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{]}} -fdump-go-spec=@var{file}
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 fmax-errors=@var{n} -pedantic @gol
233 -pedantic-errors @gol
234 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
235 -Wno-attributes -Wno-builtin-macro-redefined @gol
236 -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual @gol
237 -Wchar-subscripts -Wclobbered -Wcomment @gol
238 -Wconversion -Wcoverage-mismatch -Wcpp -Wno-deprecated @gol
239 -Wno-deprecated-declarations -Wdisabled-optimization @gol
240 -Wno-div-by-zero -Wdouble-promotion -Wempty-body -Wenum-compare @gol
241 -Wno-endif-labels -Werror -Werror=* @gol
242 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
243 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
244 -Wformat-security -Wformat-y2k @gol
245 -Wframe-larger-than=@var{len} -Wjump-misses-init -Wignored-qualifiers @gol
246 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
247 -Winit-self -Winline @gol
248 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
249 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
250 -Wlogical-op -Wlong-long @gol
251 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
252 -Wmissing-format-attribute -Wmissing-include-dirs @gol
254 -Wno-multichar -Wnonnull -Wno-overflow @gol
255 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
256 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
257 -Wpointer-arith -Wno-pointer-to-int-cast @gol
258 -Wredundant-decls @gol
259 -Wreturn-type -Wsequence-point -Wshadow @gol
260 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
261 -Wstrict-aliasing -Wstrict-aliasing=n @gol
262 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
263 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]} @gol
264 -Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand @gol
265 -Wsystem-headers -Wtrampolines -Wtrigraphs -Wtype-limits -Wundef @gol
266 -Wuninitialized -Wunknown-pragmas -Wno-pragmas @gol
267 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
268 -Wunused-label -Wunused-parameter -Wno-unused-result -Wunused-value @gol
269 -Wunused-variable -Wunused-but-set-parameter -Wunused-but-set-variable @gol
270 -Wvariadic-macros -Wvla -Wvolatile-register-var -Wwrite-strings}
272 @item C and Objective-C-only Warning Options
273 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
274 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
275 -Wold-style-declaration -Wold-style-definition @gol
276 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
277 -Wdeclaration-after-statement -Wpointer-sign}
279 @item Debugging Options
280 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
281 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
282 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
283 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
284 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
285 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
286 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
287 -fdump-statistics @gol
289 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
290 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
291 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
293 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
294 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
295 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
296 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
297 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
298 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
299 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
300 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
301 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
302 -fdump-tree-nrv -fdump-tree-vect @gol
303 -fdump-tree-sink @gol
304 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
305 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
306 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
307 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
308 -ftree-vectorizer-verbose=@var{n} @gol
309 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
310 -fdump-final-insns=@var{file} @gol
311 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
312 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
313 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
314 -fenable-icf-debug @gol
315 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
316 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
317 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
318 -fstack-usage -ftest-coverage -ftime-report -fvar-tracking @gol
319 -fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol
320 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
321 -ggdb -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
322 -gvms -gxcoff -gxcoff+ @gol
323 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
324 -fdebug-prefix-map=@var{old}=@var{new} @gol
325 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
326 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
327 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
328 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
329 -print-prog-name=@var{program} -print-search-dirs -Q @gol
330 -print-sysroot -print-sysroot-headers-suffix @gol
331 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
333 @item Optimization Options
334 @xref{Optimize Options,,Options that Control Optimization}.
335 @gccoptlist{-falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
336 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
337 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
338 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
339 -fcheck-data-deps -fcombine-stack-adjustments -fconserve-stack @gol
340 -fcprop-registers -fcrossjumping @gol
341 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
342 -fcx-limited-range @gol
343 -fdata-sections -fdce -fdce @gol
344 -fdelayed-branch -fdelete-null-pointer-checks -fdse -fdse @gol
345 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffast-math @gol
346 -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
347 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
348 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
349 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
350 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
351 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg @gol
352 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference @gol
353 -fipa-struct-reorg -fira-algorithm=@var{algorithm} @gol
354 -fira-region=@var{region} @gol
355 -fira-loop-pressure -fno-ira-share-save-slots @gol
356 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
357 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
358 -floop-block -floop-flatten -floop-interchange -floop-strip-mine @gol
359 -floop-parallelize-all -flto -flto-compression-level
360 -flto-partition=@var{alg} -flto-report -fmerge-all-constants @gol
361 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
362 -fmove-loop-invariants fmudflap -fmudflapir -fmudflapth -fno-branch-count-reg @gol
363 -fno-default-inline @gol
364 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
365 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
366 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
367 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
368 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
369 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
370 -fprefetch-loop-arrays @gol
371 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
372 -fprofile-generate=@var{path} @gol
373 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
374 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
375 -freorder-blocks-and-partition -freorder-functions @gol
376 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
377 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
378 -fsched-spec-load -fsched-spec-load-dangerous @gol
379 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
380 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
381 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
382 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
383 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
384 -fselective-scheduling -fselective-scheduling2 @gol
385 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
386 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
387 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
388 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer @gol
390 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
391 -ftree-copyrename -ftree-dce -ftree-dominator-opts -ftree-dse @gol
392 -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
393 -ftree-loop-if-convert-memory-writes -ftree-loop-im @gol
394 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
395 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
396 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-pta -ftree-reassoc @gol
397 -ftree-sink -ftree-sra -ftree-switch-conversion @gol
398 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
399 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
400 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
401 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
402 -fwhole-program -fwpa -fuse-linker-plugin @gol
403 --param @var{name}=@var{value}
404 -O -O0 -O1 -O2 -O3 -Os -Ofast}
406 @item Preprocessor Options
407 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
408 @gccoptlist{-A@var{question}=@var{answer} @gol
409 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
410 -C -dD -dI -dM -dN @gol
411 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
412 -idirafter @var{dir} @gol
413 -include @var{file} -imacros @var{file} @gol
414 -iprefix @var{file} -iwithprefix @var{dir} @gol
415 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
416 -imultilib @var{dir} -isysroot @var{dir} @gol
417 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
418 -P -fworking-directory -remap @gol
419 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
420 -Xpreprocessor @var{option}}
422 @item Assembler Option
423 @xref{Assembler Options,,Passing Options to the Assembler}.
424 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
427 @xref{Link Options,,Options for Linking}.
428 @gccoptlist{@var{object-file-name} -l@var{library} @gol
429 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
430 -s -static -static-libgcc -static-libstdc++ -shared @gol
431 -shared-libgcc -symbolic @gol
432 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
435 @item Directory Options
436 @xref{Directory Options,,Options for Directory Search}.
437 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir}}
438 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I-
441 @item Machine Dependent Options
442 @xref{Submodel Options,,Hardware Models and Configurations}.
443 @c This list is ordered alphanumerically by subsection name.
444 @c Try and put the significant identifier (CPU or system) first,
445 @c so users have a clue at guessing where the ones they want will be.
448 @gccoptlist{-EB -EL @gol
449 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
450 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
453 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
454 -mabi=@var{name} @gol
455 -mapcs-stack-check -mno-apcs-stack-check @gol
456 -mapcs-float -mno-apcs-float @gol
457 -mapcs-reentrant -mno-apcs-reentrant @gol
458 -msched-prolog -mno-sched-prolog @gol
459 -mlittle-endian -mbig-endian -mwords-little-endian @gol
460 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
461 -mfp16-format=@var{name}
462 -mthumb-interwork -mno-thumb-interwork @gol
463 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
464 -mstructure-size-boundary=@var{n} @gol
465 -mabort-on-noreturn @gol
466 -mlong-calls -mno-long-calls @gol
467 -msingle-pic-base -mno-single-pic-base @gol
468 -mpic-register=@var{reg} @gol
469 -mnop-fun-dllimport @gol
470 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
471 -mpoke-function-name @gol
473 -mtpcs-frame -mtpcs-leaf-frame @gol
474 -mcaller-super-interworking -mcallee-super-interworking @gol
476 -mword-relocations @gol
477 -mfix-cortex-m3-ldrd}
480 @gccoptlist{-mmcu=@var{mcu} -mno-interrupts @gol
481 -mcall-prologues -mtiny-stack -mint8}
483 @emph{Blackfin Options}
484 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
485 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
486 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
487 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
488 -mno-id-shared-library -mshared-library-id=@var{n} @gol
489 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
490 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
491 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
495 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
496 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
497 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
498 -mstack-align -mdata-align -mconst-align @gol
499 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
500 -melf -maout -melinux -mlinux -sim -sim2 @gol
501 -mmul-bug-workaround -mno-mul-bug-workaround}
504 @gccoptlist{-mmac -mpush-args}
506 @emph{Darwin Options}
507 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
508 -arch_only -bind_at_load -bundle -bundle_loader @gol
509 -client_name -compatibility_version -current_version @gol
511 -dependency-file -dylib_file -dylinker_install_name @gol
512 -dynamic -dynamiclib -exported_symbols_list @gol
513 -filelist -flat_namespace -force_cpusubtype_ALL @gol
514 -force_flat_namespace -headerpad_max_install_names @gol
516 -image_base -init -install_name -keep_private_externs @gol
517 -multi_module -multiply_defined -multiply_defined_unused @gol
518 -noall_load -no_dead_strip_inits_and_terms @gol
519 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
520 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
521 -private_bundle -read_only_relocs -sectalign @gol
522 -sectobjectsymbols -whyload -seg1addr @gol
523 -sectcreate -sectobjectsymbols -sectorder @gol
524 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
525 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
526 -segprot -segs_read_only_addr -segs_read_write_addr @gol
527 -single_module -static -sub_library -sub_umbrella @gol
528 -twolevel_namespace -umbrella -undefined @gol
529 -unexported_symbols_list -weak_reference_mismatches @gol
530 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
531 -mkernel -mone-byte-bool}
533 @emph{DEC Alpha Options}
534 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
535 -mieee -mieee-with-inexact -mieee-conformant @gol
536 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
537 -mtrap-precision=@var{mode} -mbuild-constants @gol
538 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
539 -mbwx -mmax -mfix -mcix @gol
540 -mfloat-vax -mfloat-ieee @gol
541 -mexplicit-relocs -msmall-data -mlarge-data @gol
542 -msmall-text -mlarge-text @gol
543 -mmemory-latency=@var{time}}
545 @emph{DEC Alpha/VMS Options}
546 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
549 @gccoptlist{-msmall-model -mno-lsim}
552 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
553 -mhard-float -msoft-float @gol
554 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
555 -mdouble -mno-double @gol
556 -mmedia -mno-media -mmuladd -mno-muladd @gol
557 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
558 -mlinked-fp -mlong-calls -malign-labels @gol
559 -mlibrary-pic -macc-4 -macc-8 @gol
560 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
561 -moptimize-membar -mno-optimize-membar @gol
562 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
563 -mvliw-branch -mno-vliw-branch @gol
564 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
565 -mno-nested-cond-exec -mtomcat-stats @gol
569 @emph{GNU/Linux Options}
570 @gccoptlist{-mglibc -muclibc -mbionic -mandroid @gol
571 -tno-android-cc -tno-android-ld}
573 @emph{H8/300 Options}
574 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
577 @gccoptlist{-march=@var{architecture-type} @gol
578 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
579 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
580 -mfixed-range=@var{register-range} @gol
581 -mjump-in-delay -mlinker-opt -mlong-calls @gol
582 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
583 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
584 -mno-jump-in-delay -mno-long-load-store @gol
585 -mno-portable-runtime -mno-soft-float @gol
586 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
587 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
588 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
589 -munix=@var{unix-std} -nolibdld -static -threads}
591 @emph{i386 and x86-64 Options}
592 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
593 -mfpmath=@var{unit} @gol
594 -masm=@var{dialect} -mno-fancy-math-387 @gol
595 -mno-fp-ret-in-387 -msoft-float @gol
596 -mno-wide-multiply -mrtd -malign-double @gol
597 -mpreferred-stack-boundary=@var{num}
598 -mincoming-stack-boundary=@var{num} @gol
599 -mcld -mcx16 -msahf -mmovbe -mcrc32 -mrecip -mvzeroupper @gol
600 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
601 -maes -mpclmul -mfsgsbase -mrdrnd -mf16c -mfused-madd @gol
602 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlwp @gol
603 -mthreads -mno-align-stringops -minline-all-stringops @gol
604 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
605 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
606 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
607 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
608 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
609 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
610 -mcmodel=@var{code-model} -mabi=@var{name} @gol
611 -m32 -m64 -mlarge-data-threshold=@var{num} @gol
612 -msse2avx -mfentry -m8bit-idiv}
615 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
616 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
617 -mconstant-gp -mauto-pic -mfused-madd @gol
618 -minline-float-divide-min-latency @gol
619 -minline-float-divide-max-throughput @gol
620 -mno-inline-float-divide @gol
621 -minline-int-divide-min-latency @gol
622 -minline-int-divide-max-throughput @gol
623 -mno-inline-int-divide @gol
624 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
625 -mno-inline-sqrt @gol
626 -mdwarf2-asm -mearly-stop-bits @gol
627 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
628 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
629 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
630 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
631 -msched-spec-ldc -msched-spec-control-ldc @gol
632 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
633 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
634 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
635 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
637 @emph{IA-64/VMS Options}
638 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
641 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
642 -msign-extend-enabled -muser-enabled}
644 @emph{M32R/D Options}
645 @gccoptlist{-m32r2 -m32rx -m32r @gol
647 -malign-loops -mno-align-loops @gol
648 -missue-rate=@var{number} @gol
649 -mbranch-cost=@var{number} @gol
650 -mmodel=@var{code-size-model-type} @gol
651 -msdata=@var{sdata-type} @gol
652 -mno-flush-func -mflush-func=@var{name} @gol
653 -mno-flush-trap -mflush-trap=@var{number} @gol
657 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
659 @emph{M680x0 Options}
660 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
661 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
662 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
663 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
664 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
665 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
666 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
667 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
670 @emph{M68hc1x Options}
671 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
672 -mauto-incdec -minmax -mlong-calls -mshort @gol
673 -msoft-reg-count=@var{count}}
676 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
677 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
678 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
679 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
680 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
683 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
684 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
685 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
686 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
689 @emph{MicroBlaze Options}
690 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
691 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
692 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
693 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
694 -mxl-mode-@var{app-model}}
697 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
698 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
699 -mips64 -mips64r2 @gol
700 -mips16 -mno-mips16 -mflip-mips16 @gol
701 -minterlink-mips16 -mno-interlink-mips16 @gol
702 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
703 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
704 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
705 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
706 -mfpu=@var{fpu-type} @gol
707 -msmartmips -mno-smartmips @gol
708 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
709 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
710 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
711 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
712 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
713 -membedded-data -mno-embedded-data @gol
714 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
715 -mcode-readable=@var{setting} @gol
716 -msplit-addresses -mno-split-addresses @gol
717 -mexplicit-relocs -mno-explicit-relocs @gol
718 -mcheck-zero-division -mno-check-zero-division @gol
719 -mdivide-traps -mdivide-breaks @gol
720 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
721 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
722 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
723 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
724 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
725 -mflush-func=@var{func} -mno-flush-func @gol
726 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
727 -mfp-exceptions -mno-fp-exceptions @gol
728 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
729 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
732 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
733 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
734 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
735 -mno-base-addresses -msingle-exit -mno-single-exit}
737 @emph{MN10300 Options}
738 @gccoptlist{-mmult-bug -mno-mult-bug @gol
739 -mno-am33 -mam33 -mam33-2 -mam34 @gol
740 -mtune=@var{cpu-type} @gol
741 -mreturn-pointer-on-d0 @gol
744 @emph{PDP-11 Options}
745 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
746 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
747 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
748 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
749 -mbranch-expensive -mbranch-cheap @gol
750 -munix-asm -mdec-asm}
752 @emph{picoChip Options}
753 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
754 -msymbol-as-address -mno-inefficient-warnings}
756 @emph{PowerPC Options}
757 See RS/6000 and PowerPC Options.
759 @emph{RS/6000 and PowerPC Options}
760 @gccoptlist{-mcpu=@var{cpu-type} @gol
761 -mtune=@var{cpu-type} @gol
762 -mcmodel=@var{code-model} @gol
763 -mpower -mno-power -mpower2 -mno-power2 @gol
764 -mpowerpc -mpowerpc64 -mno-powerpc @gol
765 -maltivec -mno-altivec @gol
766 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
767 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
768 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
769 -mfprnd -mno-fprnd @gol
770 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
771 -mnew-mnemonics -mold-mnemonics @gol
772 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
773 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
774 -malign-power -malign-natural @gol
775 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
776 -msingle-float -mdouble-float -msimple-fpu @gol
777 -mstring -mno-string -mupdate -mno-update @gol
778 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
779 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
780 -mstrict-align -mno-strict-align -mrelocatable @gol
781 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
782 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
783 -mdynamic-no-pic -maltivec -mswdiv @gol
784 -mprioritize-restricted-insns=@var{priority} @gol
785 -msched-costly-dep=@var{dependence_type} @gol
786 -minsert-sched-nops=@var{scheme} @gol
787 -mcall-sysv -mcall-netbsd @gol
788 -maix-struct-return -msvr4-struct-return @gol
789 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
790 -mblock-move-inline-limit=@var{num} @gol
791 -misel -mno-isel @gol
792 -misel=yes -misel=no @gol
794 -mspe=yes -mspe=no @gol
796 -mgen-cell-microcode -mwarn-cell-microcode @gol
797 -mvrsave -mno-vrsave @gol
798 -mmulhw -mno-mulhw @gol
799 -mdlmzb -mno-dlmzb @gol
800 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
801 -mprototype -mno-prototype @gol
802 -msim -mmvme -mads -myellowknife -memb -msdata @gol
803 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
804 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision
805 -mno-recip-precision @gol
806 -mveclibabi=@var{type} -mfriz -mno-friz}
809 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
811 -mbig-endian-data -mlittle-endian-data @gol
814 -mas100-syntax -mno-as100-syntax@gol
816 -mmax-constant-size=@gol
818 -msave-acc-in-interrupts}
820 @emph{S/390 and zSeries Options}
821 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
822 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
823 -mlong-double-64 -mlong-double-128 @gol
824 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
825 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
826 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
827 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
828 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
831 @gccoptlist{-meb -mel @gol
835 -mscore5 -mscore5u -mscore7 -mscore7d}
838 @gccoptlist{-m1 -m2 -m2e @gol
839 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
841 -m4-nofpu -m4-single-only -m4-single -m4 @gol
842 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
843 -m5-64media -m5-64media-nofpu @gol
844 -m5-32media -m5-32media-nofpu @gol
845 -m5-compact -m5-compact-nofpu @gol
846 -mb -ml -mdalign -mrelax @gol
847 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
848 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
849 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
850 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
851 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
852 -maccumulate-outgoing-args -minvalid-symbols}
854 @emph{Solaris 2 Options}
855 @gccoptlist{-mimpure-text -mno-impure-text @gol
856 -threads -pthreads -pthread}
859 @gccoptlist{-mcpu=@var{cpu-type} @gol
860 -mtune=@var{cpu-type} @gol
861 -mcmodel=@var{code-model} @gol
862 -m32 -m64 -mapp-regs -mno-app-regs @gol
863 -mfaster-structs -mno-faster-structs @gol
864 -mfpu -mno-fpu -mhard-float -msoft-float @gol
865 -mhard-quad-float -msoft-quad-float @gol
867 -mstack-bias -mno-stack-bias @gol
868 -munaligned-doubles -mno-unaligned-doubles @gol
869 -mv8plus -mno-v8plus -mvis -mno-vis}
872 @gccoptlist{-mwarn-reloc -merror-reloc @gol
873 -msafe-dma -munsafe-dma @gol
875 -msmall-mem -mlarge-mem -mstdmain @gol
876 -mfixed-range=@var{register-range} @gol
878 -maddress-space-conversion -mno-address-space-conversion @gol
879 -mcache-size=@var{cache-size} @gol
880 -matomic-updates -mno-atomic-updates}
882 @emph{System V Options}
883 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
886 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
887 -mprolog-function -mno-prolog-function -mspace @gol
888 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
889 -mapp-regs -mno-app-regs @gol
890 -mdisable-callt -mno-disable-callt @gol
898 @gccoptlist{-mg -mgnu -munix}
900 @emph{VxWorks Options}
901 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
902 -Xbind-lazy -Xbind-now}
904 @emph{x86-64 Options}
905 See i386 and x86-64 Options.
907 @emph{i386 and x86-64 Windows Options}
908 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll
909 -mnop-fun-dllimport -mthread @gol
910 -municode -mwin32 -mwindows -fno-set-stack-executable}
912 @emph{Xstormy16 Options}
915 @emph{Xtensa Options}
916 @gccoptlist{-mconst16 -mno-const16 @gol
917 -mfused-madd -mno-fused-madd @gol
919 -mserialize-volatile -mno-serialize-volatile @gol
920 -mtext-section-literals -mno-text-section-literals @gol
921 -mtarget-align -mno-target-align @gol
922 -mlongcalls -mno-longcalls}
924 @emph{zSeries Options}
925 See S/390 and zSeries Options.
927 @item Code Generation Options
928 @xref{Code Gen Options,,Options for Code Generation Conventions}.
