12358 already with @option{-ffast-math} (or the above option combination), and
12359 doesn't need @option{-mrecip}.
12361 @item -mveclibabi=@var{type}
12362 @opindex mveclibabi
12363 Specifies the ABI type to use for vectorizing intrinsics using an
12364 external library. Supported types are @code{svml} for the Intel short
12365 vector math library and @code{acml} for the AMD math core library style
12366 of interfacing. GCC will currently emit calls to @code{vmldExp2},
12367 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
12368 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
12369 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
12370 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
12371 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
12372 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
12373 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
12374 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
12375 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
12376 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
12377 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
12378 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
12379 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
12380 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
12381 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
12382 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
12383 compatible library will have to be specified at link time.
12385 @item -mabi=@var{name}
12387 Generate code for the specified calling convention. Permissible values
12388 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
12389 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
12390 ABI when targeting Windows. On all other systems, the default is the
12391 SYSV ABI. You can control this behavior for a specific function by
12392 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
12393 @xref{Function Attributes}.
12396 @itemx -mno-push-args
12397 @opindex mpush-args
12398 @opindex mno-push-args
12399 Use PUSH operations to store outgoing parameters. This method is shorter
12400 and usually equally fast as method using SUB/MOV operations and is enabled
12401 by default. In some cases disabling it may improve performance because of
12402 improved scheduling and reduced dependencies.
12404 @item -maccumulate-outgoing-args
12405 @opindex maccumulate-outgoing-args
12406 If enabled, the maximum amount of space required for outgoing arguments will be
12407 computed in the function prologue. This is faster on most modern CPUs
12408 because of reduced dependencies, improved scheduling and reduced stack usage
12409 when preferred stack boundary is not equal to 2. The drawback is a notable
12410 increase in code size. This switch implies @option{-mno-push-args}.
12414 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
12415 on thread-safe exception handling must compile and link all code with the
12416 @option{-mthreads} option. When compiling, @option{-mthreads} defines
12417 @option{-D_MT}; when linking, it links in a special thread helper library
12418 @option{-lmingwthrd} which cleans up per thread exception handling data.
12420 @item -mno-align-stringops
12421 @opindex mno-align-stringops
12422 Do not align destination of inlined string operations. This switch reduces
12423 code size and improves performance in case the destination is already aligned,
12424 but GCC doesn't know about it.
12426 @item -minline-all-stringops
12427 @opindex minline-all-stringops
12428 By default GCC inlines string operations only when destination is known to be
12429 aligned at least to 4 byte boundary. This enables more inlining, increase code
12430 size, but may improve performance of code that depends on fast memcpy, strlen
12431 and memset for short lengths.
12433 @item -minline-stringops-dynamically
12434 @opindex minline-stringops-dynamically
12435 For string operation of unknown size, inline runtime checks so for small
12436 blocks inline code is used, while for large blocks library call is used.
12438 @item -mstringop-strategy=@var{alg}
12439 @opindex mstringop-strategy=@var{alg}
12440 Overwrite internal decision heuristic about particular algorithm to inline
12441 string operation with. The allowed values are @code{rep_byte},
12442 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
12443 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
12444 expanding inline loop, @code{libcall} for always expanding library call.
12446 @item -momit-leaf-frame-pointer
12447 @opindex momit-leaf-frame-pointer
12448 Don't keep the frame pointer in a register for leaf functions. This
12449 avoids the instructions to save, set up and restore frame pointers and
12450 makes an extra register available in leaf functions. The option
12451 @option{-fomit-frame-pointer} removes the frame pointer for all functions
12452 which might make debugging harder.
12454 @item -mtls-direct-seg-refs
12455 @itemx -mno-tls-direct-seg-refs
12456 @opindex mtls-direct-seg-refs
12457 Controls whether TLS variables may be accessed with offsets from the
12458 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
12459 or whether the thread base pointer must be added. Whether or not this
12460 is legal depends on the operating system, and whether it maps the
12461 segment to cover the entire TLS area.
12463 For systems that use GNU libc, the default is on.
12466 @itemx -mno-sse2avx
12468 Specify that the assembler should encode SSE instructions with VEX
12469 prefix. The option @option{-mavx} turns this on by default.
12472 These @samp{-m} switches are supported in addition to the above
12473 on AMD x86-64 processors in 64-bit environments.
12480 Generate code for a 32-bit or 64-bit environment.
12481 The 32-bit environment sets int, long and pointer to 32 bits and
12482 generates code that runs on any i386 system.
12483 The 64-bit environment sets int to 32 bits and long and pointer
12484 to 64 bits and generates code for AMD's x86-64 architecture. For
12485 darwin only the -m64 option turns off the @option{-fno-pic} and
12486 @option{-mdynamic-no-pic} options.
12488 @item -mno-red-zone
12489 @opindex mno-red-zone
12490 Do not use a so called red zone for x86-64 code. The red zone is mandated
12491 by the x86-64 ABI, it is a 128-byte area beyond the location of the
12492 stack pointer that will not be modified by signal or interrupt handlers
12493 and therefore can be used for temporary data without adjusting the stack
12494 pointer. The flag @option{-mno-red-zone} disables this red zone.
12496 @item -mcmodel=small
12497 @opindex mcmodel=small
12498 Generate code for the small code model: the program and its symbols must
12499 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
12500 Programs can be statically or dynamically linked. This is the default
12503 @item -mcmodel=kernel
12504 @opindex mcmodel=kernel
12505 Generate code for the kernel code model. The kernel runs in the
12506 negative 2 GB of the address space.
12507 This model has to be used for Linux kernel code.
12509 @item -mcmodel=medium
12510 @opindex mcmodel=medium
12511 Generate code for the medium model: The program is linked in the lower 2
12512 GB of the address space. Small symbols are also placed there. Symbols
12513 with sizes larger than @option{-mlarge-data-threshold} are put into
12514 large data or bss sections and can be located above 2GB. Programs can
12515 be statically or dynamically linked.
12517 @item -mcmodel=large
12518 @opindex mcmodel=large
12519 Generate code for the large model: This model makes no assumptions
12520 about addresses and sizes of sections.
12523 @node IA-64 Options
12524 @subsection IA-64 Options
12525 @cindex IA-64 Options
12527 These are the @samp{-m} options defined for the Intel IA-64 architecture.
12531 @opindex mbig-endian
12532 Generate code for a big endian target. This is the default for HP-UX@.
12534 @item -mlittle-endian
12535 @opindex mlittle-endian
12536 Generate code for a little endian target. This is the default for AIX5
12542 @opindex mno-gnu-as
12543 Generate (or don't) code for the GNU assembler. This is the default.
12544 @c Also, this is the default if the configure option @option{--with-gnu-as}
12550 @opindex mno-gnu-ld
12551 Generate (or don't) code for the GNU linker. This is the default.
12552 @c Also, this is the default if the configure option @option{--with-gnu-ld}
12557 Generate code that does not use a global pointer register. The result
12558 is not position independent code, and violates the IA-64 ABI@.
12560 @item -mvolatile-asm-stop
12561 @itemx -mno-volatile-asm-stop
12562 @opindex mvolatile-asm-stop
12563 @opindex mno-volatile-asm-stop
12564 Generate (or don't) a stop bit immediately before and after volatile asm
12567 @item -mregister-names
12568 @itemx -mno-register-names
12569 @opindex mregister-names
12570 @opindex mno-register-names
12571 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
12572 the stacked registers. This may make assembler output more readable.
12578 Disable (or enable) optimizations that use the small data section. This may
12579 be useful for working around optimizer bugs.
12581 @item -mconstant-gp
12582 @opindex mconstant-gp
12583 Generate code that uses a single constant global pointer value. This is
12584 useful when compiling kernel code.
12588 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
12589 This is useful when compiling firmware code.
12591 @item -minline-float-divide-min-latency
12592 @opindex minline-float-divide-min-latency
12593 Generate code for inline divides of floating point values
12594 using the minimum latency algorithm.
12596 @item -minline-float-divide-max-throughput
12597 @opindex minline-float-divide-max-throughput
12598 Generate code for inline divides of floating point values
12599 using the maximum throughput algorithm.
12601 @item -mno-inline-float-divide
12602 @opindex mno-inline-float-divide
12603 Do not generate inline code for divides of floating point values.
12605 @item -minline-int-divide-min-latency
12606 @opindex minline-int-divide-min-latency
12607 Generate code for inline divides of integer values
12608 using the minimum latency algorithm.
12610 @item -minline-int-divide-max-throughput
12611 @opindex minline-int-divide-max-throughput
12612 Generate code for inline divides of integer values
12613 using the maximum throughput algorithm.
12615 @item -mno-inline-int-divide
12616 @opindex mno-inline-int-divide
12617 Do not generate inline code for divides of integer values.
12619 @item -minline-sqrt-min-latency
12620 @opindex minline-sqrt-min-latency
12621 Generate code for inline square roots
12622 using the minimum latency algorithm.
12624 @item -minline-sqrt-max-throughput
12625 @opindex minline-sqrt-max-throughput
12626 Generate code for inline square roots
12627 using the maximum throughput algorithm.
12629 @item -mno-inline-sqrt
12630 @opindex mno-inline-sqrt
12631 Do not generate inline code for sqrt.
12634 @itemx -mno-fused-madd
12635 @opindex mfused-madd
12636 @opindex mno-fused-madd
12637 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
12638 instructions. The default is to use these instructions.
12640 @item -mno-dwarf2-asm
12641 @itemx -mdwarf2-asm
12642 @opindex mno-dwarf2-asm
12643 @opindex mdwarf2-asm
12644 Don't (or do) generate assembler code for the DWARF2 line number debugging
12645 info. This may be useful when not using the GNU assembler.
12647 @item -mearly-stop-bits
12648 @itemx -mno-early-stop-bits
12649 @opindex mearly-stop-bits
12650 @opindex mno-early-stop-bits
12651 Allow stop bits to be placed earlier than immediately preceding the
12652 instruction that triggered the stop bit. This can improve instruction
12653 scheduling, but does not always do so.
12655 @item -mfixed-range=@var{register-range}
12656 @opindex mfixed-range
12657 Generate code treating the given register range as fixed registers.
12658 A fixed register is one that the register allocator can not use. This is
12659 useful when compiling kernel code. A register range is specified as
12660 two registers separated by a dash. Multiple register ranges can be
12661 specified separated by a comma.
12663 @item -mtls-size=@var{tls-size}
12665 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
12668 @item -mtune=@var{cpu-type}
12670 Tune the instruction scheduling for a particular CPU, Valid values are
12671 itanium, itanium1, merced, itanium2, and mckinley.
12677 Generate code for a 32-bit or 64-bit environment.
12678 The 32-bit environment sets int, long and pointer to 32 bits.
12679 The 64-bit environment sets int to 32 bits and long and pointer
12680 to 64 bits. These are HP-UX specific flags.
12682 @item -mno-sched-br-data-spec
12683 @itemx -msched-br-data-spec
12684 @opindex mno-sched-br-data-spec
12685 @opindex msched-br-data-spec
12686 (Dis/En)able data speculative scheduling before reload.
12687 This will result in generation of the ld.a instructions and
12688 the corresponding check instructions (ld.c / chk.a).
12689 The default is 'disable'.
12691 @item -msched-ar-data-spec
12692 @itemx -mno-sched-ar-data-spec
12693 @opindex msched-ar-data-spec
12694 @opindex mno-sched-ar-data-spec
12695 (En/Dis)able data speculative scheduling after reload.
12696 This will result in generation of the ld.a instructions and
12697 the corresponding check instructions (ld.c / chk.a).
12698 The default is 'enable'.
12700 @item -mno-sched-control-spec
12701 @itemx -msched-control-spec
12702 @opindex mno-sched-control-spec
12703 @opindex msched-control-spec
12704 (Dis/En)able control speculative scheduling. This feature is
12705 available only during region scheduling (i.e.@: before reload).
12706 This will result in generation of the ld.s instructions and
12707 the corresponding check instructions chk.s .
12708 The default is 'disable'.
12710 @item -msched-br-in-data-spec
12711 @itemx -mno-sched-br-in-data-spec
12712 @opindex msched-br-in-data-spec
12713 @opindex mno-sched-br-in-data-spec
12714 (En/Dis)able speculative scheduling of the instructions that
12715 are dependent on the data speculative loads before reload.
12716 This is effective only with @option{-msched-br-data-spec} enabled.
12717 The default is 'enable'.
12719 @item -msched-ar-in-data-spec
12720 @itemx -mno-sched-ar-in-data-spec
12721 @opindex msched-ar-in-data-spec
12722 @opindex mno-sched-ar-in-data-spec
12723 (En/Dis)able speculative scheduling of the instructions that
12724 are dependent on the data speculative loads after reload.
12725 This is effective only with @option{-msched-ar-data-spec} enabled.
12726 The default is 'enable'.
12728 @item -msched-in-control-spec
12729 @itemx -mno-sched-in-control-spec
12730 @opindex msched-in-control-spec
12731 @opindex mno-sched-in-control-spec
12732 (En/Dis)able speculative scheduling of the instructions that
12733 are dependent on the control speculative loads.
12734 This is effective only with @option{-msched-control-spec} enabled.
12735 The default is 'enable'.
12737 @item -mno-sched-prefer-non-data-spec-insns
12738 @itemx -msched-prefer-non-data-spec-insns
12739 @opindex mno-sched-prefer-non-data-spec-insns
12740 @opindex msched-prefer-non-data-spec-insns
12741 If enabled, data speculative instructions will be chosen for schedule
12742 only if there are no other choices at the moment. This will make
12743 the use of the data speculation much more conservative.
12744 The default is 'disable'.
12746 @item -mno-sched-prefer-non-control-spec-insns
12747 @itemx -msched-prefer-non-control-spec-insns
12748 @opindex mno-sched-prefer-non-control-spec-insns
12749 @opindex msched-prefer-non-control-spec-insns
12750 If enabled, control speculative instructions will be chosen for schedule
12751 only if there are no other choices at the moment. This will make
12752 the use of the control speculation much more conservative.
12753 The default is 'disable'.
12755 @item -mno-sched-count-spec-in-critical-path
12756 @itemx -msched-count-spec-in-critical-path
12757 @opindex mno-sched-count-spec-in-critical-path
12758 @opindex msched-count-spec-in-critical-path
12759 If enabled, speculative dependencies will be considered during
12760 computation of the instructions priorities. This will make the use of the
12761 speculation a bit more conservative.
12762 The default is 'disable'.
12764 @item -msched-spec-ldc
12765 @opindex msched-spec-ldc
12766 Use a simple data speculation check. This option is on by default.
12768 @item -msched-control-spec-ldc
12769 @opindex msched-spec-ldc
12770 Use a simple check for control speculation. This option is on by default.
12772 @item -msched-stop-bits-after-every-cycle
12773 @opindex msched-stop-bits-after-every-cycle
12774 Place a stop bit after every cycle when scheduling. This option is on
12777 @item -msched-fp-mem-deps-zero-cost
12778 @opindex msched-fp-mem-deps-zero-cost
12779 Assume that floating-point stores and loads are not likely to cause a conflict
12780 when placed into the same instruction group. This option is disabled by
12783 @item -msel-sched-dont-check-control-spec
12784 @opindex msel-sched-dont-check-control-spec
12785 Generate checks for control speculation in selective scheduling.
12786 This flag is disabled by default.
12788 @item -msched-max-memory-insns=@var{max-insns}
12789 @opindex msched-max-memory-insns
12790 Limit on the number of memory insns per instruction group, giving lower
12791 priority to subsequent memory insns attempting to schedule in the same
12792 instruction group. Frequently useful to prevent cache bank conflicts.
12793 The default value is 1.
12795 @item -msched-max-memory-insns-hard-limit
12796 @opindex msched-max-memory-insns-hard-limit
12797 Disallow more than `msched-max-memory-insns' in instruction group.
12798 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
12799 when limit is reached but may still schedule memory operations.
12803 @node IA-64/VMS Options
12804 @subsection IA-64/VMS Options
12806 These @samp{-m} options are defined for the IA-64/VMS implementations:
12809 @item -mvms-return-codes
12810 @opindex mvms-return-codes
12811 Return VMS condition codes from main. The default is to return POSIX
12812 style condition (e.g.@ error) codes.
12814 @item -mdebug-main=@var{prefix}
12815 @opindex mdebug-main=@var{prefix}
12816 Flag the first routine whose name starts with @var{prefix} as the main
12817 routine for the debugger.
12821 Default to 64bit memory allocation routines.
12825 @subsection LM32 Options
12826 @cindex LM32 options
12828 These @option{-m} options are defined for the Lattice Mico32 architecture:
12831 @item -mbarrel-shift-enabled
12832 @opindex mbarrel-shift-enabled
12833 Enable barrel-shift instructions.
12835 @item -mdivide-enabled
12836 @opindex mdivide-enabled
12837 Enable divide and modulus instructions.
12839 @item -mmultiply-enabled
12840 @opindex multiply-enabled
12841 Enable multiply instructions.
12843 @item -msign-extend-enabled
12844 @opindex msign-extend-enabled
12845 Enable sign extend instructions.
12847 @item -muser-enabled
12848 @opindex muser-enabled
12849 Enable user-defined instructions.
12854 @subsection M32C Options
12855 @cindex M32C options
12858 @item -mcpu=@var{name}
12860 Select the CPU for which code is generated. @var{name} may be one of
12861 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
12862 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
12863 the M32C/80 series.
12867 Specifies that the program will be run on the simulator. This causes
12868 an alternate runtime library to be linked in which supports, for
12869 example, file I/O@. You must not use this option when generating
12870 programs that will run on real hardware; you must provide your own
12871 runtime library for whatever I/O functions are needed.
12873 @item -memregs=@var{number}
12875 Specifies the number of memory-based pseudo-registers GCC will use
12876 during code generation. These pseudo-registers will be used like real
12877 registers, so there is a tradeoff between GCC's ability to fit the
12878 code into available registers, and the performance penalty of using
12879 memory instead of registers. Note that all modules in a program must
12880 be compiled with the same value for this option. Because of that, you
12881 must not use this option with the default runtime libraries gcc
12886 @node M32R/D Options
12887 @subsection M32R/D Options
12888 @cindex M32R/D options
12890 These @option{-m} options are defined for Renesas M32R/D architectures:
12895 Generate code for the M32R/2@.
12899 Generate code for the M32R/X@.
12903 Generate code for the M32R@. This is the default.
12905 @item -mmodel=small
12906 @opindex mmodel=small
12907 Assume all objects live in the lower 16MB of memory (so that their addresses
12908 can be loaded with the @code{ld24} instruction), and assume all subroutines
12909 are reachable with the @code{bl} instruction.
12910 This is the default.
12912 The addressability of a particular object can be set with the
12913 @code{model} attribute.
12915 @item -mmodel=medium
12916 @opindex mmodel=medium
12917 Assume objects may be anywhere in the 32-bit address space (the compiler
12918 will generate @code{seth/add3} instructions to load their addresses), and
12919 assume all subroutines are reachable with the @code{bl} instruction.
12921 @item -mmodel=large
12922 @opindex mmodel=large
12923 Assume objects may be anywhere in the 32-bit address space (the compiler
12924 will generate @code{seth/add3} instructions to load their addresses), and
12925 assume subroutines may not be reachable with the @code{bl} instruction
12926 (the compiler will generate the much slower @code{seth/add3/jl}
12927 instruction sequence).
12930 @opindex msdata=none
12931 Disable use of the small data area. Variables will be put into
12932 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
12933 @code{section} attribute has been specified).
12934 This is the default.
12936 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
12937 Objects may be explicitly put in the small data area with the
12938 @code{section} attribute using one of these sections.
12940 @item -msdata=sdata
12941 @opindex msdata=sdata
12942 Put small global and static data in the small data area, but do not
12943 generate special code to reference them.
12946 @opindex msdata=use
12947 Put small global and static data in the small data area, and generate
12948 special instructions to reference them.
12952 @cindex smaller data references
12953 Put global and static objects less than or equal to @var{num} bytes
12954 into the small data or bss sections instead of the normal data or bss
12955 sections. The default value of @var{num} is 8.
12956 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
12957 for this option to have any effect.
12959 All modules should be compiled with the same @option{-G @var{num}} value.
12960 Compiling with different values of @var{num} may or may not work; if it
12961 doesn't the linker will give an error message---incorrect code will not be
12966 Makes the M32R specific code in the compiler display some statistics
12967 that might help in debugging programs.
12969 @item -malign-loops
12970 @opindex malign-loops
12971 Align all loops to a 32-byte boundary.
12973 @item -mno-align-loops
12974 @opindex mno-align-loops
12975 Do not enforce a 32-byte alignment for loops. This is the default.
12977 @item -missue-rate=@var{number}
12978 @opindex missue-rate=@var{number}
12979 Issue @var{number} instructions per cycle. @var{number} can only be 1
12982 @item -mbranch-cost=@var{number}
12983 @opindex mbranch-cost=@var{number}
12984 @var{number} can only be 1 or 2. If it is 1 then branches will be
12985 preferred over conditional code, if it is 2, then the opposite will
12988 @item -mflush-trap=@var{number}
12989 @opindex mflush-trap=@var{number}
12990 Specifies the trap number to use to flush the cache. The default is
12991 12. Valid numbers are between 0 and 15 inclusive.
12993 @item -mno-flush-trap
12994 @opindex mno-flush-trap
12995 Specifies that the cache cannot be flushed by using a trap.
12997 @item -mflush-func=@var{name}
12998 @opindex mflush-func=@var{name}
12999 Specifies the name of the operating system function to call to flush
13000 the cache. The default is @emph{_flush_cache}, but a function call
13001 will only be used if a trap is not available.
13003 @item -mno-flush-func
13004 @opindex mno-flush-func
13005 Indicates that there is no OS function for flushing the cache.
13009 @node M680x0 Options
13010 @subsection M680x0 Options
13011 @cindex M680x0 options
13013 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
13014 The default settings depend on which architecture was selected when
13015 the compiler was configured; the defaults for the most common choices
13019 @item -march=@var{arch}
13021 Generate code for a specific M680x0 or ColdFire instruction set
13022 architecture. Permissible values of @var{arch} for M680x0
13023 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
13024 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
13025 architectures are selected according to Freescale's ISA classification
13026 and the permissible values are: @samp{isaa}, @samp{isaaplus},
13027 @samp{isab} and @samp{isac}.
13029 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
13030 code for a ColdFire target. The @var{arch} in this macro is one of the
13031 @option{-march} arguments given above.
13033 When used together, @option{-march} and @option{-mtune} select code
13034 that runs on a family of similar processors but that is optimized
13035 for a particular microarchitecture.
