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} -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 -fplan9-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 -fobjc-std=objc1 @gol
216 -freplace-objc-classes @gol
219 -Wassign-intercept @gol
220 -Wno-protocol -Wselector @gol
221 -Wstrict-selector-match @gol
222 -Wundeclared-selector}
224 @item Language Independent Options
225 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
226 @gccoptlist{-fmessage-length=@var{n} @gol
227 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
228 -fdiagnostics-show-option}
230 @item Warning Options
231 @xref{Warning Options,,Options to Request or Suppress Warnings}.
232 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
233 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
234 -Wno-attributes -Wno-builtin-macro-redefined @gol
235 -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual @gol
236 -Wchar-subscripts -Wclobbered -Wcomment @gol
237 -Wconversion -Wcoverage-mismatch -Wcpp -Wno-deprecated @gol
238 -Wno-deprecated-declarations -Wdisabled-optimization @gol
239 -Wno-div-by-zero -Wdouble-promotion -Wempty-body -Wenum-compare @gol
240 -Wno-endif-labels -Werror -Werror=* @gol
241 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
242 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
243 -Wformat-security -Wformat-y2k @gol
244 -Wframe-larger-than=@var{len} -Wjump-misses-init -Wignored-qualifiers @gol
245 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
246 -Winit-self -Winline @gol
247 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
248 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
249 -Wlogical-op -Wlong-long @gol
250 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
251 -Wmissing-format-attribute -Wmissing-include-dirs @gol
253 -Wno-multichar -Wnonnull -Wno-overflow @gol
254 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
255 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
256 -Wpointer-arith -Wno-pointer-to-int-cast @gol
257 -Wredundant-decls @gol
258 -Wreturn-type -Wsequence-point -Wshadow @gol
259 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
260 -Wstrict-aliasing -Wstrict-aliasing=n @gol
261 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
262 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]} @gol
263 -Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand @gol
264 -Wsystem-headers -Wtrampolines -Wtrigraphs -Wtype-limits -Wundef @gol
265 -Wuninitialized -Wunknown-pragmas -Wno-pragmas @gol
266 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
267 -Wunused-label -Wunused-parameter -Wno-unused-result -Wunused-value @gol
268 -Wunused-variable -Wunused-but-set-parameter -Wunused-but-set-variable @gol
269 -Wvariadic-macros -Wvla -Wvolatile-register-var -Wwrite-strings}
271 @item C and Objective-C-only Warning Options
272 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
273 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
274 -Wold-style-declaration -Wold-style-definition @gol
275 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
276 -Wdeclaration-after-statement -Wpointer-sign}
278 @item Debugging Options
279 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
280 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
281 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
282 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
283 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
284 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
285 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
286 -fdump-statistics @gol
288 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
289 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
290 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
292 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
293 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
294 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
295 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
296 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
297 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
298 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
299 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
300 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
301 -fdump-tree-nrv -fdump-tree-vect @gol
302 -fdump-tree-sink @gol
303 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
304 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
305 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
306 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
307 -ftree-vectorizer-verbose=@var{n} @gol
308 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
309 -fdump-final-insns=@var{file} @gol
310 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
311 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
312 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
313 -fenable-icf-debug @gol
314 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
315 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
316 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
317 -fstack-usage -ftest-coverage -ftime-report -fvar-tracking @gol
318 -fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol
319 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
320 -ggdb -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
321 -gvms -gxcoff -gxcoff+ @gol
322 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
323 -fdebug-prefix-map=@var{old}=@var{new} @gol
324 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
325 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
326 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
327 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
328 -print-prog-name=@var{program} -print-search-dirs -Q @gol
329 -print-sysroot -print-sysroot-headers-suffix @gol
330 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
332 @item Optimization Options
333 @xref{Optimize Options,,Options that Control Optimization}.
334 @gccoptlist{-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 -fcombine-stack-adjustments -fconserve-stack @gol
339 -fcprop-registers -fcrossjumping @gol
340 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
341 -fcx-limited-range @gol
342 -fdata-sections -fdce -fdce @gol
343 -fdelayed-branch -fdelete-null-pointer-checks -fdse -fdse @gol
344 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffast-math @gol
345 -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
346 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
347 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
348 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
349 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
350 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg @gol
351 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference @gol
352 -fipa-struct-reorg -fira-algorithm=@var{algorithm} @gol
353 -fira-region=@var{region} @gol
354 -fira-loop-pressure -fno-ira-share-save-slots @gol
355 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
356 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
357 -floop-block -floop-flatten -floop-interchange -floop-strip-mine @gol
358 -floop-parallelize-all -flto -flto-compression-level -flto-partition=@var{alg} @gol
359 -flto-report -fltrans -fltrans-output-list -fmerge-all-constants @gol
360 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
361 -fmove-loop-invariants fmudflap -fmudflapir -fmudflapth -fno-branch-count-reg @gol
362 -fno-default-inline @gol
363 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
364 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
365 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
366 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
367 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
368 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
369 -fprefetch-loop-arrays @gol
370 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
371 -fprofile-generate=@var{path} @gol
372 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
373 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
374 -freorder-blocks-and-partition -freorder-functions @gol
375 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
376 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
377 -fsched-spec-load -fsched-spec-load-dangerous @gol
378 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
379 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
380 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
381 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
382 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
383 -fselective-scheduling -fselective-scheduling2 @gol
384 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
385 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
386 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
387 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer @gol
389 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
390 -ftree-copyrename -ftree-dce -ftree-dominator-opts -ftree-dse @gol
391 -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
392 -ftree-loop-if-convert-memory-writes -ftree-loop-im @gol
393 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
394 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
395 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-pta -ftree-reassoc @gol
396 -ftree-sink -ftree-sra -ftree-switch-conversion @gol
397 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
398 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
399 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
400 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
401 -fwhole-program -fwhopr[=@var{n}] -fwpa -fuse-linker-plugin @gol
402 --param @var{name}=@var{value}
403 -O -O0 -O1 -O2 -O3 -Os -Ofast}
405 @item Preprocessor Options
406 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
407 @gccoptlist{-A@var{question}=@var{answer} @gol
408 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
409 -C -dD -dI -dM -dN @gol
410 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
411 -idirafter @var{dir} @gol
412 -include @var{file} -imacros @var{file} @gol
413 -iprefix @var{file} -iwithprefix @var{dir} @gol
414 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
415 -imultilib @var{dir} -isysroot @var{dir} @gol
416 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
417 -P -fworking-directory -remap @gol
418 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
419 -Xpreprocessor @var{option}}
421 @item Assembler Option
422 @xref{Assembler Options,,Passing Options to the Assembler}.
423 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
426 @xref{Link Options,,Options for Linking}.
427 @gccoptlist{@var{object-file-name} -l@var{library} @gol
428 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
429 -s -static -static-libgcc -static-libstdc++ -shared @gol
430 -shared-libgcc -symbolic @gol
431 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
434 @item Directory Options
435 @xref{Directory Options,,Options for Directory Search}.
436 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir}}
437 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I-
440 @item Machine Dependent Options
441 @xref{Submodel Options,,Hardware Models and Configurations}.
442 @c This list is ordered alphanumerically by subsection name.
443 @c Try and put the significant identifier (CPU or system) first,
444 @c so users have a clue at guessing where the ones they want will be.
447 @gccoptlist{-EB -EL @gol
448 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
449 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
452 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
453 -mabi=@var{name} @gol
454 -mapcs-stack-check -mno-apcs-stack-check @gol
455 -mapcs-float -mno-apcs-float @gol
456 -mapcs-reentrant -mno-apcs-reentrant @gol
457 -msched-prolog -mno-sched-prolog @gol
458 -mlittle-endian -mbig-endian -mwords-little-endian @gol
459 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
460 -mfp16-format=@var{name}
461 -mthumb-interwork -mno-thumb-interwork @gol
462 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
463 -mstructure-size-boundary=@var{n} @gol
464 -mabort-on-noreturn @gol
465 -mlong-calls -mno-long-calls @gol
466 -msingle-pic-base -mno-single-pic-base @gol
467 -mpic-register=@var{reg} @gol
468 -mnop-fun-dllimport @gol
469 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
470 -mpoke-function-name @gol
472 -mtpcs-frame -mtpcs-leaf-frame @gol
473 -mcaller-super-interworking -mcallee-super-interworking @gol
475 -mword-relocations @gol
476 -mfix-cortex-m3-ldrd}
479 @gccoptlist{-mmcu=@var{mcu} -mno-interrupts @gol
480 -mcall-prologues -mtiny-stack -mint8}
482 @emph{Blackfin Options}
483 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
484 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
485 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
486 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
487 -mno-id-shared-library -mshared-library-id=@var{n} @gol
488 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
489 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
490 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
494 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
495 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
496 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
497 -mstack-align -mdata-align -mconst-align @gol
498 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
499 -melf -maout -melinux -mlinux -sim -sim2 @gol
500 -mmul-bug-workaround -mno-mul-bug-workaround}
503 @gccoptlist{-mmac -mpush-args}
505 @emph{Darwin Options}
506 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
507 -arch_only -bind_at_load -bundle -bundle_loader @gol
508 -client_name -compatibility_version -current_version @gol
510 -dependency-file -dylib_file -dylinker_install_name @gol
511 -dynamic -dynamiclib -exported_symbols_list @gol
512 -filelist -flat_namespace -force_cpusubtype_ALL @gol
513 -force_flat_namespace -headerpad_max_install_names @gol
515 -image_base -init -install_name -keep_private_externs @gol
516 -multi_module -multiply_defined -multiply_defined_unused @gol
517 -noall_load -no_dead_strip_inits_and_terms @gol
518 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
519 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
520 -private_bundle -read_only_relocs -sectalign @gol
521 -sectobjectsymbols -whyload -seg1addr @gol
522 -sectcreate -sectobjectsymbols -sectorder @gol
523 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
524 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
525 -segprot -segs_read_only_addr -segs_read_write_addr @gol
526 -single_module -static -sub_library -sub_umbrella @gol
527 -twolevel_namespace -umbrella -undefined @gol
528 -unexported_symbols_list -weak_reference_mismatches @gol
529 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
530 -mkernel -mone-byte-bool}
532 @emph{DEC Alpha Options}
533 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
534 -mieee -mieee-with-inexact -mieee-conformant @gol
535 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
536 -mtrap-precision=@var{mode} -mbuild-constants @gol
537 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
538 -mbwx -mmax -mfix -mcix @gol
539 -mfloat-vax -mfloat-ieee @gol
540 -mexplicit-relocs -msmall-data -mlarge-data @gol
541 -msmall-text -mlarge-text @gol
542 -mmemory-latency=@var{time}}
544 @emph{DEC Alpha/VMS Options}
545 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
548 @gccoptlist{-msmall-model -mno-lsim}
551 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
552 -mhard-float -msoft-float @gol
553 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
554 -mdouble -mno-double @gol
555 -mmedia -mno-media -mmuladd -mno-muladd @gol
556 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
557 -mlinked-fp -mlong-calls -malign-labels @gol
558 -mlibrary-pic -macc-4 -macc-8 @gol
559 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
560 -moptimize-membar -mno-optimize-membar @gol
561 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
562 -mvliw-branch -mno-vliw-branch @gol
563 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
564 -mno-nested-cond-exec -mtomcat-stats @gol
568 @emph{GNU/Linux Options}
569 @gccoptlist{-mglibc -muclibc -mbionic -mandroid @gol
570 -tno-android-cc -tno-android-ld}
572 @emph{H8/300 Options}
573 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
576 @gccoptlist{-march=@var{architecture-type} @gol
577 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
578 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
579 -mfixed-range=@var{register-range} @gol
580 -mjump-in-delay -mlinker-opt -mlong-calls @gol
581 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
582 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
583 -mno-jump-in-delay -mno-long-load-store @gol
584 -mno-portable-runtime -mno-soft-float @gol
585 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
586 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
587 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
588 -munix=@var{unix-std} -nolibdld -static -threads}
590 @emph{i386 and x86-64 Options}
591 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
592 -mfpmath=@var{unit} @gol
593 -masm=@var{dialect} -mno-fancy-math-387 @gol
594 -mno-fp-ret-in-387 -msoft-float @gol
595 -mno-wide-multiply -mrtd -malign-double @gol
596 -mpreferred-stack-boundary=@var{num}
597 -mincoming-stack-boundary=@var{num} @gol
598 -mcld -mcx16 -msahf -mmovbe -mcrc32 -mrecip -mvzeroupper @gol
599 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
600 -maes -mpclmul -mfsgsbase -mrdrnd -mf16c -mfused-madd @gol
601 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlwp @gol
602 -mthreads -mno-align-stringops -minline-all-stringops @gol
603 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
604 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
605 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
606 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
607 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
608 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
609 -mcmodel=@var{code-model} -mabi=@var{name} @gol
610 -m32 -m64 -mlarge-data-threshold=@var{num} @gol
611 -msse2avx -mfentry -m8bit-idiv}
614 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
615 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
616 -mconstant-gp -mauto-pic -mfused-madd @gol
617 -minline-float-divide-min-latency @gol
618 -minline-float-divide-max-throughput @gol
619 -mno-inline-float-divide @gol
620 -minline-int-divide-min-latency @gol
621 -minline-int-divide-max-throughput @gol
622 -mno-inline-int-divide @gol
623 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
624 -mno-inline-sqrt @gol
625 -mdwarf2-asm -mearly-stop-bits @gol
626 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
627 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
628 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
629 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
630 -msched-spec-ldc -msched-spec-control-ldc @gol
631 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
632 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
633 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
634 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
636 @emph{IA-64/VMS Options}
637 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
640 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
641 -msign-extend-enabled -muser-enabled}
643 @emph{M32R/D Options}
644 @gccoptlist{-m32r2 -m32rx -m32r @gol
646 -malign-loops -mno-align-loops @gol
647 -missue-rate=@var{number} @gol
648 -mbranch-cost=@var{number} @gol
649 -mmodel=@var{code-size-model-type} @gol
650 -msdata=@var{sdata-type} @gol
651 -mno-flush-func -mflush-func=@var{name} @gol
652 -mno-flush-trap -mflush-trap=@var{number} @gol
656 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
658 @emph{M680x0 Options}
659 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
660 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
661 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
662 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
663 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
664 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
665 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
666 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
669 @emph{M68hc1x Options}
670 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
671 -mauto-incdec -minmax -mlong-calls -mshort @gol
672 -msoft-reg-count=@var{count}}
675 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
676 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
677 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
678 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
679 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
682 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
683 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
684 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
685 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
688 @emph{MicroBlaze Options}
689 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
690 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
691 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
692 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
693 -mxl-mode-@var{app-model}}
696 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
697 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
698 -mips64 -mips64r2 @gol
699 -mips16 -mno-mips16 -mflip-mips16 @gol
700 -minterlink-mips16 -mno-interlink-mips16 @gol
701 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
702 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
703 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
704 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
705 -mfpu=@var{fpu-type} @gol
706 -msmartmips -mno-smartmips @gol
707 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
708 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
709 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
710 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
711 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
712 -membedded-data -mno-embedded-data @gol
713 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
714 -mcode-readable=@var{setting} @gol
715 -msplit-addresses -mno-split-addresses @gol
716 -mexplicit-relocs -mno-explicit-relocs @gol
717 -mcheck-zero-division -mno-check-zero-division @gol
718 -mdivide-traps -mdivide-breaks @gol
719 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
720 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
721 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
722 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
723 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
724 -mflush-func=@var{func} -mno-flush-func @gol
725 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
726 -mfp-exceptions -mno-fp-exceptions @gol
727 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
728 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
731 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
732 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
733 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
734 -mno-base-addresses -msingle-exit -mno-single-exit}
736 @emph{MN10300 Options}
737 @gccoptlist{-mmult-bug -mno-mult-bug @gol
738 -mno-am33 -mam33 -mam33-2 -mam34 @gol
739 -mtune=@var{cpu-type} @gol
740 -mreturn-pointer-on-d0 @gol
743 @emph{PDP-11 Options}
744 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
745 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
746 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
747 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
748 -mbranch-expensive -mbranch-cheap @gol
749 -munix-asm -mdec-asm}
751 @emph{picoChip Options}
752 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
753 -msymbol-as-address -mno-inefficient-warnings}
755 @emph{PowerPC Options}
756 See RS/6000 and PowerPC Options.
758 @emph{RS/6000 and PowerPC Options}
759 @gccoptlist{-mcpu=@var{cpu-type} @gol
760 -mtune=@var{cpu-type} @gol
761 -mcmodel=@var{code-model} @gol
762 -mpower -mno-power -mpower2 -mno-power2 @gol
763 -mpowerpc -mpowerpc64 -mno-powerpc @gol
764 -maltivec -mno-altivec @gol
765 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
766 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
767 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
768 -mfprnd -mno-fprnd @gol
769 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
770 -mnew-mnemonics -mold-mnemonics @gol
771 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
772 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
773 -malign-power -malign-natural @gol
774 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
775 -msingle-float -mdouble-float -msimple-fpu @gol
776 -mstring -mno-string -mupdate -mno-update @gol
777 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
778 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
779 -mstrict-align -mno-strict-align -mrelocatable @gol
780 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
781 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
782 -mdynamic-no-pic -maltivec -mswdiv @gol
783 -mprioritize-restricted-insns=@var{priority} @gol
784 -msched-costly-dep=@var{dependence_type} @gol
785 -minsert-sched-nops=@var{scheme} @gol
786 -mcall-sysv -mcall-netbsd @gol
787 -maix-struct-return -msvr4-struct-return @gol
788 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
789 -mblock-move-inline-limit=@var{num} @gol
790 -misel -mno-isel @gol
791 -misel=yes -misel=no @gol
793 -mspe=yes -mspe=no @gol
795 -mgen-cell-microcode -mwarn-cell-microcode @gol
796 -mvrsave -mno-vrsave @gol
797 -mmulhw -mno-mulhw @gol
798 -mdlmzb -mno-dlmzb @gol
799 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
800 -mprototype -mno-prototype @gol
801 -msim -mmvme -mads -myellowknife -memb -msdata @gol
802 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
803 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision
804 -mno-recip-precision @gol
805 -mveclibabi=@var{type} -mfriz -mno-friz}
808 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
810 -mbig-endian-data -mlittle-endian-data @gol
813 -mas100-syntax -mno-as100-syntax@gol
815 -mmax-constant-size=@gol
817 -msave-acc-in-interrupts}
819 @emph{S/390 and zSeries Options}
820 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
821 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
822 -mlong-double-64 -mlong-double-128 @gol
823 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
824 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
825 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
826 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
827 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
830 @gccoptlist{-meb -mel @gol
834 -mscore5 -mscore5u -mscore7 -mscore7d}
837 @gccoptlist{-m1 -m2 -m2e @gol
838 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
840 -m4-nofpu -m4-single-only -m4-single -m4 @gol
841 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
842 -m5-64media -m5-64media-nofpu @gol
843 -m5-32media -m5-32media-nofpu @gol
844 -m5-compact -m5-compact-nofpu @gol
845 -mb -ml -mdalign -mrelax @gol
846 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
847 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
848 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
849 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
850 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
851 -maccumulate-outgoing-args -minvalid-symbols}
853 @emph{Solaris 2 Options}
854 @gccoptlist{-mimpure-text -mno-impure-text @gol
855 -threads -pthreads -pthread}
858 @gccoptlist{-mcpu=@var{cpu-type} @gol
859 -mtune=@var{cpu-type} @gol
860 -mcmodel=@var{code-model} @gol
861 -m32 -m64 -mapp-regs -mno-app-regs @gol
862 -mfaster-structs -mno-faster-structs @gol
863 -mfpu -mno-fpu -mhard-float -msoft-float @gol
864 -mhard-quad-float -msoft-quad-float @gol
866 -mstack-bias -mno-stack-bias @gol
867 -munaligned-doubles -mno-unaligned-doubles @gol
868 -mv8plus -mno-v8plus -mvis -mno-vis}
871 @gccoptlist{-mwarn-reloc -merror-reloc @gol
872 -msafe-dma -munsafe-dma @gol
874 -msmall-mem -mlarge-mem -mstdmain @gol
875 -mfixed-range=@var{register-range} @gol
877 -maddress-space-conversion -mno-address-space-conversion @gol
878 -mcache-size=@var{cache-size} @gol
879 -matomic-updates -mno-atomic-updates}
881 @emph{System V Options}
882 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
885 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
886 -mprolog-function -mno-prolog-function -mspace @gol
887 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
888 -mapp-regs -mno-app-regs @gol
889 -mdisable-callt -mno-disable-callt @gol
897 @gccoptlist{-mg -mgnu -munix}
899 @emph{VxWorks Options}
900 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
901 -Xbind-lazy -Xbind-now}
903 @emph{x86-64 Options}
904 See i386 and x86-64 Options.
906 @emph{i386 and x86-64 Windows Options}
907 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll
908 -mnop-fun-dllimport -mthread @gol
909 -municode -mwin32 -mwindows -fno-set-stack-executable}
911 @emph{Xstormy16 Options}
914 @emph{Xtensa Options}
915 @gccoptlist{-mconst16 -mno-const16 @gol
916 -mfused-madd -mno-fused-madd @gol
918 -mserialize-volatile -mno-serialize-volatile @gol
919 -mtext-section-literals -mno-text-section-literals @gol
920 -mtarget-align -mno-target-align @gol
921 -mlongcalls -mno-longcalls}
923 @emph{zSeries Options}
924 See S/390 and zSeries Options.
926 @item Code Generation Options
927 @xref{Code Gen Options,,Options for Code Generation Conventions}.
928 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
929 -ffixed-@var{reg} -fexceptions @gol
930 -fnon-call-exceptions -funwind-tables @gol
931 -fasynchronous-unwind-tables @gol
932 -finhibit-size-directive -finstrument-functions @gol
933 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
934 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
935 -fno-common -fno-ident @gol
936 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
937 -fno-jump-tables @gol
938 -frecord-gcc-switches @gol
939 -freg-struct-return -fshort-enums @gol
940 -fshort-double -fshort-wchar @gol
941 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
942 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
943 -fno-stack-limit -fsplit-stack @gol
944 -fleading-underscore -ftls-model=@var{model} @gol
945 -ftrapv -fwrapv -fbounds-check @gol
946 -fvisibility -fstrict-volatile-bitfields}
950 * Overall Options:: Controlling the kind of output:
951 an executable, object files, assembler files,
952 or preprocessed source.
953 * C Dialect Options:: Controlling the variant of C language compiled.
954 * C++ Dialect Options:: Variations on C++.
955 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
957 * Language Independent Options:: Controlling how diagnostics should be
959 * Warning Options:: How picky should the compiler be?
960 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
961 * Optimize Options:: How much optimization?
962 * Preprocessor Options:: Controlling header files and macro definitions.
963 Also, getting dependency information for Make.
964 * Assembler Options:: Passing options to the assembler.
965 * Link Options:: Specifying libraries and so on.
966 * Directory Options:: Where to find header files and libraries.
967 Where to find the compiler executable files.
968 * Spec Files:: How to pass switches to sub-processes.
969 * Target Options:: Running a cross-compiler, or an old version of GCC.
972 @node Overall Options
973 @section Options Controlling the Kind of Output
975 Compilation can involve up to four stages: preprocessing, compilation
976 proper, assembly and linking, always in that order. GCC is capable of
977 preprocessing and compiling several files either into several
978 assembler input files, or into one assembler input file; then each
979 assembler input file produces an object file, and linking combines all
980 the object files (those newly compiled, and those specified as input)
981 into an executable file.
983 @cindex file name suffix
984 For any given input file, the file name suffix determines what kind of
989 C source code which must be preprocessed.
992 C source code which should not be preprocessed.
995 C++ source code which should not be preprocessed.
998 Objective-C source code. Note that you must link with the @file{libobjc}
999 library to make an Objective-C program work.
1002 Objective-C source code which should not be preprocessed.
1006 Objective-C++ source code. Note that you must link with the @file{libobjc}
1007 library to make an Objective-C++ program work. Note that @samp{.M} refers
1008 to a literal capital M@.
1010 @item @var{file}.mii
1011 Objective-C++ source code which should not be preprocessed.
1014 C, C++, Objective-C or Objective-C++ header file to be turned into a
1015 precompiled header (default), or C, C++ header file to be turned into an
1016 Ada spec (via the @option{-fdump-ada-spec} switch).
1019 @itemx @var{file}.cp
1020 @itemx @var{file}.cxx
1021 @itemx @var{file}.cpp
1022 @itemx @var{file}.CPP
1023 @itemx @var{file}.c++
1025 C++ source code which must be preprocessed. Note that in @samp{.cxx},
1026 the last two letters must both be literally @samp{x}. Likewise,
1027 @samp{.C} refers to a literal capital C@.
1031 Objective-C++ source code which must be preprocessed.
1033 @item @var{file}.mii
1034 Objective-C++ source code which should not be preprocessed.
1038 @itemx @var{file}.hp
1039 @itemx @var{file}.hxx
1040 @itemx @var{file}.hpp
1041 @itemx @var{file}.HPP
1042 @itemx @var{file}.h++
1043 @itemx @var{file}.tcc
1044 C++ header file to be turned into a precompiled header or Ada spec.
1047 @itemx @var{file}.for
1048 @itemx @var{file}.ftn
1049 Fixed form Fortran source code which should not be preprocessed.
1052 @itemx @var{file}.FOR
1053 @itemx @var{file}.fpp
1054 @itemx @var{file}.FPP
1055 @itemx @var{file}.FTN
1056 Fixed form Fortran source code which must be preprocessed (with the traditional
1059 @item @var{file}.f90
1060 @itemx @var{file}.f95
1061 @itemx @var{file}.f03
1062 @itemx @var{file}.f08
1063 Free form Fortran source code which should not be preprocessed.
1065 @item @var{file}.F90
1066 @itemx @var{file}.F95
1067 @itemx @var{file}.F03
1068 @itemx @var{file}.F08
1069 Free form Fortran source code which must be preprocessed (with the
1070 traditional preprocessor).
1072 @c FIXME: Descriptions of Java file types.
1078 @item @var{file}.ads
1079 Ada source code file which contains a library unit declaration (a
1080 declaration of a package, subprogram, or generic, or a generic
1081 instantiation), or a library unit renaming declaration (a package,
1082 generic, or subprogram renaming declaration). Such files are also
1085 @item @var{file}.adb
1086 Ada source code file containing a library unit body (a subprogram or
1087 package body). Such files are also called @dfn{bodies}.
1089 @c GCC also knows about some suffixes for languages not yet included:
1100 @itemx @var{file}.sx
1101 Assembler code which must be preprocessed.
1104 An object file to be fed straight into linking.
1105 Any file name with no recognized suffix is treated this way.
1109 You can specify the input language explicitly with the @option{-x} option:
1112 @item -x @var{language}
1113 Specify explicitly the @var{language} for the following input files
1114 (rather than letting the compiler choose a default based on the file
1115 name suffix). This option applies to all following input files until
1116 the next @option{-x} option. Possible values for @var{language} are:
1118 c c-header cpp-output
1119 c++ c++-header c++-cpp-output
1120 objective-c objective-c-header objective-c-cpp-output
1121 objective-c++ objective-c++-header objective-c++-cpp-output
1122 assembler assembler-with-cpp
1124 f77 f77-cpp-input f95 f95-cpp-input
1129 Turn off any specification of a language, so that subsequent files are
1130 handled according to their file name suffixes (as they are if @option{-x}
1131 has not been used at all).
1133 @item -pass-exit-codes
1134 @opindex pass-exit-codes
1135 Normally the @command{gcc} program will exit with the code of 1 if any
1136 phase of the compiler returns a non-success return code. If you specify
1137 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1138 numerically highest error produced by any phase that returned an error
1139 indication. The C, C++, and Fortran frontends return 4, if an internal
1140 compiler error is encountered.
1143 If you only want some of the stages of compilation, you can use
1144 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1145 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1146 @command{gcc} is to stop. Note that some combinations (for example,
1147 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1152 Compile or assemble the source files, but do not link. The linking
1153 stage simply is not done. The ultimate output is in the form of an
1154 object file for each source file.
1156 By default, the object file name for a source file is made by replacing
1157 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1159 Unrecognized input files, not requiring compilation or assembly, are
1164 Stop after the stage of compilation proper; do not assemble. The output
1165 is in the form of an assembler code file for each non-assembler input
1168 By default, the assembler file name for a source file is made by
1169 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1171 Input files that don't require compilation are ignored.
1175 Stop after the preprocessing stage; do not run the compiler proper. The
1176 output is in the form of preprocessed source code, which is sent to the
1179 Input files which don't require preprocessing are ignored.
1181 @cindex output file option
1184 Place output in file @var{file}. This applies regardless to whatever
1185 sort of output is being produced, whether it be an executable file,
1186 an object file, an assembler file or preprocessed C code.
1188 If @option{-o} is not specified, the default is to put an executable
1189 file in @file{a.out}, the object file for
1190 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1191 assembler file in @file{@var{source}.s}, a precompiled header file in
1192 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1197 Print (on standard error output) the commands executed to run the stages
1198 of compilation. Also print the version number of the compiler driver
1199 program and of the preprocessor and the compiler proper.
1203 Like @option{-v} except the commands are not executed and arguments
1204 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1205 This is useful for shell scripts to capture the driver-generated command lines.
1209 Use pipes rather than temporary files for communication between the
1210 various stages of compilation. This fails to work on some systems where
1211 the assembler is unable to read from a pipe; but the GNU assembler has
1216 Print (on the standard output) a description of the command line options
1217 understood by @command{gcc}. If the @option{-v} option is also specified
1218 then @option{--help} will also be passed on to the various processes
1219 invoked by @command{gcc}, so that they can display the command line options
1220 they accept. If the @option{-Wextra} option has also been specified
1221 (prior to the @option{--help} option), then command line options which
1222 have no documentation associated with them will also be displayed.
1225 @opindex target-help
1226 Print (on the standard output) a description of target-specific command
1227 line options for each tool. For some targets extra target-specific
1228 information may also be printed.
1230 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1231 Print (on the standard output) a description of the command line
1232 options understood by the compiler that fit into all specified classes
1233 and qualifiers. These are the supported classes:
1236 @item @samp{optimizers}
1237 This will display all of the optimization options supported by the
1240 @item @samp{warnings}
1241 This will display all of the options controlling warning messages
1242 produced by the compiler.
1245 This will display target-specific options. Unlike the
1246 @option{--target-help} option however, target-specific options of the
1247 linker and assembler will not be displayed. This is because those
1248 tools do not currently support the extended @option{--help=} syntax.
1251 This will display the values recognized by the @option{--param}
1254 @item @var{language}
1255 This will display the options supported for @var{language}, where
1256 @var{language} is the name of one of the languages supported in this
1260 This will display the options that are common to all languages.
1263 These are the supported qualifiers:
1266 @item @samp{undocumented}
1267 Display only those options which are undocumented.
1270 Display options which take an argument that appears after an equal
1271 sign in the same continuous piece of text, such as:
1272 @samp{--help=target}.
1274 @item @samp{separate}
1275 Display options which take an argument that appears as a separate word
1276 following the original option, such as: @samp{-o output-file}.
1279 Thus for example to display all the undocumented target-specific
1280 switches supported by the compiler the following can be used:
1283 --help=target,undocumented
1286 The sense of a qualifier can be inverted by prefixing it with the
1287 @samp{^} character, so for example to display all binary warning
1288 options (i.e., ones that are either on or off and that do not take an
1289 argument), which have a description the following can be used:
1292 --help=warnings,^joined,^undocumented
1295 The argument to @option{--help=} should not consist solely of inverted
1298 Combining several classes is possible, although this usually
1299 restricts the output by so much that there is nothing to display. One
1300 case where it does work however is when one of the classes is
1301 @var{target}. So for example to display all the target-specific
1302 optimization options the following can be used:
1305 --help=target,optimizers
1308 The @option{--help=} option can be repeated on the command line. Each
1309 successive use will display its requested class of options, skipping
1310 those that have already been displayed.
1312 If the @option{-Q} option appears on the command line before the
1313 @option{--help=} option, then the descriptive text displayed by
1314 @option{--help=} is changed. Instead of describing the displayed
1315 options, an indication is given as to whether the option is enabled,
1316 disabled or set to a specific value (assuming that the compiler
1317 knows this at the point where the @option{--help=} option is used).
1319 Here is a truncated example from the ARM port of @command{gcc}:
1322 % gcc -Q -mabi=2 --help=target -c
1323 The following options are target specific:
1325 -mabort-on-noreturn [disabled]
1329 The output is sensitive to the effects of previous command line
1330 options, so for example it is possible to find out which optimizations
1331 are enabled at @option{-O2} by using:
1334 -Q -O2 --help=optimizers
1337 Alternatively you can discover which binary optimizations are enabled
1338 by @option{-O3} by using:
1341 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1342 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1343 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1346 @item -no-canonical-prefixes
1347 @opindex no-canonical-prefixes
1348 Do not expand any symbolic links, resolve references to @samp{/../}
1349 or @samp{/./}, or make the path absolute when generating a relative
1354 Display the version number and copyrights of the invoked GCC@.
1358 Invoke all subcommands under a wrapper program. It takes a single
1359 comma separated list as an argument, which will be used to invoke
1363 gcc -c t.c -wrapper gdb,--args
1366 This will invoke all subprograms of gcc under "gdb --args",
1367 thus cc1 invocation will be "gdb --args cc1 ...".
1369 @item -fplugin=@var{name}.so
1370 Load the plugin code in file @var{name}.so, assumed to be a
1371 shared object to be dlopen'd by the compiler. The base name of
1372 the shared object file is used to identify the plugin for the
1373 purposes of argument parsing (See
1374 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1375 Each plugin should define the callback functions specified in the
1378 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1379 Define an argument called @var{key} with a value of @var{value}
1380 for the plugin called @var{name}.
1382 @item -fdump-ada-spec@r{[}-slim@r{]}
1383 For C and C++ source and include files, generate corresponding Ada
1384 specs. @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1385 GNAT User's Guide}, which provides detailed documentation on this feature.
1387 @include @value{srcdir}/../libiberty/at-file.texi
1391 @section Compiling C++ Programs
1393 @cindex suffixes for C++ source
1394 @cindex C++ source file suffixes
1395 C++ source files conventionally use one of the suffixes @samp{.C},
1396 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1397 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1398 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1399 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1400 files with these names and compiles them as C++ programs even if you
1401 call the compiler the same way as for compiling C programs (usually
1402 with the name @command{gcc}).
1406 However, the use of @command{gcc} does not add the C++ library.
1407 @command{g++} is a program that calls GCC and treats @samp{.c},
1408 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1409 files unless @option{-x} is used, and automatically specifies linking
1410 against the C++ library. This program is also useful when
1411 precompiling a C header file with a @samp{.h} extension for use in C++
1412 compilations. On many systems, @command{g++} is also installed with
1413 the name @command{c++}.
1415 @cindex invoking @command{g++}
1416 When you compile C++ programs, you may specify many of the same
1417 command-line options that you use for compiling programs in any
1418 language; or command-line options meaningful for C and related
1419 languages; or options that are meaningful only for C++ programs.
1420 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1421 explanations of options for languages related to C@.
1422 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1423 explanations of options that are meaningful only for C++ programs.
1425 @node C Dialect Options
1426 @section Options Controlling C Dialect
1427 @cindex dialect options
1428 @cindex language dialect options
1429 @cindex options, dialect
1431 The following options control the dialect of C (or languages derived
1432 from C, such as C++, Objective-C and Objective-C++) that the compiler
1436 @cindex ANSI support
1440 In C mode, this is equivalent to @samp{-std=c90}. In C++ mode, it is
1441 equivalent to @samp{-std=c++98}.
1443 This turns off certain features of GCC that are incompatible with ISO
1444 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1445 such as the @code{asm} and @code{typeof} keywords, and
1446 predefined macros such as @code{unix} and @code{vax} that identify the
1447 type of system you are using. It also enables the undesirable and
1448 rarely used ISO trigraph feature. For the C compiler,
1449 it disables recognition of C++ style @samp{//} comments as well as
1450 the @code{inline} keyword.
1452 The alternate keywords @code{__asm__}, @code{__extension__},
1453 @code{__inline__} and @code{__typeof__} continue to work despite
1454 @option{-ansi}. You would not want to use them in an ISO C program, of
1455 course, but it is useful to put them in header files that might be included
1456 in compilations done with @option{-ansi}. Alternate predefined macros
1457 such as @code{__unix__} and @code{__vax__} are also available, with or
1458 without @option{-ansi}.
1460 The @option{-ansi} option does not cause non-ISO programs to be
1461 rejected gratuitously. For that, @option{-pedantic} is required in
1462 addition to @option{-ansi}. @xref{Warning Options}.
1464 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1465 option is used. Some header files may notice this macro and refrain
1466 from declaring certain functions or defining certain macros that the
1467 ISO standard doesn't call for; this is to avoid interfering with any
1468 programs that might use these names for other things.
1470 Functions that would normally be built in but do not have semantics
1471 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1472 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1473 built-in functions provided by GCC}, for details of the functions
1478 Determine the language standard. @xref{Standards,,Language Standards
1479 Supported by GCC}, for details of these standard versions. This option
1480 is currently only supported when compiling C or C++.
1482 The compiler can accept several base standards, such as @samp{c90} or
1483 @samp{c++98}, and GNU dialects of those standards, such as
1484 @samp{gnu90} or @samp{gnu++98}. By specifying a base standard, the
1485 compiler will accept all programs following that standard and those
1486 using GNU extensions that do not contradict it. For example,
1487 @samp{-std=c90} turns off certain features of GCC that are
1488 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1489 keywords, but not other GNU extensions that do not have a meaning in
1490 ISO C90, such as omitting the middle term of a @code{?:}
1491 expression. On the other hand, by specifying a GNU dialect of a
1492 standard, all features the compiler support are enabled, even when
1493 those features change the meaning of the base standard and some
1494 strict-conforming programs may be rejected. The particular standard
1495 is used by @option{-pedantic} to identify which features are GNU
1496 extensions given that version of the standard. For example
1497 @samp{-std=gnu90 -pedantic} would warn about C++ style @samp{//}
1498 comments, while @samp{-std=gnu99 -pedantic} would not.
1500 A value for this option must be provided; possible values are
1506 Support all ISO C90 programs (certain GNU extensions that conflict
1507 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1509 @item iso9899:199409
1510 ISO C90 as modified in amendment 1.
1516 ISO C99. Note that this standard is not yet fully supported; see
1517 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1518 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1521 ISO C1X, the draft of the next revision of the ISO C standard.
1522 Support is limited and experimental and features enabled by this
1523 option may be changed or removed if changed in or removed from the
1528 GNU dialect of ISO C90 (including some C99 features). This
1529 is the default for C code.
1533 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1534 this will become the default. The name @samp{gnu9x} is deprecated.
1537 GNU dialect of ISO C1X. Support is limited and experimental and
1538 features enabled by this option may be changed or removed if changed
1539 in or removed from the standard draft.
1542 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1546 GNU dialect of @option{-std=c++98}. This is the default for
1550 The working draft of the upcoming ISO C++0x standard. This option
1551 enables experimental features that are likely to be included in
1552 C++0x. The working draft is constantly changing, and any feature that is
1553 enabled by this flag may be removed from future versions of GCC if it is
1554 not part of the C++0x standard.
1557 GNU dialect of @option{-std=c++0x}. This option enables
1558 experimental features that may be removed in future versions of GCC.
1561 @item -fgnu89-inline
1562 @opindex fgnu89-inline
1563 The option @option{-fgnu89-inline} tells GCC to use the traditional
1564 GNU semantics for @code{inline} functions when in C99 mode.
1565 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1566 is accepted and ignored by GCC versions 4.1.3 up to but not including
1567 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1568 C99 mode. Using this option is roughly equivalent to adding the
1569 @code{gnu_inline} function attribute to all inline functions
1570 (@pxref{Function Attributes}).
1572 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1573 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1574 specifies the default behavior). This option was first supported in
1575 GCC 4.3. This option is not supported in @option{-std=c90} or
1576 @option{-std=gnu90} mode.
1578 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1579 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1580 in effect for @code{inline} functions. @xref{Common Predefined
1581 Macros,,,cpp,The C Preprocessor}.
1583 @item -aux-info @var{filename}
1585 Output to the given filename prototyped declarations for all functions
1586 declared and/or defined in a translation unit, including those in header
1587 files. This option is silently ignored in any language other than C@.
1589 Besides declarations, the file indicates, in comments, the origin of
1590 each declaration (source file and line), whether the declaration was
1591 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1592 @samp{O} for old, respectively, in the first character after the line
1593 number and the colon), and whether it came from a declaration or a
1594 definition (@samp{C} or @samp{F}, respectively, in the following
1595 character). In the case of function definitions, a K&R-style list of
1596 arguments followed by their declarations is also provided, inside
1597 comments, after the declaration.
1601 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1602 keyword, so that code can use these words as identifiers. You can use
1603 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1604 instead. @option{-ansi} implies @option{-fno-asm}.
1606 In C++, this switch only affects the @code{typeof} keyword, since
1607 @code{asm} and @code{inline} are standard keywords. You may want to
1608 use the @option{-fno-gnu-keywords} flag instead, which has the same
1609 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1610 switch only affects the @code{asm} and @code{typeof} keywords, since
1611 @code{inline} is a standard keyword in ISO C99.
1614 @itemx -fno-builtin-@var{function}
1615 @opindex fno-builtin
1616 @cindex built-in functions
1617 Don't recognize built-in functions that do not begin with
1618 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1619 functions provided by GCC}, for details of the functions affected,
1620 including those which are not built-in functions when @option{-ansi} or
1621 @option{-std} options for strict ISO C conformance are used because they
1622 do not have an ISO standard meaning.
1624 GCC normally generates special code to handle certain built-in functions
1625 more efficiently; for instance, calls to @code{alloca} may become single
1626 instructions that adjust the stack directly, and calls to @code{memcpy}
1627 may become inline copy loops. The resulting code is often both smaller
1628 and faster, but since the function calls no longer appear as such, you
1629 cannot set a breakpoint on those calls, nor can you change the behavior
1630 of the functions by linking with a different library. In addition,
1631 when a function is recognized as a built-in function, GCC may use
1632 information about that function to warn about problems with calls to
1633 that function, or to generate more efficient code, even if the
1634 resulting code still contains calls to that function. For example,
1635 warnings are given with @option{-Wformat} for bad calls to
1636 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1637 known not to modify global memory.
1639 With the @option{-fno-builtin-@var{function}} option
1640 only the built-in function @var{function} is
1641 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1642 function is named that is not built-in in this version of GCC, this
1643 option is ignored. There is no corresponding
1644 @option{-fbuiltin-@var{function}} option; if you wish to enable
1645 built-in functions selectively when using @option{-fno-builtin} or
1646 @option{-ffreestanding}, you may define macros such as:
1649 #define abs(n) __builtin_abs ((n))
1650 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1655 @cindex hosted environment
1657 Assert that compilation takes place in a hosted environment. This implies
1658 @option{-fbuiltin}. A hosted environment is one in which the
1659 entire standard library is available, and in which @code{main} has a return
1660 type of @code{int}. Examples are nearly everything except a kernel.
1661 This is equivalent to @option{-fno-freestanding}.
1663 @item -ffreestanding
1664 @opindex ffreestanding
1665 @cindex hosted environment
1667 Assert that compilation takes place in a freestanding environment. This
1668 implies @option{-fno-builtin}. A freestanding environment
1669 is one in which the standard library may not exist, and program startup may
1670 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1671 This is equivalent to @option{-fno-hosted}.
1673 @xref{Standards,,Language Standards Supported by GCC}, for details of
1674 freestanding and hosted environments.
1678 @cindex OpenMP parallel
1679 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1680 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1681 compiler generates parallel code according to the OpenMP Application
1682 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1683 implies @option{-pthread}, and thus is only supported on targets that
1684 have support for @option{-pthread}.
1686 @item -fms-extensions
1687 @opindex fms-extensions
1688 Accept some non-standard constructs used in Microsoft header files.
1690 It allows for c++ that member-names in structures can be similiar
1691 to previous types declarations.
1700 Some cases of unnamed fields in structures and unions are only
1701 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1702 fields within structs/unions}, for details.
1704 @item -fplan9-extensions
1705 Accept some non-standard constructs used in Plan 9 code.
1707 This enables @option{-fms-extensions}, permits passing pointers to
1708 structures with anonymous fields to functions which expect pointers to
1709 elements of the type of the field, and permits referring to anonymous
1710 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
1711 struct/union fields within structs/unions}, for details. This is only
1712 supported for C, not C++.
1716 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1717 options for strict ISO C conformance) implies @option{-trigraphs}.
1719 @item -no-integrated-cpp
1720 @opindex no-integrated-cpp
1721 Performs a compilation in two passes: preprocessing and compiling. This
1722 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1723 @option{-B} option. The user supplied compilation step can then add in
1724 an additional preprocessing step after normal preprocessing but before
1725 compiling. The default is to use the integrated cpp (internal cpp)
1727 The semantics of this option will change if "cc1", "cc1plus", and
1728 "cc1obj" are merged.
1730 @cindex traditional C language
1731 @cindex C language, traditional
1733 @itemx -traditional-cpp
1734 @opindex traditional-cpp
1735 @opindex traditional
1736 Formerly, these options caused GCC to attempt to emulate a pre-standard
1737 C compiler. They are now only supported with the @option{-E} switch.
1738 The preprocessor continues to support a pre-standard mode. See the GNU
1739 CPP manual for details.
1741 @item -fcond-mismatch
1742 @opindex fcond-mismatch
1743 Allow conditional expressions with mismatched types in the second and
1744 third arguments. The value of such an expression is void. This option
1745 is not supported for C++.
1747 @item -flax-vector-conversions
1748 @opindex flax-vector-conversions
1749 Allow implicit conversions between vectors with differing numbers of
1750 elements and/or incompatible element types. This option should not be
1753 @item -funsigned-char
1754 @opindex funsigned-char
1755 Let the type @code{char} be unsigned, like @code{unsigned char}.
1757 Each kind of machine has a default for what @code{char} should
1758 be. It is either like @code{unsigned char} by default or like
1759 @code{signed char} by default.
1761 Ideally, a portable program should always use @code{signed char} or
1762 @code{unsigned char} when it depends on the signedness of an object.
1763 But many programs have been written to use plain @code{char} and
1764 expect it to be signed, or expect it to be unsigned, depending on the
1765 machines they were written for. This option, and its inverse, let you
1766 make such a program work with the opposite default.
1768 The type @code{char} is always a distinct type from each of
1769 @code{signed char} or @code{unsigned char}, even though its behavior
1770 is always just like one of those two.
1773 @opindex fsigned-char
1774 Let the type @code{char} be signed, like @code{signed char}.
1776 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1777 the negative form of @option{-funsigned-char}. Likewise, the option
1778 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1780 @item -fsigned-bitfields
1781 @itemx -funsigned-bitfields
1782 @itemx -fno-signed-bitfields
1783 @itemx -fno-unsigned-bitfields
1784 @opindex fsigned-bitfields
1785 @opindex funsigned-bitfields
1786 @opindex fno-signed-bitfields
1787 @opindex fno-unsigned-bitfields
1788 These options control whether a bit-field is signed or unsigned, when the
1789 declaration does not use either @code{signed} or @code{unsigned}. By
1790 default, such a bit-field is signed, because this is consistent: the
1791 basic integer types such as @code{int} are signed types.
1794 @node C++ Dialect Options
1795 @section Options Controlling C++ Dialect
1797 @cindex compiler options, C++
1798 @cindex C++ options, command line
1799 @cindex options, C++
1800 This section describes the command-line options that are only meaningful
1801 for C++ programs; but you can also use most of the GNU compiler options
1802 regardless of what language your program is in. For example, you
1803 might compile a file @code{firstClass.C} like this:
1806 g++ -g -frepo -O -c firstClass.C
1810 In this example, only @option{-frepo} is an option meant
1811 only for C++ programs; you can use the other options with any
1812 language supported by GCC@.
1814 Here is a list of options that are @emph{only} for compiling C++ programs:
1818 @item -fabi-version=@var{n}
1819 @opindex fabi-version
1820 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1821 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1822 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1823 the version that conforms most closely to the C++ ABI specification.
1824 Therefore, the ABI obtained using version 0 will change as ABI bugs
1827 The default is version 2.
1829 Version 3 corrects an error in mangling a constant address as a
1832 Version 4 implements a standard mangling for vector types.
1834 See also @option{-Wabi}.
1836 @item -fno-access-control
1837 @opindex fno-access-control
1838 Turn off all access checking. This switch is mainly useful for working
1839 around bugs in the access control code.
1843 Check that the pointer returned by @code{operator new} is non-null
1844 before attempting to modify the storage allocated. This check is
1845 normally unnecessary because the C++ standard specifies that
1846 @code{operator new} will only return @code{0} if it is declared
1847 @samp{throw()}, in which case the compiler will always check the
1848 return value even without this option. In all other cases, when
1849 @code{operator new} has a non-empty exception specification, memory
1850 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1851 @samp{new (nothrow)}.
1853 @item -fconserve-space
1854 @opindex fconserve-space
1855 Put uninitialized or runtime-initialized global variables into the
1856 common segment, as C does. This saves space in the executable at the
1857 cost of not diagnosing duplicate definitions. If you compile with this
1858 flag and your program mysteriously crashes after @code{main()} has
1859 completed, you may have an object that is being destroyed twice because
1860 two definitions were merged.
1862 This option is no longer useful on most targets, now that support has
1863 been added for putting variables into BSS without making them common.
1865 @item -fno-deduce-init-list
1866 @opindex fno-deduce-init-list
1867 Disable deduction of a template type parameter as
1868 std::initializer_list from a brace-enclosed initializer list, i.e.
1871 template <class T> auto forward(T t) -> decltype (realfn (t))
1878 forward(@{1,2@}); // call forward<std::initializer_list<int>>
1882 This option is present because this deduction is an extension to the
1883 current specification in the C++0x working draft, and there was
1884 some concern about potential overload resolution problems.
1886 @item -ffriend-injection
1887 @opindex ffriend-injection
1888 Inject friend functions into the enclosing namespace, so that they are
1889 visible outside the scope of the class in which they are declared.
1890 Friend functions were documented to work this way in the old Annotated
1891 C++ Reference Manual, and versions of G++ before 4.1 always worked
1892 that way. However, in ISO C++ a friend function which is not declared
1893 in an enclosing scope can only be found using argument dependent
1894 lookup. This option causes friends to be injected as they were in
1897 This option is for compatibility, and may be removed in a future
1900 @item -fno-elide-constructors
1901 @opindex fno-elide-constructors
1902 The C++ standard allows an implementation to omit creating a temporary
1903 which is only used to initialize another object of the same type.
1904 Specifying this option disables that optimization, and forces G++ to
1905 call the copy constructor in all cases.
1907 @item -fno-enforce-eh-specs
1908 @opindex fno-enforce-eh-specs
1909 Don't generate code to check for violation of exception specifications
1910 at runtime. This option violates the C++ standard, but may be useful
1911 for reducing code size in production builds, much like defining
1912 @samp{NDEBUG}. This does not give user code permission to throw
1913 exceptions in violation of the exception specifications; the compiler
1914 will still optimize based on the specifications, so throwing an
1915 unexpected exception will result in undefined behavior.
1918 @itemx -fno-for-scope
1920 @opindex fno-for-scope
1921 If @option{-ffor-scope} is specified, the scope of variables declared in
1922 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1923 as specified by the C++ standard.
1924 If @option{-fno-for-scope} is specified, the scope of variables declared in
1925 a @i{for-init-statement} extends to the end of the enclosing scope,
1926 as was the case in old versions of G++, and other (traditional)
1927 implementations of C++.
1929 The default if neither flag is given to follow the standard,
1930 but to allow and give a warning for old-style code that would
1931 otherwise be invalid, or have different behavior.
1933 @item -fno-gnu-keywords
1934 @opindex fno-gnu-keywords
1935 Do not recognize @code{typeof} as a keyword, so that code can use this
1936 word as an identifier. You can use the keyword @code{__typeof__} instead.
1937 @option{-ansi} implies @option{-fno-gnu-keywords}.
1939 @item -fno-implicit-templates
1940 @opindex fno-implicit-templates
1941 Never emit code for non-inline templates which are instantiated
1942 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1943 @xref{Template Instantiation}, for more information.
1945 @item -fno-implicit-inline-templates
1946 @opindex fno-implicit-inline-templates
1947 Don't emit code for implicit instantiations of inline templates, either.
1948 The default is to handle inlines differently so that compiles with and
1949 without optimization will need the same set of explicit instantiations.
1951 @item -fno-implement-inlines
1952 @opindex fno-implement-inlines
1953 To save space, do not emit out-of-line copies of inline functions
1954 controlled by @samp{#pragma implementation}. This will cause linker
1955 errors if these functions are not inlined everywhere they are called.
1957 @item -fms-extensions
1958 @opindex fms-extensions
1959 Disable pedantic warnings about constructs used in MFC, such as implicit
1960 int and getting a pointer to member function via non-standard syntax.
1962 @item -fno-nonansi-builtins
1963 @opindex fno-nonansi-builtins
1964 Disable built-in declarations of functions that are not mandated by
1965 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1966 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1969 @opindex fnothrow-opt
1970 Treat a @code{throw()} exception specification as though it were a
1971 @code{noexcept} specification to reduce or eliminate the text size
1972 overhead relative to a function with no exception specification. If
1973 the function has local variables of types with non-trivial
1974 destructors, the exception specification will actually make the
1975 function smaller because the EH cleanups for those variables can be
1976 optimized away. The semantic effect is that an exception thrown out of
1977 a function with such an exception specification will result in a call
1978 to @code{terminate} rather than @code{unexpected}.
1980 @item -fno-operator-names
1981 @opindex fno-operator-names
1982 Do not treat the operator name keywords @code{and}, @code{bitand},
1983 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1984 synonyms as keywords.
1986 @item -fno-optional-diags
1987 @opindex fno-optional-diags
1988 Disable diagnostics that the standard says a compiler does not need to
1989 issue. Currently, the only such diagnostic issued by G++ is the one for
1990 a name having multiple meanings within a class.
1993 @opindex fpermissive
1994 Downgrade some diagnostics about nonconformant code from errors to
1995 warnings. Thus, using @option{-fpermissive} will allow some
1996 nonconforming code to compile.
1998 @item -fno-pretty-templates
1999 @opindex fno-pretty-templates
2000 When an error message refers to a specialization of a function
2001 template, the compiler will normally print the signature of the
2002 template followed by the template arguments and any typedefs or
2003 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2004 rather than @code{void f(int)}) so that it's clear which template is
2005 involved. When an error message refers to a specialization of a class
2006 template, the compiler will omit any template arguments which match
2007 the default template arguments for that template. If either of these
2008 behaviors make it harder to understand the error message rather than
2009 easier, using @option{-fno-pretty-templates} will disable them.
2013 Enable automatic template instantiation at link time. This option also
2014 implies @option{-fno-implicit-templates}. @xref{Template
2015 Instantiation}, for more information.
2019 Disable generation of information about every class with virtual
2020 functions for use by the C++ runtime type identification features
2021 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
2022 of the language, you can save some space by using this flag. Note that
2023 exception handling uses the same information, but it will generate it as
2024 needed. The @samp{dynamic_cast} operator can still be used for casts that
2025 do not require runtime type information, i.e.@: casts to @code{void *} or to
2026 unambiguous base classes.
2030 Emit statistics about front-end processing at the end of the compilation.
2031 This information is generally only useful to the G++ development team.
2033 @item -fstrict-enums
2034 @opindex fstrict-enums
2035 Allow the compiler to optimize using the assumption that a value of
2036 enumeration type can only be one of the values of the enumeration (as
2037 defined in the C++ standard; basically, a value which can be
2038 represented in the minimum number of bits needed to represent all the
2039 enumerators). This assumption may not be valid if the program uses a
2040 cast to convert an arbitrary integer value to the enumeration type.
2042 @item -ftemplate-depth=@var{n}
2043 @opindex ftemplate-depth
2044 Set the maximum instantiation depth for template classes to @var{n}.
2045 A limit on the template instantiation depth is needed to detect
2046 endless recursions during template class instantiation. ANSI/ISO C++
2047 conforming programs must not rely on a maximum depth greater than 17
2048 (changed to 1024 in C++0x).
2050 @item -fno-threadsafe-statics
2051 @opindex fno-threadsafe-statics
2052 Do not emit the extra code to use the routines specified in the C++
2053 ABI for thread-safe initialization of local statics. You can use this
2054 option to reduce code size slightly in code that doesn't need to be
2057 @item -fuse-cxa-atexit
2058 @opindex fuse-cxa-atexit
2059 Register destructors for objects with static storage duration with the
2060 @code{__cxa_atexit} function rather than the @code{atexit} function.
2061 This option is required for fully standards-compliant handling of static
2062 destructors, but will only work if your C library supports
2063 @code{__cxa_atexit}.
2065 @item -fno-use-cxa-get-exception-ptr
2066 @opindex fno-use-cxa-get-exception-ptr
2067 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2068 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
2069 if the runtime routine is not available.
2071 @item -fvisibility-inlines-hidden
2072 @opindex fvisibility-inlines-hidden
2073 This switch declares that the user does not attempt to compare
2074 pointers to inline methods where the addresses of the two functions
2075 were taken in different shared objects.
2077 The effect of this is that GCC may, effectively, mark inline methods with
2078 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2079 appear in the export table of a DSO and do not require a PLT indirection
2080 when used within the DSO@. Enabling this option can have a dramatic effect
2081 on load and link times of a DSO as it massively reduces the size of the
2082 dynamic export table when the library makes heavy use of templates.
2084 The behavior of this switch is not quite the same as marking the
2085 methods as hidden directly, because it does not affect static variables
2086 local to the function or cause the compiler to deduce that
2087 the function is defined in only one shared object.
2089 You may mark a method as having a visibility explicitly to negate the
2090 effect of the switch for that method. For example, if you do want to
2091 compare pointers to a particular inline method, you might mark it as
2092 having default visibility. Marking the enclosing class with explicit
2093 visibility will have no effect.
2095 Explicitly instantiated inline methods are unaffected by this option
2096 as their linkage might otherwise cross a shared library boundary.
2097 @xref{Template Instantiation}.
2099 @item -fvisibility-ms-compat
2100 @opindex fvisibility-ms-compat
2101 This flag attempts to use visibility settings to make GCC's C++
2102 linkage model compatible with that of Microsoft Visual Studio.
2104 The flag makes these changes to GCC's linkage model:
2108 It sets the default visibility to @code{hidden}, like
2109 @option{-fvisibility=hidden}.
2112 Types, but not their members, are not hidden by default.
2115 The One Definition Rule is relaxed for types without explicit
2116 visibility specifications which are defined in more than one different
2117 shared object: those declarations are permitted if they would have
2118 been permitted when this option was not used.
2121 In new code it is better to use @option{-fvisibility=hidden} and
2122 export those classes which are intended to be externally visible.
2123 Unfortunately it is possible for code to rely, perhaps accidentally,
2124 on the Visual Studio behavior.
2126 Among the consequences of these changes are that static data members
2127 of the same type with the same name but defined in different shared
2128 objects will be different, so changing one will not change the other;
2129 and that pointers to function members defined in different shared
2130 objects may not compare equal. When this flag is given, it is a
2131 violation of the ODR to define types with the same name differently.
2135 Do not use weak symbol support, even if it is provided by the linker.
2136 By default, G++ will use weak symbols if they are available. This
2137 option exists only for testing, and should not be used by end-users;
2138 it will result in inferior code and has no benefits. This option may
2139 be removed in a future release of G++.
2143 Do not search for header files in the standard directories specific to
2144 C++, but do still search the other standard directories. (This option
2145 is used when building the C++ library.)
2148 In addition, these optimization, warning, and code generation options
2149 have meanings only for C++ programs:
2152 @item -fno-default-inline
2153 @opindex fno-default-inline
2154 Do not assume @samp{inline} for functions defined inside a class scope.
2155 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2156 functions will have linkage like inline functions; they just won't be
2159 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2162 Warn when G++ generates code that is probably not compatible with the
2163 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2164 all such cases, there are probably some cases that are not warned about,
2165 even though G++ is generating incompatible code. There may also be
2166 cases where warnings are emitted even though the code that is generated
2169 You should rewrite your code to avoid these warnings if you are
2170 concerned about the fact that code generated by G++ may not be binary
2171 compatible with code generated by other compilers.
2173 The known incompatibilities in @option{-fabi-version=2} (the default) include:
2178 A template with a non-type template parameter of reference type is
2179 mangled incorrectly:
2182 template <int &> struct S @{@};
2186 This is fixed in @option{-fabi-version=3}.
2189 SIMD vector types declared using @code{__attribute ((vector_size))} are
2190 mangled in a non-standard way that does not allow for overloading of
2191 functions taking vectors of different sizes.
2193 The mangling is changed in @option{-fabi-version=4}.
2196 The known incompatibilities in @option{-fabi-version=1} include:
2201 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2202 pack data into the same byte as a base class. For example:
2205 struct A @{ virtual void f(); int f1 : 1; @};
2206 struct B : public A @{ int f2 : 1; @};
2210 In this case, G++ will place @code{B::f2} into the same byte
2211 as@code{A::f1}; other compilers will not. You can avoid this problem
2212 by explicitly padding @code{A} so that its size is a multiple of the
2213 byte size on your platform; that will cause G++ and other compilers to
2214 layout @code{B} identically.
2217 Incorrect handling of tail-padding for virtual bases. G++ does not use
2218 tail padding when laying out virtual bases. For example:
2221 struct A @{ virtual void f(); char c1; @};
2222 struct B @{ B(); char c2; @};
2223 struct C : public A, public virtual B @{@};
2227 In this case, G++ will not place @code{B} into the tail-padding for
2228 @code{A}; other compilers will. You can avoid this problem by
2229 explicitly padding @code{A} so that its size is a multiple of its
2230 alignment (ignoring virtual base classes); that will cause G++ and other
2231 compilers to layout @code{C} identically.
2234 Incorrect handling of bit-fields with declared widths greater than that
2235 of their underlying types, when the bit-fields appear in a union. For
2239 union U @{ int i : 4096; @};
2243 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2244 union too small by the number of bits in an @code{int}.
2247 Empty classes can be placed at incorrect offsets. For example:
2257 struct C : public B, public A @{@};
2261 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2262 it should be placed at offset zero. G++ mistakenly believes that the
2263 @code{A} data member of @code{B} is already at offset zero.
2266 Names of template functions whose types involve @code{typename} or
2267 template template parameters can be mangled incorrectly.
2270 template <typename Q>
2271 void f(typename Q::X) @{@}
2273 template <template <typename> class Q>
2274 void f(typename Q<int>::X) @{@}
2278 Instantiations of these templates may be mangled incorrectly.
2282 It also warns psABI related changes. The known psABI changes at this
2288 For SYSV/x86-64, when passing union with long double, it is changed to
2289 pass in memory as specified in psABI. For example:
2299 @code{union U} will always be passed in memory.
2303 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2304 @opindex Wctor-dtor-privacy
2305 @opindex Wno-ctor-dtor-privacy
2306 Warn when a class seems unusable because all the constructors or
2307 destructors in that class are private, and it has neither friends nor
2308 public static member functions.
2310 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2312 @opindex Wno-noexcept
2313 Warn when a noexcept-expression evaluates to false because of a call
2314 to a function that does not have a non-throwing exception
2315 specification (i.e. @samp{throw()} or @samp{noexcept}) but is known by
2316 the compiler to never throw an exception.
2318 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2319 @opindex Wnon-virtual-dtor
2320 @opindex Wno-non-virtual-dtor
2321 Warn when a class has virtual functions and accessible non-virtual
2322 destructor, in which case it would be possible but unsafe to delete
2323 an instance of a derived class through a pointer to the base class.
2324 This warning is also enabled if -Weffc++ is specified.
2326 @item -Wreorder @r{(C++ and Objective-C++ only)}
2328 @opindex Wno-reorder
2329 @cindex reordering, warning
2330 @cindex warning for reordering of member initializers
2331 Warn when the order of member initializers given in the code does not
2332 match the order in which they must be executed. For instance:
2338 A(): j (0), i (1) @{ @}
2342 The compiler will rearrange the member initializers for @samp{i}
2343 and @samp{j} to match the declaration order of the members, emitting
2344 a warning to that effect. This warning is enabled by @option{-Wall}.
2347 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2350 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2353 Warn about violations of the following style guidelines from Scott Meyers'
2354 @cite{Effective C++} book:
2358 Item 11: Define a copy constructor and an assignment operator for classes
2359 with dynamically allocated memory.
2362 Item 12: Prefer initialization to assignment in constructors.
2365 Item 14: Make destructors virtual in base classes.
2368 Item 15: Have @code{operator=} return a reference to @code{*this}.
2371 Item 23: Don't try to return a reference when you must return an object.
2375 Also warn about violations of the following style guidelines from
2376 Scott Meyers' @cite{More Effective C++} book:
2380 Item 6: Distinguish between prefix and postfix forms of increment and
2381 decrement operators.
2384 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2388 When selecting this option, be aware that the standard library
2389 headers do not obey all of these guidelines; use @samp{grep -v}
2390 to filter out those warnings.
2392 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2393 @opindex Wstrict-null-sentinel
2394 @opindex Wno-strict-null-sentinel
2395 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2396 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2397 to @code{__null}. Although it is a null pointer constant not a null pointer,
2398 it is guaranteed to be of the same size as a pointer. But this use is
2399 not portable across different compilers.
2401 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2402 @opindex Wno-non-template-friend
2403 @opindex Wnon-template-friend
2404 Disable warnings when non-templatized friend functions are declared
2405 within a template. Since the advent of explicit template specification
2406 support in G++, if the name of the friend is an unqualified-id (i.e.,
2407 @samp{friend foo(int)}), the C++ language specification demands that the
2408 friend declare or define an ordinary, nontemplate function. (Section
2409 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2410 could be interpreted as a particular specialization of a templatized
2411 function. Because this non-conforming behavior is no longer the default
2412 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2413 check existing code for potential trouble spots and is on by default.
2414 This new compiler behavior can be turned off with
2415 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2416 but disables the helpful warning.
2418 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2419 @opindex Wold-style-cast
2420 @opindex Wno-old-style-cast
2421 Warn if an old-style (C-style) cast to a non-void type is used within
2422 a C++ program. The new-style casts (@samp{dynamic_cast},
2423 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2424 less vulnerable to unintended effects and much easier to search for.
2426 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2427 @opindex Woverloaded-virtual
2428 @opindex Wno-overloaded-virtual
2429 @cindex overloaded virtual function, warning
2430 @cindex warning for overloaded virtual function
2431 Warn when a function declaration hides virtual functions from a
2432 base class. For example, in:
2439 struct B: public A @{
2444 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2452 will fail to compile.
2454 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2455 @opindex Wno-pmf-conversions
2456 @opindex Wpmf-conversions
2457 Disable the diagnostic for converting a bound pointer to member function
2460 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2461 @opindex Wsign-promo
2462 @opindex Wno-sign-promo
2463 Warn when overload resolution chooses a promotion from unsigned or
2464 enumerated type to a signed type, over a conversion to an unsigned type of
2465 the same size. Previous versions of G++ would try to preserve
2466 unsignedness, but the standard mandates the current behavior.
2471 A& operator = (int);
2481 In this example, G++ will synthesize a default @samp{A& operator =
2482 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2485 @node Objective-C and Objective-C++ Dialect Options
2486 @section Options Controlling Objective-C and Objective-C++ Dialects
2488 @cindex compiler options, Objective-C and Objective-C++
2489 @cindex Objective-C and Objective-C++ options, command line
2490 @cindex options, Objective-C and Objective-C++
2491 (NOTE: This manual does not describe the Objective-C and Objective-C++
2492 languages themselves. See @xref{Standards,,Language Standards
2493 Supported by GCC}, for references.)
2495 This section describes the command-line options that are only meaningful
2496 for Objective-C and Objective-C++ programs, but you can also use most of
2497 the language-independent GNU compiler options.
2498 For example, you might compile a file @code{some_class.m} like this:
2501 gcc -g -fgnu-runtime -O -c some_class.m
2505 In this example, @option{-fgnu-runtime} is an option meant only for
2506 Objective-C and Objective-C++ programs; you can use the other options with
2507 any language supported by GCC@.
2509 Note that since Objective-C is an extension of the C language, Objective-C
2510 compilations may also use options specific to the C front-end (e.g.,
2511 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2512 C++-specific options (e.g., @option{-Wabi}).
2514 Here is a list of options that are @emph{only} for compiling Objective-C
2515 and Objective-C++ programs:
2518 @item -fconstant-string-class=@var{class-name}
2519 @opindex fconstant-string-class
2520 Use @var{class-name} as the name of the class to instantiate for each
2521 literal string specified with the syntax @code{@@"@dots{}"}. The default
2522 class name is @code{NXConstantString} if the GNU runtime is being used, and
2523 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2524 @option{-fconstant-cfstrings} option, if also present, will override the
2525 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2526 to be laid out as constant CoreFoundation strings.
2529 @opindex fgnu-runtime
2530 Generate object code compatible with the standard GNU Objective-C
2531 runtime. This is the default for most types of systems.
2533 @item -fnext-runtime
2534 @opindex fnext-runtime
2535 Generate output compatible with the NeXT runtime. This is the default
2536 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2537 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2540 @item -fno-nil-receivers
2541 @opindex fno-nil-receivers
2542 Assume that all Objective-C message dispatches (@code{[receiver
2543 message:arg]}) in this translation unit ensure that the receiver is
2544 not @code{nil}. This allows for more efficient entry points in the
2545 runtime to be used. Currently, this option is only available in
2546 conjunction with the NeXT runtime on Mac OS X 10.3 and later.
2548 @item -fobjc-call-cxx-cdtors
2549 @opindex fobjc-call-cxx-cdtors
2550 For each Objective-C class, check if any of its instance variables is a
2551 C++ object with a non-trivial default constructor. If so, synthesize a
2552 special @code{- (id) .cxx_construct} instance method that will run
2553 non-trivial default constructors on any such instance variables, in order,
2554 and then return @code{self}. Similarly, check if any instance variable
2555 is a C++ object with a non-trivial destructor, and if so, synthesize a
2556 special @code{- (void) .cxx_destruct} method that will run
2557 all such default destructors, in reverse order.
2559 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
2560 methods thusly generated will only operate on instance variables
2561 declared in the current Objective-C class, and not those inherited
2562 from superclasses. It is the responsibility of the Objective-C
2563 runtime to invoke all such methods in an object's inheritance
2564 hierarchy. The @code{- (id) .cxx_construct} methods will be invoked
2565 by the runtime immediately after a new object instance is allocated;
2566 the @code{- (void) .cxx_destruct} methods will be invoked immediately
2567 before the runtime deallocates an object instance.
2569 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2570 support for invoking the @code{- (id) .cxx_construct} and
2571 @code{- (void) .cxx_destruct} methods.
2573 @item -fobjc-direct-dispatch
2574 @opindex fobjc-direct-dispatch
2575 Allow fast jumps to the message dispatcher. On Darwin this is
2576 accomplished via the comm page.
2578 @item -fobjc-exceptions
2579 @opindex fobjc-exceptions
2580 Enable syntactic support for structured exception handling in
2581 Objective-C, similar to what is offered by C++ and Java. This option
2582 is required to use the Objective-C keywords @code{@@try},
2583 @code{@@throw}, @code{@@catch}, @code{@@finally} and
2584 @code{@@synchronized}. This option is available with both the GNU
2585 runtime and the NeXT runtime (but not available in conjunction with
2586 the NeXT runtime on Mac OS X 10.2 and earlier).
2590 Enable garbage collection (GC) in Objective-C and Objective-C++
2591 programs. This option is only available with the NeXT runtime; the
2592 GNU runtime has a different garbage collection implementation that
2593 does not require special compiler flags.
2595 @item -fobjc-std=objc1
2597 Conform to the language syntax of Objective-C 1.0, the language
2598 recognized by GCC 4.0. This only affects the Objective-C additions to
2599 the C/C++ language; it does not affect conformance to C/C++ standards,
2600 which is controlled by the separate C/C++ dialect option flags. When
2601 this option is used with the Objective-C or Objective-C++ compiler,
2602 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
2603 This is useful if you need to make sure that your Objective-C code can
2604 be compiled with older versions of GCC.