929 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
930 -ffixed-@var{reg} -fexceptions @gol
931 -fnon-call-exceptions -funwind-tables @gol
932 -fasynchronous-unwind-tables @gol
933 -finhibit-size-directive -finstrument-functions @gol
934 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
935 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
936 -fno-common -fno-ident @gol
937 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
938 -fno-jump-tables @gol
939 -frecord-gcc-switches @gol
940 -freg-struct-return -fshort-enums @gol
941 -fshort-double -fshort-wchar @gol
942 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
943 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
944 -fno-stack-limit -fsplit-stack @gol
945 -fleading-underscore -ftls-model=@var{model} @gol
946 -ftrapv -fwrapv -fbounds-check @gol
947 -fvisibility -fstrict-volatile-bitfields}
951 * Overall Options:: Controlling the kind of output:
952 an executable, object files, assembler files,
953 or preprocessed source.
954 * C Dialect Options:: Controlling the variant of C language compiled.
955 * C++ Dialect Options:: Variations on C++.
956 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
958 * Language Independent Options:: Controlling how diagnostics should be
960 * Warning Options:: How picky should the compiler be?
961 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
962 * Optimize Options:: How much optimization?
963 * Preprocessor Options:: Controlling header files and macro definitions.
964 Also, getting dependency information for Make.
965 * Assembler Options:: Passing options to the assembler.
966 * Link Options:: Specifying libraries and so on.
967 * Directory Options:: Where to find header files and libraries.
968 Where to find the compiler executable files.
969 * Spec Files:: How to pass switches to sub-processes.
970 * Target Options:: Running a cross-compiler, or an old version of GCC.
973 @node Overall Options
974 @section Options Controlling the Kind of Output
976 Compilation can involve up to four stages: preprocessing, compilation
977 proper, assembly and linking, always in that order. GCC is capable of
978 preprocessing and compiling several files either into several
979 assembler input files, or into one assembler input file; then each
980 assembler input file produces an object file, and linking combines all
981 the object files (those newly compiled, and those specified as input)
982 into an executable file.
984 @cindex file name suffix
985 For any given input file, the file name suffix determines what kind of
990 C source code which must be preprocessed.
993 C source code which should not be preprocessed.
996 C++ source code which should not be preprocessed.
999 Objective-C source code. Note that you must link with the @file{libobjc}
1000 library to make an Objective-C program work.
1003 Objective-C source code which should not be preprocessed.
1007 Objective-C++ source code. Note that you must link with the @file{libobjc}
1008 library to make an Objective-C++ program work. Note that @samp{.M} refers
1009 to a literal capital M@.
1011 @item @var{file}.mii
1012 Objective-C++ source code which should not be preprocessed.
1015 C, C++, Objective-C or Objective-C++ header file to be turned into a
1016 precompiled header (default), or C, C++ header file to be turned into an
1017 Ada spec (via the @option{-fdump-ada-spec} switch).
1020 @itemx @var{file}.cp
1021 @itemx @var{file}.cxx
1022 @itemx @var{file}.cpp
1023 @itemx @var{file}.CPP
1024 @itemx @var{file}.c++
1026 C++ source code which must be preprocessed. Note that in @samp{.cxx},
1027 the last two letters must both be literally @samp{x}. Likewise,
1028 @samp{.C} refers to a literal capital C@.
1032 Objective-C++ source code which must be preprocessed.
1034 @item @var{file}.mii
1035 Objective-C++ source code which should not be preprocessed.
1039 @itemx @var{file}.hp
1040 @itemx @var{file}.hxx
1041 @itemx @var{file}.hpp
1042 @itemx @var{file}.HPP
1043 @itemx @var{file}.h++
1044 @itemx @var{file}.tcc
1045 C++ header file to be turned into a precompiled header or Ada spec.
1048 @itemx @var{file}.for
1049 @itemx @var{file}.ftn
1050 Fixed form Fortran source code which should not be preprocessed.
1053 @itemx @var{file}.FOR
1054 @itemx @var{file}.fpp
1055 @itemx @var{file}.FPP
1056 @itemx @var{file}.FTN
1057 Fixed form Fortran source code which must be preprocessed (with the traditional
1060 @item @var{file}.f90
1061 @itemx @var{file}.f95
1062 @itemx @var{file}.f03
1063 @itemx @var{file}.f08
1064 Free form Fortran source code which should not be preprocessed.
1066 @item @var{file}.F90
1067 @itemx @var{file}.F95
1068 @itemx @var{file}.F03
1069 @itemx @var{file}.F08
1070 Free form Fortran source code which must be preprocessed (with the
1071 traditional preprocessor).
1073 @c FIXME: Descriptions of Java file types.
1079 @item @var{file}.ads
1080 Ada source code file which contains a library unit declaration (a
1081 declaration of a package, subprogram, or generic, or a generic
1082 instantiation), or a library unit renaming declaration (a package,
1083 generic, or subprogram renaming declaration). Such files are also
1086 @item @var{file}.adb
1087 Ada source code file containing a library unit body (a subprogram or
1088 package body). Such files are also called @dfn{bodies}.
1090 @c GCC also knows about some suffixes for languages not yet included:
1101 @itemx @var{file}.sx
1102 Assembler code which must be preprocessed.
1105 An object file to be fed straight into linking.
1106 Any file name with no recognized suffix is treated this way.
1110 You can specify the input language explicitly with the @option{-x} option:
1113 @item -x @var{language}
1114 Specify explicitly the @var{language} for the following input files
1115 (rather than letting the compiler choose a default based on the file
1116 name suffix). This option applies to all following input files until
1117 the next @option{-x} option. Possible values for @var{language} are:
1119 c c-header cpp-output
1120 c++ c++-header c++-cpp-output
1121 objective-c objective-c-header objective-c-cpp-output
1122 objective-c++ objective-c++-header objective-c++-cpp-output
1123 assembler assembler-with-cpp
1125 f77 f77-cpp-input f95 f95-cpp-input
1130 Turn off any specification of a language, so that subsequent files are
1131 handled according to their file name suffixes (as they are if @option{-x}
1132 has not been used at all).
1134 @item -pass-exit-codes
1135 @opindex pass-exit-codes
1136 Normally the @command{gcc} program will exit with the code of 1 if any
1137 phase of the compiler returns a non-success return code. If you specify
1138 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1139 numerically highest error produced by any phase that returned an error
1140 indication. The C, C++, and Fortran frontends return 4, if an internal
1141 compiler error is encountered.
1144 If you only want some of the stages of compilation, you can use
1145 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1146 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1147 @command{gcc} is to stop. Note that some combinations (for example,
1148 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1153 Compile or assemble the source files, but do not link. The linking
1154 stage simply is not done. The ultimate output is in the form of an
1155 object file for each source file.
1157 By default, the object file name for a source file is made by replacing
1158 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1160 Unrecognized input files, not requiring compilation or assembly, are
1165 Stop after the stage of compilation proper; do not assemble. The output
1166 is in the form of an assembler code file for each non-assembler input
1169 By default, the assembler file name for a source file is made by
1170 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1172 Input files that don't require compilation are ignored.
1176 Stop after the preprocessing stage; do not run the compiler proper. The
1177 output is in the form of preprocessed source code, which is sent to the
1180 Input files which don't require preprocessing are ignored.
1182 @cindex output file option
1185 Place output in file @var{file}. This applies regardless to whatever
1186 sort of output is being produced, whether it be an executable file,
1187 an object file, an assembler file or preprocessed C code.
1189 If @option{-o} is not specified, the default is to put an executable
1190 file in @file{a.out}, the object file for
1191 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1192 assembler file in @file{@var{source}.s}, a precompiled header file in
1193 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1198 Print (on standard error output) the commands executed to run the stages
1199 of compilation. Also print the version number of the compiler driver
1200 program and of the preprocessor and the compiler proper.
1204 Like @option{-v} except the commands are not executed and arguments
1205 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1206 This is useful for shell scripts to capture the driver-generated command lines.
1210 Use pipes rather than temporary files for communication between the
1211 various stages of compilation. This fails to work on some systems where
1212 the assembler is unable to read from a pipe; but the GNU assembler has
1217 Print (on the standard output) a description of the command line options
1218 understood by @command{gcc}. If the @option{-v} option is also specified
1219 then @option{--help} will also be passed on to the various processes
1220 invoked by @command{gcc}, so that they can display the command line options
1221 they accept. If the @option{-Wextra} option has also been specified
1222 (prior to the @option{--help} option), then command line options which
1223 have no documentation associated with them will also be displayed.
1226 @opindex target-help
1227 Print (on the standard output) a description of target-specific command
1228 line options for each tool. For some targets extra target-specific
1229 information may also be printed.
1231 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1232 Print (on the standard output) a description of the command line
1233 options understood by the compiler that fit into all specified classes
1234 and qualifiers. These are the supported classes:
1237 @item @samp{optimizers}
1238 This will display all of the optimization options supported by the
1241 @item @samp{warnings}
1242 This will display all of the options controlling warning messages
1243 produced by the compiler.
1246 This will display target-specific options. Unlike the
1247 @option{--target-help} option however, target-specific options of the
1248 linker and assembler will not be displayed. This is because those
1249 tools do not currently support the extended @option{--help=} syntax.
1252 This will display the values recognized by the @option{--param}
1255 @item @var{language}
1256 This will display the options supported for @var{language}, where
1257 @var{language} is the name of one of the languages supported in this
1261 This will display the options that are common to all languages.
1264 These are the supported qualifiers:
1267 @item @samp{undocumented}
1268 Display only those options which are undocumented.
1271 Display options which take an argument that appears after an equal
1272 sign in the same continuous piece of text, such as:
1273 @samp{--help=target}.
1275 @item @samp{separate}
1276 Display options which take an argument that appears as a separate word
1277 following the original option, such as: @samp{-o output-file}.
1280 Thus for example to display all the undocumented target-specific
1281 switches supported by the compiler the following can be used:
1284 --help=target,undocumented
1287 The sense of a qualifier can be inverted by prefixing it with the
1288 @samp{^} character, so for example to display all binary warning
1289 options (i.e., ones that are either on or off and that do not take an
1290 argument), which have a description the following can be used:
1293 --help=warnings,^joined,^undocumented
1296 The argument to @option{--help=} should not consist solely of inverted
1299 Combining several classes is possible, although this usually
1300 restricts the output by so much that there is nothing to display. One
1301 case where it does work however is when one of the classes is
1302 @var{target}. So for example to display all the target-specific
1303 optimization options the following can be used:
1306 --help=target,optimizers
1309 The @option{--help=} option can be repeated on the command line. Each
1310 successive use will display its requested class of options, skipping
1311 those that have already been displayed.
1313 If the @option{-Q} option appears on the command line before the
1314 @option{--help=} option, then the descriptive text displayed by
1315 @option{--help=} is changed. Instead of describing the displayed
1316 options, an indication is given as to whether the option is enabled,
1317 disabled or set to a specific value (assuming that the compiler
1318 knows this at the point where the @option{--help=} option is used).
1320 Here is a truncated example from the ARM port of @command{gcc}:
1323 % gcc -Q -mabi=2 --help=target -c
1324 The following options are target specific:
1326 -mabort-on-noreturn [disabled]
1330 The output is sensitive to the effects of previous command line
1331 options, so for example it is possible to find out which optimizations
1332 are enabled at @option{-O2} by using:
1335 -Q -O2 --help=optimizers
1338 Alternatively you can discover which binary optimizations are enabled
1339 by @option{-O3} by using:
1342 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1343 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1344 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1347 @item -no-canonical-prefixes
1348 @opindex no-canonical-prefixes
1349 Do not expand any symbolic links, resolve references to @samp{/../}
1350 or @samp{/./}, or make the path absolute when generating a relative
1355 Display the version number and copyrights of the invoked GCC@.
1359 Invoke all subcommands under a wrapper program. It takes a single
1360 comma separated list as an argument, which will be used to invoke
1364 gcc -c t.c -wrapper gdb,--args
1367 This will invoke all subprograms of gcc under "gdb --args",
1368 thus cc1 invocation will be "gdb --args cc1 ...".
1370 @item -fplugin=@var{name}.so
1371 Load the plugin code in file @var{name}.so, assumed to be a
1372 shared object to be dlopen'd by the compiler. The base name of
1373 the shared object file is used to identify the plugin for the
1374 purposes of argument parsing (See
1375 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1376 Each plugin should define the callback functions specified in the
1379 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1380 Define an argument called @var{key} with a value of @var{value}
1381 for the plugin called @var{name}.
1383 @item -fdump-ada-spec@r{[}-slim@r{]}
1384 For C and C++ source and include files, generate corresponding Ada
1385 specs. @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1386 GNAT User's Guide}, which provides detailed documentation on this feature.
1388 @item -fdump-go-spec=@var{file}
1389 For input files in any language, generate corresponding Go
1390 declarations in @var{file}. This generates Go @code{const},
1391 @code{type}, @code{var}, and @code{func} declarations which may be a
1392 useful way to start writing a Go interface to code written in some
1395 @include @value{srcdir}/../libiberty/at-file.texi
1399 @section Compiling C++ Programs
1401 @cindex suffixes for C++ source
1402 @cindex C++ source file suffixes
1403 C++ source files conventionally use one of the suffixes @samp{.C},
1404 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1405 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1406 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1407 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1408 files with these names and compiles them as C++ programs even if you
1409 call the compiler the same way as for compiling C programs (usually
1410 with the name @command{gcc}).
1414 However, the use of @command{gcc} does not add the C++ library.
1415 @command{g++} is a program that calls GCC and treats @samp{.c},
1416 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1417 files unless @option{-x} is used, and automatically specifies linking
1418 against the C++ library. This program is also useful when
1419 precompiling a C header file with a @samp{.h} extension for use in C++
1420 compilations. On many systems, @command{g++} is also installed with
1421 the name @command{c++}.
1423 @cindex invoking @command{g++}
1424 When you compile C++ programs, you may specify many of the same
1425 command-line options that you use for compiling programs in any
1426 language; or command-line options meaningful for C and related
1427 languages; or options that are meaningful only for C++ programs.
1428 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1429 explanations of options for languages related to C@.
1430 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1431 explanations of options that are meaningful only for C++ programs.
1433 @node C Dialect Options
1434 @section Options Controlling C Dialect
1435 @cindex dialect options
1436 @cindex language dialect options
1437 @cindex options, dialect
1439 The following options control the dialect of C (or languages derived
1440 from C, such as C++, Objective-C and Objective-C++) that the compiler
1444 @cindex ANSI support
1448 In C mode, this is equivalent to @samp{-std=c90}. In C++ mode, it is
1449 equivalent to @samp{-std=c++98}.
1451 This turns off certain features of GCC that are incompatible with ISO
1452 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1453 such as the @code{asm} and @code{typeof} keywords, and
1454 predefined macros such as @code{unix} and @code{vax} that identify the
1455 type of system you are using. It also enables the undesirable and
1456 rarely used ISO trigraph feature. For the C compiler,
1457 it disables recognition of C++ style @samp{//} comments as well as
1458 the @code{inline} keyword.
1460 The alternate keywords @code{__asm__}, @code{__extension__},
1461 @code{__inline__} and @code{__typeof__} continue to work despite
1462 @option{-ansi}. You would not want to use them in an ISO C program, of
1463 course, but it is useful to put them in header files that might be included
1464 in compilations done with @option{-ansi}. Alternate predefined macros
1465 such as @code{__unix__} and @code{__vax__} are also available, with or
1466 without @option{-ansi}.
1468 The @option{-ansi} option does not cause non-ISO programs to be
1469 rejected gratuitously. For that, @option{-pedantic} is required in
1470 addition to @option{-ansi}. @xref{Warning Options}.
1472 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1473 option is used. Some header files may notice this macro and refrain
1474 from declaring certain functions or defining certain macros that the
1475 ISO standard doesn't call for; this is to avoid interfering with any
1476 programs that might use these names for other things.
1478 Functions that would normally be built in but do not have semantics
1479 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1480 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1481 built-in functions provided by GCC}, for details of the functions
1486 Determine the language standard. @xref{Standards,,Language Standards
1487 Supported by GCC}, for details of these standard versions. This option
1488 is currently only supported when compiling C or C++.
1490 The compiler can accept several base standards, such as @samp{c90} or
1491 @samp{c++98}, and GNU dialects of those standards, such as
1492 @samp{gnu90} or @samp{gnu++98}. By specifying a base standard, the
1493 compiler will accept all programs following that standard and those
1494 using GNU extensions that do not contradict it. For example,
1495 @samp{-std=c90} turns off certain features of GCC that are
1496 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1497 keywords, but not other GNU extensions that do not have a meaning in
1498 ISO C90, such as omitting the middle term of a @code{?:}
1499 expression. On the other hand, by specifying a GNU dialect of a
1500 standard, all features the compiler support are enabled, even when
1501 those features change the meaning of the base standard and some
1502 strict-conforming programs may be rejected. The particular standard
1503 is used by @option{-pedantic} to identify which features are GNU
1504 extensions given that version of the standard. For example
1505 @samp{-std=gnu90 -pedantic} would warn about C++ style @samp{//}
1506 comments, while @samp{-std=gnu99 -pedantic} would not.
1508 A value for this option must be provided; possible values are
1514 Support all ISO C90 programs (certain GNU extensions that conflict
1515 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1517 @item iso9899:199409
1518 ISO C90 as modified in amendment 1.
1524 ISO C99. Note that this standard is not yet fully supported; see
1525 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1526 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1529 ISO C1X, the draft of the next revision of the ISO C standard.
1530 Support is limited and experimental and features enabled by this
1531 option may be changed or removed if changed in or removed from the
1536 GNU dialect of ISO C90 (including some C99 features). This
1537 is the default for C code.
1541 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1542 this will become the default. The name @samp{gnu9x} is deprecated.
1545 GNU dialect of ISO C1X. Support is limited and experimental and
1546 features enabled by this option may be changed or removed if changed
1547 in or removed from the standard draft.
1550 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1554 GNU dialect of @option{-std=c++98}. This is the default for
1558 The working draft of the upcoming ISO C++0x standard. This option
1559 enables experimental features that are likely to be included in
1560 C++0x. The working draft is constantly changing, and any feature that is
1561 enabled by this flag may be removed from future versions of GCC if it is
1562 not part of the C++0x standard.
1565 GNU dialect of @option{-std=c++0x}. This option enables
1566 experimental features that may be removed in future versions of GCC.
1569 @item -fgnu89-inline
1570 @opindex fgnu89-inline
1571 The option @option{-fgnu89-inline} tells GCC to use the traditional
1572 GNU semantics for @code{inline} functions when in C99 mode.
1573 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1574 is accepted and ignored by GCC versions 4.1.3 up to but not including
1575 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1576 C99 mode. Using this option is roughly equivalent to adding the
1577 @code{gnu_inline} function attribute to all inline functions
1578 (@pxref{Function Attributes}).
1580 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1581 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1582 specifies the default behavior). This option was first supported in
1583 GCC 4.3. This option is not supported in @option{-std=c90} or
1584 @option{-std=gnu90} mode.
1586 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1587 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1588 in effect for @code{inline} functions. @xref{Common Predefined
1589 Macros,,,cpp,The C Preprocessor}.
1591 @item -aux-info @var{filename}
1593 Output to the given filename prototyped declarations for all functions
1594 declared and/or defined in a translation unit, including those in header
1595 files. This option is silently ignored in any language other than C@.
1597 Besides declarations, the file indicates, in comments, the origin of
1598 each declaration (source file and line), whether the declaration was
1599 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1600 @samp{O} for old, respectively, in the first character after the line
1601 number and the colon), and whether it came from a declaration or a
1602 definition (@samp{C} or @samp{F}, respectively, in the following
1603 character). In the case of function definitions, a K&R-style list of
1604 arguments followed by their declarations is also provided, inside
1605 comments, after the declaration.
1609 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1610 keyword, so that code can use these words as identifiers. You can use
1611 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1612 instead. @option{-ansi} implies @option{-fno-asm}.
1614 In C++, this switch only affects the @code{typeof} keyword, since
1615 @code{asm} and @code{inline} are standard keywords. You may want to
1616 use the @option{-fno-gnu-keywords} flag instead, which has the same
1617 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1618 switch only affects the @code{asm} and @code{typeof} keywords, since
1619 @code{inline} is a standard keyword in ISO C99.
1622 @itemx -fno-builtin-@var{function}
1623 @opindex fno-builtin
1624 @cindex built-in functions
1625 Don't recognize built-in functions that do not begin with
1626 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1627 functions provided by GCC}, for details of the functions affected,
1628 including those which are not built-in functions when @option{-ansi} or
1629 @option{-std} options for strict ISO C conformance are used because they
1630 do not have an ISO standard meaning.
1632 GCC normally generates special code to handle certain built-in functions
1633 more efficiently; for instance, calls to @code{alloca} may become single
1634 instructions that adjust the stack directly, and calls to @code{memcpy}
1635 may become inline copy loops. The resulting code is often both smaller
1636 and faster, but since the function calls no longer appear as such, you
1637 cannot set a breakpoint on those calls, nor can you change the behavior
1638 of the functions by linking with a different library. In addition,
1639 when a function is recognized as a built-in function, GCC may use
1640 information about that function to warn about problems with calls to
1641 that function, or to generate more efficient code, even if the
1642 resulting code still contains calls to that function. For example,
1643 warnings are given with @option{-Wformat} for bad calls to
1644 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1645 known not to modify global memory.