13037 @item -mcpu=@var{cpu}
13039 Generate code for a specific M680x0 or ColdFire processor.
13040 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
13041 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
13042 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
13043 below, which also classifies the CPUs into families:
13045 @multitable @columnfractions 0.20 0.80
13046 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
13047 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
13048 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
13049 @item @samp{5206e} @tab @samp{5206e}
13050 @item @samp{5208} @tab @samp{5207} @samp{5208}
13051 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
13052 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
13053 @item @samp{5216} @tab @samp{5214} @samp{5216}
13054 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
13055 @item @samp{5225} @tab @samp{5224} @samp{5225}
13056 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
13057 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
13058 @item @samp{5249} @tab @samp{5249}
13059 @item @samp{5250} @tab @samp{5250}
13060 @item @samp{5271} @tab @samp{5270} @samp{5271}
13061 @item @samp{5272} @tab @samp{5272}
13062 @item @samp{5275} @tab @samp{5274} @samp{5275}
13063 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
13064 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
13065 @item @samp{5307} @tab @samp{5307}
13066 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
13067 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
13068 @item @samp{5407} @tab @samp{5407}
13069 @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}
13072 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
13073 @var{arch} is compatible with @var{cpu}. Other combinations of
13074 @option{-mcpu} and @option{-march} are rejected.
13076 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
13077 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
13078 where the value of @var{family} is given by the table above.
13080 @item -mtune=@var{tune}
13082 Tune the code for a particular microarchitecture, within the
13083 constraints set by @option{-march} and @option{-mcpu}.
13084 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
13085 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
13086 and @samp{cpu32}. The ColdFire microarchitectures
13087 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
13089 You can also use @option{-mtune=68020-40} for code that needs
13090 to run relatively well on 68020, 68030 and 68040 targets.
13091 @option{-mtune=68020-60} is similar but includes 68060 targets
13092 as well. These two options select the same tuning decisions as
13093 @option{-m68020-40} and @option{-m68020-60} respectively.
13095 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
13096 when tuning for 680x0 architecture @var{arch}. It also defines
13097 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
13098 option is used. If gcc is tuning for a range of architectures,
13099 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
13100 it defines the macros for every architecture in the range.
13102 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
13103 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
13104 of the arguments given above.
13110 Generate output for a 68000. This is the default
13111 when the compiler is configured for 68000-based systems.
13112 It is equivalent to @option{-march=68000}.
13114 Use this option for microcontrollers with a 68000 or EC000 core,
13115 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
13119 Generate output for a 68010. This is the default
13120 when the compiler is configured for 68010-based systems.
13121 It is equivalent to @option{-march=68010}.
13127 Generate output for a 68020. This is the default
13128 when the compiler is configured for 68020-based systems.
13129 It is equivalent to @option{-march=68020}.
13133 Generate output for a 68030. This is the default when the compiler is
13134 configured for 68030-based systems. It is equivalent to
13135 @option{-march=68030}.
13139 Generate output for a 68040. This is the default when the compiler is
13140 configured for 68040-based systems. It is equivalent to
13141 @option{-march=68040}.
13143 This option inhibits the use of 68881/68882 instructions that have to be
13144 emulated by software on the 68040. Use this option if your 68040 does not
13145 have code to emulate those instructions.
13149 Generate output for a 68060. This is the default when the compiler is
13150 configured for 68060-based systems. It is equivalent to
13151 @option{-march=68060}.
13153 This option inhibits the use of 68020 and 68881/68882 instructions that
13154 have to be emulated by software on the 68060. Use this option if your 68060
13155 does not have code to emulate those instructions.
13159 Generate output for a CPU32. This is the default
13160 when the compiler is configured for CPU32-based systems.
13161 It is equivalent to @option{-march=cpu32}.
13163 Use this option for microcontrollers with a
13164 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
13165 68336, 68340, 68341, 68349 and 68360.
13169 Generate output for a 520X ColdFire CPU@. This is the default
13170 when the compiler is configured for 520X-based systems.
13171 It is equivalent to @option{-mcpu=5206}, and is now deprecated
13172 in favor of that option.
13174 Use this option for microcontroller with a 5200 core, including
13175 the MCF5202, MCF5203, MCF5204 and MCF5206.
13179 Generate output for a 5206e ColdFire CPU@. The option is now
13180 deprecated in favor of the equivalent @option{-mcpu=5206e}.
13184 Generate output for a member of the ColdFire 528X family.
13185 The option is now deprecated in favor of the equivalent
13186 @option{-mcpu=528x}.
13190 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
13191 in favor of the equivalent @option{-mcpu=5307}.
13195 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
13196 in favor of the equivalent @option{-mcpu=5407}.
13200 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
13201 This includes use of hardware floating point instructions.
13202 The option is equivalent to @option{-mcpu=547x}, and is now
13203 deprecated in favor of that option.
13207 Generate output for a 68040, without using any of the new instructions.
13208 This results in code which can run relatively efficiently on either a
13209 68020/68881 or a 68030 or a 68040. The generated code does use the
13210 68881 instructions that are emulated on the 68040.
13212 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
13216 Generate output for a 68060, without using any of the new instructions.
13217 This results in code which can run relatively efficiently on either a
13218 68020/68881 or a 68030 or a 68040. The generated code does use the
13219 68881 instructions that are emulated on the 68060.
13221 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
13225 @opindex mhard-float
13227 Generate floating-point instructions. This is the default for 68020
13228 and above, and for ColdFire devices that have an FPU@. It defines the
13229 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
13230 on ColdFire targets.
13233 @opindex msoft-float
13234 Do not generate floating-point instructions; use library calls instead.
13235 This is the default for 68000, 68010, and 68832 targets. It is also
13236 the default for ColdFire devices that have no FPU.
13242 Generate (do not generate) ColdFire hardware divide and remainder
13243 instructions. If @option{-march} is used without @option{-mcpu},
13244 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
13245 architectures. Otherwise, the default is taken from the target CPU
13246 (either the default CPU, or the one specified by @option{-mcpu}). For
13247 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
13248 @option{-mcpu=5206e}.
13250 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
13254 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13255 Additionally, parameters passed on the stack are also aligned to a
13256 16-bit boundary even on targets whose API mandates promotion to 32-bit.
13260 Do not consider type @code{int} to be 16 bits wide. This is the default.
13263 @itemx -mno-bitfield
13264 @opindex mnobitfield
13265 @opindex mno-bitfield
13266 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
13267 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
13271 Do use the bit-field instructions. The @option{-m68020} option implies
13272 @option{-mbitfield}. This is the default if you use a configuration
13273 designed for a 68020.
13277 Use a different function-calling convention, in which functions
13278 that take a fixed number of arguments return with the @code{rtd}
13279 instruction, which pops their arguments while returning. This
13280 saves one instruction in the caller since there is no need to pop
13281 the arguments there.
13283 This calling convention is incompatible with the one normally
13284 used on Unix, so you cannot use it if you need to call libraries
13285 compiled with the Unix compiler.
13287 Also, you must provide function prototypes for all functions that
13288 take variable numbers of arguments (including @code{printf});
13289 otherwise incorrect code will be generated for calls to those
13292 In addition, seriously incorrect code will result if you call a
13293 function with too many arguments. (Normally, extra arguments are
13294 harmlessly ignored.)
13296 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
13297 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
13301 Do not use the calling conventions selected by @option{-mrtd}.
13302 This is the default.
13305 @itemx -mno-align-int
13306 @opindex malign-int
13307 @opindex mno-align-int
13308 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
13309 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
13310 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
13311 Aligning variables on 32-bit boundaries produces code that runs somewhat
13312 faster on processors with 32-bit busses at the expense of more memory.
13314 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
13315 align structures containing the above types differently than
13316 most published application binary interface specifications for the m68k.
13320 Use the pc-relative addressing mode of the 68000 directly, instead of
13321 using a global offset table. At present, this option implies @option{-fpic},
13322 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
13323 not presently supported with @option{-mpcrel}, though this could be supported for
13324 68020 and higher processors.
13326 @item -mno-strict-align
13327 @itemx -mstrict-align
13328 @opindex mno-strict-align
13329 @opindex mstrict-align
13330 Do not (do) assume that unaligned memory references will be handled by
13334 Generate code that allows the data segment to be located in a different
13335 area of memory from the text segment. This allows for execute in place in
13336 an environment without virtual memory management. This option implies
13339 @item -mno-sep-data
13340 Generate code that assumes that the data segment follows the text segment.
13341 This is the default.
13343 @item -mid-shared-library
13344 Generate code that supports shared libraries via the library ID method.
13345 This allows for execute in place and shared libraries in an environment
13346 without virtual memory management. This option implies @option{-fPIC}.
13348 @item -mno-id-shared-library
13349 Generate code that doesn't assume ID based shared libraries are being used.
13350 This is the default.
13352 @item -mshared-library-id=n
13353 Specified the identification number of the ID based shared library being
13354 compiled. Specifying a value of 0 will generate more compact code, specifying
13355 other values will force the allocation of that number to the current
13356 library but is no more space or time efficient than omitting this option.
13362 When generating position-independent code for ColdFire, generate code
13363 that works if the GOT has more than 8192 entries. This code is
13364 larger and slower than code generated without this option. On M680x0
13365 processors, this option is not needed; @option{-fPIC} suffices.
13367 GCC normally uses a single instruction to load values from the GOT@.
13368 While this is relatively efficient, it only works if the GOT
13369 is smaller than about 64k. Anything larger causes the linker
13370 to report an error such as:
13372 @cindex relocation truncated to fit (ColdFire)
13374 relocation truncated to fit: R_68K_GOT16O foobar
13377 If this happens, you should recompile your code with @option{-mxgot}.
13378 It should then work with very large GOTs. However, code generated with
13379 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
13380 the value of a global symbol.
13382 Note that some linkers, including newer versions of the GNU linker,
13383 can create multiple GOTs and sort GOT entries. If you have such a linker,
13384 you should only need to use @option{-mxgot} when compiling a single
13385 object file that accesses more than 8192 GOT entries. Very few do.
13387 These options have no effect unless GCC is generating
13388 position-independent code.
13392 @node M68hc1x Options
13393 @subsection M68hc1x Options
13394 @cindex M68hc1x options
13396 These are the @samp{-m} options defined for the 68hc11 and 68hc12
13397 microcontrollers. The default values for these options depends on
13398 which style of microcontroller was selected when the compiler was configured;
13399 the defaults for the most common choices are given below.
13406 Generate output for a 68HC11. This is the default
13407 when the compiler is configured for 68HC11-based systems.
13413 Generate output for a 68HC12. This is the default
13414 when the compiler is configured for 68HC12-based systems.
13420 Generate output for a 68HCS12.
13422 @item -mauto-incdec
13423 @opindex mauto-incdec
13424 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
13431 Enable the use of 68HC12 min and max instructions.
13434 @itemx -mno-long-calls
13435 @opindex mlong-calls
13436 @opindex mno-long-calls
13437 Treat all calls as being far away (near). If calls are assumed to be
13438 far away, the compiler will use the @code{call} instruction to
13439 call a function and the @code{rtc} instruction for returning.
13443 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13445 @item -msoft-reg-count=@var{count}
13446 @opindex msoft-reg-count
13447 Specify the number of pseudo-soft registers which are used for the
13448 code generation. The maximum number is 32. Using more pseudo-soft
13449 register may or may not result in better code depending on the program.
13450 The default is 4 for 68HC11 and 2 for 68HC12.
13454 @node MCore Options
13455 @subsection MCore Options
13456 @cindex MCore options
13458 These are the @samp{-m} options defined for the Motorola M*Core
13464 @itemx -mno-hardlit
13466 @opindex mno-hardlit
13467 Inline constants into the code stream if it can be done in two
13468 instructions or less.
13474 Use the divide instruction. (Enabled by default).
13476 @item -mrelax-immediate
13477 @itemx -mno-relax-immediate
13478 @opindex mrelax-immediate
13479 @opindex mno-relax-immediate
13480 Allow arbitrary sized immediates in bit operations.
13482 @item -mwide-bitfields
13483 @itemx -mno-wide-bitfields
13484 @opindex mwide-bitfields
13485 @opindex mno-wide-bitfields
13486 Always treat bit-fields as int-sized.
13488 @item -m4byte-functions
13489 @itemx -mno-4byte-functions
13490 @opindex m4byte-functions
13491 @opindex mno-4byte-functions
13492 Force all functions to be aligned to a four byte boundary.
13494 @item -mcallgraph-data
13495 @itemx -mno-callgraph-data
13496 @opindex mcallgraph-data
13497 @opindex mno-callgraph-data
13498 Emit callgraph information.
13501 @itemx -mno-slow-bytes
13502 @opindex mslow-bytes
13503 @opindex mno-slow-bytes
13504 Prefer word access when reading byte quantities.
13506 @item -mlittle-endian
13507 @itemx -mbig-endian
13508 @opindex mlittle-endian
13509 @opindex mbig-endian
13510 Generate code for a little endian target.
13516 Generate code for the 210 processor.
13520 Assume that run-time support has been provided and so omit the
13521 simulator library (@file{libsim.a)} from the linker command line.
13523 @item -mstack-increment=@var{size}
13524 @opindex mstack-increment
13525 Set the maximum amount for a single stack increment operation. Large
13526 values can increase the speed of programs which contain functions
13527 that need a large amount of stack space, but they can also trigger a
13528 segmentation fault if the stack is extended too much. The default
13534 @subsection MeP Options
13535 @cindex MeP options
13541 Enables the @code{abs} instruction, which is the absolute difference
13542 between two registers.
13546 Enables all the optional instructions - average, multiply, divide, bit
13547 operations, leading zero, absolute difference, min/max, clip, and
13553 Enables the @code{ave} instruction, which computes the average of two
13556 @item -mbased=@var{n}
13558 Variables of size @var{n} bytes or smaller will be placed in the
13559 @code{.based} section by default. Based variables use the @code{$tp}
13560 register as a base register, and there is a 128 byte limit to the
13561 @code{.based} section.
13565 Enables the bit operation instructions - bit test (@code{btstm}), set
13566 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
13567 test-and-set (@code{tas}).
13569 @item -mc=@var{name}
13571 Selects which section constant data will be placed in. @var{name} may
13572 be @code{tiny}, @code{near}, or @code{far}.
13576 Enables the @code{clip} instruction. Note that @code{-mclip} is not
13577 useful unless you also provide @code{-mminmax}.
13579 @item -mconfig=@var{name}
13581 Selects one of the build-in core configurations. Each MeP chip has
13582 one or more modules in it; each module has a core CPU and a variety of
13583 coprocessors, optional instructions, and peripherals. The
13584 @code{MeP-Integrator} tool, not part of GCC, provides these
13585 configurations through this option; using this option is the same as
13586 using all the corresponding command line options. The default
13587 configuration is @code{default}.
13591 Enables the coprocessor instructions. By default, this is a 32-bit
13592 coprocessor. Note that the coprocessor is normally enabled via the
13593 @code{-mconfig=} option.
13597 Enables the 32-bit coprocessor's instructions.
13601 Enables the 64-bit coprocessor's instructions.
13605 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
13609 Causes constant variables to be placed in the @code{.near} section.
13613 Enables the @code{div} and @code{divu} instructions.
13617 Generate big-endian code.
13621 Generate little-endian code.
13623 @item -mio-volatile
13624 @opindex mio-volatile
13625 Tells the compiler that any variable marked with the @code{io}
13626 attribute is to be considered volatile.
13630 Causes variables to be assigned to the @code{.far} section by default.
13634 Enables the @code{leadz} (leading zero) instruction.
13638 Causes variables to be assigned to the @code{.near} section by default.
13642 Enables the @code{min} and @code{max} instructions.
13646 Enables the multiplication and multiply-accumulate instructions.
13650 Disables all the optional instructions enabled by @code{-mall-opts}.
13654 Enables the @code{repeat} and @code{erepeat} instructions, used for
13655 low-overhead looping.
13659 Causes all variables to default to the @code{.tiny} section. Note
13660 that there is a 65536 byte limit to this section. Accesses to these
13661 variables use the @code{%gp} base register.
13665 Enables the saturation instructions. Note that the compiler does not
13666 currently generate these itself, but this option is included for
13667 compatibility with other tools, like @code{as}.
13671 Link the SDRAM-based runtime instead of the default ROM-based runtime.
13675 Link the simulator runtime libraries.
13679 Link the simulator runtime libraries, excluding built-in support
13680 for reset and exception vectors and tables.
13684 Causes all functions to default to the @code{.far} section. Without
13685 this option, functions default to the @code{.near} section.
13687 @item -mtiny=@var{n}
13689 Variables that are @var{n} bytes or smaller will be allocated to the
13690 @code{.tiny} section. These variables use the @code{$gp} base
13691 register. The default for this option is 4, but note that there's a
13692 65536 byte limit to the @code{.tiny} section.
13697 @subsection MIPS Options
13698 @cindex MIPS options
13704 Generate big-endian code.
13708 Generate little-endian code. This is the default for @samp{mips*el-*-*}
13711 @item -march=@var{arch}
13713 Generate code that will run on @var{arch}, which can be the name of a
13714 generic MIPS ISA, or the name of a particular processor.
13716 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
13717 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
13718 The processor names are:
13719 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
13720 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
13721 @samp{5kc}, @samp{5kf},
13723 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
13724 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
13725 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
13726 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
13727 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
13728 @samp{loongson2e}, @samp{loongson2f},
13732 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
13733 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
13734 @samp{rm7000}, @samp{rm9000},
13735 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
13738 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
13739 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
13741 The special value @samp{from-abi} selects the
13742 most compatible architecture for the selected ABI (that is,
13743 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
13745 Native Linux/GNU toolchains also support the value @samp{native},
13746 which selects the best architecture option for the host processor.
13747 @option{-march=native} has no effect if GCC does not recognize
13750 In processor names, a final @samp{000} can be abbreviated as @samp{k}
13751 (for example, @samp{-march=r2k}). Prefixes are optional, and
13752 @samp{vr} may be written @samp{r}.
13754 Names of the form @samp{@var{n}f2_1} refer to processors with
13755 FPUs clocked at half the rate of the core, names of the form
13756 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
13757 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
13758 processors with FPUs clocked a ratio of 3:2 with respect to the core.
13759 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
13760 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
13761 accepted as synonyms for @samp{@var{n}f1_1}.
13763 GCC defines two macros based on the value of this option. The first
13764 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
13765 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
13766 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
13767 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
13768 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
13770 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
13771 above. In other words, it will have the full prefix and will not
13772 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
13773 the macro names the resolved architecture (either @samp{"mips1"} or
13774 @samp{"mips3"}). It names the default architecture when no
13775 @option{-march} option is given.
13777 @item -mtune=@var{arch}
13779 Optimize for @var{arch}. Among other things, this option controls
13780 the way instructions are scheduled, and the perceived cost of arithmetic
13781 operations. The list of @var{arch} values is the same as for
13784 When this option is not used, GCC will optimize for the processor
13785 specified by @option{-march}. By using @option{-march} and
13786 @option{-mtune} together, it is possible to generate code that will
13787 run on a family of processors, but optimize the code for one
13788 particular member of that family.
13790 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
13791 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
13792 @samp{-march} ones described above.
13796 Equivalent to @samp{-march=mips1}.
13800 Equivalent to @samp{-march=mips2}.
13804 Equivalent to @samp{-march=mips3}.
13808 Equivalent to @samp{-march=mips4}.
13812 Equivalent to @samp{-march=mips32}.
13816 Equivalent to @samp{-march=mips32r2}.
13820 Equivalent to @samp{-march=mips64}.
13824 Equivalent to @samp{-march=mips64r2}.
13829 @opindex mno-mips16
13830 Generate (do not generate) MIPS16 code. If GCC is targetting a
13831 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
13833 MIPS16 code generation can also be controlled on a per-function basis
13834 by means of @code{mips16} and @code{nomips16} attributes.
13835 @xref{Function Attributes}, for more information.
13837 @item -mflip-mips16
13838 @opindex mflip-mips16
13839 Generate MIPS16 code on alternating functions. This option is provided
13840 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
13841 not intended for ordinary use in compiling user code.
13843 @item -minterlink-mips16
13844 @itemx -mno-interlink-mips16
13845 @opindex minterlink-mips16
13846 @opindex mno-interlink-mips16
13847 Require (do not require) that non-MIPS16 code be link-compatible with
13850 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
13851 it must either use a call or an indirect jump. @option{-minterlink-mips16}
13852 therefore disables direct jumps unless GCC knows that the target of the
13853 jump is not MIPS16.
13865 Generate code for the given ABI@.
13867 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
13868 generates 64-bit code when you select a 64-bit architecture, but you
13869 can use @option{-mgp32} to get 32-bit code instead.
13871 For information about the O64 ABI, see
13872 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
13874 GCC supports a variant of the o32 ABI in which floating-point registers
13875 are 64 rather than 32 bits wide. You can select this combination with
13876 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
13877 and @samp{mfhc1} instructions and is therefore only supported for
13878 MIPS32R2 processors.
13880 The register assignments for arguments and return values remain the
13881 same, but each scalar value is passed in a single 64-bit register
13882 rather than a pair of 32-bit registers. For example, scalar
13883 floating-point values are returned in @samp{$f0} only, not a
13884 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
13885 remains the same, but all 64 bits are saved.
13888 @itemx -mno-abicalls
13890 @opindex mno-abicalls
13891 Generate (do not generate) code that is suitable for SVR4-style
13892 dynamic objects. @option{-mabicalls} is the default for SVR4-based
13897 Generate (do not generate) code that is fully position-independent,
13898 and that can therefore be linked into shared libraries. This option
13899 only affects @option{-mabicalls}.
13901 All @option{-mabicalls} code has traditionally been position-independent,
13902 regardless of options like @option{-fPIC} and @option{-fpic}. However,
13903 as an extension, the GNU toolchain allows executables to use absolute
13904 accesses for locally-binding symbols. It can also use shorter GP
13905 initialization sequences and generate direct calls to locally-defined
13906 functions. This mode is selected by @option{-mno-shared}.
13908 @option{-mno-shared} depends on binutils 2.16 or higher and generates
13909 objects that can only be linked by the GNU linker. However, the option
13910 does not affect the ABI of the final executable; it only affects the ABI
13911 of relocatable objects. Using @option{-mno-shared} will generally make
13912 executables both smaller and quicker.
13914 @option{-mshared} is the default.
13920 Assume (do not assume) that the static and dynamic linkers
13921 support PLTs and copy relocations. This option only affects
13922 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
13923 has no effect without @samp{-msym32}.
13925 You can make @option{-mplt} the default by configuring
13926 GCC with @option{--with-mips-plt}. The default is
13927 @option{-mno-plt} otherwise.