2606 @item -freplace-objc-classes
2607 @opindex freplace-objc-classes
2608 Emit a special marker instructing @command{ld(1)} not to statically link in
2609 the resulting object file, and allow @command{dyld(1)} to load it in at
2610 run time instead. This is used in conjunction with the Fix-and-Continue
2611 debugging mode, where the object file in question may be recompiled and
2612 dynamically reloaded in the course of program execution, without the need
2613 to restart the program itself. Currently, Fix-and-Continue functionality
2614 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2619 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2620 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2621 compile time) with static class references that get initialized at load time,
2622 which improves run-time performance. Specifying the @option{-fzero-link} flag
2623 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2624 to be retained. This is useful in Zero-Link debugging mode, since it allows
2625 for individual class implementations to be modified during program execution.
2626 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
2627 regardless of command line options.
2631 Dump interface declarations for all classes seen in the source file to a
2632 file named @file{@var{sourcename}.decl}.
2634 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2635 @opindex Wassign-intercept
2636 @opindex Wno-assign-intercept
2637 Warn whenever an Objective-C assignment is being intercepted by the
2640 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2641 @opindex Wno-protocol
2643 If a class is declared to implement a protocol, a warning is issued for
2644 every method in the protocol that is not implemented by the class. The
2645 default behavior is to issue a warning for every method not explicitly
2646 implemented in the class, even if a method implementation is inherited
2647 from the superclass. If you use the @option{-Wno-protocol} option, then
2648 methods inherited from the superclass are considered to be implemented,
2649 and no warning is issued for them.
2651 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2653 @opindex Wno-selector
2654 Warn if multiple methods of different types for the same selector are
2655 found during compilation. The check is performed on the list of methods
2656 in the final stage of compilation. Additionally, a check is performed
2657 for each selector appearing in a @code{@@selector(@dots{})}
2658 expression, and a corresponding method for that selector has been found
2659 during compilation. Because these checks scan the method table only at
2660 the end of compilation, these warnings are not produced if the final
2661 stage of compilation is not reached, for example because an error is
2662 found during compilation, or because the @option{-fsyntax-only} option is
2665 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2666 @opindex Wstrict-selector-match
2667 @opindex Wno-strict-selector-match
2668 Warn if multiple methods with differing argument and/or return types are
2669 found for a given selector when attempting to send a message using this
2670 selector to a receiver of type @code{id} or @code{Class}. When this flag
2671 is off (which is the default behavior), the compiler will omit such warnings
2672 if any differences found are confined to types which share the same size
2675 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2676 @opindex Wundeclared-selector
2677 @opindex Wno-undeclared-selector
2678 Warn if a @code{@@selector(@dots{})} expression referring to an
2679 undeclared selector is found. A selector is considered undeclared if no
2680 method with that name has been declared before the
2681 @code{@@selector(@dots{})} expression, either explicitly in an
2682 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2683 an @code{@@implementation} section. This option always performs its
2684 checks as soon as a @code{@@selector(@dots{})} expression is found,
2685 while @option{-Wselector} only performs its checks in the final stage of
2686 compilation. This also enforces the coding style convention
2687 that methods and selectors must be declared before being used.
2689 @item -print-objc-runtime-info
2690 @opindex print-objc-runtime-info
2691 Generate C header describing the largest structure that is passed by
2696 @node Language Independent Options
2697 @section Options to Control Diagnostic Messages Formatting
2698 @cindex options to control diagnostics formatting
2699 @cindex diagnostic messages
2700 @cindex message formatting
2702 Traditionally, diagnostic messages have been formatted irrespective of
2703 the output device's aspect (e.g.@: its width, @dots{}). The options described
2704 below can be used to control the diagnostic messages formatting
2705 algorithm, e.g.@: how many characters per line, how often source location
2706 information should be reported. Right now, only the C++ front end can
2707 honor these options. However it is expected, in the near future, that
2708 the remaining front ends would be able to digest them correctly.
2711 @item -fmessage-length=@var{n}
2712 @opindex fmessage-length
2713 Try to format error messages so that they fit on lines of about @var{n}
2714 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2715 the front ends supported by GCC@. If @var{n} is zero, then no
2716 line-wrapping will be done; each error message will appear on a single
2719 @opindex fdiagnostics-show-location
2720 @item -fdiagnostics-show-location=once
2721 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2722 reporter to emit @emph{once} source location information; that is, in
2723 case the message is too long to fit on a single physical line and has to
2724 be wrapped, the source location won't be emitted (as prefix) again,
2725 over and over, in subsequent continuation lines. This is the default
2728 @item -fdiagnostics-show-location=every-line
2729 Only meaningful in line-wrapping mode. Instructs the diagnostic
2730 messages reporter to emit the same source location information (as
2731 prefix) for physical lines that result from the process of breaking
2732 a message which is too long to fit on a single line.
2734 @item -fdiagnostics-show-option
2735 @opindex fdiagnostics-show-option
2736 This option instructs the diagnostic machinery to add text to each
2737 diagnostic emitted, which indicates which command line option directly
2738 controls that diagnostic, when such an option is known to the
2739 diagnostic machinery.
2741 @item -Wcoverage-mismatch
2742 @opindex Wcoverage-mismatch
2743 Warn if feedback profiles do not match when using the
2744 @option{-fprofile-use} option.
2745 If a source file was changed between @option{-fprofile-gen} and
2746 @option{-fprofile-use}, the files with the profile feedback can fail
2747 to match the source file and GCC can not use the profile feedback
2748 information. By default, this warning is enabled and is treated as an
2749 error. @option{-Wno-coverage-mismatch} can be used to disable the
2750 warning or @option{-Wno-error=coverage-mismatch} can be used to
2751 disable the error. Disable the error for this warning can result in
2752 poorly optimized code, so disabling the error is useful only in the
2753 case of very minor changes such as bug fixes to an existing code-base.
2754 Completely disabling the warning is not recommended.
2758 @node Warning Options
2759 @section Options to Request or Suppress Warnings
2760 @cindex options to control warnings
2761 @cindex warning messages
2762 @cindex messages, warning
2763 @cindex suppressing warnings
2765 Warnings are diagnostic messages that report constructions which
2766 are not inherently erroneous but which are risky or suggest there
2767 may have been an error.
2769 The following language-independent options do not enable specific
2770 warnings but control the kinds of diagnostics produced by GCC.
2773 @cindex syntax checking
2775 @opindex fsyntax-only
2776 Check the code for syntax errors, but don't do anything beyond that.
2780 Inhibit all warning messages.
2785 Make all warnings into errors.
2790 Make the specified warning into an error. The specifier for a warning
2791 is appended, for example @option{-Werror=switch} turns the warnings
2792 controlled by @option{-Wswitch} into errors. This switch takes a
2793 negative form, to be used to negate @option{-Werror} for specific
2794 warnings, for example @option{-Wno-error=switch} makes
2795 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2796 is in effect. You can use the @option{-fdiagnostics-show-option}
2797 option to have each controllable warning amended with the option which
2798 controls it, to determine what to use with this option.
2800 Note that specifying @option{-Werror=}@var{foo} automatically implies
2801 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2804 @item -Wfatal-errors
2805 @opindex Wfatal-errors
2806 @opindex Wno-fatal-errors
2807 This option causes the compiler to abort compilation on the first error
2808 occurred rather than trying to keep going and printing further error
2813 You can request many specific warnings with options beginning
2814 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2815 implicit declarations. Each of these specific warning options also
2816 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2817 example, @option{-Wno-implicit}. This manual lists only one of the
2818 two forms, whichever is not the default. For further,
2819 language-specific options also refer to @ref{C++ Dialect Options} and
2820 @ref{Objective-C and Objective-C++ Dialect Options}.
2822 When an unrecognized warning option is requested (e.g.,
2823 @option{-Wunknown-warning}), GCC will emit a diagnostic stating
2824 that the option is not recognized. However, if the @option{-Wno-} form
2825 is used, the behavior is slightly different: No diagnostic will be
2826 produced for @option{-Wno-unknown-warning} unless other diagnostics
2827 are being produced. This allows the use of new @option{-Wno-} options
2828 with old compilers, but if something goes wrong, the compiler will
2829 warn that an unrecognized option was used.
2834 Issue all the warnings demanded by strict ISO C and ISO C++;
2835 reject all programs that use forbidden extensions, and some other
2836 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2837 version of the ISO C standard specified by any @option{-std} option used.
2839 Valid ISO C and ISO C++ programs should compile properly with or without
2840 this option (though a rare few will require @option{-ansi} or a
2841 @option{-std} option specifying the required version of ISO C)@. However,
2842 without this option, certain GNU extensions and traditional C and C++
2843 features are supported as well. With this option, they are rejected.
2845 @option{-pedantic} does not cause warning messages for use of the
2846 alternate keywords whose names begin and end with @samp{__}. Pedantic
2847 warnings are also disabled in the expression that follows
2848 @code{__extension__}. However, only system header files should use
2849 these escape routes; application programs should avoid them.
2850 @xref{Alternate Keywords}.
2852 Some users try to use @option{-pedantic} to check programs for strict ISO
2853 C conformance. They soon find that it does not do quite what they want:
2854 it finds some non-ISO practices, but not all---only those for which
2855 ISO C @emph{requires} a diagnostic, and some others for which
2856 diagnostics have been added.
2858 A feature to report any failure to conform to ISO C might be useful in
2859 some instances, but would require considerable additional work and would
2860 be quite different from @option{-pedantic}. We don't have plans to
2861 support such a feature in the near future.
2863 Where the standard specified with @option{-std} represents a GNU
2864 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
2865 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2866 extended dialect is based. Warnings from @option{-pedantic} are given
2867 where they are required by the base standard. (It would not make sense
2868 for such warnings to be given only for features not in the specified GNU
2869 C dialect, since by definition the GNU dialects of C include all
2870 features the compiler supports with the given option, and there would be
2871 nothing to warn about.)
2873 @item -pedantic-errors
2874 @opindex pedantic-errors
2875 Like @option{-pedantic}, except that errors are produced rather than
2881 This enables all the warnings about constructions that some users
2882 consider questionable, and that are easy to avoid (or modify to
2883 prevent the warning), even in conjunction with macros. This also
2884 enables some language-specific warnings described in @ref{C++ Dialect
2885 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2887 @option{-Wall} turns on the following warning flags:
2889 @gccoptlist{-Waddress @gol
2890 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2892 -Wchar-subscripts @gol
2893 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
2894 -Wimplicit-int @r{(C and Objective-C only)} @gol
2895 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
2898 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2899 -Wmissing-braces @gol
2905 -Wsequence-point @gol
2906 -Wsign-compare @r{(only in C++)} @gol
2907 -Wstrict-aliasing @gol
2908 -Wstrict-overflow=1 @gol
2911 -Wuninitialized @gol
2912 -Wunknown-pragmas @gol
2913 -Wunused-function @gol
2916 -Wunused-variable @gol
2917 -Wvolatile-register-var @gol
2920 Note that some warning flags are not implied by @option{-Wall}. Some of
2921 them warn about constructions that users generally do not consider
2922 questionable, but which occasionally you might wish to check for;
2923 others warn about constructions that are necessary or hard to avoid in
2924 some cases, and there is no simple way to modify the code to suppress
2925 the warning. Some of them are enabled by @option{-Wextra} but many of
2926 them must be enabled individually.
2932 This enables some extra warning flags that are not enabled by
2933 @option{-Wall}. (This option used to be called @option{-W}. The older
2934 name is still supported, but the newer name is more descriptive.)
2936 @gccoptlist{-Wclobbered @gol
2938 -Wignored-qualifiers @gol
2939 -Wmissing-field-initializers @gol
2940 -Wmissing-parameter-type @r{(C only)} @gol
2941 -Wold-style-declaration @r{(C only)} @gol
2942 -Woverride-init @gol
2945 -Wuninitialized @gol
2946 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2947 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2950 The option @option{-Wextra} also prints warning messages for the
2956 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2957 @samp{>}, or @samp{>=}.
2960 (C++ only) An enumerator and a non-enumerator both appear in a
2961 conditional expression.
2964 (C++ only) Ambiguous virtual bases.
2967 (C++ only) Subscripting an array which has been declared @samp{register}.
2970 (C++ only) Taking the address of a variable which has been declared
2974 (C++ only) A base class is not initialized in a derived class' copy
2979 @item -Wchar-subscripts
2980 @opindex Wchar-subscripts
2981 @opindex Wno-char-subscripts
2982 Warn if an array subscript has type @code{char}. This is a common cause
2983 of error, as programmers often forget that this type is signed on some
2985 This warning is enabled by @option{-Wall}.
2989 @opindex Wno-comment
2990 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2991 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2992 This warning is enabled by @option{-Wall}.
2995 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
2997 Suppress warning messages emitted by @code{#warning} directives.
2999 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3000 @opindex Wdouble-promotion
3001 @opindex Wno-double-promotion
3002 Give a warning when a value of type @code{float} is implicitly
3003 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3004 floating-point unit implement @code{float} in hardware, but emulate
3005 @code{double} in software. On such a machine, doing computations
3006 using @code{double} values is much more expensive because of the
3007 overhead required for software emulation.
3009 It is easy to accidentally do computations with @code{double} because
3010 floating-point literals are implicitly of type @code{double}. For
3014 float area(float radius)
3016 return 3.14159 * radius * radius;
3020 the compiler will perform the entire computation with @code{double}
3021 because the floating-point literal is a @code{double}.
3026 @opindex ffreestanding
3027 @opindex fno-builtin
3028 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3029 the arguments supplied have types appropriate to the format string
3030 specified, and that the conversions specified in the format string make
3031 sense. This includes standard functions, and others specified by format
3032 attributes (@pxref{Function Attributes}), in the @code{printf},
3033 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3034 not in the C standard) families (or other target-specific families).
3035 Which functions are checked without format attributes having been
3036 specified depends on the standard version selected, and such checks of
3037 functions without the attribute specified are disabled by
3038 @option{-ffreestanding} or @option{-fno-builtin}.
3040 The formats are checked against the format features supported by GNU
3041 libc version 2.2. These include all ISO C90 and C99 features, as well
3042 as features from the Single Unix Specification and some BSD and GNU
3043 extensions. Other library implementations may not support all these
3044 features; GCC does not support warning about features that go beyond a
3045 particular library's limitations. However, if @option{-pedantic} is used
3046 with @option{-Wformat}, warnings will be given about format features not
3047 in the selected standard version (but not for @code{strfmon} formats,
3048 since those are not in any version of the C standard). @xref{C Dialect
3049 Options,,Options Controlling C Dialect}.
3051 Since @option{-Wformat} also checks for null format arguments for
3052 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
3054 @option{-Wformat} is included in @option{-Wall}. For more control over some
3055 aspects of format checking, the options @option{-Wformat-y2k},
3056 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
3057 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
3058 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
3061 @opindex Wformat-y2k
3062 @opindex Wno-format-y2k
3063 If @option{-Wformat} is specified, also warn about @code{strftime}
3064 formats which may yield only a two-digit year.
3066 @item -Wno-format-contains-nul
3067 @opindex Wno-format-contains-nul
3068 @opindex Wformat-contains-nul
3069 If @option{-Wformat} is specified, do not warn about format strings that
3072 @item -Wno-format-extra-args
3073 @opindex Wno-format-extra-args
3074 @opindex Wformat-extra-args
3075 If @option{-Wformat} is specified, do not warn about excess arguments to a
3076 @code{printf} or @code{scanf} format function. The C standard specifies
3077 that such arguments are ignored.
3079 Where the unused arguments lie between used arguments that are
3080 specified with @samp{$} operand number specifications, normally
3081 warnings are still given, since the implementation could not know what
3082 type to pass to @code{va_arg} to skip the unused arguments. However,
3083 in the case of @code{scanf} formats, this option will suppress the
3084 warning if the unused arguments are all pointers, since the Single
3085 Unix Specification says that such unused arguments are allowed.
3087 @item -Wno-format-zero-length @r{(C and Objective-C only)}
3088 @opindex Wno-format-zero-length
3089 @opindex Wformat-zero-length
3090 If @option{-Wformat} is specified, do not warn about zero-length formats.
3091 The C standard specifies that zero-length formats are allowed.
3093 @item -Wformat-nonliteral
3094 @opindex Wformat-nonliteral
3095 @opindex Wno-format-nonliteral
3096 If @option{-Wformat} is specified, also warn if the format string is not a
3097 string literal and so cannot be checked, unless the format function
3098 takes its format arguments as a @code{va_list}.
3100 @item -Wformat-security
3101 @opindex Wformat-security
3102 @opindex Wno-format-security
3103 If @option{-Wformat} is specified, also warn about uses of format
3104 functions that represent possible security problems. At present, this
3105 warns about calls to @code{printf} and @code{scanf} functions where the
3106 format string is not a string literal and there are no format arguments,
3107 as in @code{printf (foo);}. This may be a security hole if the format
3108 string came from untrusted input and contains @samp{%n}. (This is
3109 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3110 in future warnings may be added to @option{-Wformat-security} that are not
3111 included in @option{-Wformat-nonliteral}.)
3115 @opindex Wno-format=2
3116 Enable @option{-Wformat} plus format checks not included in
3117 @option{-Wformat}. Currently equivalent to @samp{-Wformat
3118 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
3120 @item -Wnonnull @r{(C and Objective-C only)}
3122 @opindex Wno-nonnull
3123 Warn about passing a null pointer for arguments marked as
3124 requiring a non-null value by the @code{nonnull} function attribute.
3126 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3127 can be disabled with the @option{-Wno-nonnull} option.
3129 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3131 @opindex Wno-init-self
3132 Warn about uninitialized variables which are initialized with themselves.
3133 Note this option can only be used with the @option{-Wuninitialized} option.
3135 For example, GCC will warn about @code{i} being uninitialized in the
3136 following snippet only when @option{-Winit-self} has been specified:
3147 @item -Wimplicit-int @r{(C and Objective-C only)}
3148 @opindex Wimplicit-int
3149 @opindex Wno-implicit-int
3150 Warn when a declaration does not specify a type.
3151 This warning is enabled by @option{-Wall}.
3153 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3154 @opindex Wimplicit-function-declaration
3155 @opindex Wno-implicit-function-declaration
3156 Give a warning whenever a function is used before being declared. In
3157 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3158 enabled by default and it is made into an error by
3159 @option{-pedantic-errors}. This warning is also enabled by
3162 @item -Wimplicit @r{(C and Objective-C only)}
3164 @opindex Wno-implicit
3165 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3166 This warning is enabled by @option{-Wall}.
3168 @item -Wignored-qualifiers @r{(C and C++ only)}
3169 @opindex Wignored-qualifiers
3170 @opindex Wno-ignored-qualifiers
3171 Warn if the return type of a function has a type qualifier
3172 such as @code{const}. For ISO C such a type qualifier has no effect,
3173 since the value returned by a function is not an lvalue.
3174 For C++, the warning is only emitted for scalar types or @code{void}.
3175 ISO C prohibits qualified @code{void} return types on function
3176 definitions, so such return types always receive a warning
3177 even without this option.
3179 This warning is also enabled by @option{-Wextra}.
3184 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3185 a function with external linkage, returning int, taking either zero
3186 arguments, two, or three arguments of appropriate types. This warning
3187 is enabled by default in C++ and is enabled by either @option{-Wall}
3188 or @option{-pedantic}.
3190 @item -Wmissing-braces
3191 @opindex Wmissing-braces
3192 @opindex Wno-missing-braces
3193 Warn if an aggregate or union initializer is not fully bracketed. In
3194 the following example, the initializer for @samp{a} is not fully
3195 bracketed, but that for @samp{b} is fully bracketed.
3198 int a[2][2] = @{ 0, 1, 2, 3 @};
3199 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3202 This warning is enabled by @option{-Wall}.
3204 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3205 @opindex Wmissing-include-dirs
3206 @opindex Wno-missing-include-dirs
3207 Warn if a user-supplied include directory does not exist.
3210 @opindex Wparentheses
3211 @opindex Wno-parentheses
3212 Warn if parentheses are omitted in certain contexts, such
3213 as when there is an assignment in a context where a truth value
3214 is expected, or when operators are nested whose precedence people
3215 often get confused about.
3217 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3218 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3219 interpretation from that of ordinary mathematical notation.
3221 Also warn about constructions where there may be confusion to which
3222 @code{if} statement an @code{else} branch belongs. Here is an example of
3237 In C/C++, every @code{else} branch belongs to the innermost possible
3238 @code{if} statement, which in this example is @code{if (b)}. This is
3239 often not what the programmer expected, as illustrated in the above
3240 example by indentation the programmer chose. When there is the
3241 potential for this confusion, GCC will issue a warning when this flag
3242 is specified. To eliminate the warning, add explicit braces around
3243 the innermost @code{if} statement so there is no way the @code{else}
3244 could belong to the enclosing @code{if}. The resulting code would
3261 Also warn for dangerous uses of the
3262 ?: with omitted middle operand GNU extension. When the condition
3263 in the ?: operator is a boolean expression the omitted value will
3264 be always 1. Often the user expects it to be a value computed
3265 inside the conditional expression instead.
3267 This warning is enabled by @option{-Wall}.
3269 @item -Wsequence-point
3270 @opindex Wsequence-point
3271 @opindex Wno-sequence-point
3272 Warn about code that may have undefined semantics because of violations
3273 of sequence point rules in the C and C++ standards.
3275 The C and C++ standards defines the order in which expressions in a C/C++
3276 program are evaluated in terms of @dfn{sequence points}, which represent
3277 a partial ordering between the execution of parts of the program: those
3278 executed before the sequence point, and those executed after it. These
3279 occur after the evaluation of a full expression (one which is not part
3280 of a larger expression), after the evaluation of the first operand of a
3281 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3282 function is called (but after the evaluation of its arguments and the
3283 expression denoting the called function), and in certain other places.
3284 Other than as expressed by the sequence point rules, the order of
3285 evaluation of subexpressions of an expression is not specified. All
3286 these rules describe only a partial order rather than a total order,
3287 since, for example, if two functions are called within one expression
3288 with no sequence point between them, the order in which the functions
3289 are called is not specified. However, the standards committee have
3290 ruled that function calls do not overlap.
3292 It is not specified when between sequence points modifications to the
3293 values of objects take effect. Programs whose behavior depends on this
3294 have undefined behavior; the C and C++ standards specify that ``Between
3295 the previous and next sequence point an object shall have its stored
3296 value modified at most once by the evaluation of an expression.
3297 Furthermore, the prior value shall be read only to determine the value
3298 to be stored.''. If a program breaks these rules, the results on any
3299 particular implementation are entirely unpredictable.
3301 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3302 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3303 diagnosed by this option, and it may give an occasional false positive
3304 result, but in general it has been found fairly effective at detecting
3305 this sort of problem in programs.
3307 The standard is worded confusingly, therefore there is some debate
3308 over the precise meaning of the sequence point rules in subtle cases.
3309 Links to discussions of the problem, including proposed formal
3310 definitions, may be found on the GCC readings page, at
3311 @uref{http://gcc.gnu.org/@/readings.html}.
3313 This warning is enabled by @option{-Wall} for C and C++.
3316 @opindex Wreturn-type
3317 @opindex Wno-return-type
3318 Warn whenever a function is defined with a return-type that defaults
3319 to @code{int}. Also warn about any @code{return} statement with no
3320 return-value in a function whose return-type is not @code{void}
3321 (falling off the end of the function body is considered returning
3322 without a value), and about a @code{return} statement with an
3323 expression in a function whose return-type is @code{void}.
3325 For C++, a function without return type always produces a diagnostic
3326 message, even when @option{-Wno-return-type} is specified. The only
3327 exceptions are @samp{main} and functions defined in system headers.
3329 This warning is enabled by @option{-Wall}.
3334 Warn whenever a @code{switch} statement has an index of enumerated type
3335 and lacks a @code{case} for one or more of the named codes of that
3336 enumeration. (The presence of a @code{default} label prevents this
3337 warning.) @code{case} labels outside the enumeration range also
3338 provoke warnings when this option is used (even if there is a
3339 @code{default} label).
3340 This warning is enabled by @option{-Wall}.
3342 @item -Wswitch-default
3343 @opindex Wswitch-default
3344 @opindex Wno-switch-default
3345 Warn whenever a @code{switch} statement does not have a @code{default}
3349 @opindex Wswitch-enum
3350 @opindex Wno-switch-enum
3351 Warn whenever a @code{switch} statement has an index of enumerated type
3352 and lacks a @code{case} for one or more of the named codes of that
3353 enumeration. @code{case} labels outside the enumeration range also
3354 provoke warnings when this option is used. The only difference
3355 between @option{-Wswitch} and this option is that this option gives a
3356 warning about an omitted enumeration code even if there is a
3357 @code{default} label.
3359 @item -Wsync-nand @r{(C and C++ only)}
3361 @opindex Wno-sync-nand
3362 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3363 built-in functions are used. These functions changed semantics in GCC 4.4.
3367 @opindex Wno-trigraphs
3368 Warn if any trigraphs are encountered that might change the meaning of
3369 the program (trigraphs within comments are not warned about).
3370 This warning is enabled by @option{-Wall}.
3372 @item -Wunused-but-set-parameter
3373 @opindex Wunused-but-set-parameter
3374 @opindex Wno-unused-but-set-parameter
3375 Warn whenever a function parameter is assigned to, but otherwise unused
3376 (aside from its declaration).
3378 To suppress this warning use the @samp{unused} attribute
3379 (@pxref{Variable Attributes}).
3381 This warning is also enabled by @option{-Wunused} together with
3384 @item -Wunused-but-set-variable
3385 @opindex Wunused-but-set-variable
3386 @opindex Wno-unused-but-set-variable
3387 Warn whenever a local variable is assigned to, but otherwise unused
3388 (aside from its declaration).
3389 This warning is enabled by @option{-Wall}.
3391 To suppress this warning use the @samp{unused} attribute
3392 (@pxref{Variable Attributes}).
3394 This warning is also enabled by @option{-Wunused}, which is enabled
3397 @item -Wunused-function
3398 @opindex Wunused-function
3399 @opindex Wno-unused-function
3400 Warn whenever a static function is declared but not defined or a
3401 non-inline static function is unused.
3402 This warning is enabled by @option{-Wall}.
3404 @item -Wunused-label
3405 @opindex Wunused-label
3406 @opindex Wno-unused-label
3407 Warn whenever a label is declared but not used.
3408 This warning is enabled by @option{-Wall}.
3410 To suppress this warning use the @samp{unused} attribute
3411 (@pxref{Variable Attributes}).
3413 @item -Wunused-parameter
3414 @opindex Wunused-parameter
3415 @opindex Wno-unused-parameter
3416 Warn whenever a function parameter is unused aside from its declaration.
3418 To suppress this warning use the @samp{unused} attribute
3419 (@pxref{Variable Attributes}).
3421 @item -Wno-unused-result
3422 @opindex Wunused-result
3423 @opindex Wno-unused-result
3424 Do not warn if a caller of a function marked with attribute
3425 @code{warn_unused_result} (@pxref{Variable Attributes}) does not use
3426 its return value. The default is @option{-Wunused-result}.
3428 @item -Wunused-variable
3429 @opindex Wunused-variable
3430 @opindex Wno-unused-variable
3431 Warn whenever a local variable or non-constant static variable is unused
3432 aside from its declaration.
3433 This warning is enabled by @option{-Wall}.
3435 To suppress this warning use the @samp{unused} attribute
3436 (@pxref{Variable Attributes}).
3438 @item -Wunused-value
3439 @opindex Wunused-value
3440 @opindex Wno-unused-value
3441 Warn whenever a statement computes a result that is explicitly not
3442 used. To suppress this warning cast the unused expression to
3443 @samp{void}. This includes an expression-statement or the left-hand
3444 side of a comma expression that contains no side effects. For example,
3445 an expression such as @samp{x[i,j]} will cause a warning, while
3446 @samp{x[(void)i,j]} will not.
3448 This warning is enabled by @option{-Wall}.
3453 All the above @option{-Wunused} options combined.
3455 In order to get a warning about an unused function parameter, you must
3456 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3457 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3459 @item -Wuninitialized
3460 @opindex Wuninitialized
3461 @opindex Wno-uninitialized
3462 Warn if an automatic variable is used without first being initialized
3463 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3464 warn if a non-static reference or non-static @samp{const} member
3465 appears in a class without constructors.
3467 If you want to warn about code which uses the uninitialized value of the
3468 variable in its own initializer, use the @option{-Winit-self} option.
3470 These warnings occur for individual uninitialized or clobbered
3471 elements of structure, union or array variables as well as for
3472 variables which are uninitialized or clobbered as a whole. They do
3473 not occur for variables or elements declared @code{volatile}. Because
3474 these warnings depend on optimization, the exact variables or elements
3475 for which there are warnings will depend on the precise optimization
3476 options and version of GCC used.
3478 Note that there may be no warning about a variable that is used only
3479 to compute a value that itself is never used, because such
3480 computations may be deleted by data flow analysis before the warnings
3483 These warnings are made optional because GCC is not smart
3484 enough to see all the reasons why the code might be correct
3485 despite appearing to have an error. Here is one example of how
3506 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3507 always initialized, but GCC doesn't know this. Here is
3508 another common case:
3513 if (change_y) save_y = y, y = new_y;
3515 if (change_y) y = save_y;
3520 This has no bug because @code{save_y} is used only if it is set.
3522 @cindex @code{longjmp} warnings
3523 This option also warns when a non-volatile automatic variable might be
3524 changed by a call to @code{longjmp}. These warnings as well are possible
3525 only in optimizing compilation.
3527 The compiler sees only the calls to @code{setjmp}. It cannot know
3528 where @code{longjmp} will be called; in fact, a signal handler could
3529 call it at any point in the code. As a result, you may get a warning
3530 even when there is in fact no problem because @code{longjmp} cannot
3531 in fact be called at the place which would cause a problem.
3533 Some spurious warnings can be avoided if you declare all the functions
3534 you use that never return as @code{noreturn}. @xref{Function
3537 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3539 @item -Wunknown-pragmas
3540 @opindex Wunknown-pragmas
3541 @opindex Wno-unknown-pragmas
3542 @cindex warning for unknown pragmas
3543 @cindex unknown pragmas, warning
3544 @cindex pragmas, warning of unknown
3545 Warn when a #pragma directive is encountered which is not understood by
3546 GCC@. If this command line option is used, warnings will even be issued
3547 for unknown pragmas in system header files. This is not the case if
3548 the warnings were only enabled by the @option{-Wall} command line option.
3551 @opindex Wno-pragmas
3553 Do not warn about misuses of pragmas, such as incorrect parameters,
3554 invalid syntax, or conflicts between pragmas. See also
3555 @samp{-Wunknown-pragmas}.
3557 @item -Wstrict-aliasing
3558 @opindex Wstrict-aliasing
3559 @opindex Wno-strict-aliasing
3560 This option is only active when @option{-fstrict-aliasing} is active.
3561 It warns about code which might break the strict aliasing rules that the
3562 compiler is using for optimization. The warning does not catch all
3563 cases, but does attempt to catch the more common pitfalls. It is
3564 included in @option{-Wall}.
3565 It is equivalent to @option{-Wstrict-aliasing=3}
3567 @item -Wstrict-aliasing=n
3568 @opindex Wstrict-aliasing=n
3569 @opindex Wno-strict-aliasing=n
3570 This option is only active when @option{-fstrict-aliasing} is active.
3571 It warns about code which might break the strict aliasing rules that the
3572 compiler is using for optimization.
3573 Higher levels correspond to higher accuracy (fewer false positives).
3574 Higher levels also correspond to more effort, similar to the way -O works.
3575 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3578 Level 1: Most aggressive, quick, least accurate.
3579 Possibly useful when higher levels
3580 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3581 false negatives. However, it has many false positives.
3582 Warns for all pointer conversions between possibly incompatible types,
3583 even if never dereferenced. Runs in the frontend only.
3585 Level 2: Aggressive, quick, not too precise.
3586 May still have many false positives (not as many as level 1 though),
3587 and few false negatives (but possibly more than level 1).
3588 Unlike level 1, it only warns when an address is taken. Warns about
3589 incomplete types. Runs in the frontend only.
3591 Level 3 (default for @option{-Wstrict-aliasing}):
3592 Should have very few false positives and few false
3593 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3594 Takes care of the common pun+dereference pattern in the frontend:
3595 @code{*(int*)&some_float}.
3596 If optimization is enabled, it also runs in the backend, where it deals
3597 with multiple statement cases using flow-sensitive points-to information.
3598 Only warns when the converted pointer is dereferenced.
3599 Does not warn about incomplete types.
3601 @item -Wstrict-overflow
3602 @itemx -Wstrict-overflow=@var{n}
3603 @opindex Wstrict-overflow
3604 @opindex Wno-strict-overflow
3605 This option is only active when @option{-fstrict-overflow} is active.
3606 It warns about cases where the compiler optimizes based on the
3607 assumption that signed overflow does not occur. Note that it does not
3608 warn about all cases where the code might overflow: it only warns
3609 about cases where the compiler implements some optimization. Thus
3610 this warning depends on the optimization level.
3612 An optimization which assumes that signed overflow does not occur is
3613 perfectly safe if the values of the variables involved are such that
3614 overflow never does, in fact, occur. Therefore this warning can
3615 easily give a false positive: a warning about code which is not
3616 actually a problem. To help focus on important issues, several
3617 warning levels are defined. No warnings are issued for the use of
3618 undefined signed overflow when estimating how many iterations a loop
3619 will require, in particular when determining whether a loop will be
3623 @item -Wstrict-overflow=1
3624 Warn about cases which are both questionable and easy to avoid. For
3625 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3626 compiler will simplify this to @code{1}. This level of
3627 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3628 are not, and must be explicitly requested.
3630 @item -Wstrict-overflow=2
3631 Also warn about other cases where a comparison is simplified to a
3632 constant. For example: @code{abs (x) >= 0}. This can only be
3633 simplified when @option{-fstrict-overflow} is in effect, because
3634 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3635 zero. @option{-Wstrict-overflow} (with no level) is the same as
3636 @option{-Wstrict-overflow=2}.
3638 @item -Wstrict-overflow=3
3639 Also warn about other cases where a comparison is simplified. For
3640 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3642 @item -Wstrict-overflow=4
3643 Also warn about other simplifications not covered by the above cases.
3644 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3646 @item -Wstrict-overflow=5
3647 Also warn about cases where the compiler reduces the magnitude of a
3648 constant involved in a comparison. For example: @code{x + 2 > y} will
3649 be simplified to @code{x + 1 >= y}. This is reported only at the
3650 highest warning level because this simplification applies to many
3651 comparisons, so this warning level will give a very large number of
3655 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]}
3656 @opindex Wsuggest-attribute=
3657 @opindex Wno-suggest-attribute=
3658 Warn for cases where adding an attribute may be beneficial. The
3659 attributes currently supported are listed below.
3662 @item -Wsuggest-attribute=pure
3663 @itemx -Wsuggest-attribute=const
3664 @itemx -Wsuggest-attribute=noreturn
3665 @opindex Wsuggest-attribute=pure
3666 @opindex Wno-suggest-attribute=pure
3667 @opindex Wsuggest-attribute=const
3668 @opindex Wno-suggest-attribute=const
3669 @opindex Wsuggest-attribute=noreturn
3670 @opindex Wno-suggest-attribute=noreturn
3672 Warn about functions which might be candidates for attributes
3673 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
3674 functions visible in other compilation units or (in the case of @code{pure} and
3675 @code{const}) if it cannot prove that the function returns normally. A function
3676 returns normally if it doesn't contain an infinite loop nor returns abnormally
3677 by throwing, calling @code{abort()} or trapping. This analysis requires option
3678 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
3679 higher. Higher optimization levels improve the accuracy of the analysis.
3682 @item -Warray-bounds
3683 @opindex Wno-array-bounds
3684 @opindex Warray-bounds
3685 This option is only active when @option{-ftree-vrp} is active
3686 (default for @option{-O2} and above). It warns about subscripts to arrays
3687 that are always out of bounds. This warning is enabled by @option{-Wall}.
3689 @item -Wno-div-by-zero
3690 @opindex Wno-div-by-zero
3691 @opindex Wdiv-by-zero
3692 Do not warn about compile-time integer division by zero. Floating point
3693 division by zero is not warned about, as it can be a legitimate way of
3694 obtaining infinities and NaNs.
3696 @item -Wsystem-headers
3697 @opindex Wsystem-headers
3698 @opindex Wno-system-headers
3699 @cindex warnings from system headers
3700 @cindex system headers, warnings from
3701 Print warning messages for constructs found in system header files.
3702 Warnings from system headers are normally suppressed, on the assumption
3703 that they usually do not indicate real problems and would only make the
3704 compiler output harder to read. Using this command line option tells
3705 GCC to emit warnings from system headers as if they occurred in user
3706 code. However, note that using @option{-Wall} in conjunction with this
3707 option will @emph{not} warn about unknown pragmas in system
3708 headers---for that, @option{-Wunknown-pragmas} must also be used.
3711 @opindex Wtrampolines
3712 @opindex Wno-trampolines
3713 Warn about trampolines generated for pointers to nested functions.
3715 A trampoline is a small piece of data or code that is created at run
3716 time on the stack when the address of a nested function is taken, and
3717 is used to call the nested function indirectly. For some targets, it
3718 is made up of data only and thus requires no special treatment. But,
3719 for most targets, it is made up of code and thus requires the stack
3720 to be made executable in order for the program to work properly.
3723 @opindex Wfloat-equal
3724 @opindex Wno-float-equal
3725 Warn if floating point values are used in equality comparisons.
3727 The idea behind this is that sometimes it is convenient (for the
3728 programmer) to consider floating-point values as approximations to
3729 infinitely precise real numbers. If you are doing this, then you need
3730 to compute (by analyzing the code, or in some other way) the maximum or
3731 likely maximum error that the computation introduces, and allow for it
3732 when performing comparisons (and when producing output, but that's a
3733 different problem). In particular, instead of testing for equality, you
3734 would check to see whether the two values have ranges that overlap; and
3735 this is done with the relational operators, so equality comparisons are
3738 @item -Wtraditional @r{(C and Objective-C only)}
3739 @opindex Wtraditional
3740 @opindex Wno-traditional
3741 Warn about certain constructs that behave differently in traditional and
3742 ISO C@. Also warn about ISO C constructs that have no traditional C
3743 equivalent, and/or problematic constructs which should be avoided.
3747 Macro parameters that appear within string literals in the macro body.
3748 In traditional C macro replacement takes place within string literals,
3749 but does not in ISO C@.
3752 In traditional C, some preprocessor directives did not exist.
3753 Traditional preprocessors would only consider a line to be a directive
3754 if the @samp{#} appeared in column 1 on the line. Therefore
3755 @option{-Wtraditional} warns about directives that traditional C
3756 understands but would ignore because the @samp{#} does not appear as the
3757 first character on the line. It also suggests you hide directives like
3758 @samp{#pragma} not understood by traditional C by indenting them. Some
3759 traditional implementations would not recognize @samp{#elif}, so it
3760 suggests avoiding it altogether.
3763 A function-like macro that appears without arguments.
3766 The unary plus operator.
3769 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3770 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3771 constants.) Note, these suffixes appear in macros defined in the system
3772 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3773 Use of these macros in user code might normally lead to spurious
3774 warnings, however GCC's integrated preprocessor has enough context to
3775 avoid warning in these cases.
3778 A function declared external in one block and then used after the end of
3782 A @code{switch} statement has an operand of type @code{long}.
3785 A non-@code{static} function declaration follows a @code{static} one.
3786 This construct is not accepted by some traditional C compilers.
3789 The ISO type of an integer constant has a different width or
3790 signedness from its traditional type. This warning is only issued if
3791 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3792 typically represent bit patterns, are not warned about.
3795 Usage of ISO string concatenation is detected.
3798 Initialization of automatic aggregates.
3801 Identifier conflicts with labels. Traditional C lacks a separate
3802 namespace for labels.
3805 Initialization of unions. If the initializer is zero, the warning is
3806 omitted. This is done under the assumption that the zero initializer in
3807 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3808 initializer warnings and relies on default initialization to zero in the
3812 Conversions by prototypes between fixed/floating point values and vice
3813 versa. The absence of these prototypes when compiling with traditional
3814 C would cause serious problems. This is a subset of the possible
3815 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3818 Use of ISO C style function definitions. This warning intentionally is
3819 @emph{not} issued for prototype declarations or variadic functions
3820 because these ISO C features will appear in your code when using
3821 libiberty's traditional C compatibility macros, @code{PARAMS} and
3822 @code{VPARAMS}. This warning is also bypassed for nested functions
3823 because that feature is already a GCC extension and thus not relevant to
3824 traditional C compatibility.
3827 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3828 @opindex Wtraditional-conversion
3829 @opindex Wno-traditional-conversion
3830 Warn if a prototype causes a type conversion that is different from what
3831 would happen to the same argument in the absence of a prototype. This
3832 includes conversions of fixed point to floating and vice versa, and
3833 conversions changing the width or signedness of a fixed point argument
3834 except when the same as the default promotion.
3836 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3837 @opindex Wdeclaration-after-statement
3838 @opindex Wno-declaration-after-statement
3839 Warn when a declaration is found after a statement in a block. This
3840 construct, known from C++, was introduced with ISO C99 and is by default
3841 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3842 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3847 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3849 @item -Wno-endif-labels
3850 @opindex Wno-endif-labels
3851 @opindex Wendif-labels
3852 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3857 Warn whenever a local variable or type declaration shadows another variable,
3858 parameter, type, or class member (in C++), or whenever a built-in function
3859 is shadowed. Note that in C++, the compiler will not warn if a local variable
3860 shadows a struct/class/enum, but will warn if it shadows an explicit typedef.
3862 @item -Wlarger-than=@var{len}
3863 @opindex Wlarger-than=@var{len}
3864 @opindex Wlarger-than-@var{len}
3865 Warn whenever an object of larger than @var{len} bytes is defined.
3867 @item -Wframe-larger-than=@var{len}
3868 @opindex Wframe-larger-than
3869 Warn if the size of a function frame is larger than @var{len} bytes.
3870 The computation done to determine the stack frame size is approximate
3871 and not conservative.
3872 The actual requirements may be somewhat greater than @var{len}
3873 even if you do not get a warning. In addition, any space allocated
3874 via @code{alloca}, variable-length arrays, or related constructs
3875 is not included by the compiler when determining
3876 whether or not to issue a warning.
3878 @item -Wunsafe-loop-optimizations
3879 @opindex Wunsafe-loop-optimizations
3880 @opindex Wno-unsafe-loop-optimizations
3881 Warn if the loop cannot be optimized because the compiler could not
3882 assume anything on the bounds of the loop indices. With
3883 @option{-funsafe-loop-optimizations} warn if the compiler made
3886 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
3887 @opindex Wno-pedantic-ms-format
3888 @opindex Wpedantic-ms-format
3889 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
3890 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
3891 depending on the MS runtime, when you are using the options @option{-Wformat}
3892 and @option{-pedantic} without gnu-extensions.
3894 @item -Wpointer-arith
3895 @opindex Wpointer-arith
3896 @opindex Wno-pointer-arith
3897 Warn about anything that depends on the ``size of'' a function type or
3898 of @code{void}. GNU C assigns these types a size of 1, for
3899 convenience in calculations with @code{void *} pointers and pointers
3900 to functions. In C++, warn also when an arithmetic operation involves
3901 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3904 @opindex Wtype-limits
3905 @opindex Wno-type-limits
3906 Warn if a comparison is always true or always false due to the limited
3907 range of the data type, but do not warn for constant expressions. For
3908 example, warn if an unsigned variable is compared against zero with
3909 @samp{<} or @samp{>=}. This warning is also enabled by
3912 @item -Wbad-function-cast @r{(C and Objective-C only)}
3913 @opindex Wbad-function-cast
3914 @opindex Wno-bad-function-cast
3915 Warn whenever a function call is cast to a non-matching type.
3916 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3918 @item -Wc++-compat @r{(C and Objective-C only)}
3919 Warn about ISO C constructs that are outside of the common subset of
3920 ISO C and ISO C++, e.g.@: request for implicit conversion from
3921 @code{void *} to a pointer to non-@code{void} type.
3923 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3924 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3925 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3926 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3930 @opindex Wno-cast-qual
3931 Warn whenever a pointer is cast so as to remove a type qualifier from
3932 the target type. For example, warn if a @code{const char *} is cast
3933 to an ordinary @code{char *}.
3935 Also warn when making a cast which introduces a type qualifier in an
3936 unsafe way. For example, casting @code{char **} to @code{const char **}
3937 is unsafe, as in this example:
3940 /* p is char ** value. */
3941 const char **q = (const char **) p;
3942 /* Assignment of readonly string to const char * is OK. */
3944 /* Now char** pointer points to read-only memory. */
3949 @opindex Wcast-align
3950 @opindex Wno-cast-align
3951 Warn whenever a pointer is cast such that the required alignment of the
3952 target is increased. For example, warn if a @code{char *} is cast to
3953 an @code{int *} on machines where integers can only be accessed at
3954 two- or four-byte boundaries.
3956 @item -Wwrite-strings
3957 @opindex Wwrite-strings
3958 @opindex Wno-write-strings
3959 When compiling C, give string constants the type @code{const
3960 char[@var{length}]} so that copying the address of one into a
3961 non-@code{const} @code{char *} pointer will get a warning. These
3962 warnings will help you find at compile time code that can try to write
3963 into a string constant, but only if you have been very careful about
3964 using @code{const} in declarations and prototypes. Otherwise, it will
3965 just be a nuisance. This is why we did not make @option{-Wall} request
3968 When compiling C++, warn about the deprecated conversion from string
3969 literals to @code{char *}. This warning is enabled by default for C++
3974 @opindex Wno-clobbered
3975 Warn for variables that might be changed by @samp{longjmp} or
3976 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
3979 @opindex Wconversion
3980 @opindex Wno-conversion
3981 Warn for implicit conversions that may alter a value. This includes
3982 conversions between real and integer, like @code{abs (x)} when
3983 @code{x} is @code{double}; conversions between signed and unsigned,
3984 like @code{unsigned ui = -1}; and conversions to smaller types, like
3985 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
3986 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
3987 changed by the conversion like in @code{abs (2.0)}. Warnings about
3988 conversions between signed and unsigned integers can be disabled by
3989 using @option{-Wno-sign-conversion}.
3991 For C++, also warn for confusing overload resolution for user-defined
3992 conversions; and conversions that will never use a type conversion
3993 operator: conversions to @code{void}, the same type, a base class or a
3994 reference to them. Warnings about conversions between signed and
3995 unsigned integers are disabled by default in C++ unless
3996 @option{-Wsign-conversion} is explicitly enabled.
3998 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
3999 @opindex Wconversion-null
4000 @opindex Wno-conversion-null
4001 Do not warn for conversions between @code{NULL} and non-pointer
4002 types. @option{-Wconversion-null} is enabled by default.
4005 @opindex Wempty-body
4006 @opindex Wno-empty-body
4007 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
4008 while} statement. This warning is also enabled by @option{-Wextra}.
4010 @item -Wenum-compare
4011 @opindex Wenum-compare
4012 @opindex Wno-enum-compare
4013 Warn about a comparison between values of different enum types. In C++
4014 this warning is enabled by default. In C this warning is enabled by
4017 @item -Wjump-misses-init @r{(C, Objective-C only)}
4018 @opindex Wjump-misses-init
4019 @opindex Wno-jump-misses-init
4020 Warn if a @code{goto} statement or a @code{switch} statement jumps
4021 forward across the initialization of a variable, or jumps backward to a
4022 label after the variable has been initialized. This only warns about
4023 variables which are initialized when they are declared. This warning is
4024 only supported for C and Objective C; in C++ this sort of branch is an
4027 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
4028 can be disabled with the @option{-Wno-jump-misses-init} option.
4030 @item -Wsign-compare
4031 @opindex Wsign-compare
4032 @opindex Wno-sign-compare
4033 @cindex warning for comparison of signed and unsigned values
4034 @cindex comparison of signed and unsigned values, warning
4035 @cindex signed and unsigned values, comparison warning
4036 Warn when a comparison between signed and unsigned values could produce
4037 an incorrect result when the signed value is converted to unsigned.
4038 This warning is also enabled by @option{-Wextra}; to get the other warnings
4039 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
4041 @item -Wsign-conversion
4042 @opindex Wsign-conversion
4043 @opindex Wno-sign-conversion
4044 Warn for implicit conversions that may change the sign of an integer
4045 value, like assigning a signed integer expression to an unsigned
4046 integer variable. An explicit cast silences the warning. In C, this
4047 option is enabled also by @option{-Wconversion}.
4051 @opindex Wno-address
4052 Warn about suspicious uses of memory addresses. These include using
4053 the address of a function in a conditional expression, such as
4054 @code{void func(void); if (func)}, and comparisons against the memory
4055 address of a string literal, such as @code{if (x == "abc")}. Such
4056 uses typically indicate a programmer error: the address of a function
4057 always evaluates to true, so their use in a conditional usually
4058 indicate that the programmer forgot the parentheses in a function
4059 call; and comparisons against string literals result in unspecified
4060 behavior and are not portable in C, so they usually indicate that the
4061 programmer intended to use @code{strcmp}. This warning is enabled by
4065 @opindex Wlogical-op
4066 @opindex Wno-logical-op
4067 Warn about suspicious uses of logical operators in expressions.
4068 This includes using logical operators in contexts where a
4069 bit-wise operator is likely to be expected.
4071 @item -Waggregate-return
4072 @opindex Waggregate-return
4073 @opindex Wno-aggregate-return
4074 Warn if any functions that return structures or unions are defined or
4075 called. (In languages where you can return an array, this also elicits
4078 @item -Wno-attributes
4079 @opindex Wno-attributes
4080 @opindex Wattributes
4081 Do not warn if an unexpected @code{__attribute__} is used, such as
4082 unrecognized attributes, function attributes applied to variables,
4083 etc. This will not stop errors for incorrect use of supported
4086 @item -Wno-builtin-macro-redefined
4087 @opindex Wno-builtin-macro-redefined
4088 @opindex Wbuiltin-macro-redefined
4089 Do not warn if certain built-in macros are redefined. This suppresses
4090 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
4091 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
4093 @item -Wstrict-prototypes @r{(C and Objective-C only)}
4094 @opindex Wstrict-prototypes
4095 @opindex Wno-strict-prototypes
4096 Warn if a function is declared or defined without specifying the
4097 argument types. (An old-style function definition is permitted without
4098 a warning if preceded by a declaration which specifies the argument
4101 @item -Wold-style-declaration @r{(C and Objective-C only)}
4102 @opindex Wold-style-declaration
4103 @opindex Wno-old-style-declaration
4104 Warn for obsolescent usages, according to the C Standard, in a
4105 declaration. For example, warn if storage-class specifiers like
4106 @code{static} are not the first things in a declaration. This warning
4107 is also enabled by @option{-Wextra}.
4109 @item -Wold-style-definition @r{(C and Objective-C only)}
4110 @opindex Wold-style-definition
4111 @opindex Wno-old-style-definition
4112 Warn if an old-style function definition is used. A warning is given
4113 even if there is a previous prototype.
4115 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
4116 @opindex Wmissing-parameter-type
4117 @opindex Wno-missing-parameter-type
4118 A function parameter is declared without a type specifier in K&R-style
4125 This warning is also enabled by @option{-Wextra}.
4127 @item -Wmissing-prototypes @r{(C and Objective-C only)}
4128 @opindex Wmissing-prototypes
4129 @opindex Wno-missing-prototypes
4130 Warn if a global function is defined without a previous prototype
4131 declaration. This warning is issued even if the definition itself
4132 provides a prototype. The aim is to detect global functions that fail
4133 to be declared in header files.
4135 @item -Wmissing-declarations
4136 @opindex Wmissing-declarations
4137 @opindex Wno-missing-declarations
4138 Warn if a global function is defined without a previous declaration.
4139 Do so even if the definition itself provides a prototype.
4140 Use this option to detect global functions that are not declared in
4141 header files. In C++, no warnings are issued for function templates,
4142 or for inline functions, or for functions in anonymous namespaces.
4144 @item -Wmissing-field-initializers
4145 @opindex Wmissing-field-initializers
4146 @opindex Wno-missing-field-initializers
4150 Warn if a structure's initializer has some fields missing. For
4151 example, the following code would cause such a warning, because
4152 @code{x.h} is implicitly zero:
4155 struct s @{ int f, g, h; @};
4156 struct s x = @{ 3, 4 @};
4159 This option does not warn about designated initializers, so the following
4160 modification would not trigger a warning:
4163 struct s @{ int f, g, h; @};
4164 struct s x = @{ .f = 3, .g = 4 @};
4167 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
4168 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
4170 @item -Wmissing-format-attribute
4171 @opindex Wmissing-format-attribute
4172 @opindex Wno-missing-format-attribute
4175 Warn about function pointers which might be candidates for @code{format}
4176 attributes. Note these are only possible candidates, not absolute ones.
4177 GCC will guess that function pointers with @code{format} attributes that
4178 are used in assignment, initialization, parameter passing or return
4179 statements should have a corresponding @code{format} attribute in the
4180 resulting type. I.e.@: the left-hand side of the assignment or
4181 initialization, the type of the parameter variable, or the return type
4182 of the containing function respectively should also have a @code{format}
4183 attribute to avoid the warning.
4185 GCC will also warn about function definitions which might be
4186 candidates for @code{format} attributes. Again, these are only
4187 possible candidates. GCC will guess that @code{format} attributes
4188 might be appropriate for any function that calls a function like
4189 @code{vprintf} or @code{vscanf}, but this might not always be the
4190 case, and some functions for which @code{format} attributes are
4191 appropriate may not be detected.
4193 @item -Wno-multichar
4194 @opindex Wno-multichar
4196 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4197 Usually they indicate a typo in the user's code, as they have
4198 implementation-defined values, and should not be used in portable code.
4200 @item -Wnormalized=<none|id|nfc|nfkc>
4201 @opindex Wnormalized=
4204 @cindex character set, input normalization
4205 In ISO C and ISO C++, two identifiers are different if they are
4206 different sequences of characters. However, sometimes when characters
4207 outside the basic ASCII character set are used, you can have two
4208 different character sequences that look the same. To avoid confusion,
4209 the ISO 10646 standard sets out some @dfn{normalization rules} which
4210 when applied ensure that two sequences that look the same are turned into
4211 the same sequence. GCC can warn you if you are using identifiers which
4212 have not been normalized; this option controls that warning.
4214 There are four levels of warning that GCC supports. The default is
4215 @option{-Wnormalized=nfc}, which warns about any identifier which is
4216 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4217 recommended form for most uses.
4219 Unfortunately, there are some characters which ISO C and ISO C++ allow
4220 in identifiers that when turned into NFC aren't allowable as
4221 identifiers. That is, there's no way to use these symbols in portable
4222 ISO C or C++ and have all your identifiers in NFC@.
4223 @option{-Wnormalized=id} suppresses the warning for these characters.
4224 It is hoped that future versions of the standards involved will correct
4225 this, which is why this option is not the default.
4227 You can switch the warning off for all characters by writing
4228 @option{-Wnormalized=none}. You would only want to do this if you
4229 were using some other normalization scheme (like ``D''), because
4230 otherwise you can easily create bugs that are literally impossible to see.
4232 Some characters in ISO 10646 have distinct meanings but look identical
4233 in some fonts or display methodologies, especially once formatting has
4234 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4235 LETTER N'', will display just like a regular @code{n} which has been
4236 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4237 normalization scheme to convert all these into a standard form as
4238 well, and GCC will warn if your code is not in NFKC if you use
4239 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4240 about every identifier that contains the letter O because it might be
4241 confused with the digit 0, and so is not the default, but may be
4242 useful as a local coding convention if the programming environment is
4243 unable to be fixed to display these characters distinctly.
4245 @item -Wno-deprecated
4246 @opindex Wno-deprecated
4247 @opindex Wdeprecated
4248 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4250 @item -Wno-deprecated-declarations
4251 @opindex Wno-deprecated-declarations
4252 @opindex Wdeprecated-declarations
4253 Do not warn about uses of functions (@pxref{Function Attributes}),
4254 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4255 Attributes}) marked as deprecated by using the @code{deprecated}
4259 @opindex Wno-overflow
4261 Do not warn about compile-time overflow in constant expressions.
4263 @item -Woverride-init @r{(C and Objective-C only)}
4264 @opindex Woverride-init
4265 @opindex Wno-override-init
4269 Warn if an initialized field without side effects is overridden when
4270 using designated initializers (@pxref{Designated Inits, , Designated
4273 This warning is included in @option{-Wextra}. To get other
4274 @option{-Wextra} warnings without this one, use @samp{-Wextra
4275 -Wno-override-init}.
4280 Warn if a structure is given the packed attribute, but the packed
4281 attribute has no effect on the layout or size of the structure.
4282 Such structures may be mis-aligned for little benefit. For
4283 instance, in this code, the variable @code{f.x} in @code{struct bar}
4284 will be misaligned even though @code{struct bar} does not itself
4285 have the packed attribute:
4292 @} __attribute__((packed));
4300 @item -Wpacked-bitfield-compat
4301 @opindex Wpacked-bitfield-compat
4302 @opindex Wno-packed-bitfield-compat
4303 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4304 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4305 the change can lead to differences in the structure layout. GCC
4306 informs you when the offset of such a field has changed in GCC 4.4.
4307 For example there is no longer a 4-bit padding between field @code{a}
4308 and @code{b} in this structure:
4315 @} __attribute__ ((packed));
4318 This warning is enabled by default. Use
4319 @option{-Wno-packed-bitfield-compat} to disable this warning.
4324 Warn if padding is included in a structure, either to align an element
4325 of the structure or to align the whole structure. Sometimes when this
4326 happens it is possible to rearrange the fields of the structure to
4327 reduce the padding and so make the structure smaller.
4329 @item -Wredundant-decls
4330 @opindex Wredundant-decls
4331 @opindex Wno-redundant-decls
4332 Warn if anything is declared more than once in the same scope, even in
4333 cases where multiple declaration is valid and changes nothing.
4335 @item -Wnested-externs @r{(C and Objective-C only)}
4336 @opindex Wnested-externs
4337 @opindex Wno-nested-externs
4338 Warn if an @code{extern} declaration is encountered within a function.
4343 Warn if a function can not be inlined and it was declared as inline.
4344 Even with this option, the compiler will not warn about failures to
4345 inline functions declared in system headers.
4347 The compiler uses a variety of heuristics to determine whether or not
4348 to inline a function. For example, the compiler takes into account
4349 the size of the function being inlined and the amount of inlining
4350 that has already been done in the current function. Therefore,
4351 seemingly insignificant changes in the source program can cause the
4352 warnings produced by @option{-Winline} to appear or disappear.
4354 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4355 @opindex Wno-invalid-offsetof
4356 @opindex Winvalid-offsetof
4357 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4358 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4359 to a non-POD type is undefined. In existing C++ implementations,
4360 however, @samp{offsetof} typically gives meaningful results even when
4361 applied to certain kinds of non-POD types. (Such as a simple
4362 @samp{struct} that fails to be a POD type only by virtue of having a
4363 constructor.) This flag is for users who are aware that they are
4364 writing nonportable code and who have deliberately chosen to ignore the
4367 The restrictions on @samp{offsetof} may be relaxed in a future version
4368 of the C++ standard.
4370 @item -Wno-int-to-pointer-cast
4371 @opindex Wno-int-to-pointer-cast
4372 @opindex Wint-to-pointer-cast
4373 Suppress warnings from casts to pointer type of an integer of a
4374 different size. In C++, casting to a pointer type of smaller size is
4375 an error. @option{Wint-to-pointer-cast} is enabled by default.
4378 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4379 @opindex Wno-pointer-to-int-cast
4380 @opindex Wpointer-to-int-cast
4381 Suppress warnings from casts from a pointer to an integer type of a
4385 @opindex Winvalid-pch
4386 @opindex Wno-invalid-pch
4387 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4388 the search path but can't be used.
4392 @opindex Wno-long-long
4393 Warn if @samp{long long} type is used. This is enabled by either
4394 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4395 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4397 @item -Wvariadic-macros
4398 @opindex Wvariadic-macros
4399 @opindex Wno-variadic-macros
4400 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4401 alternate syntax when in pedantic ISO C99 mode. This is default.
4402 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4407 Warn if variable length array is used in the code.
4408 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4409 the variable length array.
4411 @item -Wvolatile-register-var
4412 @opindex Wvolatile-register-var
4413 @opindex Wno-volatile-register-var
4414 Warn if a register variable is declared volatile. The volatile
4415 modifier does not inhibit all optimizations that may eliminate reads
4416 and/or writes to register variables. This warning is enabled by
4419 @item -Wdisabled-optimization
4420 @opindex Wdisabled-optimization
4421 @opindex Wno-disabled-optimization
4422 Warn if a requested optimization pass is disabled. This warning does
4423 not generally indicate that there is anything wrong with your code; it
4424 merely indicates that GCC's optimizers were unable to handle the code
4425 effectively. Often, the problem is that your code is too big or too
4426 complex; GCC will refuse to optimize programs when the optimization
4427 itself is likely to take inordinate amounts of time.
4429 @item -Wpointer-sign @r{(C and Objective-C only)}
4430 @opindex Wpointer-sign
4431 @opindex Wno-pointer-sign
4432 Warn for pointer argument passing or assignment with different signedness.
4433 This option is only supported for C and Objective-C@. It is implied by
4434 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4435 @option{-Wno-pointer-sign}.
4437 @item -Wstack-protector
4438 @opindex Wstack-protector
4439 @opindex Wno-stack-protector
4440 This option is only active when @option{-fstack-protector} is active. It
4441 warns about functions that will not be protected against stack smashing.
4444 @opindex Wno-mudflap
4445 Suppress warnings about constructs that cannot be instrumented by
4448 @item -Woverlength-strings
4449 @opindex Woverlength-strings
4450 @opindex Wno-overlength-strings
4451 Warn about string constants which are longer than the ``minimum
4452 maximum'' length specified in the C standard. Modern compilers
4453 generally allow string constants which are much longer than the
4454 standard's minimum limit, but very portable programs should avoid
4455 using longer strings.
4457 The limit applies @emph{after} string constant concatenation, and does
4458 not count the trailing NUL@. In C90, the limit was 509 characters; in
4459 C99, it was raised to 4095. C++98 does not specify a normative
4460 minimum maximum, so we do not diagnose overlength strings in C++@.
4462 This option is implied by @option{-pedantic}, and can be disabled with
4463 @option{-Wno-overlength-strings}.
4465 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
4466 @opindex Wunsuffixed-float-constants
4468 GCC will issue a warning for any floating constant that does not have
4469 a suffix. When used together with @option{-Wsystem-headers} it will
4470 warn about such constants in system header files. This can be useful
4471 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4472 from the decimal floating-point extension to C99.
4475 @node Debugging Options
4476 @section Options for Debugging Your Program or GCC
4477 @cindex options, debugging
4478 @cindex debugging information options
4480 GCC has various special options that are used for debugging
4481 either your program or GCC:
4486 Produce debugging information in the operating system's native format
4487 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4490 On most systems that use stabs format, @option{-g} enables use of extra
4491 debugging information that only GDB can use; this extra information
4492 makes debugging work better in GDB but will probably make other debuggers
4494 refuse to read the program. If you want to control for certain whether
4495 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4496 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4498 GCC allows you to use @option{-g} with
4499 @option{-O}. The shortcuts taken by optimized code may occasionally
4500 produce surprising results: some variables you declared may not exist
4501 at all; flow of control may briefly move where you did not expect it;
4502 some statements may not be executed because they compute constant
4503 results or their values were already at hand; some statements may
4504 execute in different places because they were moved out of loops.
4506 Nevertheless it proves possible to debug optimized output. This makes
4507 it reasonable to use the optimizer for programs that might have bugs.
4509 The following options are useful when GCC is generated with the
4510 capability for more than one debugging format.
4514 Produce debugging information for use by GDB@. This means to use the
4515 most expressive format available (DWARF 2, stabs, or the native format
4516 if neither of those are supported), including GDB extensions if at all
4521 Produce debugging information in stabs format (if that is supported),
4522 without GDB extensions. This is the format used by DBX on most BSD
4523 systems. On MIPS, Alpha and System V Release 4 systems this option
4524 produces stabs debugging output which is not understood by DBX or SDB@.
4525 On System V Release 4 systems this option requires the GNU assembler.
4527 @item -feliminate-unused-debug-symbols
4528 @opindex feliminate-unused-debug-symbols
4529 Produce debugging information in stabs format (if that is supported),
4530 for only symbols that are actually used.
4532 @item -femit-class-debug-always
4533 Instead of emitting debugging information for a C++ class in only one
4534 object file, emit it in all object files using the class. This option
4535 should be used only with debuggers that are unable to handle the way GCC
4536 normally emits debugging information for classes because using this
4537 option will increase the size of debugging information by as much as a
4542 Produce debugging information in stabs format (if that is supported),
4543 using GNU extensions understood only by the GNU debugger (GDB)@. The
4544 use of these extensions is likely to make other debuggers crash or
4545 refuse to read the program.
4549 Produce debugging information in COFF format (if that is supported).
4550 This is the format used by SDB on most System V systems prior to
4555 Produce debugging information in XCOFF format (if that is supported).
4556 This is the format used by the DBX debugger on IBM RS/6000 systems.
4560 Produce debugging information in XCOFF 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, and may cause assemblers other than the GNU
4564 assembler (GAS) to fail with an error.
4566 @item -gdwarf-@var{version}
4567 @opindex gdwarf-@var{version}
4568 Produce debugging information in DWARF format (if that is
4569 supported). This is the format used by DBX on IRIX 6. The value
4570 of @var{version} may be either 2, 3 or 4; the default version is 2.
4572 Note that with DWARF version 2 some ports require, and will always
4573 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4575 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
4576 for maximum benefit.
4578 @item -gstrict-dwarf
4579 @opindex gstrict-dwarf
4580 Disallow using extensions of later DWARF standard version than selected
4581 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
4582 DWARF extensions from later standard versions is allowed.
4584 @item -gno-strict-dwarf
4585 @opindex gno-strict-dwarf
4586 Allow using extensions of later DWARF standard version than selected with
4587 @option{-gdwarf-@var{version}}.
4591 Produce debugging information in VMS debug format (if that is
4592 supported). This is the format used by DEBUG on VMS systems.
4595 @itemx -ggdb@var{level}
4596 @itemx -gstabs@var{level}
4597 @itemx -gcoff@var{level}
4598 @itemx -gxcoff@var{level}
4599 @itemx -gvms@var{level}
4600 Request debugging information and also use @var{level} to specify how
4601 much information. The default level is 2.
4603 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4606 Level 1 produces minimal information, enough for making backtraces in
4607 parts of the program that you don't plan to debug. This includes
4608 descriptions of functions and external variables, but no information
4609 about local variables and no line numbers.
4611 Level 3 includes extra information, such as all the macro definitions
4612 present in the program. Some debuggers support macro expansion when
4613 you use @option{-g3}.
4615 @option{-gdwarf-2} does not accept a concatenated debug level, because
4616 GCC used to support an option @option{-gdwarf} that meant to generate
4617 debug information in version 1 of the DWARF format (which is very
4618 different from version 2), and it would have been too confusing. That
4619 debug format is long obsolete, but the option cannot be changed now.
4620 Instead use an additional @option{-g@var{level}} option to change the
4621 debug level for DWARF.
4625 Turn off generation of debug info, if leaving out this option would have
4626 generated it, or turn it on at level 2 otherwise. The position of this
4627 argument in the command line does not matter, it takes effect after all
4628 other options are processed, and it does so only once, no matter how
4629 many times it is given. This is mainly intended to be used with
4630 @option{-fcompare-debug}.
4632 @item -fdump-final-insns@r{[}=@var{file}@r{]}
4633 @opindex fdump-final-insns
4634 Dump the final internal representation (RTL) to @var{file}. If the
4635 optional argument is omitted (or if @var{file} is @code{.}), the name
4636 of the dump file will be determined by appending @code{.gkd} to the
4637 compilation output file name.
4639 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4640 @opindex fcompare-debug
4641 @opindex fno-compare-debug
4642 If no error occurs during compilation, run the compiler a second time,
4643 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4644 passed to the second compilation. Dump the final internal
4645 representation in both compilations, and print an error if they differ.
4647 If the equal sign is omitted, the default @option{-gtoggle} is used.
4649 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4650 and nonzero, implicitly enables @option{-fcompare-debug}. If
4651 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4652 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4655 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4656 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4657 of the final representation and the second compilation, preventing even
4658 @env{GCC_COMPARE_DEBUG} from taking effect.
4660 To verify full coverage during @option{-fcompare-debug} testing, set
4661 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4662 which GCC will reject as an invalid option in any actual compilation
4663 (rather than preprocessing, assembly or linking). To get just a
4664 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4665 not overridden} will do.
4667 @item -fcompare-debug-second
4668 @opindex fcompare-debug-second
4669 This option is implicitly passed to the compiler for the second
4670 compilation requested by @option{-fcompare-debug}, along with options to
4671 silence warnings, and omitting other options that would cause
4672 side-effect compiler outputs to files or to the standard output. Dump
4673 files and preserved temporary files are renamed so as to contain the
4674 @code{.gk} additional extension during the second compilation, to avoid
4675 overwriting those generated by the first.
4677 When this option is passed to the compiler driver, it causes the
4678 @emph{first} compilation to be skipped, which makes it useful for little
4679 other than debugging the compiler proper.
4681 @item -feliminate-dwarf2-dups
4682 @opindex feliminate-dwarf2-dups
4683 Compress DWARF2 debugging information by eliminating duplicated
4684 information about each symbol. This option only makes sense when
4685 generating DWARF2 debugging information with @option{-gdwarf-2}.
4687 @item -femit-struct-debug-baseonly
4688 Emit debug information for struct-like types
4689 only when the base name of the compilation source file
4690 matches the base name of file in which the struct was defined.
4692 This option substantially reduces the size of debugging information,
4693 but at significant potential loss in type information to the debugger.
4694 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4695 See @option{-femit-struct-debug-detailed} for more detailed control.
4697 This option works only with DWARF 2.
4699 @item -femit-struct-debug-reduced
4700 Emit debug information for struct-like types
4701 only when the base name of the compilation source file
4702 matches the base name of file in which the type was defined,
4703 unless the struct is a template or defined in a system header.
4705 This option significantly reduces the size of debugging information,
4706 with some potential loss in type information to the debugger.
4707 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4708 See @option{-femit-struct-debug-detailed} for more detailed control.
4710 This option works only with DWARF 2.
4712 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4713 Specify the struct-like types
4714 for which the compiler will generate debug information.
4715 The intent is to reduce duplicate struct debug information
4716 between different object files within the same program.
4718 This option is a detailed version of
4719 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4720 which will serve for most needs.
4722 A specification has the syntax@*
4723 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4725 The optional first word limits the specification to
4726 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4727 A struct type is used directly when it is the type of a variable, member.
4728 Indirect uses arise through pointers to structs.
4729 That is, when use of an incomplete struct would be legal, the use is indirect.
4731 @samp{struct one direct; struct two * indirect;}.
4733 The optional second word limits the specification to
4734 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4735 Generic structs are a bit complicated to explain.
4736 For C++, these are non-explicit specializations of template classes,
4737 or non-template classes within the above.
4738 Other programming languages have generics,
4739 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4741 The third word specifies the source files for those
4742 structs for which the compiler will emit debug information.
4743 The values @samp{none} and @samp{any} have the normal meaning.
4744 The value @samp{base} means that
4745 the base of name of the file in which the type declaration appears
4746 must match the base of the name of the main compilation file.
4747 In practice, this means that
4748 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4749 but types declared in other header will not.
4750 The value @samp{sys} means those types satisfying @samp{base}
4751 or declared in system or compiler headers.
4753 You may need to experiment to determine the best settings for your application.
4755 The default is @samp{-femit-struct-debug-detailed=all}.
4757 This option works only with DWARF 2.
4759 @item -fenable-icf-debug
4760 @opindex fenable-icf-debug
4761 Generate additional debug information to support identical code folding (ICF).
4762 This option only works with DWARF version 2 or higher.
4764 @item -fno-merge-debug-strings
4765 @opindex fmerge-debug-strings
4766 @opindex fno-merge-debug-strings
4767 Direct the linker to not merge together strings in the debugging
4768 information which are identical in different object files. Merging is
4769 not supported by all assemblers or linkers. Merging decreases the size
4770 of the debug information in the output file at the cost of increasing
4771 link processing time. Merging is enabled by default.
4773 @item -fdebug-prefix-map=@var{old}=@var{new}
4774 @opindex fdebug-prefix-map
4775 When compiling files in directory @file{@var{old}}, record debugging
4776 information describing them as in @file{@var{new}} instead.
4778 @item -fno-dwarf2-cfi-asm
4779 @opindex fdwarf2-cfi-asm
4780 @opindex fno-dwarf2-cfi-asm
4781 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4782 instead of using GAS @code{.cfi_*} directives.