1647 With the @option{-fno-builtin-@var{function}} option
1648 only the built-in function @var{function} is
1649 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1650 function is named that is not built-in in this version of GCC, this
1651 option is ignored. There is no corresponding
1652 @option{-fbuiltin-@var{function}} option; if you wish to enable
1653 built-in functions selectively when using @option{-fno-builtin} or
1654 @option{-ffreestanding}, you may define macros such as:
1657 #define abs(n) __builtin_abs ((n))
1658 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1663 @cindex hosted environment
1665 Assert that compilation takes place in a hosted environment. This implies
1666 @option{-fbuiltin}. A hosted environment is one in which the
1667 entire standard library is available, and in which @code{main} has a return
1668 type of @code{int}. Examples are nearly everything except a kernel.
1669 This is equivalent to @option{-fno-freestanding}.
1671 @item -ffreestanding
1672 @opindex ffreestanding
1673 @cindex hosted environment
1675 Assert that compilation takes place in a freestanding environment. This
1676 implies @option{-fno-builtin}. A freestanding environment
1677 is one in which the standard library may not exist, and program startup may
1678 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1679 This is equivalent to @option{-fno-hosted}.
1681 @xref{Standards,,Language Standards Supported by GCC}, for details of
1682 freestanding and hosted environments.
1686 @cindex OpenMP parallel
1687 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1688 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1689 compiler generates parallel code according to the OpenMP Application
1690 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1691 implies @option{-pthread}, and thus is only supported on targets that
1692 have support for @option{-pthread}.
1694 @item -fms-extensions
1695 @opindex fms-extensions
1696 Accept some non-standard constructs used in Microsoft header files.
1698 It allows for c++ that member-names in structures can be similiar
1699 to previous types declarations.
1708 Some cases of unnamed fields in structures and unions are only
1709 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1710 fields within structs/unions}, for details.
1712 @item -fplan9-extensions
1713 Accept some non-standard constructs used in Plan 9 code.
1715 This enables @option{-fms-extensions}, permits passing pointers to
1716 structures with anonymous fields to functions which expect pointers to
1717 elements of the type of the field, and permits referring to anonymous
1718 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
1719 struct/union fields within structs/unions}, for details. This is only
1720 supported for C, not C++.
1724 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1725 options for strict ISO C conformance) implies @option{-trigraphs}.
1727 @item -no-integrated-cpp
1728 @opindex no-integrated-cpp
1729 Performs a compilation in two passes: preprocessing and compiling. This
1730 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1731 @option{-B} option. The user supplied compilation step can then add in
1732 an additional preprocessing step after normal preprocessing but before
1733 compiling. The default is to use the integrated cpp (internal cpp)
1735 The semantics of this option will change if "cc1", "cc1plus", and
1736 "cc1obj" are merged.
1738 @cindex traditional C language
1739 @cindex C language, traditional
1741 @itemx -traditional-cpp
1742 @opindex traditional-cpp
1743 @opindex traditional
1744 Formerly, these options caused GCC to attempt to emulate a pre-standard
1745 C compiler. They are now only supported with the @option{-E} switch.
1746 The preprocessor continues to support a pre-standard mode. See the GNU
1747 CPP manual for details.
1749 @item -fcond-mismatch
1750 @opindex fcond-mismatch
1751 Allow conditional expressions with mismatched types in the second and
1752 third arguments. The value of such an expression is void. This option
1753 is not supported for C++.
1755 @item -flax-vector-conversions
1756 @opindex flax-vector-conversions
1757 Allow implicit conversions between vectors with differing numbers of
1758 elements and/or incompatible element types. This option should not be
1761 @item -funsigned-char
1762 @opindex funsigned-char
1763 Let the type @code{char} be unsigned, like @code{unsigned char}.
1765 Each kind of machine has a default for what @code{char} should
1766 be. It is either like @code{unsigned char} by default or like
1767 @code{signed char} by default.
1769 Ideally, a portable program should always use @code{signed char} or
1770 @code{unsigned char} when it depends on the signedness of an object.
1771 But many programs have been written to use plain @code{char} and
1772 expect it to be signed, or expect it to be unsigned, depending on the
1773 machines they were written for. This option, and its inverse, let you
1774 make such a program work with the opposite default.
1776 The type @code{char} is always a distinct type from each of
1777 @code{signed char} or @code{unsigned char}, even though its behavior
1778 is always just like one of those two.
1781 @opindex fsigned-char
1782 Let the type @code{char} be signed, like @code{signed char}.
1784 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1785 the negative form of @option{-funsigned-char}. Likewise, the option
1786 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1788 @item -fsigned-bitfields
1789 @itemx -funsigned-bitfields
1790 @itemx -fno-signed-bitfields
1791 @itemx -fno-unsigned-bitfields
1792 @opindex fsigned-bitfields
1793 @opindex funsigned-bitfields
1794 @opindex fno-signed-bitfields
1795 @opindex fno-unsigned-bitfields
1796 These options control whether a bit-field is signed or unsigned, when the
1797 declaration does not use either @code{signed} or @code{unsigned}. By
1798 default, such a bit-field is signed, because this is consistent: the
1799 basic integer types such as @code{int} are signed types.
1802 @node C++ Dialect Options
1803 @section Options Controlling C++ Dialect
1805 @cindex compiler options, C++
1806 @cindex C++ options, command line
1807 @cindex options, C++
1808 This section describes the command-line options that are only meaningful
1809 for C++ programs; but you can also use most of the GNU compiler options
1810 regardless of what language your program is in. For example, you
1811 might compile a file @code{firstClass.C} like this:
1814 g++ -g -frepo -O -c firstClass.C
1818 In this example, only @option{-frepo} is an option meant
1819 only for C++ programs; you can use the other options with any
1820 language supported by GCC@.
1822 Here is a list of options that are @emph{only} for compiling C++ programs:
1826 @item -fabi-version=@var{n}
1827 @opindex fabi-version
1828 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1829 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1830 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1831 the version that conforms most closely to the C++ ABI specification.
1832 Therefore, the ABI obtained using version 0 will change as ABI bugs
1835 The default is version 2.
1837 Version 3 corrects an error in mangling a constant address as a
1840 Version 4 implements a standard mangling for vector types.
1842 See also @option{-Wabi}.
1844 @item -fno-access-control
1845 @opindex fno-access-control
1846 Turn off all access checking. This switch is mainly useful for working
1847 around bugs in the access control code.
1851 Check that the pointer returned by @code{operator new} is non-null
1852 before attempting to modify the storage allocated. This check is
1853 normally unnecessary because the C++ standard specifies that
1854 @code{operator new} will only return @code{0} if it is declared
1855 @samp{throw()}, in which case the compiler will always check the
1856 return value even without this option. In all other cases, when
1857 @code{operator new} has a non-empty exception specification, memory
1858 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1859 @samp{new (nothrow)}.
1861 @item -fconserve-space
1862 @opindex fconserve-space
1863 Put uninitialized or runtime-initialized global variables into the
1864 common segment, as C does. This saves space in the executable at the
1865 cost of not diagnosing duplicate definitions. If you compile with this
1866 flag and your program mysteriously crashes after @code{main()} has
1867 completed, you may have an object that is being destroyed twice because
1868 two definitions were merged.
1870 This option is no longer useful on most targets, now that support has
1871 been added for putting variables into BSS without making them common.
1873 @item -fno-deduce-init-list
1874 @opindex fno-deduce-init-list
1875 Disable deduction of a template type parameter as
1876 std::initializer_list from a brace-enclosed initializer list, i.e.
1879 template <class T> auto forward(T t) -> decltype (realfn (t))
1886 forward(@{1,2@}); // call forward<std::initializer_list<int>>
1890 This option is present because this deduction is an extension to the
1891 current specification in the C++0x working draft, and there was
1892 some concern about potential overload resolution problems.
1894 @item -ffriend-injection
1895 @opindex ffriend-injection
1896 Inject friend functions into the enclosing namespace, so that they are
1897 visible outside the scope of the class in which they are declared.
1898 Friend functions were documented to work this way in the old Annotated
1899 C++ Reference Manual, and versions of G++ before 4.1 always worked
1900 that way. However, in ISO C++ a friend function which is not declared
1901 in an enclosing scope can only be found using argument dependent
1902 lookup. This option causes friends to be injected as they were in
1905 This option is for compatibility, and may be removed in a future
1908 @item -fno-elide-constructors
1909 @opindex fno-elide-constructors
1910 The C++ standard allows an implementation to omit creating a temporary
1911 which is only used to initialize another object of the same type.
1912 Specifying this option disables that optimization, and forces G++ to
1913 call the copy constructor in all cases.
1915 @item -fno-enforce-eh-specs
1916 @opindex fno-enforce-eh-specs
1917 Don't generate code to check for violation of exception specifications
1918 at runtime. This option violates the C++ standard, but may be useful
1919 for reducing code size in production builds, much like defining
1920 @samp{NDEBUG}. This does not give user code permission to throw
1921 exceptions in violation of the exception specifications; the compiler
1922 will still optimize based on the specifications, so throwing an
1923 unexpected exception will result in undefined behavior.
1926 @itemx -fno-for-scope
1928 @opindex fno-for-scope
1929 If @option{-ffor-scope} is specified, the scope of variables declared in
1930 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1931 as specified by the C++ standard.
1932 If @option{-fno-for-scope} is specified, the scope of variables declared in
1933 a @i{for-init-statement} extends to the end of the enclosing scope,
1934 as was the case in old versions of G++, and other (traditional)
1935 implementations of C++.
1937 The default if neither flag is given to follow the standard,
1938 but to allow and give a warning for old-style code that would
1939 otherwise be invalid, or have different behavior.
1941 @item -fno-gnu-keywords
1942 @opindex fno-gnu-keywords
1943 Do not recognize @code{typeof} as a keyword, so that code can use this
1944 word as an identifier. You can use the keyword @code{__typeof__} instead.
1945 @option{-ansi} implies @option{-fno-gnu-keywords}.
1947 @item -fno-implicit-templates
1948 @opindex fno-implicit-templates
1949 Never emit code for non-inline templates which are instantiated
1950 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1951 @xref{Template Instantiation}, for more information.
1953 @item -fno-implicit-inline-templates
1954 @opindex fno-implicit-inline-templates
1955 Don't emit code for implicit instantiations of inline templates, either.
1956 The default is to handle inlines differently so that compiles with and
1957 without optimization will need the same set of explicit instantiations.
1959 @item -fno-implement-inlines
1960 @opindex fno-implement-inlines
1961 To save space, do not emit out-of-line copies of inline functions
1962 controlled by @samp{#pragma implementation}. This will cause linker
1963 errors if these functions are not inlined everywhere they are called.
1965 @item -fms-extensions
1966 @opindex fms-extensions
1967 Disable pedantic warnings about constructs used in MFC, such as implicit
1968 int and getting a pointer to member function via non-standard syntax.
1970 @item -fno-nonansi-builtins
1971 @opindex fno-nonansi-builtins
1972 Disable built-in declarations of functions that are not mandated by
1973 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1974 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1977 @opindex fnothrow-opt
1978 Treat a @code{throw()} exception specification as though it were a
1979 @code{noexcept} specification to reduce or eliminate the text size
1980 overhead relative to a function with no exception specification. If
1981 the function has local variables of types with non-trivial
1982 destructors, the exception specification will actually make the
1983 function smaller because the EH cleanups for those variables can be
1984 optimized away. The semantic effect is that an exception thrown out of
1985 a function with such an exception specification will result in a call
1986 to @code{terminate} rather than @code{unexpected}.
1988 @item -fno-operator-names
1989 @opindex fno-operator-names
1990 Do not treat the operator name keywords @code{and}, @code{bitand},
1991 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1992 synonyms as keywords.
1994 @item -fno-optional-diags
1995 @opindex fno-optional-diags
1996 Disable diagnostics that the standard says a compiler does not need to
1997 issue. Currently, the only such diagnostic issued by G++ is the one for
1998 a name having multiple meanings within a class.
2001 @opindex fpermissive
2002 Downgrade some diagnostics about nonconformant code from errors to
2003 warnings. Thus, using @option{-fpermissive} will allow some
2004 nonconforming code to compile.
2006 @item -fno-pretty-templates
2007 @opindex fno-pretty-templates
2008 When an error message refers to a specialization of a function
2009 template, the compiler will normally print the signature of the
2010 template followed by the template arguments and any typedefs or
2011 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2012 rather than @code{void f(int)}) so that it's clear which template is
2013 involved. When an error message refers to a specialization of a class
2014 template, the compiler will omit any template arguments which match
2015 the default template arguments for that template. If either of these
2016 behaviors make it harder to understand the error message rather than
2017 easier, using @option{-fno-pretty-templates} will disable them.
2021 Enable automatic template instantiation at link time. This option also
2022 implies @option{-fno-implicit-templates}. @xref{Template
2023 Instantiation}, for more information.
2027 Disable generation of information about every class with virtual
2028 functions for use by the C++ runtime type identification features
2029 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
2030 of the language, you can save some space by using this flag. Note that
2031 exception handling uses the same information, but it will generate it as
2032 needed. The @samp{dynamic_cast} operator can still be used for casts that
2033 do not require runtime type information, i.e.@: casts to @code{void *} or to
2034 unambiguous base classes.
2038 Emit statistics about front-end processing at the end of the compilation.
2039 This information is generally only useful to the G++ development team.
2041 @item -fstrict-enums
2042 @opindex fstrict-enums
2043 Allow the compiler to optimize using the assumption that a value of
2044 enumeration type can only be one of the values of the enumeration (as
2045 defined in the C++ standard; basically, a value which can be
2046 represented in the minimum number of bits needed to represent all the
2047 enumerators). This assumption may not be valid if the program uses a
2048 cast to convert an arbitrary integer value to the enumeration type.
2050 @item -ftemplate-depth=@var{n}
2051 @opindex ftemplate-depth
2052 Set the maximum instantiation depth for template classes to @var{n}.
2053 A limit on the template instantiation depth is needed to detect
2054 endless recursions during template class instantiation. ANSI/ISO C++
2055 conforming programs must not rely on a maximum depth greater than 17
2056 (changed to 1024 in C++0x).
2058 @item -fno-threadsafe-statics
2059 @opindex fno-threadsafe-statics
2060 Do not emit the extra code to use the routines specified in the C++
2061 ABI for thread-safe initialization of local statics. You can use this
2062 option to reduce code size slightly in code that doesn't need to be
2065 @item -fuse-cxa-atexit
2066 @opindex fuse-cxa-atexit
2067 Register destructors for objects with static storage duration with the
2068 @code{__cxa_atexit} function rather than the @code{atexit} function.
2069 This option is required for fully standards-compliant handling of static
2070 destructors, but will only work if your C library supports
2071 @code{__cxa_atexit}.
2073 @item -fno-use-cxa-get-exception-ptr
2074 @opindex fno-use-cxa-get-exception-ptr
2075 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2076 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
2077 if the runtime routine is not available.
2079 @item -fvisibility-inlines-hidden
2080 @opindex fvisibility-inlines-hidden
2081 This switch declares that the user does not attempt to compare
2082 pointers to inline methods where the addresses of the two functions
2083 were taken in different shared objects.
2085 The effect of this is that GCC may, effectively, mark inline methods with
2086 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2087 appear in the export table of a DSO and do not require a PLT indirection
2088 when used within the DSO@. Enabling this option can have a dramatic effect
2089 on load and link times of a DSO as it massively reduces the size of the
2090 dynamic export table when the library makes heavy use of templates.
2092 The behavior of this switch is not quite the same as marking the
2093 methods as hidden directly, because it does not affect static variables
2094 local to the function or cause the compiler to deduce that
2095 the function is defined in only one shared object.
2097 You may mark a method as having a visibility explicitly to negate the
2098 effect of the switch for that method. For example, if you do want to
2099 compare pointers to a particular inline method, you might mark it as
2100 having default visibility. Marking the enclosing class with explicit
2101 visibility will have no effect.
2103 Explicitly instantiated inline methods are unaffected by this option
2104 as their linkage might otherwise cross a shared library boundary.
2105 @xref{Template Instantiation}.
2107 @item -fvisibility-ms-compat
2108 @opindex fvisibility-ms-compat
2109 This flag attempts to use visibility settings to make GCC's C++
2110 linkage model compatible with that of Microsoft Visual Studio.
2112 The flag makes these changes to GCC's linkage model:
2116 It sets the default visibility to @code{hidden}, like
2117 @option{-fvisibility=hidden}.
2120 Types, but not their members, are not hidden by default.
2123 The One Definition Rule is relaxed for types without explicit
2124 visibility specifications which are defined in more than one different
2125 shared object: those declarations are permitted if they would have
2126 been permitted when this option was not used.
2129 In new code it is better to use @option{-fvisibility=hidden} and
2130 export those classes which are intended to be externally visible.
2131 Unfortunately it is possible for code to rely, perhaps accidentally,
2132 on the Visual Studio behavior.
2134 Among the consequences of these changes are that static data members
2135 of the same type with the same name but defined in different shared
2136 objects will be different, so changing one will not change the other;
2137 and that pointers to function members defined in different shared
2138 objects may not compare equal. When this flag is given, it is a
2139 violation of the ODR to define types with the same name differently.
2143 Do not use weak symbol support, even if it is provided by the linker.
2144 By default, G++ will use weak symbols if they are available. This
2145 option exists only for testing, and should not be used by end-users;
2146 it will result in inferior code and has no benefits. This option may
2147 be removed in a future release of G++.
2151 Do not search for header files in the standard directories specific to
2152 C++, but do still search the other standard directories. (This option
2153 is used when building the C++ library.)
2156 In addition, these optimization, warning, and code generation options
2157 have meanings only for C++ programs:
2160 @item -fno-default-inline
2161 @opindex fno-default-inline
2162 Do not assume @samp{inline} for functions defined inside a class scope.
2163 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2164 functions will have linkage like inline functions; they just won't be
2167 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2170 Warn when G++ generates code that is probably not compatible with the
2171 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2172 all such cases, there are probably some cases that are not warned about,
2173 even though G++ is generating incompatible code. There may also be
2174 cases where warnings are emitted even though the code that is generated
2177 You should rewrite your code to avoid these warnings if you are
2178 concerned about the fact that code generated by G++ may not be binary
2179 compatible with code generated by other compilers.
2181 The known incompatibilities in @option{-fabi-version=2} (the default) include:
2186 A template with a non-type template parameter of reference type is
2187 mangled incorrectly:
2190 template <int &> struct S @{@};
2194 This is fixed in @option{-fabi-version=3}.
2197 SIMD vector types declared using @code{__attribute ((vector_size))} are
2198 mangled in a non-standard way that does not allow for overloading of
2199 functions taking vectors of different sizes.
2201 The mangling is changed in @option{-fabi-version=4}.
2204 The known incompatibilities in @option{-fabi-version=1} include:
2209 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2210 pack data into the same byte as a base class. For example:
2213 struct A @{ virtual void f(); int f1 : 1; @};
2214 struct B : public A @{ int f2 : 1; @};
2218 In this case, G++ will place @code{B::f2} into the same byte
2219 as@code{A::f1}; other compilers will not. You can avoid this problem
2220 by explicitly padding @code{A} so that its size is a multiple of the
2221 byte size on your platform; that will cause G++ and other compilers to
2222 layout @code{B} identically.
2225 Incorrect handling of tail-padding for virtual bases. G++ does not use
2226 tail padding when laying out virtual bases. For example:
2229 struct A @{ virtual void f(); char c1; @};
2230 struct B @{ B(); char c2; @};
2231 struct C : public A, public virtual B @{@};
2235 In this case, G++ will not place @code{B} into the tail-padding for
2236 @code{A}; other compilers will. You can avoid this problem by
2237 explicitly padding @code{A} so that its size is a multiple of its
2238 alignment (ignoring virtual base classes); that will cause G++ and other
2239 compilers to layout @code{C} identically.
2242 Incorrect handling of bit-fields with declared widths greater than that
2243 of their underlying types, when the bit-fields appear in a union. For
2247 union U @{ int i : 4096; @};
2251 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2252 union too small by the number of bits in an @code{int}.
2255 Empty classes can be placed at incorrect offsets. For example:
2265 struct C : public B, public A @{@};
2269 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2270 it should be placed at offset zero. G++ mistakenly believes that the
2271 @code{A} data member of @code{B} is already at offset zero.
2274 Names of template functions whose types involve @code{typename} or
2275 template template parameters can be mangled incorrectly.
2278 template <typename Q>
2279 void f(typename Q::X) @{@}
2281 template <template <typename> class Q>
2282 void f(typename Q<int>::X) @{@}
2286 Instantiations of these templates may be mangled incorrectly.
2290 It also warns psABI related changes. The known psABI changes at this
2296 For SYSV/x86-64, when passing union with long double, it is changed to
2297 pass in memory as specified in psABI. For example:
2307 @code{union U} will always be passed in memory.
2311 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2312 @opindex Wctor-dtor-privacy
2313 @opindex Wno-ctor-dtor-privacy
2314 Warn when a class seems unusable because all the constructors or
2315 destructors in that class are private, and it has neither friends nor
2316 public static member functions.
2318 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2320 @opindex Wno-noexcept
2321 Warn when a noexcept-expression evaluates to false because of a call
2322 to a function that does not have a non-throwing exception
2323 specification (i.e. @samp{throw()} or @samp{noexcept}) but is known by
2324 the compiler to never throw an exception.
2326 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2327 @opindex Wnon-virtual-dtor
2328 @opindex Wno-non-virtual-dtor
2329 Warn when a class has virtual functions and accessible non-virtual
2330 destructor, in which case it would be possible but unsafe to delete
2331 an instance of a derived class through a pointer to the base class.