13933 Lift (do not lift) the usual restrictions on the size of the global
13936 GCC normally uses a single instruction to load values from the GOT@.
13937 While this is relatively efficient, it will only work if the GOT
13938 is smaller than about 64k. Anything larger will cause the linker
13939 to report an error such as:
13941 @cindex relocation truncated to fit (MIPS)
13943 relocation truncated to fit: R_MIPS_GOT16 foobar
13946 If this happens, you should recompile your code with @option{-mxgot}.
13947 It should then work with very large GOTs, although it will also be
13948 less efficient, since it will take three instructions to fetch the
13949 value of a global symbol.
13951 Note that some linkers can create multiple GOTs. If you have such a
13952 linker, you should only need to use @option{-mxgot} when a single object
13953 file accesses more than 64k's worth of GOT entries. Very few do.
13955 These options have no effect unless GCC is generating position
13960 Assume that general-purpose registers are 32 bits wide.
13964 Assume that general-purpose registers are 64 bits wide.
13968 Assume that floating-point registers are 32 bits wide.
13972 Assume that floating-point registers are 64 bits wide.
13975 @opindex mhard-float
13976 Use floating-point coprocessor instructions.
13979 @opindex msoft-float
13980 Do not use floating-point coprocessor instructions. Implement
13981 floating-point calculations using library calls instead.
13983 @item -msingle-float
13984 @opindex msingle-float
13985 Assume that the floating-point coprocessor only supports single-precision
13988 @item -mdouble-float
13989 @opindex mdouble-float
13990 Assume that the floating-point coprocessor supports double-precision
13991 operations. This is the default.
13997 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
13998 implement atomic memory built-in functions. When neither option is
13999 specified, GCC will use the instructions if the target architecture
14002 @option{-mllsc} is useful if the runtime environment can emulate the
14003 instructions and @option{-mno-llsc} can be useful when compiling for
14004 nonstandard ISAs. You can make either option the default by
14005 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
14006 respectively. @option{--with-llsc} is the default for some
14007 configurations; see the installation documentation for details.
14013 Use (do not use) revision 1 of the MIPS DSP ASE@.
14014 @xref{MIPS DSP Built-in Functions}. This option defines the
14015 preprocessor macro @samp{__mips_dsp}. It also defines
14016 @samp{__mips_dsp_rev} to 1.
14022 Use (do not use) revision 2 of the MIPS DSP ASE@.
14023 @xref{MIPS DSP Built-in Functions}. This option defines the
14024 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
14025 It also defines @samp{__mips_dsp_rev} to 2.
14028 @itemx -mno-smartmips
14029 @opindex msmartmips
14030 @opindex mno-smartmips
14031 Use (do not use) the MIPS SmartMIPS ASE.
14033 @item -mpaired-single
14034 @itemx -mno-paired-single
14035 @opindex mpaired-single
14036 @opindex mno-paired-single
14037 Use (do not use) paired-single floating-point instructions.
14038 @xref{MIPS Paired-Single Support}. This option requires
14039 hardware floating-point support to be enabled.
14045 Use (do not use) MIPS Digital Media Extension instructions.
14046 This option can only be used when generating 64-bit code and requires
14047 hardware floating-point support to be enabled.
14052 @opindex mno-mips3d
14053 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
14054 The option @option{-mips3d} implies @option{-mpaired-single}.
14060 Use (do not use) MT Multithreading instructions.
14064 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
14065 an explanation of the default and the way that the pointer size is
14070 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
14072 The default size of @code{int}s, @code{long}s and pointers depends on
14073 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
14074 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
14075 32-bit @code{long}s. Pointers are the same size as @code{long}s,
14076 or the same size as integer registers, whichever is smaller.
14082 Assume (do not assume) that all symbols have 32-bit values, regardless
14083 of the selected ABI@. This option is useful in combination with
14084 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
14085 to generate shorter and faster references to symbolic addresses.
14089 Put definitions of externally-visible data in a small data section
14090 if that data is no bigger than @var{num} bytes. GCC can then access
14091 the data more efficiently; see @option{-mgpopt} for details.
14093 The default @option{-G} option depends on the configuration.
14095 @item -mlocal-sdata
14096 @itemx -mno-local-sdata
14097 @opindex mlocal-sdata
14098 @opindex mno-local-sdata
14099 Extend (do not extend) the @option{-G} behavior to local data too,
14100 such as to static variables in C@. @option{-mlocal-sdata} is the
14101 default for all configurations.
14103 If the linker complains that an application is using too much small data,
14104 you might want to try rebuilding the less performance-critical parts with
14105 @option{-mno-local-sdata}. You might also want to build large
14106 libraries with @option{-mno-local-sdata}, so that the libraries leave
14107 more room for the main program.
14109 @item -mextern-sdata
14110 @itemx -mno-extern-sdata
14111 @opindex mextern-sdata
14112 @opindex mno-extern-sdata
14113 Assume (do not assume) that externally-defined data will be in
14114 a small data section if that data is within the @option{-G} limit.
14115 @option{-mextern-sdata} is the default for all configurations.
14117 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
14118 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
14119 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
14120 is placed in a small data section. If @var{Var} is defined by another
14121 module, you must either compile that module with a high-enough
14122 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
14123 definition. If @var{Var} is common, you must link the application
14124 with a high-enough @option{-G} setting.
14126 The easiest way of satisfying these restrictions is to compile
14127 and link every module with the same @option{-G} option. However,
14128 you may wish to build a library that supports several different
14129 small data limits. You can do this by compiling the library with
14130 the highest supported @option{-G} setting and additionally using
14131 @option{-mno-extern-sdata} to stop the library from making assumptions
14132 about externally-defined data.
14138 Use (do not use) GP-relative accesses for symbols that are known to be
14139 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
14140 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
14143 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
14144 might not hold the value of @code{_gp}. For example, if the code is
14145 part of a library that might be used in a boot monitor, programs that
14146 call boot monitor routines will pass an unknown value in @code{$gp}.
14147 (In such situations, the boot monitor itself would usually be compiled
14148 with @option{-G0}.)
14150 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
14151 @option{-mno-extern-sdata}.
14153 @item -membedded-data
14154 @itemx -mno-embedded-data
14155 @opindex membedded-data
14156 @opindex mno-embedded-data
14157 Allocate variables to the read-only data section first if possible, then
14158 next in the small data section if possible, otherwise in data. This gives
14159 slightly slower code than the default, but reduces the amount of RAM required
14160 when executing, and thus may be preferred for some embedded systems.
14162 @item -muninit-const-in-rodata
14163 @itemx -mno-uninit-const-in-rodata
14164 @opindex muninit-const-in-rodata
14165 @opindex mno-uninit-const-in-rodata
14166 Put uninitialized @code{const} variables in the read-only data section.
14167 This option is only meaningful in conjunction with @option{-membedded-data}.
14169 @item -mcode-readable=@var{setting}
14170 @opindex mcode-readable
14171 Specify whether GCC may generate code that reads from executable sections.
14172 There are three possible settings:
14175 @item -mcode-readable=yes
14176 Instructions may freely access executable sections. This is the
14179 @item -mcode-readable=pcrel
14180 MIPS16 PC-relative load instructions can access executable sections,
14181 but other instructions must not do so. This option is useful on 4KSc
14182 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
14183 It is also useful on processors that can be configured to have a dual
14184 instruction/data SRAM interface and that, like the M4K, automatically
14185 redirect PC-relative loads to the instruction RAM.
14187 @item -mcode-readable=no
14188 Instructions must not access executable sections. This option can be
14189 useful on targets that are configured to have a dual instruction/data
14190 SRAM interface but that (unlike the M4K) do not automatically redirect
14191 PC-relative loads to the instruction RAM.
14194 @item -msplit-addresses
14195 @itemx -mno-split-addresses
14196 @opindex msplit-addresses
14197 @opindex mno-split-addresses
14198 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
14199 relocation operators. This option has been superseded by
14200 @option{-mexplicit-relocs} but is retained for backwards compatibility.
14202 @item -mexplicit-relocs
14203 @itemx -mno-explicit-relocs
14204 @opindex mexplicit-relocs
14205 @opindex mno-explicit-relocs
14206 Use (do not use) assembler relocation operators when dealing with symbolic
14207 addresses. The alternative, selected by @option{-mno-explicit-relocs},
14208 is to use assembler macros instead.
14210 @option{-mexplicit-relocs} is the default if GCC was configured
14211 to use an assembler that supports relocation operators.
14213 @item -mcheck-zero-division
14214 @itemx -mno-check-zero-division
14215 @opindex mcheck-zero-division
14216 @opindex mno-check-zero-division
14217 Trap (do not trap) on integer division by zero.
14219 The default is @option{-mcheck-zero-division}.
14221 @item -mdivide-traps
14222 @itemx -mdivide-breaks
14223 @opindex mdivide-traps
14224 @opindex mdivide-breaks
14225 MIPS systems check for division by zero by generating either a
14226 conditional trap or a break instruction. Using traps results in
14227 smaller code, but is only supported on MIPS II and later. Also, some
14228 versions of the Linux kernel have a bug that prevents trap from
14229 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
14230 allow conditional traps on architectures that support them and
14231 @option{-mdivide-breaks} to force the use of breaks.
14233 The default is usually @option{-mdivide-traps}, but this can be
14234 overridden at configure time using @option{--with-divide=breaks}.
14235 Divide-by-zero checks can be completely disabled using
14236 @option{-mno-check-zero-division}.
14241 @opindex mno-memcpy
14242 Force (do not force) the use of @code{memcpy()} for non-trivial block
14243 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
14244 most constant-sized copies.
14247 @itemx -mno-long-calls
14248 @opindex mlong-calls
14249 @opindex mno-long-calls
14250 Disable (do not disable) use of the @code{jal} instruction. Calling
14251 functions using @code{jal} is more efficient but requires the caller
14252 and callee to be in the same 256 megabyte segment.
14254 This option has no effect on abicalls code. The default is
14255 @option{-mno-long-calls}.
14261 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
14262 instructions, as provided by the R4650 ISA@.
14265 @itemx -mno-fused-madd
14266 @opindex mfused-madd
14267 @opindex mno-fused-madd
14268 Enable (disable) use of the floating point multiply-accumulate
14269 instructions, when they are available. The default is
14270 @option{-mfused-madd}.
14272 When multiply-accumulate instructions are used, the intermediate
14273 product is calculated to infinite precision and is not subject to
14274 the FCSR Flush to Zero bit. This may be undesirable in some
14279 Tell the MIPS assembler to not run its preprocessor over user
14280 assembler files (with a @samp{.s} suffix) when assembling them.
14283 @itemx -mno-fix-r4000
14284 @opindex mfix-r4000
14285 @opindex mno-fix-r4000
14286 Work around certain R4000 CPU errata:
14289 A double-word or a variable shift may give an incorrect result if executed
14290 immediately after starting an integer division.
14292 A double-word or a variable shift may give an incorrect result if executed
14293 while an integer multiplication is in progress.
14295 An integer division may give an incorrect result if started in a delay slot
14296 of a taken branch or a jump.
14300 @itemx -mno-fix-r4400
14301 @opindex mfix-r4400
14302 @opindex mno-fix-r4400
14303 Work around certain R4400 CPU errata:
14306 A double-word or a variable shift may give an incorrect result if executed
14307 immediately after starting an integer division.
14311 @itemx -mno-fix-r10000
14312 @opindex mfix-r10000
14313 @opindex mno-fix-r10000
14314 Work around certain R10000 errata:
14317 @code{ll}/@code{sc} sequences may not behave atomically on revisions
14318 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
14321 This option can only be used if the target architecture supports
14322 branch-likely instructions. @option{-mfix-r10000} is the default when
14323 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
14327 @itemx -mno-fix-vr4120
14328 @opindex mfix-vr4120
14329 Work around certain VR4120 errata:
14332 @code{dmultu} does not always produce the correct result.
14334 @code{div} and @code{ddiv} do not always produce the correct result if one
14335 of the operands is negative.
14337 The workarounds for the division errata rely on special functions in
14338 @file{libgcc.a}. At present, these functions are only provided by
14339 the @code{mips64vr*-elf} configurations.
14341 Other VR4120 errata require a nop to be inserted between certain pairs of
14342 instructions. These errata are handled by the assembler, not by GCC itself.
14345 @opindex mfix-vr4130
14346 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
14347 workarounds are implemented by the assembler rather than by GCC,
14348 although GCC will avoid using @code{mflo} and @code{mfhi} if the
14349 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
14350 instructions are available instead.
14353 @itemx -mno-fix-sb1
14355 Work around certain SB-1 CPU core errata.
14356 (This flag currently works around the SB-1 revision 2
14357 ``F1'' and ``F2'' floating point errata.)
14359 @item -mr10k-cache-barrier=@var{setting}
14360 @opindex mr10k-cache-barrier
14361 Specify whether GCC should insert cache barriers to avoid the
14362 side-effects of speculation on R10K processors.
14364 In common with many processors, the R10K tries to predict the outcome
14365 of a conditional branch and speculatively executes instructions from
14366 the ``taken'' branch. It later aborts these instructions if the
14367 predicted outcome was wrong. However, on the R10K, even aborted
14368 instructions can have side effects.
14370 This problem only affects kernel stores and, depending on the system,
14371 kernel loads. As an example, a speculatively-executed store may load
14372 the target memory into cache and mark the cache line as dirty, even if
14373 the store itself is later aborted. If a DMA operation writes to the
14374 same area of memory before the ``dirty'' line is flushed, the cached
14375 data will overwrite the DMA-ed data. See the R10K processor manual
14376 for a full description, including other potential problems.
14378 One workaround is to insert cache barrier instructions before every memory
14379 access that might be speculatively executed and that might have side
14380 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
14381 controls GCC's implementation of this workaround. It assumes that
14382 aborted accesses to any byte in the following regions will not have
14387 the memory occupied by the current function's stack frame;
14390 the memory occupied by an incoming stack argument;
14393 the memory occupied by an object with a link-time-constant address.
14396 It is the kernel's responsibility to ensure that speculative
14397 accesses to these regions are indeed safe.
14399 If the input program contains a function declaration such as:
14405 then the implementation of @code{foo} must allow @code{j foo} and
14406 @code{jal foo} to be executed speculatively. GCC honors this
14407 restriction for functions it compiles itself. It expects non-GCC
14408 functions (such as hand-written assembly code) to do the same.
14410 The option has three forms:
14413 @item -mr10k-cache-barrier=load-store
14414 Insert a cache barrier before a load or store that might be
14415 speculatively executed and that might have side effects even
14418 @item -mr10k-cache-barrier=store
14419 Insert a cache barrier before a store that might be speculatively
14420 executed and that might have side effects even if aborted.
14422 @item -mr10k-cache-barrier=none
14423 Disable the insertion of cache barriers. This is the default setting.
14426 @item -mflush-func=@var{func}
14427 @itemx -mno-flush-func
14428 @opindex mflush-func
14429 Specifies the function to call to flush the I and D caches, or to not
14430 call any such function. If called, the function must take the same
14431 arguments as the common @code{_flush_func()}, that is, the address of the
14432 memory range for which the cache is being flushed, the size of the
14433 memory range, and the number 3 (to flush both caches). The default
14434 depends on the target GCC was configured for, but commonly is either
14435 @samp{_flush_func} or @samp{__cpu_flush}.
14437 @item mbranch-cost=@var{num}
14438 @opindex mbranch-cost
14439 Set the cost of branches to roughly @var{num} ``simple'' instructions.
14440 This cost is only a heuristic and is not guaranteed to produce
14441 consistent results across releases. A zero cost redundantly selects
14442 the default, which is based on the @option{-mtune} setting.
14444 @item -mbranch-likely
14445 @itemx -mno-branch-likely
14446 @opindex mbranch-likely
14447 @opindex mno-branch-likely
14448 Enable or disable use of Branch Likely instructions, regardless of the
14449 default for the selected architecture. By default, Branch Likely
14450 instructions may be generated if they are supported by the selected
14451 architecture. An exception is for the MIPS32 and MIPS64 architectures
14452 and processors which implement those architectures; for those, Branch
14453 Likely instructions will not be generated by default because the MIPS32
14454 and MIPS64 architectures specifically deprecate their use.
14456 @item -mfp-exceptions
14457 @itemx -mno-fp-exceptions
14458 @opindex mfp-exceptions
14459 Specifies whether FP exceptions are enabled. This affects how we schedule
14460 FP instructions for some processors. The default is that FP exceptions are
14463 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
14464 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
14467 @item -mvr4130-align
14468 @itemx -mno-vr4130-align
14469 @opindex mvr4130-align
14470 The VR4130 pipeline is two-way superscalar, but can only issue two
14471 instructions together if the first one is 8-byte aligned. When this
14472 option is enabled, GCC will align pairs of instructions that it
14473 thinks should execute in parallel.
14475 This option only has an effect when optimizing for the VR4130.
14476 It normally makes code faster, but at the expense of making it bigger.
14477 It is enabled by default at optimization level @option{-O3}.
14482 Enable (disable) generation of @code{synci} instructions on
14483 architectures that support it. The @code{synci} instructions (if
14484 enabled) will be generated when @code{__builtin___clear_cache()} is
14487 This option defaults to @code{-mno-synci}, but the default can be
14488 overridden by configuring with @code{--with-synci}.
14490 When compiling code for single processor systems, it is generally safe
14491 to use @code{synci}. However, on many multi-core (SMP) systems, it
14492 will not invalidate the instruction caches on all cores and may lead
14493 to undefined behavior.
14495 @item -mrelax-pic-calls
14496 @itemx -mno-relax-pic-calls
14497 @opindex mrelax-pic-calls
14498 Try to turn PIC calls that are normally dispatched via register
14499 @code{$25} into direct calls. This is only possible if the linker can
14500 resolve the destination at link-time and if the destination is within
14501 range for a direct call.
14503 @option{-mrelax-pic-calls} is the default if GCC was configured to use
14504 an assembler and a linker that supports the @code{.reloc} assembly
14505 directive and @code{-mexplicit-relocs} is in effect. With
14506 @code{-mno-explicit-relocs}, this optimization can be performed by the
14507 assembler and the linker alone without help from the compiler.
14509 @item -mmcount-ra-address
14510 @itemx -mno-mcount-ra-address
14511 @opindex mmcount-ra-address
14512 @opindex mno-mcount-ra-address
14513 Emit (do not emit) code that allows @code{_mcount} to modify the
14514 calling function's return address. When enabled, this option extends
14515 the usual @code{_mcount} interface with a new @var{ra-address}
14516 parameter, which has type @code{intptr_t *} and is passed in register
14517 @code{$12}. @code{_mcount} can then modify the return address by
14518 doing both of the following:
14521 Returning the new address in register @code{$31}.
14523 Storing the new address in @code{*@var{ra-address}},
14524 if @var{ra-address} is nonnull.
14527 The default is @option{-mno-mcount-ra-address}.
14532 @subsection MMIX Options
14533 @cindex MMIX Options
14535 These options are defined for the MMIX:
14539 @itemx -mno-libfuncs
14541 @opindex mno-libfuncs
14542 Specify that intrinsic library functions are being compiled, passing all
14543 values in registers, no matter the size.
14546 @itemx -mno-epsilon
14548 @opindex mno-epsilon
14549 Generate floating-point comparison instructions that compare with respect
14550 to the @code{rE} epsilon register.
14552 @item -mabi=mmixware
14554 @opindex mabi=mmixware
14556 Generate code that passes function parameters and return values that (in
14557 the called function) are seen as registers @code{$0} and up, as opposed to
14558 the GNU ABI which uses global registers @code{$231} and up.
14560 @item -mzero-extend
14561 @itemx -mno-zero-extend
14562 @opindex mzero-extend
14563 @opindex mno-zero-extend
14564 When reading data from memory in sizes shorter than 64 bits, use (do not
14565 use) zero-extending load instructions by default, rather than
14566 sign-extending ones.
14569 @itemx -mno-knuthdiv
14571 @opindex mno-knuthdiv
14572 Make the result of a division yielding a remainder have the same sign as
14573 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
14574 remainder follows the sign of the dividend. Both methods are
14575 arithmetically valid, the latter being almost exclusively used.
14577 @item -mtoplevel-symbols
14578 @itemx -mno-toplevel-symbols
14579 @opindex mtoplevel-symbols
14580 @opindex mno-toplevel-symbols
14581 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
14582 code can be used with the @code{PREFIX} assembly directive.
14586 Generate an executable in the ELF format, rather than the default
14587 @samp{mmo} format used by the @command{mmix} simulator.
14589 @item -mbranch-predict
14590 @itemx -mno-branch-predict
14591 @opindex mbranch-predict
14592 @opindex mno-branch-predict
14593 Use (do not use) the probable-branch instructions, when static branch
14594 prediction indicates a probable branch.
14596 @item -mbase-addresses
14597 @itemx -mno-base-addresses
14598 @opindex mbase-addresses
14599 @opindex mno-base-addresses
14600 Generate (do not generate) code that uses @emph{base addresses}. Using a
14601 base address automatically generates a request (handled by the assembler
14602 and the linker) for a constant to be set up in a global register. The
14603 register is used for one or more base address requests within the range 0
14604 to 255 from the value held in the register. The generally leads to short
14605 and fast code, but the number of different data items that can be
14606 addressed is limited. This means that a program that uses lots of static
14607 data may require @option{-mno-base-addresses}.
14609 @item -msingle-exit
14610 @itemx -mno-single-exit
14611 @opindex msingle-exit
14612 @opindex mno-single-exit
14613 Force (do not force) generated code to have a single exit point in each
14617 @node MN10300 Options
14618 @subsection MN10300 Options
14619 @cindex MN10300 options
14621 These @option{-m} options are defined for Matsushita MN10300 architectures:
14626 Generate code to avoid bugs in the multiply instructions for the MN10300
14627 processors. This is the default.
14629 @item -mno-mult-bug
14630 @opindex mno-mult-bug
14631 Do not generate code to avoid bugs in the multiply instructions for the
14632 MN10300 processors.
14636 Generate code which uses features specific to the AM33 processor.
14640 Do not generate code which uses features specific to the AM33 processor. This
14643 @item -mreturn-pointer-on-d0
14644 @opindex mreturn-pointer-on-d0
14645 When generating a function which returns a pointer, return the pointer
14646 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
14647 only in a0, and attempts to call such functions without a prototype
14648 would result in errors. Note that this option is on by default; use
14649 @option{-mno-return-pointer-on-d0} to disable it.
14653 Do not link in the C run-time initialization object file.
14657 Indicate to the linker that it should perform a relaxation optimization pass
14658 to shorten branches, calls and absolute memory addresses. This option only
14659 has an effect when used on the command line for the final link step.