4784 @cindex @command{prof}
4787 Generate extra code to write profile information suitable for the
4788 analysis program @command{prof}. You must use this option when compiling
4789 the source files you want data about, and you must also use it when
4792 @cindex @command{gprof}
4795 Generate extra code to write profile information suitable for the
4796 analysis program @command{gprof}. You must use this option when compiling
4797 the source files you want data about, and you must also use it when
4802 Makes the compiler print out each function name as it is compiled, and
4803 print some statistics about each pass when it finishes.
4806 @opindex ftime-report
4807 Makes the compiler print some statistics about the time consumed by each
4808 pass when it finishes.
4811 @opindex fmem-report
4812 Makes the compiler print some statistics about permanent memory
4813 allocation when it finishes.
4815 @item -fpre-ipa-mem-report
4816 @opindex fpre-ipa-mem-report
4817 @item -fpost-ipa-mem-report
4818 @opindex fpost-ipa-mem-report
4819 Makes the compiler print some statistics about permanent memory
4820 allocation before or after interprocedural optimization.
4823 @opindex fstack-usage
4824 Makes the compiler output stack usage information for the program, on a
4825 per-function basis. The filename for the dump is made by appending
4826 @file{.su} to the AUXNAME. AUXNAME is generated from the name of
4827 the output file, if explicitly specified and it is not an executable,
4828 otherwise it is the basename of the source file. An entry is made up
4833 The name of the function.
4837 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
4840 The qualifier @code{static} means that the function manipulates the stack
4841 statically: a fixed number of bytes are allocated for the frame on function
4842 entry and released on function exit; no stack adjustments are otherwise made
4843 in the function. The second field is this fixed number of bytes.
4845 The qualifier @code{dynamic} means that the function manipulates the stack
4846 dynamically: in addition to the static allocation described above, stack
4847 adjustments are made in the body of the function, for example to push/pop
4848 arguments around function calls. If the qualifier @code{bounded} is also
4849 present, the amount of these adjustments is bounded at compile-time and
4850 the second field is an upper bound of the total amount of stack used by
4851 the function. If it is not present, the amount of these adjustments is
4852 not bounded at compile-time and the second field only represents the
4855 @item -fprofile-arcs
4856 @opindex fprofile-arcs
4857 Add code so that program flow @dfn{arcs} are instrumented. During
4858 execution the program records how many times each branch and call is
4859 executed and how many times it is taken or returns. When the compiled
4860 program exits it saves this data to a file called
4861 @file{@var{auxname}.gcda} for each source file. The data may be used for
4862 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4863 test coverage analysis (@option{-ftest-coverage}). Each object file's
4864 @var{auxname} is generated from the name of the output file, if
4865 explicitly specified and it is not the final executable, otherwise it is
4866 the basename of the source file. In both cases any suffix is removed
4867 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4868 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4869 @xref{Cross-profiling}.
4871 @cindex @command{gcov}
4875 This option is used to compile and link code instrumented for coverage
4876 analysis. The option is a synonym for @option{-fprofile-arcs}
4877 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4878 linking). See the documentation for those options for more details.
4883 Compile the source files with @option{-fprofile-arcs} plus optimization
4884 and code generation options. For test coverage analysis, use the
4885 additional @option{-ftest-coverage} option. You do not need to profile
4886 every source file in a program.
4889 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4890 (the latter implies the former).
4893 Run the program on a representative workload to generate the arc profile
4894 information. This may be repeated any number of times. You can run
4895 concurrent instances of your program, and provided that the file system
4896 supports locking, the data files will be correctly updated. Also
4897 @code{fork} calls are detected and correctly handled (double counting
4901 For profile-directed optimizations, compile the source files again with
4902 the same optimization and code generation options plus
4903 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4904 Control Optimization}).
4907 For test coverage analysis, use @command{gcov} to produce human readable
4908 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4909 @command{gcov} documentation for further information.
4913 With @option{-fprofile-arcs}, for each function of your program GCC
4914 creates a program flow graph, then finds a spanning tree for the graph.
4915 Only arcs that are not on the spanning tree have to be instrumented: the
4916 compiler adds code to count the number of times that these arcs are
4917 executed. When an arc is the only exit or only entrance to a block, the
4918 instrumentation code can be added to the block; otherwise, a new basic
4919 block must be created to hold the instrumentation code.
4922 @item -ftest-coverage
4923 @opindex ftest-coverage
4924 Produce a notes file that the @command{gcov} code-coverage utility
4925 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4926 show program coverage. Each source file's note file is called
4927 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4928 above for a description of @var{auxname} and instructions on how to
4929 generate test coverage data. Coverage data will match the source files
4930 more closely, if you do not optimize.
4932 @item -fdbg-cnt-list
4933 @opindex fdbg-cnt-list
4934 Print the name and the counter upperbound for all debug counters.
4936 @item -fdbg-cnt=@var{counter-value-list}
4938 Set the internal debug counter upperbound. @var{counter-value-list}
4939 is a comma-separated list of @var{name}:@var{value} pairs
4940 which sets the upperbound of each debug counter @var{name} to @var{value}.
4941 All debug counters have the initial upperbound of @var{UINT_MAX},
4942 thus dbg_cnt() returns true always unless the upperbound is set by this option.
4943 e.g. With -fdbg-cnt=dce:10,tail_call:0
4944 dbg_cnt(dce) will return true only for first 10 invocations
4945 and dbg_cnt(tail_call) will return false always.
4947 @item -d@var{letters}
4948 @itemx -fdump-rtl-@var{pass}
4950 Says to make debugging dumps during compilation at times specified by
4951 @var{letters}. This is used for debugging the RTL-based passes of the
4952 compiler. The file names for most of the dumps are made by appending
4953 a pass number and a word to the @var{dumpname}, and the files are
4954 created in the directory of the output file. @var{dumpname} is
4955 generated from the name of the output file, if explicitly specified
4956 and it is not an executable, otherwise it is the basename of the
4957 source file. These switches may have different effects when
4958 @option{-E} is used for preprocessing.
4960 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
4961 @option{-d} option @var{letters}. Here are the possible
4962 letters for use in @var{pass} and @var{letters}, and their meanings:
4966 @item -fdump-rtl-alignments
4967 @opindex fdump-rtl-alignments
4968 Dump after branch alignments have been computed.
4970 @item -fdump-rtl-asmcons
4971 @opindex fdump-rtl-asmcons
4972 Dump after fixing rtl statements that have unsatisfied in/out constraints.
4974 @item -fdump-rtl-auto_inc_dec
4975 @opindex fdump-rtl-auto_inc_dec
4976 Dump after auto-inc-dec discovery. This pass is only run on
4977 architectures that have auto inc or auto dec instructions.
4979 @item -fdump-rtl-barriers
4980 @opindex fdump-rtl-barriers
4981 Dump after cleaning up the barrier instructions.
4983 @item -fdump-rtl-bbpart
4984 @opindex fdump-rtl-bbpart
4985 Dump after partitioning hot and cold basic blocks.
4987 @item -fdump-rtl-bbro
4988 @opindex fdump-rtl-bbro
4989 Dump after block reordering.
4991 @item -fdump-rtl-btl1
4992 @itemx -fdump-rtl-btl2
4993 @opindex fdump-rtl-btl2
4994 @opindex fdump-rtl-btl2
4995 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
4996 after the two branch
4997 target load optimization passes.
4999 @item -fdump-rtl-bypass
5000 @opindex fdump-rtl-bypass
5001 Dump after jump bypassing and control flow optimizations.
5003 @item -fdump-rtl-combine
5004 @opindex fdump-rtl-combine
5005 Dump after the RTL instruction combination pass.
5007 @item -fdump-rtl-compgotos
5008 @opindex fdump-rtl-compgotos
5009 Dump after duplicating the computed gotos.
5011 @item -fdump-rtl-ce1
5012 @itemx -fdump-rtl-ce2
5013 @itemx -fdump-rtl-ce3
5014 @opindex fdump-rtl-ce1
5015 @opindex fdump-rtl-ce2
5016 @opindex fdump-rtl-ce3
5017 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
5018 @option{-fdump-rtl-ce3} enable dumping after the three
5019 if conversion passes.
5021 @itemx -fdump-rtl-cprop_hardreg
5022 @opindex fdump-rtl-cprop_hardreg
5023 Dump after hard register copy propagation.
5025 @itemx -fdump-rtl-csa
5026 @opindex fdump-rtl-csa
5027 Dump after combining stack adjustments.
5029 @item -fdump-rtl-cse1
5030 @itemx -fdump-rtl-cse2
5031 @opindex fdump-rtl-cse1
5032 @opindex fdump-rtl-cse2
5033 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
5034 the two common sub-expression elimination passes.
5036 @itemx -fdump-rtl-dce
5037 @opindex fdump-rtl-dce
5038 Dump after the standalone dead code elimination passes.
5040 @itemx -fdump-rtl-dbr
5041 @opindex fdump-rtl-dbr
5042 Dump after delayed branch scheduling.
5044 @item -fdump-rtl-dce1
5045 @itemx -fdump-rtl-dce2
5046 @opindex fdump-rtl-dce1
5047 @opindex fdump-rtl-dce2
5048 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
5049 the two dead store elimination passes.
5052 @opindex fdump-rtl-eh
5053 Dump after finalization of EH handling code.
5055 @item -fdump-rtl-eh_ranges
5056 @opindex fdump-rtl-eh_ranges
5057 Dump after conversion of EH handling range regions.
5059 @item -fdump-rtl-expand
5060 @opindex fdump-rtl-expand
5061 Dump after RTL generation.
5063 @item -fdump-rtl-fwprop1
5064 @itemx -fdump-rtl-fwprop2
5065 @opindex fdump-rtl-fwprop1
5066 @opindex fdump-rtl-fwprop2
5067 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
5068 dumping after the two forward propagation passes.
5070 @item -fdump-rtl-gcse1
5071 @itemx -fdump-rtl-gcse2
5072 @opindex fdump-rtl-gcse1
5073 @opindex fdump-rtl-gcse2
5074 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
5075 after global common subexpression elimination.
5077 @item -fdump-rtl-init-regs
5078 @opindex fdump-rtl-init-regs
5079 Dump after the initialization of the registers.
5081 @item -fdump-rtl-initvals
5082 @opindex fdump-rtl-initvals
5083 Dump after the computation of the initial value sets.
5085 @itemx -fdump-rtl-into_cfglayout
5086 @opindex fdump-rtl-into_cfglayout
5087 Dump after converting to cfglayout mode.
5089 @item -fdump-rtl-ira
5090 @opindex fdump-rtl-ira
5091 Dump after iterated register allocation.
5093 @item -fdump-rtl-jump
5094 @opindex fdump-rtl-jump
5095 Dump after the second jump optimization.
5097 @item -fdump-rtl-loop2
5098 @opindex fdump-rtl-loop2
5099 @option{-fdump-rtl-loop2} enables dumping after the rtl
5100 loop optimization passes.
5102 @item -fdump-rtl-mach
5103 @opindex fdump-rtl-mach
5104 Dump after performing the machine dependent reorganization pass, if that
5107 @item -fdump-rtl-mode_sw
5108 @opindex fdump-rtl-mode_sw
5109 Dump after removing redundant mode switches.
5111 @item -fdump-rtl-rnreg
5112 @opindex fdump-rtl-rnreg
5113 Dump after register renumbering.
5115 @itemx -fdump-rtl-outof_cfglayout
5116 @opindex fdump-rtl-outof_cfglayout
5117 Dump after converting from cfglayout mode.
5119 @item -fdump-rtl-peephole2
5120 @opindex fdump-rtl-peephole2
5121 Dump after the peephole pass.
5123 @item -fdump-rtl-postreload
5124 @opindex fdump-rtl-postreload
5125 Dump after post-reload optimizations.
5127 @itemx -fdump-rtl-pro_and_epilogue
5128 @opindex fdump-rtl-pro_and_epilogue
5129 Dump after generating the function pro and epilogues.
5131 @item -fdump-rtl-regmove
5132 @opindex fdump-rtl-regmove
5133 Dump after the register move pass.
5135 @item -fdump-rtl-sched1
5136 @itemx -fdump-rtl-sched2
5137 @opindex fdump-rtl-sched1
5138 @opindex fdump-rtl-sched2
5139 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
5140 after the basic block scheduling passes.
5142 @item -fdump-rtl-see
5143 @opindex fdump-rtl-see
5144 Dump after sign extension elimination.
5146 @item -fdump-rtl-seqabstr
5147 @opindex fdump-rtl-seqabstr
5148 Dump after common sequence discovery.
5150 @item -fdump-rtl-shorten
5151 @opindex fdump-rtl-shorten
5152 Dump after shortening branches.
5154 @item -fdump-rtl-sibling
5155 @opindex fdump-rtl-sibling
5156 Dump after sibling call optimizations.
5158 @item -fdump-rtl-split1
5159 @itemx -fdump-rtl-split2
5160 @itemx -fdump-rtl-split3
5161 @itemx -fdump-rtl-split4
5162 @itemx -fdump-rtl-split5
5163 @opindex fdump-rtl-split1
5164 @opindex fdump-rtl-split2
5165 @opindex fdump-rtl-split3
5166 @opindex fdump-rtl-split4
5167 @opindex fdump-rtl-split5
5168 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
5169 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
5170 @option{-fdump-rtl-split5} enable dumping after five rounds of
5171 instruction splitting.
5173 @item -fdump-rtl-sms
5174 @opindex fdump-rtl-sms
5175 Dump after modulo scheduling. This pass is only run on some
5178 @item -fdump-rtl-stack
5179 @opindex fdump-rtl-stack
5180 Dump after conversion from GCC's "flat register file" registers to the
5181 x87's stack-like registers. This pass is only run on x86 variants.
5183 @item -fdump-rtl-subreg1
5184 @itemx -fdump-rtl-subreg2
5185 @opindex fdump-rtl-subreg1
5186 @opindex fdump-rtl-subreg2
5187 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
5188 the two subreg expansion passes.
5190 @item -fdump-rtl-unshare
5191 @opindex fdump-rtl-unshare
5192 Dump after all rtl has been unshared.
5194 @item -fdump-rtl-vartrack
5195 @opindex fdump-rtl-vartrack
5196 Dump after variable tracking.
5198 @item -fdump-rtl-vregs
5199 @opindex fdump-rtl-vregs
5200 Dump after converting virtual registers to hard registers.
5202 @item -fdump-rtl-web
5203 @opindex fdump-rtl-web
5204 Dump after live range splitting.
5206 @item -fdump-rtl-regclass
5207 @itemx -fdump-rtl-subregs_of_mode_init
5208 @itemx -fdump-rtl-subregs_of_mode_finish
5209 @itemx -fdump-rtl-dfinit
5210 @itemx -fdump-rtl-dfinish
5211 @opindex fdump-rtl-regclass
5212 @opindex fdump-rtl-subregs_of_mode_init
5213 @opindex fdump-rtl-subregs_of_mode_finish
5214 @opindex fdump-rtl-dfinit
5215 @opindex fdump-rtl-dfinish
5216 These dumps are defined but always produce empty files.
5218 @item -fdump-rtl-all
5219 @opindex fdump-rtl-all
5220 Produce all the dumps listed above.
5224 Annotate the assembler output with miscellaneous debugging information.
5228 Dump all macro definitions, at the end of preprocessing, in addition to
5233 Produce a core dump whenever an error occurs.
5237 Print statistics on memory usage, at the end of the run, to
5242 Annotate the assembler output with a comment indicating which
5243 pattern and alternative was used. The length of each instruction is
5248 Dump the RTL in the assembler output as a comment before each instruction.
5249 Also turns on @option{-dp} annotation.
5253 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
5254 dump a representation of the control flow graph suitable for viewing with VCG
5255 to @file{@var{file}.@var{pass}.vcg}.
5259 Just generate RTL for a function instead of compiling it. Usually used
5260 with @option{-fdump-rtl-expand}.
5264 Dump debugging information during parsing, to standard error.
5268 @opindex fdump-noaddr
5269 When doing debugging dumps, suppress address output. This makes it more
5270 feasible to use diff on debugging dumps for compiler invocations with
5271 different compiler binaries and/or different
5272 text / bss / data / heap / stack / dso start locations.
5274 @item -fdump-unnumbered
5275 @opindex fdump-unnumbered
5276 When doing debugging dumps, suppress instruction numbers and address output.
5277 This makes it more feasible to use diff on debugging dumps for compiler
5278 invocations with different options, in particular with and without
5281 @item -fdump-unnumbered-links
5282 @opindex fdump-unnumbered-links
5283 When doing debugging dumps (see @option{-d} option above), suppress
5284 instruction numbers for the links to the previous and next instructions
5287 @item -fdump-translation-unit @r{(C++ only)}
5288 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5289 @opindex fdump-translation-unit
5290 Dump a representation of the tree structure for the entire translation
5291 unit to a file. The file name is made by appending @file{.tu} to the
5292 source file name, and the file is created in the same directory as the
5293 output file. If the @samp{-@var{options}} form is used, @var{options}
5294 controls the details of the dump as described for the
5295 @option{-fdump-tree} options.
5297 @item -fdump-class-hierarchy @r{(C++ only)}
5298 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5299 @opindex fdump-class-hierarchy
5300 Dump a representation of each class's hierarchy and virtual function
5301 table layout to a file. The file name is made by appending
5302 @file{.class} to the source file name, and the file is created in the
5303 same directory as the output file. If the @samp{-@var{options}} form
5304 is used, @var{options} controls the details of the dump as described
5305 for the @option{-fdump-tree} options.
5307 @item -fdump-ipa-@var{switch}
5309 Control the dumping at various stages of inter-procedural analysis
5310 language tree to a file. The file name is generated by appending a
5311 switch specific suffix to the source file name, and the file is created
5312 in the same directory as the output file. The following dumps are
5317 Enables all inter-procedural analysis dumps.
5320 Dumps information about call-graph optimization, unused function removal,
5321 and inlining decisions.
5324 Dump after function inlining.
5328 @item -fdump-statistics-@var{option}
5329 @opindex fdump-statistics
5330 Enable and control dumping of pass statistics in a separate file. The
5331 file name is generated by appending a suffix ending in
5332 @samp{.statistics} to the source file name, and the file is created in
5333 the same directory as the output file. If the @samp{-@var{option}}
5334 form is used, @samp{-stats} will cause counters to be summed over the
5335 whole compilation unit while @samp{-details} will dump every event as
5336 the passes generate them. The default with no option is to sum
5337 counters for each function compiled.
5339 @item -fdump-tree-@var{switch}
5340 @itemx -fdump-tree-@var{switch}-@var{options}
5342 Control the dumping at various stages of processing the intermediate
5343 language tree to a file. The file name is generated by appending a
5344 switch specific suffix to the source file name, and the file is
5345 created in the same directory as the output file. If the
5346 @samp{-@var{options}} form is used, @var{options} is a list of
5347 @samp{-} separated options that control the details of the dump. Not
5348 all options are applicable to all dumps, those which are not
5349 meaningful will be ignored. The following options are available
5353 Print the address of each node. Usually this is not meaningful as it
5354 changes according to the environment and source file. Its primary use
5355 is for tying up a dump file with a debug environment.
5357 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
5358 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
5359 use working backward from mangled names in the assembly file.
5361 Inhibit dumping of members of a scope or body of a function merely
5362 because that scope has been reached. Only dump such items when they
5363 are directly reachable by some other path. When dumping pretty-printed
5364 trees, this option inhibits dumping the bodies of control structures.
5366 Print a raw representation of the tree. By default, trees are
5367 pretty-printed into a C-like representation.
5369 Enable more detailed dumps (not honored by every dump option).
5371 Enable dumping various statistics about the pass (not honored by every dump
5374 Enable showing basic block boundaries (disabled in raw dumps).
5376 Enable showing virtual operands for every statement.
5378 Enable showing line numbers for statements.
5380 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5382 Enable showing the tree dump for each statement.
5384 Enable showing the EH region number holding each statement.
5386 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5387 and @option{lineno}.
5390 The following tree dumps are possible:
5394 @opindex fdump-tree-original
5395 Dump before any tree based optimization, to @file{@var{file}.original}.
5398 @opindex fdump-tree-optimized
5399 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5402 @opindex fdump-tree-gimple
5403 Dump each function before and after the gimplification pass to a file. The
5404 file name is made by appending @file{.gimple} to the source file name.
5407 @opindex fdump-tree-cfg
5408 Dump the control flow graph of each function to a file. The file name is
5409 made by appending @file{.cfg} to the source file name.
5412 @opindex fdump-tree-vcg
5413 Dump the control flow graph of each function to a file in VCG format. The
5414 file name is made by appending @file{.vcg} to the source file name. Note
5415 that if the file contains more than one function, the generated file cannot
5416 be used directly by VCG@. You will need to cut and paste each function's
5417 graph into its own separate file first.
5420 @opindex fdump-tree-ch
5421 Dump each function after copying loop headers. The file name is made by
5422 appending @file{.ch} to the source file name.
5425 @opindex fdump-tree-ssa
5426 Dump SSA related information to a file. The file name is made by appending
5427 @file{.ssa} to the source file name.
5430 @opindex fdump-tree-alias
5431 Dump aliasing information for each function. The file name is made by
5432 appending @file{.alias} to the source file name.
5435 @opindex fdump-tree-ccp
5436 Dump each function after CCP@. The file name is made by appending
5437 @file{.ccp} to the source file name.
5440 @opindex fdump-tree-storeccp
5441 Dump each function after STORE-CCP@. The file name is made by appending
5442 @file{.storeccp} to the source file name.
5445 @opindex fdump-tree-pre
5446 Dump trees after partial redundancy elimination. The file name is made
5447 by appending @file{.pre} to the source file name.
5450 @opindex fdump-tree-fre
5451 Dump trees after full redundancy elimination. The file name is made
5452 by appending @file{.fre} to the source file name.
5455 @opindex fdump-tree-copyprop
5456 Dump trees after copy propagation. The file name is made
5457 by appending @file{.copyprop} to the source file name.
5459 @item store_copyprop
5460 @opindex fdump-tree-store_copyprop
5461 Dump trees after store copy-propagation. The file name is made
5462 by appending @file{.store_copyprop} to the source file name.
5465 @opindex fdump-tree-dce
5466 Dump each function after dead code elimination. The file name is made by
5467 appending @file{.dce} to the source file name.
5470 @opindex fdump-tree-mudflap
5471 Dump each function after adding mudflap instrumentation. The file name is
5472 made by appending @file{.mudflap} to the source file name.
5475 @opindex fdump-tree-sra
5476 Dump each function after performing scalar replacement of aggregates. The
5477 file name is made by appending @file{.sra} to the source file name.
5480 @opindex fdump-tree-sink
5481 Dump each function after performing code sinking. The file name is made
5482 by appending @file{.sink} to the source file name.
5485 @opindex fdump-tree-dom
5486 Dump each function after applying dominator tree optimizations. The file
5487 name is made by appending @file{.dom} to the source file name.
5490 @opindex fdump-tree-dse
5491 Dump each function after applying dead store elimination. The file
5492 name is made by appending @file{.dse} to the source file name.
5495 @opindex fdump-tree-phiopt
5496 Dump each function after optimizing PHI nodes into straightline code. The file
5497 name is made by appending @file{.phiopt} to the source file name.
5500 @opindex fdump-tree-forwprop
5501 Dump each function after forward propagating single use variables. The file
5502 name is made by appending @file{.forwprop} to the source file name.
5505 @opindex fdump-tree-copyrename
5506 Dump each function after applying the copy rename optimization. The file
5507 name is made by appending @file{.copyrename} to the source file name.
5510 @opindex fdump-tree-nrv
5511 Dump each function after applying the named return value optimization on
5512 generic trees. The file name is made by appending @file{.nrv} to the source
5516 @opindex fdump-tree-vect
5517 Dump each function after applying vectorization of loops. The file name is
5518 made by appending @file{.vect} to the source file name.
5521 @opindex fdump-tree-slp
5522 Dump each function after applying vectorization of basic blocks. The file name
5523 is made by appending @file{.slp} to the source file name.
5526 @opindex fdump-tree-vrp
5527 Dump each function after Value Range Propagation (VRP). The file name
5528 is made by appending @file{.vrp} to the source file name.
5531 @opindex fdump-tree-all
5532 Enable all the available tree dumps with the flags provided in this option.
5535 @item -ftree-vectorizer-verbose=@var{n}
5536 @opindex ftree-vectorizer-verbose
5537 This option controls the amount of debugging output the vectorizer prints.
5538 This information is written to standard error, unless
5539 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5540 in which case it is output to the usual dump listing file, @file{.vect}.
5541 For @var{n}=0 no diagnostic information is reported.
5542 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5543 and the total number of loops that got vectorized.
5544 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5545 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5546 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5547 level that @option{-fdump-tree-vect-stats} uses.
5548 Higher verbosity levels mean either more information dumped for each
5549 reported loop, or same amount of information reported for more loops:
5550 if @var{n}=3, vectorizer cost model information is reported.
5551 If @var{n}=4, alignment related information is added to the reports.
5552 If @var{n}=5, data-references related information (e.g.@: memory dependences,
5553 memory access-patterns) is added to the reports.
5554 If @var{n}=6, the vectorizer reports also non-vectorized inner-most loops
5555 that did not pass the first analysis phase (i.e., may not be countable, or
5556 may have complicated control-flow).
5557 If @var{n}=7, the vectorizer reports also non-vectorized nested loops.
5558 If @var{n}=8, SLP related information is added to the reports.
5559 For @var{n}=9, all the information the vectorizer generates during its
5560 analysis and transformation is reported. This is the same verbosity level
5561 that @option{-fdump-tree-vect-details} uses.
5563 @item -frandom-seed=@var{string}
5564 @opindex frandom-seed
5565 This option provides a seed that GCC uses when it would otherwise use
5566 random numbers. It is used to generate certain symbol names
5567 that have to be different in every compiled file. It is also used to
5568 place unique stamps in coverage data files and the object files that
5569 produce them. You can use the @option{-frandom-seed} option to produce
5570 reproducibly identical object files.
5572 The @var{string} should be different for every file you compile.
5574 @item -fsched-verbose=@var{n}
5575 @opindex fsched-verbose
5576 On targets that use instruction scheduling, this option controls the
5577 amount of debugging output the scheduler prints. This information is
5578 written to standard error, unless @option{-fdump-rtl-sched1} or
5579 @option{-fdump-rtl-sched2} is specified, in which case it is output
5580 to the usual dump listing file, @file{.sched1} or @file{.sched2}
5581 respectively. However for @var{n} greater than nine, the output is
5582 always printed to standard error.
5584 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5585 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5586 For @var{n} greater than one, it also output basic block probabilities,
5587 detailed ready list information and unit/insn info. For @var{n} greater
5588 than two, it includes RTL at abort point, control-flow and regions info.
5589 And for @var{n} over four, @option{-fsched-verbose} also includes
5593 @itemx -save-temps=cwd
5595 Store the usual ``temporary'' intermediate files permanently; place them
5596 in the current directory and name them based on the source file. Thus,
5597 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5598 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5599 preprocessed @file{foo.i} output file even though the compiler now
5600 normally uses an integrated preprocessor.
5602 When used in combination with the @option{-x} command line option,
5603 @option{-save-temps} is sensible enough to avoid over writing an
5604 input source file with the same extension as an intermediate file.
5605 The corresponding intermediate file may be obtained by renaming the
5606 source file before using @option{-save-temps}.
5608 If you invoke GCC in parallel, compiling several different source
5609 files that share a common base name in different subdirectories or the
5610 same source file compiled for multiple output destinations, it is
5611 likely that the different parallel compilers will interfere with each
5612 other, and overwrite the temporary files. For instance:
5615 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5616 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5619 may result in @file{foo.i} and @file{foo.o} being written to
5620 simultaneously by both compilers.
5622 @item -save-temps=obj
5623 @opindex save-temps=obj
5624 Store the usual ``temporary'' intermediate files permanently. If the
5625 @option{-o} option is used, the temporary files are based on the
5626 object file. If the @option{-o} option is not used, the
5627 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5632 gcc -save-temps=obj -c foo.c
5633 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5634 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5637 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5638 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5639 @file{dir2/yfoobar.o}.
5641 @item -time@r{[}=@var{file}@r{]}
5643 Report the CPU time taken by each subprocess in the compilation
5644 sequence. For C source files, this is the compiler proper and assembler
5645 (plus the linker if linking is done).
5647 Without the specification of an output file, the output looks like this:
5654 The first number on each line is the ``user time'', that is time spent
5655 executing the program itself. The second number is ``system time'',
5656 time spent executing operating system routines on behalf of the program.
5657 Both numbers are in seconds.
5659 With the specification of an output file, the output is appended to the
5660 named file, and it looks like this:
5663 0.12 0.01 cc1 @var{options}
5664 0.00 0.01 as @var{options}
5667 The ``user time'' and the ``system time'' are moved before the program
5668 name, and the options passed to the program are displayed, so that one
5669 can later tell what file was being compiled, and with which options.
5671 @item -fvar-tracking
5672 @opindex fvar-tracking
5673 Run variable tracking pass. It computes where variables are stored at each
5674 position in code. Better debugging information is then generated
5675 (if the debugging information format supports this information).
5677 It is enabled by default when compiling with optimization (@option{-Os},
5678 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5679 the debug info format supports it.
5681 @item -fvar-tracking-assignments
5682 @opindex fvar-tracking-assignments
5683 @opindex fno-var-tracking-assignments
5684 Annotate assignments to user variables early in the compilation and
5685 attempt to carry the annotations over throughout the compilation all the
5686 way to the end, in an attempt to improve debug information while
5687 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
5689 It can be enabled even if var-tracking is disabled, in which case
5690 annotations will be created and maintained, but discarded at the end.
5692 @item -fvar-tracking-assignments-toggle
5693 @opindex fvar-tracking-assignments-toggle
5694 @opindex fno-var-tracking-assignments-toggle
5695 Toggle @option{-fvar-tracking-assignments}, in the same way that
5696 @option{-gtoggle} toggles @option{-g}.
5698 @item -print-file-name=@var{library}
5699 @opindex print-file-name
5700 Print the full absolute name of the library file @var{library} that
5701 would be used when linking---and don't do anything else. With this
5702 option, GCC does not compile or link anything; it just prints the
5705 @item -print-multi-directory
5706 @opindex print-multi-directory
5707 Print the directory name corresponding to the multilib selected by any
5708 other switches present in the command line. This directory is supposed
5709 to exist in @env{GCC_EXEC_PREFIX}.
5711 @item -print-multi-lib
5712 @opindex print-multi-lib
5713 Print the mapping from multilib directory names to compiler switches
5714 that enable them. The directory name is separated from the switches by
5715 @samp{;}, and each switch starts with an @samp{@@} instead of the
5716 @samp{-}, without spaces between multiple switches. This is supposed to
5717 ease shell-processing.
5719 @item -print-multi-os-directory
5720 @opindex print-multi-os-directory
5721 Print the path to OS libraries for the selected
5722 multilib, relative to some @file{lib} subdirectory. If OS libraries are
5723 present in the @file{lib} subdirectory and no multilibs are used, this is
5724 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
5725 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
5726 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
5727 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
5729 @item -print-prog-name=@var{program}
5730 @opindex print-prog-name
5731 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5733 @item -print-libgcc-file-name
5734 @opindex print-libgcc-file-name
5735 Same as @option{-print-file-name=libgcc.a}.
5737 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5738 but you do want to link with @file{libgcc.a}. You can do
5741 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5744 @item -print-search-dirs
5745 @opindex print-search-dirs
5746 Print the name of the configured installation directory and a list of
5747 program and library directories @command{gcc} will search---and don't do anything else.
5749 This is useful when @command{gcc} prints the error message
5750 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5751 To resolve this you either need to put @file{cpp0} and the other compiler
5752 components where @command{gcc} expects to find them, or you can set the environment
5753 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5754 Don't forget the trailing @samp{/}.
5755 @xref{Environment Variables}.
5757 @item -print-sysroot
5758 @opindex print-sysroot
5759 Print the target sysroot directory that will be used during
5760 compilation. This is the target sysroot specified either at configure
5761 time or using the @option{--sysroot} option, possibly with an extra
5762 suffix that depends on compilation options. If no target sysroot is
5763 specified, the option prints nothing.
5765 @item -print-sysroot-headers-suffix
5766 @opindex print-sysroot-headers-suffix
5767 Print the suffix added to the target sysroot when searching for
5768 headers, or give an error if the compiler is not configured with such
5769 a suffix---and don't do anything else.
5772 @opindex dumpmachine
5773 Print the compiler's target machine (for example,
5774 @samp{i686-pc-linux-gnu})---and don't do anything else.
5777 @opindex dumpversion
5778 Print the compiler version (for example, @samp{3.0})---and don't do
5783 Print the compiler's built-in specs---and don't do anything else. (This
5784 is used when GCC itself is being built.) @xref{Spec Files}.
5786 @item -feliminate-unused-debug-types
5787 @opindex feliminate-unused-debug-types
5788 Normally, when producing DWARF2 output, GCC will emit debugging
5789 information for all types declared in a compilation
5790 unit, regardless of whether or not they are actually used
5791 in that compilation unit. Sometimes this is useful, such as
5792 if, in the debugger, you want to cast a value to a type that is
5793 not actually used in your program (but is declared). More often,
5794 however, this results in a significant amount of wasted space.
5795 With this option, GCC will avoid producing debug symbol output
5796 for types that are nowhere used in the source file being compiled.
5799 @node Optimize Options
5800 @section Options That Control Optimization
5801 @cindex optimize options
5802 @cindex options, optimization
5804 These options control various sorts of optimizations.
5806 Without any optimization option, the compiler's goal is to reduce the
5807 cost of compilation and to make debugging produce the expected
5808 results. Statements are independent: if you stop the program with a
5809 breakpoint between statements, you can then assign a new value to any
5810 variable or change the program counter to any other statement in the
5811 function and get exactly the results you would expect from the source
5814 Turning on optimization flags makes the compiler attempt to improve
5815 the performance and/or code size at the expense of compilation time
5816 and possibly the ability to debug the program.
5818 The compiler performs optimization based on the knowledge it has of the
5819 program. Compiling multiple files at once to a single output file mode allows
5820 the compiler to use information gained from all of the files when compiling
5823 Not all optimizations are controlled directly by a flag. Only
5824 optimizations that have a flag are listed in this section.
5826 Most optimizations are only enabled if an @option{-O} level is set on
5827 the command line. Otherwise they are disabled, even if individual
5828 optimization flags are specified.
5830 Depending on the target and how GCC was configured, a slightly different
5831 set of optimizations may be enabled at each @option{-O} level than
5832 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
5833 to find out the exact set of optimizations that are enabled at each level.
5834 @xref{Overall Options}, for examples.
5841 Optimize. Optimizing compilation takes somewhat more time, and a lot
5842 more memory for a large function.
5844 With @option{-O}, the compiler tries to reduce code size and execution
5845 time, without performing any optimizations that take a great deal of
5848 @option{-O} turns on the following optimization flags:
5851 -fcprop-registers @gol
5854 -fdelayed-branch @gol
5856 -fguess-branch-probability @gol
5857 -fif-conversion2 @gol
5858 -fif-conversion @gol
5859 -fipa-pure-const @gol
5861 -fipa-reference @gol
5863 -fsplit-wide-types @gol
5865 -ftree-builtin-call-dce @gol
5868 -ftree-copyrename @gol
5870 -ftree-dominator-opts @gol
5872 -ftree-forwprop @gol
5880 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5881 where doing so does not interfere with debugging.
5885 Optimize even more. GCC performs nearly all supported optimizations
5886 that do not involve a space-speed tradeoff.
5887 As compared to @option{-O}, this option increases both compilation time
5888 and the performance of the generated code.
5890 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5891 also turns on the following optimization flags:
5892 @gccoptlist{-fthread-jumps @gol
5893 -falign-functions -falign-jumps @gol
5894 -falign-loops -falign-labels @gol
5897 -fcse-follow-jumps -fcse-skip-blocks @gol
5898 -fdelete-null-pointer-checks @gol
5899 -fexpensive-optimizations @gol
5900 -fgcse -fgcse-lm @gol
5901 -finline-small-functions @gol
5902 -findirect-inlining @gol
5904 -foptimize-sibling-calls @gol
5905 -fpartial-inlining @gol
5908 -freorder-blocks -freorder-functions @gol
5909 -frerun-cse-after-loop @gol
5910 -fsched-interblock -fsched-spec @gol
5911 -fschedule-insns -fschedule-insns2 @gol
5912 -fstrict-aliasing -fstrict-overflow @gol
5913 -ftree-switch-conversion @gol
5917 Please note the warning under @option{-fgcse} about
5918 invoking @option{-O2} on programs that use computed gotos.
5922 Optimize yet more. @option{-O3} turns on all optimizations specified
5923 by @option{-O2} and also turns on the @option{-finline-functions},
5924 @option{-funswitch-loops}, @option{-fpredictive-commoning},
5925 @option{-fgcse-after-reload}, @option{-ftree-vectorize} and
5926 @option{-fipa-cp-clone} options.
5930 Reduce compilation time and make debugging produce the expected
5931 results. This is the default.
5935 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5936 do not typically increase code size. It also performs further
5937 optimizations designed to reduce code size.
5939 @option{-Os} disables the following optimization flags:
5940 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5941 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5942 -fprefetch-loop-arrays -ftree-vect-loop-version}
5946 Disregard strict standards compliance. @option{-Ofast} enables all
5947 @option{-O3} optimizations. It also enables optimizations that are not
5948 valid for all standard compliant programs.
5949 It turns on @option{-ffast-math}.
5951 If you use multiple @option{-O} options, with or without level numbers,
5952 the last such option is the one that is effective.
5955 Options of the form @option{-f@var{flag}} specify machine-independent
5956 flags. Most flags have both positive and negative forms; the negative
5957 form of @option{-ffoo} would be @option{-fno-foo}. In the table
5958 below, only one of the forms is listed---the one you typically will
5959 use. You can figure out the other form by either removing @samp{no-}
5962 The following options control specific optimizations. They are either
5963 activated by @option{-O} options or are related to ones that are. You
5964 can use the following flags in the rare cases when ``fine-tuning'' of
5965 optimizations to be performed is desired.
5968 @item -fno-default-inline
5969 @opindex fno-default-inline
5970 Do not make member functions inline by default merely because they are
5971 defined inside the class scope (C++ only). Otherwise, when you specify
5972 @w{@option{-O}}, member functions defined inside class scope are compiled
5973 inline by default; i.e., you don't need to add @samp{inline} in front of
5974 the member function name.
5976 @item -fno-defer-pop
5977 @opindex fno-defer-pop
5978 Always pop the arguments to each function call as soon as that function
5979 returns. For machines which must pop arguments after a function call,
5980 the compiler normally lets arguments accumulate on the stack for several
5981 function calls and pops them all at once.
5983 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5985 @item -fforward-propagate
5986 @opindex fforward-propagate
5987 Perform a forward propagation pass on RTL@. The pass tries to combine two
5988 instructions and checks if the result can be simplified. If loop unrolling
5989 is active, two passes are performed and the second is scheduled after
5992 This option is enabled by default at optimization levels @option{-O},
5993 @option{-O2}, @option{-O3}, @option{-Os}.
5995 @item -ffp-contract=@var{style}
5996 @opindex ffp-contract
5997 @option{-ffp-contract=off} disables floating-point expression contraction.
5998 @option{-ffp-contract=fast} enables floating-point expression contraction
5999 such as forming of fused multiply-add operations if the target has
6000 native support for them.
6001 @option{-ffp-contract=on} enables floating-point expression contraction
6002 if allowed by the language standard. This is currently not implemented
6003 and treated equal to @option{-ffp-contract=off}.
6005 The default is @option{-ffp-contract=fast}.
6007 @item -fomit-frame-pointer
6008 @opindex fomit-frame-pointer
6009 Don't keep the frame pointer in a register for functions that
6010 don't need one. This avoids the instructions to save, set up and
6011 restore frame pointers; it also makes an extra register available
6012 in many functions. @strong{It also makes debugging impossible on
6015 On some machines, such as the VAX, this flag has no effect, because
6016 the standard calling sequence automatically handles the frame pointer
6017 and nothing is saved by pretending it doesn't exist. The
6018 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
6019 whether a target machine supports this flag. @xref{Registers,,Register
6020 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
6022 Starting with GCC version 4.6, the default setting (when not optimizing for
6023 size) for 32-bit Linux x86 and 32-bit Darwin x86 targets has been changed to
6024 @option{-fomit-frame-pointer}. The default can be reverted to
6025 @option{-fno-omit-frame-pointer} by configuring GCC with the
6026 @option{--enable-frame-pointer} configure option.
6028 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6030 @item -foptimize-sibling-calls
6031 @opindex foptimize-sibling-calls
6032 Optimize sibling and tail recursive calls.
6034 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6038 Don't pay attention to the @code{inline} keyword. Normally this option
6039 is used to keep the compiler from expanding any functions inline.
6040 Note that if you are not optimizing, no functions can be expanded inline.
6042 @item -finline-small-functions
6043 @opindex finline-small-functions
6044 Integrate functions into their callers when their body is smaller than expected
6045 function call code (so overall size of program gets smaller). The compiler
6046 heuristically decides which functions are simple enough to be worth integrating
6049 Enabled at level @option{-O2}.
6051 @item -findirect-inlining
6052 @opindex findirect-inlining
6053 Inline also indirect calls that are discovered to be known at compile
6054 time thanks to previous inlining. This option has any effect only
6055 when inlining itself is turned on by the @option{-finline-functions}
6056 or @option{-finline-small-functions} options.
6058 Enabled at level @option{-O2}.
6060 @item -finline-functions
6061 @opindex finline-functions
6062 Integrate all simple functions into their callers. The compiler
6063 heuristically decides which functions are simple enough to be worth
6064 integrating in this way.
6066 If all calls to a given function are integrated, and the function is
6067 declared @code{static}, then the function is normally not output as
6068 assembler code in its own right.
6070 Enabled at level @option{-O3}.
6072 @item -finline-functions-called-once
6073 @opindex finline-functions-called-once
6074 Consider all @code{static} functions called once for inlining into their
6075 caller even if they are not marked @code{inline}. If a call to a given
6076 function is integrated, then the function is not output as assembler code
6079 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
6081 @item -fearly-inlining
6082 @opindex fearly-inlining
6083 Inline functions marked by @code{always_inline} and functions whose body seems
6084 smaller than the function call overhead early before doing
6085 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
6086 makes profiling significantly cheaper and usually inlining faster on programs
6087 having large chains of nested wrapper functions.
6093 Perform interprocedural scalar replacement of aggregates, removal of
6094 unused parameters and replacement of parameters passed by reference
6095 by parameters passed by value.
6097 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
6099 @item -finline-limit=@var{n}
6100 @opindex finline-limit
6101 By default, GCC limits the size of functions that can be inlined. This flag
6102 allows coarse control of this limit. @var{n} is the size of functions that
6103 can be inlined in number of pseudo instructions.
6105 Inlining is actually controlled by a number of parameters, which may be
6106 specified individually by using @option{--param @var{name}=@var{value}}.
6107 The @option{-finline-limit=@var{n}} option sets some of these parameters
6111 @item max-inline-insns-single
6112 is set to @var{n}/2.
6113 @item max-inline-insns-auto
6114 is set to @var{n}/2.
6117 See below for a documentation of the individual
6118 parameters controlling inlining and for the defaults of these parameters.
6120 @emph{Note:} there may be no value to @option{-finline-limit} that results
6121 in default behavior.
6123 @emph{Note:} pseudo instruction represents, in this particular context, an
6124 abstract measurement of function's size. In no way does it represent a count
6125 of assembly instructions and as such its exact meaning might change from one
6126 release to an another.
6128 @item -fkeep-inline-functions
6129 @opindex fkeep-inline-functions
6130 In C, emit @code{static} functions that are declared @code{inline}
6131 into the object file, even if the function has been inlined into all
6132 of its callers. This switch does not affect functions using the
6133 @code{extern inline} extension in GNU C90@. In C++, emit any and all
6134 inline functions into the object file.
6136 @item -fkeep-static-consts
6137 @opindex fkeep-static-consts
6138 Emit variables declared @code{static const} when optimization isn't turned
6139 on, even if the variables aren't referenced.
6141 GCC enables this option by default. If you want to force the compiler to
6142 check if the variable was referenced, regardless of whether or not
6143 optimization is turned on, use the @option{-fno-keep-static-consts} option.
6145 @item -fmerge-constants
6146 @opindex fmerge-constants
6147 Attempt to merge identical constants (string constants and floating point
6148 constants) across compilation units.
6150 This option is the default for optimized compilation if the assembler and
6151 linker support it. Use @option{-fno-merge-constants} to inhibit this
6154 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6156 @item -fmerge-all-constants
6157 @opindex fmerge-all-constants
6158 Attempt to merge identical constants and identical variables.
6160 This option implies @option{-fmerge-constants}. In addition to
6161 @option{-fmerge-constants} this considers e.g.@: even constant initialized
6162 arrays or initialized constant variables with integral or floating point
6163 types. Languages like C or C++ require each variable, including multiple
6164 instances of the same variable in recursive calls, to have distinct locations,
6165 so using this option will result in non-conforming
6168 @item -fmodulo-sched
6169 @opindex fmodulo-sched
6170 Perform swing modulo scheduling immediately before the first scheduling
6171 pass. This pass looks at innermost loops and reorders their
6172 instructions by overlapping different iterations.
6174 @item -fmodulo-sched-allow-regmoves
6175 @opindex fmodulo-sched-allow-regmoves
6176 Perform more aggressive SMS based modulo scheduling with register moves
6177 allowed. By setting this flag certain anti-dependences edges will be
6178 deleted which will trigger the generation of reg-moves based on the
6179 life-range analysis. This option is effective only with
6180 @option{-fmodulo-sched} enabled.
6182 @item -fno-branch-count-reg
6183 @opindex fno-branch-count-reg
6184 Do not use ``decrement and branch'' instructions on a count register,
6185 but instead generate a sequence of instructions that decrement a
6186 register, compare it against zero, then branch based upon the result.
6187 This option is only meaningful on architectures that support such
6188 instructions, which include x86, PowerPC, IA-64 and S/390.
6190 The default is @option{-fbranch-count-reg}.
6192 @item -fno-function-cse
6193 @opindex fno-function-cse
6194 Do not put function addresses in registers; make each instruction that
6195 calls a constant function contain the function's address explicitly.
6197 This option results in less efficient code, but some strange hacks
6198 that alter the assembler output may be confused by the optimizations
6199 performed when this option is not used.
6201 The default is @option{-ffunction-cse}
6203 @item -fno-zero-initialized-in-bss
6204 @opindex fno-zero-initialized-in-bss
6205 If the target supports a BSS section, GCC by default puts variables that
6206 are initialized to zero into BSS@. This can save space in the resulting
6209 This option turns off this behavior because some programs explicitly
6210 rely on variables going to the data section. E.g., so that the
6211 resulting executable can find the beginning of that section and/or make
6212 assumptions based on that.
6214 The default is @option{-fzero-initialized-in-bss}.
6216 @item -fmudflap -fmudflapth -fmudflapir
6220 @cindex bounds checking
6222 For front-ends that support it (C and C++), instrument all risky
6223 pointer/array dereferencing operations, some standard library
6224 string/heap functions, and some other associated constructs with
6225 range/validity tests. Modules so instrumented should be immune to
6226 buffer overflows, invalid heap use, and some other classes of C/C++
6227 programming errors. The instrumentation relies on a separate runtime
6228 library (@file{libmudflap}), which will be linked into a program if
6229 @option{-fmudflap} is given at link time. Run-time behavior of the
6230 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
6231 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
6234 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
6235 link if your program is multi-threaded. Use @option{-fmudflapir}, in
6236 addition to @option{-fmudflap} or @option{-fmudflapth}, if
6237 instrumentation should ignore pointer reads. This produces less
6238 instrumentation (and therefore faster execution) and still provides
6239 some protection against outright memory corrupting writes, but allows
6240 erroneously read data to propagate within a program.
6242 @item -fthread-jumps
6243 @opindex fthread-jumps
6244 Perform optimizations where we check to see if a jump branches to a
6245 location where another comparison subsumed by the first is found. If
6246 so, the first branch is redirected to either the destination of the
6247 second branch or a point immediately following it, depending on whether
6248 the condition is known to be true or false.
6250 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6252 @item -fsplit-wide-types
6253 @opindex fsplit-wide-types
6254 When using a type that occupies multiple registers, such as @code{long
6255 long} on a 32-bit system, split the registers apart and allocate them
6256 independently. This normally generates better code for those types,
6257 but may make debugging more difficult.
6259 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
6262 @item -fcse-follow-jumps
6263 @opindex fcse-follow-jumps
6264 In common subexpression elimination (CSE), scan through jump instructions
6265 when the target of the jump is not reached by any other path. For
6266 example, when CSE encounters an @code{if} statement with an
6267 @code{else} clause, CSE will follow the jump when the condition
6270 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6272 @item -fcse-skip-blocks
6273 @opindex fcse-skip-blocks
6274 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
6275 follow jumps which conditionally skip over blocks. When CSE
6276 encounters a simple @code{if} statement with no else clause,
6277 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6278 body of the @code{if}.
6280 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6282 @item -frerun-cse-after-loop
6283 @opindex frerun-cse-after-loop
6284 Re-run common subexpression elimination after loop optimizations has been
6287 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6291 Perform a global common subexpression elimination pass.
6292 This pass also performs global constant and copy propagation.
6294 @emph{Note:} When compiling a program using computed gotos, a GCC
6295 extension, you may get better runtime performance if you disable
6296 the global common subexpression elimination pass by adding
6297 @option{-fno-gcse} to the command line.
6299 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6303 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
6304 attempt to move loads which are only killed by stores into themselves. This
6305 allows a loop containing a load/store sequence to be changed to a load outside
6306 the loop, and a copy/store within the loop.
6308 Enabled by default when gcse is enabled.
6312 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6313 global common subexpression elimination. This pass will attempt to move
6314 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6315 loops containing a load/store sequence can be changed to a load before
6316 the loop and a store after the loop.
6318 Not enabled at any optimization level.
6322 When @option{-fgcse-las} is enabled, the global common subexpression
6323 elimination pass eliminates redundant loads that come after stores to the
6324 same memory location (both partial and full redundancies).
6326 Not enabled at any optimization level.
6328 @item -fgcse-after-reload
6329 @opindex fgcse-after-reload
6330 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6331 pass is performed after reload. The purpose of this pass is to cleanup
6334 @item -funsafe-loop-optimizations
6335 @opindex funsafe-loop-optimizations
6336 If given, the loop optimizer will assume that loop indices do not
6337 overflow, and that the loops with nontrivial exit condition are not
6338 infinite. This enables a wider range of loop optimizations even if
6339 the loop optimizer itself cannot prove that these assumptions are valid.
6340 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
6341 if it finds this kind of loop.
6343 @item -fcrossjumping
6344 @opindex fcrossjumping
6345 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6346 resulting code may or may not perform better than without cross-jumping.
6348 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6350 @item -fauto-inc-dec
6351 @opindex fauto-inc-dec
6352 Combine increments or decrements of addresses with memory accesses.
6353 This pass is always skipped on architectures that do not have
6354 instructions to support this. Enabled by default at @option{-O} and
6355 higher on architectures that support this.
6359 Perform dead code elimination (DCE) on RTL@.
6360 Enabled by default at @option{-O} and higher.
6364 Perform dead store elimination (DSE) on RTL@.
6365 Enabled by default at @option{-O} and higher.
6367 @item -fif-conversion
6368 @opindex fif-conversion
6369 Attempt to transform conditional jumps into branch-less equivalents. This
6370 include use of conditional moves, min, max, set flags and abs instructions, and
6371 some tricks doable by standard arithmetics. The use of conditional execution
6372 on chips where it is available is controlled by @code{if-conversion2}.
6374 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6376 @item -fif-conversion2
6377 @opindex fif-conversion2
6378 Use conditional execution (where available) to transform conditional jumps into
6379 branch-less equivalents.
6381 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6383 @item -fdelete-null-pointer-checks
6384 @opindex fdelete-null-pointer-checks
6385 Assume that programs cannot safely dereference null pointers, and that
6386 no code or data element resides there. This enables simple constant
6387 folding optimizations at all optimization levels. In addition, other
6388 optimization passes in GCC use this flag to control global dataflow
6389 analyses that eliminate useless checks for null pointers; these assume
6390 that if a pointer is checked after it has already been dereferenced,
6393 Note however that in some environments this assumption is not true.
6394 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6395 for programs which depend on that behavior.
6397 Some targets, especially embedded ones, disable this option at all levels.
6398 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6399 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6400 are enabled independently at different optimization levels.
6402 @item -fexpensive-optimizations
6403 @opindex fexpensive-optimizations
6404 Perform a number of minor optimizations that are relatively expensive.
6406 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6408 @item -foptimize-register-move
6410 @opindex foptimize-register-move
6412 Attempt to reassign register numbers in move instructions and as
6413 operands of other simple instructions in order to maximize the amount of
6414 register tying. This is especially helpful on machines with two-operand
6417 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6420 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6422 @item -fira-algorithm=@var{algorithm}
6423 Use specified coloring algorithm for the integrated register
6424 allocator. The @var{algorithm} argument should be @code{priority} or
6425 @code{CB}. The first algorithm specifies Chow's priority coloring,
6426 the second one specifies Chaitin-Briggs coloring. The second
6427 algorithm can be unimplemented for some architectures. If it is
6428 implemented, it is the default because Chaitin-Briggs coloring as a
6429 rule generates a better code.
6431 @item -fira-region=@var{region}
6432 Use specified regions for the integrated register allocator. The
6433 @var{region} argument should be one of @code{all}, @code{mixed}, or
6434 @code{one}. The first value means using all loops as register
6435 allocation regions, the second value which is the default means using
6436 all loops except for loops with small register pressure as the
6437 regions, and third one means using all function as a single region.
6438 The first value can give best result for machines with small size and
6439 irregular register set, the third one results in faster and generates
6440 decent code and the smallest size code, and the default value usually
6441 give the best results in most cases and for most architectures.
6443 @item -fira-loop-pressure
6444 @opindex fira-loop-pressure
6445 Use IRA to evaluate register pressure in loops for decision to move
6446 loop invariants. Usage of this option usually results in generation
6447 of faster and smaller code on machines with big register files (>= 32
6448 registers) but it can slow compiler down.
6450 This option is enabled at level @option{-O3} for some targets.
6452 @item -fno-ira-share-save-slots
6453 @opindex fno-ira-share-save-slots
6454 Switch off sharing stack slots used for saving call used hard
6455 registers living through a call. Each hard register will get a
6456 separate stack slot and as a result function stack frame will be
6459 @item -fno-ira-share-spill-slots
6460 @opindex fno-ira-share-spill-slots
6461 Switch off sharing stack slots allocated for pseudo-registers. Each
6462 pseudo-register which did not get a hard register will get a separate
6463 stack slot and as a result function stack frame will be bigger.
6465 @item -fira-verbose=@var{n}
6466 @opindex fira-verbose
6467 Set up how verbose dump file for the integrated register allocator
6468 will be. Default value is 5. If the value is greater or equal to 10,
6469 the dump file will be stderr as if the value were @var{n} minus 10.
6471 @item -fdelayed-branch
6472 @opindex fdelayed-branch
6473 If supported for the target machine, attempt to reorder instructions
6474 to exploit instruction slots available after delayed branch
6477 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6479 @item -fschedule-insns
6480 @opindex fschedule-insns
6481 If supported for the target machine, attempt to reorder instructions to
6482 eliminate execution stalls due to required data being unavailable. This
6483 helps machines that have slow floating point or memory load instructions
6484 by allowing other instructions to be issued until the result of the load
6485 or floating point instruction is required.
6487 Enabled at levels @option{-O2}, @option{-O3}.
6489 @item -fschedule-insns2
6490 @opindex fschedule-insns2
6491 Similar to @option{-fschedule-insns}, but requests an additional pass of
6492 instruction scheduling after register allocation has been done. This is
6493 especially useful on machines with a relatively small number of
6494 registers and where memory load instructions take more than one cycle.
6496 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6498 @item -fno-sched-interblock
6499 @opindex fno-sched-interblock
6500 Don't schedule instructions across basic blocks. This is normally
6501 enabled by default when scheduling before register allocation, i.e.@:
6502 with @option{-fschedule-insns} or at @option{-O2} or higher.
6504 @item -fno-sched-spec
6505 @opindex fno-sched-spec
6506 Don't allow speculative motion of non-load instructions. This is normally
6507 enabled by default when scheduling before register allocation, i.e.@:
6508 with @option{-fschedule-insns} or at @option{-O2} or higher.
6510 @item -fsched-pressure
6511 @opindex fsched-pressure
6512 Enable register pressure sensitive insn scheduling before the register
6513 allocation. This only makes sense when scheduling before register
6514 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
6515 @option{-O2} or higher. Usage of this option can improve the
6516 generated code and decrease its size by preventing register pressure
6517 increase above the number of available hard registers and as a
6518 consequence register spills in the register allocation.
6520 @item -fsched-spec-load
6521 @opindex fsched-spec-load
6522 Allow speculative motion of some load instructions. This only makes
6523 sense when scheduling before register allocation, i.e.@: with
6524 @option{-fschedule-insns} or at @option{-O2} or higher.
6526 @item -fsched-spec-load-dangerous
6527 @opindex fsched-spec-load-dangerous
6528 Allow speculative motion of more load instructions. This only makes
6529 sense when scheduling before register allocation, i.e.@: with
6530 @option{-fschedule-insns} or at @option{-O2} or higher.
6532 @item -fsched-stalled-insns
6533 @itemx -fsched-stalled-insns=@var{n}
6534 @opindex fsched-stalled-insns
6535 Define how many insns (if any) can be moved prematurely from the queue
6536 of stalled insns into the ready list, during the second scheduling pass.
6537 @option{-fno-sched-stalled-insns} means that no insns will be moved
6538 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6539 on how many queued insns can be moved prematurely.
6540 @option{-fsched-stalled-insns} without a value is equivalent to
6541 @option{-fsched-stalled-insns=1}.
6543 @item -fsched-stalled-insns-dep
6544 @itemx -fsched-stalled-insns-dep=@var{n}
6545 @opindex fsched-stalled-insns-dep
6546 Define how many insn groups (cycles) will be examined for a dependency
6547 on a stalled insn that is candidate for premature removal from the queue
6548 of stalled insns. This has an effect only during the second scheduling pass,
6549 and only if @option{-fsched-stalled-insns} is used.
6550 @option{-fno-sched-stalled-insns-dep} is equivalent to
6551 @option{-fsched-stalled-insns-dep=0}.
6552 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6553 @option{-fsched-stalled-insns-dep=1}.
6555 @item -fsched2-use-superblocks
6556 @opindex fsched2-use-superblocks
6557 When scheduling after register allocation, do use superblock scheduling
6558 algorithm. Superblock scheduling allows motion across basic block boundaries
6559 resulting on faster schedules. This option is experimental, as not all machine
6560 descriptions used by GCC model the CPU closely enough to avoid unreliable
6561 results from the algorithm.
6563 This only makes sense when scheduling after register allocation, i.e.@: with
6564 @option{-fschedule-insns2} or at @option{-O2} or higher.
6566 @item -fsched-group-heuristic
6567 @opindex fsched-group-heuristic
6568 Enable the group heuristic in the scheduler. This heuristic favors
6569 the instruction that belongs to a schedule group. This is enabled
6570 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6571 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6573 @item -fsched-critical-path-heuristic
6574 @opindex fsched-critical-path-heuristic
6575 Enable the critical-path heuristic in the scheduler. This heuristic favors
6576 instructions on the critical path. This is enabled by default when
6577 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6578 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6580 @item -fsched-spec-insn-heuristic
6581 @opindex fsched-spec-insn-heuristic
6582 Enable the speculative instruction heuristic in the scheduler. This
6583 heuristic favors speculative instructions with greater dependency weakness.
6584 This is enabled by default when scheduling is enabled, i.e.@:
6585 with @option{-fschedule-insns} or @option{-fschedule-insns2}
6586 or at @option{-O2} or higher.
6588 @item -fsched-rank-heuristic
6589 @opindex fsched-rank-heuristic
6590 Enable the rank heuristic in the scheduler. This heuristic favors
6591 the instruction belonging to a basic block with greater size or frequency.
6592 This is enabled by default when scheduling is enabled, i.e.@:
6593 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6594 at @option{-O2} or higher.
6596 @item -fsched-last-insn-heuristic
6597 @opindex fsched-last-insn-heuristic
6598 Enable the last-instruction heuristic in the scheduler. This heuristic
6599 favors the instruction that is less dependent on the last instruction
6600 scheduled. This is enabled by default when scheduling is enabled,
6601 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6602 at @option{-O2} or higher.
6604 @item -fsched-dep-count-heuristic
6605 @opindex fsched-dep-count-heuristic
6606 Enable the dependent-count heuristic in the scheduler. This heuristic
6607 favors the instruction that has more instructions depending on it.
6608 This is enabled by default when scheduling is enabled, i.e.@:
6609 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6610 at @option{-O2} or higher.
6612 @item -freschedule-modulo-scheduled-loops
6613 @opindex freschedule-modulo-scheduled-loops
6614 The modulo scheduling comes before the traditional scheduling, if a loop
6615 was modulo scheduled we may want to prevent the later scheduling passes
6616 from changing its schedule, we use this option to control that.
6618 @item -fselective-scheduling
6619 @opindex fselective-scheduling
6620 Schedule instructions using selective scheduling algorithm. Selective
6621 scheduling runs instead of the first scheduler pass.
6623 @item -fselective-scheduling2
6624 @opindex fselective-scheduling2
6625 Schedule instructions using selective scheduling algorithm. Selective
6626 scheduling runs instead of the second scheduler pass.
6628 @item -fsel-sched-pipelining
6629 @opindex fsel-sched-pipelining
6630 Enable software pipelining of innermost loops during selective scheduling.
6631 This option has no effect until one of @option{-fselective-scheduling} or
6632 @option{-fselective-scheduling2} is turned on.
6634 @item -fsel-sched-pipelining-outer-loops
6635 @opindex fsel-sched-pipelining-outer-loops
6636 When pipelining loops during selective scheduling, also pipeline outer loops.
6637 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6639 @item -fcaller-saves
6640 @opindex fcaller-saves
6641 Enable values to be allocated in registers that will be clobbered by
6642 function calls, by emitting extra instructions to save and restore the
6643 registers around such calls. Such allocation is done only when it
6644 seems to result in better code than would otherwise be produced.
6646 This option is always enabled by default on certain machines, usually
6647 those which have no call-preserved registers to use instead.
6649 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6651 @item -fcombine-stack-adjustments
6652 @opindex fcombine-stack-adjustments
6653 Tracks stack adjustments (pushes and pops) and stack memory references
6654 and then tries to find ways to combine them.
6656 Enabled by default at @option{-O1} and higher.
6658 @item -fconserve-stack
6659 @opindex fconserve-stack
6660 Attempt to minimize stack usage. The compiler will attempt to use less
6661 stack space, even if that makes the program slower. This option
6662 implies setting the @option{large-stack-frame} parameter to 100
6663 and the @option{large-stack-frame-growth} parameter to 400.
6665 @item -ftree-reassoc
6666 @opindex ftree-reassoc
6667 Perform reassociation on trees. This flag is enabled by default
6668 at @option{-O} and higher.
6672 Perform partial redundancy elimination (PRE) on trees. This flag is
6673 enabled by default at @option{-O2} and @option{-O3}.
6675 @item -ftree-forwprop
6676 @opindex ftree-forwprop
6677 Perform forward propagation on trees. This flag is enabled by default
6678 at @option{-O} and higher.
6682 Perform full redundancy elimination (FRE) on trees. The difference
6683 between FRE and PRE is that FRE only considers expressions
6684 that are computed on all paths leading to the redundant computation.
6685 This analysis is faster than PRE, though it exposes fewer redundancies.
6686 This flag is enabled by default at @option{-O} and higher.
6688 @item -ftree-phiprop
6689 @opindex ftree-phiprop
6690 Perform hoisting of loads from conditional pointers on trees. This
6691 pass is enabled by default at @option{-O} and higher.
6693 @item -ftree-copy-prop
6694 @opindex ftree-copy-prop
6695 Perform copy propagation on trees. This pass eliminates unnecessary
6696 copy operations. This flag is enabled by default at @option{-O} and
6699 @item -fipa-pure-const
6700 @opindex fipa-pure-const
6701 Discover which functions are pure or constant.
6702 Enabled by default at @option{-O} and higher.
6704 @item -fipa-reference
6705 @opindex fipa-reference
6706 Discover which static variables do not escape cannot escape the
6708 Enabled by default at @option{-O} and higher.
6710 @item -fipa-struct-reorg
6711 @opindex fipa-struct-reorg
6712 Perform structure reorganization optimization, that change C-like structures
6713 layout in order to better utilize spatial locality. This transformation is
6714 affective for programs containing arrays of structures. Available in two
6715 compilation modes: profile-based (enabled with @option{-fprofile-generate})
6716 or static (which uses built-in heuristics). It works only in whole program
6717 mode, so it requires @option{-fwhole-program} to be
6718 enabled. Structures considered @samp{cold} by this transformation are not
6719 affected (see @option{--param struct-reorg-cold-struct-ratio=@var{value}}).
6721 With this flag, the program debug info reflects a new structure layout.
6725 Perform interprocedural pointer analysis and interprocedural modification
6726 and reference analysis. This option can cause excessive memory and
6727 compile-time usage on large compilation units. It is not enabled by
6728 default at any optimization level.
6731 @opindex fipa-profile
6732 Perform interprocedural profile propagation. The functions called only from
6733 cold functions are marked as cold. Also functions executed once (such as
6734 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
6735 functions and loop less parts of functions executed once are then optimized for
6737 Enabled by default at @option{-O} and higher.
6741 Perform interprocedural constant propagation.
6742 This optimization analyzes the program to determine when values passed
6743 to functions are constants and then optimizes accordingly.
6744 This optimization can substantially increase performance
6745 if the application has constants passed to functions.
6746 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
6748 @item -fipa-cp-clone
6749 @opindex fipa-cp-clone
6750 Perform function cloning to make interprocedural constant propagation stronger.
6751 When enabled, interprocedural constant propagation will perform function cloning
6752 when externally visible function can be called with constant arguments.
6753 Because this optimization can create multiple copies of functions,
6754 it may significantly increase code size
6755 (see @option{--param ipcp-unit-growth=@var{value}}).
6756 This flag is enabled by default at @option{-O3}.
6758 @item -fipa-matrix-reorg
6759 @opindex fipa-matrix-reorg
6760 Perform matrix flattening and transposing.
6761 Matrix flattening tries to replace an @math{m}-dimensional matrix
6762 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
6763 This reduces the level of indirection needed for accessing the elements
6764 of the matrix. The second optimization is matrix transposing that
6765 attempts to change the order of the matrix's dimensions in order to
6766 improve cache locality.
6767 Both optimizations need the @option{-fwhole-program} flag.
6768 Transposing is enabled only if profiling information is available.
6772 Perform forward store motion on trees. This flag is
6773 enabled by default at @option{-O} and higher.
6775 @item -ftree-bit-ccp
6776 @opindex ftree-bit-ccp
6777 Perform sparse conditional bit constant propagation on trees and propagate
6778 pointer alignment information.
6779 This pass only operates on local scalar variables and is enabled by default
6780 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
6784 Perform sparse conditional constant propagation (CCP) on trees. This
6785 pass only operates on local scalar variables and is enabled by default
6786 at @option{-O} and higher.
6788 @item -ftree-switch-conversion
6789 Perform conversion of simple initializations in a switch to
6790 initializations from a scalar array. This flag is enabled by default
6791 at @option{-O2} and higher.
6795 Perform dead code elimination (DCE) on trees. This flag is enabled by
6796 default at @option{-O} and higher.
6798 @item -ftree-builtin-call-dce
6799 @opindex ftree-builtin-call-dce
6800 Perform conditional dead code elimination (DCE) for calls to builtin functions
6801 that may set @code{errno} but are otherwise side-effect free. This flag is
6802 enabled by default at @option{-O2} and higher if @option{-Os} is not also
6805 @item -ftree-dominator-opts
6806 @opindex ftree-dominator-opts
6807 Perform a variety of simple scalar cleanups (constant/copy
6808 propagation, redundancy elimination, range propagation and expression
6809 simplification) based on a dominator tree traversal. This also
6810 performs jump threading (to reduce jumps to jumps). This flag is
6811 enabled by default at @option{-O} and higher.
6815 Perform dead store elimination (DSE) on trees. A dead store is a store into
6816 a memory location which will later be overwritten by another store without
6817 any intervening loads. In this case the earlier store can be deleted. This
6818 flag is enabled by default at @option{-O} and higher.
6822 Perform loop header copying on trees. This is beneficial since it increases
6823 effectiveness of code motion optimizations. It also saves one jump. This flag
6824 is enabled by default at @option{-O} and higher. It is not enabled
6825 for @option{-Os}, since it usually increases code size.
6827 @item -ftree-loop-optimize
6828 @opindex ftree-loop-optimize
6829 Perform loop optimizations on trees. This flag is enabled by default
6830 at @option{-O} and higher.
6832 @item -ftree-loop-linear
6833 @opindex ftree-loop-linear
6834 Perform linear loop transformations on tree. This flag can improve cache
6835 performance and allow further loop optimizations to take place.
6837 @item -floop-interchange
6838 @opindex floop-interchange
6839 Perform loop interchange transformations on loops. Interchanging two
6840 nested loops switches the inner and outer loops. For example, given a
6845 A(J, I) = A(J, I) * C
6849 loop interchange will transform the loop as if the user had written:
6853 A(J, I) = A(J, I) * C
6857 which can be beneficial when @code{N} is larger than the caches,
6858 because in Fortran, the elements of an array are stored in memory
6859 contiguously by column, and the original loop iterates over rows,
6860 potentially creating at each access a cache miss. This optimization
6861 applies to all the languages supported by GCC and is not limited to
6862 Fortran. To use this code transformation, GCC has to be configured
6863 with @option{--with-ppl} and @option{--with-cloog} to enable the
6864 Graphite loop transformation infrastructure.
6866 @item -floop-strip-mine
6867 @opindex floop-strip-mine
6868 Perform loop strip mining transformations on loops. Strip mining
6869 splits a loop into two nested loops. The outer loop has strides
6870 equal to the strip size and the inner loop has strides of the
6871 original loop within a strip. The strip length can be changed
6872 using the @option{loop-block-tile-size} parameter. For example,
6879 loop strip mining will transform the loop as if the user had written:
6882 DO I = II, min (II + 50, N)
6887 This optimization applies to all the languages supported by GCC and is
6888 not limited to Fortran. To use this code transformation, GCC has to
6889 be configured with @option{--with-ppl} and @option{--with-cloog} to
6890 enable the Graphite loop transformation infrastructure.
6893 @opindex floop-block
6894 Perform loop blocking transformations on loops. Blocking strip mines
6895 each loop in the loop nest such that the memory accesses of the
6896 element loops fit inside caches. The strip length can be changed
6897 using the @option{loop-block-tile-size} parameter. For example, given
6902 A(J, I) = B(I) + C(J)
6906 loop blocking will transform the loop as if the user had written:
6910 DO I = II, min (II + 50, N)
6911 DO J = JJ, min (JJ + 50, M)
6912 A(J, I) = B(I) + C(J)
6918 which can be beneficial when @code{M} is larger than the caches,
6919 because the innermost loop will iterate over a smaller amount of data
6920 that can be kept in the caches. This optimization applies to all the
6921 languages supported by GCC and is not limited to Fortran. To use this
6922 code transformation, GCC has to be configured with @option{--with-ppl}
6923 and @option{--with-cloog} to enable the Graphite loop transformation
6926 @item -fgraphite-identity
6927 @opindex fgraphite-identity
6928 Enable the identity transformation for graphite. For every SCoP we generate
6929 the polyhedral representation and transform it back to gimple. Using
6930 @option{-fgraphite-identity} we can check the costs or benefits of the
6931 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
6932 are also performed by the code generator CLooG, like index splitting and
6933 dead code elimination in loops.
6935 @item -floop-flatten
6936 @opindex floop-flatten
6937 Removes the loop nesting structure: transforms the loop nest into a
6938 single loop. This transformation can be useful to vectorize all the
6939 levels of the loop nest.
6941 @item -floop-parallelize-all
6942 @opindex floop-parallelize-all
6943 Use the Graphite data dependence analysis to identify loops that can
6944 be parallelized. Parallelize all the loops that can be analyzed to
6945 not contain loop carried dependences without checking that it is
6946 profitable to parallelize the loops.