2332 This warning is also enabled if -Weffc++ is specified.
2334 @item -Wreorder @r{(C++ and Objective-C++ only)}
2336 @opindex Wno-reorder
2337 @cindex reordering, warning
2338 @cindex warning for reordering of member initializers
2339 Warn when the order of member initializers given in the code does not
2340 match the order in which they must be executed. For instance:
2346 A(): j (0), i (1) @{ @}
2350 The compiler will rearrange the member initializers for @samp{i}
2351 and @samp{j} to match the declaration order of the members, emitting
2352 a warning to that effect. This warning is enabled by @option{-Wall}.
2355 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2358 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2361 Warn about violations of the following style guidelines from Scott Meyers'
2362 @cite{Effective C++} book:
2366 Item 11: Define a copy constructor and an assignment operator for classes
2367 with dynamically allocated memory.
2370 Item 12: Prefer initialization to assignment in constructors.
2373 Item 14: Make destructors virtual in base classes.
2376 Item 15: Have @code{operator=} return a reference to @code{*this}.
2379 Item 23: Don't try to return a reference when you must return an object.
2383 Also warn about violations of the following style guidelines from
2384 Scott Meyers' @cite{More Effective C++} book:
2388 Item 6: Distinguish between prefix and postfix forms of increment and
2389 decrement operators.
2392 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2396 When selecting this option, be aware that the standard library
2397 headers do not obey all of these guidelines; use @samp{grep -v}
2398 to filter out those warnings.
2400 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2401 @opindex Wstrict-null-sentinel
2402 @opindex Wno-strict-null-sentinel
2403 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2404 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2405 to @code{__null}. Although it is a null pointer constant not a null pointer,
2406 it is guaranteed to be of the same size as a pointer. But this use is
2407 not portable across different compilers.
2409 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2410 @opindex Wno-non-template-friend
2411 @opindex Wnon-template-friend
2412 Disable warnings when non-templatized friend functions are declared
2413 within a template. Since the advent of explicit template specification
2414 support in G++, if the name of the friend is an unqualified-id (i.e.,
2415 @samp{friend foo(int)}), the C++ language specification demands that the
2416 friend declare or define an ordinary, nontemplate function. (Section
2417 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2418 could be interpreted as a particular specialization of a templatized
2419 function. Because this non-conforming behavior is no longer the default
2420 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2421 check existing code for potential trouble spots and is on by default.
2422 This new compiler behavior can be turned off with
2423 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2424 but disables the helpful warning.
2426 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2427 @opindex Wold-style-cast
2428 @opindex Wno-old-style-cast
2429 Warn if an old-style (C-style) cast to a non-void type is used within
2430 a C++ program. The new-style casts (@samp{dynamic_cast},
2431 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2432 less vulnerable to unintended effects and much easier to search for.
2434 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2435 @opindex Woverloaded-virtual
2436 @opindex Wno-overloaded-virtual
2437 @cindex overloaded virtual function, warning
2438 @cindex warning for overloaded virtual function
2439 Warn when a function declaration hides virtual functions from a
2440 base class. For example, in:
2447 struct B: public A @{
2452 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2460 will fail to compile.
2462 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2463 @opindex Wno-pmf-conversions
2464 @opindex Wpmf-conversions
2465 Disable the diagnostic for converting a bound pointer to member function
2468 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2469 @opindex Wsign-promo
2470 @opindex Wno-sign-promo
2471 Warn when overload resolution chooses a promotion from unsigned or
2472 enumerated type to a signed type, over a conversion to an unsigned type of
2473 the same size. Previous versions of G++ would try to preserve
2474 unsignedness, but the standard mandates the current behavior.
2479 A& operator = (int);
2489 In this example, G++ will synthesize a default @samp{A& operator =
2490 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2493 @node Objective-C and Objective-C++ Dialect Options
2494 @section Options Controlling Objective-C and Objective-C++ Dialects
2496 @cindex compiler options, Objective-C and Objective-C++
2497 @cindex Objective-C and Objective-C++ options, command line
2498 @cindex options, Objective-C and Objective-C++
2499 (NOTE: This manual does not describe the Objective-C and Objective-C++
2500 languages themselves. See @xref{Standards,,Language Standards
2501 Supported by GCC}, for references.)
2503 This section describes the command-line options that are only meaningful
2504 for Objective-C and Objective-C++ programs, but you can also use most of
2505 the language-independent GNU compiler options.
2506 For example, you might compile a file @code{some_class.m} like this:
2509 gcc -g -fgnu-runtime -O -c some_class.m
2513 In this example, @option{-fgnu-runtime} is an option meant only for
2514 Objective-C and Objective-C++ programs; you can use the other options with
2515 any language supported by GCC@.
2517 Note that since Objective-C is an extension of the C language, Objective-C
2518 compilations may also use options specific to the C front-end (e.g.,
2519 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2520 C++-specific options (e.g., @option{-Wabi}).
2522 Here is a list of options that are @emph{only} for compiling Objective-C
2523 and Objective-C++ programs:
2526 @item -fconstant-string-class=@var{class-name}
2527 @opindex fconstant-string-class
2528 Use @var{class-name} as the name of the class to instantiate for each
2529 literal string specified with the syntax @code{@@"@dots{}"}. The default
2530 class name is @code{NXConstantString} if the GNU runtime is being used, and
2531 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2532 @option{-fconstant-cfstrings} option, if also present, will override the
2533 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2534 to be laid out as constant CoreFoundation strings.
2537 @opindex fgnu-runtime
2538 Generate object code compatible with the standard GNU Objective-C
2539 runtime. This is the default for most types of systems.
2541 @item -fnext-runtime
2542 @opindex fnext-runtime
2543 Generate output compatible with the NeXT runtime. This is the default
2544 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2545 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2548 @item -fno-nil-receivers
2549 @opindex fno-nil-receivers
2550 Assume that all Objective-C message dispatches (@code{[receiver
2551 message:arg]}) in this translation unit ensure that the receiver is
2552 not @code{nil}. This allows for more efficient entry points in the
2553 runtime to be used. Currently, this option is only available in
2554 conjunction with the NeXT runtime on Mac OS X 10.3 and later.
2556 @item -fobjc-call-cxx-cdtors
2557 @opindex fobjc-call-cxx-cdtors
2558 For each Objective-C class, check if any of its instance variables is a
2559 C++ object with a non-trivial default constructor. If so, synthesize a
2560 special @code{- (id) .cxx_construct} instance method that will run
2561 non-trivial default constructors on any such instance variables, in order,
2562 and then return @code{self}. Similarly, check if any instance variable
2563 is a C++ object with a non-trivial destructor, and if so, synthesize a
2564 special @code{- (void) .cxx_destruct} method that will run
2565 all such default destructors, in reverse order.
2567 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
2568 methods thusly generated will only operate on instance variables
2569 declared in the current Objective-C class, and not those inherited
2570 from superclasses. It is the responsibility of the Objective-C
2571 runtime to invoke all such methods in an object's inheritance
2572 hierarchy. The @code{- (id) .cxx_construct} methods will be invoked
2573 by the runtime immediately after a new object instance is allocated;
2574 the @code{- (void) .cxx_destruct} methods will be invoked immediately
2575 before the runtime deallocates an object instance.
2577 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2578 support for invoking the @code{- (id) .cxx_construct} and
2579 @code{- (void) .cxx_destruct} methods.
2581 @item -fobjc-direct-dispatch
2582 @opindex fobjc-direct-dispatch
2583 Allow fast jumps to the message dispatcher. On Darwin this is
2584 accomplished via the comm page.
2586 @item -fobjc-exceptions
2587 @opindex fobjc-exceptions
2588 Enable syntactic support for structured exception handling in
2589 Objective-C, similar to what is offered by C++ and Java. This option
2590 is required to use the Objective-C keywords @code{@@try},
2591 @code{@@throw}, @code{@@catch}, @code{@@finally} and
2592 @code{@@synchronized}. This option is available with both the GNU
2593 runtime and the NeXT runtime (but not available in conjunction with
2594 the NeXT runtime on Mac OS X 10.2 and earlier).
2598 Enable garbage collection (GC) in Objective-C and Objective-C++
2599 programs. This option is only available with the NeXT runtime; the
2600 GNU runtime has a different garbage collection implementation that
2601 does not require special compiler flags.
2603 @item -fobjc-std=objc1
2605 Conform to the language syntax of Objective-C 1.0, the language
2606 recognized by GCC 4.0. This only affects the Objective-C additions to
2607 the C/C++ language; it does not affect conformance to C/C++ standards,
2608 which is controlled by the separate C/C++ dialect option flags. When
2609 this option is used with the Objective-C or Objective-C++ compiler,
2610 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
2611 This is useful if you need to make sure that your Objective-C code can
2612 be compiled with older versions of GCC.
2614 @item -freplace-objc-classes
2615 @opindex freplace-objc-classes
2616 Emit a special marker instructing @command{ld(1)} not to statically link in
2617 the resulting object file, and allow @command{dyld(1)} to load it in at
2618 run time instead. This is used in conjunction with the Fix-and-Continue
2619 debugging mode, where the object file in question may be recompiled and
2620 dynamically reloaded in the course of program execution, without the need
2621 to restart the program itself. Currently, Fix-and-Continue functionality
2622 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2627 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2628 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2629 compile time) with static class references that get initialized at load time,
2630 which improves run-time performance. Specifying the @option{-fzero-link} flag
2631 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2632 to be retained. This is useful in Zero-Link debugging mode, since it allows
2633 for individual class implementations to be modified during program execution.
2634 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
2635 regardless of command line options.
2639 Dump interface declarations for all classes seen in the source file to a
2640 file named @file{@var{sourcename}.decl}.
2642 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2643 @opindex Wassign-intercept
2644 @opindex Wno-assign-intercept
2645 Warn whenever an Objective-C assignment is being intercepted by the
2648 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2649 @opindex Wno-protocol
2651 If a class is declared to implement a protocol, a warning is issued for
2652 every method in the protocol that is not implemented by the class. The
2653 default behavior is to issue a warning for every method not explicitly
2654 implemented in the class, even if a method implementation is inherited
2655 from the superclass. If you use the @option{-Wno-protocol} option, then
2656 methods inherited from the superclass are considered to be implemented,
2657 and no warning is issued for them.
2659 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2661 @opindex Wno-selector
2662 Warn if multiple methods of different types for the same selector are
2663 found during compilation. The check is performed on the list of methods
2664 in the final stage of compilation. Additionally, a check is performed
2665 for each selector appearing in a @code{@@selector(@dots{})}
2666 expression, and a corresponding method for that selector has been found
2667 during compilation. Because these checks scan the method table only at
2668 the end of compilation, these warnings are not produced if the final
2669 stage of compilation is not reached, for example because an error is
2670 found during compilation, or because the @option{-fsyntax-only} option is
2673 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2674 @opindex Wstrict-selector-match
2675 @opindex Wno-strict-selector-match
2676 Warn if multiple methods with differing argument and/or return types are
2677 found for a given selector when attempting to send a message using this
2678 selector to a receiver of type @code{id} or @code{Class}. When this flag
2679 is off (which is the default behavior), the compiler will omit such warnings
2680 if any differences found are confined to types which share the same size
2683 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2684 @opindex Wundeclared-selector
2685 @opindex Wno-undeclared-selector
2686 Warn if a @code{@@selector(@dots{})} expression referring to an
2687 undeclared selector is found. A selector is considered undeclared if no
2688 method with that name has been declared before the
2689 @code{@@selector(@dots{})} expression, either explicitly in an
2690 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2691 an @code{@@implementation} section. This option always performs its
2692 checks as soon as a @code{@@selector(@dots{})} expression is found,
2693 while @option{-Wselector} only performs its checks in the final stage of
2694 compilation. This also enforces the coding style convention
2695 that methods and selectors must be declared before being used.
2697 @item -print-objc-runtime-info
2698 @opindex print-objc-runtime-info
2699 Generate C header describing the largest structure that is passed by
2704 @node Language Independent Options
2705 @section Options to Control Diagnostic Messages Formatting
2706 @cindex options to control diagnostics formatting
2707 @cindex diagnostic messages
2708 @cindex message formatting
2710 Traditionally, diagnostic messages have been formatted irrespective of
2711 the output device's aspect (e.g.@: its width, @dots{}). The options described
2712 below can be used to control the diagnostic messages formatting
2713 algorithm, e.g.@: how many characters per line, how often source location
2714 information should be reported. Right now, only the C++ front end can
2715 honor these options. However it is expected, in the near future, that
2716 the remaining front ends would be able to digest them correctly.
2719 @item -fmessage-length=@var{n}
2720 @opindex fmessage-length
2721 Try to format error messages so that they fit on lines of about @var{n}
2722 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2723 the front ends supported by GCC@. If @var{n} is zero, then no
2724 line-wrapping will be done; each error message will appear on a single
2727 @opindex fdiagnostics-show-location
2728 @item -fdiagnostics-show-location=once
2729 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2730 reporter to emit @emph{once} source location information; that is, in
2731 case the message is too long to fit on a single physical line and has to
2732 be wrapped, the source location won't be emitted (as prefix) again,
2733 over and over, in subsequent continuation lines. This is the default
2736 @item -fdiagnostics-show-location=every-line
2737 Only meaningful in line-wrapping mode. Instructs the diagnostic
2738 messages reporter to emit the same source location information (as
2739 prefix) for physical lines that result from the process of breaking
2740 a message which is too long to fit on a single line.
2742 @item -fdiagnostics-show-option
2743 @opindex fdiagnostics-show-option
2744 This option instructs the diagnostic machinery to add text to each
2745 diagnostic emitted, which indicates which command line option directly
2746 controls that diagnostic, when such an option is known to the
2747 diagnostic machinery.
2749 @item -Wcoverage-mismatch
2750 @opindex Wcoverage-mismatch
2751 Warn if feedback profiles do not match when using the
2752 @option{-fprofile-use} option.
2753 If a source file was changed between @option{-fprofile-gen} and
2754 @option{-fprofile-use}, the files with the profile feedback can fail
2755 to match the source file and GCC can not use the profile feedback
2756 information. By default, this warning is enabled and is treated as an
2757 error. @option{-Wno-coverage-mismatch} can be used to disable the
2758 warning or @option{-Wno-error=coverage-mismatch} can be used to
2759 disable the error. Disable the error for this warning can result in
2760 poorly optimized code, so disabling the error is useful only in the
2761 case of very minor changes such as bug fixes to an existing code-base.
2762 Completely disabling the warning is not recommended.
2766 @node Warning Options
2767 @section Options to Request or Suppress Warnings
2768 @cindex options to control warnings
2769 @cindex warning messages
2770 @cindex messages, warning
2771 @cindex suppressing warnings
2773 Warnings are diagnostic messages that report constructions which
2774 are not inherently erroneous but which are risky or suggest there
2775 may have been an error.
2777 The following language-independent options do not enable specific
2778 warnings but control the kinds of diagnostics produced by GCC.
2781 @cindex syntax checking
2783 @opindex fsyntax-only
2784 Check the code for syntax errors, but don't do anything beyond that.
2786 @item -fmax-errors=@var{n}
2787 @opindex fmax-errors
2788 Limits the maximum number of error messages to @var{n}, at which point
2789 GCC bails out rather than attempting to continue processing the source
2790 code. If @var{n} is 0 (the default), there is no limit on the number
2791 of error messages produced. If @option{-Wfatal-errors} is also
2792 specified, then @option{-Wfatal-errors} takes precedence over this
2797 Inhibit all warning messages.
2802 Make all warnings into errors.
2807 Make the specified warning into an error. The specifier for a warning
2808 is appended, for example @option{-Werror=switch} turns the warnings
2809 controlled by @option{-Wswitch} into errors. This switch takes a
2810 negative form, to be used to negate @option{-Werror} for specific
2811 warnings, for example @option{-Wno-error=switch} makes
2812 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2813 is in effect. You can use the @option{-fdiagnostics-show-option}
2814 option to have each controllable warning amended with the option which
2815 controls it, to determine what to use with this option.
2817 Note that specifying @option{-Werror=}@var{foo} automatically implies
2818 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2821 @item -Wfatal-errors
2822 @opindex Wfatal-errors
2823 @opindex Wno-fatal-errors
2824 This option causes the compiler to abort compilation on the first error
2825 occurred rather than trying to keep going and printing further error
2830 You can request many specific warnings with options beginning
2831 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2832 implicit declarations. Each of these specific warning options also
2833 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2834 example, @option{-Wno-implicit}. This manual lists only one of the
2835 two forms, whichever is not the default. For further,
2836 language-specific options also refer to @ref{C++ Dialect Options} and
2837 @ref{Objective-C and Objective-C++ Dialect Options}.
2839 When an unrecognized warning option is requested (e.g.,
2840 @option{-Wunknown-warning}), GCC will emit a diagnostic stating
2841 that the option is not recognized. However, if the @option{-Wno-} form
2842 is used, the behavior is slightly different: No diagnostic will be
2843 produced for @option{-Wno-unknown-warning} unless other diagnostics
2844 are being produced. This allows the use of new @option{-Wno-} options
2845 with old compilers, but if something goes wrong, the compiler will
2846 warn that an unrecognized option was used.
2851 Issue all the warnings demanded by strict ISO C and ISO C++;
2852 reject all programs that use forbidden extensions, and some other
2853 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2854 version of the ISO C standard specified by any @option{-std} option used.
2856 Valid ISO C and ISO C++ programs should compile properly with or without
2857 this option (though a rare few will require @option{-ansi} or a
2858 @option{-std} option specifying the required version of ISO C)@. However,
2859 without this option, certain GNU extensions and traditional C and C++
2860 features are supported as well. With this option, they are rejected.
2862 @option{-pedantic} does not cause warning messages for use of the
2863 alternate keywords whose names begin and end with @samp{__}. Pedantic
2864 warnings are also disabled in the expression that follows
2865 @code{__extension__}. However, only system header files should use
2866 these escape routes; application programs should avoid them.
2867 @xref{Alternate Keywords}.
2869 Some users try to use @option{-pedantic} to check programs for strict ISO
2870 C conformance. They soon find that it does not do quite what they want:
2871 it finds some non-ISO practices, but not all---only those for which
2872 ISO C @emph{requires} a diagnostic, and some others for which
2873 diagnostics have been added.
2875 A feature to report any failure to conform to ISO C might be useful in
2876 some instances, but would require considerable additional work and would
2877 be quite different from @option{-pedantic}. We don't have plans to
2878 support such a feature in the near future.
2880 Where the standard specified with @option{-std} represents a GNU
2881 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
2882 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2883 extended dialect is based. Warnings from @option{-pedantic} are given
2884 where they are required by the base standard. (It would not make sense
2885 for such warnings to be given only for features not in the specified GNU
2886 C dialect, since by definition the GNU dialects of C include all
2887 features the compiler supports with the given option, and there would be
2888 nothing to warn about.)
2890 @item -pedantic-errors
2891 @opindex pedantic-errors
2892 Like @option{-pedantic}, except that errors are produced rather than
2898 This enables all the warnings about constructions that some users
2899 consider questionable, and that are easy to avoid (or modify to
2900 prevent the warning), even in conjunction with macros. This also
2901 enables some language-specific warnings described in @ref{C++ Dialect
2902 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2904 @option{-Wall} turns on the following warning flags:
2906 @gccoptlist{-Waddress @gol
2907 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2909 -Wchar-subscripts @gol
2910 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
2911 -Wimplicit-int @r{(C and Objective-C only)} @gol
2912 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
2915 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2916 -Wmissing-braces @gol
2922 -Wsequence-point @gol
2923 -Wsign-compare @r{(only in C++)} @gol
2924 -Wstrict-aliasing @gol
2925 -Wstrict-overflow=1 @gol
2928 -Wuninitialized @gol
2929 -Wunknown-pragmas @gol
2930 -Wunused-function @gol
2933 -Wunused-variable @gol
2934 -Wvolatile-register-var @gol
2937 Note that some warning flags are not implied by @option{-Wall}. Some of
2938 them warn about constructions that users generally do not consider
2939 questionable, but which occasionally you might wish to check for;
2940 others warn about constructions that are necessary or hard to avoid in
2941 some cases, and there is no simple way to modify the code to suppress
2942 the warning. Some of them are enabled by @option{-Wextra} but many of
2943 them must be enabled individually.
2949 This enables some extra warning flags that are not enabled by
2950 @option{-Wall}. (This option used to be called @option{-W}. The older
2951 name is still supported, but the newer name is more descriptive.)
2953 @gccoptlist{-Wclobbered @gol
2955 -Wignored-qualifiers @gol
2956 -Wmissing-field-initializers @gol
2957 -Wmissing-parameter-type @r{(C only)} @gol
2958 -Wold-style-declaration @r{(C only)} @gol
2959 -Woverride-init @gol
2962 -Wuninitialized @gol
2963 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2964 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2967 The option @option{-Wextra} also prints warning messages for the
2973 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2974 @samp{>}, or @samp{>=}.
2977 (C++ only) An enumerator and a non-enumerator both appear in a
2978 conditional expression.
2981 (C++ only) Ambiguous virtual bases.
2984 (C++ only) Subscripting an array which has been declared @samp{register}.
2987 (C++ only) Taking the address of a variable which has been declared
2991 (C++ only) A base class is not initialized in a derived class' copy
2996 @item -Wchar-subscripts
2997 @opindex Wchar-subscripts
2998 @opindex Wno-char-subscripts
2999 Warn if an array subscript has type @code{char}. This is a common cause
3000 of error, as programmers often forget that this type is signed on some
3002 This warning is enabled by @option{-Wall}.
3006 @opindex Wno-comment
3007 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3008 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3009 This warning is enabled by @option{-Wall}.