14661 This option makes symbolic debugging impossible.
14664 @node PDP-11 Options
14665 @subsection PDP-11 Options
14666 @cindex PDP-11 Options
14668 These options are defined for the PDP-11:
14673 Use hardware FPP floating point. This is the default. (FIS floating
14674 point on the PDP-11/40 is not supported.)
14677 @opindex msoft-float
14678 Do not use hardware floating point.
14682 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
14686 Return floating-point results in memory. This is the default.
14690 Generate code for a PDP-11/40.
14694 Generate code for a PDP-11/45. This is the default.
14698 Generate code for a PDP-11/10.
14700 @item -mbcopy-builtin
14701 @opindex mbcopy-builtin
14702 Use inline @code{movmemhi} patterns for copying memory. This is the
14707 Do not use inline @code{movmemhi} patterns for copying memory.
14713 Use 16-bit @code{int}. This is the default.
14719 Use 32-bit @code{int}.
14722 @itemx -mno-float32
14724 @opindex mno-float32
14725 Use 64-bit @code{float}. This is the default.
14728 @itemx -mno-float64
14730 @opindex mno-float64
14731 Use 32-bit @code{float}.
14735 Use @code{abshi2} pattern. This is the default.
14739 Do not use @code{abshi2} pattern.
14741 @item -mbranch-expensive
14742 @opindex mbranch-expensive
14743 Pretend that branches are expensive. This is for experimenting with
14744 code generation only.
14746 @item -mbranch-cheap
14747 @opindex mbranch-cheap
14748 Do not pretend that branches are expensive. This is the default.
14752 Generate code for a system with split I&D@.
14756 Generate code for a system without split I&D@. This is the default.
14760 Use Unix assembler syntax. This is the default when configured for
14761 @samp{pdp11-*-bsd}.
14765 Use DEC assembler syntax. This is the default when configured for any
14766 PDP-11 target other than @samp{pdp11-*-bsd}.
14769 @node picoChip Options
14770 @subsection picoChip Options
14771 @cindex picoChip options
14773 These @samp{-m} options are defined for picoChip implementations:
14777 @item -mae=@var{ae_type}
14779 Set the instruction set, register set, and instruction scheduling
14780 parameters for array element type @var{ae_type}. Supported values
14781 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
14783 @option{-mae=ANY} selects a completely generic AE type. Code
14784 generated with this option will run on any of the other AE types. The
14785 code will not be as efficient as it would be if compiled for a specific
14786 AE type, and some types of operation (e.g., multiplication) will not
14787 work properly on all types of AE.
14789 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
14790 for compiled code, and is the default.
14792 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
14793 option may suffer from poor performance of byte (char) manipulation,
14794 since the DSP AE does not provide hardware support for byte load/stores.
14796 @item -msymbol-as-address
14797 Enable the compiler to directly use a symbol name as an address in a
14798 load/store instruction, without first loading it into a
14799 register. Typically, the use of this option will generate larger
14800 programs, which run faster than when the option isn't used. However, the
14801 results vary from program to program, so it is left as a user option,
14802 rather than being permanently enabled.
14804 @item -mno-inefficient-warnings
14805 Disables warnings about the generation of inefficient code. These
14806 warnings can be generated, for example, when compiling code which
14807 performs byte-level memory operations on the MAC AE type. The MAC AE has
14808 no hardware support for byte-level memory operations, so all byte
14809 load/stores must be synthesized from word load/store operations. This is
14810 inefficient and a warning will be generated indicating to the programmer
14811 that they should rewrite the code to avoid byte operations, or to target
14812 an AE type which has the necessary hardware support. This option enables
14813 the warning to be turned off.
14817 @node PowerPC Options
14818 @subsection PowerPC Options
14819 @cindex PowerPC options
14821 These are listed under @xref{RS/6000 and PowerPC Options}.
14823 @node RS/6000 and PowerPC Options
14824 @subsection IBM RS/6000 and PowerPC Options
14825 @cindex RS/6000 and PowerPC Options
14826 @cindex IBM RS/6000 and PowerPC Options
14828 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
14835 @itemx -mno-powerpc
14836 @itemx -mpowerpc-gpopt
14837 @itemx -mno-powerpc-gpopt
14838 @itemx -mpowerpc-gfxopt
14839 @itemx -mno-powerpc-gfxopt
14841 @itemx -mno-powerpc64
14845 @itemx -mno-popcntb
14847 @itemx -mno-popcntd
14855 @itemx -mno-hard-dfp
14859 @opindex mno-power2
14861 @opindex mno-powerpc
14862 @opindex mpowerpc-gpopt
14863 @opindex mno-powerpc-gpopt
14864 @opindex mpowerpc-gfxopt
14865 @opindex mno-powerpc-gfxopt
14866 @opindex mpowerpc64
14867 @opindex mno-powerpc64
14871 @opindex mno-popcntb
14873 @opindex mno-popcntd
14879 @opindex mno-mfpgpr
14881 @opindex mno-hard-dfp
14882 GCC supports two related instruction set architectures for the
14883 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
14884 instructions supported by the @samp{rios} chip set used in the original
14885 RS/6000 systems and the @dfn{PowerPC} instruction set is the
14886 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
14887 the IBM 4xx, 6xx, and follow-on microprocessors.
14889 Neither architecture is a subset of the other. However there is a
14890 large common subset of instructions supported by both. An MQ
14891 register is included in processors supporting the POWER architecture.
14893 You use these options to specify which instructions are available on the
14894 processor you are using. The default value of these options is
14895 determined when configuring GCC@. Specifying the
14896 @option{-mcpu=@var{cpu_type}} overrides the specification of these
14897 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
14898 rather than the options listed above.
14900 The @option{-mpower} option allows GCC to generate instructions that
14901 are found only in the POWER architecture and to use the MQ register.
14902 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
14903 to generate instructions that are present in the POWER2 architecture but
14904 not the original POWER architecture.
14906 The @option{-mpowerpc} option allows GCC to generate instructions that
14907 are found only in the 32-bit subset of the PowerPC architecture.
14908 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
14909 GCC to use the optional PowerPC architecture instructions in the
14910 General Purpose group, including floating-point square root. Specifying
14911 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
14912 use the optional PowerPC architecture instructions in the Graphics
14913 group, including floating-point select.
14915 The @option{-mmfcrf} option allows GCC to generate the move from
14916 condition register field instruction implemented on the POWER4
14917 processor and other processors that support the PowerPC V2.01
14919 The @option{-mpopcntb} option allows GCC to generate the popcount and
14920 double precision FP reciprocal estimate instruction implemented on the
14921 POWER5 processor and other processors that support the PowerPC V2.02
14923 The @option{-mpopcntd} option allows GCC to generate the popcount
14924 instruction implemented on the POWER7 processor and other processors
14925 that support the PowerPC V2.06 architecture.
14926 The @option{-mfprnd} option allows GCC to generate the FP round to
14927 integer instructions implemented on the POWER5+ processor and other
14928 processors that support the PowerPC V2.03 architecture.
14929 The @option{-mcmpb} option allows GCC to generate the compare bytes
14930 instruction implemented on the POWER6 processor and other processors
14931 that support the PowerPC V2.05 architecture.
14932 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
14933 general purpose register instructions implemented on the POWER6X
14934 processor and other processors that support the extended PowerPC V2.05
14936 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
14937 point instructions implemented on some POWER processors.
14939 The @option{-mpowerpc64} option allows GCC to generate the additional
14940 64-bit instructions that are found in the full PowerPC64 architecture
14941 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
14942 @option{-mno-powerpc64}.
14944 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
14945 will use only the instructions in the common subset of both
14946 architectures plus some special AIX common-mode calls, and will not use
14947 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
14948 permits GCC to use any instruction from either architecture and to
14949 allow use of the MQ register; specify this for the Motorola MPC601.
14951 @item -mnew-mnemonics
14952 @itemx -mold-mnemonics
14953 @opindex mnew-mnemonics
14954 @opindex mold-mnemonics
14955 Select which mnemonics to use in the generated assembler code. With
14956 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
14957 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
14958 assembler mnemonics defined for the POWER architecture. Instructions
14959 defined in only one architecture have only one mnemonic; GCC uses that
14960 mnemonic irrespective of which of these options is specified.
14962 GCC defaults to the mnemonics appropriate for the architecture in
14963 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
14964 value of these option. Unless you are building a cross-compiler, you
14965 should normally not specify either @option{-mnew-mnemonics} or
14966 @option{-mold-mnemonics}, but should instead accept the default.
14968 @item -mcpu=@var{cpu_type}
14970 Set architecture type, register usage, choice of mnemonics, and
14971 instruction scheduling parameters for machine type @var{cpu_type}.
14972 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
14973 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
14974 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
14975 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
14976 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
14977 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
14978 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{ec603e}, @samp{G3},
14979 @samp{G4}, @samp{G5}, @samp{titan}, @samp{power}, @samp{power2}, @samp{power3},
14980 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x},
14981 @samp{power7}, @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
14982 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
14984 @option{-mcpu=common} selects a completely generic processor. Code
14985 generated under this option will run on any POWER or PowerPC processor.
14986 GCC will use only the instructions in the common subset of both
14987 architectures, and will not use the MQ register. GCC assumes a generic
14988 processor model for scheduling purposes.
14990 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
14991 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
14992 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
14993 types, with an appropriate, generic processor model assumed for
14994 scheduling purposes.
14996 The other options specify a specific processor. Code generated under
14997 those options will run best on that processor, and may not run at all on
15000 The @option{-mcpu} options automatically enable or disable the
15003 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
15004 -mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol
15005 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
15006 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx}
15008 The particular options set for any particular CPU will vary between
15009 compiler versions, depending on what setting seems to produce optimal
15010 code for that CPU; it doesn't necessarily reflect the actual hardware's
15011 capabilities. If you wish to set an individual option to a particular
15012 value, you may specify it after the @option{-mcpu} option, like
15013 @samp{-mcpu=970 -mno-altivec}.
15015 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
15016 not enabled or disabled by the @option{-mcpu} option at present because
15017 AIX does not have full support for these options. You may still
15018 enable or disable them individually if you're sure it'll work in your
15021 @item -mtune=@var{cpu_type}
15023 Set the instruction scheduling parameters for machine type
15024 @var{cpu_type}, but do not set the architecture type, register usage, or
15025 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
15026 values for @var{cpu_type} are used for @option{-mtune} as for
15027 @option{-mcpu}. If both are specified, the code generated will use the
15028 architecture, registers, and mnemonics set by @option{-mcpu}, but the
15029 scheduling parameters set by @option{-mtune}.
15031 @item -mcmodel=small
15032 @opindex mcmodel=small
15033 Generate PowerPC64 code for the small model: The TOC is limited to
15036 @item -mcmodel=large
15037 @opindex mcmodel=large
15038 Generate PowerPC64 code for the large model: The TOC may be up to 4G
15039 in size. Other data and code is only limited by the 64-bit address
15043 @itemx -mno-altivec
15045 @opindex mno-altivec
15046 Generate code that uses (does not use) AltiVec instructions, and also
15047 enable the use of built-in functions that allow more direct access to
15048 the AltiVec instruction set. You may also need to set
15049 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
15055 @opindex mno-vrsave
15056 Generate VRSAVE instructions when generating AltiVec code.
15058 @item -mgen-cell-microcode
15059 @opindex mgen-cell-microcode
15060 Generate Cell microcode instructions
15062 @item -mwarn-cell-microcode
15063 @opindex mwarn-cell-microcode
15064 Warning when a Cell microcode instruction is going to emitted. An example
15065 of a Cell microcode instruction is a variable shift.
15068 @opindex msecure-plt
15069 Generate code that allows ld and ld.so to build executables and shared
15070 libraries with non-exec .plt and .got sections. This is a PowerPC
15071 32-bit SYSV ABI option.
15075 Generate code that uses a BSS .plt section that ld.so fills in, and
15076 requires .plt and .got sections that are both writable and executable.
15077 This is a PowerPC 32-bit SYSV ABI option.
15083 This switch enables or disables the generation of ISEL instructions.
15085 @item -misel=@var{yes/no}
15086 This switch has been deprecated. Use @option{-misel} and
15087 @option{-mno-isel} instead.
15093 This switch enables or disables the generation of SPE simd
15099 @opindex mno-paired
15100 This switch enables or disables the generation of PAIRED simd
15103 @item -mspe=@var{yes/no}
15104 This option has been deprecated. Use @option{-mspe} and
15105 @option{-mno-spe} instead.
15111 Generate code that uses (does not use) vector/scalar (VSX)
15112 instructions, and also enable the use of built-in functions that allow
15113 more direct access to the VSX instruction set.
15115 @item -mfloat-gprs=@var{yes/single/double/no}
15116 @itemx -mfloat-gprs
15117 @opindex mfloat-gprs
15118 This switch enables or disables the generation of floating point
15119 operations on the general purpose registers for architectures that
15122 The argument @var{yes} or @var{single} enables the use of
15123 single-precision floating point operations.
15125 The argument @var{double} enables the use of single and
15126 double-precision floating point operations.
15128 The argument @var{no} disables floating point operations on the
15129 general purpose registers.
15131 This option is currently only available on the MPC854x.
15137 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
15138 targets (including GNU/Linux). The 32-bit environment sets int, long
15139 and pointer to 32 bits and generates code that runs on any PowerPC
15140 variant. The 64-bit environment sets int to 32 bits and long and
15141 pointer to 64 bits, and generates code for PowerPC64, as for
15142 @option{-mpowerpc64}.
15145 @itemx -mno-fp-in-toc
15146 @itemx -mno-sum-in-toc
15147 @itemx -mminimal-toc
15149 @opindex mno-fp-in-toc
15150 @opindex mno-sum-in-toc
15151 @opindex mminimal-toc
15152 Modify generation of the TOC (Table Of Contents), which is created for
15153 every executable file. The @option{-mfull-toc} option is selected by
15154 default. In that case, GCC will allocate at least one TOC entry for
15155 each unique non-automatic variable reference in your program. GCC
15156 will also place floating-point constants in the TOC@. However, only
15157 16,384 entries are available in the TOC@.
15159 If you receive a linker error message that saying you have overflowed
15160 the available TOC space, you can reduce the amount of TOC space used
15161 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
15162 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
15163 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
15164 generate code to calculate the sum of an address and a constant at
15165 run-time instead of putting that sum into the TOC@. You may specify one
15166 or both of these options. Each causes GCC to produce very slightly
15167 slower and larger code at the expense of conserving TOC space.
15169 If you still run out of space in the TOC even when you specify both of
15170 these options, specify @option{-mminimal-toc} instead. This option causes
15171 GCC to make only one TOC entry for every file. When you specify this
15172 option, GCC will produce code that is slower and larger but which
15173 uses extremely little TOC space. You may wish to use this option
15174 only on files that contain less frequently executed code.
15180 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
15181 @code{long} type, and the infrastructure needed to support them.
15182 Specifying @option{-maix64} implies @option{-mpowerpc64} and
15183 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
15184 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
15187 @itemx -mno-xl-compat
15188 @opindex mxl-compat
15189 @opindex mno-xl-compat
15190 Produce code that conforms more closely to IBM XL compiler semantics
15191 when using AIX-compatible ABI@. Pass floating-point arguments to
15192 prototyped functions beyond the register save area (RSA) on the stack
15193 in addition to argument FPRs. Do not assume that most significant
15194 double in 128-bit long double value is properly rounded when comparing
15195 values and converting to double. Use XL symbol names for long double
15198 The AIX calling convention was extended but not initially documented to
15199 handle an obscure K&R C case of calling a function that takes the
15200 address of its arguments with fewer arguments than declared. IBM XL
15201 compilers access floating point arguments which do not fit in the
15202 RSA from the stack when a subroutine is compiled without
15203 optimization. Because always storing floating-point arguments on the
15204 stack is inefficient and rarely needed, this option is not enabled by
15205 default and only is necessary when calling subroutines compiled by IBM
15206 XL compilers without optimization.
15210 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
15211 application written to use message passing with special startup code to
15212 enable the application to run. The system must have PE installed in the
15213 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
15214 must be overridden with the @option{-specs=} option to specify the
15215 appropriate directory location. The Parallel Environment does not
15216 support threads, so the @option{-mpe} option and the @option{-pthread}
15217 option are incompatible.
15219 @item -malign-natural
15220 @itemx -malign-power
15221 @opindex malign-natural
15222 @opindex malign-power
15223 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
15224 @option{-malign-natural} overrides the ABI-defined alignment of larger
15225 types, such as floating-point doubles, on their natural size-based boundary.
15226 The option @option{-malign-power} instructs GCC to follow the ABI-specified
15227 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
15229 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
15233 @itemx -mhard-float
15234 @opindex msoft-float
15235 @opindex mhard-float
15236 Generate code that does not use (uses) the floating-point register set.
15237 Software floating point emulation is provided if you use the
15238 @option{-msoft-float} option, and pass the option to GCC when linking.
15240 @item -msingle-float
15241 @itemx -mdouble-float
15242 @opindex msingle-float
15243 @opindex mdouble-float
15244 Generate code for single or double-precision floating point operations.
15245 @option{-mdouble-float} implies @option{-msingle-float}.
15248 @opindex msimple-fpu
15249 Do not generate sqrt and div instructions for hardware floating point unit.
15253 Specify type of floating point unit. Valid values are @var{sp_lite}
15254 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
15255 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
15256 and @var{dp_full} (equivalent to -mdouble-float).
15259 @opindex mxilinx-fpu
15260 Perform optimizations for floating point unit on Xilinx PPC 405/440.
15263 @itemx -mno-multiple
15265 @opindex mno-multiple
15266 Generate code that uses (does not use) the load multiple word
15267 instructions and the store multiple word instructions. These
15268 instructions are generated by default on POWER systems, and not
15269 generated on PowerPC systems. Do not use @option{-mmultiple} on little
15270 endian PowerPC systems, since those instructions do not work when the
15271 processor is in little endian mode. The exceptions are PPC740 and
15272 PPC750 which permit the instructions usage in little endian mode.
15277 @opindex mno-string
15278 Generate code that uses (does not use) the load string instructions
15279 and the store string word instructions to save multiple registers and
15280 do small block moves. These instructions are generated by default on
15281 POWER systems, and not generated on PowerPC systems. Do not use
15282 @option{-mstring} on little endian PowerPC systems, since those
15283 instructions do not work when the processor is in little endian mode.
15284 The exceptions are PPC740 and PPC750 which permit the instructions
15285 usage in little endian mode.
15290 @opindex mno-update
15291 Generate code that uses (does not use) the load or store instructions
15292 that update the base register to the address of the calculated memory
15293 location. These instructions are generated by default. If you use
15294 @option{-mno-update}, there is a small window between the time that the
15295 stack pointer is updated and the address of the previous frame is
15296 stored, which means code that walks the stack frame across interrupts or
15297 signals may get corrupted data.
15299 @item -mavoid-indexed-addresses
15300 @itemx -mno-avoid-indexed-addresses
15301 @opindex mavoid-indexed-addresses
15302 @opindex mno-avoid-indexed-addresses
15303 Generate code that tries to avoid (not avoid) the use of indexed load
15304 or store instructions. These instructions can incur a performance
15305 penalty on Power6 processors in certain situations, such as when
15306 stepping through large arrays that cross a 16M boundary. This option
15307 is enabled by default when targetting Power6 and disabled otherwise.
15310 @itemx -mno-fused-madd
15311 @opindex mfused-madd
15312 @opindex mno-fused-madd
15313 Generate code that uses (does not use) the floating point multiply and
15314 accumulate instructions. These instructions are generated by default if
15315 hardware floating is used.
15321 Generate code that uses (does not use) the half-word multiply and
15322 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
15323 These instructions are generated by default when targetting those
15330 Generate code that uses (does not use) the string-search @samp{dlmzb}
15331 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
15332 generated by default when targetting those processors.
15334 @item -mno-bit-align
15336 @opindex mno-bit-align
15337 @opindex mbit-align
15338 On System V.4 and embedded PowerPC systems do not (do) force structures
15339 and unions that contain bit-fields to be aligned to the base type of the
15342 For example, by default a structure containing nothing but 8
15343 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
15344 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
15345 the structure would be aligned to a 1 byte boundary and be one byte in
15348 @item -mno-strict-align
15349 @itemx -mstrict-align
15350 @opindex mno-strict-align
15351 @opindex mstrict-align
15352 On System V.4 and embedded PowerPC systems do not (do) assume that
15353 unaligned memory references will be handled by the system.
15355 @item -mrelocatable
15356 @itemx -mno-relocatable
15357 @opindex mrelocatable
15358 @opindex mno-relocatable
15359 On embedded PowerPC systems generate code that allows (does not allow)
15360 the program to be relocated to a different address at runtime. If you
15361 use @option{-mrelocatable} on any module, all objects linked together must
15362 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
15364 @item -mrelocatable-lib
15365 @itemx -mno-relocatable-lib
15366 @opindex mrelocatable-lib
15367 @opindex mno-relocatable-lib
15368 On embedded PowerPC systems generate code that allows (does not allow)
15369 the program to be relocated to a different address at runtime. Modules
15370 compiled with @option{-mrelocatable-lib} can be linked with either modules
15371 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
15372 with modules compiled with the @option{-mrelocatable} options.
15378 On System V.4 and embedded PowerPC systems do not (do) assume that
15379 register 2 contains a pointer to a global area pointing to the addresses
15380 used in the program.
15383 @itemx -mlittle-endian
15385 @opindex mlittle-endian
15386 On System V.4 and embedded PowerPC systems compile code for the
15387 processor in little endian mode. The @option{-mlittle-endian} option is
15388 the same as @option{-mlittle}.
15391 @itemx -mbig-endian
15393 @opindex mbig-endian
15394 On System V.4 and embedded PowerPC systems compile code for the
15395 processor in big endian mode. The @option{-mbig-endian} option is
15396 the same as @option{-mbig}.
15398 @item -mdynamic-no-pic
15399 @opindex mdynamic-no-pic
15400 On Darwin and Mac OS X systems, compile code so that it is not
15401 relocatable, but that its external references are relocatable. The
15402 resulting code is suitable for applications, but not shared
15405 @item -mprioritize-restricted-insns=@var{priority}
15406 @opindex mprioritize-restricted-insns
15407 This option controls the priority that is assigned to
15408 dispatch-slot restricted instructions during the second scheduling
15409 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
15410 @var{no/highest/second-highest} priority to dispatch slot restricted
15413 @item -msched-costly-dep=@var{dependence_type}
15414 @opindex msched-costly-dep
15415 This option controls which dependences are considered costly
15416 by the target during instruction scheduling. The argument
15417 @var{dependence_type} takes one of the following values:
15418 @var{no}: no dependence is costly,
15419 @var{all}: all dependences are costly,
15420 @var{true_store_to_load}: a true dependence from store to load is costly,
15421 @var{store_to_load}: any dependence from store to load is costly,
15422 @var{number}: any dependence which latency >= @var{number} is costly.