6948 @item -fcheck-data-deps
6949 @opindex fcheck-data-deps
6950 Compare the results of several data dependence analyzers. This option
6951 is used for debugging the data dependence analyzers.
6953 @item -ftree-loop-if-convert
6954 Attempt to transform conditional jumps in the innermost loops to
6955 branch-less equivalents. The intent is to remove control-flow from
6956 the innermost loops in order to improve the ability of the
6957 vectorization pass to handle these loops. This is enabled by default
6958 if vectorization is enabled.
6960 @item -ftree-loop-if-convert-stores
6961 Attempt to also if-convert conditional jumps containing memory writes.
6962 This transformation can be unsafe for multi-threaded programs as it
6963 transforms conditional memory writes into unconditional memory writes.
6966 for (i = 0; i < N; i++)
6970 would be transformed to
6972 for (i = 0; i < N; i++)
6973 A[i] = cond ? expr : A[i];
6975 potentially producing data races.
6977 @item -ftree-loop-distribution
6978 Perform loop distribution. This flag can improve cache performance on
6979 big loop bodies and allow further loop optimizations, like
6980 parallelization or vectorization, to take place. For example, the loop
6997 @item -ftree-loop-distribute-patterns
6998 Perform loop distribution of patterns that can be code generated with
6999 calls to a library. This flag is enabled by default at @option{-O3}.
7001 This pass distributes the initialization loops and generates a call to
7002 memset zero. For example, the loop
7018 and the initialization loop is transformed into a call to memset zero.
7020 @item -ftree-loop-im
7021 @opindex ftree-loop-im
7022 Perform loop invariant motion on trees. This pass moves only invariants that
7023 would be hard to handle at RTL level (function calls, operations that expand to
7024 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
7025 operands of conditions that are invariant out of the loop, so that we can use
7026 just trivial invariantness analysis in loop unswitching. The pass also includes
7029 @item -ftree-loop-ivcanon
7030 @opindex ftree-loop-ivcanon
7031 Create a canonical counter for number of iterations in the loop for that
7032 determining number of iterations requires complicated analysis. Later
7033 optimizations then may determine the number easily. Useful especially
7034 in connection with unrolling.
7038 Perform induction variable optimizations (strength reduction, induction
7039 variable merging and induction variable elimination) on trees.
7041 @item -ftree-parallelize-loops=n
7042 @opindex ftree-parallelize-loops
7043 Parallelize loops, i.e., split their iteration space to run in n threads.
7044 This is only possible for loops whose iterations are independent
7045 and can be arbitrarily reordered. The optimization is only
7046 profitable on multiprocessor machines, for loops that are CPU-intensive,
7047 rather than constrained e.g.@: by memory bandwidth. This option
7048 implies @option{-pthread}, and thus is only supported on targets
7049 that have support for @option{-pthread}.
7053 Perform function-local points-to analysis on trees. This flag is
7054 enabled by default at @option{-O} and higher.
7058 Perform scalar replacement of aggregates. This pass replaces structure
7059 references with scalars to prevent committing structures to memory too
7060 early. This flag is enabled by default at @option{-O} and higher.
7062 @item -ftree-copyrename
7063 @opindex ftree-copyrename
7064 Perform copy renaming on trees. This pass attempts to rename compiler
7065 temporaries to other variables at copy locations, usually resulting in
7066 variable names which more closely resemble the original variables. This flag
7067 is enabled by default at @option{-O} and higher.
7071 Perform temporary expression replacement during the SSA->normal phase. Single
7072 use/single def temporaries are replaced at their use location with their
7073 defining expression. This results in non-GIMPLE code, but gives the expanders
7074 much more complex trees to work on resulting in better RTL generation. This is
7075 enabled by default at @option{-O} and higher.
7077 @item -ftree-vectorize
7078 @opindex ftree-vectorize
7079 Perform loop vectorization on trees. This flag is enabled by default at
7082 @item -ftree-slp-vectorize
7083 @opindex ftree-slp-vectorize
7084 Perform basic block vectorization on trees. This flag is enabled by default at
7085 @option{-O3} and when @option{-ftree-vectorize} is enabled.
7087 @item -ftree-vect-loop-version
7088 @opindex ftree-vect-loop-version
7089 Perform loop versioning when doing loop vectorization on trees. When a loop
7090 appears to be vectorizable except that data alignment or data dependence cannot
7091 be determined at compile time then vectorized and non-vectorized versions of
7092 the loop are generated along with runtime checks for alignment or dependence
7093 to control which version is executed. This option is enabled by default
7094 except at level @option{-Os} where it is disabled.
7096 @item -fvect-cost-model
7097 @opindex fvect-cost-model
7098 Enable cost model for vectorization.
7102 Perform Value Range Propagation on trees. This is similar to the
7103 constant propagation pass, but instead of values, ranges of values are
7104 propagated. This allows the optimizers to remove unnecessary range
7105 checks like array bound checks and null pointer checks. This is
7106 enabled by default at @option{-O2} and higher. Null pointer check
7107 elimination is only done if @option{-fdelete-null-pointer-checks} is
7112 Perform tail duplication to enlarge superblock size. This transformation
7113 simplifies the control flow of the function allowing other optimizations to do
7116 @item -funroll-loops
7117 @opindex funroll-loops
7118 Unroll loops whose number of iterations can be determined at compile
7119 time or upon entry to the loop. @option{-funroll-loops} implies
7120 @option{-frerun-cse-after-loop}. This option makes code larger,
7121 and may or may not make it run faster.
7123 @item -funroll-all-loops
7124 @opindex funroll-all-loops
7125 Unroll all loops, even if their number of iterations is uncertain when
7126 the loop is entered. This usually makes programs run more slowly.
7127 @option{-funroll-all-loops} implies the same options as
7128 @option{-funroll-loops},
7130 @item -fsplit-ivs-in-unroller
7131 @opindex fsplit-ivs-in-unroller
7132 Enables expressing of values of induction variables in later iterations
7133 of the unrolled loop using the value in the first iteration. This breaks
7134 long dependency chains, thus improving efficiency of the scheduling passes.
7136 Combination of @option{-fweb} and CSE is often sufficient to obtain the
7137 same effect. However in cases the loop body is more complicated than
7138 a single basic block, this is not reliable. It also does not work at all
7139 on some of the architectures due to restrictions in the CSE pass.
7141 This optimization is enabled by default.
7143 @item -fvariable-expansion-in-unroller
7144 @opindex fvariable-expansion-in-unroller
7145 With this option, the compiler will create multiple copies of some
7146 local variables when unrolling a loop which can result in superior code.
7148 @item -fpartial-inlining
7149 @opindex fpartial-inlining
7150 Inline parts of functions. This option has any effect only
7151 when inlining itself is turned on by the @option{-finline-functions}
7152 or @option{-finline-small-functions} options.
7154 Enabled at level @option{-O2}.
7156 @item -fpredictive-commoning
7157 @opindex fpredictive-commoning
7158 Perform predictive commoning optimization, i.e., reusing computations
7159 (especially memory loads and stores) performed in previous
7160 iterations of loops.
7162 This option is enabled at level @option{-O3}.
7164 @item -fprefetch-loop-arrays
7165 @opindex fprefetch-loop-arrays
7166 If supported by the target machine, generate instructions to prefetch
7167 memory to improve the performance of loops that access large arrays.
7169 This option may generate better or worse code; results are highly
7170 dependent on the structure of loops within the source code.
7172 Disabled at level @option{-Os}.
7175 @itemx -fno-peephole2
7176 @opindex fno-peephole
7177 @opindex fno-peephole2
7178 Disable any machine-specific peephole optimizations. The difference
7179 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
7180 are implemented in the compiler; some targets use one, some use the
7181 other, a few use both.
7183 @option{-fpeephole} is enabled by default.
7184 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7186 @item -fno-guess-branch-probability
7187 @opindex fno-guess-branch-probability
7188 Do not guess branch probabilities using heuristics.
7190 GCC will use heuristics to guess branch probabilities if they are
7191 not provided by profiling feedback (@option{-fprofile-arcs}). These
7192 heuristics are based on the control flow graph. If some branch probabilities
7193 are specified by @samp{__builtin_expect}, then the heuristics will be
7194 used to guess branch probabilities for the rest of the control flow graph,
7195 taking the @samp{__builtin_expect} info into account. The interactions
7196 between the heuristics and @samp{__builtin_expect} can be complex, and in
7197 some cases, it may be useful to disable the heuristics so that the effects
7198 of @samp{__builtin_expect} are easier to understand.
7200 The default is @option{-fguess-branch-probability} at levels
7201 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7203 @item -freorder-blocks
7204 @opindex freorder-blocks
7205 Reorder basic blocks in the compiled function in order to reduce number of
7206 taken branches and improve code locality.
7208 Enabled at levels @option{-O2}, @option{-O3}.
7210 @item -freorder-blocks-and-partition
7211 @opindex freorder-blocks-and-partition
7212 In addition to reordering basic blocks in the compiled function, in order
7213 to reduce number of taken branches, partitions hot and cold basic blocks
7214 into separate sections of the assembly and .o files, to improve
7215 paging and cache locality performance.
7217 This optimization is automatically turned off in the presence of
7218 exception handling, for linkonce sections, for functions with a user-defined
7219 section attribute and on any architecture that does not support named
7222 @item -freorder-functions
7223 @opindex freorder-functions
7224 Reorder functions in the object file in order to
7225 improve code locality. This is implemented by using special
7226 subsections @code{.text.hot} for most frequently executed functions and
7227 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
7228 the linker so object file format must support named sections and linker must
7229 place them in a reasonable way.
7231 Also profile feedback must be available in to make this option effective. See
7232 @option{-fprofile-arcs} for details.
7234 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7236 @item -fstrict-aliasing
7237 @opindex fstrict-aliasing
7238 Allow the compiler to assume the strictest aliasing rules applicable to
7239 the language being compiled. For C (and C++), this activates
7240 optimizations based on the type of expressions. In particular, an
7241 object of one type is assumed never to reside at the same address as an
7242 object of a different type, unless the types are almost the same. For
7243 example, an @code{unsigned int} can alias an @code{int}, but not a
7244 @code{void*} or a @code{double}. A character type may alias any other
7247 @anchor{Type-punning}Pay special attention to code like this:
7260 The practice of reading from a different union member than the one most
7261 recently written to (called ``type-punning'') is common. Even with
7262 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
7263 is accessed through the union type. So, the code above will work as
7264 expected. @xref{Structures unions enumerations and bit-fields
7265 implementation}. However, this code might not:
7276 Similarly, access by taking the address, casting the resulting pointer
7277 and dereferencing the result has undefined behavior, even if the cast
7278 uses a union type, e.g.:
7282 return ((union a_union *) &d)->i;
7286 The @option{-fstrict-aliasing} option is enabled at levels
7287 @option{-O2}, @option{-O3}, @option{-Os}.
7289 @item -fstrict-overflow
7290 @opindex fstrict-overflow
7291 Allow the compiler to assume strict signed overflow rules, depending
7292 on the language being compiled. For C (and C++) this means that
7293 overflow when doing arithmetic with signed numbers is undefined, which
7294 means that the compiler may assume that it will not happen. This
7295 permits various optimizations. For example, the compiler will assume
7296 that an expression like @code{i + 10 > i} will always be true for
7297 signed @code{i}. This assumption is only valid if signed overflow is
7298 undefined, as the expression is false if @code{i + 10} overflows when
7299 using twos complement arithmetic. When this option is in effect any
7300 attempt to determine whether an operation on signed numbers will
7301 overflow must be written carefully to not actually involve overflow.
7303 This option also allows the compiler to assume strict pointer
7304 semantics: given a pointer to an object, if adding an offset to that
7305 pointer does not produce a pointer to the same object, the addition is
7306 undefined. This permits the compiler to conclude that @code{p + u >
7307 p} is always true for a pointer @code{p} and unsigned integer
7308 @code{u}. This assumption is only valid because pointer wraparound is
7309 undefined, as the expression is false if @code{p + u} overflows using
7310 twos complement arithmetic.
7312 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
7313 that integer signed overflow is fully defined: it wraps. When
7314 @option{-fwrapv} is used, there is no difference between
7315 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
7316 integers. With @option{-fwrapv} certain types of overflow are
7317 permitted. For example, if the compiler gets an overflow when doing
7318 arithmetic on constants, the overflowed value can still be used with
7319 @option{-fwrapv}, but not otherwise.
7321 The @option{-fstrict-overflow} option is enabled at levels
7322 @option{-O2}, @option{-O3}, @option{-Os}.
7324 @item -falign-functions
7325 @itemx -falign-functions=@var{n}
7326 @opindex falign-functions
7327 Align the start of functions to the next power-of-two greater than
7328 @var{n}, skipping up to @var{n} bytes. For instance,
7329 @option{-falign-functions=32} aligns functions to the next 32-byte
7330 boundary, but @option{-falign-functions=24} would align to the next
7331 32-byte boundary only if this can be done by skipping 23 bytes or less.
7333 @option{-fno-align-functions} and @option{-falign-functions=1} are
7334 equivalent and mean that functions will not be aligned.
7336 Some assemblers only support this flag when @var{n} is a power of two;
7337 in that case, it is rounded up.
7339 If @var{n} is not specified or is zero, use a machine-dependent default.
7341 Enabled at levels @option{-O2}, @option{-O3}.
7343 @item -falign-labels
7344 @itemx -falign-labels=@var{n}
7345 @opindex falign-labels
7346 Align all branch targets to a power-of-two boundary, skipping up to
7347 @var{n} bytes like @option{-falign-functions}. This option can easily
7348 make code slower, because it must insert dummy operations for when the
7349 branch target is reached in the usual flow of the code.
7351 @option{-fno-align-labels} and @option{-falign-labels=1} are
7352 equivalent and mean that labels will not be aligned.
7354 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
7355 are greater than this value, then their values are used instead.
7357 If @var{n} is not specified or is zero, use a machine-dependent default
7358 which is very likely to be @samp{1}, meaning no alignment.
7360 Enabled at levels @option{-O2}, @option{-O3}.
7363 @itemx -falign-loops=@var{n}
7364 @opindex falign-loops
7365 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7366 like @option{-falign-functions}. The hope is that the loop will be
7367 executed many times, which will make up for any execution of the dummy
7370 @option{-fno-align-loops} and @option{-falign-loops=1} are
7371 equivalent and mean that loops will not be aligned.
7373 If @var{n} is not specified or is zero, use a machine-dependent default.
7375 Enabled at levels @option{-O2}, @option{-O3}.
7378 @itemx -falign-jumps=@var{n}
7379 @opindex falign-jumps
7380 Align branch targets to a power-of-two boundary, for branch targets
7381 where the targets can only be reached by jumping, skipping up to @var{n}
7382 bytes like @option{-falign-functions}. In this case, no dummy operations
7385 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7386 equivalent and mean that loops will not be aligned.
7388 If @var{n} is not specified or is zero, use a machine-dependent default.
7390 Enabled at levels @option{-O2}, @option{-O3}.
7392 @item -funit-at-a-time
7393 @opindex funit-at-a-time
7394 This option is left for compatibility reasons. @option{-funit-at-a-time}
7395 has no effect, while @option{-fno-unit-at-a-time} implies
7396 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7400 @item -fno-toplevel-reorder
7401 @opindex fno-toplevel-reorder
7402 Do not reorder top-level functions, variables, and @code{asm}
7403 statements. Output them in the same order that they appear in the
7404 input file. When this option is used, unreferenced static variables
7405 will not be removed. This option is intended to support existing code
7406 which relies on a particular ordering. For new code, it is better to
7409 Enabled at level @option{-O0}. When disabled explicitly, it also imply
7410 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
7415 Constructs webs as commonly used for register allocation purposes and assign
7416 each web individual pseudo register. This allows the register allocation pass
7417 to operate on pseudos directly, but also strengthens several other optimization
7418 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7419 however, make debugging impossible, since variables will no longer stay in a
7422 Enabled by default with @option{-funroll-loops}.
7424 @item -fwhole-program
7425 @opindex fwhole-program
7426 Assume that the current compilation unit represents the whole program being
7427 compiled. All public functions and variables with the exception of @code{main}
7428 and those merged by attribute @code{externally_visible} become static functions
7429 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.
7430 While this option is equivalent to proper use of the @code{static} keyword for
7431 programs consisting of a single file, in combination with option
7432 @option{-flto} this flag can be used to
7433 compile many smaller scale programs since the functions and variables become
7434 local for the whole combined compilation unit, not for the single source file
7437 This option implies @option{-fwhole-file} for Fortran programs.
7439 @item -flto[=@var{n}]
7441 This option runs the standard link-time optimizer. When invoked
7442 with source code, it generates GIMPLE (one of GCC's internal
7443 representations) and writes it to special ELF sections in the object
7444 file. When the object files are linked together, all the function
7445 bodies are read from these ELF sections and instantiated as if they
7446 had been part of the same translation unit.
7448 To use the link-timer optimizer, @option{-flto} needs to be specified at
7449 compile time and during the final link. For example,
7452 gcc -c -O2 -flto foo.c
7453 gcc -c -O2 -flto bar.c
7454 gcc -o myprog -flto -O2 foo.o bar.o
7457 The first two invocations to GCC will save a bytecode representation
7458 of GIMPLE into special ELF sections inside @file{foo.o} and
7459 @file{bar.o}. The final invocation will read the GIMPLE bytecode from
7460 @file{foo.o} and @file{bar.o}, merge the two files into a single
7461 internal image, and compile the result as usual. Since both
7462 @file{foo.o} and @file{bar.o} are merged into a single image, this
7463 causes all the inter-procedural analyses and optimizations in GCC to
7464 work across the two files as if they were a single one. This means,
7465 for example, that the inliner will be able to inline functions in
7466 @file{bar.o} into functions in @file{foo.o} and vice-versa.
7468 Another (simpler) way to enable link-time optimization is,
7471 gcc -o myprog -flto -O2 foo.c bar.c
7474 The above will generate bytecode for @file{foo.c} and @file{bar.c},
7475 merge them together into a single GIMPLE representation and optimize
7476 them as usual to produce @file{myprog}.
7478 The only important thing to keep in mind is that to enable link-time
7479 optimizations the @option{-flto} flag needs to be passed to both the
7480 compile and the link commands.
7482 Note that when a file is compiled with @option{-flto}, the generated
7483 object file will be larger than a regular object file because it will
7484 contain GIMPLE bytecodes and the usual final code. This means that
7485 object files with LTO information can be linked as a normal object
7486 file. So, in the previous example, if the final link is done with
7489 gcc -o myprog foo.o bar.o
7492 The only difference will be that no inter-procedural optimizations
7493 will be applied to produce @file{myprog}. The two object files
7494 @file{foo.o} and @file{bar.o} will be simply sent to the regular
7497 Additionally, the optimization flags used to compile individual files
7498 are not necessarily related to those used at link-time. For instance,
7501 gcc -c -O0 -flto foo.c
7502 gcc -c -O0 -flto bar.c
7503 gcc -o myprog -flto -O3 foo.o bar.o
7506 This will produce individual object files with unoptimized assembler
7507 code, but the resulting binary @file{myprog} will be optimized at
7508 @option{-O3}. Now, if the final binary is generated without
7509 @option{-flto}, then @file{myprog} will not be optimized.
7511 When producing the final binary with @option{-flto}, GCC will only
7512 apply link-time optimizations to those files that contain bytecode.
7513 Therefore, you can mix and match object files and libraries with
7514 GIMPLE bytecodes and final object code. GCC will automatically select
7515 which files to optimize in LTO mode and which files to link without
7518 There are some code generation flags that GCC will preserve when
7519 generating bytecodes, as they need to be used during the final link
7520 stage. Currently, the following options are saved into the GIMPLE
7521 bytecode files: @option{-fPIC}, @option{-fcommon} and all the
7522 @option{-m} target flags.
7524 At link time, these options are read-in and reapplied. Note that the
7525 current implementation makes no attempt at recognizing conflicting
7526 values for these options. If two or more files have a conflicting
7527 value (e.g., one file is compiled with @option{-fPIC} and another
7528 isn't), the compiler will simply use the last value read from the
7529 bytecode files. It is recommended, then, that all the files
7530 participating in the same link be compiled with the same options.
7532 Another feature of LTO is that it is possible to apply interprocedural
7533 optimizations on files written in different languages. This requires
7534 some support in the language front end. Currently, the C, C++ and
7535 Fortran front ends are capable of emitting GIMPLE bytecodes, so
7536 something like this should work
7541 gfortran -c -flto baz.f90
7542 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
7545 Notice that the final link is done with @command{g++} to get the C++
7546 runtime libraries and @option{-lgfortran} is added to get the Fortran
7547 runtime libraries. In general, when mixing languages in LTO mode, you
7548 should use the same link command used when mixing languages in a
7549 regular (non-LTO) compilation. This means that if your build process
7550 was mixing languages before, all you need to add is @option{-flto} to
7551 all the compile and link commands.
7553 If LTO encounters objects with C linkage declared with incompatible
7554 types in separate translation units to be linked together (undefined
7555 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
7556 issued. The behavior is still undefined at runtime.
7558 If object files containing GIMPLE bytecode are stored in a library
7559 archive, say @file{libfoo.a}, it is possible to extract and use them
7560 in an LTO link if you are using @command{gold} as the linker (which,
7561 in turn requires GCC to be configured with @option{--enable-gold}).
7562 To enable this feature, use the flag @option{-fuse-linker-plugin} at
7566 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
7569 With the linker plugin enabled, @command{gold} will extract the needed
7570 GIMPLE files from @file{libfoo.a} and pass them on to the running GCC
7571 to make them part of the aggregated GIMPLE image to be optimized.
7573 If you are not using @command{gold} and/or do not specify
7574 @option{-fuse-linker-plugin} then the objects inside @file{libfoo.a}
7575 will be extracted and linked as usual, but they will not participate
7576 in the LTO optimization process.
7578 Link time optimizations do not require the presence of the whole program to
7579 operate. If the program does not require any symbols to be exported, it is
7580 possible to combine @option{-flto} and with @option{-fwhole-program} to allow
7581 the interprocedural optimizers to use more aggressive assumptions which may
7582 lead to improved optimization opportunities.
7583 Use of @option{-fwhole-program} is not needed when linker plugin is
7584 active (see @option{-fuse-linker-plugin}).
7586 Regarding portability: the current implementation of LTO makes no
7587 attempt at generating bytecode that can be ported between different
7588 types of hosts. The bytecode files are versioned and there is a
7589 strict version check, so bytecode files generated in one version of
7590 GCC will not work with an older/newer version of GCC.
7592 Link time optimization does not play well with generating debugging
7593 information. Combining @option{-flto} with
7594 @option{-g} is experimental.
7596 If you specify the optional @var{n} the link stage is executed in
7597 parallel using @var{n} parallel jobs by utilizing an installed
7598 @command{make} program. The environment variable @env{MAKE} may be
7599 used to override the program used.
7601 You can also specify @option{-fwhopr=jobserver} to use GNU make's
7602 job server mode to determine the number of parallel jobs. This
7603 is useful when the Makefile calling GCC is already parallel.
7604 The parent Makefile will need a @samp{+} prepended to the command recipe
7605 for this to work. This will likely only work if @env{MAKE} is
7608 This option is disabled by default.
7610 @item -flto-partition=@var{alg}
7611 @opindex flto-partition
7612 Specify partitioning algorithm used by @option{-fwhopr} mode. The value is
7613 either @code{1to1} to specify partitioning corresponding to source files
7614 or @code{balanced} to specify partitioning into, if possible, equally sized
7615 chunks. Specifying @code{none} as an algorithm disables partitioning
7616 and streaming completely.
7617 The default value is @code{balanced}.
7621 This is an internal option used by GCC when compiling with
7622 @option{-fwhopr}. You should never need to use it.
7624 This option runs the link-time optimizer in the whole-program-analysis
7625 (WPA) mode, which reads in summary information from all inputs and
7626 performs a whole-program analysis based on summary information only.
7627 It generates object files for subsequent runs of the link-time
7628 optimizer where individual object files are optimized using both
7629 summary information from the WPA mode and the actual function bodies.
7630 It then drives the LTRANS phase.
7632 Disabled by default.
7636 This is an internal option used by GCC when compiling with
7637 @option{-fwhopr}. You should never need to use it.
7639 This option runs the link-time optimizer in the local-transformation (LTRANS)
7640 mode, which reads in output from a previous run of the LTO in WPA mode.
7641 In the LTRANS mode, LTO optimizes an object and produces the final assembly.
7643 Disabled by default.
7645 @item -fltrans-output-list=@var{file}
7646 @opindex fltrans-output-list
7647 This is an internal option used by GCC when compiling with
7648 @option{-fwhopr}. You should never need to use it.
7650 This option specifies a file to which the names of LTRANS output files are
7651 written. This option is only meaningful in conjunction with @option{-fwpa}.
7653 Disabled by default.
7655 @item -flto-compression-level=@var{n}
7656 This option specifies the level of compression used for intermediate
7657 language written to LTO object files, and is only meaningful in
7658 conjunction with LTO mode (@option{-fwhopr}, @option{-flto}). Valid
7659 values are 0 (no compression) to 9 (maximum compression). Values
7660 outside this range are clamped to either 0 or 9. If the option is not
7661 given, a default balanced compression setting is used.
7664 Prints a report with internal details on the workings of the link-time
7665 optimizer. The contents of this report vary from version to version,
7666 it is meant to be useful to GCC developers when processing object
7667 files in LTO mode (via @option{-fwhopr} or @option{-flto}).
7669 Disabled by default.
7671 @item -fuse-linker-plugin
7672 Enables the extraction of objects with GIMPLE bytecode information
7673 from library archives. This option relies on features available only
7674 in @command{gold}, so to use this you must configure GCC with
7675 @option{--enable-gold}. See @option{-flto} for a description on the
7676 effect of this flag and how to use it.
7678 Disabled by default.
7680 @item -fcprop-registers
7681 @opindex fcprop-registers
7682 After register allocation and post-register allocation instruction splitting,
7683 we perform a copy-propagation pass to try to reduce scheduling dependencies
7684 and occasionally eliminate the copy.
7686 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7688 @item -fprofile-correction
7689 @opindex fprofile-correction
7690 Profiles collected using an instrumented binary for multi-threaded programs may
7691 be inconsistent due to missed counter updates. When this option is specified,
7692 GCC will use heuristics to correct or smooth out such inconsistencies. By
7693 default, GCC will emit an error message when an inconsistent profile is detected.
7695 @item -fprofile-dir=@var{path}
7696 @opindex fprofile-dir
7698 Set the directory to search the profile data files in to @var{path}.
7699 This option affects only the profile data generated by
7700 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
7701 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
7702 and its related options.
7703 By default, GCC will use the current directory as @var{path}
7704 thus the profile data file will appear in the same directory as the object file.
7706 @item -fprofile-generate
7707 @itemx -fprofile-generate=@var{path}
7708 @opindex fprofile-generate
7710 Enable options usually used for instrumenting application to produce
7711 profile useful for later recompilation with profile feedback based
7712 optimization. You must use @option{-fprofile-generate} both when
7713 compiling and when linking your program.
7715 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
7717 If @var{path} is specified, GCC will look at the @var{path} to find
7718 the profile feedback data files. See @option{-fprofile-dir}.
7721 @itemx -fprofile-use=@var{path}
7722 @opindex fprofile-use
7723 Enable profile feedback directed optimizations, and optimizations
7724 generally profitable only with profile feedback available.
7726 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
7727 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
7729 By default, GCC emits an error message if the feedback profiles do not
7730 match the source code. This error can be turned into a warning by using
7731 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
7734 If @var{path} is specified, GCC will look at the @var{path} to find
7735 the profile feedback data files. See @option{-fprofile-dir}.
7738 The following options control compiler behavior regarding floating
7739 point arithmetic. These options trade off between speed and
7740 correctness. All must be specifically enabled.
7744 @opindex ffloat-store
7745 Do not store floating point variables in registers, and inhibit other
7746 options that might change whether a floating point value is taken from a
7749 @cindex floating point precision
7750 This option prevents undesirable excess precision on machines such as
7751 the 68000 where the floating registers (of the 68881) keep more
7752 precision than a @code{double} is supposed to have. Similarly for the
7753 x86 architecture. For most programs, the excess precision does only
7754 good, but a few programs rely on the precise definition of IEEE floating
7755 point. Use @option{-ffloat-store} for such programs, after modifying
7756 them to store all pertinent intermediate computations into variables.
7758 @item -fexcess-precision=@var{style}
7759 @opindex fexcess-precision
7760 This option allows further control over excess precision on machines
7761 where floating-point registers have more precision than the IEEE
7762 @code{float} and @code{double} types and the processor does not
7763 support operations rounding to those types. By default,
7764 @option{-fexcess-precision=fast} is in effect; this means that
7765 operations are carried out in the precision of the registers and that
7766 it is unpredictable when rounding to the types specified in the source
7767 code takes place. When compiling C, if
7768 @option{-fexcess-precision=standard} is specified then excess
7769 precision will follow the rules specified in ISO C99; in particular,
7770 both casts and assignments cause values to be rounded to their
7771 semantic types (whereas @option{-ffloat-store} only affects
7772 assignments). This option is enabled by default for C if a strict
7773 conformance option such as @option{-std=c99} is used.
7776 @option{-fexcess-precision=standard} is not implemented for languages
7777 other than C, and has no effect if
7778 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
7779 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
7780 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
7781 semantics apply without excess precision, and in the latter, rounding
7786 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
7787 @option{-ffinite-math-only}, @option{-fno-rounding-math},
7788 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
7790 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
7792 This option is not turned on by any @option{-O} option since
7793 it can result in incorrect output for programs which depend on
7794 an exact implementation of IEEE or ISO rules/specifications for
7795 math functions. It may, however, yield faster code for programs
7796 that do not require the guarantees of these specifications.
7798 @item -fno-math-errno
7799 @opindex fno-math-errno
7800 Do not set ERRNO after calling math functions that are executed
7801 with a single instruction, e.g., sqrt. A program that relies on
7802 IEEE exceptions for math error handling may want to use this flag
7803 for speed while maintaining IEEE arithmetic compatibility.
7805 This option is not turned on by any @option{-O} option since
7806 it can result in incorrect output for programs which depend on
7807 an exact implementation of IEEE or ISO rules/specifications for
7808 math functions. It may, however, yield faster code for programs
7809 that do not require the guarantees of these specifications.
7811 The default is @option{-fmath-errno}.
7813 On Darwin systems, the math library never sets @code{errno}. There is
7814 therefore no reason for the compiler to consider the possibility that
7815 it might, and @option{-fno-math-errno} is the default.
7817 @item -funsafe-math-optimizations
7818 @opindex funsafe-math-optimizations
7820 Allow optimizations for floating-point arithmetic that (a) assume
7821 that arguments and results are valid and (b) may violate IEEE or
7822 ANSI standards. When used at link-time, it may include libraries
7823 or startup files that change the default FPU control word or other
7824 similar optimizations.
7826 This option is not turned on by any @option{-O} option since
7827 it can result in incorrect output for programs which depend on
7828 an exact implementation of IEEE or ISO rules/specifications for
7829 math functions. It may, however, yield faster code for programs
7830 that do not require the guarantees of these specifications.
7831 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
7832 @option{-fassociative-math} and @option{-freciprocal-math}.
7834 The default is @option{-fno-unsafe-math-optimizations}.
7836 @item -fassociative-math
7837 @opindex fassociative-math
7839 Allow re-association of operands in series of floating-point operations.
7840 This violates the ISO C and C++ language standard by possibly changing
7841 computation result. NOTE: re-ordering may change the sign of zero as
7842 well as ignore NaNs and inhibit or create underflow or overflow (and
7843 thus cannot be used on a code which relies on rounding behavior like
7844 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
7845 and thus may not be used when ordered comparisons are required.
7846 This option requires that both @option{-fno-signed-zeros} and
7847 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
7848 much sense with @option{-frounding-math}. For Fortran the option
7849 is automatically enabled when both @option{-fno-signed-zeros} and
7850 @option{-fno-trapping-math} are in effect.
7852 The default is @option{-fno-associative-math}.
7854 @item -freciprocal-math
7855 @opindex freciprocal-math
7857 Allow the reciprocal of a value to be used instead of dividing by
7858 the value if this enables optimizations. For example @code{x / y}
7859 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
7860 is subject to common subexpression elimination. Note that this loses
7861 precision and increases the number of flops operating on the value.
7863 The default is @option{-fno-reciprocal-math}.
7865 @item -ffinite-math-only
7866 @opindex ffinite-math-only
7867 Allow optimizations for floating-point arithmetic that assume
7868 that arguments and results are not NaNs or +-Infs.
7870 This option is not turned on by any @option{-O} option since
7871 it can result in incorrect output for programs which depend on
7872 an exact implementation of IEEE or ISO rules/specifications for
7873 math functions. It may, however, yield faster code for programs
7874 that do not require the guarantees of these specifications.
7876 The default is @option{-fno-finite-math-only}.
7878 @item -fno-signed-zeros
7879 @opindex fno-signed-zeros
7880 Allow optimizations for floating point arithmetic that ignore the
7881 signedness of zero. IEEE arithmetic specifies the behavior of
7882 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
7883 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
7884 This option implies that the sign of a zero result isn't significant.
7886 The default is @option{-fsigned-zeros}.
7888 @item -fno-trapping-math
7889 @opindex fno-trapping-math
7890 Compile code assuming that floating-point operations cannot generate
7891 user-visible traps. These traps include division by zero, overflow,
7892 underflow, inexact result and invalid operation. This option requires
7893 that @option{-fno-signaling-nans} be in effect. Setting this option may
7894 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
7896 This option should never be turned on by any @option{-O} option since
7897 it can result in incorrect output for programs which depend on
7898 an exact implementation of IEEE or ISO rules/specifications for
7901 The default is @option{-ftrapping-math}.
7903 @item -frounding-math
7904 @opindex frounding-math
7905 Disable transformations and optimizations that assume default floating
7906 point rounding behavior. This is round-to-zero for all floating point
7907 to integer conversions, and round-to-nearest for all other arithmetic
7908 truncations. This option should be specified for programs that change
7909 the FP rounding mode dynamically, or that may be executed with a
7910 non-default rounding mode. This option disables constant folding of
7911 floating point expressions at compile-time (which may be affected by
7912 rounding mode) and arithmetic transformations that are unsafe in the
7913 presence of sign-dependent rounding modes.
7915 The default is @option{-fno-rounding-math}.
7917 This option is experimental and does not currently guarantee to
7918 disable all GCC optimizations that are affected by rounding mode.
7919 Future versions of GCC may provide finer control of this setting
7920 using C99's @code{FENV_ACCESS} pragma. This command line option
7921 will be used to specify the default state for @code{FENV_ACCESS}.
7923 @item -fsignaling-nans
7924 @opindex fsignaling-nans
7925 Compile code assuming that IEEE signaling NaNs may generate user-visible
7926 traps during floating-point operations. Setting this option disables
7927 optimizations that may change the number of exceptions visible with
7928 signaling NaNs. This option implies @option{-ftrapping-math}.
7930 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
7933 The default is @option{-fno-signaling-nans}.
7935 This option is experimental and does not currently guarantee to
7936 disable all GCC optimizations that affect signaling NaN behavior.
7938 @item -fsingle-precision-constant
7939 @opindex fsingle-precision-constant
7940 Treat floating point constant as single precision constant instead of
7941 implicitly converting it to double precision constant.
7943 @item -fcx-limited-range
7944 @opindex fcx-limited-range
7945 When enabled, this option states that a range reduction step is not
7946 needed when performing complex division. Also, there is no checking
7947 whether the result of a complex multiplication or division is @code{NaN
7948 + I*NaN}, with an attempt to rescue the situation in that case. The
7949 default is @option{-fno-cx-limited-range}, but is enabled by
7950 @option{-ffast-math}.
7952 This option controls the default setting of the ISO C99
7953 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
7956 @item -fcx-fortran-rules
7957 @opindex fcx-fortran-rules
7958 Complex multiplication and division follow Fortran rules. Range
7959 reduction is done as part of complex division, but there is no checking
7960 whether the result of a complex multiplication or division is @code{NaN
7961 + I*NaN}, with an attempt to rescue the situation in that case.
7963 The default is @option{-fno-cx-fortran-rules}.
7967 The following options control optimizations that may improve
7968 performance, but are not enabled by any @option{-O} options. This
7969 section includes experimental options that may produce broken code.
7972 @item -fbranch-probabilities
7973 @opindex fbranch-probabilities
7974 After running a program compiled with @option{-fprofile-arcs}
7975 (@pxref{Debugging Options,, Options for Debugging Your Program or
7976 @command{gcc}}), you can compile it a second time using
7977 @option{-fbranch-probabilities}, to improve optimizations based on
7978 the number of times each branch was taken. When the program
7979 compiled with @option{-fprofile-arcs} exits it saves arc execution
7980 counts to a file called @file{@var{sourcename}.gcda} for each source
7981 file. The information in this data file is very dependent on the
7982 structure of the generated code, so you must use the same source code
7983 and the same optimization options for both compilations.
7985 With @option{-fbranch-probabilities}, GCC puts a
7986 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
7987 These can be used to improve optimization. Currently, they are only
7988 used in one place: in @file{reorg.c}, instead of guessing which path a
7989 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
7990 exactly determine which path is taken more often.
7992 @item -fprofile-values
7993 @opindex fprofile-values
7994 If combined with @option{-fprofile-arcs}, it adds code so that some
7995 data about values of expressions in the program is gathered.
7997 With @option{-fbranch-probabilities}, it reads back the data gathered
7998 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
7999 notes to instructions for their later usage in optimizations.
8001 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
8005 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
8006 a code to gather information about values of expressions.
8008 With @option{-fbranch-probabilities}, it reads back the data gathered
8009 and actually performs the optimizations based on them.
8010 Currently the optimizations include specialization of division operation
8011 using the knowledge about the value of the denominator.
8013 @item -frename-registers
8014 @opindex frename-registers
8015 Attempt to avoid false dependencies in scheduled code by making use
8016 of registers left over after register allocation. This optimization
8017 will most benefit processors with lots of registers. Depending on the
8018 debug information format adopted by the target, however, it can
8019 make debugging impossible, since variables will no longer stay in
8020 a ``home register''.
8022 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
8026 Perform tail duplication to enlarge superblock size. This transformation
8027 simplifies the control flow of the function allowing other optimizations to do
8030 Enabled with @option{-fprofile-use}.
8032 @item -funroll-loops
8033 @opindex funroll-loops
8034 Unroll loops whose number of iterations can be determined at compile time or
8035 upon entry to the loop. @option{-funroll-loops} implies
8036 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
8037 It also turns on complete loop peeling (i.e.@: complete removal of loops with
8038 small constant number of iterations). This option makes code larger, and may
8039 or may not make it run faster.
8041 Enabled with @option{-fprofile-use}.
8043 @item -funroll-all-loops
8044 @opindex funroll-all-loops
8045 Unroll all loops, even if their number of iterations is uncertain when
8046 the loop is entered. This usually makes programs run more slowly.
8047 @option{-funroll-all-loops} implies the same options as
8048 @option{-funroll-loops}.
8051 @opindex fpeel-loops
8052 Peels the loops for that there is enough information that they do not
8053 roll much (from profile feedback). It also turns on complete loop peeling
8054 (i.e.@: complete removal of loops with small constant number of iterations).
8056 Enabled with @option{-fprofile-use}.
8058 @item -fmove-loop-invariants
8059 @opindex fmove-loop-invariants
8060 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
8061 at level @option{-O1}
8063 @item -funswitch-loops
8064 @opindex funswitch-loops
8065 Move branches with loop invariant conditions out of the loop, with duplicates
8066 of the loop on both branches (modified according to result of the condition).
8068 @item -ffunction-sections
8069 @itemx -fdata-sections
8070 @opindex ffunction-sections
8071 @opindex fdata-sections
8072 Place each function or data item into its own section in the output
8073 file if the target supports arbitrary sections. The name of the
8074 function or the name of the data item determines the section's name
8077 Use these options on systems where the linker can perform optimizations
8078 to improve locality of reference in the instruction space. Most systems
8079 using the ELF object format and SPARC processors running Solaris 2 have
8080 linkers with such optimizations. AIX may have these optimizations in
8083 Only use these options when there are significant benefits from doing
8084 so. When you specify these options, the assembler and linker will
8085 create larger object and executable files and will also be slower.
8086 You will not be able to use @code{gprof} on all systems if you
8087 specify this option and you may have problems with debugging if
8088 you specify both this option and @option{-g}.
8090 @item -fbranch-target-load-optimize
8091 @opindex fbranch-target-load-optimize
8092 Perform branch target register load optimization before prologue / epilogue
8094 The use of target registers can typically be exposed only during reload,
8095 thus hoisting loads out of loops and doing inter-block scheduling needs
8096 a separate optimization pass.
8098 @item -fbranch-target-load-optimize2
8099 @opindex fbranch-target-load-optimize2
8100 Perform branch target register load optimization after prologue / epilogue
8103 @item -fbtr-bb-exclusive
8104 @opindex fbtr-bb-exclusive
8105 When performing branch target register load optimization, don't reuse
8106 branch target registers in within any basic block.
8108 @item -fstack-protector
8109 @opindex fstack-protector
8110 Emit extra code to check for buffer overflows, such as stack smashing
8111 attacks. This is done by adding a guard variable to functions with
8112 vulnerable objects. This includes functions that call alloca, and
8113 functions with buffers larger than 8 bytes. The guards are initialized
8114 when a function is entered and then checked when the function exits.
8115 If a guard check fails, an error message is printed and the program exits.
8117 @item -fstack-protector-all
8118 @opindex fstack-protector-all
8119 Like @option{-fstack-protector} except that all functions are protected.
8121 @item -fsection-anchors
8122 @opindex fsection-anchors
8123 Try to reduce the number of symbolic address calculations by using
8124 shared ``anchor'' symbols to address nearby objects. This transformation
8125 can help to reduce the number of GOT entries and GOT accesses on some
8128 For example, the implementation of the following function @code{foo}:
8132 int foo (void) @{ return a + b + c; @}
8135 would usually calculate the addresses of all three variables, but if you
8136 compile it with @option{-fsection-anchors}, it will access the variables
8137 from a common anchor point instead. The effect is similar to the
8138 following pseudocode (which isn't valid C):
8143 register int *xr = &x;
8144 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
8148 Not all targets support this option.
8150 @item --param @var{name}=@var{value}
8152 In some places, GCC uses various constants to control the amount of
8153 optimization that is done. For example, GCC will not inline functions
8154 that contain more that a certain number of instructions. You can
8155 control some of these constants on the command-line using the
8156 @option{--param} option.
8158 The names of specific parameters, and the meaning of the values, are
8159 tied to the internals of the compiler, and are subject to change
8160 without notice in future releases.
8162 In each case, the @var{value} is an integer. The allowable choices for
8163 @var{name} are given in the following table:
8166 @item struct-reorg-cold-struct-ratio
8167 The threshold ratio (as a percentage) between a structure frequency
8168 and the frequency of the hottest structure in the program. This parameter
8169 is used by struct-reorg optimization enabled by @option{-fipa-struct-reorg}.
8170 We say that if the ratio of a structure frequency, calculated by profiling,
8171 to the hottest structure frequency in the program is less than this
8172 parameter, then structure reorganization is not applied to this structure.
8175 @item predictable-branch-outcome
8176 When branch is predicted to be taken with probability lower than this threshold
8177 (in percent), then it is considered well predictable. The default is 10.
8179 @item max-crossjump-edges
8180 The maximum number of incoming edges to consider for crossjumping.
8181 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
8182 the number of edges incoming to each block. Increasing values mean
8183 more aggressive optimization, making the compile time increase with
8184 probably small improvement in executable size.
8186 @item min-crossjump-insns
8187 The minimum number of instructions which must be matched at the end
8188 of two blocks before crossjumping will be performed on them. This
8189 value is ignored in the case where all instructions in the block being
8190 crossjumped from are matched. The default value is 5.
8192 @item max-grow-copy-bb-insns
8193 The maximum code size expansion factor when copying basic blocks
8194 instead of jumping. The expansion is relative to a jump instruction.
8195 The default value is 8.
8197 @item max-goto-duplication-insns
8198 The maximum number of instructions to duplicate to a block that jumps
8199 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
8200 passes, GCC factors computed gotos early in the compilation process,
8201 and unfactors them as late as possible. Only computed jumps at the
8202 end of a basic blocks with no more than max-goto-duplication-insns are
8203 unfactored. The default value is 8.
8205 @item max-delay-slot-insn-search
8206 The maximum number of instructions to consider when looking for an
8207 instruction to fill a delay slot. If more than this arbitrary number of
8208 instructions is searched, the time savings from filling the delay slot
8209 will be minimal so stop searching. Increasing values mean more
8210 aggressive optimization, making the compile time increase with probably
8211 small improvement in executable run time.
8213 @item max-delay-slot-live-search
8214 When trying to fill delay slots, the maximum number of instructions to
8215 consider when searching for a block with valid live register
8216 information. Increasing this arbitrarily chosen value means more
8217 aggressive optimization, increasing the compile time. This parameter
8218 should be removed when the delay slot code is rewritten to maintain the
8221 @item max-gcse-memory
8222 The approximate maximum amount of memory that will be allocated in
8223 order to perform the global common subexpression elimination
8224 optimization. If more memory than specified is required, the
8225 optimization will not be done.
8227 @item max-pending-list-length
8228 The maximum number of pending dependencies scheduling will allow
8229 before flushing the current state and starting over. Large functions
8230 with few branches or calls can create excessively large lists which
8231 needlessly consume memory and resources.
8233 @item max-inline-insns-single
8234 Several parameters control the tree inliner used in gcc.
8235 This number sets the maximum number of instructions (counted in GCC's
8236 internal representation) in a single function that the tree inliner
8237 will consider for inlining. This only affects functions declared
8238 inline and methods implemented in a class declaration (C++).
8239 The default value is 300.
8241 @item max-inline-insns-auto
8242 When you use @option{-finline-functions} (included in @option{-O3}),
8243 a lot of functions that would otherwise not be considered for inlining
8244 by the compiler will be investigated. To those functions, a different
8245 (more restrictive) limit compared to functions declared inline can
8247 The default value is 40.
8249 @item large-function-insns
8250 The limit specifying really large functions. For functions larger than this
8251 limit after inlining, inlining is constrained by
8252 @option{--param large-function-growth}. This parameter is useful primarily
8253 to avoid extreme compilation time caused by non-linear algorithms used by the
8255 The default value is 2700.
8257 @item large-function-growth
8258 Specifies maximal growth of large function caused by inlining in percents.
8259 The default value is 100 which limits large function growth to 2.0 times
8262 @item large-unit-insns
8263 The limit specifying large translation unit. Growth caused by inlining of
8264 units larger than this limit is limited by @option{--param inline-unit-growth}.
8265 For small units this might be too tight (consider unit consisting of function A
8266 that is inline and B that just calls A three time. If B is small relative to
8267 A, the growth of unit is 300\% and yet such inlining is very sane. For very
8268 large units consisting of small inlineable functions however the overall unit
8269 growth limit is needed to avoid exponential explosion of code size. Thus for
8270 smaller units, the size is increased to @option{--param large-unit-insns}
8271 before applying @option{--param inline-unit-growth}. The default is 10000
8273 @item inline-unit-growth
8274 Specifies maximal overall growth of the compilation unit caused by inlining.
8275 The default value is 30 which limits unit growth to 1.3 times the original
8278 @item ipcp-unit-growth
8279 Specifies maximal overall growth of the compilation unit caused by
8280 interprocedural constant propagation. The default value is 10 which limits
8281 unit growth to 1.1 times the original size.
8283 @item large-stack-frame
8284 The limit specifying large stack frames. While inlining the algorithm is trying
8285 to not grow past this limit too much. Default value is 256 bytes.
8287 @item large-stack-frame-growth
8288 Specifies maximal growth of large stack frames caused by inlining in percents.
8289 The default value is 1000 which limits large stack frame growth to 11 times
8292 @item max-inline-insns-recursive
8293 @itemx max-inline-insns-recursive-auto
8294 Specifies maximum number of instructions out-of-line copy of self recursive inline
8295 function can grow into by performing recursive inlining.
8297 For functions declared inline @option{--param max-inline-insns-recursive} is
8298 taken into account. For function not declared inline, recursive inlining
8299 happens only when @option{-finline-functions} (included in @option{-O3}) is
8300 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
8301 default value is 450.
8303 @item max-inline-recursive-depth
8304 @itemx max-inline-recursive-depth-auto
8305 Specifies maximum recursion depth used by the recursive inlining.
8307 For functions declared inline @option{--param max-inline-recursive-depth} is
8308 taken into account. For function not declared inline, recursive inlining
8309 happens only when @option{-finline-functions} (included in @option{-O3}) is
8310 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
8313 @item min-inline-recursive-probability
8314 Recursive inlining is profitable only for function having deep recursion
8315 in average and can hurt for function having little recursion depth by
8316 increasing the prologue size or complexity of function body to other
8319 When profile feedback is available (see @option{-fprofile-generate}) the actual
8320 recursion depth can be guessed from probability that function will recurse via
8321 given call expression. This parameter limits inlining only to call expression
8322 whose probability exceeds given threshold (in percents). The default value is
8325 @item early-inlining-insns
8326 Specify growth that early inliner can make. In effect it increases amount of
8327 inlining for code having large abstraction penalty. The default value is 8.
8329 @item max-early-inliner-iterations
8330 @itemx max-early-inliner-iterations
8331 Limit of iterations of early inliner. This basically bounds number of nested
8332 indirect calls early inliner can resolve. Deeper chains are still handled by
8335 @item comdat-sharing-probability
8336 @itemx comdat-sharing-probability
8337 Probability (in percent) that C++ inline function with comdat visibility
8338 will be shared acroess multiple compilation units. The default value is 20.
8340 @item min-vect-loop-bound
8341 The minimum number of iterations under which a loop will not get vectorized
8342 when @option{-ftree-vectorize} is used. The number of iterations after
8343 vectorization needs to be greater than the value specified by this option
8344 to allow vectorization. The default value is 0.
8346 @item gcse-cost-distance-ratio
8347 Scaling factor in calculation of maximum distance an expression
8348 can be moved by GCSE optimizations. This is currently supported only in
8349 code hoisting pass. The bigger the ratio, the more agressive code hoisting
8350 will be with simple expressions, i.e., the expressions which have cost
8351 less than @option{gcse-unrestricted-cost}. Specifying 0 will disable
8352 hoisting of simple expressions. The default value is 10.
8354 @item gcse-unrestricted-cost
8355 Cost, roughly measured as the cost of a single typical machine
8356 instruction, at which GCSE optimizations will not constrain
8357 the distance an expression can travel. This is currently
8358 supported only in code hoisting pass. The lesser the cost,
8359 the more aggressive code hoisting will be. Specifying 0 will
8360 allow all expressions to travel unrestricted distances.
8361 The default value is 3.
8363 @item max-hoist-depth
8364 The depth of search in the dominator tree for expressions to hoist.
8365 This is used to avoid quadratic behavior in hoisting algorithm.
8366 The value of 0 will avoid limiting the search, but may slow down compilation
8367 of huge functions. The default value is 30.
8369 @item max-unrolled-insns
8370 The maximum number of instructions that a loop should have if that loop
8371 is unrolled, and if the loop is unrolled, it determines how many times
8372 the loop code is unrolled.
8374 @item max-average-unrolled-insns
8375 The maximum number of instructions biased by probabilities of their execution
8376 that a loop should have if that loop is unrolled, and if the loop is unrolled,
8377 it determines how many times the loop code is unrolled.
8379 @item max-unroll-times
8380 The maximum number of unrollings of a single loop.
8382 @item max-peeled-insns
8383 The maximum number of instructions that a loop should have if that loop
8384 is peeled, and if the loop is peeled, it determines how many times
8385 the loop code is peeled.
8387 @item max-peel-times
8388 The maximum number of peelings of a single loop.
8390 @item max-completely-peeled-insns
8391 The maximum number of insns of a completely peeled loop.
8393 @item max-completely-peel-times
8394 The maximum number of iterations of a loop to be suitable for complete peeling.
8396 @item max-completely-peel-loop-nest-depth
8397 The maximum depth of a loop nest suitable for complete peeling.
8399 @item max-unswitch-insns
8400 The maximum number of insns of an unswitched loop.
8402 @item max-unswitch-level
8403 The maximum number of branches unswitched in a single loop.
8406 The minimum cost of an expensive expression in the loop invariant motion.
8408 @item iv-consider-all-candidates-bound
8409 Bound on number of candidates for induction variables below that
8410 all candidates are considered for each use in induction variable
8411 optimizations. Only the most relevant candidates are considered
8412 if there are more candidates, to avoid quadratic time complexity.
8414 @item iv-max-considered-uses
8415 The induction variable optimizations give up on loops that contain more
8416 induction variable uses.
8418 @item iv-always-prune-cand-set-bound
8419 If number of candidates in the set is smaller than this value,
8420 we always try to remove unnecessary ivs from the set during its
8421 optimization when a new iv is added to the set.
8423 @item scev-max-expr-size
8424 Bound on size of expressions used in the scalar evolutions analyzer.
8425 Large expressions slow the analyzer.
8427 @item omega-max-vars
8428 The maximum number of variables in an Omega constraint system.
8429 The default value is 128.
8431 @item omega-max-geqs
8432 The maximum number of inequalities in an Omega constraint system.
8433 The default value is 256.
8436 The maximum number of equalities in an Omega constraint system.
8437 The default value is 128.
8439 @item omega-max-wild-cards
8440 The maximum number of wildcard variables that the Omega solver will
8441 be able to insert. The default value is 18.
8443 @item omega-hash-table-size
8444 The size of the hash table in the Omega solver. The default value is
8447 @item omega-max-keys
8448 The maximal number of keys used by the Omega solver. The default
8451 @item omega-eliminate-redundant-constraints
8452 When set to 1, use expensive methods to eliminate all redundant
8453 constraints. The default value is 0.
8455 @item vect-max-version-for-alignment-checks
8456 The maximum number of runtime checks that can be performed when
8457 doing loop versioning for alignment in the vectorizer. See option
8458 ftree-vect-loop-version for more information.
8460 @item vect-max-version-for-alias-checks
8461 The maximum number of runtime checks that can be performed when
8462 doing loop versioning for alias in the vectorizer. See option
8463 ftree-vect-loop-version for more information.
8465 @item max-iterations-to-track
8467 The maximum number of iterations of a loop the brute force algorithm
8468 for analysis of # of iterations of the loop tries to evaluate.
8470 @item hot-bb-count-fraction
8471 Select fraction of the maximal count of repetitions of basic block in program
8472 given basic block needs to have to be considered hot.
8474 @item hot-bb-frequency-fraction
8475 Select fraction of the maximal frequency of executions of basic block in
8476 function given basic block needs to have to be considered hot
8478 @item max-predicted-iterations
8479 The maximum number of loop iterations we predict statically. This is useful
8480 in cases where function contain single loop with known bound and other loop
8481 with unknown. We predict the known number of iterations correctly, while
8482 the unknown number of iterations average to roughly 10. This means that the
8483 loop without bounds would appear artificially cold relative to the other one.
8485 @item align-threshold
8487 Select fraction of the maximal frequency of executions of basic block in
8488 function given basic block will get aligned.
8490 @item align-loop-iterations
8492 A loop expected to iterate at lest the selected number of iterations will get
8495 @item tracer-dynamic-coverage
8496 @itemx tracer-dynamic-coverage-feedback
8498 This value is used to limit superblock formation once the given percentage of
8499 executed instructions is covered. This limits unnecessary code size
8502 The @option{tracer-dynamic-coverage-feedback} is used only when profile
8503 feedback is available. The real profiles (as opposed to statically estimated
8504 ones) are much less balanced allowing the threshold to be larger value.
8506 @item tracer-max-code-growth
8507 Stop tail duplication once code growth has reached given percentage. This is
8508 rather hokey argument, as most of the duplicates will be eliminated later in
8509 cross jumping, so it may be set to much higher values than is the desired code
8512 @item tracer-min-branch-ratio
8514 Stop reverse growth when the reverse probability of best edge is less than this
8515 threshold (in percent).
8517 @item tracer-min-branch-ratio
8518 @itemx tracer-min-branch-ratio-feedback
8520 Stop forward growth if the best edge do have probability lower than this
8523 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
8524 compilation for profile feedback and one for compilation without. The value
8525 for compilation with profile feedback needs to be more conservative (higher) in
8526 order to make tracer effective.
8528 @item max-cse-path-length
8530 Maximum number of basic blocks on path that cse considers. The default is 10.
8533 The maximum instructions CSE process before flushing. The default is 1000.
8535 @item ggc-min-expand
8537 GCC uses a garbage collector to manage its own memory allocation. This
8538 parameter specifies the minimum percentage by which the garbage
8539 collector's heap should be allowed to expand between collections.
8540 Tuning this may improve compilation speed; it has no effect on code
8543 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
8544 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
8545 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
8546 GCC is not able to calculate RAM on a particular platform, the lower
8547 bound of 30% is used. Setting this parameter and
8548 @option{ggc-min-heapsize} to zero causes a full collection to occur at
8549 every opportunity. This is extremely slow, but can be useful for
8552 @item ggc-min-heapsize
8554 Minimum size of the garbage collector's heap before it begins bothering
8555 to collect garbage. The first collection occurs after the heap expands
8556 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
8557 tuning this may improve compilation speed, and has no effect on code
8560 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
8561 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
8562 with a lower bound of 4096 (four megabytes) and an upper bound of
8563 131072 (128 megabytes). If GCC is not able to calculate RAM on a
8564 particular platform, the lower bound is used. Setting this parameter
8565 very large effectively disables garbage collection. Setting this
8566 parameter and @option{ggc-min-expand} to zero causes a full collection
8567 to occur at every opportunity.
8569 @item max-reload-search-insns
8570 The maximum number of instruction reload should look backward for equivalent
8571 register. Increasing values mean more aggressive optimization, making the
8572 compile time increase with probably slightly better performance. The default
8575 @item max-cselib-memory-locations
8576 The maximum number of memory locations cselib should take into account.
8577 Increasing values mean more aggressive optimization, making the compile time
8578 increase with probably slightly better performance. The default value is 500.
8580 @item reorder-blocks-duplicate
8581 @itemx reorder-blocks-duplicate-feedback
8583 Used by basic block reordering pass to decide whether to use unconditional
8584 branch or duplicate the code on its destination. Code is duplicated when its
8585 estimated size is smaller than this value multiplied by the estimated size of
8586 unconditional jump in the hot spots of the program.
8588 The @option{reorder-block-duplicate-feedback} is used only when profile
8589 feedback is available and may be set to higher values than
8590 @option{reorder-block-duplicate} since information about the hot spots is more
8593 @item max-sched-ready-insns
8594 The maximum number of instructions ready to be issued the scheduler should
8595 consider at any given time during the first scheduling pass. Increasing
8596 values mean more thorough searches, making the compilation time increase
8597 with probably little benefit. The default value is 100.
8599 @item max-sched-region-blocks
8600 The maximum number of blocks in a region to be considered for
8601 interblock scheduling. The default value is 10.
8603 @item max-pipeline-region-blocks
8604 The maximum number of blocks in a region to be considered for
8605 pipelining in the selective scheduler. The default value is 15.
8607 @item max-sched-region-insns
8608 The maximum number of insns in a region to be considered for
8609 interblock scheduling. The default value is 100.
8611 @item max-pipeline-region-insns
8612 The maximum number of insns in a region to be considered for
8613 pipelining in the selective scheduler. The default value is 200.
8616 The minimum probability (in percents) of reaching a source block
8617 for interblock speculative scheduling. The default value is 40.
8619 @item max-sched-extend-regions-iters
8620 The maximum number of iterations through CFG to extend regions.
8621 0 - disable region extension,
8622 N - do at most N iterations.
8623 The default value is 0.
8625 @item max-sched-insn-conflict-delay
8626 The maximum conflict delay for an insn to be considered for speculative motion.
8627 The default value is 3.
8629 @item sched-spec-prob-cutoff
8630 The minimal probability of speculation success (in percents), so that
8631 speculative insn will be scheduled.
8632 The default value is 40.
8634 @item sched-mem-true-dep-cost
8635 Minimal distance (in CPU cycles) between store and load targeting same
8636 memory locations. The default value is 1.
8638 @item selsched-max-lookahead
8639 The maximum size of the lookahead window of selective scheduling. It is a
8640 depth of search for available instructions.
8641 The default value is 50.
8643 @item selsched-max-sched-times
8644 The maximum number of times that an instruction will be scheduled during
8645 selective scheduling. This is the limit on the number of iterations
8646 through which the instruction may be pipelined. The default value is 2.
8648 @item selsched-max-insns-to-rename
8649 The maximum number of best instructions in the ready list that are considered
8650 for renaming in the selective scheduler. The default value is 2.
8652 @item max-last-value-rtl
8653 The maximum size measured as number of RTLs that can be recorded in an expression
8654 in combiner for a pseudo register as last known value of that register. The default
8657 @item integer-share-limit
8658 Small integer constants can use a shared data structure, reducing the
8659 compiler's memory usage and increasing its speed. This sets the maximum
8660 value of a shared integer constant. The default value is 256.
8662 @item min-virtual-mappings
8663 Specifies the minimum number of virtual mappings in the incremental
8664 SSA updater that should be registered to trigger the virtual mappings
8665 heuristic defined by virtual-mappings-ratio. The default value is
8668 @item virtual-mappings-ratio
8669 If the number of virtual mappings is virtual-mappings-ratio bigger
8670 than the number of virtual symbols to be updated, then the incremental
8671 SSA updater switches to a full update for those symbols. The default
8674 @item ssp-buffer-size
8675 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
8676 protection when @option{-fstack-protection} is used.
8678 @item max-jump-thread-duplication-stmts
8679 Maximum number of statements allowed in a block that needs to be
8680 duplicated when threading jumps.
8682 @item max-fields-for-field-sensitive
8683 Maximum number of fields in a structure we will treat in
8684 a field sensitive manner during pointer analysis. The default is zero
8685 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
8687 @item prefetch-latency
8688 Estimate on average number of instructions that are executed before
8689 prefetch finishes. The distance we prefetch ahead is proportional
8690 to this constant. Increasing this number may also lead to less
8691 streams being prefetched (see @option{simultaneous-prefetches}).
8693 @item simultaneous-prefetches
8694 Maximum number of prefetches that can run at the same time.
8696 @item l1-cache-line-size
8697 The size of cache line in L1 cache, in bytes.
8700 The size of L1 cache, in kilobytes.
8703 The size of L2 cache, in kilobytes.
8705 @item min-insn-to-prefetch-ratio
8706 The minimum ratio between the number of instructions and the
8707 number of prefetches to enable prefetching in a loop.
8709 @item prefetch-min-insn-to-mem-ratio
8710 The minimum ratio between the number of instructions and the
8711 number of memory references to enable prefetching in a loop.
8713 @item use-canonical-types
8714 Whether the compiler should use the ``canonical'' type system. By
8715 default, this should always be 1, which uses a more efficient internal
8716 mechanism for comparing types in C++ and Objective-C++. However, if
8717 bugs in the canonical type system are causing compilation failures,
8718 set this value to 0 to disable canonical types.
8720 @item switch-conversion-max-branch-ratio
8721 Switch initialization conversion will refuse to create arrays that are
8722 bigger than @option{switch-conversion-max-branch-ratio} times the number of
8723 branches in the switch.
8725 @item max-partial-antic-length
8726 Maximum length of the partial antic set computed during the tree
8727 partial redundancy elimination optimization (@option{-ftree-pre}) when
8728 optimizing at @option{-O3} and above. For some sorts of source code
8729 the enhanced partial redundancy elimination optimization can run away,
8730 consuming all of the memory available on the host machine. This
8731 parameter sets a limit on the length of the sets that are computed,
8732 which prevents the runaway behavior. Setting a value of 0 for
8733 this parameter will allow an unlimited set length.
8735 @item sccvn-max-scc-size
8736 Maximum size of a strongly connected component (SCC) during SCCVN
8737 processing. If this limit is hit, SCCVN processing for the whole
8738 function will not be done and optimizations depending on it will
8739 be disabled. The default maximum SCC size is 10000.
8741 @item ira-max-loops-num
8742 IRA uses a regional register allocation by default. If a function
8743 contains loops more than number given by the parameter, only at most
8744 given number of the most frequently executed loops will form regions
8745 for the regional register allocation. The default value of the
8748 @item ira-max-conflict-table-size
8749 Although IRA uses a sophisticated algorithm of compression conflict
8750 table, the table can be still big for huge functions. If the conflict
8751 table for a function could be more than size in MB given by the
8752 parameter, the conflict table is not built and faster, simpler, and
8753 lower quality register allocation algorithm will be used. The
8754 algorithm do not use pseudo-register conflicts. The default value of
8755 the parameter is 2000.
8757 @item ira-loop-reserved-regs
8758 IRA can be used to evaluate more accurate register pressure in loops
8759 for decision to move loop invariants (see @option{-O3}). The number
8760 of available registers reserved for some other purposes is described
8761 by this parameter. The default value of the parameter is 2 which is
8762 minimal number of registers needed for execution of typical
8763 instruction. This value is the best found from numerous experiments.
8765 @item loop-invariant-max-bbs-in-loop
8766 Loop invariant motion can be very expensive, both in compile time and
8767 in amount of needed compile time memory, with very large loops. Loops
8768 with more basic blocks than this parameter won't have loop invariant
8769 motion optimization performed on them. The default value of the
8770 parameter is 1000 for -O1 and 10000 for -O2 and above.
8772 @item max-vartrack-size
8773 Sets a maximum number of hash table slots to use during variable
8774 tracking dataflow analysis of any function. If this limit is exceeded
8775 with variable tracking at assignments enabled, analysis for that
8776 function is retried without it, after removing all debug insns from
8777 the function. If the limit is exceeded even without debug insns, var
8778 tracking analysis is completely disabled for the function. Setting
8779 the parameter to zero makes it unlimited.
8781 @item min-nondebug-insn-uid
8782 Use uids starting at this parameter for nondebug insns. The range below
8783 the parameter is reserved exclusively for debug insns created by
8784 @option{-fvar-tracking-assignments}, but debug insns may get
8785 (non-overlapping) uids above it if the reserved range is exhausted.
8787 @item ipa-sra-ptr-growth-factor
8788 IPA-SRA will replace a pointer to an aggregate with one or more new
8789 parameters only when their cumulative size is less or equal to
8790 @option{ipa-sra-ptr-growth-factor} times the size of the original
8793 @item graphite-max-nb-scop-params
8794 To avoid exponential effects in the Graphite loop transforms, the
8795 number of parameters in a Static Control Part (SCoP) is bounded. The
8796 default value is 10 parameters. A variable whose value is unknown at
8797 compile time and defined outside a SCoP is a parameter of the SCoP.
8799 @item graphite-max-bbs-per-function
8800 To avoid exponential effects in the detection of SCoPs, the size of
8801 the functions analyzed by Graphite is bounded. The default value is
8804 @item loop-block-tile-size
8805 Loop blocking or strip mining transforms, enabled with
8806 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
8807 loop in the loop nest by a given number of iterations. The strip
8808 length can be changed using the @option{loop-block-tile-size}
8809 parameter. The default value is 51 iterations.
8811 @item devirt-type-list-size
8812 IPA-CP attempts to track all possible types passed to a function's
8813 parameter in order to perform devirtualization.
8814 @option{devirt-type-list-size} is the maximum number of types it
8815 stores per a single formal parameter of a function.
8817 @item lto-partitions
8818 Specify desired nuber of partitions produced during WHOPR copmilation.
8819 Number of partitions should exceed number of CPUs used for compilatoin.
8820 Default value is 32.
8822 @item lto-minpartition
8823 Size of minimal paritition for WHOPR (in estimated instructions).
8824 This prevents expenses of splitting very small programs into too many
8830 @node Preprocessor Options
8831 @section Options Controlling the Preprocessor
8832 @cindex preprocessor options
8833 @cindex options, preprocessor
8835 These options control the C preprocessor, which is run on each C source
8836 file before actual compilation.
8838 If you use the @option{-E} option, nothing is done except preprocessing.
8839 Some of these options make sense only together with @option{-E} because
8840 they cause the preprocessor output to be unsuitable for actual
8844 @item -Wp,@var{option}
8846 You can use @option{-Wp,@var{option}} to bypass the compiler driver
8847 and pass @var{option} directly through to the preprocessor. If
8848 @var{option} contains commas, it is split into multiple options at the
8849 commas. However, many options are modified, translated or interpreted
8850 by the compiler driver before being passed to the preprocessor, and
8851 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
8852 interface is undocumented and subject to change, so whenever possible
8853 you should avoid using @option{-Wp} and let the driver handle the
8856 @item -Xpreprocessor @var{option}
8857 @opindex Xpreprocessor
8858 Pass @var{option} as an option to the preprocessor. You can use this to
8859 supply system-specific preprocessor options which GCC does not know how to
8862 If you want to pass an option that takes an argument, you must use
8863 @option{-Xpreprocessor} twice, once for the option and once for the argument.
8866 @include cppopts.texi
8868 @node Assembler Options
8869 @section Passing Options to the Assembler
8871 @c prevent bad page break with this line
8872 You can pass options to the assembler.
8875 @item -Wa,@var{option}
8877 Pass @var{option} as an option to the assembler. If @var{option}
8878 contains commas, it is split into multiple options at the commas.
8880 @item -Xassembler @var{option}
8882 Pass @var{option} as an option to the assembler. You can use this to
8883 supply system-specific assembler options which GCC does not know how to
8886 If you want to pass an option that takes an argument, you must use
8887 @option{-Xassembler} twice, once for the option and once for the argument.
8892 @section Options for Linking
8893 @cindex link options
8894 @cindex options, linking
8896 These options come into play when the compiler links object files into
8897 an executable output file. They are meaningless if the compiler is
8898 not doing a link step.
8902 @item @var{object-file-name}
8903 A file name that does not end in a special recognized suffix is
8904 considered to name an object file or library. (Object files are
8905 distinguished from libraries by the linker according to the file
8906 contents.) If linking is done, these object files are used as input
8915 If any of these options is used, then the linker is not run, and
8916 object file names should not be used as arguments. @xref{Overall
8920 @item -l@var{library}
8921 @itemx -l @var{library}
8923 Search the library named @var{library} when linking. (The second
8924 alternative with the library as a separate argument is only for
8925 POSIX compliance and is not recommended.)
8927 It makes a difference where in the command you write this option; the
8928 linker searches and processes libraries and object files in the order they
8929 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
8930 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
8931 to functions in @samp{z}, those functions may not be loaded.
8933 The linker searches a standard list of directories for the library,
8934 which is actually a file named @file{lib@var{library}.a}. The linker
8935 then uses this file as if it had been specified precisely by name.
8937 The directories searched include several standard system directories
8938 plus any that you specify with @option{-L}.
8940 Normally the files found this way are library files---archive files
8941 whose members are object files. The linker handles an archive file by
8942 scanning through it for members which define symbols that have so far
8943 been referenced but not defined. But if the file that is found is an
8944 ordinary object file, it is linked in the usual fashion. The only
8945 difference between using an @option{-l} option and specifying a file name
8946 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
8947 and searches several directories.
8951 You need this special case of the @option{-l} option in order to
8952 link an Objective-C or Objective-C++ program.
8955 @opindex nostartfiles
8956 Do not use the standard system startup files when linking.
8957 The standard system libraries are used normally, unless @option{-nostdlib}
8958 or @option{-nodefaultlibs} is used.
8960 @item -nodefaultlibs
8961 @opindex nodefaultlibs
8962 Do not use the standard system libraries when linking.
8963 Only the libraries you specify will be passed to the linker, options
8964 specifying linkage of the system libraries, such as @code{-static-libgcc}
8965 or @code{-shared-libgcc}, will be ignored.
8966 The standard startup files are used normally, unless @option{-nostartfiles}
8967 is used. The compiler may generate calls to @code{memcmp},
8968 @code{memset}, @code{memcpy} and @code{memmove}.
8969 These entries are usually resolved by entries in
8970 libc. These entry points should be supplied through some other
8971 mechanism when this option is specified.
8975 Do not use the standard system startup files or libraries when linking.
8976 No startup files and only the libraries you specify will be passed to
8977 the linker, options specifying linkage of the system libraries, such as
8978 @code{-static-libgcc} or @code{-shared-libgcc}, will be ignored.
8979 The compiler may generate calls to @code{memcmp}, @code{memset},
8980 @code{memcpy} and @code{memmove}.
8981 These entries are usually resolved by entries in
8982 libc. These entry points should be supplied through some other
8983 mechanism when this option is specified.