3012 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3014 Suppress warning messages emitted by @code{#warning} directives.
3016 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3017 @opindex Wdouble-promotion
3018 @opindex Wno-double-promotion
3019 Give a warning when a value of type @code{float} is implicitly
3020 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3021 floating-point unit implement @code{float} in hardware, but emulate
3022 @code{double} in software. On such a machine, doing computations
3023 using @code{double} values is much more expensive because of the
3024 overhead required for software emulation.
3026 It is easy to accidentally do computations with @code{double} because
3027 floating-point literals are implicitly of type @code{double}. For
3031 float area(float radius)
3033 return 3.14159 * radius * radius;
3037 the compiler will perform the entire computation with @code{double}
3038 because the floating-point literal is a @code{double}.
3043 @opindex ffreestanding
3044 @opindex fno-builtin
3045 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3046 the arguments supplied have types appropriate to the format string
3047 specified, and that the conversions specified in the format string make
3048 sense. This includes standard functions, and others specified by format
3049 attributes (@pxref{Function Attributes}), in the @code{printf},
3050 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3051 not in the C standard) families (or other target-specific families).
3052 Which functions are checked without format attributes having been
3053 specified depends on the standard version selected, and such checks of
3054 functions without the attribute specified are disabled by
3055 @option{-ffreestanding} or @option{-fno-builtin}.
3057 The formats are checked against the format features supported by GNU
3058 libc version 2.2. These include all ISO C90 and C99 features, as well
3059 as features from the Single Unix Specification and some BSD and GNU
3060 extensions. Other library implementations may not support all these
3061 features; GCC does not support warning about features that go beyond a
3062 particular library's limitations. However, if @option{-pedantic} is used
3063 with @option{-Wformat}, warnings will be given about format features not
3064 in the selected standard version (but not for @code{strfmon} formats,
3065 since those are not in any version of the C standard). @xref{C Dialect
3066 Options,,Options Controlling C Dialect}.
3068 Since @option{-Wformat} also checks for null format arguments for
3069 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
3071 @option{-Wformat} is included in @option{-Wall}. For more control over some
3072 aspects of format checking, the options @option{-Wformat-y2k},
3073 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
3074 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
3075 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
3078 @opindex Wformat-y2k
3079 @opindex Wno-format-y2k
3080 If @option{-Wformat} is specified, also warn about @code{strftime}
3081 formats which may yield only a two-digit year.
3083 @item -Wno-format-contains-nul
3084 @opindex Wno-format-contains-nul
3085 @opindex Wformat-contains-nul
3086 If @option{-Wformat} is specified, do not warn about format strings that
3089 @item -Wno-format-extra-args
3090 @opindex Wno-format-extra-args
3091 @opindex Wformat-extra-args
3092 If @option{-Wformat} is specified, do not warn about excess arguments to a
3093 @code{printf} or @code{scanf} format function. The C standard specifies
3094 that such arguments are ignored.
3096 Where the unused arguments lie between used arguments that are
3097 specified with @samp{$} operand number specifications, normally
3098 warnings are still given, since the implementation could not know what
3099 type to pass to @code{va_arg} to skip the unused arguments. However,
3100 in the case of @code{scanf} formats, this option will suppress the
3101 warning if the unused arguments are all pointers, since the Single
3102 Unix Specification says that such unused arguments are allowed.
3104 @item -Wno-format-zero-length @r{(C and Objective-C only)}
3105 @opindex Wno-format-zero-length
3106 @opindex Wformat-zero-length
3107 If @option{-Wformat} is specified, do not warn about zero-length formats.
3108 The C standard specifies that zero-length formats are allowed.
3110 @item -Wformat-nonliteral
3111 @opindex Wformat-nonliteral
3112 @opindex Wno-format-nonliteral
3113 If @option{-Wformat} is specified, also warn if the format string is not a
3114 string literal and so cannot be checked, unless the format function
3115 takes its format arguments as a @code{va_list}.
3117 @item -Wformat-security
3118 @opindex Wformat-security
3119 @opindex Wno-format-security
3120 If @option{-Wformat} is specified, also warn about uses of format
3121 functions that represent possible security problems. At present, this
3122 warns about calls to @code{printf} and @code{scanf} functions where the
3123 format string is not a string literal and there are no format arguments,
3124 as in @code{printf (foo);}. This may be a security hole if the format
3125 string came from untrusted input and contains @samp{%n}. (This is
3126 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3127 in future warnings may be added to @option{-Wformat-security} that are not
3128 included in @option{-Wformat-nonliteral}.)
3132 @opindex Wno-format=2
3133 Enable @option{-Wformat} plus format checks not included in
3134 @option{-Wformat}. Currently equivalent to @samp{-Wformat
3135 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
3137 @item -Wnonnull @r{(C and Objective-C only)}
3139 @opindex Wno-nonnull
3140 Warn about passing a null pointer for arguments marked as
3141 requiring a non-null value by the @code{nonnull} function attribute.
3143 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3144 can be disabled with the @option{-Wno-nonnull} option.
3146 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3148 @opindex Wno-init-self
3149 Warn about uninitialized variables which are initialized with themselves.
3150 Note this option can only be used with the @option{-Wuninitialized} option.
3152 For example, GCC will warn about @code{i} being uninitialized in the
3153 following snippet only when @option{-Winit-self} has been specified:
3164 @item -Wimplicit-int @r{(C and Objective-C only)}
3165 @opindex Wimplicit-int
3166 @opindex Wno-implicit-int
3167 Warn when a declaration does not specify a type.
3168 This warning is enabled by @option{-Wall}.
3170 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3171 @opindex Wimplicit-function-declaration
3172 @opindex Wno-implicit-function-declaration
3173 Give a warning whenever a function is used before being declared. In
3174 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3175 enabled by default and it is made into an error by
3176 @option{-pedantic-errors}. This warning is also enabled by
3179 @item -Wimplicit @r{(C and Objective-C only)}
3181 @opindex Wno-implicit
3182 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3183 This warning is enabled by @option{-Wall}.
3185 @item -Wignored-qualifiers @r{(C and C++ only)}
3186 @opindex Wignored-qualifiers
3187 @opindex Wno-ignored-qualifiers
3188 Warn if the return type of a function has a type qualifier
3189 such as @code{const}. For ISO C such a type qualifier has no effect,
3190 since the value returned by a function is not an lvalue.
3191 For C++, the warning is only emitted for scalar types or @code{void}.
3192 ISO C prohibits qualified @code{void} return types on function
3193 definitions, so such return types always receive a warning
3194 even without this option.
3196 This warning is also enabled by @option{-Wextra}.
3201 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3202 a function with external linkage, returning int, taking either zero
3203 arguments, two, or three arguments of appropriate types. This warning
3204 is enabled by default in C++ and is enabled by either @option{-Wall}
3205 or @option{-pedantic}.
3207 @item -Wmissing-braces
3208 @opindex Wmissing-braces
3209 @opindex Wno-missing-braces
3210 Warn if an aggregate or union initializer is not fully bracketed. In
3211 the following example, the initializer for @samp{a} is not fully
3212 bracketed, but that for @samp{b} is fully bracketed.
3215 int a[2][2] = @{ 0, 1, 2, 3 @};
3216 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3219 This warning is enabled by @option{-Wall}.
3221 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3222 @opindex Wmissing-include-dirs
3223 @opindex Wno-missing-include-dirs
3224 Warn if a user-supplied include directory does not exist.
3227 @opindex Wparentheses
3228 @opindex Wno-parentheses
3229 Warn if parentheses are omitted in certain contexts, such
3230 as when there is an assignment in a context where a truth value
3231 is expected, or when operators are nested whose precedence people
3232 often get confused about.
3234 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3235 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3236 interpretation from that of ordinary mathematical notation.
3238 Also warn about constructions where there may be confusion to which
3239 @code{if} statement an @code{else} branch belongs. Here is an example of
3254 In C/C++, every @code{else} branch belongs to the innermost possible
3255 @code{if} statement, which in this example is @code{if (b)}. This is
3256 often not what the programmer expected, as illustrated in the above
3257 example by indentation the programmer chose. When there is the
3258 potential for this confusion, GCC will issue a warning when this flag
3259 is specified. To eliminate the warning, add explicit braces around
3260 the innermost @code{if} statement so there is no way the @code{else}
3261 could belong to the enclosing @code{if}. The resulting code would
3278 Also warn for dangerous uses of the
3279 ?: with omitted middle operand GNU extension. When the condition
3280 in the ?: operator is a boolean expression the omitted value will
3281 be always 1. Often the user expects it to be a value computed
3282 inside the conditional expression instead.
3284 This warning is enabled by @option{-Wall}.
3286 @item -Wsequence-point
3287 @opindex Wsequence-point
3288 @opindex Wno-sequence-point
3289 Warn about code that may have undefined semantics because of violations
3290 of sequence point rules in the C and C++ standards.
3292 The C and C++ standards defines the order in which expressions in a C/C++
3293 program are evaluated in terms of @dfn{sequence points}, which represent
3294 a partial ordering between the execution of parts of the program: those
3295 executed before the sequence point, and those executed after it. These
3296 occur after the evaluation of a full expression (one which is not part
3297 of a larger expression), after the evaluation of the first operand of a
3298 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3299 function is called (but after the evaluation of its arguments and the
3300 expression denoting the called function), and in certain other places.
3301 Other than as expressed by the sequence point rules, the order of
3302 evaluation of subexpressions of an expression is not specified. All
3303 these rules describe only a partial order rather than a total order,
3304 since, for example, if two functions are called within one expression
3305 with no sequence point between them, the order in which the functions
3306 are called is not specified. However, the standards committee have
3307 ruled that function calls do not overlap.
3309 It is not specified when between sequence points modifications to the
3310 values of objects take effect. Programs whose behavior depends on this
3311 have undefined behavior; the C and C++ standards specify that ``Between
3312 the previous and next sequence point an object shall have its stored
3313 value modified at most once by the evaluation of an expression.
3314 Furthermore, the prior value shall be read only to determine the value
3315 to be stored.''. If a program breaks these rules, the results on any
3316 particular implementation are entirely unpredictable.
3318 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3319 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3320 diagnosed by this option, and it may give an occasional false positive
3321 result, but in general it has been found fairly effective at detecting
3322 this sort of problem in programs.
3324 The standard is worded confusingly, therefore there is some debate
3325 over the precise meaning of the sequence point rules in subtle cases.
3326 Links to discussions of the problem, including proposed formal
3327 definitions, may be found on the GCC readings page, at
3328 @uref{http://gcc.gnu.org/@/readings.html}.
3330 This warning is enabled by @option{-Wall} for C and C++.
3333 @opindex Wreturn-type
3334 @opindex Wno-return-type
3335 Warn whenever a function is defined with a return-type that defaults
3336 to @code{int}. Also warn about any @code{return} statement with no
3337 return-value in a function whose return-type is not @code{void}
3338 (falling off the end of the function body is considered returning
3339 without a value), and about a @code{return} statement with an
3340 expression in a function whose return-type is @code{void}.
3342 For C++, a function without return type always produces a diagnostic
3343 message, even when @option{-Wno-return-type} is specified. The only
3344 exceptions are @samp{main} and functions defined in system headers.
3346 This warning is enabled by @option{-Wall}.
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. (The presence of a @code{default} label prevents this
3354 warning.) @code{case} labels outside the enumeration range also
3355 provoke warnings when this option is used (even if there is a
3356 @code{default} label).
3357 This warning is enabled by @option{-Wall}.
3359 @item -Wswitch-default
3360 @opindex Wswitch-default
3361 @opindex Wno-switch-default
3362 Warn whenever a @code{switch} statement does not have a @code{default}
3366 @opindex Wswitch-enum
3367 @opindex Wno-switch-enum
3368 Warn whenever a @code{switch} statement has an index of enumerated type
3369 and lacks a @code{case} for one or more of the named codes of that
3370 enumeration. @code{case} labels outside the enumeration range also
3371 provoke warnings when this option is used. The only difference
3372 between @option{-Wswitch} and this option is that this option gives a
3373 warning about an omitted enumeration code even if there is a
3374 @code{default} label.
3376 @item -Wsync-nand @r{(C and C++ only)}
3378 @opindex Wno-sync-nand
3379 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3380 built-in functions are used. These functions changed semantics in GCC 4.4.
3384 @opindex Wno-trigraphs
3385 Warn if any trigraphs are encountered that might change the meaning of
3386 the program (trigraphs within comments are not warned about).
3387 This warning is enabled by @option{-Wall}.
3389 @item -Wunused-but-set-parameter
3390 @opindex Wunused-but-set-parameter
3391 @opindex Wno-unused-but-set-parameter
3392 Warn whenever a function parameter is assigned to, but otherwise unused
3393 (aside from its declaration).
3395 To suppress this warning use the @samp{unused} attribute
3396 (@pxref{Variable Attributes}).
3398 This warning is also enabled by @option{-Wunused} together with
3401 @item -Wunused-but-set-variable
3402 @opindex Wunused-but-set-variable
3403 @opindex Wno-unused-but-set-variable
3404 Warn whenever a local variable is assigned to, but otherwise unused
3405 (aside from its declaration).
3406 This warning is enabled by @option{-Wall}.
3408 To suppress this warning use the @samp{unused} attribute
3409 (@pxref{Variable Attributes}).
3411 This warning is also enabled by @option{-Wunused}, which is enabled
3414 @item -Wunused-function
3415 @opindex Wunused-function
3416 @opindex Wno-unused-function
3417 Warn whenever a static function is declared but not defined or a
3418 non-inline static function is unused.
3419 This warning is enabled by @option{-Wall}.
3421 @item -Wunused-label
3422 @opindex Wunused-label
3423 @opindex Wno-unused-label
3424 Warn whenever a label is declared but not used.
3425 This warning is enabled by @option{-Wall}.
3427 To suppress this warning use the @samp{unused} attribute
3428 (@pxref{Variable Attributes}).
3430 @item -Wunused-parameter
3431 @opindex Wunused-parameter
3432 @opindex Wno-unused-parameter
3433 Warn whenever a function parameter is unused aside from its declaration.
3435 To suppress this warning use the @samp{unused} attribute
3436 (@pxref{Variable Attributes}).
3438 @item -Wno-unused-result
3439 @opindex Wunused-result
3440 @opindex Wno-unused-result
3441 Do not warn if a caller of a function marked with attribute
3442 @code{warn_unused_result} (@pxref{Variable Attributes}) does not use
3443 its return value. The default is @option{-Wunused-result}.
3445 @item -Wunused-variable
3446 @opindex Wunused-variable
3447 @opindex Wno-unused-variable
3448 Warn whenever a local variable or non-constant static variable is unused
3449 aside from its declaration.
3450 This warning is enabled by @option{-Wall}.
3452 To suppress this warning use the @samp{unused} attribute
3453 (@pxref{Variable Attributes}).
3455 @item -Wunused-value
3456 @opindex Wunused-value
3457 @opindex Wno-unused-value
3458 Warn whenever a statement computes a result that is explicitly not
3459 used. To suppress this warning cast the unused expression to
3460 @samp{void}. This includes an expression-statement or the left-hand
3461 side of a comma expression that contains no side effects. For example,
3462 an expression such as @samp{x[i,j]} will cause a warning, while
3463 @samp{x[(void)i,j]} will not.
3465 This warning is enabled by @option{-Wall}.
3470 All the above @option{-Wunused} options combined.
3472 In order to get a warning about an unused function parameter, you must
3473 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3474 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3476 @item -Wuninitialized
3477 @opindex Wuninitialized
3478 @opindex Wno-uninitialized
3479 Warn if an automatic variable is used without first being initialized
3480 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3481 warn if a non-static reference or non-static @samp{const} member
3482 appears in a class without constructors.
3484 If you want to warn about code which uses the uninitialized value of the
3485 variable in its own initializer, use the @option{-Winit-self} option.
3487 These warnings occur for individual uninitialized or clobbered
3488 elements of structure, union or array variables as well as for
3489 variables which are uninitialized or clobbered as a whole. They do
3490 not occur for variables or elements declared @code{volatile}. Because
3491 these warnings depend on optimization, the exact variables or elements
3492 for which there are warnings will depend on the precise optimization
3493 options and version of GCC used.
3495 Note that there may be no warning about a variable that is used only
3496 to compute a value that itself is never used, because such
3497 computations may be deleted by data flow analysis before the warnings
3500 These warnings are made optional because GCC is not smart
3501 enough to see all the reasons why the code might be correct
3502 despite appearing to have an error. Here is one example of how
3523 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3524 always initialized, but GCC doesn't know this. Here is
3525 another common case:
3530 if (change_y) save_y = y, y = new_y;
3532 if (change_y) y = save_y;
3537 This has no bug because @code{save_y} is used only if it is set.
3539 @cindex @code{longjmp} warnings
3540 This option also warns when a non-volatile automatic variable might be
3541 changed by a call to @code{longjmp}. These warnings as well are possible
3542 only in optimizing compilation.
3544 The compiler sees only the calls to @code{setjmp}. It cannot know
3545 where @code{longjmp} will be called; in fact, a signal handler could
3546 call it at any point in the code. As a result, you may get a warning
3547 even when there is in fact no problem because @code{longjmp} cannot
3548 in fact be called at the place which would cause a problem.
3550 Some spurious warnings can be avoided if you declare all the functions
3551 you use that never return as @code{noreturn}. @xref{Function
3554 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3556 @item -Wunknown-pragmas
3557 @opindex Wunknown-pragmas
3558 @opindex Wno-unknown-pragmas
3559 @cindex warning for unknown pragmas
3560 @cindex unknown pragmas, warning
3561 @cindex pragmas, warning of unknown
3562 Warn when a #pragma directive is encountered which is not understood by
3563 GCC@. If this command line option is used, warnings will even be issued
3564 for unknown pragmas in system header files. This is not the case if
3565 the warnings were only enabled by the @option{-Wall} command line option.
3568 @opindex Wno-pragmas
3570 Do not warn about misuses of pragmas, such as incorrect parameters,
3571 invalid syntax, or conflicts between pragmas. See also
3572 @samp{-Wunknown-pragmas}.
3574 @item -Wstrict-aliasing
3575 @opindex Wstrict-aliasing
3576 @opindex Wno-strict-aliasing
3577 This option is only active when @option{-fstrict-aliasing} is active.
3578 It warns about code which might break the strict aliasing rules that the
3579 compiler is using for optimization. The warning does not catch all
3580 cases, but does attempt to catch the more common pitfalls. It is
3581 included in @option{-Wall}.
3582 It is equivalent to @option{-Wstrict-aliasing=3}
3584 @item -Wstrict-aliasing=n
3585 @opindex Wstrict-aliasing=n
3586 @opindex Wno-strict-aliasing=n
3587 This option is only active when @option{-fstrict-aliasing} is active.
3588 It warns about code which might break the strict aliasing rules that the
3589 compiler is using for optimization.
3590 Higher levels correspond to higher accuracy (fewer false positives).
3591 Higher levels also correspond to more effort, similar to the way -O works.
3592 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3595 Level 1: Most aggressive, quick, least accurate.
3596 Possibly useful when higher levels
3597 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3598 false negatives. However, it has many false positives.
3599 Warns for all pointer conversions between possibly incompatible types,
3600 even if never dereferenced. Runs in the frontend only.
3602 Level 2: Aggressive, quick, not too precise.
3603 May still have many false positives (not as many as level 1 though),
3604 and few false negatives (but possibly more than level 1).
3605 Unlike level 1, it only warns when an address is taken. Warns about
3606 incomplete types. Runs in the frontend only.
3608 Level 3 (default for @option{-Wstrict-aliasing}):
3609 Should have very few false positives and few false
3610 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3611 Takes care of the common pun+dereference pattern in the frontend:
3612 @code{*(int*)&some_float}.
3613 If optimization is enabled, it also runs in the backend, where it deals
3614 with multiple statement cases using flow-sensitive points-to information.
3615 Only warns when the converted pointer is dereferenced.
3616 Does not warn about incomplete types.
3618 @item -Wstrict-overflow
3619 @itemx -Wstrict-overflow=@var{n}
3620 @opindex Wstrict-overflow
3621 @opindex Wno-strict-overflow
3622 This option is only active when @option{-fstrict-overflow} is active.
3623 It warns about cases where the compiler optimizes based on the
3624 assumption that signed overflow does not occur. Note that it does not
3625 warn about all cases where the code might overflow: it only warns
3626 about cases where the compiler implements some optimization. Thus
3627 this warning depends on the optimization level.
3629 An optimization which assumes that signed overflow does not occur is
3630 perfectly safe if the values of the variables involved are such that
3631 overflow never does, in fact, occur. Therefore this warning can
3632 easily give a false positive: a warning about code which is not
3633 actually a problem. To help focus on important issues, several
3634 warning levels are defined. No warnings are issued for the use of
3635 undefined signed overflow when estimating how many iterations a loop
3636 will require, in particular when determining whether a loop will be
3640 @item -Wstrict-overflow=1
3641 Warn about cases which are both questionable and easy to avoid. For
3642 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3643 compiler will simplify this to @code{1}. This level of
3644 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3645 are not, and must be explicitly requested.
3647 @item -Wstrict-overflow=2
3648 Also warn about other cases where a comparison is simplified to a
3649 constant. For example: @code{abs (x) >= 0}. This can only be
3650 simplified when @option{-fstrict-overflow} is in effect, because
3651 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3652 zero. @option{-Wstrict-overflow} (with no level) is the same as
3653 @option{-Wstrict-overflow=2}.