15424 @item -minsert-sched-nops=@var{scheme}
15425 @opindex minsert-sched-nops
15426 This option controls which nop insertion scheme will be used during
15427 the second scheduling pass. The argument @var{scheme} takes one of the
15429 @var{no}: Don't insert nops.
15430 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
15431 according to the scheduler's grouping.
15432 @var{regroup_exact}: Insert nops to force costly dependent insns into
15433 separate groups. Insert exactly as many nops as needed to force an insn
15434 to a new group, according to the estimated processor grouping.
15435 @var{number}: Insert nops to force costly dependent insns into
15436 separate groups. Insert @var{number} nops to force an insn to a new group.
15439 @opindex mcall-sysv
15440 On System V.4 and embedded PowerPC systems compile code using calling
15441 conventions that adheres to the March 1995 draft of the System V
15442 Application Binary Interface, PowerPC processor supplement. This is the
15443 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
15445 @item -mcall-sysv-eabi
15447 @opindex mcall-sysv-eabi
15448 @opindex mcall-eabi
15449 Specify both @option{-mcall-sysv} and @option{-meabi} options.
15451 @item -mcall-sysv-noeabi
15452 @opindex mcall-sysv-noeabi
15453 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
15455 @item -mcall-aixdesc
15457 On System V.4 and embedded PowerPC systems compile code for the AIX
15461 @opindex mcall-linux
15462 On System V.4 and embedded PowerPC systems compile code for the
15463 Linux-based GNU system.
15467 On System V.4 and embedded PowerPC systems compile code for the
15468 Hurd-based GNU system.
15470 @item -mcall-freebsd
15471 @opindex mcall-freebsd
15472 On System V.4 and embedded PowerPC systems compile code for the
15473 FreeBSD operating system.
15475 @item -mcall-netbsd
15476 @opindex mcall-netbsd
15477 On System V.4 and embedded PowerPC systems compile code for the
15478 NetBSD operating system.
15480 @item -mcall-openbsd
15481 @opindex mcall-netbsd
15482 On System V.4 and embedded PowerPC systems compile code for the
15483 OpenBSD operating system.
15485 @item -maix-struct-return
15486 @opindex maix-struct-return
15487 Return all structures in memory (as specified by the AIX ABI)@.
15489 @item -msvr4-struct-return
15490 @opindex msvr4-struct-return
15491 Return structures smaller than 8 bytes in registers (as specified by the
15494 @item -mabi=@var{abi-type}
15496 Extend the current ABI with a particular extension, or remove such extension.
15497 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
15498 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
15502 Extend the current ABI with SPE ABI extensions. This does not change
15503 the default ABI, instead it adds the SPE ABI extensions to the current
15507 @opindex mabi=no-spe
15508 Disable Booke SPE ABI extensions for the current ABI@.
15510 @item -mabi=ibmlongdouble
15511 @opindex mabi=ibmlongdouble
15512 Change the current ABI to use IBM extended precision long double.
15513 This is a PowerPC 32-bit SYSV ABI option.
15515 @item -mabi=ieeelongdouble
15516 @opindex mabi=ieeelongdouble
15517 Change the current ABI to use IEEE extended precision long double.
15518 This is a PowerPC 32-bit Linux ABI option.
15521 @itemx -mno-prototype
15522 @opindex mprototype
15523 @opindex mno-prototype
15524 On System V.4 and embedded PowerPC systems assume that all calls to
15525 variable argument functions are properly prototyped. Otherwise, the
15526 compiler must insert an instruction before every non prototyped call to
15527 set or clear bit 6 of the condition code register (@var{CR}) to
15528 indicate whether floating point values were passed in the floating point
15529 registers in case the function takes a variable arguments. With
15530 @option{-mprototype}, only calls to prototyped variable argument functions
15531 will set or clear the bit.
15535 On embedded PowerPC systems, assume that the startup module is called
15536 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
15537 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
15542 On embedded PowerPC systems, assume that the startup module is called
15543 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
15548 On embedded PowerPC systems, assume that the startup module is called
15549 @file{crt0.o} and the standard C libraries are @file{libads.a} and
15552 @item -myellowknife
15553 @opindex myellowknife
15554 On embedded PowerPC systems, assume that the startup module is called
15555 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
15560 On System V.4 and embedded PowerPC systems, specify that you are
15561 compiling for a VxWorks system.
15565 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
15566 header to indicate that @samp{eabi} extended relocations are used.
15572 On System V.4 and embedded PowerPC systems do (do not) adhere to the
15573 Embedded Applications Binary Interface (eabi) which is a set of
15574 modifications to the System V.4 specifications. Selecting @option{-meabi}
15575 means that the stack is aligned to an 8 byte boundary, a function
15576 @code{__eabi} is called to from @code{main} to set up the eabi
15577 environment, and the @option{-msdata} option can use both @code{r2} and
15578 @code{r13} to point to two separate small data areas. Selecting
15579 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
15580 do not call an initialization function from @code{main}, and the
15581 @option{-msdata} option will only use @code{r13} to point to a single
15582 small data area. The @option{-meabi} option is on by default if you
15583 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
15586 @opindex msdata=eabi
15587 On System V.4 and embedded PowerPC systems, put small initialized
15588 @code{const} global and static data in the @samp{.sdata2} section, which
15589 is pointed to by register @code{r2}. Put small initialized
15590 non-@code{const} global and static data in the @samp{.sdata} section,
15591 which is pointed to by register @code{r13}. Put small uninitialized
15592 global and static data in the @samp{.sbss} section, which is adjacent to
15593 the @samp{.sdata} section. The @option{-msdata=eabi} option is
15594 incompatible with the @option{-mrelocatable} option. The
15595 @option{-msdata=eabi} option also sets the @option{-memb} option.
15598 @opindex msdata=sysv
15599 On System V.4 and embedded PowerPC systems, put small global and static
15600 data in the @samp{.sdata} section, which is pointed to by register
15601 @code{r13}. Put small uninitialized global and static data in the
15602 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
15603 The @option{-msdata=sysv} option is incompatible with the
15604 @option{-mrelocatable} option.
15606 @item -msdata=default
15608 @opindex msdata=default
15610 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
15611 compile code the same as @option{-msdata=eabi}, otherwise compile code the
15612 same as @option{-msdata=sysv}.
15615 @opindex msdata=data
15616 On System V.4 and embedded PowerPC systems, put small global
15617 data in the @samp{.sdata} section. Put small uninitialized global
15618 data in the @samp{.sbss} section. Do not use register @code{r13}
15619 to address small data however. This is the default behavior unless
15620 other @option{-msdata} options are used.
15624 @opindex msdata=none
15626 On embedded PowerPC systems, put all initialized global and static data
15627 in the @samp{.data} section, and all uninitialized data in the
15628 @samp{.bss} section.
15630 @item -mblock-move-inline-limit=@var{num}
15631 @opindex mblock-move-inline-limit
15632 Inline all block moves (such as calls to @code{memcpy} or structure
15633 copies) less than or equal to @var{num} bytes. The minimum value for
15634 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
15635 targets. The default value is target-specific.
15639 @cindex smaller data references (PowerPC)
15640 @cindex .sdata/.sdata2 references (PowerPC)
15641 On embedded PowerPC systems, put global and static items less than or
15642 equal to @var{num} bytes into the small data or bss sections instead of
15643 the normal data or bss section. By default, @var{num} is 8. The
15644 @option{-G @var{num}} switch is also passed to the linker.
15645 All modules should be compiled with the same @option{-G @var{num}} value.
15648 @itemx -mno-regnames
15650 @opindex mno-regnames
15651 On System V.4 and embedded PowerPC systems do (do not) emit register
15652 names in the assembly language output using symbolic forms.
15655 @itemx -mno-longcall
15657 @opindex mno-longcall
15658 By default assume that all calls are far away so that a longer more
15659 expensive calling sequence is required. This is required for calls
15660 further than 32 megabytes (33,554,432 bytes) from the current location.
15661 A short call will be generated if the compiler knows
15662 the call cannot be that far away. This setting can be overridden by
15663 the @code{shortcall} function attribute, or by @code{#pragma
15666 Some linkers are capable of detecting out-of-range calls and generating
15667 glue code on the fly. On these systems, long calls are unnecessary and
15668 generate slower code. As of this writing, the AIX linker can do this,
15669 as can the GNU linker for PowerPC/64. It is planned to add this feature
15670 to the GNU linker for 32-bit PowerPC systems as well.
15672 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
15673 callee, L42'', plus a ``branch island'' (glue code). The two target
15674 addresses represent the callee and the ``branch island''. The
15675 Darwin/PPC linker will prefer the first address and generate a ``bl
15676 callee'' if the PPC ``bl'' instruction will reach the callee directly;
15677 otherwise, the linker will generate ``bl L42'' to call the ``branch
15678 island''. The ``branch island'' is appended to the body of the
15679 calling function; it computes the full 32-bit address of the callee
15682 On Mach-O (Darwin) systems, this option directs the compiler emit to
15683 the glue for every direct call, and the Darwin linker decides whether
15684 to use or discard it.
15686 In the future, we may cause GCC to ignore all longcall specifications
15687 when the linker is known to generate glue.
15689 @item -mtls-markers
15690 @itemx -mno-tls-markers
15691 @opindex mtls-markers
15692 @opindex mno-tls-markers
15693 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
15694 specifying the function argument. The relocation allows ld to
15695 reliably associate function call with argument setup instructions for
15696 TLS optimization, which in turn allows gcc to better schedule the
15701 Adds support for multithreading with the @dfn{pthreads} library.
15702 This option sets flags for both the preprocessor and linker.
15707 This option will enable GCC to use the reciprocal estimate and
15708 reciprocal square root estimate instructions with additional
15709 Newton-Raphson steps to increase precision instead of doing a divide or
15710 square root and divide for floating point arguments. You should use
15711 the @option{-ffast-math} option when using @option{-mrecip} (or at
15712 least @option{-funsafe-math-optimizations},
15713 @option{-finite-math-only}, @option{-freciprocal-math} and
15714 @option{-fno-trapping-math}). Note that while the throughput of the
15715 sequence is generally higher than the throughput of the non-reciprocal
15716 instruction, the precision of the sequence can be decreased by up to 2
15717 ulp (i.e. the inverse of 1.0 equals 0.99999994) for reciprocal square
15720 @item -mrecip=@var{opt}
15721 @opindex mrecip=opt
15722 This option allows to control which reciprocal estimate instructions
15723 may be used. @var{opt} is a comma separated list of options, that may
15724 be preceeded by a @code{!} to invert the option:
15725 @code{all}: enable all estimate instructions,
15726 @code{default}: enable the default instructions, equvalent to @option{-mrecip},
15727 @code{none}: disable all estimate instructions, equivalent to @option{-mno-recip};
15728 @code{div}: enable the reciprocal approximation instructions for both single and double precision;
15729 @code{divf}: enable the single precision reciprocal approximation instructions;
15730 @code{divd}: enable the double precision reciprocal approximation instructions;
15731 @code{rsqrt}: enable the reciprocal square root approximation instructions for both single and double precision;
15732 @code{rsqrtf}: enable the single precision reciprocal square root approximation instructions;
15733 @code{rsqrtd}: enable the double precision reciprocal square root approximation instructions;
15735 So for example, @option{-mrecip=all,!rsqrtd} would enable the
15736 all of the reciprocal estimate instructions, except for the
15737 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
15738 which handle the double precision reciprocal square root calculations.
15740 @item -mrecip-precision
15741 @itemx -mno-recip-precision
15742 @opindex mrecip-precision
15743 Assume (do not assume) that the reciprocal estimate instructions
15744 provide higher precision estimates than is mandated by the powerpc
15745 ABI. Selecting @option{-mcpu=power6} or @option{-mcpu=power7}
15746 automatically selects @option{-mrecip-precision}. The double
15747 precision square root estimate instructions are not generated by
15748 default on low precision machines, since they do not provide an
15749 estimate that converges after three steps.
15753 @subsection RX Options
15756 These command line options are defined for RX targets:
15759 @item -m64bit-doubles
15760 @itemx -m32bit-doubles
15761 @opindex m64bit-doubles
15762 @opindex m32bit-doubles
15763 Make the @code{double} data type be 64-bits (@option{-m64bit-doubles})
15764 or 32-bits (@option{-m32bit-doubles}) in size. The default is
15765 @option{-m32bit-doubles}. @emph{Note} RX floating point hardware only
15766 works on 32-bit values, which is why the default is
15767 @option{-m32bit-doubles}.
15773 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
15774 floating point hardware. The default is enabled for the @var{RX600}
15775 series and disabled for the @var{RX200} series.
15777 Floating point instructions will only be generated for 32-bit floating
15778 point values however, so if the @option{-m64bit-doubles} option is in
15779 use then the FPU hardware will not be used for doubles.
15781 @emph{Note} If the @option{-fpu} option is enabled then
15782 @option{-funsafe-math-optimizations} is also enabled automatically.
15783 This is because the RX FPU instructions are themselves unsafe.
15785 @item -mcpu=@var{name}
15786 @itemx -patch=@var{name}
15789 Selects the type of RX CPU to be targeted. Currently three types are
15790 supported, the generic @var{RX600} and @var{RX200} series hardware and
15791 the specific @var{RX610} cpu. The default is @var{RX600}.
15793 The only difference between @var{RX600} and @var{RX610} is that the
15794 @var{RX610} does not support the @code{MVTIPL} instruction.
15796 The @var{RX200} series does not have a hardware floating point unit
15797 and so @option{-nofpu} is enabled by default when this type is
15800 @item -mbig-endian-data
15801 @itemx -mlittle-endian-data
15802 @opindex mbig-endian-data
15803 @opindex mlittle-endian-data
15804 Store data (but not code) in the big-endian format. The default is
15805 @option{-mlittle-endian-data}, ie to store data in the little endian
15808 @item -msmall-data-limit=@var{N}
15809 @opindex msmall-data-limit
15810 Specifies the maximum size in bytes of global and static variables
15811 which can be placed into the small data area. Using the small data
15812 area can lead to smaller and faster code, but the size of area is
15813 limited and it is up to the programmer to ensure that the area does
15814 not overflow. Also when the small data area is used one of the RX's
15815 registers (@code{r13}) is reserved for use pointing to this area, so
15816 it is no longer available for use by the compiler. This could result
15817 in slower and/or larger code if variables which once could have been
15818 held in @code{r13} are now pushed onto the stack.
15820 Note, common variables (variables which have not been initialised) and
15821 constants are not placed into the small data area as they are assigned
15822 to other sections in the output executable.
15824 The default value is zero, which disables this feature. Note, this
15825 feature is not enabled by default with higher optimization levels
15826 (@option{-O2} etc) because of the potentially detrimental effects of
15827 reserving register @code{r13}. It is up to the programmer to
15828 experiment and discover whether this feature is of benefit to their
15835 Use the simulator runtime. The default is to use the libgloss board
15838 @item -mas100-syntax
15839 @itemx -mno-as100-syntax
15840 @opindex mas100-syntax
15841 @opindex mno-as100-syntax
15842 When generating assembler output use a syntax that is compatible with
15843 Renesas's AS100 assembler. This syntax can also be handled by the GAS
15844 assembler but it has some restrictions so generating it is not the
15847 @item -mmax-constant-size=@var{N}
15848 @opindex mmax-constant-size
15849 Specifies the maximum size, in bytes, of a constant that can be used as
15850 an operand in a RX instruction. Although the RX instruction set does
15851 allow constants of up to 4 bytes in length to be used in instructions,
15852 a longer value equates to a longer instruction. Thus in some
15853 circumstances it can be beneficial to restrict the size of constants
15854 that are used in instructions. Constants that are too big are instead
15855 placed into a constant pool and referenced via register indirection.
15857 The value @var{N} can be between 0 and 4. A value of 0 (the default)
15858 or 4 means that constants of any size are allowed.
15862 Enable linker relaxation. Linker relaxation is a process whereby the
15863 linker will attempt to reduce the size of a program by finding shorter
15864 versions of various instructions. Disabled by default.
15866 @item -mint-register=@var{N}
15867 @opindex mint-register
15868 Specify the number of registers to reserve for fast interrupt handler
15869 functions. The value @var{N} can be between 0 and 4. A value of 1
15870 means that register @code{r13} will be reserved for the exclusive use
15871 of fast interrupt handlers. A value of 2 reserves @code{r13} and
15872 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
15873 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
15874 A value of 0, the default, does not reserve any registers.
15876 @item -msave-acc-in-interrupts
15877 @opindex msave-acc-in-interrupts
15878 Specifies that interrupt handler functions should preserve the
15879 accumulator register. This is only necessary if normal code might use
15880 the accumulator register, for example because it performs 64-bit
15881 multiplications. The default is to ignore the accumulator as this
15882 makes the interrupt handlers faster.
15886 @emph{Note:} The generic GCC command line @option{-ffixed-@var{reg}}
15887 has special significance to the RX port when used with the
15888 @code{interrupt} function attribute. This attribute indicates a
15889 function intended to process fast interrupts. GCC will will ensure
15890 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
15891 and/or @code{r13} and only provided that the normal use of the
15892 corresponding registers have been restricted via the
15893 @option{-ffixed-@var{reg}} or @option{-mint-register} command line
15896 @node S/390 and zSeries Options
15897 @subsection S/390 and zSeries Options
15898 @cindex S/390 and zSeries Options
15900 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
15904 @itemx -msoft-float
15905 @opindex mhard-float
15906 @opindex msoft-float
15907 Use (do not use) the hardware floating-point instructions and registers
15908 for floating-point operations. When @option{-msoft-float} is specified,
15909 functions in @file{libgcc.a} will be used to perform floating-point
15910 operations. When @option{-mhard-float} is specified, the compiler
15911 generates IEEE floating-point instructions. This is the default.
15914 @itemx -mno-hard-dfp
15916 @opindex mno-hard-dfp
15917 Use (do not use) the hardware decimal-floating-point instructions for
15918 decimal-floating-point operations. When @option{-mno-hard-dfp} is
15919 specified, functions in @file{libgcc.a} will be used to perform
15920 decimal-floating-point operations. When @option{-mhard-dfp} is
15921 specified, the compiler generates decimal-floating-point hardware
15922 instructions. This is the default for @option{-march=z9-ec} or higher.
15924 @item -mlong-double-64
15925 @itemx -mlong-double-128
15926 @opindex mlong-double-64
15927 @opindex mlong-double-128
15928 These switches control the size of @code{long double} type. A size
15929 of 64bit makes the @code{long double} type equivalent to the @code{double}
15930 type. This is the default.
15933 @itemx -mno-backchain
15934 @opindex mbackchain
15935 @opindex mno-backchain
15936 Store (do not store) the address of the caller's frame as backchain pointer
15937 into the callee's stack frame.
15938 A backchain may be needed to allow debugging using tools that do not understand
15939 DWARF-2 call frame information.
15940 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
15941 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
15942 the backchain is placed into the topmost word of the 96/160 byte register
15945 In general, code compiled with @option{-mbackchain} is call-compatible with
15946 code compiled with @option{-mmo-backchain}; however, use of the backchain
15947 for debugging purposes usually requires that the whole binary is built with
15948 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
15949 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
15950 to build a linux kernel use @option{-msoft-float}.
15952 The default is to not maintain the backchain.
15954 @item -mpacked-stack
15955 @itemx -mno-packed-stack
15956 @opindex mpacked-stack
15957 @opindex mno-packed-stack
15958 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
15959 specified, the compiler uses the all fields of the 96/160 byte register save
15960 area only for their default purpose; unused fields still take up stack space.
15961 When @option{-mpacked-stack} is specified, register save slots are densely
15962 packed at the top of the register save area; unused space is reused for other
15963 purposes, allowing for more efficient use of the available stack space.
15964 However, when @option{-mbackchain} is also in effect, the topmost word of
15965 the save area is always used to store the backchain, and the return address
15966 register is always saved two words below the backchain.
15968 As long as the stack frame backchain is not used, code generated with
15969 @option{-mpacked-stack} is call-compatible with code generated with
15970 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
15971 S/390 or zSeries generated code that uses the stack frame backchain at run
15972 time, not just for debugging purposes. Such code is not call-compatible
15973 with code compiled with @option{-mpacked-stack}. Also, note that the
15974 combination of @option{-mbackchain},
15975 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
15976 to build a linux kernel use @option{-msoft-float}.
15978 The default is to not use the packed stack layout.
15981 @itemx -mno-small-exec
15982 @opindex msmall-exec
15983 @opindex mno-small-exec
15984 Generate (or do not generate) code using the @code{bras} instruction
15985 to do subroutine calls.
15986 This only works reliably if the total executable size does not
15987 exceed 64k. The default is to use the @code{basr} instruction instead,
15988 which does not have this limitation.
15994 When @option{-m31} is specified, generate code compliant to the
15995 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
15996 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
15997 particular to generate 64-bit instructions. For the @samp{s390}
15998 targets, the default is @option{-m31}, while the @samp{s390x}
15999 targets default to @option{-m64}.
16005 When @option{-mzarch} is specified, generate code using the
16006 instructions available on z/Architecture.
16007 When @option{-mesa} is specified, generate code using the
16008 instructions available on ESA/390. Note that @option{-mesa} is
16009 not possible with @option{-m64}.
16010 When generating code compliant to the GNU/Linux for S/390 ABI,
16011 the default is @option{-mesa}. When generating code compliant
16012 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
16018 Generate (or do not generate) code using the @code{mvcle} instruction
16019 to perform block moves. When @option{-mno-mvcle} is specified,
16020 use a @code{mvc} loop instead. This is the default unless optimizing for
16027 Print (or do not print) additional debug information when compiling.
16028 The default is to not print debug information.
16030 @item -march=@var{cpu-type}
16032 Generate code that will run on @var{cpu-type}, which is the name of a system
16033 representing a certain processor type. Possible values for
16034 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
16035 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
16036 When generating code using the instructions available on z/Architecture,
16037 the default is @option{-march=z900}. Otherwise, the default is
16038 @option{-march=g5}.
16040 @item -mtune=@var{cpu-type}
16042 Tune to @var{cpu-type} everything applicable about the generated code,
16043 except for the ABI and the set of available instructions.