8985 @cindex @option{-lgcc}, use with @option{-nostdlib}
8986 @cindex @option{-nostdlib} and unresolved references
8987 @cindex unresolved references and @option{-nostdlib}
8988 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
8989 @cindex @option{-nodefaultlibs} and unresolved references
8990 @cindex unresolved references and @option{-nodefaultlibs}
8991 One of the standard libraries bypassed by @option{-nostdlib} and
8992 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
8993 that GCC uses to overcome shortcomings of particular machines, or special
8994 needs for some languages.
8995 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
8996 Collection (GCC) Internals},
8997 for more discussion of @file{libgcc.a}.)
8998 In most cases, you need @file{libgcc.a} even when you want to avoid
8999 other standard libraries. In other words, when you specify @option{-nostdlib}
9000 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
9001 This ensures that you have no unresolved references to internal GCC
9002 library subroutines. (For example, @samp{__main}, used to ensure C++
9003 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
9004 GNU Compiler Collection (GCC) Internals}.)
9008 Produce a position independent executable on targets which support it.
9009 For predictable results, you must also specify the same set of options
9010 that were used to generate code (@option{-fpie}, @option{-fPIE},
9011 or model suboptions) when you specify this option.
9015 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
9016 that support it. This instructs the linker to add all symbols, not
9017 only used ones, to the dynamic symbol table. This option is needed
9018 for some uses of @code{dlopen} or to allow obtaining backtraces
9019 from within a program.
9023 Remove all symbol table and relocation information from the executable.
9027 On systems that support dynamic linking, this prevents linking with the shared
9028 libraries. On other systems, this option has no effect.
9032 Produce a shared object which can then be linked with other objects to
9033 form an executable. Not all systems support this option. For predictable
9034 results, you must also specify the same set of options that were used to
9035 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
9036 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
9037 needs to build supplementary stub code for constructors to work. On
9038 multi-libbed systems, @samp{gcc -shared} must select the correct support
9039 libraries to link against. Failing to supply the correct flags may lead
9040 to subtle defects. Supplying them in cases where they are not necessary
9043 @item -shared-libgcc
9044 @itemx -static-libgcc
9045 @opindex shared-libgcc
9046 @opindex static-libgcc
9047 On systems that provide @file{libgcc} as a shared library, these options
9048 force the use of either the shared or static version respectively.
9049 If no shared version of @file{libgcc} was built when the compiler was
9050 configured, these options have no effect.
9052 There are several situations in which an application should use the
9053 shared @file{libgcc} instead of the static version. The most common
9054 of these is when the application wishes to throw and catch exceptions
9055 across different shared libraries. In that case, each of the libraries
9056 as well as the application itself should use the shared @file{libgcc}.
9058 Therefore, the G++ and GCJ drivers automatically add
9059 @option{-shared-libgcc} whenever you build a shared library or a main
9060 executable, because C++ and Java programs typically use exceptions, so
9061 this is the right thing to do.
9063 If, instead, you use the GCC driver to create shared libraries, you may
9064 find that they will not always be linked with the shared @file{libgcc}.
9065 If GCC finds, at its configuration time, that you have a non-GNU linker
9066 or a GNU linker that does not support option @option{--eh-frame-hdr},
9067 it will link the shared version of @file{libgcc} into shared libraries
9068 by default. Otherwise, it will take advantage of the linker and optimize
9069 away the linking with the shared version of @file{libgcc}, linking with
9070 the static version of libgcc by default. This allows exceptions to
9071 propagate through such shared libraries, without incurring relocation
9072 costs at library load time.
9074 However, if a library or main executable is supposed to throw or catch
9075 exceptions, you must link it using the G++ or GCJ driver, as appropriate
9076 for the languages used in the program, or using the option
9077 @option{-shared-libgcc}, such that it is linked with the shared
9080 @item -static-libstdc++
9081 When the @command{g++} program is used to link a C++ program, it will
9082 normally automatically link against @option{libstdc++}. If
9083 @file{libstdc++} is available as a shared library, and the
9084 @option{-static} option is not used, then this will link against the
9085 shared version of @file{libstdc++}. That is normally fine. However, it
9086 is sometimes useful to freeze the version of @file{libstdc++} used by
9087 the program without going all the way to a fully static link. The
9088 @option{-static-libstdc++} option directs the @command{g++} driver to
9089 link @file{libstdc++} statically, without necessarily linking other
9090 libraries statically.
9094 Bind references to global symbols when building a shared object. Warn
9095 about any unresolved references (unless overridden by the link editor
9096 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
9099 @item -T @var{script}
9101 @cindex linker script
9102 Use @var{script} as the linker script. This option is supported by most
9103 systems using the GNU linker. On some targets, such as bare-board
9104 targets without an operating system, the @option{-T} option may be required
9105 when linking to avoid references to undefined symbols.
9107 @item -Xlinker @var{option}
9109 Pass @var{option} as an option to the linker. You can use this to
9110 supply system-specific linker options which GCC does not know how to
9113 If you want to pass an option that takes a separate argument, you must use
9114 @option{-Xlinker} twice, once for the option and once for the argument.
9115 For example, to pass @option{-assert definitions}, you must write
9116 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
9117 @option{-Xlinker "-assert definitions"}, because this passes the entire
9118 string as a single argument, which is not what the linker expects.
9120 When using the GNU linker, it is usually more convenient to pass
9121 arguments to linker options using the @option{@var{option}=@var{value}}
9122 syntax than as separate arguments. For example, you can specify
9123 @samp{-Xlinker -Map=output.map} rather than
9124 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
9125 this syntax for command-line options.
9127 @item -Wl,@var{option}
9129 Pass @var{option} as an option to the linker. If @var{option} contains
9130 commas, it is split into multiple options at the commas. You can use this
9131 syntax to pass an argument to the option.
9132 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
9133 linker. When using the GNU linker, you can also get the same effect with
9134 @samp{-Wl,-Map=output.map}.
9136 @item -u @var{symbol}
9138 Pretend the symbol @var{symbol} is undefined, to force linking of
9139 library modules to define it. You can use @option{-u} multiple times with
9140 different symbols to force loading of additional library modules.
9143 @node Directory Options
9144 @section Options for Directory Search
9145 @cindex directory options
9146 @cindex options, directory search
9149 These options specify directories to search for header files, for
9150 libraries and for parts of the compiler:
9155 Add the directory @var{dir} to the head of the list of directories to be
9156 searched for header files. This can be used to override a system header
9157 file, substituting your own version, since these directories are
9158 searched before the system header file directories. However, you should
9159 not use this option to add directories that contain vendor-supplied
9160 system header files (use @option{-isystem} for that). If you use more than
9161 one @option{-I} option, the directories are scanned in left-to-right
9162 order; the standard system directories come after.
9164 If a standard system include directory, or a directory specified with
9165 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
9166 option will be ignored. The directory will still be searched but as a
9167 system directory at its normal position in the system include chain.
9168 This is to ensure that GCC's procedure to fix buggy system headers and
9169 the ordering for the include_next directive are not inadvertently changed.
9170 If you really need to change the search order for system directories,
9171 use the @option{-nostdinc} and/or @option{-isystem} options.
9173 @item -iplugindir=@var{dir}
9174 Set the directory to search for plugins which are passed
9175 by @option{-fplugin=@var{name}} instead of
9176 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
9177 to be used by the user, but only passed by the driver.
9179 @item -iquote@var{dir}
9181 Add the directory @var{dir} to the head of the list of directories to
9182 be searched for header files only for the case of @samp{#include
9183 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
9184 otherwise just like @option{-I}.
9188 Add directory @var{dir} to the list of directories to be searched
9191 @item -B@var{prefix}
9193 This option specifies where to find the executables, libraries,
9194 include files, and data files of the compiler itself.
9196 The compiler driver program runs one or more of the subprograms
9197 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
9198 @var{prefix} as a prefix for each program it tries to run, both with and
9199 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
9201 For each subprogram to be run, the compiler driver first tries the
9202 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
9203 was not specified, the driver tries two standard prefixes, which are
9204 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
9205 those results in a file name that is found, the unmodified program
9206 name is searched for using the directories specified in your
9207 @env{PATH} environment variable.
9209 The compiler will check to see if the path provided by the @option{-B}
9210 refers to a directory, and if necessary it will add a directory
9211 separator character at the end of the path.
9213 @option{-B} prefixes that effectively specify directory names also apply
9214 to libraries in the linker, because the compiler translates these
9215 options into @option{-L} options for the linker. They also apply to
9216 includes files in the preprocessor, because the compiler translates these
9217 options into @option{-isystem} options for the preprocessor. In this case,
9218 the compiler appends @samp{include} to the prefix.
9220 The run-time support file @file{libgcc.a} can also be searched for using
9221 the @option{-B} prefix, if needed. If it is not found there, the two
9222 standard prefixes above are tried, and that is all. The file is left
9223 out of the link if it is not found by those means.
9225 Another way to specify a prefix much like the @option{-B} prefix is to use
9226 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
9229 As a special kludge, if the path provided by @option{-B} is
9230 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
9231 9, then it will be replaced by @file{[dir/]include}. This is to help
9232 with boot-strapping the compiler.
9234 @item -specs=@var{file}
9236 Process @var{file} after the compiler reads in the standard @file{specs}
9237 file, in order to override the defaults that the @file{gcc} driver
9238 program uses when determining what switches to pass to @file{cc1},
9239 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
9240 @option{-specs=@var{file}} can be specified on the command line, and they
9241 are processed in order, from left to right.
9243 @item --sysroot=@var{dir}
9245 Use @var{dir} as the logical root directory for headers and libraries.
9246 For example, if the compiler would normally search for headers in
9247 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
9248 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
9250 If you use both this option and the @option{-isysroot} option, then
9251 the @option{--sysroot} option will apply to libraries, but the
9252 @option{-isysroot} option will apply to header files.
9254 The GNU linker (beginning with version 2.16) has the necessary support
9255 for this option. If your linker does not support this option, the
9256 header file aspect of @option{--sysroot} will still work, but the
9257 library aspect will not.
9261 This option has been deprecated. Please use @option{-iquote} instead for
9262 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
9263 Any directories you specify with @option{-I} options before the @option{-I-}
9264 option are searched only for the case of @samp{#include "@var{file}"};
9265 they are not searched for @samp{#include <@var{file}>}.
9267 If additional directories are specified with @option{-I} options after
9268 the @option{-I-}, these directories are searched for all @samp{#include}
9269 directives. (Ordinarily @emph{all} @option{-I} directories are used
9272 In addition, the @option{-I-} option inhibits the use of the current
9273 directory (where the current input file came from) as the first search
9274 directory for @samp{#include "@var{file}"}. There is no way to
9275 override this effect of @option{-I-}. With @option{-I.} you can specify
9276 searching the directory which was current when the compiler was
9277 invoked. That is not exactly the same as what the preprocessor does
9278 by default, but it is often satisfactory.
9280 @option{-I-} does not inhibit the use of the standard system directories
9281 for header files. Thus, @option{-I-} and @option{-nostdinc} are
9288 @section Specifying subprocesses and the switches to pass to them
9291 @command{gcc} is a driver program. It performs its job by invoking a
9292 sequence of other programs to do the work of compiling, assembling and
9293 linking. GCC interprets its command-line parameters and uses these to
9294 deduce which programs it should invoke, and which command-line options
9295 it ought to place on their command lines. This behavior is controlled
9296 by @dfn{spec strings}. In most cases there is one spec string for each
9297 program that GCC can invoke, but a few programs have multiple spec
9298 strings to control their behavior. The spec strings built into GCC can
9299 be overridden by using the @option{-specs=} command-line switch to specify
9302 @dfn{Spec files} are plaintext files that are used to construct spec
9303 strings. They consist of a sequence of directives separated by blank
9304 lines. The type of directive is determined by the first non-whitespace
9305 character on the line and it can be one of the following:
9308 @item %@var{command}
9309 Issues a @var{command} to the spec file processor. The commands that can
9313 @item %include <@var{file}>
9314 @cindex @code{%include}
9315 Search for @var{file} and insert its text at the current point in the
9318 @item %include_noerr <@var{file}>
9319 @cindex @code{%include_noerr}
9320 Just like @samp{%include}, but do not generate an error message if the include
9321 file cannot be found.
9323 @item %rename @var{old_name} @var{new_name}
9324 @cindex @code{%rename}
9325 Rename the spec string @var{old_name} to @var{new_name}.
9329 @item *[@var{spec_name}]:
9330 This tells the compiler to create, override or delete the named spec
9331 string. All lines after this directive up to the next directive or
9332 blank line are considered to be the text for the spec string. If this
9333 results in an empty string then the spec will be deleted. (Or, if the
9334 spec did not exist, then nothing will happened.) Otherwise, if the spec
9335 does not currently exist a new spec will be created. If the spec does
9336 exist then its contents will be overridden by the text of this
9337 directive, unless the first character of that text is the @samp{+}
9338 character, in which case the text will be appended to the spec.
9340 @item [@var{suffix}]:
9341 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
9342 and up to the next directive or blank line are considered to make up the
9343 spec string for the indicated suffix. When the compiler encounters an
9344 input file with the named suffix, it will processes the spec string in
9345 order to work out how to compile that file. For example:
9352 This says that any input file whose name ends in @samp{.ZZ} should be
9353 passed to the program @samp{z-compile}, which should be invoked with the
9354 command-line switch @option{-input} and with the result of performing the
9355 @samp{%i} substitution. (See below.)
9357 As an alternative to providing a spec string, the text that follows a
9358 suffix directive can be one of the following:
9361 @item @@@var{language}
9362 This says that the suffix is an alias for a known @var{language}. This is
9363 similar to using the @option{-x} command-line switch to GCC to specify a
9364 language explicitly. For example:
9371 Says that .ZZ files are, in fact, C++ source files.
9374 This causes an error messages saying:
9377 @var{name} compiler not installed on this system.
9381 GCC already has an extensive list of suffixes built into it.
9382 This directive will add an entry to the end of the list of suffixes, but
9383 since the list is searched from the end backwards, it is effectively
9384 possible to override earlier entries using this technique.
9388 GCC has the following spec strings built into it. Spec files can
9389 override these strings or create their own. Note that individual
9390 targets can also add their own spec strings to this list.
9393 asm Options to pass to the assembler
9394 asm_final Options to pass to the assembler post-processor
9395 cpp Options to pass to the C preprocessor
9396 cc1 Options to pass to the C compiler
9397 cc1plus Options to pass to the C++ compiler
9398 endfile Object files to include at the end of the link
9399 link Options to pass to the linker
9400 lib Libraries to include on the command line to the linker
9401 libgcc Decides which GCC support library to pass to the linker
9402 linker Sets the name of the linker
9403 predefines Defines to be passed to the C preprocessor
9404 signed_char Defines to pass to CPP to say whether @code{char} is signed
9406 startfile Object files to include at the start of the link
9409 Here is a small example of a spec file:
9415 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
9418 This example renames the spec called @samp{lib} to @samp{old_lib} and
9419 then overrides the previous definition of @samp{lib} with a new one.
9420 The new definition adds in some extra command-line options before
9421 including the text of the old definition.
9423 @dfn{Spec strings} are a list of command-line options to be passed to their
9424 corresponding program. In addition, the spec strings can contain
9425 @samp{%}-prefixed sequences to substitute variable text or to
9426 conditionally insert text into the command line. Using these constructs
9427 it is possible to generate quite complex command lines.
9429 Here is a table of all defined @samp{%}-sequences for spec
9430 strings. Note that spaces are not generated automatically around the
9431 results of expanding these sequences. Therefore you can concatenate them
9432 together or combine them with constant text in a single argument.
9436 Substitute one @samp{%} into the program name or argument.
9439 Substitute the name of the input file being processed.
9442 Substitute the basename of the input file being processed.
9443 This is the substring up to (and not including) the last period
9444 and not including the directory.
9447 This is the same as @samp{%b}, but include the file suffix (text after
9451 Marks the argument containing or following the @samp{%d} as a
9452 temporary file name, so that that file will be deleted if GCC exits
9453 successfully. Unlike @samp{%g}, this contributes no text to the
9456 @item %g@var{suffix}
9457 Substitute a file name that has suffix @var{suffix} and is chosen
9458 once per compilation, and mark the argument in the same way as
9459 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
9460 name is now chosen in a way that is hard to predict even when previously
9461 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
9462 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
9463 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
9464 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
9465 was simply substituted with a file name chosen once per compilation,
9466 without regard to any appended suffix (which was therefore treated
9467 just like ordinary text), making such attacks more likely to succeed.
9469 @item %u@var{suffix}
9470 Like @samp{%g}, but generates a new temporary file name even if
9471 @samp{%u@var{suffix}} was already seen.
9473 @item %U@var{suffix}
9474 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
9475 new one if there is no such last file name. In the absence of any
9476 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
9477 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
9478 would involve the generation of two distinct file names, one
9479 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
9480 simply substituted with a file name chosen for the previous @samp{%u},
9481 without regard to any appended suffix.
9483 @item %j@var{suffix}
9484 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
9485 writable, and if save-temps is off; otherwise, substitute the name
9486 of a temporary file, just like @samp{%u}. This temporary file is not
9487 meant for communication between processes, but rather as a junk
9490 @item %|@var{suffix}
9491 @itemx %m@var{suffix}
9492 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
9493 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
9494 all. These are the two most common ways to instruct a program that it
9495 should read from standard input or write to standard output. If you
9496 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
9497 construct: see for example @file{f/lang-specs.h}.
9499 @item %.@var{SUFFIX}
9500 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
9501 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
9502 terminated by the next space or %.
9505 Marks the argument containing or following the @samp{%w} as the
9506 designated output file of this compilation. This puts the argument
9507 into the sequence of arguments that @samp{%o} will substitute later.
9510 Substitutes the names of all the output files, with spaces
9511 automatically placed around them. You should write spaces
9512 around the @samp{%o} as well or the results are undefined.
9513 @samp{%o} is for use in the specs for running the linker.
9514 Input files whose names have no recognized suffix are not compiled
9515 at all, but they are included among the output files, so they will
9519 Substitutes the suffix for object files. Note that this is
9520 handled specially when it immediately follows @samp{%g, %u, or %U},
9521 because of the need for those to form complete file names. The
9522 handling is such that @samp{%O} is treated exactly as if it had already
9523 been substituted, except that @samp{%g, %u, and %U} do not currently
9524 support additional @var{suffix} characters following @samp{%O} as they would
9525 following, for example, @samp{.o}.
9528 Substitutes the standard macro predefinitions for the
9529 current target machine. Use this when running @code{cpp}.
9532 Like @samp{%p}, but puts @samp{__} before and after the name of each
9533 predefined macro, except for macros that start with @samp{__} or with
9534 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
9538 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
9539 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
9540 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
9541 and @option{-imultilib} as necessary.
9544 Current argument is the name of a library or startup file of some sort.
9545 Search for that file in a standard list of directories and substitute
9546 the full name found. The current working directory is included in the
9547 list of directories scanned.
9550 Current argument is the name of a linker script. Search for that file
9551 in the current list of directories to scan for libraries. If the file
9552 is located insert a @option{--script} option into the command line
9553 followed by the full path name found. If the file is not found then
9554 generate an error message. Note: the current working directory is not
9558 Print @var{str} as an error message. @var{str} is terminated by a newline.
9559 Use this when inconsistent options are detected.
9562 Substitute the contents of spec string @var{name} at this point.
9565 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
9567 @item %x@{@var{option}@}
9568 Accumulate an option for @samp{%X}.
9571 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
9575 Output the accumulated assembler options specified by @option{-Wa}.
9578 Output the accumulated preprocessor options specified by @option{-Wp}.
9581 Process the @code{asm} spec. This is used to compute the
9582 switches to be passed to the assembler.
9585 Process the @code{asm_final} spec. This is a spec string for
9586 passing switches to an assembler post-processor, if such a program is
9590 Process the @code{link} spec. This is the spec for computing the
9591 command line passed to the linker. Typically it will make use of the
9592 @samp{%L %G %S %D and %E} sequences.
9595 Dump out a @option{-L} option for each directory that GCC believes might
9596 contain startup files. If the target supports multilibs then the
9597 current multilib directory will be prepended to each of these paths.
9600 Process the @code{lib} spec. This is a spec string for deciding which
9601 libraries should be included on the command line to the linker.
9604 Process the @code{libgcc} spec. This is a spec string for deciding
9605 which GCC support library should be included on the command line to the linker.
9608 Process the @code{startfile} spec. This is a spec for deciding which
9609 object files should be the first ones passed to the linker. Typically
9610 this might be a file named @file{crt0.o}.
9613 Process the @code{endfile} spec. This is a spec string that specifies
9614 the last object files that will be passed to the linker.
9617 Process the @code{cpp} spec. This is used to construct the arguments
9618 to be passed to the C preprocessor.
9621 Process the @code{cc1} spec. This is used to construct the options to be
9622 passed to the actual C compiler (@samp{cc1}).
9625 Process the @code{cc1plus} spec. This is used to construct the options to be
9626 passed to the actual C++ compiler (@samp{cc1plus}).
9629 Substitute the variable part of a matched option. See below.
9630 Note that each comma in the substituted string is replaced by
9634 Remove all occurrences of @code{-S} from the command line. Note---this
9635 command is position dependent. @samp{%} commands in the spec string
9636 before this one will see @code{-S}, @samp{%} commands in the spec string
9637 after this one will not.
9639 @item %:@var{function}(@var{args})
9640 Call the named function @var{function}, passing it @var{args}.
9641 @var{args} is first processed as a nested spec string, then split
9642 into an argument vector in the usual fashion. The function returns
9643 a string which is processed as if it had appeared literally as part
9644 of the current spec.
9646 The following built-in spec functions are provided:
9650 The @code{getenv} spec function takes two arguments: an environment
9651 variable name and a string. If the environment variable is not
9652 defined, a fatal error is issued. Otherwise, the return value is the
9653 value of the environment variable concatenated with the string. For
9654 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
9657 %:getenv(TOPDIR /include)
9660 expands to @file{/path/to/top/include}.
9662 @item @code{if-exists}
9663 The @code{if-exists} spec function takes one argument, an absolute
9664 pathname to a file. If the file exists, @code{if-exists} returns the
9665 pathname. Here is a small example of its usage:
9669 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
9672 @item @code{if-exists-else}
9673 The @code{if-exists-else} spec function is similar to the @code{if-exists}
9674 spec function, except that it takes two arguments. The first argument is
9675 an absolute pathname to a file. If the file exists, @code{if-exists-else}
9676 returns the pathname. If it does not exist, it returns the second argument.
9677 This way, @code{if-exists-else} can be used to select one file or another,
9678 based on the existence of the first. Here is a small example of its usage:
9682 crt0%O%s %:if-exists(crti%O%s) \
9683 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
9686 @item @code{replace-outfile}
9687 The @code{replace-outfile} spec function takes two arguments. It looks for the
9688 first argument in the outfiles array and replaces it with the second argument. Here
9689 is a small example of its usage:
9692 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
9695 @item @code{remove-outfile}
9696 The @code{remove-outfile} spec function takes one argument. It looks for the
9697 first argument in the outfiles array and removes it. Here is a small example
9701 %:remove-outfile(-lm)
9704 @item @code{print-asm-header}
9705 The @code{print-asm-header} function takes no arguments and simply
9706 prints a banner like:
9712 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
9715 It is used to separate compiler options from assembler options
9716 in the @option{--target-help} output.
9720 Substitutes the @code{-S} switch, if that switch was given to GCC@.
9721 If that switch was not specified, this substitutes nothing. Note that
9722 the leading dash is omitted when specifying this option, and it is
9723 automatically inserted if the substitution is performed. Thus the spec
9724 string @samp{%@{foo@}} would match the command-line option @option{-foo}
9725 and would output the command line option @option{-foo}.
9727 @item %W@{@code{S}@}
9728 Like %@{@code{S}@} but mark last argument supplied within as a file to be
9731 @item %@{@code{S}*@}
9732 Substitutes all the switches specified to GCC whose names start
9733 with @code{-S}, but which also take an argument. This is used for
9734 switches like @option{-o}, @option{-D}, @option{-I}, etc.
9735 GCC considers @option{-o foo} as being
9736 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
9737 text, including the space. Thus two arguments would be generated.
9739 @item %@{@code{S}*&@code{T}*@}
9740 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
9741 (the order of @code{S} and @code{T} in the spec is not significant).
9742 There can be any number of ampersand-separated variables; for each the
9743 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
9745 @item %@{@code{S}:@code{X}@}
9746 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
9748 @item %@{!@code{S}:@code{X}@}
9749 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
9751 @item %@{@code{S}*:@code{X}@}
9752 Substitutes @code{X} if one or more switches whose names start with
9753 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
9754 once, no matter how many such switches appeared. However, if @code{%*}
9755 appears somewhere in @code{X}, then @code{X} will be substituted once
9756 for each matching switch, with the @code{%*} replaced by the part of
9757 that switch that matched the @code{*}.
9759 @item %@{.@code{S}:@code{X}@}
9760 Substitutes @code{X}, if processing a file with suffix @code{S}.
9762 @item %@{!.@code{S}:@code{X}@}
9763 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
9765 @item %@{,@code{S}:@code{X}@}
9766 Substitutes @code{X}, if processing a file for language @code{S}.
9768 @item %@{!,@code{S}:@code{X}@}
9769 Substitutes @code{X}, if not processing a file for language @code{S}.
9771 @item %@{@code{S}|@code{P}:@code{X}@}
9772 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
9773 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
9774 @code{*} sequences as well, although they have a stronger binding than
9775 the @samp{|}. If @code{%*} appears in @code{X}, all of the
9776 alternatives must be starred, and only the first matching alternative
9779 For example, a spec string like this:
9782 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
9785 will output the following command-line options from the following input
9786 command-line options:
9791 -d fred.c -foo -baz -boggle
9792 -d jim.d -bar -baz -boggle
9795 @item %@{S:X; T:Y; :D@}
9797 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
9798 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
9799 be as many clauses as you need. This may be combined with @code{.},
9800 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
9805 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
9806 construct may contain other nested @samp{%} constructs or spaces, or
9807 even newlines. They are processed as usual, as described above.
9808 Trailing white space in @code{X} is ignored. White space may also
9809 appear anywhere on the left side of the colon in these constructs,
9810 except between @code{.} or @code{*} and the corresponding word.
9812 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
9813 handled specifically in these constructs. If another value of
9814 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
9815 @option{-W} switch is found later in the command line, the earlier
9816 switch value is ignored, except with @{@code{S}*@} where @code{S} is
9817 just one letter, which passes all matching options.
9819 The character @samp{|} at the beginning of the predicate text is used to
9820 indicate that a command should be piped to the following command, but
9821 only if @option{-pipe} is specified.
9823 It is built into GCC which switches take arguments and which do not.
9824 (You might think it would be useful to generalize this to allow each
9825 compiler's spec to say which switches take arguments. But this cannot
9826 be done in a consistent fashion. GCC cannot even decide which input
9827 files have been specified without knowing which switches take arguments,
9828 and it must know which input files to compile in order to tell which
9831 GCC also knows implicitly that arguments starting in @option{-l} are to be
9832 treated as compiler output files, and passed to the linker in their
9833 proper position among the other output files.
9835 @c man begin OPTIONS
9837 @node Target Options
9838 @section Specifying Target Machine and Compiler Version
9839 @cindex target options
9840 @cindex cross compiling
9841 @cindex specifying machine version
9842 @cindex specifying compiler version and target machine
9843 @cindex compiler version, specifying
9844 @cindex target machine, specifying
9846 The usual way to run GCC is to run the executable called @file{gcc}, or
9847 @file{<machine>-gcc} when cross-compiling, or
9848 @file{<machine>-gcc-<version>} to run a version other than the one that
9851 @node Submodel Options
9852 @section Hardware Models and Configurations
9853 @cindex submodel options
9854 @cindex specifying hardware config
9855 @cindex hardware models and configurations, specifying
9856 @cindex machine dependent options
9858 Each target machine types can have its own
9859 special options, starting with @samp{-m}, to choose among various
9860 hardware models or configurations---for example, 68010 vs 68020,
9861 floating coprocessor or none. A single installed version of the
9862 compiler can compile for any model or configuration, according to the
9865 Some configurations of the compiler also support additional special
9866 options, usually for compatibility with other compilers on the same
9869 @c This list is ordered alphanumerically by subsection name.
9870 @c It should be the same order and spelling as these options are listed
9871 @c in Machine Dependent Options
9877 * Blackfin Options::
9881 * DEC Alpha Options::
9882 * DEC Alpha/VMS Options::
9885 * GNU/Linux Options::
9888 * i386 and x86-64 Options::
9889 * i386 and x86-64 Windows Options::
9891 * IA-64/VMS Options::
9899 * MicroBlaze Options::
9904 * picoChip Options::
9906 * RS/6000 and PowerPC Options::
9908 * S/390 and zSeries Options::
9911 * Solaris 2 Options::
9914 * System V Options::
9919 * Xstormy16 Options::
9925 @subsection ARC Options
9928 These options are defined for ARC implementations:
9933 Compile code for little endian mode. This is the default.
9937 Compile code for big endian mode.
9940 @opindex mmangle-cpu
9941 Prepend the name of the cpu to all public symbol names.
9942 In multiple-processor systems, there are many ARC variants with different
9943 instruction and register set characteristics. This flag prevents code
9944 compiled for one cpu to be linked with code compiled for another.
9945 No facility exists for handling variants that are ``almost identical''.
9946 This is an all or nothing option.
9948 @item -mcpu=@var{cpu}
9950 Compile code for ARC variant @var{cpu}.
9951 Which variants are supported depend on the configuration.
9952 All variants support @option{-mcpu=base}, this is the default.
9954 @item -mtext=@var{text-section}
9955 @itemx -mdata=@var{data-section}
9956 @itemx -mrodata=@var{readonly-data-section}
9960 Put functions, data, and readonly data in @var{text-section},
9961 @var{data-section}, and @var{readonly-data-section} respectively
9962 by default. This can be overridden with the @code{section} attribute.
9963 @xref{Variable Attributes}.
9968 @subsection ARM Options
9971 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
9975 @item -mabi=@var{name}
9977 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
9978 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
9981 @opindex mapcs-frame
9982 Generate a stack frame that is compliant with the ARM Procedure Call
9983 Standard for all functions, even if this is not strictly necessary for
9984 correct execution of the code. Specifying @option{-fomit-frame-pointer}
9985 with this option will cause the stack frames not to be generated for
9986 leaf functions. The default is @option{-mno-apcs-frame}.
9990 This is a synonym for @option{-mapcs-frame}.
9993 @c not currently implemented
9994 @item -mapcs-stack-check
9995 @opindex mapcs-stack-check
9996 Generate code to check the amount of stack space available upon entry to
9997 every function (that actually uses some stack space). If there is
9998 insufficient space available then either the function
9999 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
10000 called, depending upon the amount of stack space required. The run time
10001 system is required to provide these functions. The default is
10002 @option{-mno-apcs-stack-check}, since this produces smaller code.
10004 @c not currently implemented
10006 @opindex mapcs-float
10007 Pass floating point arguments using the float point registers. This is
10008 one of the variants of the APCS@. This option is recommended if the
10009 target hardware has a floating point unit or if a lot of floating point
10010 arithmetic is going to be performed by the code. The default is
10011 @option{-mno-apcs-float}, since integer only code is slightly increased in
10012 size if @option{-mapcs-float} is used.
10014 @c not currently implemented
10015 @item -mapcs-reentrant
10016 @opindex mapcs-reentrant
10017 Generate reentrant, position independent code. The default is
10018 @option{-mno-apcs-reentrant}.
10021 @item -mthumb-interwork
10022 @opindex mthumb-interwork
10023 Generate code which supports calling between the ARM and Thumb
10024 instruction sets. Without this option the two instruction sets cannot
10025 be reliably used inside one program. The default is
10026 @option{-mno-thumb-interwork}, since slightly larger code is generated
10027 when @option{-mthumb-interwork} is specified.
10029 @item -mno-sched-prolog
10030 @opindex mno-sched-prolog
10031 Prevent the reordering of instructions in the function prolog, or the
10032 merging of those instruction with the instructions in the function's
10033 body. This means that all functions will start with a recognizable set
10034 of instructions (or in fact one of a choice from a small set of
10035 different function prologues), and this information can be used to
10036 locate the start if functions inside an executable piece of code. The
10037 default is @option{-msched-prolog}.
10039 @item -mfloat-abi=@var{name}
10040 @opindex mfloat-abi
10041 Specifies which floating-point ABI to use. Permissible values
10042 are: @samp{soft}, @samp{softfp} and @samp{hard}.
10044 Specifying @samp{soft} causes GCC to generate output containing
10045 library calls for floating-point operations.
10046 @samp{softfp} allows the generation of code using hardware floating-point
10047 instructions, but still uses the soft-float calling conventions.
10048 @samp{hard} allows generation of floating-point instructions
10049 and uses FPU-specific calling conventions.
10051 The default depends on the specific target configuration. Note that
10052 the hard-float and soft-float ABIs are not link-compatible; you must
10053 compile your entire program with the same ABI, and link with a
10054 compatible set of libraries.
10057 @opindex mhard-float
10058 Equivalent to @option{-mfloat-abi=hard}.
10061 @opindex msoft-float
10062 Equivalent to @option{-mfloat-abi=soft}.
10064 @item -mlittle-endian
10065 @opindex mlittle-endian
10066 Generate code for a processor running in little-endian mode. This is
10067 the default for all standard configurations.
10070 @opindex mbig-endian
10071 Generate code for a processor running in big-endian mode; the default is
10072 to compile code for a little-endian processor.
10074 @item -mwords-little-endian
10075 @opindex mwords-little-endian
10076 This option only applies when generating code for big-endian processors.
10077 Generate code for a little-endian word order but a big-endian byte
10078 order. That is, a byte order of the form @samp{32107654}. Note: this
10079 option should only be used if you require compatibility with code for
10080 big-endian ARM processors generated by versions of the compiler prior to
10083 @item -mcpu=@var{name}
10085 This specifies the name of the target ARM processor. GCC uses this name
10086 to determine what kind of instructions it can emit when generating
10087 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
10088 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
10089 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
10090 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
10091 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
10093 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
10094 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
10095 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
10096 @samp{strongarm1110},
10097 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
10098 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
10099 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
10100 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
10101 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
10102 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
10103 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
10104 @samp{cortex-a5}, @samp{cortex-a8}, @samp{cortex-a9}, @samp{cortex-a15},
10105 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-m4}, @samp{cortex-m3},
10108 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
10110 @item -mtune=@var{name}
10112 This option is very similar to the @option{-mcpu=} option, except that
10113 instead of specifying the actual target processor type, and hence
10114 restricting which instructions can be used, it specifies that GCC should
10115 tune the performance of the code as if the target were of the type
10116 specified in this option, but still choosing the instructions that it
10117 will generate based on the cpu specified by a @option{-mcpu=} option.
10118 For some ARM implementations better performance can be obtained by using
10121 @item -march=@var{name}
10123 This specifies the name of the target ARM architecture. GCC uses this
10124 name to determine what kind of instructions it can emit when generating
10125 assembly code. This option can be used in conjunction with or instead
10126 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
10127 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
10128 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
10129 @samp{armv6}, @samp{armv6j},
10130 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
10131 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
10132 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
10134 @item -mfpu=@var{name}
10135 @itemx -mfpe=@var{number}
10136 @itemx -mfp=@var{number}
10140 This specifies what floating point hardware (or hardware emulation) is
10141 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
10142 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-fp16},
10143 @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd}, @samp{vfpv3xd-fp16},
10144 @samp{neon}, @samp{neon-fp16}, @samp{vfpv4}, @samp{vfpv4-d16},
10145 @samp{fpv4-sp-d16} and @samp{neon-vfpv4}.
10146 @option{-mfp} and @option{-mfpe} are synonyms for
10147 @option{-mfpu}=@samp{fpe}@var{number}, for compatibility with older versions
10150 If @option{-msoft-float} is specified this specifies the format of
10151 floating point values.
10153 If the selected floating-point hardware includes the NEON extension
10154 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
10155 operations will not be used by GCC's auto-vectorization pass unless
10156 @option{-funsafe-math-optimizations} is also specified. This is
10157 because NEON hardware does not fully implement the IEEE 754 standard for
10158 floating-point arithmetic (in particular denormal values are treated as
10159 zero), so the use of NEON instructions may lead to a loss of precision.
10161 @item -mfp16-format=@var{name}
10162 @opindex mfp16-format
10163 Specify the format of the @code{__fp16} half-precision floating-point type.
10164 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
10165 the default is @samp{none}, in which case the @code{__fp16} type is not
10166 defined. @xref{Half-Precision}, for more information.
10168 @item -mstructure-size-boundary=@var{n}
10169 @opindex mstructure-size-boundary
10170 The size of all structures and unions will be rounded up to a multiple
10171 of the number of bits set by this option. Permissible values are 8, 32
10172 and 64. The default value varies for different toolchains. For the COFF
10173 targeted toolchain the default value is 8. A value of 64 is only allowed
10174 if the underlying ABI supports it.
10176 Specifying the larger number can produce faster, more efficient code, but
10177 can also increase the size of the program. Different values are potentially
10178 incompatible. Code compiled with one value cannot necessarily expect to
10179 work with code or libraries compiled with another value, if they exchange
10180 information using structures or unions.
10182 @item -mabort-on-noreturn
10183 @opindex mabort-on-noreturn
10184 Generate a call to the function @code{abort} at the end of a
10185 @code{noreturn} function. It will be executed if the function tries to
10189 @itemx -mno-long-calls
10190 @opindex mlong-calls
10191 @opindex mno-long-calls
10192 Tells the compiler to perform function calls by first loading the
10193 address of the function into a register and then performing a subroutine
10194 call on this register. This switch is needed if the target function
10195 will lie outside of the 64 megabyte addressing range of the offset based
10196 version of subroutine call instruction.
10198 Even if this switch is enabled, not all function calls will be turned
10199 into long calls. The heuristic is that static functions, functions
10200 which have the @samp{short-call} attribute, functions that are inside
10201 the scope of a @samp{#pragma no_long_calls} directive and functions whose
10202 definitions have already been compiled within the current compilation
10203 unit, will not be turned into long calls. The exception to this rule is
10204 that weak function definitions, functions with the @samp{long-call}
10205 attribute or the @samp{section} attribute, and functions that are within
10206 the scope of a @samp{#pragma long_calls} directive, will always be
10207 turned into long calls.
10209 This feature is not enabled by default. Specifying
10210 @option{-mno-long-calls} will restore the default behavior, as will
10211 placing the function calls within the scope of a @samp{#pragma
10212 long_calls_off} directive. Note these switches have no effect on how
10213 the compiler generates code to handle function calls via function
10216 @item -msingle-pic-base
10217 @opindex msingle-pic-base
10218 Treat the register used for PIC addressing as read-only, rather than
10219 loading it in the prologue for each function. The run-time system is
10220 responsible for initializing this register with an appropriate value
10221 before execution begins.
10223 @item -mpic-register=@var{reg}
10224 @opindex mpic-register
10225 Specify the register to be used for PIC addressing. The default is R10
10226 unless stack-checking is enabled, when R9 is used.
10228 @item -mcirrus-fix-invalid-insns
10229 @opindex mcirrus-fix-invalid-insns
10230 @opindex mno-cirrus-fix-invalid-insns
10231 Insert NOPs into the instruction stream to in order to work around
10232 problems with invalid Maverick instruction combinations. This option
10233 is only valid if the @option{-mcpu=ep9312} option has been used to
10234 enable generation of instructions for the Cirrus Maverick floating
10235 point co-processor. This option is not enabled by default, since the
10236 problem is only present in older Maverick implementations. The default
10237 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
10240 @item -mpoke-function-name
10241 @opindex mpoke-function-name
10242 Write the name of each function into the text section, directly
10243 preceding the function prologue. The generated code is similar to this:
10247 .ascii "arm_poke_function_name", 0
10250 .word 0xff000000 + (t1 - t0)
10251 arm_poke_function_name
10253 stmfd sp!, @{fp, ip, lr, pc@}
10257 When performing a stack backtrace, code can inspect the value of
10258 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
10259 location @code{pc - 12} and the top 8 bits are set, then we know that
10260 there is a function name embedded immediately preceding this location
10261 and has length @code{((pc[-3]) & 0xff000000)}.
10265 Generate code for the Thumb instruction set. The default is to
10266 use the 32-bit ARM instruction set.
10267 This option automatically enables either 16-bit Thumb-1 or
10268 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
10269 and @option{-march=@var{name}} options. This option is not passed to the
10270 assembler. If you want to force assembler files to be interpreted as Thumb code,
10271 either add a @samp{.thumb} directive to the source or pass the @option{-mthumb}
10272 option directly to the assembler by prefixing it with @option{-Wa}.
10275 @opindex mtpcs-frame
10276 Generate a stack frame that is compliant with the Thumb Procedure Call
10277 Standard for all non-leaf functions. (A leaf function is one that does
10278 not call any other functions.) The default is @option{-mno-tpcs-frame}.
10280 @item -mtpcs-leaf-frame
10281 @opindex mtpcs-leaf-frame
10282 Generate a stack frame that is compliant with the Thumb Procedure Call
10283 Standard for all leaf functions. (A leaf function is one that does
10284 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
10286 @item -mcallee-super-interworking
10287 @opindex mcallee-super-interworking
10288 Gives all externally visible functions in the file being compiled an ARM
10289 instruction set header which switches to Thumb mode before executing the
10290 rest of the function. This allows these functions to be called from
10291 non-interworking code. This option is not valid in AAPCS configurations
10292 because interworking is enabled by default.
10294 @item -mcaller-super-interworking
10295 @opindex mcaller-super-interworking
10296 Allows calls via function pointers (including virtual functions) to
10297 execute correctly regardless of whether the target code has been
10298 compiled for interworking or not. There is a small overhead in the cost
10299 of executing a function pointer if this option is enabled. This option
10300 is not valid in AAPCS configurations because interworking is enabled
10303 @item -mtp=@var{name}
10305 Specify the access model for the thread local storage pointer. The valid
10306 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
10307 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
10308 (supported in the arm6k architecture), and @option{auto}, which uses the
10309 best available method for the selected processor. The default setting is
10312 @item -mword-relocations
10313 @opindex mword-relocations
10314 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
10315 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
10316 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
10319 @item -mfix-cortex-m3-ldrd
10320 @opindex mfix-cortex-m3-ldrd
10321 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
10322 with overlapping destination and base registers are used. This option avoids
10323 generating these instructions. This option is enabled by default when
10324 @option{-mcpu=cortex-m3} is specified.
10329 @subsection AVR Options
10330 @cindex AVR Options
10332 These options are defined for AVR implementations:
10335 @item -mmcu=@var{mcu}
10337 Specify ATMEL AVR instruction set or MCU type.
10339 Instruction set avr1 is for the minimal AVR core, not supported by the C
10340 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
10341 attiny11, attiny12, attiny15, attiny28).
10343 Instruction set avr2 (default) is for the classic AVR core with up to
10344 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
10345 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
10346 at90c8534, at90s8535).
10348 Instruction set avr3 is for the classic AVR core with up to 128K program
10349 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
10351 Instruction set avr4 is for the enhanced AVR core with up to 8K program
10352 memory space (MCU types: atmega8, atmega83, atmega85).
10354 Instruction set avr5 is for the enhanced AVR core with up to 128K program
10355 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
10356 atmega64, atmega128, at43usb355, at94k).
10358 @item -mno-interrupts
10359 @opindex mno-interrupts
10360 Generated code is not compatible with hardware interrupts.
10361 Code size will be smaller.
10363 @item -mcall-prologues
10364 @opindex mcall-prologues
10365 Functions prologues/epilogues expanded as call to appropriate
10366 subroutines. Code size will be smaller.
10369 @opindex mtiny-stack
10370 Change only the low 8 bits of the stack pointer.
10374 Assume int to be 8 bit integer. This affects the sizes of all types: A
10375 char will be 1 byte, an int will be 1 byte, a long will be 2 bytes
10376 and long long will be 4 bytes. Please note that this option does not
10377 comply to the C standards, but it will provide you with smaller code
10381 @node Blackfin Options
10382 @subsection Blackfin Options
10383 @cindex Blackfin Options
10386 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
10388 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
10389 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
10390 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
10391 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
10392 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
10393 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
10394 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
10396 The optional @var{sirevision} specifies the silicon revision of the target
10397 Blackfin processor. Any workarounds available for the targeted silicon revision
10398 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
10399 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
10400 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
10401 hexadecimal digits representing the major and minor numbers in the silicon
10402 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
10403 is not defined. If @var{sirevision} is @samp{any}, the
10404 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
10405 If this optional @var{sirevision} is not used, GCC assumes the latest known
10406 silicon revision of the targeted Blackfin processor.
10408 Support for @samp{bf561} is incomplete. For @samp{bf561},
10409 Only the processor macro is defined.
10410 Without this option, @samp{bf532} is used as the processor by default.
10411 The corresponding predefined processor macros for @var{cpu} is to
10412 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
10413 provided by libgloss to be linked in if @option{-msim} is not given.
10417 Specifies that the program will be run on the simulator. This causes
10418 the simulator BSP provided by libgloss to be linked in. This option
10419 has effect only for @samp{bfin-elf} toolchain.
10420 Certain other options, such as @option{-mid-shared-library} and
10421 @option{-mfdpic}, imply @option{-msim}.
10423 @item -momit-leaf-frame-pointer
10424 @opindex momit-leaf-frame-pointer
10425 Don't keep the frame pointer in a register for leaf functions. This
10426 avoids the instructions to save, set up and restore frame pointers and
10427 makes an extra register available in leaf functions. The option
10428 @option{-fomit-frame-pointer} removes the frame pointer for all functions
10429 which might make debugging harder.
10431 @item -mspecld-anomaly
10432 @opindex mspecld-anomaly
10433 When enabled, the compiler will ensure that the generated code does not
10434 contain speculative loads after jump instructions. If this option is used,
10435 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
10437 @item -mno-specld-anomaly
10438 @opindex mno-specld-anomaly
10439 Don't generate extra code to prevent speculative loads from occurring.
10441 @item -mcsync-anomaly
10442 @opindex mcsync-anomaly
10443 When enabled, the compiler will ensure that the generated code does not
10444 contain CSYNC or SSYNC instructions too soon after conditional branches.
10445 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
10447 @item -mno-csync-anomaly
10448 @opindex mno-csync-anomaly
10449 Don't generate extra code to prevent CSYNC or SSYNC instructions from
10450 occurring too soon after a conditional branch.
10454 When enabled, the compiler is free to take advantage of the knowledge that
10455 the entire program fits into the low 64k of memory.
10458 @opindex mno-low-64k
10459 Assume that the program is arbitrarily large. This is the default.
10461 @item -mstack-check-l1
10462 @opindex mstack-check-l1
10463 Do stack checking using information placed into L1 scratchpad memory by the
10466 @item -mid-shared-library
10467 @opindex mid-shared-library
10468 Generate code that supports shared libraries via the library ID method.
10469 This allows for execute in place and shared libraries in an environment
10470 without virtual memory management. This option implies @option{-fPIC}.
10471 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10473 @item -mno-id-shared-library
10474 @opindex mno-id-shared-library
10475 Generate code that doesn't assume ID based shared libraries are being used.
10476 This is the default.
10478 @item -mleaf-id-shared-library
10479 @opindex mleaf-id-shared-library
10480 Generate code that supports shared libraries via the library ID method,
10481 but assumes that this library or executable won't link against any other
10482 ID shared libraries. That allows the compiler to use faster code for jumps
10485 @item -mno-leaf-id-shared-library
10486 @opindex mno-leaf-id-shared-library
10487 Do not assume that the code being compiled won't link against any ID shared
10488 libraries. Slower code will be generated for jump and call insns.
10490 @item -mshared-library-id=n
10491 @opindex mshared-library-id
10492 Specified the identification number of the ID based shared library being
10493 compiled. Specifying a value of 0 will generate more compact code, specifying
10494 other values will force the allocation of that number to the current
10495 library but is no more space or time efficient than omitting this option.
10499 Generate code that allows the data segment to be located in a different
10500 area of memory from the text segment. This allows for execute in place in
10501 an environment without virtual memory management by eliminating relocations
10502 against the text section.
10504 @item -mno-sep-data
10505 @opindex mno-sep-data
10506 Generate code that assumes that the data segment follows the text segment.
10507 This is the default.
10510 @itemx -mno-long-calls
10511 @opindex mlong-calls
10512 @opindex mno-long-calls
10513 Tells the compiler to perform function calls by first loading the
10514 address of the function into a register and then performing a subroutine
10515 call on this register. This switch is needed if the target function
10516 will lie outside of the 24 bit addressing range of the offset based
10517 version of subroutine call instruction.
10519 This feature is not enabled by default. Specifying
10520 @option{-mno-long-calls} will restore the default behavior. Note these
10521 switches have no effect on how the compiler generates code to handle
10522 function calls via function pointers.
10526 Link with the fast floating-point library. This library relaxes some of
10527 the IEEE floating-point standard's rules for checking inputs against
10528 Not-a-Number (NAN), in the interest of performance.
10531 @opindex minline-plt
10532 Enable inlining of PLT entries in function calls to functions that are
10533 not known to bind locally. It has no effect without @option{-mfdpic}.
10536 @opindex mmulticore
10537 Build standalone application for multicore Blackfin processor. Proper
10538 start files and link scripts will be used to support multicore.
10539 This option defines @code{__BFIN_MULTICORE}. It can only be used with
10540 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
10541 @option{-mcorea} or @option{-mcoreb}. If it's used without
10542 @option{-mcorea} or @option{-mcoreb}, single application/dual core
10543 programming model is used. In this model, the main function of Core B
10544 should be named as coreb_main. If it's used with @option{-mcorea} or
10545 @option{-mcoreb}, one application per core programming model is used.
10546 If this option is not used, single core application programming
10551 Build standalone application for Core A of BF561 when using
10552 one application per core programming model. Proper start files
10553 and link scripts will be used to support Core A. This option
10554 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
10558 Build standalone application for Core B of BF561 when using
10559 one application per core programming model. Proper start files
10560 and link scripts will be used to support Core B. This option
10561 defines @code{__BFIN_COREB}. When this option is used, coreb_main
10562 should be used instead of main. It must be used with
10563 @option{-mmulticore}.
10567 Build standalone application for SDRAM. Proper start files and
10568 link scripts will be used to put the application into SDRAM.
10569 Loader should initialize SDRAM before loading the application
10570 into SDRAM. This option defines @code{__BFIN_SDRAM}.
10574 Assume that ICPLBs are enabled at runtime. This has an effect on certain
10575 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
10576 are enabled; for standalone applications the default is off.
10580 @subsection CRIS Options
10581 @cindex CRIS Options
10583 These options are defined specifically for the CRIS ports.
10586 @item -march=@var{architecture-type}
10587 @itemx -mcpu=@var{architecture-type}
10590 Generate code for the specified architecture. The choices for
10591 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
10592 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
10593 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
10596 @item -mtune=@var{architecture-type}
10598 Tune to @var{architecture-type} everything applicable about the generated
10599 code, except for the ABI and the set of available instructions. The
10600 choices for @var{architecture-type} are the same as for
10601 @option{-march=@var{architecture-type}}.
10603 @item -mmax-stack-frame=@var{n}
10604 @opindex mmax-stack-frame
10605 Warn when the stack frame of a function exceeds @var{n} bytes.
10611 The options @option{-metrax4} and @option{-metrax100} are synonyms for
10612 @option{-march=v3} and @option{-march=v8} respectively.
10614 @item -mmul-bug-workaround
10615 @itemx -mno-mul-bug-workaround
10616 @opindex mmul-bug-workaround
10617 @opindex mno-mul-bug-workaround
10618 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
10619 models where it applies. This option is active by default.
10623 Enable CRIS-specific verbose debug-related information in the assembly
10624 code. This option also has the effect to turn off the @samp{#NO_APP}
10625 formatted-code indicator to the assembler at the beginning of the
10630 Do not use condition-code results from previous instruction; always emit
10631 compare and test instructions before use of condition codes.
10633 @item -mno-side-effects
10634 @opindex mno-side-effects
10635 Do not emit instructions with side-effects in addressing modes other than
10638 @item -mstack-align
10639 @itemx -mno-stack-align
10640 @itemx -mdata-align
10641 @itemx -mno-data-align
10642 @itemx -mconst-align
10643 @itemx -mno-const-align
10644 @opindex mstack-align
10645 @opindex mno-stack-align
10646 @opindex mdata-align
10647 @opindex mno-data-align
10648 @opindex mconst-align
10649 @opindex mno-const-align
10650 These options (no-options) arranges (eliminate arrangements) for the
10651 stack-frame, individual data and constants to be aligned for the maximum
10652 single data access size for the chosen CPU model. The default is to
10653 arrange for 32-bit alignment. ABI details such as structure layout are
10654 not affected by these options.
10662 Similar to the stack- data- and const-align options above, these options
10663 arrange for stack-frame, writable data and constants to all be 32-bit,
10664 16-bit or 8-bit aligned. The default is 32-bit alignment.
10666 @item -mno-prologue-epilogue
10667 @itemx -mprologue-epilogue
10668 @opindex mno-prologue-epilogue
10669 @opindex mprologue-epilogue
10670 With @option{-mno-prologue-epilogue}, the normal function prologue and
10671 epilogue that sets up the stack-frame are omitted and no return
10672 instructions or return sequences are generated in the code. Use this
10673 option only together with visual inspection of the compiled code: no
10674 warnings or errors are generated when call-saved registers must be saved,
10675 or storage for local variable needs to be allocated.
10679 @opindex mno-gotplt
10681 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
10682 instruction sequences that load addresses for functions from the PLT part
10683 of the GOT rather than (traditional on other architectures) calls to the
10684 PLT@. The default is @option{-mgotplt}.
10688 Legacy no-op option only recognized with the cris-axis-elf and
10689 cris-axis-linux-gnu targets.
10693 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
10697 This option, recognized for the cris-axis-elf arranges
10698 to link with input-output functions from a simulator library. Code,
10699 initialized data and zero-initialized data are allocated consecutively.
10703 Like @option{-sim}, but pass linker options to locate initialized data at
10704 0x40000000 and zero-initialized data at 0x80000000.
10708 @subsection CRX Options
10709 @cindex CRX Options
10711 These options are defined specifically for the CRX ports.
10717 Enable the use of multiply-accumulate instructions. Disabled by default.
10720 @opindex mpush-args
10721 Push instructions will be used to pass outgoing arguments when functions
10722 are called. Enabled by default.
10725 @node Darwin Options
10726 @subsection Darwin Options
10727 @cindex Darwin options
10729 These options are defined for all architectures running the Darwin operating
10732 FSF GCC on Darwin does not create ``fat'' object files; it will create
10733 an object file for the single architecture that it was built to
10734 target. Apple's GCC on Darwin does create ``fat'' files if multiple
10735 @option{-arch} options are used; it does so by running the compiler or
10736 linker multiple times and joining the results together with
10739 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
10740 @samp{i686}) is determined by the flags that specify the ISA
10741 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
10742 @option{-force_cpusubtype_ALL} option can be used to override this.
10744 The Darwin tools vary in their behavior when presented with an ISA
10745 mismatch. The assembler, @file{as}, will only permit instructions to
10746 be used that are valid for the subtype of the file it is generating,
10747 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
10748 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
10749 and print an error if asked to create a shared library with a less
10750 restrictive subtype than its input files (for instance, trying to put
10751 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
10752 for executables, @file{ld}, will quietly give the executable the most
10753 restrictive subtype of any of its input files.
10758 Add the framework directory @var{dir} to the head of the list of
10759 directories to be searched for header files. These directories are
10760 interleaved with those specified by @option{-I} options and are
10761 scanned in a left-to-right order.
10763 A framework directory is a directory with frameworks in it. A
10764 framework is a directory with a @samp{"Headers"} and/or
10765 @samp{"PrivateHeaders"} directory contained directly in it that ends
10766 in @samp{".framework"}. The name of a framework is the name of this
10767 directory excluding the @samp{".framework"}. Headers associated with
10768 the framework are found in one of those two directories, with
10769 @samp{"Headers"} being searched first. A subframework is a framework
10770 directory that is in a framework's @samp{"Frameworks"} directory.
10771 Includes of subframework headers can only appear in a header of a
10772 framework that contains the subframework, or in a sibling subframework
10773 header. Two subframeworks are siblings if they occur in the same
10774 framework. A subframework should not have the same name as a
10775 framework, a warning will be issued if this is violated. Currently a
10776 subframework cannot have subframeworks, in the future, the mechanism
10777 may be extended to support this. The standard frameworks can be found
10778 in @samp{"/System/Library/Frameworks"} and
10779 @samp{"/Library/Frameworks"}. An example include looks like
10780 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
10781 the name of the framework and header.h is found in the
10782 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
10784 @item -iframework@var{dir}
10785 @opindex iframework
10786 Like @option{-F} except the directory is a treated as a system
10787 directory. The main difference between this @option{-iframework} and
10788 @option{-F} is that with @option{-iframework} the compiler does not
10789 warn about constructs contained within header files found via
10790 @var{dir}. This option is valid only for the C family of languages.
10794 Emit debugging information for symbols that are used. For STABS
10795 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
10796 This is by default ON@.
10800 Emit debugging information for all symbols and types.
10802 @item -mmacosx-version-min=@var{version}
10803 The earliest version of MacOS X that this executable will run on
10804 is @var{version}. Typical values of @var{version} include @code{10.1},
10805 @code{10.2}, and @code{10.3.9}.
10807 If the compiler was built to use the system's headers by default,
10808 then the default for this option is the system version on which the
10809 compiler is running, otherwise the default is to make choices which
10810 are compatible with as many systems and code bases as possible.
10814 Enable kernel development mode. The @option{-mkernel} option sets
10815 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
10816 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
10817 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
10818 applicable. This mode also sets @option{-mno-altivec},
10819 @option{-msoft-float}, @option{-fno-builtin} and
10820 @option{-mlong-branch} for PowerPC targets.
10822 @item -mone-byte-bool
10823 @opindex mone-byte-bool
10824 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
10825 By default @samp{sizeof(bool)} is @samp{4} when compiling for
10826 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
10827 option has no effect on x86.
10829 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
10830 to generate code that is not binary compatible with code generated
10831 without that switch. Using this switch may require recompiling all
10832 other modules in a program, including system libraries. Use this
10833 switch to conform to a non-default data model.
10835 @item -mfix-and-continue
10836 @itemx -ffix-and-continue
10837 @itemx -findirect-data
10838 @opindex mfix-and-continue
10839 @opindex ffix-and-continue
10840 @opindex findirect-data
10841 Generate code suitable for fast turn around development. Needed to
10842 enable gdb to dynamically load @code{.o} files into already running
10843 programs. @option{-findirect-data} and @option{-ffix-and-continue}
10844 are provided for backwards compatibility.
10848 Loads all members of static archive libraries.
10849 See man ld(1) for more information.
10851 @item -arch_errors_fatal
10852 @opindex arch_errors_fatal
10853 Cause the errors having to do with files that have the wrong architecture
10856 @item -bind_at_load
10857 @opindex bind_at_load
10858 Causes the output file to be marked such that the dynamic linker will
10859 bind all undefined references when the file is loaded or launched.
10863 Produce a Mach-o bundle format file.
10864 See man ld(1) for more information.
10866 @item -bundle_loader @var{executable}
10867 @opindex bundle_loader
10868 This option specifies the @var{executable} that will be loading the build
10869 output file being linked. See man ld(1) for more information.
10872 @opindex dynamiclib
10873 When passed this option, GCC will produce a dynamic library instead of
10874 an executable when linking, using the Darwin @file{libtool} command.
10876 @item -force_cpusubtype_ALL
10877 @opindex force_cpusubtype_ALL
10878 This causes GCC's output file to have the @var{ALL} subtype, instead of
10879 one controlled by the @option{-mcpu} or @option{-march} option.
10881 @item -allowable_client @var{client_name}
10882 @itemx -client_name
10883 @itemx -compatibility_version
10884 @itemx -current_version
10886 @itemx -dependency-file
10888 @itemx -dylinker_install_name
10890 @itemx -exported_symbols_list
10893 @itemx -flat_namespace
10894 @itemx -force_flat_namespace
10895 @itemx -headerpad_max_install_names
10898 @itemx -install_name
10899 @itemx -keep_private_externs
10900 @itemx -multi_module
10901 @itemx -multiply_defined
10902 @itemx -multiply_defined_unused
10905 @itemx -no_dead_strip_inits_and_terms
10906 @itemx -nofixprebinding
10907 @itemx -nomultidefs
10909 @itemx -noseglinkedit
10910 @itemx -pagezero_size
10912 @itemx -prebind_all_twolevel_modules
10913 @itemx -private_bundle
10915 @itemx -read_only_relocs
10917 @itemx -sectobjectsymbols
10921 @itemx -sectobjectsymbols
10924 @itemx -segs_read_only_addr
10926 @itemx -segs_read_write_addr
10927 @itemx -seg_addr_table
10928 @itemx -seg_addr_table_filename
10929 @itemx -seglinkedit
10931 @itemx -segs_read_only_addr
10932 @itemx -segs_read_write_addr
10933 @itemx -single_module
10935 @itemx -sub_library
10937 @itemx -sub_umbrella
10938 @itemx -twolevel_namespace
10941 @itemx -unexported_symbols_list
10942 @itemx -weak_reference_mismatches
10943 @itemx -whatsloaded
10944 @opindex allowable_client
10945 @opindex client_name
10946 @opindex compatibility_version
10947 @opindex current_version
10948 @opindex dead_strip
10949 @opindex dependency-file
10950 @opindex dylib_file
10951 @opindex dylinker_install_name
10953 @opindex exported_symbols_list
10955 @opindex flat_namespace
10956 @opindex force_flat_namespace
10957 @opindex headerpad_max_install_names
10958 @opindex image_base
10960 @opindex install_name
10961 @opindex keep_private_externs
10962 @opindex multi_module
10963 @opindex multiply_defined
10964 @opindex multiply_defined_unused
10965 @opindex noall_load
10966 @opindex no_dead_strip_inits_and_terms
10967 @opindex nofixprebinding
10968 @opindex nomultidefs
10970 @opindex noseglinkedit
10971 @opindex pagezero_size
10973 @opindex prebind_all_twolevel_modules
10974 @opindex private_bundle
10975 @opindex read_only_relocs
10977 @opindex sectobjectsymbols
10980 @opindex sectcreate
10981 @opindex sectobjectsymbols
10984 @opindex segs_read_only_addr
10985 @opindex segs_read_write_addr
10986 @opindex seg_addr_table
10987 @opindex seg_addr_table_filename
10988 @opindex seglinkedit
10990 @opindex segs_read_only_addr
10991 @opindex segs_read_write_addr
10992 @opindex single_module
10994 @opindex sub_library
10995 @opindex sub_umbrella
10996 @opindex twolevel_namespace
10999 @opindex unexported_symbols_list
11000 @opindex weak_reference_mismatches
11001 @opindex whatsloaded
11002 These options are passed to the Darwin linker. The Darwin linker man page
11003 describes them in detail.
11006 @node DEC Alpha Options
11007 @subsection DEC Alpha Options
11009 These @samp{-m} options are defined for the DEC Alpha implementations:
11012 @item -mno-soft-float
11013 @itemx -msoft-float
11014 @opindex mno-soft-float
11015 @opindex msoft-float
11016 Use (do not use) the hardware floating-point instructions for
11017 floating-point operations. When @option{-msoft-float} is specified,
11018 functions in @file{libgcc.a} will be used to perform floating-point
11019 operations. Unless they are replaced by routines that emulate the
11020 floating-point operations, or compiled in such a way as to call such
11021 emulations routines, these routines will issue floating-point
11022 operations. If you are compiling for an Alpha without floating-point
11023 operations, you must ensure that the library is built so as not to call
11026 Note that Alpha implementations without floating-point operations are
11027 required to have floating-point registers.
11030 @itemx -mno-fp-regs
11032 @opindex mno-fp-regs
11033 Generate code that uses (does not use) the floating-point register set.
11034 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
11035 register set is not used, floating point operands are passed in integer
11036 registers as if they were integers and floating-point results are passed
11037 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
11038 so any function with a floating-point argument or return value called by code
11039 compiled with @option{-mno-fp-regs} must also be compiled with that
11042 A typical use of this option is building a kernel that does not use,
11043 and hence need not save and restore, any floating-point registers.
11047 The Alpha architecture implements floating-point hardware optimized for
11048 maximum performance. It is mostly compliant with the IEEE floating
11049 point standard. However, for full compliance, software assistance is
11050 required. This option generates code fully IEEE compliant code
11051 @emph{except} that the @var{inexact-flag} is not maintained (see below).
11052 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
11053 defined during compilation. The resulting code is less efficient but is
11054 able to correctly support denormalized numbers and exceptional IEEE
11055 values such as not-a-number and plus/minus infinity. Other Alpha
11056 compilers call this option @option{-ieee_with_no_inexact}.
11058 @item -mieee-with-inexact
11059 @opindex mieee-with-inexact
11060 This is like @option{-mieee} except the generated code also maintains
11061 the IEEE @var{inexact-flag}. Turning on this option causes the
11062 generated code to implement fully-compliant IEEE math. In addition to
11063 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
11064 macro. On some Alpha implementations the resulting code may execute
11065 significantly slower than the code generated by default. Since there is
11066 very little code that depends on the @var{inexact-flag}, you should
11067 normally not specify this option. Other Alpha compilers call this
11068 option @option{-ieee_with_inexact}.
11070 @item -mfp-trap-mode=@var{trap-mode}
11071 @opindex mfp-trap-mode
11072 This option controls what floating-point related traps are enabled.
11073 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
11074 The trap mode can be set to one of four values:
11078 This is the default (normal) setting. The only traps that are enabled
11079 are the ones that cannot be disabled in software (e.g., division by zero
11083 In addition to the traps enabled by @samp{n}, underflow traps are enabled
11087 Like @samp{u}, but the instructions are marked to be safe for software
11088 completion (see Alpha architecture manual for details).
11091 Like @samp{su}, but inexact traps are enabled as well.
11094 @item -mfp-rounding-mode=@var{rounding-mode}
11095 @opindex mfp-rounding-mode
11096 Selects the IEEE rounding mode. Other Alpha compilers call this option
11097 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
11102 Normal IEEE rounding mode. Floating point numbers are rounded towards
11103 the nearest machine number or towards the even machine number in case
11107 Round towards minus infinity.
11110 Chopped rounding mode. Floating point numbers are rounded towards zero.
11113 Dynamic rounding mode. A field in the floating point control register
11114 (@var{fpcr}, see Alpha architecture reference manual) controls the
11115 rounding mode in effect. The C library initializes this register for
11116 rounding towards plus infinity. Thus, unless your program modifies the
11117 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
11120 @item -mtrap-precision=@var{trap-precision}
11121 @opindex mtrap-precision
11122 In the Alpha architecture, floating point traps are imprecise. This
11123 means without software assistance it is impossible to recover from a
11124 floating trap and program execution normally needs to be terminated.
11125 GCC can generate code that can assist operating system trap handlers
11126 in determining the exact location that caused a floating point trap.
11127 Depending on the requirements of an application, different levels of
11128 precisions can be selected:
11132 Program precision. This option is the default and means a trap handler
11133 can only identify which program caused a floating point exception.
11136 Function precision. The trap handler can determine the function that
11137 caused a floating point exception.
11140 Instruction precision. The trap handler can determine the exact
11141 instruction that caused a floating point exception.
11144 Other Alpha compilers provide the equivalent options called
11145 @option{-scope_safe} and @option{-resumption_safe}.
11147 @item -mieee-conformant
11148 @opindex mieee-conformant
11149 This option marks the generated code as IEEE conformant. You must not
11150 use this option unless you also specify @option{-mtrap-precision=i} and either
11151 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
11152 is to emit the line @samp{.eflag 48} in the function prologue of the
11153 generated assembly file. Under DEC Unix, this has the effect that
11154 IEEE-conformant math library routines will be linked in.
11156 @item -mbuild-constants
11157 @opindex mbuild-constants
11158 Normally GCC examines a 32- or 64-bit integer constant to
11159 see if it can construct it from smaller constants in two or three
11160 instructions. If it cannot, it will output the constant as a literal and
11161 generate code to load it from the data segment at runtime.
11163 Use this option to require GCC to construct @emph{all} integer constants
11164 using code, even if it takes more instructions (the maximum is six).
11166 You would typically use this option to build a shared library dynamic
11167 loader. Itself a shared library, it must relocate itself in memory
11168 before it can find the variables and constants in its own data segment.
11174 Select whether to generate code to be assembled by the vendor-supplied
11175 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
11193 Indicate whether GCC should generate code to use the optional BWX,
11194 CIX, FIX and MAX instruction sets. The default is to use the instruction
11195 sets supported by the CPU type specified via @option{-mcpu=} option or that
11196 of the CPU on which GCC was built if none was specified.
11199 @itemx -mfloat-ieee
11200 @opindex mfloat-vax
11201 @opindex mfloat-ieee
11202 Generate code that uses (does not use) VAX F and G floating point
11203 arithmetic instead of IEEE single and double precision.
11205 @item -mexplicit-relocs
11206 @itemx -mno-explicit-relocs
11207 @opindex mexplicit-relocs
11208 @opindex mno-explicit-relocs
11209 Older Alpha assemblers provided no way to generate symbol relocations
11210 except via assembler macros. Use of these macros does not allow
11211 optimal instruction scheduling. GNU binutils as of version 2.12
11212 supports a new syntax that allows the compiler to explicitly mark
11213 which relocations should apply to which instructions. This option
11214 is mostly useful for debugging, as GCC detects the capabilities of
11215 the assembler when it is built and sets the default accordingly.
11218 @itemx -mlarge-data
11219 @opindex msmall-data
11220 @opindex mlarge-data
11221 When @option{-mexplicit-relocs} is in effect, static data is
11222 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
11223 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
11224 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
11225 16-bit relocations off of the @code{$gp} register. This limits the
11226 size of the small data area to 64KB, but allows the variables to be
11227 directly accessed via a single instruction.
11229 The default is @option{-mlarge-data}. With this option the data area
11230 is limited to just below 2GB@. Programs that require more than 2GB of
11231 data must use @code{malloc} or @code{mmap} to allocate the data in the
11232 heap instead of in the program's data segment.
11234 When generating code for shared libraries, @option{-fpic} implies
11235 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
11238 @itemx -mlarge-text
11239 @opindex msmall-text
11240 @opindex mlarge-text
11241 When @option{-msmall-text} is used, the compiler assumes that the
11242 code of the entire program (or shared library) fits in 4MB, and is
11243 thus reachable with a branch instruction. When @option{-msmall-data}
11244 is used, the compiler can assume that all local symbols share the
11245 same @code{$gp} value, and thus reduce the number of instructions
11246 required for a function call from 4 to 1.
11248 The default is @option{-mlarge-text}.
11250 @item -mcpu=@var{cpu_type}
11252 Set the instruction set and instruction scheduling parameters for
11253 machine type @var{cpu_type}. You can specify either the @samp{EV}
11254 style name or the corresponding chip number. GCC supports scheduling
11255 parameters for the EV4, EV5 and EV6 family of processors and will
11256 choose the default values for the instruction set from the processor
11257 you specify. If you do not specify a processor type, GCC will default
11258 to the processor on which the compiler was built.
11260 Supported values for @var{cpu_type} are
11266 Schedules as an EV4 and has no instruction set extensions.
11270 Schedules as an EV5 and has no instruction set extensions.
11274 Schedules as an EV5 and supports the BWX extension.
11279 Schedules as an EV5 and supports the BWX and MAX extensions.
11283 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
11287 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
11290 Native Linux/GNU toolchains also support the value @samp{native},
11291 which selects the best architecture option for the host processor.
11292 @option{-mcpu=native} has no effect if GCC does not recognize
11295 @item -mtune=@var{cpu_type}
11297 Set only the instruction scheduling parameters for machine type
11298 @var{cpu_type}. The instruction set is not changed.
11300 Native Linux/GNU toolchains also support the value @samp{native},
11301 which selects the best architecture option for the host processor.
11302 @option{-mtune=native} has no effect if GCC does not recognize
11305 @item -mmemory-latency=@var{time}
11306 @opindex mmemory-latency
11307 Sets the latency the scheduler should assume for typical memory
11308 references as seen by the application. This number is highly
11309 dependent on the memory access patterns used by the application
11310 and the size of the external cache on the machine.
11312 Valid options for @var{time} are
11316 A decimal number representing clock cycles.
11322 The compiler contains estimates of the number of clock cycles for
11323 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
11324 (also called Dcache, Scache, and Bcache), as well as to main memory.