3655 @item -Wstrict-overflow=3
3656 Also warn about other cases where a comparison is simplified. For
3657 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3659 @item -Wstrict-overflow=4
3660 Also warn about other simplifications not covered by the above cases.
3661 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3663 @item -Wstrict-overflow=5
3664 Also warn about cases where the compiler reduces the magnitude of a
3665 constant involved in a comparison. For example: @code{x + 2 > y} will
3666 be simplified to @code{x + 1 >= y}. This is reported only at the
3667 highest warning level because this simplification applies to many
3668 comparisons, so this warning level will give a very large number of
3672 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]}
3673 @opindex Wsuggest-attribute=
3674 @opindex Wno-suggest-attribute=
3675 Warn for cases where adding an attribute may be beneficial. The
3676 attributes currently supported are listed below.
3679 @item -Wsuggest-attribute=pure
3680 @itemx -Wsuggest-attribute=const
3681 @itemx -Wsuggest-attribute=noreturn
3682 @opindex Wsuggest-attribute=pure
3683 @opindex Wno-suggest-attribute=pure
3684 @opindex Wsuggest-attribute=const
3685 @opindex Wno-suggest-attribute=const
3686 @opindex Wsuggest-attribute=noreturn
3687 @opindex Wno-suggest-attribute=noreturn
3689 Warn about functions which might be candidates for attributes
3690 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
3691 functions visible in other compilation units or (in the case of @code{pure} and
3692 @code{const}) if it cannot prove that the function returns normally. A function
3693 returns normally if it doesn't contain an infinite loop nor returns abnormally
3694 by throwing, calling @code{abort()} or trapping. This analysis requires option
3695 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
3696 higher. Higher optimization levels improve the accuracy of the analysis.
3699 @item -Warray-bounds
3700 @opindex Wno-array-bounds
3701 @opindex Warray-bounds
3702 This option is only active when @option{-ftree-vrp} is active
3703 (default for @option{-O2} and above). It warns about subscripts to arrays
3704 that are always out of bounds. This warning is enabled by @option{-Wall}.
3706 @item -Wno-div-by-zero
3707 @opindex Wno-div-by-zero
3708 @opindex Wdiv-by-zero
3709 Do not warn about compile-time integer division by zero. Floating point
3710 division by zero is not warned about, as it can be a legitimate way of
3711 obtaining infinities and NaNs.
3713 @item -Wsystem-headers
3714 @opindex Wsystem-headers
3715 @opindex Wno-system-headers
3716 @cindex warnings from system headers
3717 @cindex system headers, warnings from
3718 Print warning messages for constructs found in system header files.
3719 Warnings from system headers are normally suppressed, on the assumption
3720 that they usually do not indicate real problems and would only make the
3721 compiler output harder to read. Using this command line option tells
3722 GCC to emit warnings from system headers as if they occurred in user
3723 code. However, note that using @option{-Wall} in conjunction with this
3724 option will @emph{not} warn about unknown pragmas in system
3725 headers---for that, @option{-Wunknown-pragmas} must also be used.
3728 @opindex Wtrampolines
3729 @opindex Wno-trampolines
3730 Warn about trampolines generated for pointers to nested functions.
3732 A trampoline is a small piece of data or code that is created at run
3733 time on the stack when the address of a nested function is taken, and
3734 is used to call the nested function indirectly. For some targets, it
3735 is made up of data only and thus requires no special treatment. But,
3736 for most targets, it is made up of code and thus requires the stack
3737 to be made executable in order for the program to work properly.
3740 @opindex Wfloat-equal
3741 @opindex Wno-float-equal
3742 Warn if floating point values are used in equality comparisons.
3744 The idea behind this is that sometimes it is convenient (for the
3745 programmer) to consider floating-point values as approximations to
3746 infinitely precise real numbers. If you are doing this, then you need
3747 to compute (by analyzing the code, or in some other way) the maximum or
3748 likely maximum error that the computation introduces, and allow for it
3749 when performing comparisons (and when producing output, but that's a
3750 different problem). In particular, instead of testing for equality, you
3751 would check to see whether the two values have ranges that overlap; and
3752 this is done with the relational operators, so equality comparisons are
3755 @item -Wtraditional @r{(C and Objective-C only)}
3756 @opindex Wtraditional
3757 @opindex Wno-traditional
3758 Warn about certain constructs that behave differently in traditional and
3759 ISO C@. Also warn about ISO C constructs that have no traditional C
3760 equivalent, and/or problematic constructs which should be avoided.
3764 Macro parameters that appear within string literals in the macro body.
3765 In traditional C macro replacement takes place within string literals,
3766 but does not in ISO C@.
3769 In traditional C, some preprocessor directives did not exist.
3770 Traditional preprocessors would only consider a line to be a directive
3771 if the @samp{#} appeared in column 1 on the line. Therefore
3772 @option{-Wtraditional} warns about directives that traditional C
3773 understands but would ignore because the @samp{#} does not appear as the
3774 first character on the line. It also suggests you hide directives like
3775 @samp{#pragma} not understood by traditional C by indenting them. Some
3776 traditional implementations would not recognize @samp{#elif}, so it
3777 suggests avoiding it altogether.
3780 A function-like macro that appears without arguments.
3783 The unary plus operator.
3786 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3787 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3788 constants.) Note, these suffixes appear in macros defined in the system
3789 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3790 Use of these macros in user code might normally lead to spurious
3791 warnings, however GCC's integrated preprocessor has enough context to
3792 avoid warning in these cases.
3795 A function declared external in one block and then used after the end of
3799 A @code{switch} statement has an operand of type @code{long}.
3802 A non-@code{static} function declaration follows a @code{static} one.
3803 This construct is not accepted by some traditional C compilers.
3806 The ISO type of an integer constant has a different width or
3807 signedness from its traditional type. This warning is only issued if
3808 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3809 typically represent bit patterns, are not warned about.
3812 Usage of ISO string concatenation is detected.
3815 Initialization of automatic aggregates.
3818 Identifier conflicts with labels. Traditional C lacks a separate
3819 namespace for labels.
3822 Initialization of unions. If the initializer is zero, the warning is
3823 omitted. This is done under the assumption that the zero initializer in
3824 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3825 initializer warnings and relies on default initialization to zero in the
3829 Conversions by prototypes between fixed/floating point values and vice
3830 versa. The absence of these prototypes when compiling with traditional
3831 C would cause serious problems. This is a subset of the possible
3832 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3835 Use of ISO C style function definitions. This warning intentionally is
3836 @emph{not} issued for prototype declarations or variadic functions
3837 because these ISO C features will appear in your code when using
3838 libiberty's traditional C compatibility macros, @code{PARAMS} and
3839 @code{VPARAMS}. This warning is also bypassed for nested functions
3840 because that feature is already a GCC extension and thus not relevant to
3841 traditional C compatibility.
3844 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3845 @opindex Wtraditional-conversion
3846 @opindex Wno-traditional-conversion
3847 Warn if a prototype causes a type conversion that is different from what
3848 would happen to the same argument in the absence of a prototype. This
3849 includes conversions of fixed point to floating and vice versa, and
3850 conversions changing the width or signedness of a fixed point argument
3851 except when the same as the default promotion.
3853 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3854 @opindex Wdeclaration-after-statement
3855 @opindex Wno-declaration-after-statement
3856 Warn when a declaration is found after a statement in a block. This
3857 construct, known from C++, was introduced with ISO C99 and is by default
3858 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3859 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3864 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3866 @item -Wno-endif-labels
3867 @opindex Wno-endif-labels
3868 @opindex Wendif-labels
3869 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3874 Warn whenever a local variable or type declaration shadows another variable,
3875 parameter, type, or class member (in C++), or whenever a built-in function
3876 is shadowed. Note that in C++, the compiler will not warn if a local variable
3877 shadows a struct/class/enum, but will warn if it shadows an explicit typedef.
3879 @item -Wlarger-than=@var{len}
3880 @opindex Wlarger-than=@var{len}
3881 @opindex Wlarger-than-@var{len}
3882 Warn whenever an object of larger than @var{len} bytes is defined.
3884 @item -Wframe-larger-than=@var{len}
3885 @opindex Wframe-larger-than
3886 Warn if the size of a function frame is larger than @var{len} bytes.
3887 The computation done to determine the stack frame size is approximate
3888 and not conservative.
3889 The actual requirements may be somewhat greater than @var{len}
3890 even if you do not get a warning. In addition, any space allocated
3891 via @code{alloca}, variable-length arrays, or related constructs
3892 is not included by the compiler when determining
3893 whether or not to issue a warning.
3895 @item -Wunsafe-loop-optimizations
3896 @opindex Wunsafe-loop-optimizations
3897 @opindex Wno-unsafe-loop-optimizations
3898 Warn if the loop cannot be optimized because the compiler could not
3899 assume anything on the bounds of the loop indices. With
3900 @option{-funsafe-loop-optimizations} warn if the compiler made
3903 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
3904 @opindex Wno-pedantic-ms-format
3905 @opindex Wpedantic-ms-format
3906 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
3907 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
3908 depending on the MS runtime, when you are using the options @option{-Wformat}
3909 and @option{-pedantic} without gnu-extensions.
3911 @item -Wpointer-arith
3912 @opindex Wpointer-arith
3913 @opindex Wno-pointer-arith
3914 Warn about anything that depends on the ``size of'' a function type or
3915 of @code{void}. GNU C assigns these types a size of 1, for
3916 convenience in calculations with @code{void *} pointers and pointers
3917 to functions. In C++, warn also when an arithmetic operation involves
3918 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3921 @opindex Wtype-limits
3922 @opindex Wno-type-limits
3923 Warn if a comparison is always true or always false due to the limited
3924 range of the data type, but do not warn for constant expressions. For
3925 example, warn if an unsigned variable is compared against zero with
3926 @samp{<} or @samp{>=}. This warning is also enabled by
3929 @item -Wbad-function-cast @r{(C and Objective-C only)}
3930 @opindex Wbad-function-cast
3931 @opindex Wno-bad-function-cast
3932 Warn whenever a function call is cast to a non-matching type.
3933 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3935 @item -Wc++-compat @r{(C and Objective-C only)}
3936 Warn about ISO C constructs that are outside of the common subset of
3937 ISO C and ISO C++, e.g.@: request for implicit conversion from
3938 @code{void *} to a pointer to non-@code{void} type.
3940 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3941 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3942 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3943 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3947 @opindex Wno-cast-qual
3948 Warn whenever a pointer is cast so as to remove a type qualifier from
3949 the target type. For example, warn if a @code{const char *} is cast
3950 to an ordinary @code{char *}.
3952 Also warn when making a cast which introduces a type qualifier in an
3953 unsafe way. For example, casting @code{char **} to @code{const char **}
3954 is unsafe, as in this example:
3957 /* p is char ** value. */
3958 const char **q = (const char **) p;
3959 /* Assignment of readonly string to const char * is OK. */
3961 /* Now char** pointer points to read-only memory. */
3966 @opindex Wcast-align
3967 @opindex Wno-cast-align
3968 Warn whenever a pointer is cast such that the required alignment of the
3969 target is increased. For example, warn if a @code{char *} is cast to
3970 an @code{int *} on machines where integers can only be accessed at
3971 two- or four-byte boundaries.
3973 @item -Wwrite-strings
3974 @opindex Wwrite-strings
3975 @opindex Wno-write-strings
3976 When compiling C, give string constants the type @code{const
3977 char[@var{length}]} so that copying the address of one into a
3978 non-@code{const} @code{char *} pointer will get a warning. These
3979 warnings will help you find at compile time code that can try to write
3980 into a string constant, but only if you have been very careful about
3981 using @code{const} in declarations and prototypes. Otherwise, it will
3982 just be a nuisance. This is why we did not make @option{-Wall} request
3985 When compiling C++, warn about the deprecated conversion from string
3986 literals to @code{char *}. This warning is enabled by default for C++
3991 @opindex Wno-clobbered
3992 Warn for variables that might be changed by @samp{longjmp} or
3993 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
3996 @opindex Wconversion
3997 @opindex Wno-conversion
3998 Warn for implicit conversions that may alter a value. This includes
3999 conversions between real and integer, like @code{abs (x)} when
4000 @code{x} is @code{double}; conversions between signed and unsigned,
4001 like @code{unsigned ui = -1}; and conversions to smaller types, like
4002 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
4003 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
4004 changed by the conversion like in @code{abs (2.0)}. Warnings about
4005 conversions between signed and unsigned integers can be disabled by
4006 using @option{-Wno-sign-conversion}.
4008 For C++, also warn for confusing overload resolution for user-defined
4009 conversions; and conversions that will never use a type conversion
4010 operator: conversions to @code{void}, the same type, a base class or a
4011 reference to them. Warnings about conversions between signed and
4012 unsigned integers are disabled by default in C++ unless
4013 @option{-Wsign-conversion} is explicitly enabled.
4015 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
4016 @opindex Wconversion-null
4017 @opindex Wno-conversion-null
4018 Do not warn for conversions between @code{NULL} and non-pointer
4019 types. @option{-Wconversion-null} is enabled by default.
4022 @opindex Wempty-body
4023 @opindex Wno-empty-body
4024 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
4025 while} statement. This warning is also enabled by @option{-Wextra}.
4027 @item -Wenum-compare
4028 @opindex Wenum-compare
4029 @opindex Wno-enum-compare
4030 Warn about a comparison between values of different enum types. In C++
4031 this warning is enabled by default. In C this warning is enabled by
4034 @item -Wjump-misses-init @r{(C, Objective-C only)}
4035 @opindex Wjump-misses-init
4036 @opindex Wno-jump-misses-init
4037 Warn if a @code{goto} statement or a @code{switch} statement jumps
4038 forward across the initialization of a variable, or jumps backward to a
4039 label after the variable has been initialized. This only warns about
4040 variables which are initialized when they are declared. This warning is
4041 only supported for C and Objective C; in C++ this sort of branch is an
4044 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
4045 can be disabled with the @option{-Wno-jump-misses-init} option.
4047 @item -Wsign-compare
4048 @opindex Wsign-compare
4049 @opindex Wno-sign-compare
4050 @cindex warning for comparison of signed and unsigned values
4051 @cindex comparison of signed and unsigned values, warning
4052 @cindex signed and unsigned values, comparison warning
4053 Warn when a comparison between signed and unsigned values could produce
4054 an incorrect result when the signed value is converted to unsigned.
4055 This warning is also enabled by @option{-Wextra}; to get the other warnings
4056 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
4058 @item -Wsign-conversion
4059 @opindex Wsign-conversion
4060 @opindex Wno-sign-conversion
4061 Warn for implicit conversions that may change the sign of an integer
4062 value, like assigning a signed integer expression to an unsigned
4063 integer variable. An explicit cast silences the warning. In C, this
4064 option is enabled also by @option{-Wconversion}.
4068 @opindex Wno-address
4069 Warn about suspicious uses of memory addresses. These include using
4070 the address of a function in a conditional expression, such as
4071 @code{void func(void); if (func)}, and comparisons against the memory
4072 address of a string literal, such as @code{if (x == "abc")}. Such
4073 uses typically indicate a programmer error: the address of a function
4074 always evaluates to true, so their use in a conditional usually
4075 indicate that the programmer forgot the parentheses in a function
4076 call; and comparisons against string literals result in unspecified
4077 behavior and are not portable in C, so they usually indicate that the
4078 programmer intended to use @code{strcmp}. This warning is enabled by
4082 @opindex Wlogical-op
4083 @opindex Wno-logical-op
4084 Warn about suspicious uses of logical operators in expressions.
4085 This includes using logical operators in contexts where a
4086 bit-wise operator is likely to be expected.
4088 @item -Waggregate-return
4089 @opindex Waggregate-return
4090 @opindex Wno-aggregate-return
4091 Warn if any functions that return structures or unions are defined or
4092 called. (In languages where you can return an array, this also elicits
4095 @item -Wno-attributes
4096 @opindex Wno-attributes
4097 @opindex Wattributes
4098 Do not warn if an unexpected @code{__attribute__} is used, such as
4099 unrecognized attributes, function attributes applied to variables,
4100 etc. This will not stop errors for incorrect use of supported
4103 @item -Wno-builtin-macro-redefined
4104 @opindex Wno-builtin-macro-redefined
4105 @opindex Wbuiltin-macro-redefined
4106 Do not warn if certain built-in macros are redefined. This suppresses
4107 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
4108 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
4110 @item -Wstrict-prototypes @r{(C and Objective-C only)}
4111 @opindex Wstrict-prototypes
4112 @opindex Wno-strict-prototypes
4113 Warn if a function is declared or defined without specifying the
4114 argument types. (An old-style function definition is permitted without
4115 a warning if preceded by a declaration which specifies the argument
4118 @item -Wold-style-declaration @r{(C and Objective-C only)}
4119 @opindex Wold-style-declaration
4120 @opindex Wno-old-style-declaration
4121 Warn for obsolescent usages, according to the C Standard, in a
4122 declaration. For example, warn if storage-class specifiers like
4123 @code{static} are not the first things in a declaration. This warning
4124 is also enabled by @option{-Wextra}.
4126 @item -Wold-style-definition @r{(C and Objective-C only)}
4127 @opindex Wold-style-definition
4128 @opindex Wno-old-style-definition
4129 Warn if an old-style function definition is used. A warning is given
4130 even if there is a previous prototype.
4132 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
4133 @opindex Wmissing-parameter-type
4134 @opindex Wno-missing-parameter-type
4135 A function parameter is declared without a type specifier in K&R-style
4142 This warning is also enabled by @option{-Wextra}.
4144 @item -Wmissing-prototypes @r{(C and Objective-C only)}
4145 @opindex Wmissing-prototypes
4146 @opindex Wno-missing-prototypes
4147 Warn if a global function is defined without a previous prototype
4148 declaration. This warning is issued even if the definition itself
4149 provides a prototype. The aim is to detect global functions that fail
4150 to be declared in header files.
4152 @item -Wmissing-declarations
4153 @opindex Wmissing-declarations
4154 @opindex Wno-missing-declarations
4155 Warn if a global function is defined without a previous declaration.
4156 Do so even if the definition itself provides a prototype.
4157 Use this option to detect global functions that are not declared in
4158 header files. In C++, no warnings are issued for function templates,
4159 or for inline functions, or for functions in anonymous namespaces.
4161 @item -Wmissing-field-initializers
4162 @opindex Wmissing-field-initializers
4163 @opindex Wno-missing-field-initializers
4167 Warn if a structure's initializer has some fields missing. For
4168 example, the following code would cause such a warning, because
4169 @code{x.h} is implicitly zero:
4172 struct s @{ int f, g, h; @};
4173 struct s x = @{ 3, 4 @};
4176 This option does not warn about designated initializers, so the following
4177 modification would not trigger a warning:
4180 struct s @{ int f, g, h; @};
4181 struct s x = @{ .f = 3, .g = 4 @};
4184 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
4185 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
4187 @item -Wmissing-format-attribute
4188 @opindex Wmissing-format-attribute
4189 @opindex Wno-missing-format-attribute
4192 Warn about function pointers which might be candidates for @code{format}
4193 attributes. Note these are only possible candidates, not absolute ones.
4194 GCC will guess that function pointers with @code{format} attributes that
4195 are used in assignment, initialization, parameter passing or return
4196 statements should have a corresponding @code{format} attribute in the
4197 resulting type. I.e.@: the left-hand side of the assignment or
4198 initialization, the type of the parameter variable, or the return type
4199 of the containing function respectively should also have a @code{format}
4200 attribute to avoid the warning.
4202 GCC will also warn about function definitions which might be
4203 candidates for @code{format} attributes. Again, these are only
4204 possible candidates. GCC will guess that @code{format} attributes
4205 might be appropriate for any function that calls a function like
4206 @code{vprintf} or @code{vscanf}, but this might not always be the
4207 case, and some functions for which @code{format} attributes are
4208 appropriate may not be detected.
4210 @item -Wno-multichar
4211 @opindex Wno-multichar
4213 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4214 Usually they indicate a typo in the user's code, as they have
4215 implementation-defined values, and should not be used in portable code.
4217 @item -Wnormalized=<none|id|nfc|nfkc>
4218 @opindex Wnormalized=
4221 @cindex character set, input normalization
4222 In ISO C and ISO C++, two identifiers are different if they are
4223 different sequences of characters. However, sometimes when characters
4224 outside the basic ASCII character set are used, you can have two
4225 different character sequences that look the same. To avoid confusion,
4226 the ISO 10646 standard sets out some @dfn{normalization rules} which
4227 when applied ensure that two sequences that look the same are turned into
4228 the same sequence. GCC can warn you if you are using identifiers which
4229 have not been normalized; this option controls that warning.
4231 There are four levels of warning that GCC supports. The default is
4232 @option{-Wnormalized=nfc}, which warns about any identifier which is
4233 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4234 recommended form for most uses.
4236 Unfortunately, there are some characters which ISO C and ISO C++ allow
4237 in identifiers that when turned into NFC aren't allowable as
4238 identifiers. That is, there's no way to use these symbols in portable
4239 ISO C or C++ and have all your identifiers in NFC@.
4240 @option{-Wnormalized=id} suppresses the warning for these characters.
4241 It is hoped that future versions of the standards involved will correct
4242 this, which is why this option is not the default.
4244 You can switch the warning off for all characters by writing
4245 @option{-Wnormalized=none}. You would only want to do this if you
4246 were using some other normalization scheme (like ``D''), because
4247 otherwise you can easily create bugs that are literally impossible to see.