16044 The list of @var{cpu-type} values is the same as for @option{-march}.
16045 The default is the value used for @option{-march}.
16048 @itemx -mno-tpf-trace
16049 @opindex mtpf-trace
16050 @opindex mno-tpf-trace
16051 Generate code that adds (does not add) in TPF OS specific branches to trace
16052 routines in the operating system. This option is off by default, even
16053 when compiling for the TPF OS@.
16056 @itemx -mno-fused-madd
16057 @opindex mfused-madd
16058 @opindex mno-fused-madd
16059 Generate code that uses (does not use) the floating point multiply and
16060 accumulate instructions. These instructions are generated by default if
16061 hardware floating point is used.
16063 @item -mwarn-framesize=@var{framesize}
16064 @opindex mwarn-framesize
16065 Emit a warning if the current function exceeds the given frame size. Because
16066 this is a compile time check it doesn't need to be a real problem when the program
16067 runs. It is intended to identify functions which most probably cause
16068 a stack overflow. It is useful to be used in an environment with limited stack
16069 size e.g.@: the linux kernel.
16071 @item -mwarn-dynamicstack
16072 @opindex mwarn-dynamicstack
16073 Emit a warning if the function calls alloca or uses dynamically
16074 sized arrays. This is generally a bad idea with a limited stack size.
16076 @item -mstack-guard=@var{stack-guard}
16077 @itemx -mstack-size=@var{stack-size}
16078 @opindex mstack-guard
16079 @opindex mstack-size
16080 If these options are provided the s390 back end emits additional instructions in
16081 the function prologue which trigger a trap if the stack size is @var{stack-guard}
16082 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
16083 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
16084 the frame size of the compiled function is chosen.
16085 These options are intended to be used to help debugging stack overflow problems.
16086 The additionally emitted code causes only little overhead and hence can also be
16087 used in production like systems without greater performance degradation. The given
16088 values have to be exact powers of 2 and @var{stack-size} has to be greater than
16089 @var{stack-guard} without exceeding 64k.
16090 In order to be efficient the extra code makes the assumption that the stack starts
16091 at an address aligned to the value given by @var{stack-size}.
16092 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
16095 @node Score Options
16096 @subsection Score Options
16097 @cindex Score Options
16099 These options are defined for Score implementations:
16104 Compile code for big endian mode. This is the default.
16108 Compile code for little endian mode.
16112 Disable generate bcnz instruction.
16116 Enable generate unaligned load and store instruction.
16120 Enable the use of multiply-accumulate instructions. Disabled by default.
16124 Specify the SCORE5 as the target architecture.
16128 Specify the SCORE5U of the target architecture.
16132 Specify the SCORE7 as the target architecture. This is the default.
16136 Specify the SCORE7D as the target architecture.
16140 @subsection SH Options
16142 These @samp{-m} options are defined for the SH implementations:
16147 Generate code for the SH1.
16151 Generate code for the SH2.
16154 Generate code for the SH2e.
16158 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
16159 that the floating-point unit is not used.
16161 @item -m2a-single-only
16162 @opindex m2a-single-only
16163 Generate code for the SH2a-FPU, in such a way that no double-precision
16164 floating point operations are used.
16167 @opindex m2a-single
16168 Generate code for the SH2a-FPU assuming the floating-point unit is in
16169 single-precision mode by default.
16173 Generate code for the SH2a-FPU assuming the floating-point unit is in
16174 double-precision mode by default.
16178 Generate code for the SH3.
16182 Generate code for the SH3e.
16186 Generate code for the SH4 without a floating-point unit.
16188 @item -m4-single-only
16189 @opindex m4-single-only
16190 Generate code for the SH4 with a floating-point unit that only
16191 supports single-precision arithmetic.
16195 Generate code for the SH4 assuming the floating-point unit is in
16196 single-precision mode by default.
16200 Generate code for the SH4.
16204 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
16205 floating-point unit is not used.
16207 @item -m4a-single-only
16208 @opindex m4a-single-only
16209 Generate code for the SH4a, in such a way that no double-precision
16210 floating point operations are used.
16213 @opindex m4a-single
16214 Generate code for the SH4a assuming the floating-point unit is in
16215 single-precision mode by default.
16219 Generate code for the SH4a.
16223 Same as @option{-m4a-nofpu}, except that it implicitly passes
16224 @option{-dsp} to the assembler. GCC doesn't generate any DSP
16225 instructions at the moment.
16229 Compile code for the processor in big endian mode.
16233 Compile code for the processor in little endian mode.
16237 Align doubles at 64-bit boundaries. Note that this changes the calling
16238 conventions, and thus some functions from the standard C library will
16239 not work unless you recompile it first with @option{-mdalign}.
16243 Shorten some address references at link time, when possible; uses the
16244 linker option @option{-relax}.
16248 Use 32-bit offsets in @code{switch} tables. The default is to use
16253 Enable the use of bit manipulation instructions on SH2A.
16257 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
16258 alignment constraints.
16262 Comply with the calling conventions defined by Renesas.
16266 Comply with the calling conventions defined by Renesas.
16270 Comply with the calling conventions defined for GCC before the Renesas
16271 conventions were available. This option is the default for all
16272 targets of the SH toolchain except for @samp{sh-symbianelf}.
16275 @opindex mnomacsave
16276 Mark the @code{MAC} register as call-clobbered, even if
16277 @option{-mhitachi} is given.
16281 Increase IEEE-compliance of floating-point code.
16282 At the moment, this is equivalent to @option{-fno-finite-math-only}.
16283 When generating 16 bit SH opcodes, getting IEEE-conforming results for
16284 comparisons of NANs / infinities incurs extra overhead in every
16285 floating point comparison, therefore the default is set to
16286 @option{-ffinite-math-only}.
16288 @item -minline-ic_invalidate
16289 @opindex minline-ic_invalidate
16290 Inline code to invalidate instruction cache entries after setting up
16291 nested function trampolines.
16292 This option has no effect if -musermode is in effect and the selected
16293 code generation option (e.g. -m4) does not allow the use of the icbi
16295 If the selected code generation option does not allow the use of the icbi
16296 instruction, and -musermode is not in effect, the inlined code will
16297 manipulate the instruction cache address array directly with an associative
16298 write. This not only requires privileged mode, but it will also
16299 fail if the cache line had been mapped via the TLB and has become unmapped.
16303 Dump instruction size and location in the assembly code.
16306 @opindex mpadstruct
16307 This option is deprecated. It pads structures to multiple of 4 bytes,
16308 which is incompatible with the SH ABI@.
16312 Optimize for space instead of speed. Implied by @option{-Os}.
16315 @opindex mprefergot
16316 When generating position-independent code, emit function calls using
16317 the Global Offset Table instead of the Procedure Linkage Table.
16321 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
16322 if the inlined code would not work in user mode.
16323 This is the default when the target is @code{sh-*-linux*}.
16325 @item -multcost=@var{number}
16326 @opindex multcost=@var{number}
16327 Set the cost to assume for a multiply insn.
16329 @item -mdiv=@var{strategy}
16330 @opindex mdiv=@var{strategy}
16331 Set the division strategy to use for SHmedia code. @var{strategy} must be
16332 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
16333 inv:call2, inv:fp .
16334 "fp" performs the operation in floating point. This has a very high latency,
16335 but needs only a few instructions, so it might be a good choice if
16336 your code has enough easily exploitable ILP to allow the compiler to
16337 schedule the floating point instructions together with other instructions.
16338 Division by zero causes a floating point exception.
16339 "inv" uses integer operations to calculate the inverse of the divisor,
16340 and then multiplies the dividend with the inverse. This strategy allows
16341 cse and hoisting of the inverse calculation. Division by zero calculates
16342 an unspecified result, but does not trap.
16343 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
16344 have been found, or if the entire operation has been hoisted to the same
16345 place, the last stages of the inverse calculation are intertwined with the
16346 final multiply to reduce the overall latency, at the expense of using a few
16347 more instructions, and thus offering fewer scheduling opportunities with
16349 "call" calls a library function that usually implements the inv:minlat
16351 This gives high code density for m5-*media-nofpu compilations.
16352 "call2" uses a different entry point of the same library function, where it
16353 assumes that a pointer to a lookup table has already been set up, which
16354 exposes the pointer load to cse / code hoisting optimizations.
16355 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
16356 code generation, but if the code stays unoptimized, revert to the "call",
16357 "call2", or "fp" strategies, respectively. Note that the
16358 potentially-trapping side effect of division by zero is carried by a
16359 separate instruction, so it is possible that all the integer instructions
16360 are hoisted out, but the marker for the side effect stays where it is.
16361 A recombination to fp operations or a call is not possible in that case.
16362 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
16363 that the inverse calculation was nor separated from the multiply, they speed
16364 up division where the dividend fits into 20 bits (plus sign where applicable),
16365 by inserting a test to skip a number of operations in this case; this test
16366 slows down the case of larger dividends. inv20u assumes the case of a such
16367 a small dividend to be unlikely, and inv20l assumes it to be likely.
16369 @item -maccumulate-outgoing-args
16370 @opindex maccumulate-outgoing-args
16371 Reserve space once for outgoing arguments in the function prologue rather
16372 than around each call. Generally beneficial for performance and size. Also
16373 needed for unwinding to avoid changing the stack frame around conditional code.
16375 @item -mdivsi3_libfunc=@var{name}
16376 @opindex mdivsi3_libfunc=@var{name}
16377 Set the name of the library function used for 32 bit signed division to
16378 @var{name}. This only affect the name used in the call and inv:call
16379 division strategies, and the compiler will still expect the same
16380 sets of input/output/clobbered registers as if this option was not present.
16382 @item -mfixed-range=@var{register-range}
16383 @opindex mfixed-range
16384 Generate code treating the given register range as fixed registers.
16385 A fixed register is one that the register allocator can not use. This is
16386 useful when compiling kernel code. A register range is specified as
16387 two registers separated by a dash. Multiple register ranges can be
16388 specified separated by a comma.
16390 @item -madjust-unroll
16391 @opindex madjust-unroll
16392 Throttle unrolling to avoid thrashing target registers.
16393 This option only has an effect if the gcc code base supports the
16394 TARGET_ADJUST_UNROLL_MAX target hook.
16396 @item -mindexed-addressing
16397 @opindex mindexed-addressing
16398 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
16399 This is only safe if the hardware and/or OS implement 32 bit wrap-around
16400 semantics for the indexed addressing mode. The architecture allows the
16401 implementation of processors with 64 bit MMU, which the OS could use to
16402 get 32 bit addressing, but since no current hardware implementation supports
16403 this or any other way to make the indexed addressing mode safe to use in
16404 the 32 bit ABI, the default is -mno-indexed-addressing.
16406 @item -mgettrcost=@var{number}
16407 @opindex mgettrcost=@var{number}
16408 Set the cost assumed for the gettr instruction to @var{number}.
16409 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
16413 Assume pt* instructions won't trap. This will generally generate better
16414 scheduled code, but is unsafe on current hardware. The current architecture
16415 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
16416 This has the unintentional effect of making it unsafe to schedule ptabs /
16417 ptrel before a branch, or hoist it out of a loop. For example,
16418 __do_global_ctors, a part of libgcc that runs constructors at program
16419 startup, calls functions in a list which is delimited by @minus{}1. With the
16420 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
16421 That means that all the constructors will be run a bit quicker, but when
16422 the loop comes to the end of the list, the program crashes because ptabs
16423 loads @minus{}1 into a target register. Since this option is unsafe for any
16424 hardware implementing the current architecture specification, the default
16425 is -mno-pt-fixed. Unless the user specifies a specific cost with
16426 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
16427 this deters register allocation using target registers for storing
16430 @item -minvalid-symbols
16431 @opindex minvalid-symbols
16432 Assume symbols might be invalid. Ordinary function symbols generated by
16433 the compiler will always be valid to load with movi/shori/ptabs or
16434 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
16435 to generate symbols that will cause ptabs / ptrel to trap.
16436 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
16437 It will then prevent cross-basic-block cse, hoisting and most scheduling
16438 of symbol loads. The default is @option{-mno-invalid-symbols}.
16441 @node SPARC Options
16442 @subsection SPARC Options
16443 @cindex SPARC options
16445 These @samp{-m} options are supported on the SPARC:
16448 @item -mno-app-regs
16450 @opindex mno-app-regs
16452 Specify @option{-mapp-regs} to generate output using the global registers
16453 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
16456 To be fully SVR4 ABI compliant at the cost of some performance loss,
16457 specify @option{-mno-app-regs}. You should compile libraries and system
16458 software with this option.
16461 @itemx -mhard-float
16463 @opindex mhard-float
16464 Generate output containing floating point instructions. This is the
16468 @itemx -msoft-float
16470 @opindex msoft-float
16471 Generate output containing library calls for floating point.
16472 @strong{Warning:} the requisite libraries are not available for all SPARC
16473 targets. Normally the facilities of the machine's usual C compiler are
16474 used, but this cannot be done directly in cross-compilation. You must make
16475 your own arrangements to provide suitable library functions for
16476 cross-compilation. The embedded targets @samp{sparc-*-aout} and
16477 @samp{sparclite-*-*} do provide software floating point support.
16479 @option{-msoft-float} changes the calling convention in the output file;
16480 therefore, it is only useful if you compile @emph{all} of a program with
16481 this option. In particular, you need to compile @file{libgcc.a}, the
16482 library that comes with GCC, with @option{-msoft-float} in order for
16485 @item -mhard-quad-float
16486 @opindex mhard-quad-float
16487 Generate output containing quad-word (long double) floating point
16490 @item -msoft-quad-float
16491 @opindex msoft-quad-float
16492 Generate output containing library calls for quad-word (long double)
16493 floating point instructions. The functions called are those specified
16494 in the SPARC ABI@. This is the default.
16496 As of this writing, there are no SPARC implementations that have hardware
16497 support for the quad-word floating point instructions. They all invoke
16498 a trap handler for one of these instructions, and then the trap handler
16499 emulates the effect of the instruction. Because of the trap handler overhead,
16500 this is much slower than calling the ABI library routines. Thus the
16501 @option{-msoft-quad-float} option is the default.
16503 @item -mno-unaligned-doubles
16504 @itemx -munaligned-doubles
16505 @opindex mno-unaligned-doubles
16506 @opindex munaligned-doubles
16507 Assume that doubles have 8 byte alignment. This is the default.
16509 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
16510 alignment only if they are contained in another type, or if they have an
16511 absolute address. Otherwise, it assumes they have 4 byte alignment.
16512 Specifying this option avoids some rare compatibility problems with code
16513 generated by other compilers. It is not the default because it results
16514 in a performance loss, especially for floating point code.
16516 @item -mno-faster-structs
16517 @itemx -mfaster-structs
16518 @opindex mno-faster-structs
16519 @opindex mfaster-structs
16520 With @option{-mfaster-structs}, the compiler assumes that structures
16521 should have 8 byte alignment. This enables the use of pairs of
16522 @code{ldd} and @code{std} instructions for copies in structure
16523 assignment, in place of twice as many @code{ld} and @code{st} pairs.
16524 However, the use of this changed alignment directly violates the SPARC
16525 ABI@. Thus, it's intended only for use on targets where the developer
16526 acknowledges that their resulting code will not be directly in line with
16527 the rules of the ABI@.
16529 @item -mimpure-text
16530 @opindex mimpure-text
16531 @option{-mimpure-text}, used in addition to @option{-shared}, tells
16532 the compiler to not pass @option{-z text} to the linker when linking a
16533 shared object. Using this option, you can link position-dependent
16534 code into a shared object.
16536 @option{-mimpure-text} suppresses the ``relocations remain against
16537 allocatable but non-writable sections'' linker error message.
16538 However, the necessary relocations will trigger copy-on-write, and the
16539 shared object is not actually shared across processes. Instead of
16540 using @option{-mimpure-text}, you should compile all source code with
16541 @option{-fpic} or @option{-fPIC}.
16543 This option is only available on SunOS and Solaris.
16545 @item -mcpu=@var{cpu_type}
16547 Set the instruction set, register set, and instruction scheduling parameters
16548 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
16549 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
16550 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
16551 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
16552 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
16554 Default instruction scheduling parameters are used for values that select
16555 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
16556 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
16558 Here is a list of each supported architecture and their supported
16563 v8: supersparc, hypersparc
16564 sparclite: f930, f934, sparclite86x
16566 v9: ultrasparc, ultrasparc3, niagara, niagara2
16569 By default (unless configured otherwise), GCC generates code for the V7
16570 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
16571 additionally optimizes it for the Cypress CY7C602 chip, as used in the
16572 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
16573 SPARCStation 1, 2, IPX etc.
16575 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
16576 architecture. The only difference from V7 code is that the compiler emits
16577 the integer multiply and integer divide instructions which exist in SPARC-V8
16578 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
16579 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
16582 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
16583 the SPARC architecture. This adds the integer multiply, integer divide step
16584 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
16585 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
16586 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
16587 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
16588 MB86934 chip, which is the more recent SPARClite with FPU@.
16590 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
16591 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
16592 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
16593 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
16594 optimizes it for the TEMIC SPARClet chip.
16596 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
16597 architecture. This adds 64-bit integer and floating-point move instructions,
16598 3 additional floating-point condition code registers and conditional move
16599 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
16600 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
16601 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
16602 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
16603 @option{-mcpu=niagara}, the compiler additionally optimizes it for
16604 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
16605 additionally optimizes it for Sun UltraSPARC T2 chips.
16607 @item -mtune=@var{cpu_type}
16609 Set the instruction scheduling parameters for machine type
16610 @var{cpu_type}, but do not set the instruction set or register set that the
16611 option @option{-mcpu=@var{cpu_type}} would.
16613 The same values for @option{-mcpu=@var{cpu_type}} can be used for
16614 @option{-mtune=@var{cpu_type}}, but the only useful values are those
16615 that select a particular cpu implementation. Those are @samp{cypress},
16616 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
16617 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
16618 @samp{ultrasparc3}, @samp{niagara}, and @samp{niagara2}.
16623 @opindex mno-v8plus
16624 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
16625 difference from the V8 ABI is that the global and out registers are
16626 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
16627 mode for all SPARC-V9 processors.
16633 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
16634 Visual Instruction Set extensions. The default is @option{-mno-vis}.
16637 These @samp{-m} options are supported in addition to the above
16638 on SPARC-V9 processors in 64-bit environments:
16641 @item -mlittle-endian
16642 @opindex mlittle-endian
16643 Generate code for a processor running in little-endian mode. It is only
16644 available for a few configurations and most notably not on Solaris and Linux.
16650 Generate code for a 32-bit or 64-bit environment.
16651 The 32-bit environment sets int, long and pointer to 32 bits.
16652 The 64-bit environment sets int to 32 bits and long and pointer
16655 @item -mcmodel=medlow
16656 @opindex mcmodel=medlow
16657 Generate code for the Medium/Low code model: 64-bit addresses, programs
16658 must be linked in the low 32 bits of memory. Programs can be statically
16659 or dynamically linked.
16661 @item -mcmodel=medmid
16662 @opindex mcmodel=medmid
16663 Generate code for the Medium/Middle code model: 64-bit addresses, programs
16664 must be linked in the low 44 bits of memory, the text and data segments must
16665 be less than 2GB in size and the data segment must be located within 2GB of
16668 @item -mcmodel=medany
16669 @opindex mcmodel=medany
16670 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
16671 may be linked anywhere in memory, the text and data segments must be less
16672 than 2GB in size and the data segment must be located within 2GB of the
16675 @item -mcmodel=embmedany
16676 @opindex mcmodel=embmedany
16677 Generate code for the Medium/Anywhere code model for embedded systems:
16678 64-bit addresses, the text and data segments must be less than 2GB in
16679 size, both starting anywhere in memory (determined at link time). The
16680 global register %g4 points to the base of the data segment. Programs
16681 are statically linked and PIC is not supported.
16684 @itemx -mno-stack-bias
16685 @opindex mstack-bias
16686 @opindex mno-stack-bias
16687 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
16688 frame pointer if present, are offset by @minus{}2047 which must be added back
16689 when making stack frame references. This is the default in 64-bit mode.
16690 Otherwise, assume no such offset is present.
16693 These switches are supported in addition to the above on Solaris:
16698 Add support for multithreading using the Solaris threads library. This
16699 option sets flags for both the preprocessor and linker. This option does
16700 not affect the thread safety of object code produced by the compiler or
16701 that of libraries supplied with it.
16705 Add support for multithreading using the POSIX threads library. This
16706 option sets flags for both the preprocessor and linker. This option does
16707 not affect the thread safety of object code produced by the compiler or
16708 that of libraries supplied with it.
16712 This is a synonym for @option{-pthreads}.
16716 @subsection SPU Options
16717 @cindex SPU options
16719 These @samp{-m} options are supported on the SPU:
16723 @itemx -merror-reloc
16724 @opindex mwarn-reloc
16725 @opindex merror-reloc
16727 The loader for SPU does not handle dynamic relocations. By default, GCC
16728 will give an error when it generates code that requires a dynamic
16729 relocation. @option{-mno-error-reloc} disables the error,
16730 @option{-mwarn-reloc} will generate a warning instead.
16733 @itemx -munsafe-dma
16735 @opindex munsafe-dma
16737 Instructions which initiate or test completion of DMA must not be
16738 reordered with respect to loads and stores of the memory which is being
16739 accessed. Users typically address this problem using the volatile
16740 keyword, but that can lead to inefficient code in places where the
16741 memory is known to not change. Rather than mark the memory as volatile
16742 we treat the DMA instructions as potentially effecting all memory. With
16743 @option{-munsafe-dma} users must use the volatile keyword to protect
16746 @item -mbranch-hints
16747 @opindex mbranch-hints
16749 By default, GCC will generate a branch hint instruction to avoid
16750 pipeline stalls for always taken or probably taken branches. A hint
16751 will not be generated closer than 8 instructions away from its branch.
16752 There is little reason to disable them, except for debugging purposes,
16753 or to make an object a little bit smaller.
16757 @opindex msmall-mem
16758 @opindex mlarge-mem
16760 By default, GCC generates code assuming that addresses are never larger
16761 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
16762 a full 32 bit address.
16767 By default, GCC links against startup code that assumes the SPU-style
16768 main function interface (which has an unconventional parameter list).
16769 With @option{-mstdmain}, GCC will link your program against startup
16770 code that assumes a C99-style interface to @code{main}, including a
16771 local copy of @code{argv} strings.
16773 @item -mfixed-range=@var{register-range}
16774 @opindex mfixed-range
16775 Generate code treating the given register range as fixed registers.
16776 A fixed register is one that the register allocator can not use. This is
16777 useful when compiling kernel code. A register range is specified as
16778 two registers separated by a dash. Multiple register ranges can be
16779 specified separated by a comma.