11325 Note that L3 is only valid for EV5.
11330 @node DEC Alpha/VMS Options
11331 @subsection DEC Alpha/VMS Options
11333 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
11336 @item -mvms-return-codes
11337 @opindex mvms-return-codes
11338 Return VMS condition codes from main. The default is to return POSIX
11339 style condition (e.g.@: error) codes.
11341 @item -mdebug-main=@var{prefix}
11342 @opindex mdebug-main=@var{prefix}
11343 Flag the first routine whose name starts with @var{prefix} as the main
11344 routine for the debugger.
11348 Default to 64bit memory allocation routines.
11352 @subsection FR30 Options
11353 @cindex FR30 Options
11355 These options are defined specifically for the FR30 port.
11359 @item -msmall-model
11360 @opindex msmall-model
11361 Use the small address space model. This can produce smaller code, but
11362 it does assume that all symbolic values and addresses will fit into a
11367 Assume that run-time support has been provided and so there is no need
11368 to include the simulator library (@file{libsim.a}) on the linker
11374 @subsection FRV Options
11375 @cindex FRV Options
11381 Only use the first 32 general purpose registers.
11386 Use all 64 general purpose registers.
11391 Use only the first 32 floating point registers.
11396 Use all 64 floating point registers
11399 @opindex mhard-float
11401 Use hardware instructions for floating point operations.
11404 @opindex msoft-float
11406 Use library routines for floating point operations.
11411 Dynamically allocate condition code registers.
11416 Do not try to dynamically allocate condition code registers, only
11417 use @code{icc0} and @code{fcc0}.
11422 Change ABI to use double word insns.
11427 Do not use double word instructions.
11432 Use floating point double instructions.
11435 @opindex mno-double
11437 Do not use floating point double instructions.
11442 Use media instructions.
11447 Do not use media instructions.
11452 Use multiply and add/subtract instructions.
11455 @opindex mno-muladd
11457 Do not use multiply and add/subtract instructions.
11462 Select the FDPIC ABI, that uses function descriptors to represent
11463 pointers to functions. Without any PIC/PIE-related options, it
11464 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
11465 assumes GOT entries and small data are within a 12-bit range from the
11466 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
11467 are computed with 32 bits.
11468 With a @samp{bfin-elf} target, this option implies @option{-msim}.
11471 @opindex minline-plt
11473 Enable inlining of PLT entries in function calls to functions that are
11474 not known to bind locally. It has no effect without @option{-mfdpic}.
11475 It's enabled by default if optimizing for speed and compiling for
11476 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
11477 optimization option such as @option{-O3} or above is present in the
11483 Assume a large TLS segment when generating thread-local code.
11488 Do not assume a large TLS segment when generating thread-local code.
11493 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
11494 that is known to be in read-only sections. It's enabled by default,
11495 except for @option{-fpic} or @option{-fpie}: even though it may help
11496 make the global offset table smaller, it trades 1 instruction for 4.
11497 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
11498 one of which may be shared by multiple symbols, and it avoids the need
11499 for a GOT entry for the referenced symbol, so it's more likely to be a
11500 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
11502 @item -multilib-library-pic
11503 @opindex multilib-library-pic
11505 Link with the (library, not FD) pic libraries. It's implied by
11506 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
11507 @option{-fpic} without @option{-mfdpic}. You should never have to use
11511 @opindex mlinked-fp
11513 Follow the EABI requirement of always creating a frame pointer whenever
11514 a stack frame is allocated. This option is enabled by default and can
11515 be disabled with @option{-mno-linked-fp}.
11518 @opindex mlong-calls
11520 Use indirect addressing to call functions outside the current
11521 compilation unit. This allows the functions to be placed anywhere
11522 within the 32-bit address space.
11524 @item -malign-labels
11525 @opindex malign-labels
11527 Try to align labels to an 8-byte boundary by inserting nops into the
11528 previous packet. This option only has an effect when VLIW packing
11529 is enabled. It doesn't create new packets; it merely adds nops to
11532 @item -mlibrary-pic
11533 @opindex mlibrary-pic
11535 Generate position-independent EABI code.
11540 Use only the first four media accumulator registers.
11545 Use all eight media accumulator registers.
11550 Pack VLIW instructions.
11555 Do not pack VLIW instructions.
11558 @opindex mno-eflags
11560 Do not mark ABI switches in e_flags.
11563 @opindex mcond-move
11565 Enable the use of conditional-move instructions (default).
11567 This switch is mainly for debugging the compiler and will likely be removed
11568 in a future version.
11570 @item -mno-cond-move
11571 @opindex mno-cond-move
11573 Disable the use of conditional-move instructions.
11575 This switch is mainly for debugging the compiler and will likely be removed
11576 in a future version.
11581 Enable the use of conditional set instructions (default).
11583 This switch is mainly for debugging the compiler and will likely be removed
11584 in a future version.
11589 Disable the use of conditional set instructions.
11591 This switch is mainly for debugging the compiler and will likely be removed
11592 in a future version.
11595 @opindex mcond-exec
11597 Enable the use of conditional execution (default).
11599 This switch is mainly for debugging the compiler and will likely be removed
11600 in a future version.
11602 @item -mno-cond-exec
11603 @opindex mno-cond-exec
11605 Disable the use of conditional execution.
11607 This switch is mainly for debugging the compiler and will likely be removed
11608 in a future version.
11610 @item -mvliw-branch
11611 @opindex mvliw-branch
11613 Run a pass to pack branches into VLIW instructions (default).
11615 This switch is mainly for debugging the compiler and will likely be removed
11616 in a future version.
11618 @item -mno-vliw-branch
11619 @opindex mno-vliw-branch
11621 Do not run a pass to pack branches into VLIW instructions.
11623 This switch is mainly for debugging the compiler and will likely be removed
11624 in a future version.
11626 @item -mmulti-cond-exec
11627 @opindex mmulti-cond-exec
11629 Enable optimization of @code{&&} and @code{||} in conditional execution
11632 This switch is mainly for debugging the compiler and will likely be removed
11633 in a future version.
11635 @item -mno-multi-cond-exec
11636 @opindex mno-multi-cond-exec
11638 Disable optimization of @code{&&} and @code{||} in conditional execution.
11640 This switch is mainly for debugging the compiler and will likely be removed
11641 in a future version.
11643 @item -mnested-cond-exec
11644 @opindex mnested-cond-exec
11646 Enable nested conditional execution optimizations (default).
11648 This switch is mainly for debugging the compiler and will likely be removed
11649 in a future version.
11651 @item -mno-nested-cond-exec
11652 @opindex mno-nested-cond-exec
11654 Disable nested conditional execution optimizations.
11656 This switch is mainly for debugging the compiler and will likely be removed
11657 in a future version.
11659 @item -moptimize-membar
11660 @opindex moptimize-membar
11662 This switch removes redundant @code{membar} instructions from the
11663 compiler generated code. It is enabled by default.
11665 @item -mno-optimize-membar
11666 @opindex mno-optimize-membar
11668 This switch disables the automatic removal of redundant @code{membar}
11669 instructions from the generated code.
11671 @item -mtomcat-stats
11672 @opindex mtomcat-stats
11674 Cause gas to print out tomcat statistics.
11676 @item -mcpu=@var{cpu}
11679 Select the processor type for which to generate code. Possible values are
11680 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
11681 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
11685 @node GNU/Linux Options
11686 @subsection GNU/Linux Options
11688 These @samp{-m} options are defined for GNU/Linux targets:
11693 Use the GNU C library. This is the default except
11694 on @samp{*-*-linux-*uclibc*} and @samp{*-*-linux-*android*} targets.
11698 Use uClibc C library. This is the default on
11699 @samp{*-*-linux-*uclibc*} targets.
11703 Use Bionic C library. This is the default on
11704 @samp{*-*-linux-*android*} targets.
11708 Compile code compatible with Android platform. This is the default on
11709 @samp{*-*-linux-*android*} targets.
11711 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
11712 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
11713 this option makes the GCC driver pass Android-specific options to the linker.
11714 Finally, this option causes the preprocessor macro @code{__ANDROID__}
11717 @item -tno-android-cc
11718 @opindex tno-android-cc
11719 Disable compilation effects of @option{-mandroid}, i.e., do not enable
11720 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
11721 @option{-fno-rtti} by default.
11723 @item -tno-android-ld
11724 @opindex tno-android-ld
11725 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
11726 linking options to the linker.
11730 @node H8/300 Options
11731 @subsection H8/300 Options
11733 These @samp{-m} options are defined for the H8/300 implementations:
11738 Shorten some address references at link time, when possible; uses the
11739 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
11740 ld, Using ld}, for a fuller description.
11744 Generate code for the H8/300H@.
11748 Generate code for the H8S@.
11752 Generate code for the H8S and H8/300H in the normal mode. This switch
11753 must be used either with @option{-mh} or @option{-ms}.
11757 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
11761 Make @code{int} data 32 bits by default.
11764 @opindex malign-300
11765 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
11766 The default for the H8/300H and H8S is to align longs and floats on 4
11768 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
11769 This option has no effect on the H8/300.
11773 @subsection HPPA Options
11774 @cindex HPPA Options
11776 These @samp{-m} options are defined for the HPPA family of computers:
11779 @item -march=@var{architecture-type}
11781 Generate code for the specified architecture. The choices for
11782 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
11783 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
11784 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
11785 architecture option for your machine. Code compiled for lower numbered
11786 architectures will run on higher numbered architectures, but not the
11789 @item -mpa-risc-1-0
11790 @itemx -mpa-risc-1-1
11791 @itemx -mpa-risc-2-0
11792 @opindex mpa-risc-1-0
11793 @opindex mpa-risc-1-1
11794 @opindex mpa-risc-2-0
11795 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
11798 @opindex mbig-switch
11799 Generate code suitable for big switch tables. Use this option only if
11800 the assembler/linker complain about out of range branches within a switch
11803 @item -mjump-in-delay
11804 @opindex mjump-in-delay
11805 Fill delay slots of function calls with unconditional jump instructions
11806 by modifying the return pointer for the function call to be the target
11807 of the conditional jump.
11809 @item -mdisable-fpregs
11810 @opindex mdisable-fpregs
11811 Prevent floating point registers from being used in any manner. This is
11812 necessary for compiling kernels which perform lazy context switching of
11813 floating point registers. If you use this option and attempt to perform
11814 floating point operations, the compiler will abort.
11816 @item -mdisable-indexing
11817 @opindex mdisable-indexing
11818 Prevent the compiler from using indexing address modes. This avoids some
11819 rather obscure problems when compiling MIG generated code under MACH@.
11821 @item -mno-space-regs
11822 @opindex mno-space-regs
11823 Generate code that assumes the target has no space registers. This allows
11824 GCC to generate faster indirect calls and use unscaled index address modes.
11826 Such code is suitable for level 0 PA systems and kernels.
11828 @item -mfast-indirect-calls
11829 @opindex mfast-indirect-calls
11830 Generate code that assumes calls never cross space boundaries. This
11831 allows GCC to emit code which performs faster indirect calls.
11833 This option will not work in the presence of shared libraries or nested
11836 @item -mfixed-range=@var{register-range}
11837 @opindex mfixed-range
11838 Generate code treating the given register range as fixed registers.
11839 A fixed register is one that the register allocator can not use. This is
11840 useful when compiling kernel code. A register range is specified as
11841 two registers separated by a dash. Multiple register ranges can be
11842 specified separated by a comma.
11844 @item -mlong-load-store
11845 @opindex mlong-load-store
11846 Generate 3-instruction load and store sequences as sometimes required by
11847 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
11850 @item -mportable-runtime
11851 @opindex mportable-runtime
11852 Use the portable calling conventions proposed by HP for ELF systems.
11856 Enable the use of assembler directives only GAS understands.
11858 @item -mschedule=@var{cpu-type}
11860 Schedule code according to the constraints for the machine type
11861 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
11862 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
11863 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
11864 proper scheduling option for your machine. The default scheduling is
11868 @opindex mlinker-opt
11869 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
11870 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
11871 linkers in which they give bogus error messages when linking some programs.
11874 @opindex msoft-float
11875 Generate output containing library calls for floating point.
11876 @strong{Warning:} the requisite libraries are not available for all HPPA
11877 targets. Normally the facilities of the machine's usual C compiler are
11878 used, but this cannot be done directly in cross-compilation. You must make
11879 your own arrangements to provide suitable library functions for
11882 @option{-msoft-float} changes the calling convention in the output file;
11883 therefore, it is only useful if you compile @emph{all} of a program with
11884 this option. In particular, you need to compile @file{libgcc.a}, the
11885 library that comes with GCC, with @option{-msoft-float} in order for
11890 Generate the predefine, @code{_SIO}, for server IO@. The default is
11891 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
11892 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
11893 options are available under HP-UX and HI-UX@.
11897 Use GNU ld specific options. This passes @option{-shared} to ld when
11898 building a shared library. It is the default when GCC is configured,
11899 explicitly or implicitly, with the GNU linker. This option does not
11900 have any affect on which ld is called, it only changes what parameters
11901 are passed to that ld. The ld that is called is determined by the
11902 @option{--with-ld} configure option, GCC's program search path, and
11903 finally by the user's @env{PATH}. The linker used by GCC can be printed
11904 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
11905 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11909 Use HP ld specific options. This passes @option{-b} to ld when building
11910 a shared library and passes @option{+Accept TypeMismatch} to ld on all
11911 links. It is the default when GCC is configured, explicitly or
11912 implicitly, with the HP linker. This option does not have any affect on
11913 which ld is called, it only changes what parameters are passed to that
11914 ld. The ld that is called is determined by the @option{--with-ld}
11915 configure option, GCC's program search path, and finally by the user's
11916 @env{PATH}. The linker used by GCC can be printed using @samp{which
11917 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
11918 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11921 @opindex mno-long-calls
11922 Generate code that uses long call sequences. This ensures that a call
11923 is always able to reach linker generated stubs. The default is to generate
11924 long calls only when the distance from the call site to the beginning
11925 of the function or translation unit, as the case may be, exceeds a
11926 predefined limit set by the branch type being used. The limits for
11927 normal calls are 7,600,000 and 240,000 bytes, respectively for the
11928 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
11931 Distances are measured from the beginning of functions when using the
11932 @option{-ffunction-sections} option, or when using the @option{-mgas}
11933 and @option{-mno-portable-runtime} options together under HP-UX with
11936 It is normally not desirable to use this option as it will degrade
11937 performance. However, it may be useful in large applications,
11938 particularly when partial linking is used to build the application.
11940 The types of long calls used depends on the capabilities of the
11941 assembler and linker, and the type of code being generated. The
11942 impact on systems that support long absolute calls, and long pic
11943 symbol-difference or pc-relative calls should be relatively small.
11944 However, an indirect call is used on 32-bit ELF systems in pic code
11945 and it is quite long.
11947 @item -munix=@var{unix-std}
11949 Generate compiler predefines and select a startfile for the specified
11950 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
11951 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
11952 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
11953 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
11954 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
11957 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
11958 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
11959 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
11960 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
11961 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
11962 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
11964 It is @emph{important} to note that this option changes the interfaces
11965 for various library routines. It also affects the operational behavior
11966 of the C library. Thus, @emph{extreme} care is needed in using this
11969 Library code that is intended to operate with more than one UNIX
11970 standard must test, set and restore the variable @var{__xpg4_extended_mask}
11971 as appropriate. Most GNU software doesn't provide this capability.
11975 Suppress the generation of link options to search libdld.sl when the
11976 @option{-static} option is specified on HP-UX 10 and later.
11980 The HP-UX implementation of setlocale in libc has a dependency on
11981 libdld.sl. There isn't an archive version of libdld.sl. Thus,
11982 when the @option{-static} option is specified, special link options
11983 are needed to resolve this dependency.
11985 On HP-UX 10 and later, the GCC driver adds the necessary options to
11986 link with libdld.sl when the @option{-static} option is specified.
11987 This causes the resulting binary to be dynamic. On the 64-bit port,
11988 the linkers generate dynamic binaries by default in any case. The
11989 @option{-nolibdld} option can be used to prevent the GCC driver from
11990 adding these link options.
11994 Add support for multithreading with the @dfn{dce thread} library
11995 under HP-UX@. This option sets flags for both the preprocessor and
11999 @node i386 and x86-64 Options
12000 @subsection Intel 386 and AMD x86-64 Options
12001 @cindex i386 Options
12002 @cindex x86-64 Options
12003 @cindex Intel 386 Options
12004 @cindex AMD x86-64 Options
12006 These @samp{-m} options are defined for the i386 and x86-64 family of
12010 @item -mtune=@var{cpu-type}
12012 Tune to @var{cpu-type} everything applicable about the generated code, except
12013 for the ABI and the set of available instructions. The choices for
12014 @var{cpu-type} are:
12017 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
12018 If you know the CPU on which your code will run, then you should use
12019 the corresponding @option{-mtune} option instead of
12020 @option{-mtune=generic}. But, if you do not know exactly what CPU users
12021 of your application will have, then you should use this option.
12023 As new processors are deployed in the marketplace, the behavior of this
12024 option will change. Therefore, if you upgrade to a newer version of
12025 GCC, the code generated option will change to reflect the processors
12026 that were most common when that version of GCC was released.
12028 There is no @option{-march=generic} option because @option{-march}
12029 indicates the instruction set the compiler can use, and there is no
12030 generic instruction set applicable to all processors. In contrast,
12031 @option{-mtune} indicates the processor (or, in this case, collection of
12032 processors) for which the code is optimized.
12034 This selects the CPU to tune for at compilation time by determining
12035 the processor type of the compiling machine. Using @option{-mtune=native}
12036 will produce code optimized for the local machine under the constraints
12037 of the selected instruction set. Using @option{-march=native} will
12038 enable all instruction subsets supported by the local machine (hence
12039 the result might not run on different machines).
12041 Original Intel's i386 CPU@.
12043 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
12044 @item i586, pentium
12045 Intel Pentium CPU with no MMX support.
12047 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
12049 Intel PentiumPro CPU@.
12051 Same as @code{generic}, but when used as @code{march} option, PentiumPro
12052 instruction set will be used, so the code will run on all i686 family chips.
12054 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
12055 @item pentium3, pentium3m
12056 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
12059 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
12060 support. Used by Centrino notebooks.
12061 @item pentium4, pentium4m
12062 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
12064 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
12067 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
12068 SSE2 and SSE3 instruction set support.
12070 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
12071 instruction set support.
12073 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1
12074 and SSE4.2 instruction set support.
12076 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
12077 instruction set support.
12079 AMD K6 CPU with MMX instruction set support.
12081 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
12082 @item athlon, athlon-tbird
12083 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
12085 @item athlon-4, athlon-xp, athlon-mp
12086 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
12087 instruction set support.
12088 @item k8, opteron, athlon64, athlon-fx
12089 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
12090 MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit instruction set extensions.)
12091 @item k8-sse3, opteron-sse3, athlon64-sse3
12092 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
12093 @item amdfam10, barcelona
12094 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
12095 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
12096 instruction set extensions.)
12098 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
12101 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
12102 instruction set support.
12104 Via C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
12105 implemented for this chip.)
12107 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
12108 implemented for this chip.)
12110 Embedded AMD CPU with MMX and 3DNow!@: instruction set support.
12113 While picking a specific @var{cpu-type} will schedule things appropriately
12114 for that particular chip, the compiler will not generate any code that
12115 does not run on the i386 without the @option{-march=@var{cpu-type}} option
12118 @item -march=@var{cpu-type}
12120 Generate instructions for the machine type @var{cpu-type}. The choices
12121 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
12122 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
12124 @item -mcpu=@var{cpu-type}
12126 A deprecated synonym for @option{-mtune}.
12128 @item -mfpmath=@var{unit}
12130 Generate floating point arithmetics for selected unit @var{unit}. The choices
12131 for @var{unit} are:
12135 Use the standard 387 floating point coprocessor present majority of chips and
12136 emulated otherwise. Code compiled with this option will run almost everywhere.
12137 The temporary results are computed in 80bit precision instead of precision
12138 specified by the type resulting in slightly different results compared to most
12139 of other chips. See @option{-ffloat-store} for more detailed description.
12141 This is the default choice for i386 compiler.
12144 Use scalar floating point instructions present in the SSE instruction set.
12145 This instruction set is supported by Pentium3 and newer chips, in the AMD line
12146 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
12147 instruction set supports only single precision arithmetics, thus the double and
12148 extended precision arithmetics is still done using 387. Later version, present
12149 only in Pentium4 and the future AMD x86-64 chips supports double precision
12152 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
12153 or @option{-msse2} switches to enable SSE extensions and make this option
12154 effective. For the x86-64 compiler, these extensions are enabled by default.
12156 The resulting code should be considerably faster in the majority of cases and avoid
12157 the numerical instability problems of 387 code, but may break some existing
12158 code that expects temporaries to be 80bit.
12160 This is the default choice for the x86-64 compiler.
12165 Attempt to utilize both instruction sets at once. This effectively double the
12166 amount of available registers and on chips with separate execution units for
12167 387 and SSE the execution resources too. Use this option with care, as it is
12168 still experimental, because the GCC register allocator does not model separate
12169 functional units well resulting in instable performance.
12172 @item -masm=@var{dialect}
12173 @opindex masm=@var{dialect}
12174 Output asm instructions using selected @var{dialect}. Supported
12175 choices are @samp{intel} or @samp{att} (the default one). Darwin does
12176 not support @samp{intel}.
12179 @itemx -mno-ieee-fp
12181 @opindex mno-ieee-fp
12182 Control whether or not the compiler uses IEEE floating point
12183 comparisons. These handle correctly the case where the result of a
12184 comparison is unordered.
12187 @opindex msoft-float
12188 Generate output containing library calls for floating point.
12189 @strong{Warning:} the requisite libraries are not part of GCC@.
12190 Normally the facilities of the machine's usual C compiler are used, but
12191 this can't be done directly in cross-compilation. You must make your
12192 own arrangements to provide suitable library functions for
12195 On machines where a function returns floating point results in the 80387
12196 register stack, some floating point opcodes may be emitted even if
12197 @option{-msoft-float} is used.
12199 @item -mno-fp-ret-in-387
12200 @opindex mno-fp-ret-in-387
12201 Do not use the FPU registers for return values of functions.
12203 The usual calling convention has functions return values of types
12204 @code{float} and @code{double} in an FPU register, even if there
12205 is no FPU@. The idea is that the operating system should emulate
12208 The option @option{-mno-fp-ret-in-387} causes such values to be returned
12209 in ordinary CPU registers instead.
12211 @item -mno-fancy-math-387
12212 @opindex mno-fancy-math-387
12213 Some 387 emulators do not support the @code{sin}, @code{cos} and
12214 @code{sqrt} instructions for the 387. Specify this option to avoid
12215 generating those instructions. This option is the default on FreeBSD,
12216 OpenBSD and NetBSD@. This option is overridden when @option{-march}
12217 indicates that the target cpu will always have an FPU and so the
12218 instruction will not need emulation. As of revision 2.6.1, these
12219 instructions are not generated unless you also use the
12220 @option{-funsafe-math-optimizations} switch.
12222 @item -malign-double
12223 @itemx -mno-align-double
12224 @opindex malign-double
12225 @opindex mno-align-double
12226 Control whether GCC aligns @code{double}, @code{long double}, and
12227 @code{long long} variables on a two word boundary or a one word
12228 boundary. Aligning @code{double} variables on a two word boundary will
12229 produce code that runs somewhat faster on a @samp{Pentium} at the
12230 expense of more memory.
12232 On x86-64, @option{-malign-double} is enabled by default.
12234 @strong{Warning:} if you use the @option{-malign-double} switch,
12235 structures containing the above types will be aligned differently than
12236 the published application binary interface specifications for the 386
12237 and will not be binary compatible with structures in code compiled
12238 without that switch.
12240 @item -m96bit-long-double
12241 @itemx -m128bit-long-double
12242 @opindex m96bit-long-double
12243 @opindex m128bit-long-double
12244 These switches control the size of @code{long double} type. The i386
12245 application binary interface specifies the size to be 96 bits,
12246 so @option{-m96bit-long-double} is the default in 32 bit mode.
12248 Modern architectures (Pentium and newer) would prefer @code{long double}
12249 to be aligned to an 8 or 16 byte boundary. In arrays or structures
12250 conforming to the ABI, this would not be possible. So specifying a
12251 @option{-m128bit-long-double} will align @code{long double}
12252 to a 16 byte boundary by padding the @code{long double} with an additional
12255 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
12256 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
12258 Notice that neither of these options enable any extra precision over the x87
12259 standard of 80 bits for a @code{long double}.
12261 @strong{Warning:} if you override the default value for your target ABI, the
12262 structures and arrays containing @code{long double} variables will change
12263 their size as well as function calling convention for function taking
12264 @code{long double} will be modified. Hence they will not be binary
12265 compatible with arrays or structures in code compiled without that switch.
12267 @item -mlarge-data-threshold=@var{number}
12268 @opindex mlarge-data-threshold=@var{number}
12269 When @option{-mcmodel=medium} is specified, the data greater than
12270 @var{threshold} are placed in large data section. This value must be the
12271 same across all object linked into the binary and defaults to 65535.
12275 Use a different function-calling convention, in which functions that
12276 take a fixed number of arguments return with the @code{ret} @var{num}
12277 instruction, which pops their arguments while returning. This saves one
12278 instruction in the caller since there is no need to pop the arguments
12281 You can specify that an individual function is called with this calling
12282 sequence with the function attribute @samp{stdcall}. You can also
12283 override the @option{-mrtd} option by using the function attribute
12284 @samp{cdecl}. @xref{Function Attributes}.
12286 @strong{Warning:} this calling convention is incompatible with the one
12287 normally used on Unix, so you cannot use it if you need to call
12288 libraries compiled with the Unix compiler.
12290 Also, you must provide function prototypes for all functions that
12291 take variable numbers of arguments (including @code{printf});
12292 otherwise incorrect code will be generated for calls to those
12295 In addition, seriously incorrect code will result if you call a
12296 function with too many arguments. (Normally, extra arguments are
12297 harmlessly ignored.)
12299 @item -mregparm=@var{num}
12301 Control how many registers are used to pass integer arguments. By
12302 default, no registers are used to pass arguments, and at most 3
12303 registers can be used. You can control this behavior for a specific
12304 function by using the function attribute @samp{regparm}.
12305 @xref{Function Attributes}.
12307 @strong{Warning:} if you use this switch, and
12308 @var{num} is nonzero, then you must build all modules with the same
12309 value, including any libraries. This includes the system libraries and
12313 @opindex msseregparm
12314 Use SSE register passing conventions for float and double arguments
12315 and return values. You can control this behavior for a specific
12316 function by using the function attribute @samp{sseregparm}.
12317 @xref{Function Attributes}.
12319 @strong{Warning:} if you use this switch then you must build all
12320 modules with the same value, including any libraries. This includes
12321 the system libraries and startup modules.
12323 @item -mvect8-ret-in-mem
12324 @opindex mvect8-ret-in-mem
12325 Return 8-byte vectors in memory instead of MMX registers. This is the
12326 default on Solaris~8 and 9 and VxWorks to match the ABI of the Sun
12327 Studio compilers until version 12. Later compiler versions (starting
12328 with Studio 12 Update~1) follow the ABI used by other x86 targets, which
12329 is the default on Solaris~10 and later. @emph{Only} use this option if
12330 you need to remain compatible with existing code produced by those
12331 previous compiler versions or older versions of GCC.
12340 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
12341 is specified, the significands of results of floating-point operations are
12342 rounded to 24 bits (single precision); @option{-mpc64} rounds the
12343 significands of results of floating-point operations to 53 bits (double
12344 precision) and @option{-mpc80} rounds the significands of results of
12345 floating-point operations to 64 bits (extended double precision), which is
12346 the default. When this option is used, floating-point operations in higher
12347 precisions are not available to the programmer without setting the FPU
12348 control word explicitly.
12350 Setting the rounding of floating-point operations to less than the default
12351 80 bits can speed some programs by 2% or more. Note that some mathematical
12352 libraries assume that extended precision (80 bit) floating-point operations
12353 are enabled by default; routines in such libraries could suffer significant
12354 loss of accuracy, typically through so-called "catastrophic cancellation",
12355 when this option is used to set the precision to less than extended precision.
12357 @item -mstackrealign
12358 @opindex mstackrealign
12359 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
12360 option will generate an alternate prologue and epilogue that realigns the
12361 runtime stack if necessary. This supports mixing legacy codes that keep
12362 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
12363 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
12364 applicable to individual functions.
12366 @item -mpreferred-stack-boundary=@var{num}
12367 @opindex mpreferred-stack-boundary
12368 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
12369 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
12370 the default is 4 (16 bytes or 128 bits).
12372 @item -mincoming-stack-boundary=@var{num}
12373 @opindex mincoming-stack-boundary
12374 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
12375 boundary. If @option{-mincoming-stack-boundary} is not specified,
12376 the one specified by @option{-mpreferred-stack-boundary} will be used.
12378 On Pentium and PentiumPro, @code{double} and @code{long double} values
12379 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
12380 suffer significant run time performance penalties. On Pentium III, the
12381 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
12382 properly if it is not 16 byte aligned.
12384 To ensure proper alignment of this values on the stack, the stack boundary
12385 must be as aligned as that required by any value stored on the stack.
12386 Further, every function must be generated such that it keeps the stack
12387 aligned. Thus calling a function compiled with a higher preferred
12388 stack boundary from a function compiled with a lower preferred stack
12389 boundary will most likely misalign the stack. It is recommended that
12390 libraries that use callbacks always use the default setting.
12392 This extra alignment does consume extra stack space, and generally
12393 increases code size. Code that is sensitive to stack space usage, such
12394 as embedded systems and operating system kernels, may want to reduce the
12395 preferred alignment to @option{-mpreferred-stack-boundary=2}.
12422 @itemx -mno-fsgsbase
12452 These switches enable or disable the use of instructions in the MMX,
12453 SSE, SSE2, SSE3, SSSE3, SSE4.1, AVX, AES, PCLMUL, FSGSBASE, RDRND,
12454 F16C, SSE4A, FMA4, XOP, LWP, ABM, BMI, or 3DNow!@: extended instruction sets.
12455 These extensions are also available as built-in functions: see
12456 @ref{X86 Built-in Functions}, for details of the functions enabled and
12457 disabled by these switches.
12459 To have SSE/SSE2 instructions generated automatically from floating-point
12460 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
12462 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
12463 generates new AVX instructions or AVX equivalence for all SSEx instructions
12466 These options will enable GCC to use these extended instructions in
12467 generated code, even without @option{-mfpmath=sse}. Applications which
12468 perform runtime CPU detection must compile separate files for each
12469 supported architecture, using the appropriate flags. In particular,
12470 the file containing the CPU detection code should be compiled without
12474 @itemx -mno-fused-madd
12475 @opindex mfused-madd
12476 @opindex mno-fused-madd
12477 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
12478 instructions. The default is to use these instructions.
12482 This option instructs GCC to emit a @code{cld} instruction in the prologue
12483 of functions that use string instructions. String instructions depend on
12484 the DF flag to select between autoincrement or autodecrement mode. While the
12485 ABI specifies the DF flag to be cleared on function entry, some operating
12486 systems violate this specification by not clearing the DF flag in their
12487 exception dispatchers. The exception handler can be invoked with the DF flag
12488 set which leads to wrong direction mode, when string instructions are used.
12489 This option can be enabled by default on 32-bit x86 targets by configuring
12490 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
12491 instructions can be suppressed with the @option{-mno-cld} compiler option
12495 @opindex mvzeroupper
12496 This option instructs GCC to emit a @code{vzeroupper} instruction
12497 before a transfer of control flow out of the function to minimize
12498 AVX to SSE transition penalty as well as remove unnecessary zeroupper
12503 This option will enable GCC to use CMPXCHG16B instruction in generated code.
12504 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
12505 data types. This is useful for high resolution counters that could be updated
12506 by multiple processors (or cores). This instruction is generated as part of
12507 atomic built-in functions: see @ref{Atomic Builtins} for details.
12511 This option will enable GCC to use SAHF instruction in generated 64-bit code.
12512 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
12513 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
12514 SAHF are load and store instructions, respectively, for certain status flags.
12515 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
12516 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
12520 This option will enable GCC to use movbe instruction to implement
12521 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
12525 This option will enable built-in functions, @code{__builtin_ia32_crc32qi},
12526 @code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and
12527 @code{__builtin_ia32_crc32di} to generate the crc32 machine instruction.
12531 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
12532 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
12533 to increase precision instead of DIVSS and SQRTSS (and their vectorized
12534 variants) for single precision floating point arguments. These instructions
12535 are generated only when @option{-funsafe-math-optimizations} is enabled
12536 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
12537 Note that while the throughput of the sequence is higher than the throughput
12538 of the non-reciprocal instruction, the precision of the sequence can be
12539 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
12541 Note that GCC implements 1.0f/sqrtf(x) in terms of RSQRTSS (or RSQRTPS)
12542 already with @option{-ffast-math} (or the above option combination), and
12543 doesn't need @option{-mrecip}.
12545 @item -mveclibabi=@var{type}
12546 @opindex mveclibabi
12547 Specifies the ABI type to use for vectorizing intrinsics using an
12548 external library. Supported types are @code{svml} for the Intel short
12549 vector math library and @code{acml} for the AMD math core library style
12550 of interfacing. GCC will currently emit calls to @code{vmldExp2},
12551 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
12552 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
12553 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
12554 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
12555 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
12556 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
12557 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
12558 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
12559 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
12560 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
12561 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
12562 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
12563 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
12564 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
12565 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
12566 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
12567 compatible library will have to be specified at link time.
12569 @item -mabi=@var{name}
12571 Generate code for the specified calling convention. Permissible values
12572 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
12573 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
12574 ABI when targeting Windows. On all other systems, the default is the
12575 SYSV ABI. You can control this behavior for a specific function by
12576 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
12577 @xref{Function Attributes}.
12580 @itemx -mno-push-args
12581 @opindex mpush-args
12582 @opindex mno-push-args
12583 Use PUSH operations to store outgoing parameters. This method is shorter
12584 and usually equally fast as method using SUB/MOV operations and is enabled
12585 by default. In some cases disabling it may improve performance because of
12586 improved scheduling and reduced dependencies.
12588 @item -maccumulate-outgoing-args
12589 @opindex maccumulate-outgoing-args
12590 If enabled, the maximum amount of space required for outgoing arguments will be
12591 computed in the function prologue. This is faster on most modern CPUs
12592 because of reduced dependencies, improved scheduling and reduced stack usage
12593 when preferred stack boundary is not equal to 2. The drawback is a notable
12594 increase in code size. This switch implies @option{-mno-push-args}.
12598 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
12599 on thread-safe exception handling must compile and link all code with the
12600 @option{-mthreads} option. When compiling, @option{-mthreads} defines
12601 @option{-D_MT}; when linking, it links in a special thread helper library
12602 @option{-lmingwthrd} which cleans up per thread exception handling data.
12604 @item -mno-align-stringops
12605 @opindex mno-align-stringops
12606 Do not align destination of inlined string operations. This switch reduces
12607 code size and improves performance in case the destination is already aligned,
12608 but GCC doesn't know about it.
12610 @item -minline-all-stringops
12611 @opindex minline-all-stringops
12612 By default GCC inlines string operations only when destination is known to be
12613 aligned at least to 4 byte boundary. This enables more inlining, increase code
12614 size, but may improve performance of code that depends on fast memcpy, strlen
12615 and memset for short lengths.
12617 @item -minline-stringops-dynamically
12618 @opindex minline-stringops-dynamically
12619 For string operation of unknown size, inline runtime checks so for small
12620 blocks inline code is used, while for large blocks library call is used.
12622 @item -mstringop-strategy=@var{alg}
12623 @opindex mstringop-strategy=@var{alg}
12624 Overwrite internal decision heuristic about particular algorithm to inline
12625 string operation with. The allowed values are @code{rep_byte},
12626 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
12627 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
12628 expanding inline loop, @code{libcall} for always expanding library call.
12630 @item -momit-leaf-frame-pointer
12631 @opindex momit-leaf-frame-pointer
12632 Don't keep the frame pointer in a register for leaf functions. This
12633 avoids the instructions to save, set up and restore frame pointers and
12634 makes an extra register available in leaf functions. The option
12635 @option{-fomit-frame-pointer} removes the frame pointer for all functions
12636 which might make debugging harder.
12638 @item -mtls-direct-seg-refs
12639 @itemx -mno-tls-direct-seg-refs
12640 @opindex mtls-direct-seg-refs
12641 Controls whether TLS variables may be accessed with offsets from the
12642 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
12643 or whether the thread base pointer must be added. Whether or not this
12644 is legal depends on the operating system, and whether it maps the
12645 segment to cover the entire TLS area.
12647 For systems that use GNU libc, the default is on.
12650 @itemx -mno-sse2avx
12652 Specify that the assembler should encode SSE instructions with VEX
12653 prefix. The option @option{-mavx} turns this on by default.
12658 If profiling is active @option{-pg} put the profiling
12659 counter call before prologue.
12660 Note: On x86 architectures the attribute @code{ms_hook_prologue}
12661 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
12664 @itemx -mno-8bit-idiv
12666 On some processors, like Intel Atom, 8bit unsigned integer divide is
12667 much faster than 32bit/64bit integer divide. This option will generate a
12668 runt-time check. If both dividend and divisor are within range of 0
12669 to 255, 8bit unsigned integer divide will be used instead of
12670 32bit/64bit integer divide.
12674 These @samp{-m} switches are supported in addition to the above
12675 on AMD x86-64 processors in 64-bit environments.
12682 Generate code for a 32-bit or 64-bit environment.
12683 The 32-bit environment sets int, long and pointer to 32 bits and
12684 generates code that runs on any i386 system.
12685 The 64-bit environment sets int to 32 bits and long and pointer
12686 to 64 bits and generates code for AMD's x86-64 architecture. For
12687 darwin only the -m64 option turns off the @option{-fno-pic} and
12688 @option{-mdynamic-no-pic} options.
12690 @item -mno-red-zone
12691 @opindex mno-red-zone
12692 Do not use a so called red zone for x86-64 code. The red zone is mandated
12693 by the x86-64 ABI, it is a 128-byte area beyond the location of the
12694 stack pointer that will not be modified by signal or interrupt handlers
12695 and therefore can be used for temporary data without adjusting the stack
12696 pointer. The flag @option{-mno-red-zone} disables this red zone.
12698 @item -mcmodel=small
12699 @opindex mcmodel=small
12700 Generate code for the small code model: the program and its symbols must
12701 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
12702 Programs can be statically or dynamically linked. This is the default
12705 @item -mcmodel=kernel
12706 @opindex mcmodel=kernel
12707 Generate code for the kernel code model. The kernel runs in the
12708 negative 2 GB of the address space.
12709 This model has to be used for Linux kernel code.
12711 @item -mcmodel=medium
12712 @opindex mcmodel=medium
12713 Generate code for the medium model: The program is linked in the lower 2
12714 GB of the address space. Small symbols are also placed there. Symbols
12715 with sizes larger than @option{-mlarge-data-threshold} are put into
12716 large data or bss sections and can be located above 2GB. Programs can
12717 be statically or dynamically linked.
12719 @item -mcmodel=large
12720 @opindex mcmodel=large
12721 Generate code for the large model: This model makes no assumptions
12722 about addresses and sizes of sections.
12725 @node IA-64 Options
12726 @subsection IA-64 Options
12727 @cindex IA-64 Options
12729 These are the @samp{-m} options defined for the Intel IA-64 architecture.
12733 @opindex mbig-endian
12734 Generate code for a big endian target. This is the default for HP-UX@.
12736 @item -mlittle-endian
12737 @opindex mlittle-endian
12738 Generate code for a little endian target. This is the default for AIX5
12744 @opindex mno-gnu-as
12745 Generate (or don't) code for the GNU assembler. This is the default.
12746 @c Also, this is the default if the configure option @option{--with-gnu-as}
12752 @opindex mno-gnu-ld
12753 Generate (or don't) code for the GNU linker. This is the default.
12754 @c Also, this is the default if the configure option @option{--with-gnu-ld}
12759 Generate code that does not use a global pointer register. The result
12760 is not position independent code, and violates the IA-64 ABI@.
12762 @item -mvolatile-asm-stop
12763 @itemx -mno-volatile-asm-stop
12764 @opindex mvolatile-asm-stop
12765 @opindex mno-volatile-asm-stop
12766 Generate (or don't) a stop bit immediately before and after volatile asm
12769 @item -mregister-names
12770 @itemx -mno-register-names
12771 @opindex mregister-names
12772 @opindex mno-register-names
12773 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
12774 the stacked registers. This may make assembler output more readable.
12780 Disable (or enable) optimizations that use the small data section. This may
12781 be useful for working around optimizer bugs.
12783 @item -mconstant-gp
12784 @opindex mconstant-gp
12785 Generate code that uses a single constant global pointer value. This is
12786 useful when compiling kernel code.
12790 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
12791 This is useful when compiling firmware code.
12793 @item -minline-float-divide-min-latency
12794 @opindex minline-float-divide-min-latency
12795 Generate code for inline divides of floating point values
12796 using the minimum latency algorithm.
12798 @item -minline-float-divide-max-throughput
12799 @opindex minline-float-divide-max-throughput
12800 Generate code for inline divides of floating point values
12801 using the maximum throughput algorithm.
12803 @item -mno-inline-float-divide
12804 @opindex mno-inline-float-divide
12805 Do not generate inline code for divides of floating point values.
12807 @item -minline-int-divide-min-latency
12808 @opindex minline-int-divide-min-latency
12809 Generate code for inline divides of integer values
12810 using the minimum latency algorithm.
12812 @item -minline-int-divide-max-throughput
12813 @opindex minline-int-divide-max-throughput
12814 Generate code for inline divides of integer values
12815 using the maximum throughput algorithm.
12817 @item -mno-inline-int-divide
12818 @opindex mno-inline-int-divide
12819 Do not generate inline code for divides of integer values.
12821 @item -minline-sqrt-min-latency
12822 @opindex minline-sqrt-min-latency
12823 Generate code for inline square roots
12824 using the minimum latency algorithm.
12826 @item -minline-sqrt-max-throughput
12827 @opindex minline-sqrt-max-throughput
12828 Generate code for inline square roots
12829 using the maximum throughput algorithm.
12831 @item -mno-inline-sqrt
12832 @opindex mno-inline-sqrt
12833 Do not generate inline code for sqrt.
12836 @itemx -mno-fused-madd
12837 @opindex mfused-madd
12838 @opindex mno-fused-madd
12839 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
12840 instructions. The default is to use these instructions.
12842 @item -mno-dwarf2-asm
12843 @itemx -mdwarf2-asm
12844 @opindex mno-dwarf2-asm
12845 @opindex mdwarf2-asm
12846 Don't (or do) generate assembler code for the DWARF2 line number debugging
12847 info. This may be useful when not using the GNU assembler.
12849 @item -mearly-stop-bits
12850 @itemx -mno-early-stop-bits
12851 @opindex mearly-stop-bits
12852 @opindex mno-early-stop-bits
12853 Allow stop bits to be placed earlier than immediately preceding the
12854 instruction that triggered the stop bit. This can improve instruction
12855 scheduling, but does not always do so.
12857 @item -mfixed-range=@var{register-range}
12858 @opindex mfixed-range
12859 Generate code treating the given register range as fixed registers.
12860 A fixed register is one that the register allocator can not use. This is
12861 useful when compiling kernel code. A register range is specified as
12862 two registers separated by a dash. Multiple register ranges can be
12863 specified separated by a comma.
12865 @item -mtls-size=@var{tls-size}
12867 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
12870 @item -mtune=@var{cpu-type}
12872 Tune the instruction scheduling for a particular CPU, Valid values are
12873 itanium, itanium1, merced, itanium2, and mckinley.
12879 Generate code for a 32-bit or 64-bit environment.
12880 The 32-bit environment sets int, long and pointer to 32 bits.
12881 The 64-bit environment sets int to 32 bits and long and pointer
12882 to 64 bits. These are HP-UX specific flags.
12884 @item -mno-sched-br-data-spec
12885 @itemx -msched-br-data-spec
12886 @opindex mno-sched-br-data-spec
12887 @opindex msched-br-data-spec
12888 (Dis/En)able data speculative scheduling before reload.
12889 This will result in generation of the ld.a instructions and
12890 the corresponding check instructions (ld.c / chk.a).
12891 The default is 'disable'.
12893 @item -msched-ar-data-spec
12894 @itemx -mno-sched-ar-data-spec
12895 @opindex msched-ar-data-spec
12896 @opindex mno-sched-ar-data-spec
12897 (En/Dis)able data speculative scheduling after reload.
12898 This will result in generation of the ld.a instructions and
12899 the corresponding check instructions (ld.c / chk.a).
12900 The default is 'enable'.
12902 @item -mno-sched-control-spec
12903 @itemx -msched-control-spec
12904 @opindex mno-sched-control-spec
12905 @opindex msched-control-spec
12906 (Dis/En)able control speculative scheduling. This feature is
12907 available only during region scheduling (i.e.@: before reload).
12908 This will result in generation of the ld.s instructions and
12909 the corresponding check instructions chk.s .
12910 The default is 'disable'.
12912 @item -msched-br-in-data-spec
12913 @itemx -mno-sched-br-in-data-spec
12914 @opindex msched-br-in-data-spec
12915 @opindex mno-sched-br-in-data-spec
12916 (En/Dis)able speculative scheduling of the instructions that
12917 are dependent on the data speculative loads before reload.
12918 This is effective only with @option{-msched-br-data-spec} enabled.
12919 The default is 'enable'.
12921 @item -msched-ar-in-data-spec
12922 @itemx -mno-sched-ar-in-data-spec
12923 @opindex msched-ar-in-data-spec
12924 @opindex mno-sched-ar-in-data-spec
12925 (En/Dis)able speculative scheduling of the instructions that
12926 are dependent on the data speculative loads after reload.
12927 This is effective only with @option{-msched-ar-data-spec} enabled.
12928 The default is 'enable'.
12930 @item -msched-in-control-spec
12931 @itemx -mno-sched-in-control-spec
12932 @opindex msched-in-control-spec
12933 @opindex mno-sched-in-control-spec
12934 (En/Dis)able speculative scheduling of the instructions that
12935 are dependent on the control speculative loads.
12936 This is effective only with @option{-msched-control-spec} enabled.
12937 The default is 'enable'.
12939 @item -mno-sched-prefer-non-data-spec-insns
12940 @itemx -msched-prefer-non-data-spec-insns
12941 @opindex mno-sched-prefer-non-data-spec-insns
12942 @opindex msched-prefer-non-data-spec-insns
12943 If enabled, data speculative instructions will be chosen for schedule
12944 only if there are no other choices at the moment. This will make
12945 the use of the data speculation much more conservative.
12946 The default is 'disable'.
12948 @item -mno-sched-prefer-non-control-spec-insns
12949 @itemx -msched-prefer-non-control-spec-insns
12950 @opindex mno-sched-prefer-non-control-spec-insns
12951 @opindex msched-prefer-non-control-spec-insns
12952 If enabled, control speculative instructions will be chosen for schedule
12953 only if there are no other choices at the moment. This will make
12954 the use of the control speculation much more conservative.
12955 The default is 'disable'.
12957 @item -mno-sched-count-spec-in-critical-path
12958 @itemx -msched-count-spec-in-critical-path
12959 @opindex mno-sched-count-spec-in-critical-path
12960 @opindex msched-count-spec-in-critical-path
12961 If enabled, speculative dependencies will be considered during
12962 computation of the instructions priorities. This will make the use of the
12963 speculation a bit more conservative.
12964 The default is 'disable'.
12966 @item -msched-spec-ldc
12967 @opindex msched-spec-ldc
12968 Use a simple data speculation check. This option is on by default.
12970 @item -msched-control-spec-ldc
12971 @opindex msched-spec-ldc
12972 Use a simple check for control speculation. This option is on by default.
12974 @item -msched-stop-bits-after-every-cycle
12975 @opindex msched-stop-bits-after-every-cycle
12976 Place a stop bit after every cycle when scheduling. This option is on
12979 @item -msched-fp-mem-deps-zero-cost
12980 @opindex msched-fp-mem-deps-zero-cost
12981 Assume that floating-point stores and loads are not likely to cause a conflict
12982 when placed into the same instruction group. This option is disabled by
12985 @item -msel-sched-dont-check-control-spec
12986 @opindex msel-sched-dont-check-control-spec
12987 Generate checks for control speculation in selective scheduling.
12988 This flag is disabled by default.
12990 @item -msched-max-memory-insns=@var{max-insns}
12991 @opindex msched-max-memory-insns
12992 Limit on the number of memory insns per instruction group, giving lower
12993 priority to subsequent memory insns attempting to schedule in the same
12994 instruction group. Frequently useful to prevent cache bank conflicts.
12995 The default value is 1.
12997 @item -msched-max-memory-insns-hard-limit
12998 @opindex msched-max-memory-insns-hard-limit
12999 Disallow more than `msched-max-memory-insns' in instruction group.
13000 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
13001 when limit is reached but may still schedule memory operations.
13005 @node IA-64/VMS Options
13006 @subsection IA-64/VMS Options
13008 These @samp{-m} options are defined for the IA-64/VMS implementations:
13011 @item -mvms-return-codes
13012 @opindex mvms-return-codes
13013 Return VMS condition codes from main. The default is to return POSIX
13014 style condition (e.g.@ error) codes.
13016 @item -mdebug-main=@var{prefix}
13017 @opindex mdebug-main=@var{prefix}
13018 Flag the first routine whose name starts with @var{prefix} as the main
13019 routine for the debugger.
13023 Default to 64bit memory allocation routines.
13027 @subsection LM32 Options
13028 @cindex LM32 options
13030 These @option{-m} options are defined for the Lattice Mico32 architecture:
13033 @item -mbarrel-shift-enabled
13034 @opindex mbarrel-shift-enabled
13035 Enable barrel-shift instructions.
13037 @item -mdivide-enabled
13038 @opindex mdivide-enabled
13039 Enable divide and modulus instructions.
13041 @item -mmultiply-enabled
13042 @opindex multiply-enabled
13043 Enable multiply instructions.
13045 @item -msign-extend-enabled
13046 @opindex msign-extend-enabled
13047 Enable sign extend instructions.
13049 @item -muser-enabled
13050 @opindex muser-enabled
13051 Enable user-defined instructions.
13056 @subsection M32C Options
13057 @cindex M32C options
13060 @item -mcpu=@var{name}
13062 Select the CPU for which code is generated. @var{name} may be one of
13063 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
13064 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
13065 the M32C/80 series.
13069 Specifies that the program will be run on the simulator. This causes
13070 an alternate runtime library to be linked in which supports, for
13071 example, file I/O@. You must not use this option when generating
13072 programs that will run on real hardware; you must provide your own
13073 runtime library for whatever I/O functions are needed.
13075 @item -memregs=@var{number}
13077 Specifies the number of memory-based pseudo-registers GCC will use
13078 during code generation. These pseudo-registers will be used like real
13079 registers, so there is a tradeoff between GCC's ability to fit the
13080 code into available registers, and the performance penalty of using
13081 memory instead of registers. Note that all modules in a program must
13082 be compiled with the same value for this option. Because of that, you
13083 must not use this option with the default runtime libraries gcc
13088 @node M32R/D Options
13089 @subsection M32R/D Options
13090 @cindex M32R/D options
13092 These @option{-m} options are defined for Renesas M32R/D architectures:
13097 Generate code for the M32R/2@.
13101 Generate code for the M32R/X@.
13105 Generate code for the M32R@. This is the default.
13107 @item -mmodel=small
13108 @opindex mmodel=small
13109 Assume all objects live in the lower 16MB of memory (so that their addresses
13110 can be loaded with the @code{ld24} instruction), and assume all subroutines
13111 are reachable with the @code{bl} instruction.
13112 This is the default.
13114 The addressability of a particular object can be set with the
13115 @code{model} attribute.
13117 @item -mmodel=medium
13118 @opindex mmodel=medium
13119 Assume objects may be anywhere in the 32-bit address space (the compiler
13120 will generate @code{seth/add3} instructions to load their addresses), and
13121 assume all subroutines are reachable with the @code{bl} instruction.
13123 @item -mmodel=large
13124 @opindex mmodel=large
13125 Assume objects may be anywhere in the 32-bit address space (the compiler
13126 will generate @code{seth/add3} instructions to load their addresses), and
13127 assume subroutines may not be reachable with the @code{bl} instruction
13128 (the compiler will generate the much slower @code{seth/add3/jl}
13129 instruction sequence).
13132 @opindex msdata=none
13133 Disable use of the small data area. Variables will be put into
13134 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
13135 @code{section} attribute has been specified).
13136 This is the default.
13138 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
13139 Objects may be explicitly put in the small data area with the
13140 @code{section} attribute using one of these sections.
13142 @item -msdata=sdata
13143 @opindex msdata=sdata
13144 Put small global and static data in the small data area, but do not
13145 generate special code to reference them.
13148 @opindex msdata=use
13149 Put small global and static data in the small data area, and generate
13150 special instructions to reference them.
13154 @cindex smaller data references
13155 Put global and static objects less than or equal to @var{num} bytes
13156 into the small data or bss sections instead of the normal data or bss
13157 sections. The default value of @var{num} is 8.
13158 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
13159 for this option to have any effect.
13161 All modules should be compiled with the same @option{-G @var{num}} value.
13162 Compiling with different values of @var{num} may or may not work; if it
13163 doesn't the linker will give an error message---incorrect code will not be
13168 Makes the M32R specific code in the compiler display some statistics
13169 that might help in debugging programs.
13171 @item -malign-loops
13172 @opindex malign-loops
13173 Align all loops to a 32-byte boundary.
13175 @item -mno-align-loops
13176 @opindex mno-align-loops
13177 Do not enforce a 32-byte alignment for loops. This is the default.
13179 @item -missue-rate=@var{number}
13180 @opindex missue-rate=@var{number}
13181 Issue @var{number} instructions per cycle. @var{number} can only be 1
13184 @item -mbranch-cost=@var{number}
13185 @opindex mbranch-cost=@var{number}
13186 @var{number} can only be 1 or 2. If it is 1 then branches will be
13187 preferred over conditional code, if it is 2, then the opposite will
13190 @item -mflush-trap=@var{number}
13191 @opindex mflush-trap=@var{number}
13192 Specifies the trap number to use to flush the cache. The default is
13193 12. Valid numbers are between 0 and 15 inclusive.
13195 @item -mno-flush-trap
13196 @opindex mno-flush-trap
13197 Specifies that the cache cannot be flushed by using a trap.
13199 @item -mflush-func=@var{name}
13200 @opindex mflush-func=@var{name}
13201 Specifies the name of the operating system function to call to flush
13202 the cache. The default is @emph{_flush_cache}, but a function call
13203 will only be used if a trap is not available.
13205 @item -mno-flush-func
13206 @opindex mno-flush-func
13207 Indicates that there is no OS function for flushing the cache.
13211 @node M680x0 Options
13212 @subsection M680x0 Options
13213 @cindex M680x0 options
13215 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
13216 The default settings depend on which architecture was selected when
13217 the compiler was configured; the defaults for the most common choices
13221 @item -march=@var{arch}
13223 Generate code for a specific M680x0 or ColdFire instruction set
13224 architecture. Permissible values of @var{arch} for M680x0
13225 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
13226 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
13227 architectures are selected according to Freescale's ISA classification
13228 and the permissible values are: @samp{isaa}, @samp{isaaplus},
13229 @samp{isab} and @samp{isac}.
13231 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
13232 code for a ColdFire target. The @var{arch} in this macro is one of the
13233 @option{-march} arguments given above.
13235 When used together, @option{-march} and @option{-mtune} select code
13236 that runs on a family of similar processors but that is optimized
13237 for a particular microarchitecture.
13239 @item -mcpu=@var{cpu}
13241 Generate code for a specific M680x0 or ColdFire processor.
13242 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
13243 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
13244 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
13245 below, which also classifies the CPUs into families:
13247 @multitable @columnfractions 0.20 0.80
13248 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
13249 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
13250 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
13251 @item @samp{5206e} @tab @samp{5206e}
13252 @item @samp{5208} @tab @samp{5207} @samp{5208}
13253 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
13254 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
13255 @item @samp{5216} @tab @samp{5214} @samp{5216}
13256 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
13257 @item @samp{5225} @tab @samp{5224} @samp{5225}
13258 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
13259 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
13260 @item @samp{5249} @tab @samp{5249}
13261 @item @samp{5250} @tab @samp{5250}
13262 @item @samp{5271} @tab @samp{5270} @samp{5271}
13263 @item @samp{5272} @tab @samp{5272}
13264 @item @samp{5275} @tab @samp{5274} @samp{5275}
13265 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
13266 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
13267 @item @samp{5307} @tab @samp{5307}
13268 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
13269 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
13270 @item @samp{5407} @tab @samp{5407}
13271 @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}
13274 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
13275 @var{arch} is compatible with @var{cpu}. Other combinations of
13276 @option{-mcpu} and @option{-march} are rejected.
13278 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
13279 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
13280 where the value of @var{family} is given by the table above.
13282 @item -mtune=@var{tune}
13284 Tune the code for a particular microarchitecture, within the
13285 constraints set by @option{-march} and @option{-mcpu}.
13286 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
13287 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
13288 and @samp{cpu32}. The ColdFire microarchitectures
13289 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
13291 You can also use @option{-mtune=68020-40} for code that needs
13292 to run relatively well on 68020, 68030 and 68040 targets.
13293 @option{-mtune=68020-60} is similar but includes 68060 targets
13294 as well. These two options select the same tuning decisions as
13295 @option{-m68020-40} and @option{-m68020-60} respectively.
13297 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
13298 when tuning for 680x0 architecture @var{arch}. It also defines
13299 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
13300 option is used. If gcc is tuning for a range of architectures,
13301 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
13302 it defines the macros for every architecture in the range.
13304 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
13305 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
13306 of the arguments given above.
13312 Generate output for a 68000. This is the default
13313 when the compiler is configured for 68000-based systems.
13314 It is equivalent to @option{-march=68000}.
13316 Use this option for microcontrollers with a 68000 or EC000 core,
13317 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
13321 Generate output for a 68010. This is the default
13322 when the compiler is configured for 68010-based systems.
13323 It is equivalent to @option{-march=68010}.
13329 Generate output for a 68020. This is the default
13330 when the compiler is configured for 68020-based systems.
13331 It is equivalent to @option{-march=68020}.
13335 Generate output for a 68030. This is the default when the compiler is
13336 configured for 68030-based systems. It is equivalent to
13337 @option{-march=68030}.
13341 Generate output for a 68040. This is the default when the compiler is
13342 configured for 68040-based systems. It is equivalent to
13343 @option{-march=68040}.
13345 This option inhibits the use of 68881/68882 instructions that have to be
13346 emulated by software on the 68040. Use this option if your 68040 does not
13347 have code to emulate those instructions.
13351 Generate output for a 68060. This is the default when the compiler is
13352 configured for 68060-based systems. It is equivalent to
13353 @option{-march=68060}.
13355 This option inhibits the use of 68020 and 68881/68882 instructions that
13356 have to be emulated by software on the 68060. Use this option if your 68060
13357 does not have code to emulate those instructions.
13361 Generate output for a CPU32. This is the default
13362 when the compiler is configured for CPU32-based systems.
13363 It is equivalent to @option{-march=cpu32}.
13365 Use this option for microcontrollers with a
13366 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
13367 68336, 68340, 68341, 68349 and 68360.
13371 Generate output for a 520X ColdFire CPU@. This is the default
13372 when the compiler is configured for 520X-based systems.
13373 It is equivalent to @option{-mcpu=5206}, and is now deprecated
13374 in favor of that option.
13376 Use this option for microcontroller with a 5200 core, including
13377 the MCF5202, MCF5203, MCF5204 and MCF5206.
13381 Generate output for a 5206e ColdFire CPU@. The option is now
13382 deprecated in favor of the equivalent @option{-mcpu=5206e}.
13386 Generate output for a member of the ColdFire 528X family.
13387 The option is now deprecated in favor of the equivalent
13388 @option{-mcpu=528x}.
13392 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
13393 in favor of the equivalent @option{-mcpu=5307}.
13397 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
13398 in favor of the equivalent @option{-mcpu=5407}.
13402 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
13403 This includes use of hardware floating point instructions.
13404 The option is equivalent to @option{-mcpu=547x}, and is now
13405 deprecated in favor of that option.
13409 Generate output for a 68040, without using any of the new instructions.
13410 This results in code which can run relatively efficiently on either a
13411 68020/68881 or a 68030 or a 68040. The generated code does use the
13412 68881 instructions that are emulated on the 68040.
13414 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
13418 Generate output for a 68060, without using any of the new instructions.
13419 This results in code which can run relatively efficiently on either a
13420 68020/68881 or a 68030 or a 68040. The generated code does use the
13421 68881 instructions that are emulated on the 68060.
13423 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
13427 @opindex mhard-float
13429 Generate floating-point instructions. This is the default for 68020
13430 and above, and for ColdFire devices that have an FPU@. It defines the
13431 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
13432 on ColdFire targets.
13435 @opindex msoft-float
13436 Do not generate floating-point instructions; use library calls instead.
13437 This is the default for 68000, 68010, and 68832 targets. It is also
13438 the default for ColdFire devices that have no FPU.
13444 Generate (do not generate) ColdFire hardware divide and remainder
13445 instructions. If @option{-march} is used without @option{-mcpu},
13446 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
13447 architectures. Otherwise, the default is taken from the target CPU
13448 (either the default CPU, or the one specified by @option{-mcpu}). For
13449 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
13450 @option{-mcpu=5206e}.
13452 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
13456 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13457 Additionally, parameters passed on the stack are also aligned to a
13458 16-bit boundary even on targets whose API mandates promotion to 32-bit.
13462 Do not consider type @code{int} to be 16 bits wide. This is the default.
13465 @itemx -mno-bitfield
13466 @opindex mnobitfield
13467 @opindex mno-bitfield
13468 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
13469 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
13473 Do use the bit-field instructions. The @option{-m68020} option implies
13474 @option{-mbitfield}. This is the default if you use a configuration
13475 designed for a 68020.
13479 Use a different function-calling convention, in which functions
13480 that take a fixed number of arguments return with the @code{rtd}
13481 instruction, which pops their arguments while returning. This
13482 saves one instruction in the caller since there is no need to pop
13483 the arguments there.
13485 This calling convention is incompatible with the one normally
13486 used on Unix, so you cannot use it if you need to call libraries
13487 compiled with the Unix compiler.
13489 Also, you must provide function prototypes for all functions that
13490 take variable numbers of arguments (including @code{printf});
13491 otherwise incorrect code will be generated for calls to those
13494 In addition, seriously incorrect code will result if you call a
13495 function with too many arguments. (Normally, extra arguments are
13496 harmlessly ignored.)
13498 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
13499 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
13503 Do not use the calling conventions selected by @option{-mrtd}.
13504 This is the default.
13507 @itemx -mno-align-int
13508 @opindex malign-int
13509 @opindex mno-align-int
13510 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
13511 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
13512 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
13513 Aligning variables on 32-bit boundaries produces code that runs somewhat
13514 faster on processors with 32-bit busses at the expense of more memory.
13516 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
13517 align structures containing the above types differently than
13518 most published application binary interface specifications for the m68k.
13522 Use the pc-relative addressing mode of the 68000 directly, instead of
13523 using a global offset table. At present, this option implies @option{-fpic},
13524 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
13525 not presently supported with @option{-mpcrel}, though this could be supported for
13526 68020 and higher processors.
13528 @item -mno-strict-align
13529 @itemx -mstrict-align
13530 @opindex mno-strict-align
13531 @opindex mstrict-align
13532 Do not (do) assume that unaligned memory references will be handled by
13536 Generate code that allows the data segment to be located in a different
13537 area of memory from the text segment. This allows for execute in place in
13538 an environment without virtual memory management. This option implies
13541 @item -mno-sep-data
13542 Generate code that assumes that the data segment follows the text segment.
13543 This is the default.
13545 @item -mid-shared-library
13546 Generate code that supports shared libraries via the library ID method.
13547 This allows for execute in place and shared libraries in an environment
13548 without virtual memory management. This option implies @option{-fPIC}.
13550 @item -mno-id-shared-library
13551 Generate code that doesn't assume ID based shared libraries are being used.
13552 This is the default.
13554 @item -mshared-library-id=n
13555 Specified the identification number of the ID based shared library being
13556 compiled. Specifying a value of 0 will generate more compact code, specifying
13557 other values will force the allocation of that number to the current
13558 library but is no more space or time efficient than omitting this option.
13564 When generating position-independent code for ColdFire, generate code
13565 that works if the GOT has more than 8192 entries. This code is
13566 larger and slower than code generated without this option. On M680x0
13567 processors, this option is not needed; @option{-fPIC} suffices.
13569 GCC normally uses a single instruction to load values from the GOT@.
13570 While this is relatively efficient, it only works if the GOT
13571 is smaller than about 64k. Anything larger causes the linker
13572 to report an error such as:
13574 @cindex relocation truncated to fit (ColdFire)
13576 relocation truncated to fit: R_68K_GOT16O foobar
13579 If this happens, you should recompile your code with @option{-mxgot}.
13580 It should then work with very large GOTs. However, code generated with
13581 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
13582 the value of a global symbol.
13584 Note that some linkers, including newer versions of the GNU linker,
13585 can create multiple GOTs and sort GOT entries. If you have such a linker,
13586 you should only need to use @option{-mxgot} when compiling a single
13587 object file that accesses more than 8192 GOT entries. Very few do.
13589 These options have no effect unless GCC is generating
13590 position-independent code.
13594 @node M68hc1x Options
13595 @subsection M68hc1x Options
13596 @cindex M68hc1x options
13598 These are the @samp{-m} options defined for the 68hc11 and 68hc12
13599 microcontrollers. The default values for these options depends on
13600 which style of microcontroller was selected when the compiler was configured;
13601 the defaults for the most common choices are given below.
13608 Generate output for a 68HC11. This is the default
13609 when the compiler is configured for 68HC11-based systems.
13615 Generate output for a 68HC12. This is the default
13616 when the compiler is configured for 68HC12-based systems.
13622 Generate output for a 68HCS12.
13624 @item -mauto-incdec
13625 @opindex mauto-incdec
13626 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
13633 Enable the use of 68HC12 min and max instructions.
13636 @itemx -mno-long-calls
13637 @opindex mlong-calls
13638 @opindex mno-long-calls
13639 Treat all calls as being far away (near). If calls are assumed to be
13640 far away, the compiler will use the @code{call} instruction to
13641 call a function and the @code{rtc} instruction for returning.
13645 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13647 @item -msoft-reg-count=@var{count}
13648 @opindex msoft-reg-count
13649 Specify the number of pseudo-soft registers which are used for the
13650 code generation. The maximum number is 32. Using more pseudo-soft
13651 register may or may not result in better code depending on the program.
13652 The default is 4 for 68HC11 and 2 for 68HC12.
13656 @node MCore Options
13657 @subsection MCore Options
13658 @cindex MCore options
13660 These are the @samp{-m} options defined for the Motorola M*Core
13666 @itemx -mno-hardlit
13668 @opindex mno-hardlit
13669 Inline constants into the code stream if it can be done in two
13670 instructions or less.
13676 Use the divide instruction. (Enabled by default).
13678 @item -mrelax-immediate
13679 @itemx -mno-relax-immediate
13680 @opindex mrelax-immediate
13681 @opindex mno-relax-immediate
13682 Allow arbitrary sized immediates in bit operations.
13684 @item -mwide-bitfields
13685 @itemx -mno-wide-bitfields
13686 @opindex mwide-bitfields
13687 @opindex mno-wide-bitfields
13688 Always treat bit-fields as int-sized.
13690 @item -m4byte-functions
13691 @itemx -mno-4byte-functions
13692 @opindex m4byte-functions
13693 @opindex mno-4byte-functions
13694 Force all functions to be aligned to a four byte boundary.
13696 @item -mcallgraph-data
13697 @itemx -mno-callgraph-data
13698 @opindex mcallgraph-data
13699 @opindex mno-callgraph-data
13700 Emit callgraph information.
13703 @itemx -mno-slow-bytes
13704 @opindex mslow-bytes
13705 @opindex mno-slow-bytes
13706 Prefer word access when reading byte quantities.
13708 @item -mlittle-endian
13709 @itemx -mbig-endian
13710 @opindex mlittle-endian
13711 @opindex mbig-endian
13712 Generate code for a little endian target.
13718 Generate code for the 210 processor.
13722 Assume that run-time support has been provided and so omit the
13723 simulator library (@file{libsim.a)} from the linker command line.
13725 @item -mstack-increment=@var{size}
13726 @opindex mstack-increment
13727 Set the maximum amount for a single stack increment operation. Large
13728 values can increase the speed of programs which contain functions
13729 that need a large amount of stack space, but they can also trigger a
13730 segmentation fault if the stack is extended too much. The default
13736 @subsection MeP Options
13737 @cindex MeP options
13743 Enables the @code{abs} instruction, which is the absolute difference
13744 between two registers.
13748 Enables all the optional instructions - average, multiply, divide, bit
13749 operations, leading zero, absolute difference, min/max, clip, and
13755 Enables the @code{ave} instruction, which computes the average of two
13758 @item -mbased=@var{n}
13760 Variables of size @var{n} bytes or smaller will be placed in the
13761 @code{.based} section by default. Based variables use the @code{$tp}
13762 register as a base register, and there is a 128 byte limit to the
13763 @code{.based} section.
13767 Enables the bit operation instructions - bit test (@code{btstm}), set
13768 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
13769 test-and-set (@code{tas}).
13771 @item -mc=@var{name}
13773 Selects which section constant data will be placed in. @var{name} may
13774 be @code{tiny}, @code{near}, or @code{far}.
13778 Enables the @code{clip} instruction. Note that @code{-mclip} is not
13779 useful unless you also provide @code{-mminmax}.
13781 @item -mconfig=@var{name}
13783 Selects one of the build-in core configurations. Each MeP chip has
13784 one or more modules in it; each module has a core CPU and a variety of
13785 coprocessors, optional instructions, and peripherals. The
13786 @code{MeP-Integrator} tool, not part of GCC, provides these
13787 configurations through this option; using this option is the same as
13788 using all the corresponding command line options. The default
13789 configuration is @code{default}.
13793 Enables the coprocessor instructions. By default, this is a 32-bit
13794 coprocessor. Note that the coprocessor is normally enabled via the
13795 @code{-mconfig=} option.
13799 Enables the 32-bit coprocessor's instructions.
13803 Enables the 64-bit coprocessor's instructions.
13807 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
13811 Causes constant variables to be placed in the @code{.near} section.
13815 Enables the @code{div} and @code{divu} instructions.
13819 Generate big-endian code.
13823 Generate little-endian code.
13825 @item -mio-volatile
13826 @opindex mio-volatile
13827 Tells the compiler that any variable marked with the @code{io}
13828 attribute is to be considered volatile.
13832 Causes variables to be assigned to the @code{.far} section by default.
13836 Enables the @code{leadz} (leading zero) instruction.
13840 Causes variables to be assigned to the @code{.near} section by default.
13844 Enables the @code{min} and @code{max} instructions.
13848 Enables the multiplication and multiply-accumulate instructions.
13852 Disables all the optional instructions enabled by @code{-mall-opts}.
13856 Enables the @code{repeat} and @code{erepeat} instructions, used for
13857 low-overhead looping.
13861 Causes all variables to default to the @code{.tiny} section. Note
13862 that there is a 65536 byte limit to this section. Accesses to these
13863 variables use the @code{%gp} base register.
13867 Enables the saturation instructions. Note that the compiler does not
13868 currently generate these itself, but this option is included for
13869 compatibility with other tools, like @code{as}.
13873 Link the SDRAM-based runtime instead of the default ROM-based runtime.
13877 Link the simulator runtime libraries.
13881 Link the simulator runtime libraries, excluding built-in support
13882 for reset and exception vectors and tables.
13886 Causes all functions to default to the @code{.far} section. Without
13887 this option, functions default to the @code{.near} section.
13889 @item -mtiny=@var{n}
13891 Variables that are @var{n} bytes or smaller will be allocated to the
13892 @code{.tiny} section. These variables use the @code{$gp} base
13893 register. The default for this option is 4, but note that there's a
13894 65536 byte limit to the @code{.tiny} section.