4249 Some characters in ISO 10646 have distinct meanings but look identical
4250 in some fonts or display methodologies, especially once formatting has
4251 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4252 LETTER N'', will display just like a regular @code{n} which has been
4253 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4254 normalization scheme to convert all these into a standard form as
4255 well, and GCC will warn if your code is not in NFKC if you use
4256 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4257 about every identifier that contains the letter O because it might be
4258 confused with the digit 0, and so is not the default, but may be
4259 useful as a local coding convention if the programming environment is
4260 unable to be fixed to display these characters distinctly.
4262 @item -Wno-deprecated
4263 @opindex Wno-deprecated
4264 @opindex Wdeprecated
4265 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4267 @item -Wno-deprecated-declarations
4268 @opindex Wno-deprecated-declarations
4269 @opindex Wdeprecated-declarations
4270 Do not warn about uses of functions (@pxref{Function Attributes}),
4271 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4272 Attributes}) marked as deprecated by using the @code{deprecated}
4276 @opindex Wno-overflow
4278 Do not warn about compile-time overflow in constant expressions.
4280 @item -Woverride-init @r{(C and Objective-C only)}
4281 @opindex Woverride-init
4282 @opindex Wno-override-init
4286 Warn if an initialized field without side effects is overridden when
4287 using designated initializers (@pxref{Designated Inits, , Designated
4290 This warning is included in @option{-Wextra}. To get other
4291 @option{-Wextra} warnings without this one, use @samp{-Wextra
4292 -Wno-override-init}.
4297 Warn if a structure is given the packed attribute, but the packed
4298 attribute has no effect on the layout or size of the structure.
4299 Such structures may be mis-aligned for little benefit. For
4300 instance, in this code, the variable @code{f.x} in @code{struct bar}
4301 will be misaligned even though @code{struct bar} does not itself
4302 have the packed attribute:
4309 @} __attribute__((packed));
4317 @item -Wpacked-bitfield-compat
4318 @opindex Wpacked-bitfield-compat
4319 @opindex Wno-packed-bitfield-compat
4320 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4321 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4322 the change can lead to differences in the structure layout. GCC
4323 informs you when the offset of such a field has changed in GCC 4.4.
4324 For example there is no longer a 4-bit padding between field @code{a}
4325 and @code{b} in this structure:
4332 @} __attribute__ ((packed));
4335 This warning is enabled by default. Use
4336 @option{-Wno-packed-bitfield-compat} to disable this warning.
4341 Warn if padding is included in a structure, either to align an element
4342 of the structure or to align the whole structure. Sometimes when this
4343 happens it is possible to rearrange the fields of the structure to
4344 reduce the padding and so make the structure smaller.
4346 @item -Wredundant-decls
4347 @opindex Wredundant-decls
4348 @opindex Wno-redundant-decls
4349 Warn if anything is declared more than once in the same scope, even in
4350 cases where multiple declaration is valid and changes nothing.
4352 @item -Wnested-externs @r{(C and Objective-C only)}
4353 @opindex Wnested-externs
4354 @opindex Wno-nested-externs
4355 Warn if an @code{extern} declaration is encountered within a function.
4360 Warn if a function can not be inlined and it was declared as inline.
4361 Even with this option, the compiler will not warn about failures to
4362 inline functions declared in system headers.
4364 The compiler uses a variety of heuristics to determine whether or not
4365 to inline a function. For example, the compiler takes into account
4366 the size of the function being inlined and the amount of inlining
4367 that has already been done in the current function. Therefore,
4368 seemingly insignificant changes in the source program can cause the
4369 warnings produced by @option{-Winline} to appear or disappear.
4371 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4372 @opindex Wno-invalid-offsetof
4373 @opindex Winvalid-offsetof
4374 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4375 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4376 to a non-POD type is undefined. In existing C++ implementations,
4377 however, @samp{offsetof} typically gives meaningful results even when
4378 applied to certain kinds of non-POD types. (Such as a simple
4379 @samp{struct} that fails to be a POD type only by virtue of having a
4380 constructor.) This flag is for users who are aware that they are
4381 writing nonportable code and who have deliberately chosen to ignore the
4384 The restrictions on @samp{offsetof} may be relaxed in a future version
4385 of the C++ standard.
4387 @item -Wno-int-to-pointer-cast
4388 @opindex Wno-int-to-pointer-cast
4389 @opindex Wint-to-pointer-cast
4390 Suppress warnings from casts to pointer type of an integer of a
4391 different size. In C++, casting to a pointer type of smaller size is
4392 an error. @option{Wint-to-pointer-cast} is enabled by default.
4395 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4396 @opindex Wno-pointer-to-int-cast
4397 @opindex Wpointer-to-int-cast
4398 Suppress warnings from casts from a pointer to an integer type of a
4402 @opindex Winvalid-pch
4403 @opindex Wno-invalid-pch
4404 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4405 the search path but can't be used.
4409 @opindex Wno-long-long
4410 Warn if @samp{long long} type is used. This is enabled by either
4411 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4412 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4414 @item -Wvariadic-macros
4415 @opindex Wvariadic-macros
4416 @opindex Wno-variadic-macros
4417 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4418 alternate syntax when in pedantic ISO C99 mode. This is default.
4419 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4424 Warn if variable length array is used in the code.
4425 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4426 the variable length array.
4428 @item -Wvolatile-register-var
4429 @opindex Wvolatile-register-var
4430 @opindex Wno-volatile-register-var
4431 Warn if a register variable is declared volatile. The volatile
4432 modifier does not inhibit all optimizations that may eliminate reads
4433 and/or writes to register variables. This warning is enabled by
4436 @item -Wdisabled-optimization
4437 @opindex Wdisabled-optimization
4438 @opindex Wno-disabled-optimization
4439 Warn if a requested optimization pass is disabled. This warning does
4440 not generally indicate that there is anything wrong with your code; it
4441 merely indicates that GCC's optimizers were unable to handle the code
4442 effectively. Often, the problem is that your code is too big or too
4443 complex; GCC will refuse to optimize programs when the optimization
4444 itself is likely to take inordinate amounts of time.
4446 @item -Wpointer-sign @r{(C and Objective-C only)}
4447 @opindex Wpointer-sign
4448 @opindex Wno-pointer-sign
4449 Warn for pointer argument passing or assignment with different signedness.
4450 This option is only supported for C and Objective-C@. It is implied by
4451 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4452 @option{-Wno-pointer-sign}.
4454 @item -Wstack-protector
4455 @opindex Wstack-protector
4456 @opindex Wno-stack-protector
4457 This option is only active when @option{-fstack-protector} is active. It
4458 warns about functions that will not be protected against stack smashing.
4461 @opindex Wno-mudflap
4462 Suppress warnings about constructs that cannot be instrumented by
4465 @item -Woverlength-strings
4466 @opindex Woverlength-strings
4467 @opindex Wno-overlength-strings
4468 Warn about string constants which are longer than the ``minimum
4469 maximum'' length specified in the C standard. Modern compilers
4470 generally allow string constants which are much longer than the
4471 standard's minimum limit, but very portable programs should avoid
4472 using longer strings.
4474 The limit applies @emph{after} string constant concatenation, and does
4475 not count the trailing NUL@. In C90, the limit was 509 characters; in
4476 C99, it was raised to 4095. C++98 does not specify a normative
4477 minimum maximum, so we do not diagnose overlength strings in C++@.
4479 This option is implied by @option{-pedantic}, and can be disabled with
4480 @option{-Wno-overlength-strings}.
4482 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
4483 @opindex Wunsuffixed-float-constants
4485 GCC will issue a warning for any floating constant that does not have
4486 a suffix. When used together with @option{-Wsystem-headers} it will
4487 warn about such constants in system header files. This can be useful
4488 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4489 from the decimal floating-point extension to C99.
4492 @node Debugging Options
4493 @section Options for Debugging Your Program or GCC
4494 @cindex options, debugging
4495 @cindex debugging information options
4497 GCC has various special options that are used for debugging
4498 either your program or GCC:
4503 Produce debugging information in the operating system's native format
4504 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4507 On most systems that use stabs format, @option{-g} enables use of extra
4508 debugging information that only GDB can use; this extra information
4509 makes debugging work better in GDB but will probably make other debuggers
4511 refuse to read the program. If you want to control for certain whether
4512 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4513 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4515 GCC allows you to use @option{-g} with
4516 @option{-O}. The shortcuts taken by optimized code may occasionally
4517 produce surprising results: some variables you declared may not exist
4518 at all; flow of control may briefly move where you did not expect it;
4519 some statements may not be executed because they compute constant
4520 results or their values were already at hand; some statements may
4521 execute in different places because they were moved out of loops.
4523 Nevertheless it proves possible to debug optimized output. This makes
4524 it reasonable to use the optimizer for programs that might have bugs.
4526 The following options are useful when GCC is generated with the
4527 capability for more than one debugging format.
4531 Produce debugging information for use by GDB@. This means to use the
4532 most expressive format available (DWARF 2, stabs, or the native format
4533 if neither of those are supported), including GDB extensions if at all
4538 Produce debugging information in stabs format (if that is supported),
4539 without GDB extensions. This is the format used by DBX on most BSD
4540 systems. On MIPS, Alpha and System V Release 4 systems this option
4541 produces stabs debugging output which is not understood by DBX or SDB@.
4542 On System V Release 4 systems this option requires the GNU assembler.
4544 @item -feliminate-unused-debug-symbols
4545 @opindex feliminate-unused-debug-symbols
4546 Produce debugging information in stabs format (if that is supported),
4547 for only symbols that are actually used.
4549 @item -femit-class-debug-always
4550 Instead of emitting debugging information for a C++ class in only one
4551 object file, emit it in all object files using the class. This option
4552 should be used only with debuggers that are unable to handle the way GCC
4553 normally emits debugging information for classes because using this
4554 option will increase the size of debugging information by as much as a
4559 Produce debugging information in stabs format (if that is supported),
4560 using GNU extensions understood only by the GNU debugger (GDB)@. The
4561 use of these extensions is likely to make other debuggers crash or
4562 refuse to read the program.
4566 Produce debugging information in COFF format (if that is supported).
4567 This is the format used by SDB on most System V systems prior to
4572 Produce debugging information in XCOFF format (if that is supported).
4573 This is the format used by the DBX debugger on IBM RS/6000 systems.
4577 Produce debugging information in XCOFF format (if that is supported),
4578 using GNU extensions understood only by the GNU debugger (GDB)@. The
4579 use of these extensions is likely to make other debuggers crash or
4580 refuse to read the program, and may cause assemblers other than the GNU
4581 assembler (GAS) to fail with an error.
4583 @item -gdwarf-@var{version}
4584 @opindex gdwarf-@var{version}
4585 Produce debugging information in DWARF format (if that is
4586 supported). This is the format used by DBX on IRIX 6. The value
4587 of @var{version} may be either 2, 3 or 4; the default version is 2.
4589 Note that with DWARF version 2 some ports require, and will always
4590 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4592 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
4593 for maximum benefit.
4595 @item -gstrict-dwarf
4596 @opindex gstrict-dwarf
4597 Disallow using extensions of later DWARF standard version than selected
4598 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
4599 DWARF extensions from later standard versions is allowed.
4601 @item -gno-strict-dwarf
4602 @opindex gno-strict-dwarf
4603 Allow using extensions of later DWARF standard version than selected with
4604 @option{-gdwarf-@var{version}}.
4608 Produce debugging information in VMS debug format (if that is
4609 supported). This is the format used by DEBUG on VMS systems.
4612 @itemx -ggdb@var{level}
4613 @itemx -gstabs@var{level}
4614 @itemx -gcoff@var{level}
4615 @itemx -gxcoff@var{level}
4616 @itemx -gvms@var{level}
4617 Request debugging information and also use @var{level} to specify how
4618 much information. The default level is 2.
4620 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4623 Level 1 produces minimal information, enough for making backtraces in
4624 parts of the program that you don't plan to debug. This includes
4625 descriptions of functions and external variables, but no information
4626 about local variables and no line numbers.
4628 Level 3 includes extra information, such as all the macro definitions
4629 present in the program. Some debuggers support macro expansion when
4630 you use @option{-g3}.
4632 @option{-gdwarf-2} does not accept a concatenated debug level, because
4633 GCC used to support an option @option{-gdwarf} that meant to generate
4634 debug information in version 1 of the DWARF format (which is very
4635 different from version 2), and it would have been too confusing. That
4636 debug format is long obsolete, but the option cannot be changed now.
4637 Instead use an additional @option{-g@var{level}} option to change the
4638 debug level for DWARF.
4642 Turn off generation of debug info, if leaving out this option would have
4643 generated it, or turn it on at level 2 otherwise. The position of this
4644 argument in the command line does not matter, it takes effect after all
4645 other options are processed, and it does so only once, no matter how
4646 many times it is given. This is mainly intended to be used with
4647 @option{-fcompare-debug}.
4649 @item -fdump-final-insns@r{[}=@var{file}@r{]}
4650 @opindex fdump-final-insns
4651 Dump the final internal representation (RTL) to @var{file}. If the
4652 optional argument is omitted (or if @var{file} is @code{.}), the name
4653 of the dump file will be determined by appending @code{.gkd} to the
4654 compilation output file name.
4656 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4657 @opindex fcompare-debug
4658 @opindex fno-compare-debug
4659 If no error occurs during compilation, run the compiler a second time,
4660 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4661 passed to the second compilation. Dump the final internal
4662 representation in both compilations, and print an error if they differ.
4664 If the equal sign is omitted, the default @option{-gtoggle} is used.
4666 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4667 and nonzero, implicitly enables @option{-fcompare-debug}. If
4668 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4669 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4672 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4673 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4674 of the final representation and the second compilation, preventing even
4675 @env{GCC_COMPARE_DEBUG} from taking effect.
4677 To verify full coverage during @option{-fcompare-debug} testing, set
4678 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4679 which GCC will reject as an invalid option in any actual compilation
4680 (rather than preprocessing, assembly or linking). To get just a
4681 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4682 not overridden} will do.
4684 @item -fcompare-debug-second
4685 @opindex fcompare-debug-second
4686 This option is implicitly passed to the compiler for the second
4687 compilation requested by @option{-fcompare-debug}, along with options to
4688 silence warnings, and omitting other options that would cause
4689 side-effect compiler outputs to files or to the standard output. Dump
4690 files and preserved temporary files are renamed so as to contain the
4691 @code{.gk} additional extension during the second compilation, to avoid
4692 overwriting those generated by the first.
4694 When this option is passed to the compiler driver, it causes the
4695 @emph{first} compilation to be skipped, which makes it useful for little
4696 other than debugging the compiler proper.
4698 @item -feliminate-dwarf2-dups
4699 @opindex feliminate-dwarf2-dups
4700 Compress DWARF2 debugging information by eliminating duplicated
4701 information about each symbol. This option only makes sense when
4702 generating DWARF2 debugging information with @option{-gdwarf-2}.
4704 @item -femit-struct-debug-baseonly
4705 Emit debug information for struct-like types
4706 only when the base name of the compilation source file
4707 matches the base name of file in which the struct was defined.
4709 This option substantially reduces the size of debugging information,
4710 but at significant potential loss in type information to the debugger.
4711 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4712 See @option{-femit-struct-debug-detailed} for more detailed control.
4714 This option works only with DWARF 2.
4716 @item -femit-struct-debug-reduced
4717 Emit debug information for struct-like types
4718 only when the base name of the compilation source file
4719 matches the base name of file in which the type was defined,
4720 unless the struct is a template or defined in a system header.
4722 This option significantly reduces the size of debugging information,
4723 with some potential loss in type information to the debugger.
4724 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4725 See @option{-femit-struct-debug-detailed} for more detailed control.
4727 This option works only with DWARF 2.
4729 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4730 Specify the struct-like types
4731 for which the compiler will generate debug information.
4732 The intent is to reduce duplicate struct debug information
4733 between different object files within the same program.
4735 This option is a detailed version of
4736 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4737 which will serve for most needs.
4739 A specification has the syntax@*
4740 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4742 The optional first word limits the specification to
4743 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4744 A struct type is used directly when it is the type of a variable, member.
4745 Indirect uses arise through pointers to structs.
4746 That is, when use of an incomplete struct would be legal, the use is indirect.
4748 @samp{struct one direct; struct two * indirect;}.
4750 The optional second word limits the specification to
4751 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4752 Generic structs are a bit complicated to explain.
4753 For C++, these are non-explicit specializations of template classes,
4754 or non-template classes within the above.
4755 Other programming languages have generics,
4756 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4758 The third word specifies the source files for those
4759 structs for which the compiler will emit debug information.
4760 The values @samp{none} and @samp{any} have the normal meaning.
4761 The value @samp{base} means that
4762 the base of name of the file in which the type declaration appears
4763 must match the base of the name of the main compilation file.
4764 In practice, this means that
4765 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4766 but types declared in other header will not.
4767 The value @samp{sys} means those types satisfying @samp{base}
4768 or declared in system or compiler headers.
4770 You may need to experiment to determine the best settings for your application.
4772 The default is @samp{-femit-struct-debug-detailed=all}.
4774 This option works only with DWARF 2.
4776 @item -fenable-icf-debug
4777 @opindex fenable-icf-debug
4778 Generate additional debug information to support identical code folding (ICF).
4779 This option only works with DWARF version 2 or higher.
4781 @item -fno-merge-debug-strings
4782 @opindex fmerge-debug-strings
4783 @opindex fno-merge-debug-strings
4784 Direct the linker to not merge together strings in the debugging
4785 information which are identical in different object files. Merging is
4786 not supported by all assemblers or linkers. Merging decreases the size
4787 of the debug information in the output file at the cost of increasing
4788 link processing time. Merging is enabled by default.
4790 @item -fdebug-prefix-map=@var{old}=@var{new}
4791 @opindex fdebug-prefix-map
4792 When compiling files in directory @file{@var{old}}, record debugging
4793 information describing them as in @file{@var{new}} instead.
4795 @item -fno-dwarf2-cfi-asm
4796 @opindex fdwarf2-cfi-asm
4797 @opindex fno-dwarf2-cfi-asm
4798 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4799 instead of using GAS @code{.cfi_*} directives.
4801 @cindex @command{prof}
4804 Generate extra code to write profile information suitable for the
4805 analysis program @command{prof}. You must use this option when compiling
4806 the source files you want data about, and you must also use it when
4809 @cindex @command{gprof}
4812 Generate extra code to write profile information suitable for the
4813 analysis program @command{gprof}. You must use this option when compiling
4814 the source files you want data about, and you must also use it when
4819 Makes the compiler print out each function name as it is compiled, and
4820 print some statistics about each pass when it finishes.
4823 @opindex ftime-report
4824 Makes the compiler print some statistics about the time consumed by each
4825 pass when it finishes.
4828 @opindex fmem-report
4829 Makes the compiler print some statistics about permanent memory
4830 allocation when it finishes.
4832 @item -fpre-ipa-mem-report
4833 @opindex fpre-ipa-mem-report
4834 @item -fpost-ipa-mem-report
4835 @opindex fpost-ipa-mem-report
4836 Makes the compiler print some statistics about permanent memory
4837 allocation before or after interprocedural optimization.
4840 @opindex fstack-usage
4841 Makes the compiler output stack usage information for the program, on a
4842 per-function basis. The filename for the dump is made by appending
4843 @file{.su} to the AUXNAME. AUXNAME is generated from the name of
4844 the output file, if explicitly specified and it is not an executable,
4845 otherwise it is the basename of the source file. An entry is made up
4850 The name of the function.
4854 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
4857 The qualifier @code{static} means that the function manipulates the stack
4858 statically: a fixed number of bytes are allocated for the frame on function
4859 entry and released on function exit; no stack adjustments are otherwise made
4860 in the function. The second field is this fixed number of bytes.
4862 The qualifier @code{dynamic} means that the function manipulates the stack
4863 dynamically: in addition to the static allocation described above, stack
4864 adjustments are made in the body of the function, for example to push/pop
4865 arguments around function calls. If the qualifier @code{bounded} is also
4866 present, the amount of these adjustments is bounded at compile-time and
4867 the second field is an upper bound of the total amount of stack used by
4868 the function. If it is not present, the amount of these adjustments is
4869 not bounded at compile-time and the second field only represents the
4872 @item -fprofile-arcs
4873 @opindex fprofile-arcs
4874 Add code so that program flow @dfn{arcs} are instrumented. During
4875 execution the program records how many times each branch and call is
4876 executed and how many times it is taken or returns. When the compiled
4877 program exits it saves this data to a file called
4878 @file{@var{auxname}.gcda} for each source file. The data may be used for
4879 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4880 test coverage analysis (@option{-ftest-coverage}). Each object file's
4881 @var{auxname} is generated from the name of the output file, if
4882 explicitly specified and it is not the final executable, otherwise it is
4883 the basename of the source file. In both cases any suffix is removed
4884 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4885 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4886 @xref{Cross-profiling}.
4888 @cindex @command{gcov}
4892 This option is used to compile and link code instrumented for coverage
4893 analysis. The option is a synonym for @option{-fprofile-arcs}
4894 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4895 linking). See the documentation for those options for more details.
4900 Compile the source files with @option{-fprofile-arcs} plus optimization
4901 and code generation options. For test coverage analysis, use the
4902 additional @option{-ftest-coverage} option. You do not need to profile
4903 every source file in a program.
4906 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4907 (the latter implies the former).