16785 Compile code assuming that pointers to the PPU address space accessed
16786 via the @code{__ea} named address space qualifier are either 32 or 64
16787 bits wide. The default is 32 bits. As this is an ABI changing option,
16788 all object code in an executable must be compiled with the same setting.
16790 @item -maddress-space-conversion
16791 @itemx -mno-address-space-conversion
16792 @opindex maddress-space-conversion
16793 @opindex mno-address-space-conversion
16794 Allow/disallow treating the @code{__ea} address space as superset
16795 of the generic address space. This enables explicit type casts
16796 between @code{__ea} and generic pointer as well as implicit
16797 conversions of generic pointers to @code{__ea} pointers. The
16798 default is to allow address space pointer conversions.
16800 @item -mcache-size=@var{cache-size}
16801 @opindex mcache-size
16802 This option controls the version of libgcc that the compiler links to an
16803 executable and selects a software-managed cache for accessing variables
16804 in the @code{__ea} address space with a particular cache size. Possible
16805 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
16806 and @samp{128}. The default cache size is 64KB.
16808 @item -matomic-updates
16809 @itemx -mno-atomic-updates
16810 @opindex matomic-updates
16811 @opindex mno-atomic-updates
16812 This option controls the version of libgcc that the compiler links to an
16813 executable and selects whether atomic updates to the software-managed
16814 cache of PPU-side variables are used. If you use atomic updates, changes
16815 to a PPU variable from SPU code using the @code{__ea} named address space
16816 qualifier will not interfere with changes to other PPU variables residing
16817 in the same cache line from PPU code. If you do not use atomic updates,
16818 such interference may occur; however, writing back cache lines will be
16819 more efficient. The default behavior is to use atomic updates.
16822 @itemx -mdual-nops=@var{n}
16823 @opindex mdual-nops
16824 By default, GCC will insert nops to increase dual issue when it expects
16825 it to increase performance. @var{n} can be a value from 0 to 10. A
16826 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
16827 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
16829 @item -mhint-max-nops=@var{n}
16830 @opindex mhint-max-nops
16831 Maximum number of nops to insert for a branch hint. A branch hint must
16832 be at least 8 instructions away from the branch it is effecting. GCC
16833 will insert up to @var{n} nops to enforce this, otherwise it will not
16834 generate the branch hint.
16836 @item -mhint-max-distance=@var{n}
16837 @opindex mhint-max-distance
16838 The encoding of the branch hint instruction limits the hint to be within
16839 256 instructions of the branch it is effecting. By default, GCC makes
16840 sure it is within 125.
16843 @opindex msafe-hints
16844 Work around a hardware bug which causes the SPU to stall indefinitely.
16845 By default, GCC will insert the @code{hbrp} instruction to make sure
16846 this stall won't happen.
16850 @node System V Options
16851 @subsection Options for System V
16853 These additional options are available on System V Release 4 for
16854 compatibility with other compilers on those systems:
16859 Create a shared object.
16860 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
16864 Identify the versions of each tool used by the compiler, in a
16865 @code{.ident} assembler directive in the output.
16869 Refrain from adding @code{.ident} directives to the output file (this is
16872 @item -YP,@var{dirs}
16874 Search the directories @var{dirs}, and no others, for libraries
16875 specified with @option{-l}.
16877 @item -Ym,@var{dir}
16879 Look in the directory @var{dir} to find the M4 preprocessor.
16880 The assembler uses this option.
16881 @c This is supposed to go with a -Yd for predefined M4 macro files, but
16882 @c the generic assembler that comes with Solaris takes just -Ym.
16886 @subsection V850 Options
16887 @cindex V850 Options
16889 These @samp{-m} options are defined for V850 implementations:
16893 @itemx -mno-long-calls
16894 @opindex mlong-calls
16895 @opindex mno-long-calls
16896 Treat all calls as being far away (near). If calls are assumed to be
16897 far away, the compiler will always load the functions address up into a
16898 register, and call indirect through the pointer.
16904 Do not optimize (do optimize) basic blocks that use the same index
16905 pointer 4 or more times to copy pointer into the @code{ep} register, and
16906 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
16907 option is on by default if you optimize.
16909 @item -mno-prolog-function
16910 @itemx -mprolog-function
16911 @opindex mno-prolog-function
16912 @opindex mprolog-function
16913 Do not use (do use) external functions to save and restore registers
16914 at the prologue and epilogue of a function. The external functions
16915 are slower, but use less code space if more than one function saves
16916 the same number of registers. The @option{-mprolog-function} option
16917 is on by default if you optimize.
16921 Try to make the code as small as possible. At present, this just turns
16922 on the @option{-mep} and @option{-mprolog-function} options.
16924 @item -mtda=@var{n}
16926 Put static or global variables whose size is @var{n} bytes or less into
16927 the tiny data area that register @code{ep} points to. The tiny data
16928 area can hold up to 256 bytes in total (128 bytes for byte references).
16930 @item -msda=@var{n}
16932 Put static or global variables whose size is @var{n} bytes or less into
16933 the small data area that register @code{gp} points to. The small data
16934 area can hold up to 64 kilobytes.
16936 @item -mzda=@var{n}
16938 Put static or global variables whose size is @var{n} bytes or less into
16939 the first 32 kilobytes of memory.
16943 Specify that the target processor is the V850.
16946 @opindex mbig-switch
16947 Generate code suitable for big switch tables. Use this option only if
16948 the assembler/linker complain about out of range branches within a switch
16953 This option will cause r2 and r5 to be used in the code generated by
16954 the compiler. This setting is the default.
16956 @item -mno-app-regs
16957 @opindex mno-app-regs
16958 This option will cause r2 and r5 to be treated as fixed registers.
16962 Specify that the target processor is the V850E1. The preprocessor
16963 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
16964 this option is used.
16968 Specify that the target processor is the V850E@. The preprocessor
16969 constant @samp{__v850e__} will be defined if this option is used.
16971 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
16972 are defined then a default target processor will be chosen and the
16973 relevant @samp{__v850*__} preprocessor constant will be defined.
16975 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
16976 defined, regardless of which processor variant is the target.
16978 @item -mdisable-callt
16979 @opindex mdisable-callt
16980 This option will suppress generation of the CALLT instruction for the
16981 v850e and v850e1 flavors of the v850 architecture. The default is
16982 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
16987 @subsection VAX Options
16988 @cindex VAX options
16990 These @samp{-m} options are defined for the VAX:
16995 Do not output certain jump instructions (@code{aobleq} and so on)
16996 that the Unix assembler for the VAX cannot handle across long
17001 Do output those jump instructions, on the assumption that you
17002 will assemble with the GNU assembler.
17006 Output code for g-format floating point numbers instead of d-format.
17009 @node VxWorks Options
17010 @subsection VxWorks Options
17011 @cindex VxWorks Options
17013 The options in this section are defined for all VxWorks targets.
17014 Options specific to the target hardware are listed with the other
17015 options for that target.
17020 GCC can generate code for both VxWorks kernels and real time processes
17021 (RTPs). This option switches from the former to the latter. It also
17022 defines the preprocessor macro @code{__RTP__}.
17025 @opindex non-static
17026 Link an RTP executable against shared libraries rather than static
17027 libraries. The options @option{-static} and @option{-shared} can
17028 also be used for RTPs (@pxref{Link Options}); @option{-static}
17035 These options are passed down to the linker. They are defined for
17036 compatibility with Diab.
17039 @opindex Xbind-lazy
17040 Enable lazy binding of function calls. This option is equivalent to
17041 @option{-Wl,-z,now} and is defined for compatibility with Diab.
17045 Disable lazy binding of function calls. This option is the default and
17046 is defined for compatibility with Diab.
17049 @node x86-64 Options
17050 @subsection x86-64 Options
17051 @cindex x86-64 options
17053 These are listed under @xref{i386 and x86-64 Options}.
17055 @node i386 and x86-64 Windows Options
17056 @subsection i386 and x86-64 Windows Options
17057 @cindex i386 and x86-64 Windows Options
17059 These additional options are available for Windows targets:
17064 This option is available for Cygwin and MinGW targets. It
17065 specifies that a console application is to be generated, by
17066 instructing the linker to set the PE header subsystem type
17067 required for console applications.
17068 This is the default behavior for Cygwin and MinGW targets.
17072 This option is available for Cygwin targets. It specifies that
17073 the Cygwin internal interface is to be used for predefined
17074 preprocessor macros, C runtime libraries and related linker
17075 paths and options. For Cygwin targets this is the default behavior.
17076 This option is deprecated and will be removed in a future release.
17079 @opindex mno-cygwin
17080 This option is available for Cygwin targets. It specifies that
17081 the MinGW internal interface is to be used instead of Cygwin's, by
17082 setting MinGW-related predefined macros and linker paths and default
17084 This option is deprecated and will be removed in a future release.
17088 This option is available for Cygwin and MinGW targets. It
17089 specifies that a DLL - a dynamic link library - is to be
17090 generated, enabling the selection of the required runtime
17091 startup object and entry point.
17093 @item -mnop-fun-dllimport
17094 @opindex mnop-fun-dllimport
17095 This option is available for Cygwin and MinGW targets. It
17096 specifies that the dllimport attribute should be ignored.
17100 This option is available for MinGW targets. It specifies
17101 that MinGW-specific thread support is to be used.
17105 This option is available for mingw-w64 targets. It specifies
17106 that the UNICODE macro is getting pre-defined and that the
17107 unicode capable runtime startup code is chosen.
17111 This option is available for Cygwin and MinGW targets. It
17112 specifies that the typical Windows pre-defined macros are to
17113 be set in the pre-processor, but does not influence the choice
17114 of runtime library/startup code.
17118 This option is available for Cygwin and MinGW targets. It
17119 specifies that a GUI application is to be generated by
17120 instructing the linker to set the PE header subsystem type
17123 @item -fno-set-stack-executable
17124 @opindex fno-set-stack-executable
17125 This option is available for MinGW targets. It specifies that
17126 the executable flag for stack used by nested functions isn't
17127 set. This is necessary for binaries running in kernel mode of
17128 Windows, as there the user32 API, which is used to set executable
17129 privileges, isn't available.
17131 @item -mpe-aligned-commons
17132 @opindex mpe-aligned-commons
17133 This option is available for Cygwin and MinGW targets. It
17134 specifies that the GNU extension to the PE file format that
17135 permits the correct alignment of COMMON variables should be
17136 used when generating code. It will be enabled by default if
17137 GCC detects that the target assembler found during configuration
17138 supports the feature.
17141 See also under @ref{i386 and x86-64 Options} for standard options.
17143 @node Xstormy16 Options
17144 @subsection Xstormy16 Options
17145 @cindex Xstormy16 Options
17147 These options are defined for Xstormy16:
17152 Choose startup files and linker script suitable for the simulator.
17155 @node Xtensa Options
17156 @subsection Xtensa Options
17157 @cindex Xtensa Options
17159 These options are supported for Xtensa targets:
17163 @itemx -mno-const16
17165 @opindex mno-const16
17166 Enable or disable use of @code{CONST16} instructions for loading
17167 constant values. The @code{CONST16} instruction is currently not a
17168 standard option from Tensilica. When enabled, @code{CONST16}
17169 instructions are always used in place of the standard @code{L32R}
17170 instructions. The use of @code{CONST16} is enabled by default only if
17171 the @code{L32R} instruction is not available.
17174 @itemx -mno-fused-madd
17175 @opindex mfused-madd
17176 @opindex mno-fused-madd
17177 Enable or disable use of fused multiply/add and multiply/subtract
17178 instructions in the floating-point option. This has no effect if the
17179 floating-point option is not also enabled. Disabling fused multiply/add
17180 and multiply/subtract instructions forces the compiler to use separate
17181 instructions for the multiply and add/subtract operations. This may be
17182 desirable in some cases where strict IEEE 754-compliant results are
17183 required: the fused multiply add/subtract instructions do not round the
17184 intermediate result, thereby producing results with @emph{more} bits of
17185 precision than specified by the IEEE standard. Disabling fused multiply
17186 add/subtract instructions also ensures that the program output is not
17187 sensitive to the compiler's ability to combine multiply and add/subtract
17190 @item -mserialize-volatile
17191 @itemx -mno-serialize-volatile
17192 @opindex mserialize-volatile
17193 @opindex mno-serialize-volatile
17194 When this option is enabled, GCC inserts @code{MEMW} instructions before
17195 @code{volatile} memory references to guarantee sequential consistency.
17196 The default is @option{-mserialize-volatile}. Use
17197 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
17199 @item -mforce-no-pic
17200 @opindex mforce-no-pic
17201 For targets, like GNU/Linux, where all user-mode Xtensa code must be
17202 position-independent code (PIC), this option disables PIC for compiling
17205 @item -mtext-section-literals
17206 @itemx -mno-text-section-literals
17207 @opindex mtext-section-literals
17208 @opindex mno-text-section-literals
17209 Control the treatment of literal pools. The default is
17210 @option{-mno-text-section-literals}, which places literals in a separate
17211 section in the output file. This allows the literal pool to be placed
17212 in a data RAM/ROM, and it also allows the linker to combine literal
17213 pools from separate object files to remove redundant literals and
17214 improve code size. With @option{-mtext-section-literals}, the literals
17215 are interspersed in the text section in order to keep them as close as
17216 possible to their references. This may be necessary for large assembly
17219 @item -mtarget-align
17220 @itemx -mno-target-align
17221 @opindex mtarget-align
17222 @opindex mno-target-align
17223 When this option is enabled, GCC instructs the assembler to
17224 automatically align instructions to reduce branch penalties at the
17225 expense of some code density. The assembler attempts to widen density
17226 instructions to align branch targets and the instructions following call
17227 instructions. If there are not enough preceding safe density
17228 instructions to align a target, no widening will be performed. The
17229 default is @option{-mtarget-align}. These options do not affect the
17230 treatment of auto-aligned instructions like @code{LOOP}, which the
17231 assembler will always align, either by widening density instructions or
17232 by inserting no-op instructions.
17235 @itemx -mno-longcalls
17236 @opindex mlongcalls
17237 @opindex mno-longcalls
17238 When this option is enabled, GCC instructs the assembler to translate
17239 direct calls to indirect calls unless it can determine that the target
17240 of a direct call is in the range allowed by the call instruction. This
17241 translation typically occurs for calls to functions in other source
17242 files. Specifically, the assembler translates a direct @code{CALL}
17243 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
17244 The default is @option{-mno-longcalls}. This option should be used in
17245 programs where the call target can potentially be out of range. This
17246 option is implemented in the assembler, not the compiler, so the
17247 assembly code generated by GCC will still show direct call
17248 instructions---look at the disassembled object code to see the actual
17249 instructions. Note that the assembler will use an indirect call for
17250 every cross-file call, not just those that really will be out of range.
17253 @node zSeries Options
17254 @subsection zSeries Options
17255 @cindex zSeries options
17257 These are listed under @xref{S/390 and zSeries Options}.
17259 @node Code Gen Options
17260 @section Options for Code Generation Conventions
17261 @cindex code generation conventions
17262 @cindex options, code generation
17263 @cindex run-time options
17265 These machine-independent options control the interface conventions
17266 used in code generation.
17268 Most of them have both positive and negative forms; the negative form
17269 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
17270 one of the forms is listed---the one which is not the default. You
17271 can figure out the other form by either removing @samp{no-} or adding
17275 @item -fbounds-check
17276 @opindex fbounds-check
17277 For front-ends that support it, generate additional code to check that
17278 indices used to access arrays are within the declared range. This is
17279 currently only supported by the Java and Fortran front-ends, where
17280 this option defaults to true and false respectively.
17284 This option generates traps for signed overflow on addition, subtraction,
17285 multiplication operations.
17289 This option instructs the compiler to assume that signed arithmetic
17290 overflow of addition, subtraction and multiplication wraps around
17291 using twos-complement representation. This flag enables some optimizations
17292 and disables others. This option is enabled by default for the Java
17293 front-end, as required by the Java language specification.
17296 @opindex fexceptions
17297 Enable exception handling. Generates extra code needed to propagate
17298 exceptions. For some targets, this implies GCC will generate frame
17299 unwind information for all functions, which can produce significant data
17300 size overhead, although it does not affect execution. If you do not
17301 specify this option, GCC will enable it by default for languages like
17302 C++ which normally require exception handling, and disable it for
17303 languages like C that do not normally require it. However, you may need
17304 to enable this option when compiling C code that needs to interoperate
17305 properly with exception handlers written in C++. You may also wish to
17306 disable this option if you are compiling older C++ programs that don't
17307 use exception handling.
17309 @item -fnon-call-exceptions
17310 @opindex fnon-call-exceptions
17311 Generate code that allows trapping instructions to throw exceptions.
17312 Note that this requires platform-specific runtime support that does
17313 not exist everywhere. Moreover, it only allows @emph{trapping}
17314 instructions to throw exceptions, i.e.@: memory references or floating
17315 point instructions. It does not allow exceptions to be thrown from
17316 arbitrary signal handlers such as @code{SIGALRM}.
17318 @item -funwind-tables
17319 @opindex funwind-tables
17320 Similar to @option{-fexceptions}, except that it will just generate any needed
17321 static data, but will not affect the generated code in any other way.
17322 You will normally not enable this option; instead, a language processor
17323 that needs this handling would enable it on your behalf.
17325 @item -fasynchronous-unwind-tables
17326 @opindex fasynchronous-unwind-tables
17327 Generate unwind table in dwarf2 format, if supported by target machine. The
17328 table is exact at each instruction boundary, so it can be used for stack
17329 unwinding from asynchronous events (such as debugger or garbage collector).
17331 @item -fpcc-struct-return
17332 @opindex fpcc-struct-return
17333 Return ``short'' @code{struct} and @code{union} values in memory like
17334 longer ones, rather than in registers. This convention is less
17335 efficient, but it has the advantage of allowing intercallability between
17336 GCC-compiled files and files compiled with other compilers, particularly
17337 the Portable C Compiler (pcc).
17339 The precise convention for returning structures in memory depends
17340 on the target configuration macros.
17342 Short structures and unions are those whose size and alignment match
17343 that of some integer type.
17345 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
17346 switch is not binary compatible with code compiled with the
17347 @option{-freg-struct-return} switch.
17348 Use it to conform to a non-default application binary interface.
17350 @item -freg-struct-return
17351 @opindex freg-struct-return
17352 Return @code{struct} and @code{union} values in registers when possible.
17353 This is more efficient for small structures than
17354 @option{-fpcc-struct-return}.
17356 If you specify neither @option{-fpcc-struct-return} nor
17357 @option{-freg-struct-return}, GCC defaults to whichever convention is
17358 standard for the target. If there is no standard convention, GCC
17359 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
17360 the principal compiler. In those cases, we can choose the standard, and
17361 we chose the more efficient register return alternative.
17363 @strong{Warning:} code compiled with the @option{-freg-struct-return}
17364 switch is not binary compatible with code compiled with the
17365 @option{-fpcc-struct-return} switch.
17366 Use it to conform to a non-default application binary interface.
17368 @item -fshort-enums
17369 @opindex fshort-enums
17370 Allocate to an @code{enum} type only as many bytes as it needs for the
17371 declared range of possible values. Specifically, the @code{enum} type
17372 will be equivalent to the smallest integer type which has enough room.
17374 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
17375 code that is not binary compatible with code generated without that switch.
17376 Use it to conform to a non-default application binary interface.
17378 @item -fshort-double
17379 @opindex fshort-double
17380 Use the same size for @code{double} as for @code{float}.
17382 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
17383 code that is not binary compatible with code generated without that switch.
17384 Use it to conform to a non-default application binary interface.
17386 @item -fshort-wchar
17387 @opindex fshort-wchar
17388 Override the underlying type for @samp{wchar_t} to be @samp{short
17389 unsigned int} instead of the default for the target. This option is
17390 useful for building programs to run under WINE@.
17392 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
17393 code that is not binary compatible with code generated without that switch.
17394 Use it to conform to a non-default application binary interface.
17397 @opindex fno-common
17398 In C code, controls the placement of uninitialized global variables.
17399 Unix C compilers have traditionally permitted multiple definitions of
17400 such variables in different compilation units by placing the variables
17402 This is the behavior specified by @option{-fcommon}, and is the default
17403 for GCC on most targets.
17404 On the other hand, this behavior is not required by ISO C, and on some
17405 targets may carry a speed or code size penalty on variable references.
17406 The @option{-fno-common} option specifies that the compiler should place
17407 uninitialized global variables in the data section of the object file,
17408 rather than generating them as common blocks.
17409 This has the effect that if the same variable is declared
17410 (without @code{extern}) in two different compilations,
17411 you will get a multiple-definition error when you link them.
17412 In this case, you must compile with @option{-fcommon} instead.
17413 Compiling with @option{-fno-common} is useful on targets for which
17414 it provides better performance, or if you wish to verify that the
17415 program will work on other systems which always treat uninitialized
17416 variable declarations this way.
17420 Ignore the @samp{#ident} directive.
17422 @item -finhibit-size-directive
17423 @opindex finhibit-size-directive
17424 Don't output a @code{.size} assembler directive, or anything else that
17425 would cause trouble if the function is split in the middle, and the
17426 two halves are placed at locations far apart in memory. This option is
17427 used when compiling @file{crtstuff.c}; you should not need to use it
17430 @item -fverbose-asm
17431 @opindex fverbose-asm
17432 Put extra commentary information in the generated assembly code to
17433 make it more readable. This option is generally only of use to those
17434 who actually need to read the generated assembly code (perhaps while
17435 debugging the compiler itself).
17437 @option{-fno-verbose-asm}, the default, causes the
17438 extra information to be omitted and is useful when comparing two assembler
17441 @item -frecord-gcc-switches
17442 @opindex frecord-gcc-switches
17443 This switch causes the command line that was used to invoke the
17444 compiler to be recorded into the object file that is being created.
17445 This switch is only implemented on some targets and the exact format
17446 of the recording is target and binary file format dependent, but it
17447 usually takes the form of a section containing ASCII text. This
17448 switch is related to the @option{-fverbose-asm} switch, but that
17449 switch only records information in the assembler output file as
17450 comments, so it never reaches the object file.
17454 @cindex global offset table
17456 Generate position-independent code (PIC) suitable for use in a shared
17457 library, if supported for the target machine. Such code accesses all
17458 constant addresses through a global offset table (GOT)@. The dynamic
17459 loader resolves the GOT entries when the program starts (the dynamic
17460 loader is not part of GCC; it is part of the operating system). If
17461 the GOT size for the linked executable exceeds a machine-specific
17462 maximum size, you get an error message from the linker indicating that
17463 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
17464 instead. (These maximums are 8k on the SPARC and 32k
17465 on the m68k and RS/6000. The 386 has no such limit.)