13898 @node MicroBlaze Options
13899 @subsection MicroBlaze Options
13900 @cindex MicroBlaze Options
13905 @opindex msoft-float
13906 Use software emulation for floating point (default).
13909 @opindex mhard-float
13910 Use hardware floating point instructions.
13914 Do not optimize block moves, use @code{memcpy}.
13916 @item -mno-clearbss
13917 @opindex mno-clearbss
13918 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
13920 @item -mcpu=@var{cpu-type}
13922 Use features of and schedule code for given CPU.
13923 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
13924 where @var{X} is a major version, @var{YY} is the minor version, and
13925 @var{Z} is compatiblity code. Example values are @samp{v3.00.a},
13926 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
13928 @item -mxl-soft-mul
13929 @opindex mxl-soft-mul
13930 Use software multiply emulation (default).
13932 @item -mxl-soft-div
13933 @opindex mxl-soft-div
13934 Use software emulation for divides (default).
13936 @item -mxl-barrel-shift
13937 @opindex mxl-barrel-shift
13938 Use the hardware barrel shifter.
13940 @item -mxl-pattern-compare
13941 @opindex mxl-pattern-compare
13942 Use pattern compare instructions.
13944 @item -msmall-divides
13945 @opindex msmall-divides
13946 Use table lookup optimization for small signed integer divisions.
13948 @item -mxl-stack-check
13949 @opindex mxl-stack-check
13950 This option is deprecated. Use -fstack-check instead.
13953 @opindex mxl-gp-opt
13954 Use GP relative sdata/sbss sections.
13956 @item -mxl-multiply-high
13957 @opindex mxl-multiply-high
13958 Use multiply high instructions for high part of 32x32 multiply.
13960 @item -mxl-float-convert
13961 @opindex mxl-float-convert
13962 Use hardware floating point converstion instructions.
13964 @item -mxl-float-sqrt
13965 @opindex mxl-float-sqrt
13966 Use hardware floating point square root instruction.
13968 @item -mxl-mode-@var{app-model}
13969 Select application model @var{app-model}. Valid models are
13972 normal executable (default), uses startup code @file{crt0.o}.
13975 for use with Xilinx Microprocessor Debugger (XMD) based
13976 software intrusive debug agent called xmdstub. This uses startup file
13977 @file{crt1.o} and sets the start address of the program to be 0x800.
13980 for applications that are loaded using a bootloader.
13981 This model uses startup file @file{crt2.o} which does not contain a processor
13982 reset vector handler. This is suitable for transferring control on a
13983 processor reset to the bootloader rather than the application.
13986 for applications that do not require any of the
13987 MicroBlaze vectors. This option may be useful for applications running
13988 within a monitoring application. This model uses @file{crt3.o} as a startup file.
13991 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
13992 @option{-mxl-mode-@var{app-model}}.
13997 @subsection MIPS Options
13998 @cindex MIPS options
14004 Generate big-endian code.
14008 Generate little-endian code. This is the default for @samp{mips*el-*-*}
14011 @item -march=@var{arch}
14013 Generate code that will run on @var{arch}, which can be the name of a
14014 generic MIPS ISA, or the name of a particular processor.
14016 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
14017 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
14018 The processor names are:
14019 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
14020 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
14021 @samp{5kc}, @samp{5kf},
14023 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
14024 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
14025 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
14026 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
14027 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
14028 @samp{loongson2e}, @samp{loongson2f},
14032 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
14033 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
14034 @samp{rm7000}, @samp{rm9000},
14035 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
14038 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
14039 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
14041 The special value @samp{from-abi} selects the
14042 most compatible architecture for the selected ABI (that is,
14043 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
14045 Native Linux/GNU toolchains also support the value @samp{native},
14046 which selects the best architecture option for the host processor.
14047 @option{-march=native} has no effect if GCC does not recognize
14050 In processor names, a final @samp{000} can be abbreviated as @samp{k}
14051 (for example, @samp{-march=r2k}). Prefixes are optional, and
14052 @samp{vr} may be written @samp{r}.
14054 Names of the form @samp{@var{n}f2_1} refer to processors with
14055 FPUs clocked at half the rate of the core, names of the form
14056 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
14057 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
14058 processors with FPUs clocked a ratio of 3:2 with respect to the core.
14059 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
14060 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
14061 accepted as synonyms for @samp{@var{n}f1_1}.
14063 GCC defines two macros based on the value of this option. The first
14064 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
14065 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
14066 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
14067 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
14068 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
14070 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
14071 above. In other words, it will have the full prefix and will not
14072 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
14073 the macro names the resolved architecture (either @samp{"mips1"} or
14074 @samp{"mips3"}). It names the default architecture when no
14075 @option{-march} option is given.
14077 @item -mtune=@var{arch}
14079 Optimize for @var{arch}. Among other things, this option controls
14080 the way instructions are scheduled, and the perceived cost of arithmetic
14081 operations. The list of @var{arch} values is the same as for
14084 When this option is not used, GCC will optimize for the processor
14085 specified by @option{-march}. By using @option{-march} and
14086 @option{-mtune} together, it is possible to generate code that will
14087 run on a family of processors, but optimize the code for one
14088 particular member of that family.
14090 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
14091 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
14092 @samp{-march} ones described above.
14096 Equivalent to @samp{-march=mips1}.
14100 Equivalent to @samp{-march=mips2}.
14104 Equivalent to @samp{-march=mips3}.
14108 Equivalent to @samp{-march=mips4}.
14112 Equivalent to @samp{-march=mips32}.
14116 Equivalent to @samp{-march=mips32r2}.
14120 Equivalent to @samp{-march=mips64}.
14124 Equivalent to @samp{-march=mips64r2}.
14129 @opindex mno-mips16
14130 Generate (do not generate) MIPS16 code. If GCC is targetting a
14131 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
14133 MIPS16 code generation can also be controlled on a per-function basis
14134 by means of @code{mips16} and @code{nomips16} attributes.
14135 @xref{Function Attributes}, for more information.
14137 @item -mflip-mips16
14138 @opindex mflip-mips16
14139 Generate MIPS16 code on alternating functions. This option is provided
14140 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
14141 not intended for ordinary use in compiling user code.
14143 @item -minterlink-mips16
14144 @itemx -mno-interlink-mips16
14145 @opindex minterlink-mips16
14146 @opindex mno-interlink-mips16
14147 Require (do not require) that non-MIPS16 code be link-compatible with
14150 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
14151 it must either use a call or an indirect jump. @option{-minterlink-mips16}
14152 therefore disables direct jumps unless GCC knows that the target of the
14153 jump is not MIPS16.
14165 Generate code for the given ABI@.
14167 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
14168 generates 64-bit code when you select a 64-bit architecture, but you
14169 can use @option{-mgp32} to get 32-bit code instead.
14171 For information about the O64 ABI, see
14172 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
14174 GCC supports a variant of the o32 ABI in which floating-point registers
14175 are 64 rather than 32 bits wide. You can select this combination with
14176 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
14177 and @samp{mfhc1} instructions and is therefore only supported for
14178 MIPS32R2 processors.
14180 The register assignments for arguments and return values remain the
14181 same, but each scalar value is passed in a single 64-bit register
14182 rather than a pair of 32-bit registers. For example, scalar
14183 floating-point values are returned in @samp{$f0} only, not a
14184 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
14185 remains the same, but all 64 bits are saved.
14188 @itemx -mno-abicalls
14190 @opindex mno-abicalls
14191 Generate (do not generate) code that is suitable for SVR4-style
14192 dynamic objects. @option{-mabicalls} is the default for SVR4-based
14197 Generate (do not generate) code that is fully position-independent,
14198 and that can therefore be linked into shared libraries. This option
14199 only affects @option{-mabicalls}.
14201 All @option{-mabicalls} code has traditionally been position-independent,
14202 regardless of options like @option{-fPIC} and @option{-fpic}. However,
14203 as an extension, the GNU toolchain allows executables to use absolute
14204 accesses for locally-binding symbols. It can also use shorter GP
14205 initialization sequences and generate direct calls to locally-defined
14206 functions. This mode is selected by @option{-mno-shared}.
14208 @option{-mno-shared} depends on binutils 2.16 or higher and generates
14209 objects that can only be linked by the GNU linker. However, the option
14210 does not affect the ABI of the final executable; it only affects the ABI
14211 of relocatable objects. Using @option{-mno-shared} will generally make
14212 executables both smaller and quicker.
14214 @option{-mshared} is the default.
14220 Assume (do not assume) that the static and dynamic linkers
14221 support PLTs and copy relocations. This option only affects
14222 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
14223 has no effect without @samp{-msym32}.
14225 You can make @option{-mplt} the default by configuring
14226 GCC with @option{--with-mips-plt}. The default is
14227 @option{-mno-plt} otherwise.
14233 Lift (do not lift) the usual restrictions on the size of the global
14236 GCC normally uses a single instruction to load values from the GOT@.
14237 While this is relatively efficient, it will only work if the GOT
14238 is smaller than about 64k. Anything larger will cause the linker
14239 to report an error such as:
14241 @cindex relocation truncated to fit (MIPS)
14243 relocation truncated to fit: R_MIPS_GOT16 foobar
14246 If this happens, you should recompile your code with @option{-mxgot}.
14247 It should then work with very large GOTs, although it will also be
14248 less efficient, since it will take three instructions to fetch the
14249 value of a global symbol.
14251 Note that some linkers can create multiple GOTs. If you have such a
14252 linker, you should only need to use @option{-mxgot} when a single object
14253 file accesses more than 64k's worth of GOT entries. Very few do.
14255 These options have no effect unless GCC is generating position
14260 Assume that general-purpose registers are 32 bits wide.
14264 Assume that general-purpose registers are 64 bits wide.
14268 Assume that floating-point registers are 32 bits wide.
14272 Assume that floating-point registers are 64 bits wide.
14275 @opindex mhard-float
14276 Use floating-point coprocessor instructions.
14279 @opindex msoft-float
14280 Do not use floating-point coprocessor instructions. Implement
14281 floating-point calculations using library calls instead.
14283 @item -msingle-float
14284 @opindex msingle-float
14285 Assume that the floating-point coprocessor only supports single-precision
14288 @item -mdouble-float
14289 @opindex mdouble-float
14290 Assume that the floating-point coprocessor supports double-precision
14291 operations. This is the default.
14297 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
14298 implement atomic memory built-in functions. When neither option is
14299 specified, GCC will use the instructions if the target architecture
14302 @option{-mllsc} is useful if the runtime environment can emulate the
14303 instructions and @option{-mno-llsc} can be useful when compiling for
14304 nonstandard ISAs. You can make either option the default by
14305 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
14306 respectively. @option{--with-llsc} is the default for some
14307 configurations; see the installation documentation for details.
14313 Use (do not use) revision 1 of the MIPS DSP ASE@.
14314 @xref{MIPS DSP Built-in Functions}. This option defines the
14315 preprocessor macro @samp{__mips_dsp}. It also defines
14316 @samp{__mips_dsp_rev} to 1.
14322 Use (do not use) revision 2 of the MIPS DSP ASE@.
14323 @xref{MIPS DSP Built-in Functions}. This option defines the
14324 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
14325 It also defines @samp{__mips_dsp_rev} to 2.
14328 @itemx -mno-smartmips
14329 @opindex msmartmips
14330 @opindex mno-smartmips
14331 Use (do not use) the MIPS SmartMIPS ASE.
14333 @item -mpaired-single
14334 @itemx -mno-paired-single
14335 @opindex mpaired-single
14336 @opindex mno-paired-single
14337 Use (do not use) paired-single floating-point instructions.
14338 @xref{MIPS Paired-Single Support}. This option requires
14339 hardware floating-point support to be enabled.
14345 Use (do not use) MIPS Digital Media Extension instructions.
14346 This option can only be used when generating 64-bit code and requires
14347 hardware floating-point support to be enabled.
14352 @opindex mno-mips3d
14353 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
14354 The option @option{-mips3d} implies @option{-mpaired-single}.
14360 Use (do not use) MT Multithreading instructions.
14364 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
14365 an explanation of the default and the way that the pointer size is
14370 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
14372 The default size of @code{int}s, @code{long}s and pointers depends on
14373 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
14374 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
14375 32-bit @code{long}s. Pointers are the same size as @code{long}s,
14376 or the same size as integer registers, whichever is smaller.
14382 Assume (do not assume) that all symbols have 32-bit values, regardless
14383 of the selected ABI@. This option is useful in combination with
14384 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
14385 to generate shorter and faster references to symbolic addresses.
14389 Put definitions of externally-visible data in a small data section
14390 if that data is no bigger than @var{num} bytes. GCC can then access
14391 the data more efficiently; see @option{-mgpopt} for details.
14393 The default @option{-G} option depends on the configuration.
14395 @item -mlocal-sdata
14396 @itemx -mno-local-sdata
14397 @opindex mlocal-sdata
14398 @opindex mno-local-sdata
14399 Extend (do not extend) the @option{-G} behavior to local data too,
14400 such as to static variables in C@. @option{-mlocal-sdata} is the
14401 default for all configurations.
14403 If the linker complains that an application is using too much small data,
14404 you might want to try rebuilding the less performance-critical parts with
14405 @option{-mno-local-sdata}. You might also want to build large
14406 libraries with @option{-mno-local-sdata}, so that the libraries leave
14407 more room for the main program.
14409 @item -mextern-sdata
14410 @itemx -mno-extern-sdata
14411 @opindex mextern-sdata
14412 @opindex mno-extern-sdata
14413 Assume (do not assume) that externally-defined data will be in
14414 a small data section if that data is within the @option{-G} limit.
14415 @option{-mextern-sdata} is the default for all configurations.
14417 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
14418 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
14419 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
14420 is placed in a small data section. If @var{Var} is defined by another
14421 module, you must either compile that module with a high-enough
14422 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
14423 definition. If @var{Var} is common, you must link the application
14424 with a high-enough @option{-G} setting.
14426 The easiest way of satisfying these restrictions is to compile
14427 and link every module with the same @option{-G} option. However,
14428 you may wish to build a library that supports several different
14429 small data limits. You can do this by compiling the library with
14430 the highest supported @option{-G} setting and additionally using
14431 @option{-mno-extern-sdata} to stop the library from making assumptions
14432 about externally-defined data.
14438 Use (do not use) GP-relative accesses for symbols that are known to be
14439 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
14440 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
14443 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
14444 might not hold the value of @code{_gp}. For example, if the code is
14445 part of a library that might be used in a boot monitor, programs that
14446 call boot monitor routines will pass an unknown value in @code{$gp}.
14447 (In such situations, the boot monitor itself would usually be compiled
14448 with @option{-G0}.)
14450 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
14451 @option{-mno-extern-sdata}.
14453 @item -membedded-data
14454 @itemx -mno-embedded-data
14455 @opindex membedded-data
14456 @opindex mno-embedded-data
14457 Allocate variables to the read-only data section first if possible, then
14458 next in the small data section if possible, otherwise in data. This gives
14459 slightly slower code than the default, but reduces the amount of RAM required
14460 when executing, and thus may be preferred for some embedded systems.
14462 @item -muninit-const-in-rodata
14463 @itemx -mno-uninit-const-in-rodata
14464 @opindex muninit-const-in-rodata
14465 @opindex mno-uninit-const-in-rodata
14466 Put uninitialized @code{const} variables in the read-only data section.
14467 This option is only meaningful in conjunction with @option{-membedded-data}.
14469 @item -mcode-readable=@var{setting}
14470 @opindex mcode-readable
14471 Specify whether GCC may generate code that reads from executable sections.
14472 There are three possible settings:
14475 @item -mcode-readable=yes
14476 Instructions may freely access executable sections. This is the
14479 @item -mcode-readable=pcrel
14480 MIPS16 PC-relative load instructions can access executable sections,
14481 but other instructions must not do so. This option is useful on 4KSc
14482 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
14483 It is also useful on processors that can be configured to have a dual
14484 instruction/data SRAM interface and that, like the M4K, automatically
14485 redirect PC-relative loads to the instruction RAM.
14487 @item -mcode-readable=no
14488 Instructions must not access executable sections. This option can be
14489 useful on targets that are configured to have a dual instruction/data
14490 SRAM interface but that (unlike the M4K) do not automatically redirect
14491 PC-relative loads to the instruction RAM.
14494 @item -msplit-addresses
14495 @itemx -mno-split-addresses
14496 @opindex msplit-addresses
14497 @opindex mno-split-addresses
14498 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
14499 relocation operators. This option has been superseded by
14500 @option{-mexplicit-relocs} but is retained for backwards compatibility.
14502 @item -mexplicit-relocs
14503 @itemx -mno-explicit-relocs
14504 @opindex mexplicit-relocs
14505 @opindex mno-explicit-relocs
14506 Use (do not use) assembler relocation operators when dealing with symbolic
14507 addresses. The alternative, selected by @option{-mno-explicit-relocs},
14508 is to use assembler macros instead.
14510 @option{-mexplicit-relocs} is the default if GCC was configured
14511 to use an assembler that supports relocation operators.
14513 @item -mcheck-zero-division
14514 @itemx -mno-check-zero-division
14515 @opindex mcheck-zero-division
14516 @opindex mno-check-zero-division
14517 Trap (do not trap) on integer division by zero.
14519 The default is @option{-mcheck-zero-division}.
14521 @item -mdivide-traps
14522 @itemx -mdivide-breaks
14523 @opindex mdivide-traps
14524 @opindex mdivide-breaks
14525 MIPS systems check for division by zero by generating either a
14526 conditional trap or a break instruction. Using traps results in
14527 smaller code, but is only supported on MIPS II and later. Also, some
14528 versions of the Linux kernel have a bug that prevents trap from
14529 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
14530 allow conditional traps on architectures that support them and
14531 @option{-mdivide-breaks} to force the use of breaks.
14533 The default is usually @option{-mdivide-traps}, but this can be
14534 overridden at configure time using @option{--with-divide=breaks}.
14535 Divide-by-zero checks can be completely disabled using
14536 @option{-mno-check-zero-division}.
14541 @opindex mno-memcpy
14542 Force (do not force) the use of @code{memcpy()} for non-trivial block
14543 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
14544 most constant-sized copies.
14547 @itemx -mno-long-calls
14548 @opindex mlong-calls
14549 @opindex mno-long-calls
14550 Disable (do not disable) use of the @code{jal} instruction. Calling
14551 functions using @code{jal} is more efficient but requires the caller
14552 and callee to be in the same 256 megabyte segment.
14554 This option has no effect on abicalls code. The default is
14555 @option{-mno-long-calls}.
14561 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
14562 instructions, as provided by the R4650 ISA@.
14565 @itemx -mno-fused-madd
14566 @opindex mfused-madd
14567 @opindex mno-fused-madd
14568 Enable (disable) use of the floating point multiply-accumulate
14569 instructions, when they are available. The default is
14570 @option{-mfused-madd}.
14572 When multiply-accumulate instructions are used, the intermediate
14573 product is calculated to infinite precision and is not subject to
14574 the FCSR Flush to Zero bit. This may be undesirable in some
14579 Tell the MIPS assembler to not run its preprocessor over user
14580 assembler files (with a @samp{.s} suffix) when assembling them.
14583 @itemx -mno-fix-r4000
14584 @opindex mfix-r4000
14585 @opindex mno-fix-r4000
14586 Work around certain R4000 CPU errata:
14589 A double-word or a variable shift may give an incorrect result if executed
14590 immediately after starting an integer division.
14592 A double-word or a variable shift may give an incorrect result if executed
14593 while an integer multiplication is in progress.
14595 An integer division may give an incorrect result if started in a delay slot
14596 of a taken branch or a jump.
14600 @itemx -mno-fix-r4400
14601 @opindex mfix-r4400
14602 @opindex mno-fix-r4400
14603 Work around certain R4400 CPU errata:
14606 A double-word or a variable shift may give an incorrect result if executed
14607 immediately after starting an integer division.
14611 @itemx -mno-fix-r10000
14612 @opindex mfix-r10000
14613 @opindex mno-fix-r10000
14614 Work around certain R10000 errata:
14617 @code{ll}/@code{sc} sequences may not behave atomically on revisions
14618 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
14621 This option can only be used if the target architecture supports
14622 branch-likely instructions. @option{-mfix-r10000} is the default when
14623 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
14627 @itemx -mno-fix-vr4120
14628 @opindex mfix-vr4120
14629 Work around certain VR4120 errata:
14632 @code{dmultu} does not always produce the correct result.
14634 @code{div} and @code{ddiv} do not always produce the correct result if one
14635 of the operands is negative.
14637 The workarounds for the division errata rely on special functions in
14638 @file{libgcc.a}. At present, these functions are only provided by
14639 the @code{mips64vr*-elf} configurations.
14641 Other VR4120 errata require a nop to be inserted between certain pairs of
14642 instructions. These errata are handled by the assembler, not by GCC itself.
14645 @opindex mfix-vr4130
14646 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
14647 workarounds are implemented by the assembler rather than by GCC,
14648 although GCC will avoid using @code{mflo} and @code{mfhi} if the
14649 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
14650 instructions are available instead.
14653 @itemx -mno-fix-sb1
14655 Work around certain SB-1 CPU core errata.
14656 (This flag currently works around the SB-1 revision 2
14657 ``F1'' and ``F2'' floating point errata.)
14659 @item -mr10k-cache-barrier=@var{setting}
14660 @opindex mr10k-cache-barrier
14661 Specify whether GCC should insert cache barriers to avoid the
14662 side-effects of speculation on R10K processors.
14664 In common with many processors, the R10K tries to predict the outcome
14665 of a conditional branch and speculatively executes instructions from
14666 the ``taken'' branch. It later aborts these instructions if the
14667 predicted outcome was wrong. However, on the R10K, even aborted
14668 instructions can have side effects.
14670 This problem only affects kernel stores and, depending on the system,
14671 kernel loads. As an example, a speculatively-executed store may load
14672 the target memory into cache and mark the cache line as dirty, even if
14673 the store itself is later aborted. If a DMA operation writes to the
14674 same area of memory before the ``dirty'' line is flushed, the cached
14675 data will overwrite the DMA-ed data. See the R10K processor manual
14676 for a full description, including other potential problems.
14678 One workaround is to insert cache barrier instructions before every memory
14679 access that might be speculatively executed and that might have side
14680 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
14681 controls GCC's implementation of this workaround. It assumes that
14682 aborted accesses to any byte in the following regions will not have
14687 the memory occupied by the current function's stack frame;
14690 the memory occupied by an incoming stack argument;
14693 the memory occupied by an object with a link-time-constant address.
14696 It is the kernel's responsibility to ensure that speculative
14697 accesses to these regions are indeed safe.
14699 If the input program contains a function declaration such as:
14705 then the implementation of @code{foo} must allow @code{j foo} and
14706 @code{jal foo} to be executed speculatively. GCC honors this
14707 restriction for functions it compiles itself. It expects non-GCC
14708 functions (such as hand-written assembly code) to do the same.
14710 The option has three forms:
14713 @item -mr10k-cache-barrier=load-store
14714 Insert a cache barrier before a load or store that might be
14715 speculatively executed and that might have side effects even
14718 @item -mr10k-cache-barrier=store
14719 Insert a cache barrier before a store that might be speculatively
14720 executed and that might have side effects even if aborted.
14722 @item -mr10k-cache-barrier=none
14723 Disable the insertion of cache barriers. This is the default setting.
14726 @item -mflush-func=@var{func}
14727 @itemx -mno-flush-func
14728 @opindex mflush-func
14729 Specifies the function to call to flush the I and D caches, or to not
14730 call any such function. If called, the function must take the same
14731 arguments as the common @code{_flush_func()}, that is, the address of the
14732 memory range for which the cache is being flushed, the size of the
14733 memory range, and the number 3 (to flush both caches). The default
14734 depends on the target GCC was configured for, but commonly is either
14735 @samp{_flush_func} or @samp{__cpu_flush}.
14737 @item mbranch-cost=@var{num}
14738 @opindex mbranch-cost
14739 Set the cost of branches to roughly @var{num} ``simple'' instructions.
14740 This cost is only a heuristic and is not guaranteed to produce
14741 consistent results across releases. A zero cost redundantly selects
14742 the default, which is based on the @option{-mtune} setting.
14744 @item -mbranch-likely
14745 @itemx -mno-branch-likely
14746 @opindex mbranch-likely
14747 @opindex mno-branch-likely
14748 Enable or disable use of Branch Likely instructions, regardless of the
14749 default for the selected architecture. By default, Branch Likely
14750 instructions may be generated if they are supported by the selected
14751 architecture. An exception is for the MIPS32 and MIPS64 architectures
14752 and processors which implement those architectures; for those, Branch
14753 Likely instructions will not be generated by default because the MIPS32
14754 and MIPS64 architectures specifically deprecate their use.
14756 @item -mfp-exceptions
14757 @itemx -mno-fp-exceptions
14758 @opindex mfp-exceptions
14759 Specifies whether FP exceptions are enabled. This affects how we schedule
14760 FP instructions for some processors. The default is that FP exceptions are
14763 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
14764 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
14767 @item -mvr4130-align
14768 @itemx -mno-vr4130-align
14769 @opindex mvr4130-align
14770 The VR4130 pipeline is two-way superscalar, but can only issue two
14771 instructions together if the first one is 8-byte aligned. When this
14772 option is enabled, GCC will align pairs of instructions that it
14773 thinks should execute in parallel.
14775 This option only has an effect when optimizing for the VR4130.
14776 It normally makes code faster, but at the expense of making it bigger.
14777 It is enabled by default at optimization level @option{-O3}.
14782 Enable (disable) generation of @code{synci} instructions on
14783 architectures that support it. The @code{synci} instructions (if
14784 enabled) will be generated when @code{__builtin___clear_cache()} is
14787 This option defaults to @code{-mno-synci}, but the default can be
14788 overridden by configuring with @code{--with-synci}.
14790 When compiling code for single processor systems, it is generally safe
14791 to use @code{synci}. However, on many multi-core (SMP) systems, it
14792 will not invalidate the instruction caches on all cores and may lead
14793 to undefined behavior.
14795 @item -mrelax-pic-calls
14796 @itemx -mno-relax-pic-calls
14797 @opindex mrelax-pic-calls
14798 Try to turn PIC calls that are normally dispatched via register
14799 @code{$25} into direct calls. This is only possible if the linker can
14800 resolve the destination at link-time and if the destination is within
14801 range for a direct call.
14803 @option{-mrelax-pic-calls} is the default if GCC was configured to use
14804 an assembler and a linker that supports the @code{.reloc} assembly
14805 directive and @code{-mexplicit-relocs} is in effect. With
14806 @code{-mno-explicit-relocs}, this optimization can be performed by the
14807 assembler and the linker alone without help from the compiler.
14809 @item -mmcount-ra-address
14810 @itemx -mno-mcount-ra-address
14811 @opindex mmcount-ra-address
14812 @opindex mno-mcount-ra-address
14813 Emit (do not emit) code that allows @code{_mcount} to modify the
14814 calling function's return address. When enabled, this option extends
14815 the usual @code{_mcount} interface with a new @var{ra-address}
14816 parameter, which has type @code{intptr_t *} and is passed in register
14817 @code{$12}. @code{_mcount} can then modify the return address by
14818 doing both of the following:
14821 Returning the new address in register @code{$31}.
14823 Storing the new address in @code{*@var{ra-address}},
14824 if @var{ra-address} is nonnull.
14827 The default is @option{-mno-mcount-ra-address}.
14832 @subsection MMIX Options
14833 @cindex MMIX Options
14835 These options are defined for the MMIX:
14839 @itemx -mno-libfuncs
14841 @opindex mno-libfuncs
14842 Specify that intrinsic library functions are being compiled, passing all
14843 values in registers, no matter the size.
14846 @itemx -mno-epsilon
14848 @opindex mno-epsilon
14849 Generate floating-point comparison instructions that compare with respect
14850 to the @code{rE} epsilon register.
14852 @item -mabi=mmixware
14854 @opindex mabi=mmixware
14856 Generate code that passes function parameters and return values that (in
14857 the called function) are seen as registers @code{$0} and up, as opposed to
14858 the GNU ABI which uses global registers @code{$231} and up.
14860 @item -mzero-extend
14861 @itemx -mno-zero-extend
14862 @opindex mzero-extend
14863 @opindex mno-zero-extend
14864 When reading data from memory in sizes shorter than 64 bits, use (do not
14865 use) zero-extending load instructions by default, rather than
14866 sign-extending ones.
14869 @itemx -mno-knuthdiv
14871 @opindex mno-knuthdiv
14872 Make the result of a division yielding a remainder have the same sign as
14873 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
14874 remainder follows the sign of the dividend. Both methods are
14875 arithmetically valid, the latter being almost exclusively used.
14877 @item -mtoplevel-symbols
14878 @itemx -mno-toplevel-symbols
14879 @opindex mtoplevel-symbols
14880 @opindex mno-toplevel-symbols
14881 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
14882 code can be used with the @code{PREFIX} assembly directive.
14886 Generate an executable in the ELF format, rather than the default
14887 @samp{mmo} format used by the @command{mmix} simulator.
14889 @item -mbranch-predict
14890 @itemx -mno-branch-predict
14891 @opindex mbranch-predict
14892 @opindex mno-branch-predict
14893 Use (do not use) the probable-branch instructions, when static branch
14894 prediction indicates a probable branch.
14896 @item -mbase-addresses
14897 @itemx -mno-base-addresses
14898 @opindex mbase-addresses
14899 @opindex mno-base-addresses
14900 Generate (do not generate) code that uses @emph{base addresses}. Using a
14901 base address automatically generates a request (handled by the assembler
14902 and the linker) for a constant to be set up in a global register. The
14903 register is used for one or more base address requests within the range 0
14904 to 255 from the value held in the register. The generally leads to short
14905 and fast code, but the number of different data items that can be
14906 addressed is limited. This means that a program that uses lots of static
14907 data may require @option{-mno-base-addresses}.
14909 @item -msingle-exit
14910 @itemx -mno-single-exit
14911 @opindex msingle-exit
14912 @opindex mno-single-exit
14913 Force (do not force) generated code to have a single exit point in each
14917 @node MN10300 Options
14918 @subsection MN10300 Options
14919 @cindex MN10300 options
14921 These @option{-m} options are defined for Matsushita MN10300 architectures:
14926 Generate code to avoid bugs in the multiply instructions for the MN10300
14927 processors. This is the default.
14929 @item -mno-mult-bug
14930 @opindex mno-mult-bug
14931 Do not generate code to avoid bugs in the multiply instructions for the
14932 MN10300 processors.
14936 Generate code which uses features specific to the AM33 processor.
14940 Do not generate code which uses features specific to the AM33 processor. This
14945 Generate code which uses features specific to the AM33/2.0 processor.
14949 Generate code which uses features specific to the AM34 processor.
14951 @item -mtune=@var{cpu-type}
14953 Use the timing characteristics of the indicated CPU type when
14954 scheduling instructions. This does not change the targeted processor
14955 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
14956 @samp{am33-2} or @samp{am34}.
14958 @item -mreturn-pointer-on-d0
14959 @opindex mreturn-pointer-on-d0
14960 When generating a function which returns a pointer, return the pointer
14961 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
14962 only in a0, and attempts to call such functions without a prototype
14963 would result in errors. Note that this option is on by default; use
14964 @option{-mno-return-pointer-on-d0} to disable it.
14968 Do not link in the C run-time initialization object file.
14972 Indicate to the linker that it should perform a relaxation optimization pass
14973 to shorten branches, calls and absolute memory addresses. This option only
14974 has an effect when used on the command line for the final link step.
14976 This option makes symbolic debugging impossible.
14979 @node PDP-11 Options
14980 @subsection PDP-11 Options
14981 @cindex PDP-11 Options
14983 These options are defined for the PDP-11:
14988 Use hardware FPP floating point. This is the default. (FIS floating
14989 point on the PDP-11/40 is not supported.)
14992 @opindex msoft-float
14993 Do not use hardware floating point.
14997 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
15001 Return floating-point results in memory. This is the default.
15005 Generate code for a PDP-11/40.
15009 Generate code for a PDP-11/45. This is the default.
15013 Generate code for a PDP-11/10.
15015 @item -mbcopy-builtin
15016 @opindex mbcopy-builtin
15017 Use inline @code{movmemhi} patterns for copying memory. This is the
15022 Do not use inline @code{movmemhi} patterns for copying memory.
15028 Use 16-bit @code{int}. This is the default.
15034 Use 32-bit @code{int}.
15037 @itemx -mno-float32
15039 @opindex mno-float32
15040 Use 64-bit @code{float}. This is the default.
15043 @itemx -mno-float64
15045 @opindex mno-float64
15046 Use 32-bit @code{float}.
15050 Use @code{abshi2} pattern. This is the default.
15054 Do not use @code{abshi2} pattern.
15056 @item -mbranch-expensive
15057 @opindex mbranch-expensive
15058 Pretend that branches are expensive. This is for experimenting with
15059 code generation only.
15061 @item -mbranch-cheap
15062 @opindex mbranch-cheap
15063 Do not pretend that branches are expensive. This is the default.
15067 Use Unix assembler syntax. This is the default when configured for
15068 @samp{pdp11-*-bsd}.
15072 Use DEC assembler syntax. This is the default when configured for any
15073 PDP-11 target other than @samp{pdp11-*-bsd}.
15076 @node picoChip Options
15077 @subsection picoChip Options
15078 @cindex picoChip options
15080 These @samp{-m} options are defined for picoChip implementations:
15084 @item -mae=@var{ae_type}
15086 Set the instruction set, register set, and instruction scheduling
15087 parameters for array element type @var{ae_type}. Supported values
15088 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
15090 @option{-mae=ANY} selects a completely generic AE type. Code
15091 generated with this option will run on any of the other AE types. The
15092 code will not be as efficient as it would be if compiled for a specific
15093 AE type, and some types of operation (e.g., multiplication) will not
15094 work properly on all types of AE.
15096 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
15097 for compiled code, and is the default.
15099 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
15100 option may suffer from poor performance of byte (char) manipulation,
15101 since the DSP AE does not provide hardware support for byte load/stores.
15103 @item -msymbol-as-address
15104 Enable the compiler to directly use a symbol name as an address in a
15105 load/store instruction, without first loading it into a
15106 register. Typically, the use of this option will generate larger
15107 programs, which run faster than when the option isn't used. However, the
15108 results vary from program to program, so it is left as a user option,
15109 rather than being permanently enabled.
15111 @item -mno-inefficient-warnings
15112 Disables warnings about the generation of inefficient code. These
15113 warnings can be generated, for example, when compiling code which
15114 performs byte-level memory operations on the MAC AE type. The MAC AE has
15115 no hardware support for byte-level memory operations, so all byte
15116 load/stores must be synthesized from word load/store operations. This is
15117 inefficient and a warning will be generated indicating to the programmer
15118 that they should rewrite the code to avoid byte operations, or to target
15119 an AE type which has the necessary hardware support. This option enables
15120 the warning to be turned off.
15124 @node PowerPC Options
15125 @subsection PowerPC Options
15126 @cindex PowerPC options
15128 These are listed under @xref{RS/6000 and PowerPC Options}.
15130 @node RS/6000 and PowerPC Options
15131 @subsection IBM RS/6000 and PowerPC Options
15132 @cindex RS/6000 and PowerPC Options
15133 @cindex IBM RS/6000 and PowerPC Options
15135 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
15142 @itemx -mno-powerpc
15143 @itemx -mpowerpc-gpopt
15144 @itemx -mno-powerpc-gpopt
15145 @itemx -mpowerpc-gfxopt
15146 @itemx -mno-powerpc-gfxopt
15149 @itemx -mno-powerpc64
15153 @itemx -mno-popcntb
15155 @itemx -mno-popcntd
15164 @itemx -mno-hard-dfp
15168 @opindex mno-power2
15170 @opindex mno-powerpc
15171 @opindex mpowerpc-gpopt
15172 @opindex mno-powerpc-gpopt
15173 @opindex mpowerpc-gfxopt
15174 @opindex mno-powerpc-gfxopt
15175 @opindex mpowerpc64
15176 @opindex mno-powerpc64
15180 @opindex mno-popcntb
15182 @opindex mno-popcntd
15188 @opindex mno-mfpgpr
15190 @opindex mno-hard-dfp
15191 GCC supports two related instruction set architectures for the
15192 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
15193 instructions supported by the @samp{rios} chip set used in the original
15194 RS/6000 systems and the @dfn{PowerPC} instruction set is the
15195 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
15196 the IBM 4xx, 6xx, and follow-on microprocessors.
15198 Neither architecture is a subset of the other. However there is a
15199 large common subset of instructions supported by both. An MQ
15200 register is included in processors supporting the POWER architecture.
15202 You use these options to specify which instructions are available on the
15203 processor you are using. The default value of these options is
15204 determined when configuring GCC@. Specifying the
15205 @option{-mcpu=@var{cpu_type}} overrides the specification of these
15206 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
15207 rather than the options listed above.
15209 The @option{-mpower} option allows GCC to generate instructions that
15210 are found only in the POWER architecture and to use the MQ register.
15211 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
15212 to generate instructions that are present in the POWER2 architecture but
15213 not the original POWER architecture.
15215 The @option{-mpowerpc} option allows GCC to generate instructions that
15216 are found only in the 32-bit subset of the PowerPC architecture.
15217 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
15218 GCC to use the optional PowerPC architecture instructions in the
15219 General Purpose group, including floating-point square root. Specifying
15220 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
15221 use the optional PowerPC architecture instructions in the Graphics
15222 group, including floating-point select.
15224 The @option{-mmfcrf} option allows GCC to generate the move from
15225 condition register field instruction implemented on the POWER4
15226 processor and other processors that support the PowerPC V2.01
15228 The @option{-mpopcntb} option allows GCC to generate the popcount and
15229 double precision FP reciprocal estimate instruction implemented on the
15230 POWER5 processor and other processors that support the PowerPC V2.02
15232 The @option{-mpopcntd} option allows GCC to generate the popcount
15233 instruction implemented on the POWER7 processor and other processors
15234 that support the PowerPC V2.06 architecture.
15235 The @option{-mfprnd} option allows GCC to generate the FP round to
15236 integer instructions implemented on the POWER5+ processor and other
15237 processors that support the PowerPC V2.03 architecture.
15238 The @option{-mcmpb} option allows GCC to generate the compare bytes
15239 instruction implemented on the POWER6 processor and other processors
15240 that support the PowerPC V2.05 architecture.
15241 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
15242 general purpose register instructions implemented on the POWER6X
15243 processor and other processors that support the extended PowerPC V2.05
15245 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
15246 point instructions implemented on some POWER processors.
15248 The @option{-mpowerpc64} option allows GCC to generate the additional
15249 64-bit instructions that are found in the full PowerPC64 architecture
15250 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
15251 @option{-mno-powerpc64}.
15253 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
15254 will use only the instructions in the common subset of both
15255 architectures plus some special AIX common-mode calls, and will not use
15256 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
15257 permits GCC to use any instruction from either architecture and to
15258 allow use of the MQ register; specify this for the Motorola MPC601.
15260 @item -mnew-mnemonics
15261 @itemx -mold-mnemonics
15262 @opindex mnew-mnemonics
15263 @opindex mold-mnemonics
15264 Select which mnemonics to use in the generated assembler code. With
15265 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
15266 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
15267 assembler mnemonics defined for the POWER architecture. Instructions
15268 defined in only one architecture have only one mnemonic; GCC uses that
15269 mnemonic irrespective of which of these options is specified.
15271 GCC defaults to the mnemonics appropriate for the architecture in
15272 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
15273 value of these option. Unless you are building a cross-compiler, you
15274 should normally not specify either @option{-mnew-mnemonics} or
15275 @option{-mold-mnemonics}, but should instead accept the default.
15277 @item -mcpu=@var{cpu_type}
15279 Set architecture type, register usage, choice of mnemonics, and
15280 instruction scheduling parameters for machine type @var{cpu_type}.
15281 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
15282 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
15283 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
15284 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
15285 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
15286 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
15287 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{ec603e}, @samp{G3},
15288 @samp{G4}, @samp{G5}, @samp{titan}, @samp{power}, @samp{power2}, @samp{power3},
15289 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x},
15290 @samp{power7}, @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
15291 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
15293 @option{-mcpu=common} selects a completely generic processor. Code
15294 generated under this option will run on any POWER or PowerPC processor.
15295 GCC will use only the instructions in the common subset of both
15296 architectures, and will not use the MQ register. GCC assumes a generic
15297 processor model for scheduling purposes.
15299 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
15300 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
15301 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
15302 types, with an appropriate, generic processor model assumed for
15303 scheduling purposes.
15305 The other options specify a specific processor. Code generated under
15306 those options will run best on that processor, and may not run at all on
15309 The @option{-mcpu} options automatically enable or disable the
15312 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
15313 -mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol
15314 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
15315 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx}
15317 The particular options set for any particular CPU will vary between
15318 compiler versions, depending on what setting seems to produce optimal
15319 code for that CPU; it doesn't necessarily reflect the actual hardware's
15320 capabilities. If you wish to set an individual option to a particular
15321 value, you may specify it after the @option{-mcpu} option, like
15322 @samp{-mcpu=970 -mno-altivec}.
15324 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
15325 not enabled or disabled by the @option{-mcpu} option at present because
15326 AIX does not have full support for these options. You may still
15327 enable or disable them individually if you're sure it'll work in your
15330 @item -mtune=@var{cpu_type}
15332 Set the instruction scheduling parameters for machine type
15333 @var{cpu_type}, but do not set the architecture type, register usage, or
15334 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
15335 values for @var{cpu_type} are used for @option{-mtune} as for
15336 @option{-mcpu}. If both are specified, the code generated will use the
15337 architecture, registers, and mnemonics set by @option{-mcpu}, but the
15338 scheduling parameters set by @option{-mtune}.
15340 @item -mcmodel=small
15341 @opindex mcmodel=small
15342 Generate PowerPC64 code for the small model: The TOC is limited to
15345 @item -mcmodel=medium
15346 @opindex mcmodel=medium
15347 Generate PowerPC64 code for the medium model: The TOC and other static
15348 data may be up to a total of 4G in size.
15350 @item -mcmodel=large
15351 @opindex mcmodel=large
15352 Generate PowerPC64 code for the large model: The TOC may be up to 4G
15353 in size. Other data and code is only limited by the 64-bit address
15357 @itemx -mno-altivec
15359 @opindex mno-altivec
15360 Generate code that uses (does not use) AltiVec instructions, and also
15361 enable the use of built-in functions that allow more direct access to
15362 the AltiVec instruction set. You may also need to set
15363 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
15369 @opindex mno-vrsave
15370 Generate VRSAVE instructions when generating AltiVec code.
15372 @item -mgen-cell-microcode
15373 @opindex mgen-cell-microcode
15374 Generate Cell microcode instructions
15376 @item -mwarn-cell-microcode
15377 @opindex mwarn-cell-microcode
15378 Warning when a Cell microcode instruction is going to emitted. An example
15379 of a Cell microcode instruction is a variable shift.
15382 @opindex msecure-plt
15383 Generate code that allows ld and ld.so to build executables and shared
15384 libraries with non-exec .plt and .got sections. This is a PowerPC
15385 32-bit SYSV ABI option.
15389 Generate code that uses a BSS .plt section that ld.so fills in, and
15390 requires .plt and .got sections that are both writable and executable.
15391 This is a PowerPC 32-bit SYSV ABI option.
15397 This switch enables or disables the generation of ISEL instructions.
15399 @item -misel=@var{yes/no}
15400 This switch has been deprecated. Use @option{-misel} and
15401 @option{-mno-isel} instead.
15407 This switch enables or disables the generation of SPE simd
15413 @opindex mno-paired
15414 This switch enables or disables the generation of PAIRED simd
15417 @item -mspe=@var{yes/no}
15418 This option has been deprecated. Use @option{-mspe} and
15419 @option{-mno-spe} instead.
15425 Generate code that uses (does not use) vector/scalar (VSX)
15426 instructions, and also enable the use of built-in functions that allow
15427 more direct access to the VSX instruction set.
15429 @item -mfloat-gprs=@var{yes/single/double/no}
15430 @itemx -mfloat-gprs
15431 @opindex mfloat-gprs
15432 This switch enables or disables the generation of floating point
15433 operations on the general purpose registers for architectures that
15436 The argument @var{yes} or @var{single} enables the use of
15437 single-precision floating point operations.
15439 The argument @var{double} enables the use of single and
15440 double-precision floating point operations.
15442 The argument @var{no} disables floating point operations on the
15443 general purpose registers.
15445 This option is currently only available on the MPC854x.
15451 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
15452 targets (including GNU/Linux). The 32-bit environment sets int, long
15453 and pointer to 32 bits and generates code that runs on any PowerPC
15454 variant. The 64-bit environment sets int to 32 bits and long and
15455 pointer to 64 bits, and generates code for PowerPC64, as for
15456 @option{-mpowerpc64}.
15459 @itemx -mno-fp-in-toc
15460 @itemx -mno-sum-in-toc
15461 @itemx -mminimal-toc
15463 @opindex mno-fp-in-toc
15464 @opindex mno-sum-in-toc
15465 @opindex mminimal-toc
15466 Modify generation of the TOC (Table Of Contents), which is created for
15467 every executable file. The @option{-mfull-toc} option is selected by
15468 default. In that case, GCC will allocate at least one TOC entry for
15469 each unique non-automatic variable reference in your program. GCC
15470 will also place floating-point constants in the TOC@. However, only
15471 16,384 entries are available in the TOC@.
15473 If you receive a linker error message that saying you have overflowed
15474 the available TOC space, you can reduce the amount of TOC space used
15475 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
15476 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
15477 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
15478 generate code to calculate the sum of an address and a constant at
15479 run-time instead of putting that sum into the TOC@. You may specify one
15480 or both of these options. Each causes GCC to produce very slightly
15481 slower and larger code at the expense of conserving TOC space.
15483 If you still run out of space in the TOC even when you specify both of
15484 these options, specify @option{-mminimal-toc} instead. This option causes
15485 GCC to make only one TOC entry for every file. When you specify this
15486 option, GCC will produce code that is slower and larger but which
15487 uses extremely little TOC space. You may wish to use this option
15488 only on files that contain less frequently executed code.
15494 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
15495 @code{long} type, and the infrastructure needed to support them.
15496 Specifying @option{-maix64} implies @option{-mpowerpc64} and
15497 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
15498 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
15501 @itemx -mno-xl-compat
15502 @opindex mxl-compat
15503 @opindex mno-xl-compat
15504 Produce code that conforms more closely to IBM XL compiler semantics
15505 when using AIX-compatible ABI@. Pass floating-point arguments to
15506 prototyped functions beyond the register save area (RSA) on the stack
15507 in addition to argument FPRs. Do not assume that most significant
15508 double in 128-bit long double value is properly rounded when comparing
15509 values and converting to double. Use XL symbol names for long double
15512 The AIX calling convention was extended but not initially documented to
15513 handle an obscure K&R C case of calling a function that takes the
15514 address of its arguments with fewer arguments than declared. IBM XL
15515 compilers access floating point arguments which do not fit in the
15516 RSA from the stack when a subroutine is compiled without
15517 optimization. Because always storing floating-point arguments on the
15518 stack is inefficient and rarely needed, this option is not enabled by
15519 default and only is necessary when calling subroutines compiled by IBM
15520 XL compilers without optimization.
15524 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
15525 application written to use message passing with special startup code to
15526 enable the application to run. The system must have PE installed in the
15527 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
15528 must be overridden with the @option{-specs=} option to specify the
15529 appropriate directory location. The Parallel Environment does not
15530 support threads, so the @option{-mpe} option and the @option{-pthread}
15531 option are incompatible.
15533 @item -malign-natural
15534 @itemx -malign-power
15535 @opindex malign-natural
15536 @opindex malign-power
15537 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
15538 @option{-malign-natural} overrides the ABI-defined alignment of larger
15539 types, such as floating-point doubles, on their natural size-based boundary.
15540 The option @option{-malign-power} instructs GCC to follow the ABI-specified
15541 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
15543 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
15547 @itemx -mhard-float
15548 @opindex msoft-float
15549 @opindex mhard-float
15550 Generate code that does not use (uses) the floating-point register set.
15551 Software floating point emulation is provided if you use the
15552 @option{-msoft-float} option, and pass the option to GCC when linking.
15554 @item -msingle-float
15555 @itemx -mdouble-float
15556 @opindex msingle-float
15557 @opindex mdouble-float
15558 Generate code for single or double-precision floating point operations.
15559 @option{-mdouble-float} implies @option{-msingle-float}.
15562 @opindex msimple-fpu
15563 Do not generate sqrt and div instructions for hardware floating point unit.
15567 Specify type of floating point unit. Valid values are @var{sp_lite}
15568 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
15569 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
15570 and @var{dp_full} (equivalent to -mdouble-float).
15573 @opindex mxilinx-fpu
15574 Perform optimizations for floating point unit on Xilinx PPC 405/440.
15577 @itemx -mno-multiple
15579 @opindex mno-multiple
15580 Generate code that uses (does not use) the load multiple word
15581 instructions and the store multiple word instructions. These
15582 instructions are generated by default on POWER systems, and not
15583 generated on PowerPC systems. Do not use @option{-mmultiple} on little
15584 endian PowerPC systems, since those instructions do not work when the
15585 processor is in little endian mode. The exceptions are PPC740 and
15586 PPC750 which permit the instructions usage in little endian mode.
15591 @opindex mno-string
15592 Generate code that uses (does not use) the load string instructions
15593 and the store string word instructions to save multiple registers and
15594 do small block moves. These instructions are generated by default on
15595 POWER systems, and not generated on PowerPC systems. Do not use
15596 @option{-mstring} on little endian PowerPC systems, since those
15597 instructions do not work when the processor is in little endian mode.
15598 The exceptions are PPC740 and PPC750 which permit the instructions
15599 usage in little endian mode.
15604 @opindex mno-update
15605 Generate code that uses (does not use) the load or store instructions
15606 that update the base register to the address of the calculated memory
15607 location. These instructions are generated by default. If you use
15608 @option{-mno-update}, there is a small window between the time that the
15609 stack pointer is updated and the address of the previous frame is
15610 stored, which means code that walks the stack frame across interrupts or
15611 signals may get corrupted data.
15613 @item -mavoid-indexed-addresses
15614 @itemx -mno-avoid-indexed-addresses
15615 @opindex mavoid-indexed-addresses
15616 @opindex mno-avoid-indexed-addresses
15617 Generate code that tries to avoid (not avoid) the use of indexed load
15618 or store instructions. These instructions can incur a performance
15619 penalty on Power6 processors in certain situations, such as when
15620 stepping through large arrays that cross a 16M boundary. This option
15621 is enabled by default when targetting Power6 and disabled otherwise.
15624 @itemx -mno-fused-madd
15625 @opindex mfused-madd
15626 @opindex mno-fused-madd
15627 Generate code that uses (does not use) the floating point multiply and
15628 accumulate instructions. These instructions are generated by default if
15629 hardware floating is used.
15635 Generate code that uses (does not use) the half-word multiply and
15636 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
15637 These instructions are generated by default when targetting those
15644 Generate code that uses (does not use) the string-search @samp{dlmzb}
15645 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
15646 generated by default when targetting those processors.
15648 @item -mno-bit-align
15650 @opindex mno-bit-align
15651 @opindex mbit-align
15652 On System V.4 and embedded PowerPC systems do not (do) force structures
15653 and unions that contain bit-fields to be aligned to the base type of the
15656 For example, by default a structure containing nothing but 8
15657 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
15658 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
15659 the structure would be aligned to a 1 byte boundary and be one byte in
15662 @item -mno-strict-align
15663 @itemx -mstrict-align
15664 @opindex mno-strict-align
15665 @opindex mstrict-align
15666 On System V.4 and embedded PowerPC systems do not (do) assume that
15667 unaligned memory references will be handled by the system.
15669 @item -mrelocatable
15670 @itemx -mno-relocatable
15671 @opindex mrelocatable
15672 @opindex mno-relocatable
15673 Generate code that allows (does not allow) a static executable to be
15674 relocated to a different address at runtime. A simple embedded
15675 PowerPC system loader should relocate the entire contents of
15676 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
15677 a table of 32-bit addresses generated by this option. For this to
15678 work, all objects linked together must be compiled with
15679 @option{-mrelocatable} or @option{-mrelocatable-lib}.
15680 @option{-mrelocatable} code aligns the stack to an 8 byte boundary.
15682 @item -mrelocatable-lib
15683 @itemx -mno-relocatable-lib
15684 @opindex mrelocatable-lib
15685 @opindex mno-relocatable-lib
15686 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
15687 @code{.fixup} section to allow static executables to be relocated at
15688 runtime, but @option{-mrelocatable-lib} does not use the smaller stack
15689 alignment of @option{-mrelocatable}. Objects compiled with
15690 @option{-mrelocatable-lib} may be linked with objects compiled with
15691 any combination of the @option{-mrelocatable} options.
15697 On System V.4 and embedded PowerPC systems do not (do) assume that
15698 register 2 contains a pointer to a global area pointing to the addresses
15699 used in the program.
15702 @itemx -mlittle-endian
15704 @opindex mlittle-endian
15705 On System V.4 and embedded PowerPC systems compile code for the
15706 processor in little endian mode. The @option{-mlittle-endian} option is
15707 the same as @option{-mlittle}.
15710 @itemx -mbig-endian
15712 @opindex mbig-endian
15713 On System V.4 and embedded PowerPC systems compile code for the
15714 processor in big endian mode. The @option{-mbig-endian} option is
15715 the same as @option{-mbig}.
15717 @item -mdynamic-no-pic
15718 @opindex mdynamic-no-pic
15719 On Darwin and Mac OS X systems, compile code so that it is not
15720 relocatable, but that its external references are relocatable. The
15721 resulting code is suitable for applications, but not shared
15724 @item -mprioritize-restricted-insns=@var{priority}
15725 @opindex mprioritize-restricted-insns
15726 This option controls the priority that is assigned to
15727 dispatch-slot restricted instructions during the second scheduling
15728 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
15729 @var{no/highest/second-highest} priority to dispatch slot restricted
15732 @item -msched-costly-dep=@var{dependence_type}
15733 @opindex msched-costly-dep
15734 This option controls which dependences are considered costly
15735 by the target during instruction scheduling. The argument
15736 @var{dependence_type} takes one of the following values:
15737 @var{no}: no dependence is costly,
15738 @var{all}: all dependences are costly,
15739 @var{true_store_to_load}: a true dependence from store to load is costly,
15740 @var{store_to_load}: any dependence from store to load is costly,
15741 @var{number}: any dependence which latency >= @var{number} is costly.
15743 @item -minsert-sched-nops=@var{scheme}
15744 @opindex minsert-sched-nops
15745 This option controls which nop insertion scheme will be used during
15746 the second scheduling pass. The argument @var{scheme} takes one of the
15748 @var{no}: Don't insert nops.
15749 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
15750 according to the scheduler's grouping.
15751 @var{regroup_exact}: Insert nops to force costly dependent insns into
15752 separate groups. Insert exactly as many nops as needed to force an insn
15753 to a new group, according to the estimated processor grouping.
15754 @var{number}: Insert nops to force costly dependent insns into
15755 separate groups. Insert @var{number} nops to force an insn to a new group.
15758 @opindex mcall-sysv
15759 On System V.4 and embedded PowerPC systems compile code using calling
15760 conventions that adheres to the March 1995 draft of the System V
15761 Application Binary Interface, PowerPC processor supplement. This is the
15762 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
15764 @item -mcall-sysv-eabi
15766 @opindex mcall-sysv-eabi
15767 @opindex mcall-eabi
15768 Specify both @option{-mcall-sysv} and @option{-meabi} options.
15770 @item -mcall-sysv-noeabi
15771 @opindex mcall-sysv-noeabi
15772 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
15774 @item -mcall-aixdesc
15776 On System V.4 and embedded PowerPC systems compile code for the AIX
15780 @opindex mcall-linux
15781 On System V.4 and embedded PowerPC systems compile code for the
15782 Linux-based GNU system.
15786 On System V.4 and embedded PowerPC systems compile code for the
15787 Hurd-based GNU system.
15789 @item -mcall-freebsd
15790 @opindex mcall-freebsd
15791 On System V.4 and embedded PowerPC systems compile code for the
15792 FreeBSD operating system.
15794 @item -mcall-netbsd
15795 @opindex mcall-netbsd
15796 On System V.4 and embedded PowerPC systems compile code for the
15797 NetBSD operating system.
15799 @item -mcall-openbsd
15800 @opindex mcall-netbsd
15801 On System V.4 and embedded PowerPC systems compile code for the
15802 OpenBSD operating system.
15804 @item -maix-struct-return
15805 @opindex maix-struct-return
15806 Return all structures in memory (as specified by the AIX ABI)@.
15808 @item -msvr4-struct-return
15809 @opindex msvr4-struct-return
15810 Return structures smaller than 8 bytes in registers (as specified by the
15813 @item -mabi=@var{abi-type}
15815 Extend the current ABI with a particular extension, or remove such extension.
15816 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
15817 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
15821 Extend the current ABI with SPE ABI extensions. This does not change
15822 the default ABI, instead it adds the SPE ABI extensions to the current
15826 @opindex mabi=no-spe
15827 Disable Booke SPE ABI extensions for the current ABI@.
15829 @item -mabi=ibmlongdouble
15830 @opindex mabi=ibmlongdouble
15831 Change the current ABI to use IBM extended precision long double.
15832 This is a PowerPC 32-bit SYSV ABI option.
15834 @item -mabi=ieeelongdouble
15835 @opindex mabi=ieeelongdouble
15836 Change the current ABI to use IEEE extended precision long double.
15837 This is a PowerPC 32-bit Linux ABI option.
15840 @itemx -mno-prototype
15841 @opindex mprototype
15842 @opindex mno-prototype
15843 On System V.4 and embedded PowerPC systems assume that all calls to
15844 variable argument functions are properly prototyped. Otherwise, the
15845 compiler must insert an instruction before every non prototyped call to
15846 set or clear bit 6 of the condition code register (@var{CR}) to
15847 indicate whether floating point values were passed in the floating point
15848 registers in case the function takes a variable arguments. With
15849 @option{-mprototype}, only calls to prototyped variable argument functions
15850 will set or clear the bit.
15854 On embedded PowerPC systems, assume that the startup module is called
15855 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
15856 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
15861 On embedded PowerPC systems, assume that the startup module is called
15862 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
15867 On embedded PowerPC systems, assume that the startup module is called
15868 @file{crt0.o} and the standard C libraries are @file{libads.a} and
15871 @item -myellowknife
15872 @opindex myellowknife
15873 On embedded PowerPC systems, assume that the startup module is called
15874 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
15879 On System V.4 and embedded PowerPC systems, specify that you are
15880 compiling for a VxWorks system.
15884 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
15885 header to indicate that @samp{eabi} extended relocations are used.
15891 On System V.4 and embedded PowerPC systems do (do not) adhere to the
15892 Embedded Applications Binary Interface (eabi) which is a set of
15893 modifications to the System V.4 specifications. Selecting @option{-meabi}
15894 means that the stack is aligned to an 8 byte boundary, a function
15895 @code{__eabi} is called to from @code{main} to set up the eabi
15896 environment, and the @option{-msdata} option can use both @code{r2} and
15897 @code{r13} to point to two separate small data areas. Selecting
15898 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
15899 do not call an initialization function from @code{main}, and the
15900 @option{-msdata} option will only use @code{r13} to point to a single
15901 small data area. The @option{-meabi} option is on by default if you
15902 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
15905 @opindex msdata=eabi
15906 On System V.4 and embedded PowerPC systems, put small initialized
15907 @code{const} global and static data in the @samp{.sdata2} section, which
15908 is pointed to by register @code{r2}. Put small initialized
15909 non-@code{const} global and static data in the @samp{.sdata} section,
15910 which is pointed to by register @code{r13}. Put small uninitialized
15911 global and static data in the @samp{.sbss} section, which is adjacent to
15912 the @samp{.sdata} section. The @option{-msdata=eabi} option is
15913 incompatible with the @option{-mrelocatable} option. The
15914 @option{-msdata=eabi} option also sets the @option{-memb} option.
15917 @opindex msdata=sysv
15918 On System V.4 and embedded PowerPC systems, put small global and static
15919 data in the @samp{.sdata} section, which is pointed to by register
15920 @code{r13}. Put small uninitialized global and static data in the
15921 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
15922 The @option{-msdata=sysv} option is incompatible with the
15923 @option{-mrelocatable} option.
15925 @item -msdata=default
15927 @opindex msdata=default
15929 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
15930 compile code the same as @option{-msdata=eabi}, otherwise compile code the
15931 same as @option{-msdata=sysv}.
15934 @opindex msdata=data
15935 On System V.4 and embedded PowerPC systems, put small global
15936 data in the @samp{.sdata} section. Put small uninitialized global
15937 data in the @samp{.sbss} section. Do not use register @code{r13}
15938 to address small data however. This is the default behavior unless
15939 other @option{-msdata} options are used.
15943 @opindex msdata=none
15945 On embedded PowerPC systems, put all initialized global and static data
15946 in the @samp{.data} section, and all uninitialized data in the
15947 @samp{.bss} section.
15949 @item -mblock-move-inline-limit=@var{num}
15950 @opindex mblock-move-inline-limit
15951 Inline all block moves (such as calls to @code{memcpy} or structure
15952 copies) less than or equal to @var{num} bytes. The minimum value for
15953 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
15954 targets. The default value is target-specific.
15958 @cindex smaller data references (PowerPC)
15959 @cindex .sdata/.sdata2 references (PowerPC)
15960 On embedded PowerPC systems, put global and static items less than or
15961 equal to @var{num} bytes into the small data or bss sections instead of
15962 the normal data or bss section. By default, @var{num} is 8. The
15963 @option{-G @var{num}} switch is also passed to the linker.
15964 All modules should be compiled with the same @option{-G @var{num}} value.
15967 @itemx -mno-regnames
15969 @opindex mno-regnames
15970 On System V.4 and embedded PowerPC systems do (do not) emit register
15971 names in the assembly language output using symbolic forms.
15974 @itemx -mno-longcall
15976 @opindex mno-longcall
15977 By default assume that all calls are far away so that a longer more
15978 expensive calling sequence is required. This is required for calls
15979 further than 32 megabytes (33,554,432 bytes) from the current location.
15980 A short call will be generated if the compiler knows
15981 the call cannot be that far away. This setting can be overridden by
15982 the @code{shortcall} function attribute, or by @code{#pragma
15985 Some linkers are capable of detecting out-of-range calls and generating
15986 glue code on the fly. On these systems, long calls are unnecessary and
15987 generate slower code. As of this writing, the AIX linker can do this,
15988 as can the GNU linker for PowerPC/64. It is planned to add this feature
15989 to the GNU linker for 32-bit PowerPC systems as well.
15991 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
15992 callee, L42'', plus a ``branch island'' (glue code). The two target
15993 addresses represent the callee and the ``branch island''. The
15994 Darwin/PPC linker will prefer the first address and generate a ``bl
15995 callee'' if the PPC ``bl'' instruction will reach the callee directly;
15996 otherwise, the linker will generate ``bl L42'' to call the ``branch
15997 island''. The ``branch island'' is appended to the body of the
15998 calling function; it computes the full 32-bit address of the callee
16001 On Mach-O (Darwin) systems, this option directs the compiler emit to
16002 the glue for every direct call, and the Darwin linker decides whether
16003 to use or discard it.
16005 In the future, we may cause GCC to ignore all longcall specifications
16006 when the linker is known to generate glue.
16008 @item -mtls-markers
16009 @itemx -mno-tls-markers
16010 @opindex mtls-markers
16011 @opindex mno-tls-markers
16012 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
16013 specifying the function argument. The relocation allows ld to
16014 reliably associate function call with argument setup instructions for
16015 TLS optimization, which in turn allows gcc to better schedule the
16020 Adds support for multithreading with the @dfn{pthreads} library.
16021 This option sets flags for both the preprocessor and linker.
16026 This option will enable GCC to use the reciprocal estimate and
16027 reciprocal square root estimate instructions with additional
16028 Newton-Raphson steps to increase precision instead of doing a divide or
16029 square root and divide for floating point arguments. You should use
16030 the @option{-ffast-math} option when using @option{-mrecip} (or at
16031 least @option{-funsafe-math-optimizations},
16032 @option{-finite-math-only}, @option{-freciprocal-math} and
16033 @option{-fno-trapping-math}). Note that while the throughput of the
16034 sequence is generally higher than the throughput of the non-reciprocal
16035 instruction, the precision of the sequence can be decreased by up to 2
16036 ulp (i.e. the inverse of 1.0 equals 0.99999994) for reciprocal square
16039 @item -mrecip=@var{opt}
16040 @opindex mrecip=opt
16041 This option allows to control which reciprocal estimate instructions
16042 may be used. @var{opt} is a comma separated list of options, that may
16043 be preceeded by a @code{!} to invert the option:
16044 @code{all}: enable all estimate instructions,
16045 @code{default}: enable the default instructions, equvalent to @option{-mrecip},
16046 @code{none}: disable all estimate instructions, equivalent to @option{-mno-recip};
16047 @code{div}: enable the reciprocal approximation instructions for both single and double precision;
16048 @code{divf}: enable the single precision reciprocal approximation instructions;
16049 @code{divd}: enable the double precision reciprocal approximation instructions;
16050 @code{rsqrt}: enable the reciprocal square root approximation instructions for both single and double precision;
16051 @code{rsqrtf}: enable the single precision reciprocal square root approximation instructions;
16052 @code{rsqrtd}: enable the double precision reciprocal square root approximation instructions;
16054 So for example, @option{-mrecip=all,!rsqrtd} would enable the
16055 all of the reciprocal estimate instructions, except for the
16056 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
16057 which handle the double precision reciprocal square root calculations.
16059 @item -mrecip-precision
16060 @itemx -mno-recip-precision
16061 @opindex mrecip-precision
16062 Assume (do not assume) that the reciprocal estimate instructions
16063 provide higher precision estimates than is mandated by the powerpc
16064 ABI. Selecting @option{-mcpu=power6} or @option{-mcpu=power7}
16065 automatically selects @option{-mrecip-precision}. The double
16066 precision square root estimate instructions are not generated by
16067 default on low precision machines, since they do not provide an
16068 estimate that converges after three steps.
16070 @item -mveclibabi=@var{type}
16071 @opindex mveclibabi
16072 Specifies the ABI type to use for vectorizing intrinsics using an
16073 external library. The only type supported at present is @code{mass},
16074 which specifies to use IBM's Mathematical Acceleration Subsystem
16075 (MASS) libraries for vectorizing intrinsics using external libraries.
16076 GCC will currently emit calls to @code{acosd2}, @code{acosf4},
16077 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
16078 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
16079 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
16080 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
16081 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
16082 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
16083 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
16084 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
16085 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
16086 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
16087 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
16088 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
16089 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
16090 for power7. Both @option{-ftree-vectorize} and
16091 @option{-funsafe-math-optimizations} have to be enabled. The MASS
16092 libraries will have to be specified at link time.
16097 Generate (do not generate) the @code{friz} instruction when the
16098 @option{-funsafe-math-optimizations} option is used to optimize
16099 rounding a floating point value to 64-bit integer and back to floating
16100 point. The @code{friz} instruction does not return the same value if
16101 the floating point number is too large to fit in an integer.
16105 @subsection RX Options
16108 These command line options are defined for RX targets:
16111 @item -m64bit-doubles
16112 @itemx -m32bit-doubles
16113 @opindex m64bit-doubles
16114 @opindex m32bit-doubles
16115 Make the @code{double} data type be 64-bits (@option{-m64bit-doubles})
16116 or 32-bits (@option{-m32bit-doubles}) in size. The default is
16117 @option{-m32bit-doubles}. @emph{Note} RX floating point hardware only
16118 works on 32-bit values, which is why the default is
16119 @option{-m32bit-doubles}.
16125 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
16126 floating point hardware. The default is enabled for the @var{RX600}
16127 series and disabled for the @var{RX200} series.
16129 Floating point instructions will only be generated for 32-bit floating
16130 point values however, so if the @option{-m64bit-doubles} option is in
16131 use then the FPU hardware will not be used for doubles.
16133 @emph{Note} If the @option{-fpu} option is enabled then
16134 @option{-funsafe-math-optimizations} is also enabled automatically.
16135 This is because the RX FPU instructions are themselves unsafe.
16137 @item -mcpu=@var{name}
16139 Selects the type of RX CPU to be targeted. Currently three types are
16140 supported, the generic @var{RX600} and @var{RX200} series hardware and
16141 the specific @var{RX610} cpu. The default is @var{RX600}.
16143 The only difference between @var{RX600} and @var{RX610} is that the
16144 @var{RX610} does not support the @code{MVTIPL} instruction.
16146 The @var{RX200} series does not have a hardware floating point unit
16147 and so @option{-nofpu} is enabled by default when this type is
16150 @item -mbig-endian-data
16151 @itemx -mlittle-endian-data
16152 @opindex mbig-endian-data
16153 @opindex mlittle-endian-data
16154 Store data (but not code) in the big-endian format. The default is
16155 @option{-mlittle-endian-data}, ie to store data in the little endian
16158 @item -msmall-data-limit=@var{N}
16159 @opindex msmall-data-limit
16160 Specifies the maximum size in bytes of global and static variables
16161 which can be placed into the small data area. Using the small data
16162 area can lead to smaller and faster code, but the size of area is
16163 limited and it is up to the programmer to ensure that the area does
16164 not overflow. Also when the small data area is used one of the RX's
16165 registers (@code{r13}) is reserved for use pointing to this area, so
16166 it is no longer available for use by the compiler. This could result
16167 in slower and/or larger code if variables which once could have been
16168 held in @code{r13} are now pushed onto the stack.
16170 Note, common variables (variables which have not been initialised) and
16171 constants are not placed into the small data area as they are assigned
16172 to other sections in the output executable.
16174 The default value is zero, which disables this feature. Note, this
16175 feature is not enabled by default with higher optimization levels
16176 (@option{-O2} etc) because of the potentially detrimental effects of
16177 reserving register @code{r13}. It is up to the programmer to
16178 experiment and discover whether this feature is of benefit to their
16185 Use the simulator runtime. The default is to use the libgloss board
16188 @item -mas100-syntax
16189 @itemx -mno-as100-syntax
16190 @opindex mas100-syntax
16191 @opindex mno-as100-syntax
16192 When generating assembler output use a syntax that is compatible with
16193 Renesas's AS100 assembler. This syntax can also be handled by the GAS
16194 assembler but it has some restrictions so generating it is not the
16197 @item -mmax-constant-size=@var{N}
16198 @opindex mmax-constant-size
16199 Specifies the maximum size, in bytes, of a constant that can be used as
16200 an operand in a RX instruction. Although the RX instruction set does
16201 allow constants of up to 4 bytes in length to be used in instructions,
16202 a longer value equates to a longer instruction. Thus in some
16203 circumstances it can be beneficial to restrict the size of constants
16204 that are used in instructions. Constants that are too big are instead
16205 placed into a constant pool and referenced via register indirection.
16207 The value @var{N} can be between 0 and 4. A value of 0 (the default)
16208 or 4 means that constants of any size are allowed.
16212 Enable linker relaxation. Linker relaxation is a process whereby the
16213 linker will attempt to reduce the size of a program by finding shorter
16214 versions of various instructions. Disabled by default.
16216 @item -mint-register=@var{N}
16217 @opindex mint-register
16218 Specify the number of registers to reserve for fast interrupt handler
16219 functions. The value @var{N} can be between 0 and 4. A value of 1
16220 means that register @code{r13} will be reserved for the exclusive use
16221 of fast interrupt handlers. A value of 2 reserves @code{r13} and
16222 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
16223 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
16224 A value of 0, the default, does not reserve any registers.
16226 @item -msave-acc-in-interrupts
16227 @opindex msave-acc-in-interrupts
16228 Specifies that interrupt handler functions should preserve the
16229 accumulator register. This is only necessary if normal code might use
16230 the accumulator register, for example because it performs 64-bit
16231 multiplications. The default is to ignore the accumulator as this
16232 makes the interrupt handlers faster.
16236 @emph{Note:} The generic GCC command line @option{-ffixed-@var{reg}}
16237 has special significance to the RX port when used with the
16238 @code{interrupt} function attribute. This attribute indicates a
16239 function intended to process fast interrupts. GCC will will ensure
16240 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
16241 and/or @code{r13} and only provided that the normal use of the
16242 corresponding registers have been restricted via the
16243 @option{-ffixed-@var{reg}} or @option{-mint-register} command line
16246 @node S/390 and zSeries Options
16247 @subsection S/390 and zSeries Options
16248 @cindex S/390 and zSeries Options
16250 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
16254 @itemx -msoft-float
16255 @opindex mhard-float
16256 @opindex msoft-float
16257 Use (do not use) the hardware floating-point instructions and registers
16258 for floating-point operations. When @option{-msoft-float} is specified,
16259 functions in @file{libgcc.a} will be used to perform floating-point
16260 operations. When @option{-mhard-float} is specified, the compiler
16261 generates IEEE floating-point instructions. This is the default.
16264 @itemx -mno-hard-dfp
16266 @opindex mno-hard-dfp
16267 Use (do not use) the hardware decimal-floating-point instructions for
16268 decimal-floating-point operations. When @option{-mno-hard-dfp} is
16269 specified, functions in @file{libgcc.a} will be used to perform
16270 decimal-floating-point operations. When @option{-mhard-dfp} is
16271 specified, the compiler generates decimal-floating-point hardware
16272 instructions. This is the default for @option{-march=z9-ec} or higher.
16274 @item -mlong-double-64
16275 @itemx -mlong-double-128
16276 @opindex mlong-double-64
16277 @opindex mlong-double-128
16278 These switches control the size of @code{long double} type. A size
16279 of 64bit makes the @code{long double} type equivalent to the @code{double}
16280 type. This is the default.
16283 @itemx -mno-backchain
16284 @opindex mbackchain
16285 @opindex mno-backchain
16286 Store (do not store) the address of the caller's frame as backchain pointer
16287 into the callee's stack frame.
16288 A backchain may be needed to allow debugging using tools that do not understand
16289 DWARF-2 call frame information.
16290 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
16291 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
16292 the backchain is placed into the topmost word of the 96/160 byte register
16295 In general, code compiled with @option{-mbackchain} is call-compatible with
16296 code compiled with @option{-mmo-backchain}; however, use of the backchain
16297 for debugging purposes usually requires that the whole binary is built with
16298 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
16299 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
16300 to build a linux kernel use @option{-msoft-float}.
16302 The default is to not maintain the backchain.
16304 @item -mpacked-stack
16305 @itemx -mno-packed-stack
16306 @opindex mpacked-stack
16307 @opindex mno-packed-stack
16308 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
16309 specified, the compiler uses the all fields of the 96/160 byte register save
16310 area only for their default purpose; unused fields still take up stack space.
16311 When @option{-mpacked-stack} is specified, register save slots are densely
16312 packed at the top of the register save area; unused space is reused for other
16313 purposes, allowing for more efficient use of the available stack space.
16314 However, when @option{-mbackchain} is also in effect, the topmost word of
16315 the save area is always used to store the backchain, and the return address
16316 register is always saved two words below the backchain.
16318 As long as the stack frame backchain is not used, code generated with
16319 @option{-mpacked-stack} is call-compatible with code generated with
16320 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
16321 S/390 or zSeries generated code that uses the stack frame backchain at run
16322 time, not just for debugging purposes. Such code is not call-compatible
16323 with code compiled with @option{-mpacked-stack}. Also, note that the
16324 combination of @option{-mbackchain},
16325 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
16326 to build a linux kernel use @option{-msoft-float}.
16328 The default is to not use the packed stack layout.
16331 @itemx -mno-small-exec
16332 @opindex msmall-exec
16333 @opindex mno-small-exec
16334 Generate (or do not generate) code using the @code{bras} instruction
16335 to do subroutine calls.
16336 This only works reliably if the total executable size does not
16337 exceed 64k. The default is to use the @code{basr} instruction instead,
16338 which does not have this limitation.
16344 When @option{-m31} is specified, generate code compliant to the
16345 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
16346 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
16347 particular to generate 64-bit instructions. For the @samp{s390}
16348 targets, the default is @option{-m31}, while the @samp{s390x}
16349 targets default to @option{-m64}.
16355 When @option{-mzarch} is specified, generate code using the
16356 instructions available on z/Architecture.
16357 When @option{-mesa} is specified, generate code using the
16358 instructions available on ESA/390. Note that @option{-mesa} is
16359 not possible with @option{-m64}.
16360 When generating code compliant to the GNU/Linux for S/390 ABI,
16361 the default is @option{-mesa}. When generating code compliant
16362 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
16368 Generate (or do not generate) code using the @code{mvcle} instruction
16369 to perform block moves. When @option{-mno-mvcle} is specified,
16370 use a @code{mvc} loop instead. This is the default unless optimizing for
16377 Print (or do not print) additional debug information when compiling.
16378 The default is to not print debug information.
16380 @item -march=@var{cpu-type}
16382 Generate code that will run on @var{cpu-type}, which is the name of a system
16383 representing a certain processor type. Possible values for
16384 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
16385 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
16386 When generating code using the instructions available on z/Architecture,
16387 the default is @option{-march=z900}. Otherwise, the default is
16388 @option{-march=g5}.
16390 @item -mtune=@var{cpu-type}
16392 Tune to @var{cpu-type} everything applicable about the generated code,
16393 except for the ABI and the set of available instructions.
16394 The list of @var{cpu-type} values is the same as for @option{-march}.
16395 The default is the value used for @option{-march}.
16398 @itemx -mno-tpf-trace
16399 @opindex mtpf-trace
16400 @opindex mno-tpf-trace
16401 Generate code that adds (does not add) in TPF OS specific branches to trace
16402 routines in the operating system. This option is off by default, even
16403 when compiling for the TPF OS@.
16406 @itemx -mno-fused-madd
16407 @opindex mfused-madd
16408 @opindex mno-fused-madd
16409 Generate code that uses (does not use) the floating point multiply and
16410 accumulate instructions. These instructions are generated by default if
16411 hardware floating point is used.
16413 @item -mwarn-framesize=@var{framesize}
16414 @opindex mwarn-framesize
16415 Emit a warning if the current function exceeds the given frame size. Because
16416 this is a compile time check it doesn't need to be a real problem when the program
16417 runs. It is intended to identify functions which most probably cause
16418 a stack overflow. It is useful to be used in an environment with limited stack
16419 size e.g.@: the linux kernel.
16421 @item -mwarn-dynamicstack
16422 @opindex mwarn-dynamicstack
16423 Emit a warning if the function calls alloca or uses dynamically
16424 sized arrays. This is generally a bad idea with a limited stack size.
16426 @item -mstack-guard=@var{stack-guard}
16427 @itemx -mstack-size=@var{stack-size}
16428 @opindex mstack-guard
16429 @opindex mstack-size
16430 If these options are provided the s390 back end emits additional instructions in
16431 the function prologue which trigger a trap if the stack size is @var{stack-guard}
16432 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
16433 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
16434 the frame size of the compiled function is chosen.
16435 These options are intended to be used to help debugging stack overflow problems.
16436 The additionally emitted code causes only little overhead and hence can also be
16437 used in production like systems without greater performance degradation. The given
16438 values have to be exact powers of 2 and @var{stack-size} has to be greater than
16439 @var{stack-guard} without exceeding 64k.
16440 In order to be efficient the extra code makes the assumption that the stack starts
16441 at an address aligned to the value given by @var{stack-size}.
16442 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
16445 @node Score Options
16446 @subsection Score Options
16447 @cindex Score Options
16449 These options are defined for Score implementations:
16454 Compile code for big endian mode. This is the default.
16458 Compile code for little endian mode.
16462 Disable generate bcnz instruction.
16466 Enable generate unaligned load and store instruction.
16470 Enable the use of multiply-accumulate instructions. Disabled by default.
16474 Specify the SCORE5 as the target architecture.
16478 Specify the SCORE5U of the target architecture.
16482 Specify the SCORE7 as the target architecture. This is the default.
16486 Specify the SCORE7D as the target architecture.
16490 @subsection SH Options
16492 These @samp{-m} options are defined for the SH implementations:
16497 Generate code for the SH1.
16501 Generate code for the SH2.
16504 Generate code for the SH2e.
16508 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
16509 that the floating-point unit is not used.
16511 @item -m2a-single-only
16512 @opindex m2a-single-only
16513 Generate code for the SH2a-FPU, in such a way that no double-precision
16514 floating point operations are used.
16517 @opindex m2a-single
16518 Generate code for the SH2a-FPU assuming the floating-point unit is in
16519 single-precision mode by default.
16523 Generate code for the SH2a-FPU assuming the floating-point unit is in
16524 double-precision mode by default.
16528 Generate code for the SH3.
16532 Generate code for the SH3e.
16536 Generate code for the SH4 without a floating-point unit.
16538 @item -m4-single-only
16539 @opindex m4-single-only
16540 Generate code for the SH4 with a floating-point unit that only
16541 supports single-precision arithmetic.
16545 Generate code for the SH4 assuming the floating-point unit is in
16546 single-precision mode by default.
16550 Generate code for the SH4.
16554 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
16555 floating-point unit is not used.
16557 @item -m4a-single-only
16558 @opindex m4a-single-only
16559 Generate code for the SH4a, in such a way that no double-precision
16560 floating point operations are used.
16563 @opindex m4a-single
16564 Generate code for the SH4a assuming the floating-point unit is in
16565 single-precision mode by default.
16569 Generate code for the SH4a.
16573 Same as @option{-m4a-nofpu}, except that it implicitly passes
16574 @option{-dsp} to the assembler. GCC doesn't generate any DSP
16575 instructions at the moment.
16579 Compile code for the processor in big endian mode.
16583 Compile code for the processor in little endian mode.
16587 Align doubles at 64-bit boundaries. Note that this changes the calling
16588 conventions, and thus some functions from the standard C library will
16589 not work unless you recompile it first with @option{-mdalign}.
16593 Shorten some address references at link time, when possible; uses the
16594 linker option @option{-relax}.
16598 Use 32-bit offsets in @code{switch} tables. The default is to use
16603 Enable the use of bit manipulation instructions on SH2A.
16607 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
16608 alignment constraints.
16612 Comply with the calling conventions defined by Renesas.
16616 Comply with the calling conventions defined by Renesas.
16620 Comply with the calling conventions defined for GCC before the Renesas
16621 conventions were available. This option is the default for all
16622 targets of the SH toolchain except for @samp{sh-symbianelf}.
16625 @opindex mnomacsave
16626 Mark the @code{MAC} register as call-clobbered, even if
16627 @option{-mhitachi} is given.
16631 Increase IEEE-compliance of floating-point code.
16632 At the moment, this is equivalent to @option{-fno-finite-math-only}.
16633 When generating 16 bit SH opcodes, getting IEEE-conforming results for
16634 comparisons of NANs / infinities incurs extra overhead in every
16635 floating point comparison, therefore the default is set to
16636 @option{-ffinite-math-only}.
16638 @item -minline-ic_invalidate
16639 @opindex minline-ic_invalidate
16640 Inline code to invalidate instruction cache entries after setting up
16641 nested function trampolines.
16642 This option has no effect if -musermode is in effect and the selected
16643 code generation option (e.g. -m4) does not allow the use of the icbi
16645 If the selected code generation option does not allow the use of the icbi
16646 instruction, and -musermode is not in effect, the inlined code will
16647 manipulate the instruction cache address array directly with an associative
16648 write. This not only requires privileged mode, but it will also
16649 fail if the cache line had been mapped via the TLB and has become unmapped.
16653 Dump instruction size and location in the assembly code.
16656 @opindex mpadstruct
16657 This option is deprecated. It pads structures to multiple of 4 bytes,
16658 which is incompatible with the SH ABI@.
16662 Optimize for space instead of speed. Implied by @option{-Os}.
16665 @opindex mprefergot
16666 When generating position-independent code, emit function calls using
16667 the Global Offset Table instead of the Procedure Linkage Table.
16671 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
16672 if the inlined code would not work in user mode.
16673 This is the default when the target is @code{sh-*-linux*}.
16675 @item -multcost=@var{number}
16676 @opindex multcost=@var{number}
16677 Set the cost to assume for a multiply insn.
16679 @item -mdiv=@var{strategy}
16680 @opindex mdiv=@var{strategy}
16681 Set the division strategy to use for SHmedia code. @var{strategy} must be
16682 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
16683 inv:call2, inv:fp .
16684 "fp" performs the operation in floating point. This has a very high latency,
16685 but needs only a few instructions, so it might be a good choice if
16686 your code has enough easily exploitable ILP to allow the compiler to
16687 schedule the floating point instructions together with other instructions.
16688 Division by zero causes a floating point exception.
16689 "inv" uses integer operations to calculate the inverse of the divisor,
16690 and then multiplies the dividend with the inverse. This strategy allows
16691 cse and hoisting of the inverse calculation. Division by zero calculates
16692 an unspecified result, but does not trap.
16693 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
16694 have been found, or if the entire operation has been hoisted to the same
16695 place, the last stages of the inverse calculation are intertwined with the
16696 final multiply to reduce the overall latency, at the expense of using a few
16697 more instructions, and thus offering fewer scheduling opportunities with
16699 "call" calls a library function that usually implements the inv:minlat
16701 This gives high code density for m5-*media-nofpu compilations.
16702 "call2" uses a different entry point of the same library function, where it
16703 assumes that a pointer to a lookup table has already been set up, which
16704 exposes the pointer load to cse / code hoisting optimizations.
16705 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
16706 code generation, but if the code stays unoptimized, revert to the "call",
16707 "call2", or "fp" strategies, respectively. Note that the
16708 potentially-trapping side effect of division by zero is carried by a
16709 separate instruction, so it is possible that all the integer instructions
16710 are hoisted out, but the marker for the side effect stays where it is.
16711 A recombination to fp operations or a call is not possible in that case.
16712 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
16713 that the inverse calculation was nor separated from the multiply, they speed
16714 up division where the dividend fits into 20 bits (plus sign where applicable),
16715 by inserting a test to skip a number of operations in this case; this test
16716 slows down the case of larger dividends. inv20u assumes the case of a such
16717 a small dividend to be unlikely, and inv20l assumes it to be likely.
16719 @item -maccumulate-outgoing-args
16720 @opindex maccumulate-outgoing-args
16721 Reserve space once for outgoing arguments in the function prologue rather
16722 than around each call. Generally beneficial for performance and size. Also
16723 needed for unwinding to avoid changing the stack frame around conditional code.
16725 @item -mdivsi3_libfunc=@var{name}
16726 @opindex mdivsi3_libfunc=@var{name}
16727 Set the name of the library function used for 32 bit signed division to
16728 @var{name}. This only affect the name used in the call and inv:call
16729 division strategies, and the compiler will still expect the same
16730 sets of input/output/clobbered registers as if this option was not present.
16732 @item -mfixed-range=@var{register-range}
16733 @opindex mfixed-range
16734 Generate code treating the given register range as fixed registers.
16735 A fixed register is one that the register allocator can not use. This is
16736 useful when compiling kernel code. A register range is specified as
16737 two registers separated by a dash. Multiple register ranges can be
16738 specified separated by a comma.
16740 @item -madjust-unroll
16741 @opindex madjust-unroll
16742 Throttle unrolling to avoid thrashing target registers.
16743 This option only has an effect if the gcc code base supports the
16744 TARGET_ADJUST_UNROLL_MAX target hook.
16746 @item -mindexed-addressing
16747 @opindex mindexed-addressing
16748 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
16749 This is only safe if the hardware and/or OS implement 32 bit wrap-around
16750 semantics for the indexed addressing mode. The architecture allows the
16751 implementation of processors with 64 bit MMU, which the OS could use to
16752 get 32 bit addressing, but since no current hardware implementation supports
16753 this or any other way to make the indexed addressing mode safe to use in
16754 the 32 bit ABI, the default is -mno-indexed-addressing.
16756 @item -mgettrcost=@var{number}
16757 @opindex mgettrcost=@var{number}
16758 Set the cost assumed for the gettr instruction to @var{number}.
16759 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
16763 Assume pt* instructions won't trap. This will generally generate better
16764 scheduled code, but is unsafe on current hardware. The current architecture
16765 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
16766 This has the unintentional effect of making it unsafe to schedule ptabs /
16767 ptrel before a branch, or hoist it out of a loop. For example,
16768 __do_global_ctors, a part of libgcc that runs constructors at program
16769 startup, calls functions in a list which is delimited by @minus{}1. With the
16770 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
16771 That means that all the constructors will be run a bit quicker, but when
16772 the loop comes to the end of the list, the program crashes because ptabs
16773 loads @minus{}1 into a target register. Since this option is unsafe for any
16774 hardware implementing the current architecture specification, the default
16775 is -mno-pt-fixed. Unless the user specifies a specific cost with
16776 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
16777 this deters register allocation using target registers for storing
16780 @item -minvalid-symbols
16781 @opindex minvalid-symbols
16782 Assume symbols might be invalid. Ordinary function symbols generated by
16783 the compiler will always be valid to load with movi/shori/ptabs or
16784 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
16785 to generate symbols that will cause ptabs / ptrel to trap.
16786 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
16787 It will then prevent cross-basic-block cse, hoisting and most scheduling
16788 of symbol loads. The default is @option{-mno-invalid-symbols}.
16791 @node Solaris 2 Options
16792 @subsection Solaris 2 Options
16793 @cindex Solaris 2 options
16795 These @samp{-m} options are supported on Solaris 2:
16798 @item -mimpure-text
16799 @opindex mimpure-text
16800 @option{-mimpure-text}, used in addition to @option{-shared}, tells
16801 the compiler to not pass @option{-z text} to the linker when linking a
16802 shared object. Using this option, you can link position-dependent
16803 code into a shared object.
16805 @option{-mimpure-text} suppresses the ``relocations remain against
16806 allocatable but non-writable sections'' linker error message.
16807 However, the necessary relocations will trigger copy-on-write, and the
16808 shared object is not actually shared across processes. Instead of
16809 using @option{-mimpure-text}, you should compile all source code with
16810 @option{-fpic} or @option{-fPIC}.
16814 These switches are supported in addition to the above on Solaris 2:
16819 Add support for multithreading using the Solaris threads library. This
16820 option sets flags for both the preprocessor and linker. This option does
16821 not affect the thread safety of object code produced by the compiler or
16822 that of libraries supplied with it.
16826 Add support for multithreading using the POSIX threads library. This
16827 option sets flags for both the preprocessor and linker. This option does
16828 not affect the thread safety of object code produced by the compiler or
16829 that of libraries supplied with it.
16833 This is a synonym for @option{-pthreads}.
16836 @node SPARC Options
16837 @subsection SPARC Options
16838 @cindex SPARC options
16840 These @samp{-m} options are supported on the SPARC:
16843 @item -mno-app-regs
16845 @opindex mno-app-regs
16847 Specify @option{-mapp-regs} to generate output using the global registers
16848 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
16851 To be fully SVR4 ABI compliant at the cost of some performance loss,
16852 specify @option{-mno-app-regs}. You should compile libraries and system
16853 software with this option.
16856 @itemx -mhard-float
16858 @opindex mhard-float
16859 Generate output containing floating point instructions. This is the
16863 @itemx -msoft-float
16865 @opindex msoft-float
16866 Generate output containing library calls for floating point.
16867 @strong{Warning:} the requisite libraries are not available for all SPARC
16868 targets. Normally the facilities of the machine's usual C compiler are
16869 used, but this cannot be done directly in cross-compilation. You must make
16870 your own arrangements to provide suitable library functions for
16871 cross-compilation. The embedded targets @samp{sparc-*-aout} and
16872 @samp{sparclite-*-*} do provide software floating point support.
16874 @option{-msoft-float} changes the calling convention in the output file;
16875 therefore, it is only useful if you compile @emph{all} of a program with
16876 this option. In particular, you need to compile @file{libgcc.a}, the
16877 library that comes with GCC, with @option{-msoft-float} in order for
16880 @item -mhard-quad-float
16881 @opindex mhard-quad-float
16882 Generate output containing quad-word (long double) floating point
16885 @item -msoft-quad-float
16886 @opindex msoft-quad-float
16887 Generate output containing library calls for quad-word (long double)
16888 floating point instructions. The functions called are those specified
16889 in the SPARC ABI@. This is the default.
16891 As of this writing, there are no SPARC implementations that have hardware
16892 support for the quad-word floating point instructions. They all invoke
16893 a trap handler for one of these instructions, and then the trap handler
16894 emulates the effect of the instruction. Because of the trap handler overhead,
16895 this is much slower than calling the ABI library routines. Thus the
16896 @option{-msoft-quad-float} option is the default.
16898 @item -mno-unaligned-doubles
16899 @itemx -munaligned-doubles
16900 @opindex mno-unaligned-doubles
16901 @opindex munaligned-doubles
16902 Assume that doubles have 8 byte alignment. This is the default.
16904 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
16905 alignment only if they are contained in another type, or if they have an
16906 absolute address. Otherwise, it assumes they have 4 byte alignment.
16907 Specifying this option avoids some rare compatibility problems with code
16908 generated by other compilers. It is not the default because it results
16909 in a performance loss, especially for floating point code.
16911 @item -mno-faster-structs
16912 @itemx -mfaster-structs
16913 @opindex mno-faster-structs
16914 @opindex mfaster-structs
16915 With @option{-mfaster-structs}, the compiler assumes that structures
16916 should have 8 byte alignment. This enables the use of pairs of
16917 @code{ldd} and @code{std} instructions for copies in structure
16918 assignment, in place of twice as many @code{ld} and @code{st} pairs.
16919 However, the use of this changed alignment directly violates the SPARC
16920 ABI@. Thus, it's intended only for use on targets where the developer
16921 acknowledges that their resulting code will not be directly in line with
16922 the rules of the ABI@.
16924 @item -mcpu=@var{cpu_type}
16926 Set the instruction set, register set, and instruction scheduling parameters
16927 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
16928 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
16929 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
16930 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
16931 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
16933 Default instruction scheduling parameters are used for values that select
16934 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
16935 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
16937 Here is a list of each supported architecture and their supported
16942 v8: supersparc, hypersparc
16943 sparclite: f930, f934, sparclite86x
16945 v9: ultrasparc, ultrasparc3, niagara, niagara2
16948 By default (unless configured otherwise), GCC generates code for the V7
16949 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
16950 additionally optimizes it for the Cypress CY7C602 chip, as used in the
16951 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
16952 SPARCStation 1, 2, IPX etc.
16954 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
16955 architecture. The only difference from V7 code is that the compiler emits
16956 the integer multiply and integer divide instructions which exist in SPARC-V8
16957 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
16958 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
16961 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
16962 the SPARC architecture. This adds the integer multiply, integer divide step
16963 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
16964 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
16965 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
16966 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
16967 MB86934 chip, which is the more recent SPARClite with FPU@.
16969 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
16970 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
16971 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
16972 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
16973 optimizes it for the TEMIC SPARClet chip.
16975 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
16976 architecture. This adds 64-bit integer and floating-point move instructions,
16977 3 additional floating-point condition code registers and conditional move
16978 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
16979 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
16980 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
16981 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
16982 @option{-mcpu=niagara}, the compiler additionally optimizes it for
16983 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
16984 additionally optimizes it for Sun UltraSPARC T2 chips.
16986 @item -mtune=@var{cpu_type}
16988 Set the instruction scheduling parameters for machine type
16989 @var{cpu_type}, but do not set the instruction set or register set that the
16990 option @option{-mcpu=@var{cpu_type}} would.
16992 The same values for @option{-mcpu=@var{cpu_type}} can be used for
16993 @option{-mtune=@var{cpu_type}}, but the only useful values are those
16994 that select a particular cpu implementation. Those are @samp{cypress},
16995 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
16996 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
16997 @samp{ultrasparc3}, @samp{niagara}, and @samp{niagara2}.
17002 @opindex mno-v8plus
17003 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
17004 difference from the V8 ABI is that the global and out registers are
17005 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
17006 mode for all SPARC-V9 processors.
17012 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
17013 Visual Instruction Set extensions. The default is @option{-mno-vis}.
17016 These @samp{-m} options are supported in addition to the above
17017 on SPARC-V9 processors in 64-bit environments:
17020 @item -mlittle-endian
17021 @opindex mlittle-endian
17022 Generate code for a processor running in little-endian mode. It is only
17023 available for a few configurations and most notably not on Solaris and Linux.
17029 Generate code for a 32-bit or 64-bit environment.
17030 The 32-bit environment sets int, long and pointer to 32 bits.
17031 The 64-bit environment sets int to 32 bits and long and pointer
17034 @item -mcmodel=medlow
17035 @opindex mcmodel=medlow
17036 Generate code for the Medium/Low code model: 64-bit addresses, programs
17037 must be linked in the low 32 bits of memory. Programs can be statically
17038 or dynamically linked.
17040 @item -mcmodel=medmid
17041 @opindex mcmodel=medmid
17042 Generate code for the Medium/Middle code model: 64-bit addresses, programs
17043 must be linked in the low 44 bits of memory, the text and data segments must
17044 be less than 2GB in size and the data segment must be located within 2GB of
17047 @item -mcmodel=medany
17048 @opindex mcmodel=medany
17049 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
17050 may be linked anywhere in memory, the text and data segments must be less
17051 than 2GB in size and the data segment must be located within 2GB of the
17054 @item -mcmodel=embmedany
17055 @opindex mcmodel=embmedany
17056 Generate code for the Medium/Anywhere code model for embedded systems:
17057 64-bit addresses, the text and data segments must be less than 2GB in
17058 size, both starting anywhere in memory (determined at link time). The
17059 global register %g4 points to the base of the data segment. Programs
17060 are statically linked and PIC is not supported.
17063 @itemx -mno-stack-bias
17064 @opindex mstack-bias
17065 @opindex mno-stack-bias
17066 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
17067 frame pointer if present, are offset by @minus{}2047 which must be added back
17068 when making stack frame references. This is the default in 64-bit mode.
17069 Otherwise, assume no such offset is present.
17073 @subsection SPU Options
17074 @cindex SPU options
17076 These @samp{-m} options are supported on the SPU:
17080 @itemx -merror-reloc
17081 @opindex mwarn-reloc
17082 @opindex merror-reloc
17084 The loader for SPU does not handle dynamic relocations. By default, GCC
17085 will give an error when it generates code that requires a dynamic
17086 relocation. @option{-mno-error-reloc} disables the error,
17087 @option{-mwarn-reloc} will generate a warning instead.
17090 @itemx -munsafe-dma
17092 @opindex munsafe-dma
17094 Instructions which initiate or test completion of DMA must not be
17095 reordered with respect to loads and stores of the memory which is being
17096 accessed. Users typically address this problem using the volatile
17097 keyword, but that can lead to inefficient code in places where the
17098 memory is known to not change. Rather than mark the memory as volatile
17099 we treat the DMA instructions as potentially effecting all memory. With
17100 @option{-munsafe-dma} users must use the volatile keyword to protect
17103 @item -mbranch-hints
17104 @opindex mbranch-hints
17106 By default, GCC will generate a branch hint instruction to avoid
17107 pipeline stalls for always taken or probably taken branches. A hint
17108 will not be generated closer than 8 instructions away from its branch.
17109 There is little reason to disable them, except for debugging purposes,
17110 or to make an object a little bit smaller.
17114 @opindex msmall-mem
17115 @opindex mlarge-mem
17117 By default, GCC generates code assuming that addresses are never larger
17118 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
17119 a full 32 bit address.
17124 By default, GCC links against startup code that assumes the SPU-style
17125 main function interface (which has an unconventional parameter list).
17126 With @option{-mstdmain}, GCC will link your program against startup
17127 code that assumes a C99-style interface to @code{main}, including a
17128 local copy of @code{argv} strings.
17130 @item -mfixed-range=@var{register-range}
17131 @opindex mfixed-range
17132 Generate code treating the given register range as fixed registers.
17133 A fixed register is one that the register allocator can not use. This is
17134 useful when compiling kernel code. A register range is specified as
17135 two registers separated by a dash. Multiple register ranges can be
17136 specified separated by a comma.
17142 Compile code assuming that pointers to the PPU address space accessed
17143 via the @code{__ea} named address space qualifier are either 32 or 64
17144 bits wide. The default is 32 bits. As this is an ABI changing option,
17145 all object code in an executable must be compiled with the same setting.
17147 @item -maddress-space-conversion
17148 @itemx -mno-address-space-conversion
17149 @opindex maddress-space-conversion
17150 @opindex mno-address-space-conversion
17151 Allow/disallow treating the @code{__ea} address space as superset
17152 of the generic address space. This enables explicit type casts
17153 between @code{__ea} and generic pointer as well as implicit
17154 conversions of generic pointers to @code{__ea} pointers. The
17155 default is to allow address space pointer conversions.
17157 @item -mcache-size=@var{cache-size}
17158 @opindex mcache-size
17159 This option controls the version of libgcc that the compiler links to an
17160 executable and selects a software-managed cache for accessing variables
17161 in the @code{__ea} address space with a particular cache size. Possible
17162 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
17163 and @samp{128}. The default cache size is 64KB.
17165 @item -matomic-updates
17166 @itemx -mno-atomic-updates
17167 @opindex matomic-updates
17168 @opindex mno-atomic-updates
17169 This option controls the version of libgcc that the compiler links to an
17170 executable and selects whether atomic updates to the software-managed
17171 cache of PPU-side variables are used. If you use atomic updates, changes
17172 to a PPU variable from SPU code using the @code{__ea} named address space
17173 qualifier will not interfere with changes to other PPU variables residing
17174 in the same cache line from PPU code. If you do not use atomic updates,
17175 such interference may occur; however, writing back cache lines will be
17176 more efficient. The default behavior is to use atomic updates.
17179 @itemx -mdual-nops=@var{n}
17180 @opindex mdual-nops
17181 By default, GCC will insert nops to increase dual issue when it expects
17182 it to increase performance. @var{n} can be a value from 0 to 10. A
17183 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
17184 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
17186 @item -mhint-max-nops=@var{n}
17187 @opindex mhint-max-nops
17188 Maximum number of nops to insert for a branch hint. A branch hint must
17189 be at least 8 instructions away from the branch it is effecting. GCC
17190 will insert up to @var{n} nops to enforce this, otherwise it will not
17191 generate the branch hint.
17193 @item -mhint-max-distance=@var{n}
17194 @opindex mhint-max-distance
17195 The encoding of the branch hint instruction limits the hint to be within
17196 256 instructions of the branch it is effecting. By default, GCC makes
17197 sure it is within 125.
17200 @opindex msafe-hints
17201 Work around a hardware bug which causes the SPU to stall indefinitely.
17202 By default, GCC will insert the @code{hbrp} instruction to make sure
17203 this stall won't happen.
17207 @node System V Options
17208 @subsection Options for System V
17210 These additional options are available on System V Release 4 for
17211 compatibility with other compilers on those systems:
17216 Create a shared object.
17217 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
17221 Identify the versions of each tool used by the compiler, in a
17222 @code{.ident} assembler directive in the output.
17226 Refrain from adding @code{.ident} directives to the output file (this is
17229 @item -YP,@var{dirs}
17231 Search the directories @var{dirs}, and no others, for libraries
17232 specified with @option{-l}.
17234 @item -Ym,@var{dir}
17236 Look in the directory @var{dir} to find the M4 preprocessor.
17237 The assembler uses this option.
17238 @c This is supposed to go with a -Yd for predefined M4 macro files, but
17239 @c the generic assembler that comes with Solaris takes just -Ym.
17243 @subsection V850 Options
17244 @cindex V850 Options
17246 These @samp{-m} options are defined for V850 implementations:
17250 @itemx -mno-long-calls
17251 @opindex mlong-calls
17252 @opindex mno-long-calls
17253 Treat all calls as being far away (near). If calls are assumed to be
17254 far away, the compiler will always load the functions address up into a
17255 register, and call indirect through the pointer.
17261 Do not optimize (do optimize) basic blocks that use the same index
17262 pointer 4 or more times to copy pointer into the @code{ep} register, and
17263 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
17264 option is on by default if you optimize.
17266 @item -mno-prolog-function
17267 @itemx -mprolog-function
17268 @opindex mno-prolog-function
17269 @opindex mprolog-function
17270 Do not use (do use) external functions to save and restore registers
17271 at the prologue and epilogue of a function. The external functions
17272 are slower, but use less code space if more than one function saves
17273 the same number of registers. The @option{-mprolog-function} option
17274 is on by default if you optimize.
17278 Try to make the code as small as possible. At present, this just turns
17279 on the @option{-mep} and @option{-mprolog-function} options.
17281 @item -mtda=@var{n}
17283 Put static or global variables whose size is @var{n} bytes or less into
17284 the tiny data area that register @code{ep} points to. The tiny data
17285 area can hold up to 256 bytes in total (128 bytes for byte references).
17287 @item -msda=@var{n}
17289 Put static or global variables whose size is @var{n} bytes or less into
17290 the small data area that register @code{gp} points to. The small data
17291 area can hold up to 64 kilobytes.
17293 @item -mzda=@var{n}
17295 Put static or global variables whose size is @var{n} bytes or less into
17296 the first 32 kilobytes of memory.
17300 Specify that the target processor is the V850.
17303 @opindex mbig-switch
17304 Generate code suitable for big switch tables. Use this option only if
17305 the assembler/linker complain about out of range branches within a switch
17310 This option will cause r2 and r5 to be used in the code generated by
17311 the compiler. This setting is the default.
17313 @item -mno-app-regs
17314 @opindex mno-app-regs
17315 This option will cause r2 and r5 to be treated as fixed registers.
17319 Specify that the target processor is the V850E2V3. The preprocessor
17320 constants @samp{__v850e2v3__} will be defined if
17321 this option is used.
17325 Specify that the target processor is the V850E2. The preprocessor
17326 constants @samp{__v850e2__} will be defined if
17330 Specify that the target processor is the V850E1. The preprocessor
17331 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
17335 Specify that the target processor is the V850E@. The preprocessor
17336 constant @samp{__v850e__} will be defined if this option is used.
17338 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
17339 nor @option{-mv850e2} nor @option{-mv850e2v3}
17340 are defined then a default target processor will be chosen and the
17341 relevant @samp{__v850*__} preprocessor constant will be defined.
17343 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
17344 defined, regardless of which processor variant is the target.
17346 @item -mdisable-callt
17347 @opindex mdisable-callt
17348 This option will suppress generation of the CALLT instruction for the
17349 v850e, v850e1, v850e2 and v850e2v3 flavors of the v850 architecture. The default is
17350 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
17355 @subsection VAX Options
17356 @cindex VAX options
17358 These @samp{-m} options are defined for the VAX:
17363 Do not output certain jump instructions (@code{aobleq} and so on)
17364 that the Unix assembler for the VAX cannot handle across long
17369 Do output those jump instructions, on the assumption that you
17370 will assemble with the GNU assembler.
17374 Output code for g-format floating point numbers instead of d-format.
17377 @node VxWorks Options
17378 @subsection VxWorks Options
17379 @cindex VxWorks Options
17381 The options in this section are defined for all VxWorks targets.
17382 Options specific to the target hardware are listed with the other
17383 options for that target.
17388 GCC can generate code for both VxWorks kernels and real time processes
17389 (RTPs). This option switches from the former to the latter. It also
17390 defines the preprocessor macro @code{__RTP__}.
17393 @opindex non-static
17394 Link an RTP executable against shared libraries rather than static
17395 libraries. The options @option{-static} and @option{-shared} can
17396 also be used for RTPs (@pxref{Link Options}); @option{-static}
17403 These options are passed down to the linker. They are defined for
17404 compatibility with Diab.
17407 @opindex Xbind-lazy
17408 Enable lazy binding of function calls. This option is equivalent to
17409 @option{-Wl,-z,now} and is defined for compatibility with Diab.
17413 Disable lazy binding of function calls. This option is the default and
17414 is defined for compatibility with Diab.
17417 @node x86-64 Options
17418 @subsection x86-64 Options
17419 @cindex x86-64 options
17421 These are listed under @xref{i386 and x86-64 Options}.
17423 @node i386 and x86-64 Windows Options
17424 @subsection i386 and x86-64 Windows Options
17425 @cindex i386 and x86-64 Windows Options
17427 These additional options are available for Windows targets:
17432 This option is available for Cygwin and MinGW targets. It
17433 specifies that a console application is to be generated, by
17434 instructing the linker to set the PE header subsystem type
17435 required for console applications.
17436 This is the default behavior for Cygwin and MinGW targets.
17440 This option is available for Cygwin and MinGW targets. It
17441 specifies that a DLL - a dynamic link library - is to be
17442 generated, enabling the selection of the required runtime
17443 startup object and entry point.
17445 @item -mnop-fun-dllimport
17446 @opindex mnop-fun-dllimport
17447 This option is available for Cygwin and MinGW targets. It
17448 specifies that the dllimport attribute should be ignored.
17452 This option is available for MinGW targets. It specifies
17453 that MinGW-specific thread support is to be used.
17457 This option is available for mingw-w64 targets. It specifies
17458 that the UNICODE macro is getting pre-defined and that the
17459 unicode capable runtime startup code is chosen.
17463 This option is available for Cygwin and MinGW targets. It
17464 specifies that the typical Windows pre-defined macros are to
17465 be set in the pre-processor, but does not influence the choice
17466 of runtime library/startup code.
17470 This option is available for Cygwin and MinGW targets. It
17471 specifies that a GUI application is to be generated by
17472 instructing the linker to set the PE header subsystem type
17475 @item -fno-set-stack-executable
17476 @opindex fno-set-stack-executable
17477 This option is available for MinGW targets. It specifies that
17478 the executable flag for stack used by nested functions isn't
17479 set. This is necessary for binaries running in kernel mode of
17480 Windows, as there the user32 API, which is used to set executable
17481 privileges, isn't available.
17483 @item -mpe-aligned-commons
17484 @opindex mpe-aligned-commons
17485 This option is available for Cygwin and MinGW targets. It
17486 specifies that the GNU extension to the PE file format that
17487 permits the correct alignment of COMMON variables should be
17488 used when generating code. It will be enabled by default if
17489 GCC detects that the target assembler found during configuration
17490 supports the feature.
17493 See also under @ref{i386 and x86-64 Options} for standard options.
17495 @node Xstormy16 Options
17496 @subsection Xstormy16 Options
17497 @cindex Xstormy16 Options
17499 These options are defined for Xstormy16:
17504 Choose startup files and linker script suitable for the simulator.
17507 @node Xtensa Options
17508 @subsection Xtensa Options
17509 @cindex Xtensa Options
17511 These options are supported for Xtensa targets:
17515 @itemx -mno-const16
17517 @opindex mno-const16
17518 Enable or disable use of @code{CONST16} instructions for loading
17519 constant values. The @code{CONST16} instruction is currently not a
17520 standard option from Tensilica. When enabled, @code{CONST16}
17521 instructions are always used in place of the standard @code{L32R}
17522 instructions. The use of @code{CONST16} is enabled by default only if
17523 the @code{L32R} instruction is not available.
17526 @itemx -mno-fused-madd
17527 @opindex mfused-madd
17528 @opindex mno-fused-madd
17529 Enable or disable use of fused multiply/add and multiply/subtract
17530 instructions in the floating-point option. This has no effect if the
17531 floating-point option is not also enabled. Disabling fused multiply/add
17532 and multiply/subtract instructions forces the compiler to use separate
17533 instructions for the multiply and add/subtract operations. This may be
17534 desirable in some cases where strict IEEE 754-compliant results are
17535 required: the fused multiply add/subtract instructions do not round the
17536 intermediate result, thereby producing results with @emph{more} bits of
17537 precision than specified by the IEEE standard. Disabling fused multiply
17538 add/subtract instructions also ensures that the program output is not
17539 sensitive to the compiler's ability to combine multiply and add/subtract
17542 @item -mserialize-volatile
17543 @itemx -mno-serialize-volatile
17544 @opindex mserialize-volatile
17545 @opindex mno-serialize-volatile
17546 When this option is enabled, GCC inserts @code{MEMW} instructions before
17547 @code{volatile} memory references to guarantee sequential consistency.
17548 The default is @option{-mserialize-volatile}. Use
17549 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
17551 @item -mforce-no-pic
17552 @opindex mforce-no-pic
17553 For targets, like GNU/Linux, where all user-mode Xtensa code must be
17554 position-independent code (PIC), this option disables PIC for compiling
17557 @item -mtext-section-literals
17558 @itemx -mno-text-section-literals
17559 @opindex mtext-section-literals
17560 @opindex mno-text-section-literals
17561 Control the treatment of literal pools. The default is
17562 @option{-mno-text-section-literals}, which places literals in a separate
17563 section in the output file. This allows the literal pool to be placed
17564 in a data RAM/ROM, and it also allows the linker to combine literal
17565 pools from separate object files to remove redundant literals and
17566 improve code size. With @option{-mtext-section-literals}, the literals
17567 are interspersed in the text section in order to keep them as close as
17568 possible to their references. This may be necessary for large assembly
17571 @item -mtarget-align
17572 @itemx -mno-target-align
17573 @opindex mtarget-align
17574 @opindex mno-target-align
17575 When this option is enabled, GCC instructs the assembler to
17576 automatically align instructions to reduce branch penalties at the
17577 expense of some code density. The assembler attempts to widen density
17578 instructions to align branch targets and the instructions following call
17579 instructions. If there are not enough preceding safe density
17580 instructions to align a target, no widening will be performed. The
17581 default is @option{-mtarget-align}. These options do not affect the
17582 treatment of auto-aligned instructions like @code{LOOP}, which the
17583 assembler will always align, either by widening density instructions or
17584 by inserting no-op instructions.
17587 @itemx -mno-longcalls
17588 @opindex mlongcalls
17589 @opindex mno-longcalls
17590 When this option is enabled, GCC instructs the assembler to translate
17591 direct calls to indirect calls unless it can determine that the target
17592 of a direct call is in the range allowed by the call instruction. This
17593 translation typically occurs for calls to functions in other source
17594 files. Specifically, the assembler translates a direct @code{CALL}
17595 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
17596 The default is @option{-mno-longcalls}. This option should be used in
17597 programs where the call target can potentially be out of range. This
17598 option is implemented in the assembler, not the compiler, so the
17599 assembly code generated by GCC will still show direct call
17600 instructions---look at the disassembled object code to see the actual
17601 instructions. Note that the assembler will use an indirect call for
17602 every cross-file call, not just those that really will be out of range.
17605 @node zSeries Options
17606 @subsection zSeries Options
17607 @cindex zSeries options
17609 These are listed under @xref{S/390 and zSeries Options}.
17611 @node Code Gen Options
17612 @section Options for Code Generation Conventions
17613 @cindex code generation conventions
17614 @cindex options, code generation
17615 @cindex run-time options
17617 These machine-independent options control the interface conventions
17618 used in code generation.
17620 Most of them have both positive and negative forms; the negative form
17621 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
17622 one of the forms is listed---the one which is not the default. You
17623 can figure out the other form by either removing @samp{no-} or adding
17627 @item -fbounds-check
17628 @opindex fbounds-check
17629 For front-ends that support it, generate additional code to check that
17630 indices used to access arrays are within the declared range. This is
17631 currently only supported by the Java and Fortran front-ends, where
17632 this option defaults to true and false respectively.
17636 This option generates traps for signed overflow on addition, subtraction,
17637 multiplication operations.
17641 This option instructs the compiler to assume that signed arithmetic
17642 overflow of addition, subtraction and multiplication wraps around
17643 using twos-complement representation. This flag enables some optimizations
17644 and disables others. This option is enabled by default for the Java
17645 front-end, as required by the Java language specification.
17648 @opindex fexceptions
17649 Enable exception handling. Generates extra code needed to propagate
17650 exceptions. For some targets, this implies GCC will generate frame
17651 unwind information for all functions, which can produce significant data
17652 size overhead, although it does not affect execution. If you do not
17653 specify this option, GCC will enable it by default for languages like
17654 C++ which normally require exception handling, and disable it for
17655 languages like C that do not normally require it. However, you may need
17656 to enable this option when compiling C code that needs to interoperate
17657 properly with exception handlers written in C++. You may also wish to
17658 disable this option if you are compiling older C++ programs that don't
17659 use exception handling.
17661 @item -fnon-call-exceptions
17662 @opindex fnon-call-exceptions
17663 Generate code that allows trapping instructions to throw exceptions.
17664 Note that this requires platform-specific runtime support that does
17665 not exist everywhere. Moreover, it only allows @emph{trapping}
17666 instructions to throw exceptions, i.e.@: memory references or floating
17667 point instructions. It does not allow exceptions to be thrown from
17668 arbitrary signal handlers such as @code{SIGALRM}.
17670 @item -funwind-tables
17671 @opindex funwind-tables
17672 Similar to @option{-fexceptions}, except that it will just generate any needed
17673 static data, but will not affect the generated code in any other way.
17674 You will normally not enable this option; instead, a language processor
17675 that needs this handling would enable it on your behalf.
17677 @item -fasynchronous-unwind-tables
17678 @opindex fasynchronous-unwind-tables
17679 Generate unwind table in dwarf2 format, if supported by target machine. The
17680 table is exact at each instruction boundary, so it can be used for stack
17681 unwinding from asynchronous events (such as debugger or garbage collector).
17683 @item -fpcc-struct-return
17684 @opindex fpcc-struct-return
17685 Return ``short'' @code{struct} and @code{union} values in memory like
17686 longer ones, rather than in registers. This convention is less
17687 efficient, but it has the advantage of allowing intercallability between
17688 GCC-compiled files and files compiled with other compilers, particularly
17689 the Portable C Compiler (pcc).
17691 The precise convention for returning structures in memory depends
17692 on the target configuration macros.
17694 Short structures and unions are those whose size and alignment match
17695 that of some integer type.
17697 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
17698 switch is not binary compatible with code compiled with the
17699 @option{-freg-struct-return} switch.
17700 Use it to conform to a non-default application binary interface.
17702 @item -freg-struct-return
17703 @opindex freg-struct-return
17704 Return @code{struct} and @code{union} values in registers when possible.
17705 This is more efficient for small structures than
17706 @option{-fpcc-struct-return}.
17708 If you specify neither @option{-fpcc-struct-return} nor
17709 @option{-freg-struct-return}, GCC defaults to whichever convention is
17710 standard for the target. If there is no standard convention, GCC
17711 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
17712 the principal compiler. In those cases, we can choose the standard, and
17713 we chose the more efficient register return alternative.
17715 @strong{Warning:} code compiled with the @option{-freg-struct-return}
17716 switch is not binary compatible with code compiled with the
17717 @option{-fpcc-struct-return} switch.
17718 Use it to conform to a non-default application binary interface.
17720 @item -fshort-enums
17721 @opindex fshort-enums
17722 Allocate to an @code{enum} type only as many bytes as it needs for the
17723 declared range of possible values. Specifically, the @code{enum} type
17724 will be equivalent to the smallest integer type which has enough room.
17726 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
17727 code that is not binary compatible with code generated without that switch.
17728 Use it to conform to a non-default application binary interface.
17730 @item -fshort-double
17731 @opindex fshort-double
17732 Use the same size for @code{double} as for @code{float}.
17734 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
17735 code that is not binary compatible with code generated without that switch.
17736 Use it to conform to a non-default application binary interface.
17738 @item -fshort-wchar
17739 @opindex fshort-wchar
17740 Override the underlying type for @samp{wchar_t} to be @samp{short
17741 unsigned int} instead of the default for the target. This option is
17742 useful for building programs to run under WINE@.
17744 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
17745 code that is not binary compatible with code generated without that switch.
17746 Use it to conform to a non-default application binary interface.
17749 @opindex fno-common
17750 In C code, controls the placement of uninitialized global variables.
17751 Unix C compilers have traditionally permitted multiple definitions of
17752 such variables in different compilation units by placing the variables
17754 This is the behavior specified by @option{-fcommon}, and is the default
17755 for GCC on most targets.
17756 On the other hand, this behavior is not required by ISO C, and on some
17757 targets may carry a speed or code size penalty on variable references.
17758 The @option{-fno-common} option specifies that the compiler should place
17759 uninitialized global variables in the data section of the object file,
17760 rather than generating them as common blocks.
17761 This has the effect that if the same variable is declared
17762 (without @code{extern}) in two different compilations,
17763 you will get a multiple-definition error when you link them.
17764 In this case, you must compile with @option{-fcommon} instead.
17765 Compiling with @option{-fno-common} is useful on targets for which
17766 it provides better performance, or if you wish to verify that the
17767 program will work on other systems which always treat uninitialized
17768 variable declarations this way.
17772 Ignore the @samp{#ident} directive.
17774 @item -finhibit-size-directive
17775 @opindex finhibit-size-directive
17776 Don't output a @code{.size} assembler directive, or anything else that
17777 would cause trouble if the function is split in the middle, and the
17778 two halves are placed at locations far apart in memory. This option is
17779 used when compiling @file{crtstuff.c}; you should not need to use it
17782 @item -fverbose-asm
17783 @opindex fverbose-asm
17784 Put extra commentary information in the generated assembly code to
17785 make it more readable. This option is generally only of use to those
17786 who actually need to read the generated assembly code (perhaps while
17787 debugging the compiler itself).
17789 @option{-fno-verbose-asm}, the default, causes the
17790 extra information to be omitted and is useful when comparing two assembler
17793 @item -frecord-gcc-switches
17794 @opindex frecord-gcc-switches
17795 This switch causes the command line that was used to invoke the
17796 compiler to be recorded into the object file that is being created.
17797 This switch is only implemented on some targets and the exact format
17798 of the recording is target and binary file format dependent, but it
17799 usually takes the form of a section containing ASCII text. This
17800 switch is related to the @option{-fverbose-asm} switch, but that
17801 switch only records information in the assembler output file as
17802 comments, so it never reaches the object file.
17806 @cindex global offset table
17808 Generate position-independent code (PIC) suitable for use in a shared
17809 library, if supported for the target machine. Such code accesses all
17810 constant addresses through a global offset table (GOT)@. The dynamic
17811 loader resolves the GOT entries when the program starts (the dynamic
17812 loader is not part of GCC; it is part of the operating system). If
17813 the GOT size for the linked executable exceeds a machine-specific
17814 maximum size, you get an error message from the linker indicating that
17815 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
17816 instead. (These maximums are 8k on the SPARC and 32k
17817 on the m68k and RS/6000. The 386 has no such limit.)
17819 Position-independent code requires special support, and therefore works
17820 only on certain machines. For the 386, GCC supports PIC for System V
17821 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
17822 position-independent.
17824 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
17829 If supported for the target machine, emit position-independent code,
17830 suitable for dynamic linking and avoiding any limit on the size of the
17831 global offset table. This option makes a difference on the m68k,
17832 PowerPC and SPARC@.
17834 Position-independent code requires special support, and therefore works
17835 only on certain machines.
17837 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
17844 These options are similar to @option{-fpic} and @option{-fPIC}, but
17845 generated position independent code can be only linked into executables.
17846 Usually these options are used when @option{-pie} GCC option will be
17847 used during linking.
17849 @option{-fpie} and @option{-fPIE} both define the macros
17850 @code{__pie__} and @code{__PIE__}. The macros have the value 1
17851 for @option{-fpie} and 2 for @option{-fPIE}.
17853 @item -fno-jump-tables
17854 @opindex fno-jump-tables
17855 Do not use jump tables for switch statements even where it would be
17856 more efficient than other code generation strategies. This option is
17857 of use in conjunction with @option{-fpic} or @option{-fPIC} for
17858 building code which forms part of a dynamic linker and cannot
17859 reference the address of a jump table. On some targets, jump tables
17860 do not require a GOT and this option is not needed.
17862 @item -ffixed-@var{reg}
17864 Treat the register named @var{reg} as a fixed register; generated code
17865 should never refer to it (except perhaps as a stack pointer, frame
17866 pointer or in some other fixed role).
17868 @var{reg} must be the name of a register. The register names accepted
17869 are machine-specific and are defined in the @code{REGISTER_NAMES}
17870 macro in the machine description macro file.
17872 This flag does not have a negative form, because it specifies a
17875 @item -fcall-used-@var{reg}
17876 @opindex fcall-used
17877 Treat the register named @var{reg} as an allocable register that is
17878 clobbered by function calls. It may be allocated for temporaries or
17879 variables that do not live across a call. Functions compiled this way
17880 will not save and restore the register @var{reg}.
17882 It is an error to used this flag with the frame pointer or stack pointer.
17883 Use of this flag for other registers that have fixed pervasive roles in
17884 the machine's execution model will produce disastrous results.
17886 This flag does not have a negative form, because it specifies a
17889 @item -fcall-saved-@var{reg}
17890 @opindex fcall-saved
17891 Treat the register named @var{reg} as an allocable register saved by
17892 functions. It may be allocated even for temporaries or variables that
17893 live across a call. Functions compiled this way will save and restore
17894 the register @var{reg} if they use it.
17896 It is an error to used this flag with the frame pointer or stack pointer.
17897 Use of this flag for other registers that have fixed pervasive roles in
17898 the machine's execution model will produce disastrous results.
17900 A different sort of disaster will result from the use of this flag for
17901 a register in which function values may be returned.
17903 This flag does not have a negative form, because it specifies a
17906 @item -fpack-struct[=@var{n}]
17907 @opindex fpack-struct
17908 Without a value specified, pack all structure members together without
17909 holes. When a value is specified (which must be a small power of two), pack
17910 structure members according to this value, representing the maximum
17911 alignment (that is, objects with default alignment requirements larger than
17912 this will be output potentially unaligned at the next fitting location.
17914 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
17915 code that is not binary compatible with code generated without that switch.
17916 Additionally, it makes the code suboptimal.
17917 Use it to conform to a non-default application binary interface.
17919 @item -finstrument-functions
17920 @opindex finstrument-functions
17921 Generate instrumentation calls for entry and exit to functions. Just
17922 after function entry and just before function exit, the following
17923 profiling functions will be called with the address of the current
17924 function and its call site. (On some platforms,
17925 @code{__builtin_return_address} does not work beyond the current
17926 function, so the call site information may not be available to the
17927 profiling functions otherwise.)
17930 void __cyg_profile_func_enter (void *this_fn,
17932 void __cyg_profile_func_exit (void *this_fn,
17936 The first argument is the address of the start of the current function,
17937 which may be looked up exactly in the symbol table.
17939 This instrumentation is also done for functions expanded inline in other
17940 functions. The profiling calls will indicate where, conceptually, the
17941 inline function is entered and exited. This means that addressable
17942 versions of such functions must be available. If all your uses of a
17943 function are expanded inline, this may mean an additional expansion of
17944 code size. If you use @samp{extern inline} in your C code, an
17945 addressable version of such functions must be provided. (This is
17946 normally the case anyways, but if you get lucky and the optimizer always
17947 expands the functions inline, you might have gotten away without
17948 providing static copies.)
17950 A function may be given the attribute @code{no_instrument_function}, in
17951 which case this instrumentation will not be done. This can be used, for
17952 example, for the profiling functions listed above, high-priority
17953 interrupt routines, and any functions from which the profiling functions
17954 cannot safely be called (perhaps signal handlers, if the profiling
17955 routines generate output or allocate memory).
17957 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
17958 @opindex finstrument-functions-exclude-file-list
17960 Set the list of functions that are excluded from instrumentation (see
17961 the description of @code{-finstrument-functions}). If the file that
17962 contains a function definition matches with one of @var{file}, then
17963 that function is not instrumented. The match is done on substrings:
17964 if the @var{file} parameter is a substring of the file name, it is
17965 considered to be a match.
17970 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
17974 will exclude any inline function defined in files whose pathnames
17975 contain @code{/bits/stl} or @code{include/sys}.
17977 If, for some reason, you want to include letter @code{','} in one of
17978 @var{sym}, write @code{'\,'}. For example,
17979 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
17980 (note the single quote surrounding the option).
17982 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
17983 @opindex finstrument-functions-exclude-function-list
17985 This is similar to @code{-finstrument-functions-exclude-file-list},
17986 but this option sets the list of function names to be excluded from
17987 instrumentation. The function name to be matched is its user-visible
17988 name, such as @code{vector<int> blah(const vector<int> &)}, not the
17989 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
17990 match is done on substrings: if the @var{sym} parameter is a substring
17991 of the function name, it is considered to be a match. For C99 and C++
17992 extended identifiers, the function name must be given in UTF-8, not
17993 using universal character names.
17995 @item -fstack-check
17996 @opindex fstack-check
17997 Generate code to verify that you do not go beyond the boundary of the
17998 stack. You should specify this flag if you are running in an
17999 environment with multiple threads, but only rarely need to specify it in
18000 a single-threaded environment since stack overflow is automatically
18001 detected on nearly all systems if there is only one stack.
18003 Note that this switch does not actually cause checking to be done; the
18004 operating system or the language runtime must do that. The switch causes
18005 generation of code to ensure that they see the stack being extended.
18007 You can additionally specify a string parameter: @code{no} means no
18008 checking, @code{generic} means force the use of old-style checking,
18009 @code{specific} means use the best checking method and is equivalent
18010 to bare @option{-fstack-check}.
18012 Old-style checking is a generic mechanism that requires no specific
18013 target support in the compiler but comes with the following drawbacks:
18017 Modified allocation strategy for large objects: they will always be
18018 allocated dynamically if their size exceeds a fixed threshold.
18021 Fixed limit on the size of the static frame of functions: when it is
18022 topped by a particular function, stack checking is not reliable and
18023 a warning is issued by the compiler.
18026 Inefficiency: because of both the modified allocation strategy and the
18027 generic implementation, the performances of the code are hampered.
18030 Note that old-style stack checking is also the fallback method for
18031 @code{specific} if no target support has been added in the compiler.
18033 @item -fstack-limit-register=@var{reg}
18034 @itemx -fstack-limit-symbol=@var{sym}
18035 @itemx -fno-stack-limit
18036 @opindex fstack-limit-register
18037 @opindex fstack-limit-symbol
18038 @opindex fno-stack-limit
18039 Generate code to ensure that the stack does not grow beyond a certain value,
18040 either the value of a register or the address of a symbol. If the stack
18041 would grow beyond the value, a signal is raised. For most targets,
18042 the signal is raised before the stack overruns the boundary, so
18043 it is possible to catch the signal without taking special precautions.
18045 For instance, if the stack starts at absolute address @samp{0x80000000}
18046 and grows downwards, you can use the flags
18047 @option{-fstack-limit-symbol=__stack_limit} and
18048 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
18049 of 128KB@. Note that this may only work with the GNU linker.
18051 @item -fsplit-stack
18052 @opindex fsplit-stack
18053 Generate code to automatically split the stack before it overflows.
18054 The resulting program has a discontiguous stack which can only
18055 overflow if the program is unable to allocate any more memory. This
18056 is most useful when running threaded programs, as it is no longer
18057 necessary to calculate a good stack size to use for each thread. This
18058 is currently only implemented for the i386 and x86_64 backends running
18061 When code compiled with @option{-fsplit-stack} calls code compiled
18062 without @option{-fsplit-stack}, there may not be much stack space
18063 available for the latter code to run. If compiling all code,
18064 including library code, with @option{-fsplit-stack} is not an option,
18065 then the linker can fix up these calls so that the code compiled
18066 without @option{-fsplit-stack} always has a large stack. Support for
18067 this is implemented in the gold linker in GNU binutils release 2.21
18070 @item -fleading-underscore
18071 @opindex fleading-underscore
18072 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
18073 change the way C symbols are represented in the object file. One use
18074 is to help link with legacy assembly code.
18076 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
18077 generate code that is not binary compatible with code generated without that
18078 switch. Use it to conform to a non-default application binary interface.
18079 Not all targets provide complete support for this switch.
18081 @item -ftls-model=@var{model}
18082 @opindex ftls-model
18083 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
18084 The @var{model} argument should be one of @code{global-dynamic},
18085 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
18087 The default without @option{-fpic} is @code{initial-exec}; with
18088 @option{-fpic} the default is @code{global-dynamic}.
18090 @item -fvisibility=@var{default|internal|hidden|protected}
18091 @opindex fvisibility
18092 Set the default ELF image symbol visibility to the specified option---all
18093 symbols will be marked with this unless overridden within the code.
18094 Using this feature can very substantially improve linking and
18095 load times of shared object libraries, produce more optimized
18096 code, provide near-perfect API export and prevent symbol clashes.
18097 It is @strong{strongly} recommended that you use this in any shared objects
18100 Despite the nomenclature, @code{default} always means public ie;
18101 available to be linked against from outside the shared object.
18102 @code{protected} and @code{internal} are pretty useless in real-world
18103 usage so the only other commonly used option will be @code{hidden}.
18104 The default if @option{-fvisibility} isn't specified is
18105 @code{default}, i.e., make every
18106 symbol public---this causes the same behavior as previous versions of
18109 A good explanation of the benefits offered by ensuring ELF
18110 symbols have the correct visibility is given by ``How To Write
18111 Shared Libraries'' by Ulrich Drepper (which can be found at
18112 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
18113 solution made possible by this option to marking things hidden when
18114 the default is public is to make the default hidden and mark things
18115 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
18116 and @code{__attribute__ ((visibility("default")))} instead of
18117 @code{__declspec(dllexport)} you get almost identical semantics with
18118 identical syntax. This is a great boon to those working with
18119 cross-platform projects.
18121 For those adding visibility support to existing code, you may find
18122 @samp{#pragma GCC visibility} of use. This works by you enclosing
18123 the declarations you wish to set visibility for with (for example)
18124 @samp{#pragma GCC visibility push(hidden)} and
18125 @samp{#pragma GCC visibility pop}.
18126 Bear in mind that symbol visibility should be viewed @strong{as
18127 part of the API interface contract} and thus all new code should
18128 always specify visibility when it is not the default ie; declarations
18129 only for use within the local DSO should @strong{always} be marked explicitly
18130 as hidden as so to avoid PLT indirection overheads---making this
18131 abundantly clear also aids readability and self-documentation of the code.
18132 Note that due to ISO C++ specification requirements, operator new and
18133 operator delete must always be of default visibility.
18135 Be aware that headers from outside your project, in particular system
18136 headers and headers from any other library you use, may not be
18137 expecting to be compiled with visibility other than the default. You
18138 may need to explicitly say @samp{#pragma GCC visibility push(default)}
18139 before including any such headers.
18141 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
18142 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
18143 no modifications. However, this means that calls to @samp{extern}
18144 functions with no explicit visibility will use the PLT, so it is more
18145 effective to use @samp{__attribute ((visibility))} and/or
18146 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
18147 declarations should be treated as hidden.
18149 Note that @samp{-fvisibility} does affect C++ vague linkage
18150 entities. This means that, for instance, an exception class that will
18151 be thrown between DSOs must be explicitly marked with default
18152 visibility so that the @samp{type_info} nodes will be unified between
18155 An overview of these techniques, their benefits and how to use them
18156 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
18158 @item -fstrict-volatile-bitfields
18159 @opindex fstrict-volatile-bitfields
18160 This option should be used if accesses to volatile bitfields (or other
18161 structure fields, although the compiler usually honors those types
18162 anyway) should use a single access of the width of the
18163 field's type, aligned to a natural alignment if possible. For
18164 example, targets with memory-mapped peripheral registers might require
18165 all such accesses to be 16 bits wide; with this flag the user could
18166 declare all peripheral bitfields as ``unsigned short'' (assuming short
18167 is 16 bits on these targets) to force GCC to use 16 bit accesses
18168 instead of, perhaps, a more efficient 32 bit access.
18170 If this option is disabled, the compiler will use the most efficient
18171 instruction. In the previous example, that might be a 32-bit load
18172 instruction, even though that will access bytes that do not contain
18173 any portion of the bitfield, or memory-mapped registers unrelated to
18174 the one being updated.
18176 If the target requires strict alignment, and honoring the field
18177 type would require violating this alignment, a warning is issued.
18178 If the field has @code{packed} attribute, the access is done without
18179 honoring the field type. If the field doesn't have @code{packed}
18180 attribute, the access is done honoring the field type. In both cases,
18181 GCC assumes that the user knows something about the target hardware
18182 that it is unaware of.
18184 The default value of this option is determined by the application binary
18185 interface for the target processor.
18191 @node Environment Variables
18192 @section Environment Variables Affecting GCC
18193 @cindex environment variables
18195 @c man begin ENVIRONMENT
18196 This section describes several environment variables that affect how GCC
18197 operates. Some of them work by specifying directories or prefixes to use
18198 when searching for various kinds of files. Some are used to specify other
18199 aspects of the compilation environment.
18201 Note that you can also specify places to search using options such as
18202 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
18203 take precedence over places specified using environment variables, which
18204 in turn take precedence over those specified by the configuration of GCC@.
18205 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
18206 GNU Compiler Collection (GCC) Internals}.
18211 @c @itemx LC_COLLATE
18213 @c @itemx LC_MONETARY
18214 @c @itemx LC_NUMERIC
18219 @c @findex LC_COLLATE
18220 @findex LC_MESSAGES
18221 @c @findex LC_MONETARY
18222 @c @findex LC_NUMERIC
18226 These environment variables control the way that GCC uses
18227 localization information that allow GCC to work with different
18228 national conventions. GCC inspects the locale categories
18229 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
18230 so. These locale categories can be set to any value supported by your
18231 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
18232 Kingdom encoded in UTF-8.
18234 The @env{LC_CTYPE} environment variable specifies character
18235 classification. GCC uses it to determine the character boundaries in
18236 a string; this is needed for some multibyte encodings that contain quote
18237 and escape characters that would otherwise be interpreted as a string
18240 The @env{LC_MESSAGES} environment variable specifies the language to
18241 use in diagnostic messages.
18243 If the @env{LC_ALL} environment variable is set, it overrides the value
18244 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
18245 and @env{LC_MESSAGES} default to the value of the @env{LANG}
18246 environment variable. If none of these variables are set, GCC
18247 defaults to traditional C English behavior.
18251 If @env{TMPDIR} is set, it specifies the directory to use for temporary
18252 files. GCC uses temporary files to hold the output of one stage of
18253 compilation which is to be used as input to the next stage: for example,
18254 the output of the preprocessor, which is the input to the compiler
18257 @item GCC_EXEC_PREFIX
18258 @findex GCC_EXEC_PREFIX
18259 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
18260 names of the subprograms executed by the compiler. No slash is added
18261 when this prefix is combined with the name of a subprogram, but you can
18262 specify a prefix that ends with a slash if you wish.
18264 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
18265 an appropriate prefix to use based on the pathname it was invoked with.
18267 If GCC cannot find the subprogram using the specified prefix, it
18268 tries looking in the usual places for the subprogram.
18270 The default value of @env{GCC_EXEC_PREFIX} is
18271 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
18272 the installed compiler. In many cases @var{prefix} is the value
18273 of @code{prefix} when you ran the @file{configure} script.
18275 Other prefixes specified with @option{-B} take precedence over this prefix.
18277 This prefix is also used for finding files such as @file{crt0.o} that are
18280 In addition, the prefix is used in an unusual way in finding the
18281 directories to search for header files. For each of the standard
18282 directories whose name normally begins with @samp{/usr/local/lib/gcc}
18283 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
18284 replacing that beginning with the specified prefix to produce an
18285 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
18286 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
18287 These alternate directories are searched first; the standard directories
18288 come next. If a standard directory begins with the configured
18289 @var{prefix} then the value of @var{prefix} is replaced by
18290 @env{GCC_EXEC_PREFIX} when looking for header files.
18292 @item COMPILER_PATH
18293 @findex COMPILER_PATH
18294 The value of @env{COMPILER_PATH} is a colon-separated list of
18295 directories, much like @env{PATH}. GCC tries the directories thus
18296 specified when searching for subprograms, if it can't find the
18297 subprograms using @env{GCC_EXEC_PREFIX}.
18300 @findex LIBRARY_PATH
18301 The value of @env{LIBRARY_PATH} is a colon-separated list of
18302 directories, much like @env{PATH}. When configured as a native compiler,
18303 GCC tries the directories thus specified when searching for special
18304 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
18305 using GCC also uses these directories when searching for ordinary
18306 libraries for the @option{-l} option (but directories specified with
18307 @option{-L} come first).
18311 @cindex locale definition
18312 This variable is used to pass locale information to the compiler. One way in
18313 which this information is used is to determine the character set to be used
18314 when character literals, string literals and comments are parsed in C and C++.
18315 When the compiler is configured to allow multibyte characters,
18316 the following values for @env{LANG} are recognized:
18320 Recognize JIS characters.
18322 Recognize SJIS characters.
18324 Recognize EUCJP characters.
18327 If @env{LANG} is not defined, or if it has some other value, then the
18328 compiler will use mblen and mbtowc as defined by the default locale to
18329 recognize and translate multibyte characters.
18333 Some additional environments variables affect the behavior of the
18336 @include cppenv.texi
18340 @node Precompiled Headers
18341 @section Using Precompiled Headers
18342 @cindex precompiled headers
18343 @cindex speed of compilation
18345 Often large projects have many header files that are included in every
18346 source file. The time the compiler takes to process these header files
18347 over and over again can account for nearly all of the time required to
18348 build the project. To make builds faster, GCC allows users to
18349 `precompile' a header file; then, if builds can use the precompiled
18350 header file they will be much faster.
18352 To create a precompiled header file, simply compile it as you would any
18353 other file, if necessary using the @option{-x} option to make the driver
18354 treat it as a C or C++ header file. You will probably want to use a
18355 tool like @command{make} to keep the precompiled header up-to-date when
18356 the headers it contains change.
18358 A precompiled header file will be searched for when @code{#include} is
18359 seen in the compilation. As it searches for the included file
18360 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
18361 compiler looks for a precompiled header in each directory just before it
18362 looks for the include file in that directory. The name searched for is
18363 the name specified in the @code{#include} with @samp{.gch} appended. If
18364 the precompiled header file can't be used, it is ignored.
18366 For instance, if you have @code{#include "all.h"}, and you have
18367 @file{all.h.gch} in the same directory as @file{all.h}, then the
18368 precompiled header file will be used if possible, and the original
18369 header will be used otherwise.
18371 Alternatively, you might decide to put the precompiled header file in a
18372 directory and use @option{-I} to ensure that directory is searched
18373 before (or instead of) the directory containing the original header.
18374 Then, if you want to check that the precompiled header file is always
18375 used, you can put a file of the same name as the original header in this
18376 directory containing an @code{#error} command.
18378 This also works with @option{-include}. So yet another way to use
18379 precompiled headers, good for projects not designed with precompiled
18380 header files in mind, is to simply take most of the header files used by
18381 a project, include them from another header file, precompile that header
18382 file, and @option{-include} the precompiled header. If the header files
18383 have guards against multiple inclusion, they will be skipped because
18384 they've already been included (in the precompiled header).
18386 If you need to precompile the same header file for different
18387 languages, targets, or compiler options, you can instead make a
18388 @emph{directory} named like @file{all.h.gch}, and put each precompiled
18389 header in the directory, perhaps using @option{-o}. It doesn't matter
18390 what you call the files in the directory, every precompiled header in
18391 the directory will be considered. The first precompiled header
18392 encountered in the directory that is valid for this compilation will
18393 be used; they're searched in no particular order.
18395 There are many other possibilities, limited only by your imagination,
18396 good sense, and the constraints of your build system.
18398 A precompiled header file can be used only when these conditions apply:
18402 Only one precompiled header can be used in a particular compilation.
18405 A precompiled header can't be used once the first C token is seen. You
18406 can have preprocessor directives before a precompiled header; you can
18407 even include a precompiled header from inside another header, so long as
18408 there are no C tokens before the @code{#include}.
18411 The precompiled header file must be produced for the same language as
18412 the current compilation. You can't use a C precompiled header for a C++
18416 The precompiled header file must have been produced by the same compiler
18417 binary as the current compilation is using.
18420 Any macros defined before the precompiled header is included must
18421 either be defined in the same way as when the precompiled header was
18422 generated, or must not affect the precompiled header, which usually
18423 means that they don't appear in the precompiled header at all.
18425 The @option{-D} option is one way to define a macro before a
18426 precompiled header is included; using a @code{#define} can also do it.
18427 There are also some options that define macros implicitly, like
18428 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
18431 @item If debugging information is output when using the precompiled
18432 header, using @option{-g} or similar, the same kind of debugging information
18433 must have been output when building the precompiled header. However,
18434 a precompiled header built using @option{-g} can be used in a compilation
18435 when no debugging information is being output.
18437 @item The same @option{-m} options must generally be used when building
18438 and using the precompiled header. @xref{Submodel Options},
18439 for any cases where this rule is relaxed.
18441 @item Each of the following options must be the same when building and using
18442 the precompiled header:
18444 @gccoptlist{-fexceptions}
18447 Some other command-line options starting with @option{-f},
18448 @option{-p}, or @option{-O} must be defined in the same way as when
18449 the precompiled header was generated. At present, it's not clear
18450 which options are safe to change and which are not; the safest choice
18451 is to use exactly the same options when generating and using the
18452 precompiled header. The following are known to be safe:
18454 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
18455 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
18456 -fsched-verbose=<number> -fschedule-insns -fvisibility= @gol
18461 For all of these except the last, the compiler will automatically
18462 ignore the precompiled header if the conditions aren't met. If you
18463 find an option combination that doesn't work and doesn't cause the
18464 precompiled header to be ignored, please consider filing a bug report,
18467 If you do use differing options when generating and using the
18468 precompiled header, the actual behavior will be a mixture of the
18469 behavior for the options. For instance, if you use @option{-g} to
18470 generate the precompiled header but not when using it, you may or may
18471 not get debugging information for routines in the precompiled header.