17467 Position-independent code requires special support, and therefore works
17468 only on certain machines. For the 386, GCC supports PIC for System V
17469 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
17470 position-independent.
17472 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
17477 If supported for the target machine, emit position-independent code,
17478 suitable for dynamic linking and avoiding any limit on the size of the
17479 global offset table. This option makes a difference on the m68k,
17480 PowerPC and SPARC@.
17482 Position-independent code requires special support, and therefore works
17483 only on certain machines.
17485 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
17492 These options are similar to @option{-fpic} and @option{-fPIC}, but
17493 generated position independent code can be only linked into executables.
17494 Usually these options are used when @option{-pie} GCC option will be
17495 used during linking.
17497 @option{-fpie} and @option{-fPIE} both define the macros
17498 @code{__pie__} and @code{__PIE__}. The macros have the value 1
17499 for @option{-fpie} and 2 for @option{-fPIE}.
17501 @item -fno-jump-tables
17502 @opindex fno-jump-tables
17503 Do not use jump tables for switch statements even where it would be
17504 more efficient than other code generation strategies. This option is
17505 of use in conjunction with @option{-fpic} or @option{-fPIC} for
17506 building code which forms part of a dynamic linker and cannot
17507 reference the address of a jump table. On some targets, jump tables
17508 do not require a GOT and this option is not needed.
17510 @item -ffixed-@var{reg}
17512 Treat the register named @var{reg} as a fixed register; generated code
17513 should never refer to it (except perhaps as a stack pointer, frame
17514 pointer or in some other fixed role).
17516 @var{reg} must be the name of a register. The register names accepted
17517 are machine-specific and are defined in the @code{REGISTER_NAMES}
17518 macro in the machine description macro file.
17520 This flag does not have a negative form, because it specifies a
17523 @item -fcall-used-@var{reg}
17524 @opindex fcall-used
17525 Treat the register named @var{reg} as an allocable register that is
17526 clobbered by function calls. It may be allocated for temporaries or
17527 variables that do not live across a call. Functions compiled this way
17528 will not save and restore the register @var{reg}.
17530 It is an error to used this flag with the frame pointer or stack pointer.
17531 Use of this flag for other registers that have fixed pervasive roles in
17532 the machine's execution model will produce disastrous results.
17534 This flag does not have a negative form, because it specifies a
17537 @item -fcall-saved-@var{reg}
17538 @opindex fcall-saved
17539 Treat the register named @var{reg} as an allocable register saved by
17540 functions. It may be allocated even for temporaries or variables that
17541 live across a call. Functions compiled this way will save and restore
17542 the register @var{reg} if they use it.
17544 It is an error to used this flag with the frame pointer or stack pointer.
17545 Use of this flag for other registers that have fixed pervasive roles in
17546 the machine's execution model will produce disastrous results.
17548 A different sort of disaster will result from the use of this flag for
17549 a register in which function values may be returned.
17551 This flag does not have a negative form, because it specifies a
17554 @item -fpack-struct[=@var{n}]
17555 @opindex fpack-struct
17556 Without a value specified, pack all structure members together without
17557 holes. When a value is specified (which must be a small power of two), pack
17558 structure members according to this value, representing the maximum
17559 alignment (that is, objects with default alignment requirements larger than
17560 this will be output potentially unaligned at the next fitting location.
17562 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
17563 code that is not binary compatible with code generated without that switch.
17564 Additionally, it makes the code suboptimal.
17565 Use it to conform to a non-default application binary interface.
17567 @item -finstrument-functions
17568 @opindex finstrument-functions
17569 Generate instrumentation calls for entry and exit to functions. Just
17570 after function entry and just before function exit, the following
17571 profiling functions will be called with the address of the current
17572 function and its call site. (On some platforms,
17573 @code{__builtin_return_address} does not work beyond the current
17574 function, so the call site information may not be available to the
17575 profiling functions otherwise.)
17578 void __cyg_profile_func_enter (void *this_fn,
17580 void __cyg_profile_func_exit (void *this_fn,
17584 The first argument is the address of the start of the current function,
17585 which may be looked up exactly in the symbol table.
17587 This instrumentation is also done for functions expanded inline in other
17588 functions. The profiling calls will indicate where, conceptually, the
17589 inline function is entered and exited. This means that addressable
17590 versions of such functions must be available. If all your uses of a
17591 function are expanded inline, this may mean an additional expansion of
17592 code size. If you use @samp{extern inline} in your C code, an
17593 addressable version of such functions must be provided. (This is
17594 normally the case anyways, but if you get lucky and the optimizer always
17595 expands the functions inline, you might have gotten away without
17596 providing static copies.)
17598 A function may be given the attribute @code{no_instrument_function}, in
17599 which case this instrumentation will not be done. This can be used, for
17600 example, for the profiling functions listed above, high-priority
17601 interrupt routines, and any functions from which the profiling functions
17602 cannot safely be called (perhaps signal handlers, if the profiling
17603 routines generate output or allocate memory).
17605 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
17606 @opindex finstrument-functions-exclude-file-list
17608 Set the list of functions that are excluded from instrumentation (see
17609 the description of @code{-finstrument-functions}). If the file that
17610 contains a function definition matches with one of @var{file}, then
17611 that function is not instrumented. The match is done on substrings:
17612 if the @var{file} parameter is a substring of the file name, it is
17613 considered to be a match.
17616 @code{-finstrument-functions-exclude-file-list=/bits/stl,include/sys}
17617 will exclude any inline function defined in files whose pathnames
17618 contain @code{/bits/stl} or @code{include/sys}.
17620 If, for some reason, you want to include letter @code{','} in one of
17621 @var{sym}, write @code{'\,'}. For example,
17622 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
17623 (note the single quote surrounding the option).
17625 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
17626 @opindex finstrument-functions-exclude-function-list
17628 This is similar to @code{-finstrument-functions-exclude-file-list},
17629 but this option sets the list of function names to be excluded from
17630 instrumentation. The function name to be matched is its user-visible
17631 name, such as @code{vector<int> blah(const vector<int> &)}, not the
17632 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
17633 match is done on substrings: if the @var{sym} parameter is a substring
17634 of the function name, it is considered to be a match. For C99 and C++
17635 extended identifiers, the function name must be given in UTF-8, not
17636 using universal character names.
17638 @item -fstack-check
17639 @opindex fstack-check
17640 Generate code to verify that you do not go beyond the boundary of the
17641 stack. You should specify this flag if you are running in an
17642 environment with multiple threads, but only rarely need to specify it in
17643 a single-threaded environment since stack overflow is automatically
17644 detected on nearly all systems if there is only one stack.
17646 Note that this switch does not actually cause checking to be done; the
17647 operating system or the language runtime must do that. The switch causes
17648 generation of code to ensure that they see the stack being extended.
17650 You can additionally specify a string parameter: @code{no} means no
17651 checking, @code{generic} means force the use of old-style checking,
17652 @code{specific} means use the best checking method and is equivalent
17653 to bare @option{-fstack-check}.
17655 Old-style checking is a generic mechanism that requires no specific
17656 target support in the compiler but comes with the following drawbacks:
17660 Modified allocation strategy for large objects: they will always be
17661 allocated dynamically if their size exceeds a fixed threshold.
17664 Fixed limit on the size of the static frame of functions: when it is
17665 topped by a particular function, stack checking is not reliable and
17666 a warning is issued by the compiler.
17669 Inefficiency: because of both the modified allocation strategy and the
17670 generic implementation, the performances of the code are hampered.
17673 Note that old-style stack checking is also the fallback method for
17674 @code{specific} if no target support has been added in the compiler.
17676 @item -fstack-limit-register=@var{reg}
17677 @itemx -fstack-limit-symbol=@var{sym}
17678 @itemx -fno-stack-limit
17679 @opindex fstack-limit-register
17680 @opindex fstack-limit-symbol
17681 @opindex fno-stack-limit
17682 Generate code to ensure that the stack does not grow beyond a certain value,
17683 either the value of a register or the address of a symbol. If the stack
17684 would grow beyond the value, a signal is raised. For most targets,
17685 the signal is raised before the stack overruns the boundary, so
17686 it is possible to catch the signal without taking special precautions.
17688 For instance, if the stack starts at absolute address @samp{0x80000000}
17689 and grows downwards, you can use the flags
17690 @option{-fstack-limit-symbol=__stack_limit} and
17691 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
17692 of 128KB@. Note that this may only work with the GNU linker.
17694 @item -fleading-underscore
17695 @opindex fleading-underscore
17696 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
17697 change the way C symbols are represented in the object file. One use
17698 is to help link with legacy assembly code.
17700 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
17701 generate code that is not binary compatible with code generated without that
17702 switch. Use it to conform to a non-default application binary interface.
17703 Not all targets provide complete support for this switch.
17705 @item -ftls-model=@var{model}
17706 @opindex ftls-model
17707 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
17708 The @var{model} argument should be one of @code{global-dynamic},
17709 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
17711 The default without @option{-fpic} is @code{initial-exec}; with
17712 @option{-fpic} the default is @code{global-dynamic}.
17714 @item -fvisibility=@var{default|internal|hidden|protected}
17715 @opindex fvisibility
17716 Set the default ELF image symbol visibility to the specified option---all
17717 symbols will be marked with this unless overridden within the code.
17718 Using this feature can very substantially improve linking and
17719 load times of shared object libraries, produce more optimized
17720 code, provide near-perfect API export and prevent symbol clashes.
17721 It is @strong{strongly} recommended that you use this in any shared objects
17724 Despite the nomenclature, @code{default} always means public ie;
17725 available to be linked against from outside the shared object.
17726 @code{protected} and @code{internal} are pretty useless in real-world
17727 usage so the only other commonly used option will be @code{hidden}.
17728 The default if @option{-fvisibility} isn't specified is
17729 @code{default}, i.e., make every
17730 symbol public---this causes the same behavior as previous versions of
17733 A good explanation of the benefits offered by ensuring ELF
17734 symbols have the correct visibility is given by ``How To Write
17735 Shared Libraries'' by Ulrich Drepper (which can be found at
17736 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
17737 solution made possible by this option to marking things hidden when
17738 the default is public is to make the default hidden and mark things
17739 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
17740 and @code{__attribute__ ((visibility("default")))} instead of
17741 @code{__declspec(dllexport)} you get almost identical semantics with
17742 identical syntax. This is a great boon to those working with
17743 cross-platform projects.
17745 For those adding visibility support to existing code, you may find
17746 @samp{#pragma GCC visibility} of use. This works by you enclosing
17747 the declarations you wish to set visibility for with (for example)
17748 @samp{#pragma GCC visibility push(hidden)} and
17749 @samp{#pragma GCC visibility pop}.
17750 Bear in mind that symbol visibility should be viewed @strong{as
17751 part of the API interface contract} and thus all new code should
17752 always specify visibility when it is not the default ie; declarations
17753 only for use within the local DSO should @strong{always} be marked explicitly
17754 as hidden as so to avoid PLT indirection overheads---making this
17755 abundantly clear also aids readability and self-documentation of the code.
17756 Note that due to ISO C++ specification requirements, operator new and
17757 operator delete must always be of default visibility.
17759 Be aware that headers from outside your project, in particular system
17760 headers and headers from any other library you use, may not be
17761 expecting to be compiled with visibility other than the default. You
17762 may need to explicitly say @samp{#pragma GCC visibility push(default)}
17763 before including any such headers.
17765 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
17766 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
17767 no modifications. However, this means that calls to @samp{extern}
17768 functions with no explicit visibility will use the PLT, so it is more
17769 effective to use @samp{__attribute ((visibility))} and/or
17770 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
17771 declarations should be treated as hidden.
17773 Note that @samp{-fvisibility} does affect C++ vague linkage
17774 entities. This means that, for instance, an exception class that will
17775 be thrown between DSOs must be explicitly marked with default
17776 visibility so that the @samp{type_info} nodes will be unified between
17779 An overview of these techniques, their benefits and how to use them
17780 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
17782 @item -fstrict-volatile-bitfields
17783 This option should be used if accesses to volatile bitfields (or other
17784 structure fields, although the compiler usually honors those types
17785 anyway) should use a single access in a mode of the same size as the
17786 container's type, aligned to a natural alignment if possible. For
17787 example, targets with memory-mapped peripheral registers might require
17788 all such accesses to be 16 bits wide; with this flag the user could
17789 declare all peripheral bitfields as ``unsigned short'' (assuming short
17790 is 16 bits on these targets) to force GCC to use 16 bit accesses
17791 instead of, perhaps, a more efficient 32 bit access.
17793 If this option is disabled, the compiler will use the most efficient
17794 instruction. In the previous example, that might be a 32-bit load
17795 instruction, even though that will access bytes that do not contain
17796 any portion of the bitfield, or memory-mapped registers unrelated to
17797 the one being updated.
17799 If the target requires strict alignment, and honoring the container
17800 type would require violating this alignment, a warning is issued.
17801 However, the access happens as the user requested, under the
17802 assumption that the user knows something about the target hardware
17803 that GCC is unaware of.
17805 The default value of this option is determined by the application binary
17806 interface for the target processor.
17812 @node Environment Variables
17813 @section Environment Variables Affecting GCC
17814 @cindex environment variables
17816 @c man begin ENVIRONMENT
17817 This section describes several environment variables that affect how GCC
17818 operates. Some of them work by specifying directories or prefixes to use
17819 when searching for various kinds of files. Some are used to specify other
17820 aspects of the compilation environment.
17822 Note that you can also specify places to search using options such as
17823 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
17824 take precedence over places specified using environment variables, which
17825 in turn take precedence over those specified by the configuration of GCC@.
17826 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
17827 GNU Compiler Collection (GCC) Internals}.
17832 @c @itemx LC_COLLATE
17834 @c @itemx LC_MONETARY
17835 @c @itemx LC_NUMERIC
17840 @c @findex LC_COLLATE
17841 @findex LC_MESSAGES
17842 @c @findex LC_MONETARY
17843 @c @findex LC_NUMERIC
17847 These environment variables control the way that GCC uses
17848 localization information that allow GCC to work with different
17849 national conventions. GCC inspects the locale categories
17850 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
17851 so. These locale categories can be set to any value supported by your
17852 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
17853 Kingdom encoded in UTF-8.
17855 The @env{LC_CTYPE} environment variable specifies character
17856 classification. GCC uses it to determine the character boundaries in
17857 a string; this is needed for some multibyte encodings that contain quote
17858 and escape characters that would otherwise be interpreted as a string
17861 The @env{LC_MESSAGES} environment variable specifies the language to
17862 use in diagnostic messages.
17864 If the @env{LC_ALL} environment variable is set, it overrides the value
17865 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
17866 and @env{LC_MESSAGES} default to the value of the @env{LANG}
17867 environment variable. If none of these variables are set, GCC
17868 defaults to traditional C English behavior.
17872 If @env{TMPDIR} is set, it specifies the directory to use for temporary
17873 files. GCC uses temporary files to hold the output of one stage of
17874 compilation which is to be used as input to the next stage: for example,
17875 the output of the preprocessor, which is the input to the compiler
17878 @item GCC_EXEC_PREFIX
17879 @findex GCC_EXEC_PREFIX
17880 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
17881 names of the subprograms executed by the compiler. No slash is added
17882 when this prefix is combined with the name of a subprogram, but you can
17883 specify a prefix that ends with a slash if you wish.
17885 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
17886 an appropriate prefix to use based on the pathname it was invoked with.
17888 If GCC cannot find the subprogram using the specified prefix, it
17889 tries looking in the usual places for the subprogram.
17891 The default value of @env{GCC_EXEC_PREFIX} is
17892 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
17893 the installed compiler. In many cases @var{prefix} is the value
17894 of @code{prefix} when you ran the @file{configure} script.
17896 Other prefixes specified with @option{-B} take precedence over this prefix.
17898 This prefix is also used for finding files such as @file{crt0.o} that are
17901 In addition, the prefix is used in an unusual way in finding the
17902 directories to search for header files. For each of the standard
17903 directories whose name normally begins with @samp{/usr/local/lib/gcc}
17904 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
17905 replacing that beginning with the specified prefix to produce an
17906 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
17907 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
17908 These alternate directories are searched first; the standard directories
17909 come next. If a standard directory begins with the configured
17910 @var{prefix} then the value of @var{prefix} is replaced by
17911 @env{GCC_EXEC_PREFIX} when looking for header files.
17913 @item COMPILER_PATH
17914 @findex COMPILER_PATH
17915 The value of @env{COMPILER_PATH} is a colon-separated list of
17916 directories, much like @env{PATH}. GCC tries the directories thus
17917 specified when searching for subprograms, if it can't find the
17918 subprograms using @env{GCC_EXEC_PREFIX}.
17921 @findex LIBRARY_PATH
17922 The value of @env{LIBRARY_PATH} is a colon-separated list of
17923 directories, much like @env{PATH}. When configured as a native compiler,
17924 GCC tries the directories thus specified when searching for special
17925 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
17926 using GCC also uses these directories when searching for ordinary
17927 libraries for the @option{-l} option (but directories specified with
17928 @option{-L} come first).
17932 @cindex locale definition
17933 This variable is used to pass locale information to the compiler. One way in
17934 which this information is used is to determine the character set to be used
17935 when character literals, string literals and comments are parsed in C and C++.
17936 When the compiler is configured to allow multibyte characters,
17937 the following values for @env{LANG} are recognized:
17941 Recognize JIS characters.
17943 Recognize SJIS characters.
17945 Recognize EUCJP characters.
17948 If @env{LANG} is not defined, or if it has some other value, then the
17949 compiler will use mblen and mbtowc as defined by the default locale to
17950 recognize and translate multibyte characters.
17954 Some additional environments variables affect the behavior of the
17957 @include cppenv.texi
17961 @node Precompiled Headers
17962 @section Using Precompiled Headers
17963 @cindex precompiled headers
17964 @cindex speed of compilation
17966 Often large projects have many header files that are included in every
17967 source file. The time the compiler takes to process these header files
17968 over and over again can account for nearly all of the time required to
17969 build the project. To make builds faster, GCC allows users to
17970 `precompile' a header file; then, if builds can use the precompiled
17971 header file they will be much faster.
17973 To create a precompiled header file, simply compile it as you would any
17974 other file, if necessary using the @option{-x} option to make the driver
17975 treat it as a C or C++ header file. You will probably want to use a
17976 tool like @command{make} to keep the precompiled header up-to-date when
17977 the headers it contains change.
17979 A precompiled header file will be searched for when @code{#include} is
17980 seen in the compilation. As it searches for the included file
17981 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
17982 compiler looks for a precompiled header in each directory just before it
17983 looks for the include file in that directory. The name searched for is
17984 the name specified in the @code{#include} with @samp{.gch} appended. If
17985 the precompiled header file can't be used, it is ignored.
17987 For instance, if you have @code{#include "all.h"}, and you have
17988 @file{all.h.gch} in the same directory as @file{all.h}, then the
17989 precompiled header file will be used if possible, and the original
17990 header will be used otherwise.
17992 Alternatively, you might decide to put the precompiled header file in a
17993 directory and use @option{-I} to ensure that directory is searched
17994 before (or instead of) the directory containing the original header.
17995 Then, if you want to check that the precompiled header file is always
17996 used, you can put a file of the same name as the original header in this
17997 directory containing an @code{#error} command.
17999 This also works with @option{-include}. So yet another way to use
18000 precompiled headers, good for projects not designed with precompiled
18001 header files in mind, is to simply take most of the header files used by
18002 a project, include them from another header file, precompile that header
18003 file, and @option{-include} the precompiled header. If the header files
18004 have guards against multiple inclusion, they will be skipped because
18005 they've already been included (in the precompiled header).
18007 If you need to precompile the same header file for different
18008 languages, targets, or compiler options, you can instead make a
18009 @emph{directory} named like @file{all.h.gch}, and put each precompiled
18010 header in the directory, perhaps using @option{-o}. It doesn't matter
18011 what you call the files in the directory, every precompiled header in
18012 the directory will be considered. The first precompiled header
18013 encountered in the directory that is valid for this compilation will
18014 be used; they're searched in no particular order.
18016 There are many other possibilities, limited only by your imagination,
18017 good sense, and the constraints of your build system.
18019 A precompiled header file can be used only when these conditions apply:
18023 Only one precompiled header can be used in a particular compilation.
18026 A precompiled header can't be used once the first C token is seen. You
18027 can have preprocessor directives before a precompiled header; you can
18028 even include a precompiled header from inside another header, so long as
18029 there are no C tokens before the @code{#include}.
18032 The precompiled header file must be produced for the same language as
18033 the current compilation. You can't use a C precompiled header for a C++
18037 The precompiled header file must have been produced by the same compiler
18038 binary as the current compilation is using.
18041 Any macros defined before the precompiled header is included must
18042 either be defined in the same way as when the precompiled header was
18043 generated, or must not affect the precompiled header, which usually
18044 means that they don't appear in the precompiled header at all.
18046 The @option{-D} option is one way to define a macro before a
18047 precompiled header is included; using a @code{#define} can also do it.
18048 There are also some options that define macros implicitly, like
18049 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
18052 @item If debugging information is output when using the precompiled
18053 header, using @option{-g} or similar, the same kind of debugging information
18054 must have been output when building the precompiled header. However,
18055 a precompiled header built using @option{-g} can be used in a compilation
18056 when no debugging information is being output.
18058 @item The same @option{-m} options must generally be used when building
18059 and using the precompiled header. @xref{Submodel Options},
18060 for any cases where this rule is relaxed.
18062 @item Each of the following options must be the same when building and using
18063 the precompiled header:
18065 @gccoptlist{-fexceptions}
18068 Some other command-line options starting with @option{-f},
18069 @option{-p}, or @option{-O} must be defined in the same way as when
18070 the precompiled header was generated. At present, it's not clear
18071 which options are safe to change and which are not; the safest choice
18072 is to use exactly the same options when generating and using the
18073 precompiled header. The following are known to be safe:
18075 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
18076 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
18077 -fsched-verbose=<number> -fschedule-insns -fvisibility= @gol
18082 For all of these except the last, the compiler will automatically
18083 ignore the precompiled header if the conditions aren't met. If you
18084 find an option combination that doesn't work and doesn't cause the
18085 precompiled header to be ignored, please consider filing a bug report,
18088 If you do use differing options when generating and using the
18089 precompiled header, the actual behavior will be a mixture of the
18090 behavior for the options. For instance, if you use @option{-g} to
18091 generate the precompiled header but not when using it, you may or may
18092 not get debugging information for routines in the precompiled header.