4910 Run the program on a representative workload to generate the arc profile
4911 information. This may be repeated any number of times. You can run
4912 concurrent instances of your program, and provided that the file system
4913 supports locking, the data files will be correctly updated. Also
4914 @code{fork} calls are detected and correctly handled (double counting
4918 For profile-directed optimizations, compile the source files again with
4919 the same optimization and code generation options plus
4920 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4921 Control Optimization}).
4924 For test coverage analysis, use @command{gcov} to produce human readable
4925 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4926 @command{gcov} documentation for further information.
4930 With @option{-fprofile-arcs}, for each function of your program GCC
4931 creates a program flow graph, then finds a spanning tree for the graph.
4932 Only arcs that are not on the spanning tree have to be instrumented: the
4933 compiler adds code to count the number of times that these arcs are
4934 executed. When an arc is the only exit or only entrance to a block, the
4935 instrumentation code can be added to the block; otherwise, a new basic
4936 block must be created to hold the instrumentation code.
4939 @item -ftest-coverage
4940 @opindex ftest-coverage
4941 Produce a notes file that the @command{gcov} code-coverage utility
4942 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4943 show program coverage. Each source file's note file is called
4944 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4945 above for a description of @var{auxname} and instructions on how to
4946 generate test coverage data. Coverage data will match the source files
4947 more closely, if you do not optimize.
4949 @item -fdbg-cnt-list
4950 @opindex fdbg-cnt-list
4951 Print the name and the counter upperbound for all debug counters.
4953 @item -fdbg-cnt=@var{counter-value-list}
4955 Set the internal debug counter upperbound. @var{counter-value-list}
4956 is a comma-separated list of @var{name}:@var{value} pairs
4957 which sets the upperbound of each debug counter @var{name} to @var{value}.
4958 All debug counters have the initial upperbound of @var{UINT_MAX},
4959 thus dbg_cnt() returns true always unless the upperbound is set by this option.
4960 e.g. With -fdbg-cnt=dce:10,tail_call:0
4961 dbg_cnt(dce) will return true only for first 10 invocations
4962 and dbg_cnt(tail_call) will return false always.
4964 @item -d@var{letters}
4965 @itemx -fdump-rtl-@var{pass}
4967 Says to make debugging dumps during compilation at times specified by
4968 @var{letters}. This is used for debugging the RTL-based passes of the
4969 compiler. The file names for most of the dumps are made by appending
4970 a pass number and a word to the @var{dumpname}, and the files are
4971 created in the directory of the output file. Note that the pass
4972 number is computed statically as passes get registered into the pass
4973 manager. Thus the numbering is not related to the dynamic order of
4974 execution of passes. In particular, a pass installed by a plugin
4975 could have a number over 200 even if it executed quite early.
4976 @var{dumpname} is generated from the name of the output file, if
4977 explicitly specified and it is not an executable, otherwise it is the
4978 basename of the source file. These switches may have different effects
4979 when @option{-E} is used for preprocessing.
4981 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
4982 @option{-d} option @var{letters}. Here are the possible
4983 letters for use in @var{pass} and @var{letters}, and their meanings:
4987 @item -fdump-rtl-alignments
4988 @opindex fdump-rtl-alignments
4989 Dump after branch alignments have been computed.
4991 @item -fdump-rtl-asmcons
4992 @opindex fdump-rtl-asmcons
4993 Dump after fixing rtl statements that have unsatisfied in/out constraints.
4995 @item -fdump-rtl-auto_inc_dec
4996 @opindex fdump-rtl-auto_inc_dec
4997 Dump after auto-inc-dec discovery. This pass is only run on
4998 architectures that have auto inc or auto dec instructions.
5000 @item -fdump-rtl-barriers
5001 @opindex fdump-rtl-barriers
5002 Dump after cleaning up the barrier instructions.
5004 @item -fdump-rtl-bbpart
5005 @opindex fdump-rtl-bbpart
5006 Dump after partitioning hot and cold basic blocks.
5008 @item -fdump-rtl-bbro
5009 @opindex fdump-rtl-bbro
5010 Dump after block reordering.
5012 @item -fdump-rtl-btl1
5013 @itemx -fdump-rtl-btl2
5014 @opindex fdump-rtl-btl2
5015 @opindex fdump-rtl-btl2
5016 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
5017 after the two branch
5018 target load optimization passes.
5020 @item -fdump-rtl-bypass
5021 @opindex fdump-rtl-bypass
5022 Dump after jump bypassing and control flow optimizations.
5024 @item -fdump-rtl-combine
5025 @opindex fdump-rtl-combine
5026 Dump after the RTL instruction combination pass.
5028 @item -fdump-rtl-compgotos
5029 @opindex fdump-rtl-compgotos
5030 Dump after duplicating the computed gotos.
5032 @item -fdump-rtl-ce1
5033 @itemx -fdump-rtl-ce2
5034 @itemx -fdump-rtl-ce3
5035 @opindex fdump-rtl-ce1
5036 @opindex fdump-rtl-ce2
5037 @opindex fdump-rtl-ce3
5038 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
5039 @option{-fdump-rtl-ce3} enable dumping after the three
5040 if conversion passes.
5042 @itemx -fdump-rtl-cprop_hardreg
5043 @opindex fdump-rtl-cprop_hardreg
5044 Dump after hard register copy propagation.
5046 @itemx -fdump-rtl-csa
5047 @opindex fdump-rtl-csa
5048 Dump after combining stack adjustments.
5050 @item -fdump-rtl-cse1
5051 @itemx -fdump-rtl-cse2
5052 @opindex fdump-rtl-cse1
5053 @opindex fdump-rtl-cse2
5054 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
5055 the two common sub-expression elimination passes.
5057 @itemx -fdump-rtl-dce
5058 @opindex fdump-rtl-dce
5059 Dump after the standalone dead code elimination passes.
5061 @itemx -fdump-rtl-dbr
5062 @opindex fdump-rtl-dbr
5063 Dump after delayed branch scheduling.
5065 @item -fdump-rtl-dce1
5066 @itemx -fdump-rtl-dce2
5067 @opindex fdump-rtl-dce1
5068 @opindex fdump-rtl-dce2
5069 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
5070 the two dead store elimination passes.
5073 @opindex fdump-rtl-eh
5074 Dump after finalization of EH handling code.
5076 @item -fdump-rtl-eh_ranges
5077 @opindex fdump-rtl-eh_ranges
5078 Dump after conversion of EH handling range regions.
5080 @item -fdump-rtl-expand
5081 @opindex fdump-rtl-expand
5082 Dump after RTL generation.
5084 @item -fdump-rtl-fwprop1
5085 @itemx -fdump-rtl-fwprop2
5086 @opindex fdump-rtl-fwprop1
5087 @opindex fdump-rtl-fwprop2
5088 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
5089 dumping after the two forward propagation passes.
5091 @item -fdump-rtl-gcse1
5092 @itemx -fdump-rtl-gcse2
5093 @opindex fdump-rtl-gcse1
5094 @opindex fdump-rtl-gcse2
5095 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
5096 after global common subexpression elimination.
5098 @item -fdump-rtl-init-regs
5099 @opindex fdump-rtl-init-regs
5100 Dump after the initialization of the registers.
5102 @item -fdump-rtl-initvals
5103 @opindex fdump-rtl-initvals
5104 Dump after the computation of the initial value sets.
5106 @itemx -fdump-rtl-into_cfglayout
5107 @opindex fdump-rtl-into_cfglayout
5108 Dump after converting to cfglayout mode.
5110 @item -fdump-rtl-ira
5111 @opindex fdump-rtl-ira
5112 Dump after iterated register allocation.
5114 @item -fdump-rtl-jump
5115 @opindex fdump-rtl-jump
5116 Dump after the second jump optimization.
5118 @item -fdump-rtl-loop2
5119 @opindex fdump-rtl-loop2
5120 @option{-fdump-rtl-loop2} enables dumping after the rtl
5121 loop optimization passes.
5123 @item -fdump-rtl-mach
5124 @opindex fdump-rtl-mach
5125 Dump after performing the machine dependent reorganization pass, if that
5128 @item -fdump-rtl-mode_sw
5129 @opindex fdump-rtl-mode_sw
5130 Dump after removing redundant mode switches.
5132 @item -fdump-rtl-rnreg
5133 @opindex fdump-rtl-rnreg
5134 Dump after register renumbering.
5136 @itemx -fdump-rtl-outof_cfglayout
5137 @opindex fdump-rtl-outof_cfglayout
5138 Dump after converting from cfglayout mode.
5140 @item -fdump-rtl-peephole2
5141 @opindex fdump-rtl-peephole2
5142 Dump after the peephole pass.
5144 @item -fdump-rtl-postreload
5145 @opindex fdump-rtl-postreload
5146 Dump after post-reload optimizations.
5148 @itemx -fdump-rtl-pro_and_epilogue
5149 @opindex fdump-rtl-pro_and_epilogue
5150 Dump after generating the function pro and epilogues.
5152 @item -fdump-rtl-regmove
5153 @opindex fdump-rtl-regmove
5154 Dump after the register move pass.
5156 @item -fdump-rtl-sched1
5157 @itemx -fdump-rtl-sched2
5158 @opindex fdump-rtl-sched1
5159 @opindex fdump-rtl-sched2
5160 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
5161 after the basic block scheduling passes.
5163 @item -fdump-rtl-see
5164 @opindex fdump-rtl-see
5165 Dump after sign extension elimination.
5167 @item -fdump-rtl-seqabstr
5168 @opindex fdump-rtl-seqabstr
5169 Dump after common sequence discovery.
5171 @item -fdump-rtl-shorten
5172 @opindex fdump-rtl-shorten
5173 Dump after shortening branches.
5175 @item -fdump-rtl-sibling
5176 @opindex fdump-rtl-sibling
5177 Dump after sibling call optimizations.
5179 @item -fdump-rtl-split1
5180 @itemx -fdump-rtl-split2
5181 @itemx -fdump-rtl-split3
5182 @itemx -fdump-rtl-split4
5183 @itemx -fdump-rtl-split5
5184 @opindex fdump-rtl-split1
5185 @opindex fdump-rtl-split2
5186 @opindex fdump-rtl-split3
5187 @opindex fdump-rtl-split4
5188 @opindex fdump-rtl-split5
5189 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
5190 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
5191 @option{-fdump-rtl-split5} enable dumping after five rounds of
5192 instruction splitting.
5194 @item -fdump-rtl-sms
5195 @opindex fdump-rtl-sms
5196 Dump after modulo scheduling. This pass is only run on some
5199 @item -fdump-rtl-stack
5200 @opindex fdump-rtl-stack
5201 Dump after conversion from GCC's "flat register file" registers to the
5202 x87's stack-like registers. This pass is only run on x86 variants.
5204 @item -fdump-rtl-subreg1
5205 @itemx -fdump-rtl-subreg2
5206 @opindex fdump-rtl-subreg1
5207 @opindex fdump-rtl-subreg2
5208 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
5209 the two subreg expansion passes.
5211 @item -fdump-rtl-unshare
5212 @opindex fdump-rtl-unshare
5213 Dump after all rtl has been unshared.
5215 @item -fdump-rtl-vartrack
5216 @opindex fdump-rtl-vartrack
5217 Dump after variable tracking.
5219 @item -fdump-rtl-vregs
5220 @opindex fdump-rtl-vregs
5221 Dump after converting virtual registers to hard registers.
5223 @item -fdump-rtl-web
5224 @opindex fdump-rtl-web
5225 Dump after live range splitting.
5227 @item -fdump-rtl-regclass
5228 @itemx -fdump-rtl-subregs_of_mode_init
5229 @itemx -fdump-rtl-subregs_of_mode_finish
5230 @itemx -fdump-rtl-dfinit
5231 @itemx -fdump-rtl-dfinish
5232 @opindex fdump-rtl-regclass
5233 @opindex fdump-rtl-subregs_of_mode_init
5234 @opindex fdump-rtl-subregs_of_mode_finish
5235 @opindex fdump-rtl-dfinit
5236 @opindex fdump-rtl-dfinish
5237 These dumps are defined but always produce empty files.
5239 @item -fdump-rtl-all
5240 @opindex fdump-rtl-all
5241 Produce all the dumps listed above.
5245 Annotate the assembler output with miscellaneous debugging information.
5249 Dump all macro definitions, at the end of preprocessing, in addition to
5254 Produce a core dump whenever an error occurs.
5258 Print statistics on memory usage, at the end of the run, to
5263 Annotate the assembler output with a comment indicating which
5264 pattern and alternative was used. The length of each instruction is
5269 Dump the RTL in the assembler output as a comment before each instruction.
5270 Also turns on @option{-dp} annotation.
5274 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
5275 dump a representation of the control flow graph suitable for viewing with VCG
5276 to @file{@var{file}.@var{pass}.vcg}.
5280 Just generate RTL for a function instead of compiling it. Usually used
5281 with @option{-fdump-rtl-expand}.
5285 @opindex fdump-noaddr
5286 When doing debugging dumps, suppress address output. This makes it more
5287 feasible to use diff on debugging dumps for compiler invocations with
5288 different compiler binaries and/or different
5289 text / bss / data / heap / stack / dso start locations.
5291 @item -fdump-unnumbered
5292 @opindex fdump-unnumbered
5293 When doing debugging dumps, suppress instruction numbers and address output.
5294 This makes it more feasible to use diff on debugging dumps for compiler
5295 invocations with different options, in particular with and without
5298 @item -fdump-unnumbered-links
5299 @opindex fdump-unnumbered-links
5300 When doing debugging dumps (see @option{-d} option above), suppress
5301 instruction numbers for the links to the previous and next instructions
5304 @item -fdump-translation-unit @r{(C++ only)}
5305 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5306 @opindex fdump-translation-unit
5307 Dump a representation of the tree structure for the entire translation
5308 unit to a file. The file name is made by appending @file{.tu} to the
5309 source file name, and the file is created in the same directory as the
5310 output file. If the @samp{-@var{options}} form is used, @var{options}
5311 controls the details of the dump as described for the
5312 @option{-fdump-tree} options.
5314 @item -fdump-class-hierarchy @r{(C++ only)}
5315 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5316 @opindex fdump-class-hierarchy
5317 Dump a representation of each class's hierarchy and virtual function
5318 table layout to a file. The file name is made by appending
5319 @file{.class} to the source file name, and the file is created in the
5320 same directory as the output file. If the @samp{-@var{options}} form
5321 is used, @var{options} controls the details of the dump as described
5322 for the @option{-fdump-tree} options.
5324 @item -fdump-ipa-@var{switch}
5326 Control the dumping at various stages of inter-procedural analysis
5327 language tree to a file. The file name is generated by appending a
5328 switch specific suffix to the source file name, and the file is created
5329 in the same directory as the output file. The following dumps are
5334 Enables all inter-procedural analysis dumps.
5337 Dumps information about call-graph optimization, unused function removal,
5338 and inlining decisions.
5341 Dump after function inlining.
5345 @item -fdump-statistics-@var{option}
5346 @opindex fdump-statistics
5347 Enable and control dumping of pass statistics in a separate file. The
5348 file name is generated by appending a suffix ending in
5349 @samp{.statistics} to the source file name, and the file is created in
5350 the same directory as the output file. If the @samp{-@var{option}}
5351 form is used, @samp{-stats} will cause counters to be summed over the
5352 whole compilation unit while @samp{-details} will dump every event as
5353 the passes generate them. The default with no option is to sum
5354 counters for each function compiled.
5356 @item -fdump-tree-@var{switch}
5357 @itemx -fdump-tree-@var{switch}-@var{options}
5359 Control the dumping at various stages of processing the intermediate
5360 language tree to a file. The file name is generated by appending a
5361 switch specific suffix to the source file name, and the file is
5362 created in the same directory as the output file. If the
5363 @samp{-@var{options}} form is used, @var{options} is a list of
5364 @samp{-} separated options that control the details of the dump. Not
5365 all options are applicable to all dumps, those which are not
5366 meaningful will be ignored. The following options are available
5370 Print the address of each node. Usually this is not meaningful as it
5371 changes according to the environment and source file. Its primary use
5372 is for tying up a dump file with a debug environment.
5374 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
5375 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
5376 use working backward from mangled names in the assembly file.
5378 Inhibit dumping of members of a scope or body of a function merely
5379 because that scope has been reached. Only dump such items when they
5380 are directly reachable by some other path. When dumping pretty-printed
5381 trees, this option inhibits dumping the bodies of control structures.
5383 Print a raw representation of the tree. By default, trees are
5384 pretty-printed into a C-like representation.
5386 Enable more detailed dumps (not honored by every dump option).
5388 Enable dumping various statistics about the pass (not honored by every dump
5391 Enable showing basic block boundaries (disabled in raw dumps).
5393 Enable showing virtual operands for every statement.
5395 Enable showing line numbers for statements.
5397 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5399 Enable showing the tree dump for each statement.
5401 Enable showing the EH region number holding each statement.
5403 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5404 and @option{lineno}.
5407 The following tree dumps are possible:
5411 @opindex fdump-tree-original
5412 Dump before any tree based optimization, to @file{@var{file}.original}.
5415 @opindex fdump-tree-optimized
5416 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5419 @opindex fdump-tree-gimple
5420 Dump each function before and after the gimplification pass to a file. The
5421 file name is made by appending @file{.gimple} to the source file name.
5424 @opindex fdump-tree-cfg
5425 Dump the control flow graph of each function to a file. The file name is
5426 made by appending @file{.cfg} to the source file name.
5429 @opindex fdump-tree-vcg
5430 Dump the control flow graph of each function to a file in VCG format. The
5431 file name is made by appending @file{.vcg} to the source file name. Note
5432 that if the file contains more than one function, the generated file cannot
5433 be used directly by VCG@. You will need to cut and paste each function's
5434 graph into its own separate file first.
5437 @opindex fdump-tree-ch
5438 Dump each function after copying loop headers. The file name is made by
5439 appending @file{.ch} to the source file name.
5442 @opindex fdump-tree-ssa
5443 Dump SSA related information to a file. The file name is made by appending
5444 @file{.ssa} to the source file name.
5447 @opindex fdump-tree-alias
5448 Dump aliasing information for each function. The file name is made by
5449 appending @file{.alias} to the source file name.
5452 @opindex fdump-tree-ccp
5453 Dump each function after CCP@. The file name is made by appending
5454 @file{.ccp} to the source file name.
5457 @opindex fdump-tree-storeccp
5458 Dump each function after STORE-CCP@. The file name is made by appending
5459 @file{.storeccp} to the source file name.
5462 @opindex fdump-tree-pre
5463 Dump trees after partial redundancy elimination. The file name is made
5464 by appending @file{.pre} to the source file name.
5467 @opindex fdump-tree-fre
5468 Dump trees after full redundancy elimination. The file name is made
5469 by appending @file{.fre} to the source file name.
5472 @opindex fdump-tree-copyprop
5473 Dump trees after copy propagation. The file name is made
5474 by appending @file{.copyprop} to the source file name.
5476 @item store_copyprop
5477 @opindex fdump-tree-store_copyprop
5478 Dump trees after store copy-propagation. The file name is made
5479 by appending @file{.store_copyprop} to the source file name.
5482 @opindex fdump-tree-dce
5483 Dump each function after dead code elimination. The file name is made by
5484 appending @file{.dce} to the source file name.
5487 @opindex fdump-tree-mudflap
5488 Dump each function after adding mudflap instrumentation. The file name is
5489 made by appending @file{.mudflap} to the source file name.
5492 @opindex fdump-tree-sra
5493 Dump each function after performing scalar replacement of aggregates. The
5494 file name is made by appending @file{.sra} to the source file name.
5497 @opindex fdump-tree-sink
5498 Dump each function after performing code sinking. The file name is made
5499 by appending @file{.sink} to the source file name.
5502 @opindex fdump-tree-dom
5503 Dump each function after applying dominator tree optimizations. The file
5504 name is made by appending @file{.dom} to the source file name.
5507 @opindex fdump-tree-dse
5508 Dump each function after applying dead store elimination. The file
5509 name is made by appending @file{.dse} to the source file name.
5512 @opindex fdump-tree-phiopt
5513 Dump each function after optimizing PHI nodes into straightline code. The file
5514 name is made by appending @file{.phiopt} to the source file name.
5517 @opindex fdump-tree-forwprop
5518 Dump each function after forward propagating single use variables. The file
5519 name is made by appending @file{.forwprop} to the source file name.
5522 @opindex fdump-tree-copyrename
5523 Dump each function after applying the copy rename optimization. The file
5524 name is made by appending @file{.copyrename} to the source file name.
5527 @opindex fdump-tree-nrv
5528 Dump each function after applying the named return value optimization on
5529 generic trees. The file name is made by appending @file{.nrv} to the source
5533 @opindex fdump-tree-vect
5534 Dump each function after applying vectorization of loops. The file name is
5535 made by appending @file{.vect} to the source file name.
5538 @opindex fdump-tree-slp
5539 Dump each function after applying vectorization of basic blocks. The file name
5540 is made by appending @file{.slp} to the source file name.
5543 @opindex fdump-tree-vrp
5544 Dump each function after Value Range Propagation (VRP). The file name
5545 is made by appending @file{.vrp} to the source file name.
5548 @opindex fdump-tree-all
5549 Enable all the available tree dumps with the flags provided in this option.
5552 @item -ftree-vectorizer-verbose=@var{n}
5553 @opindex ftree-vectorizer-verbose
5554 This option controls the amount of debugging output the vectorizer prints.
5555 This information is written to standard error, unless
5556 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5557 in which case it is output to the usual dump listing file, @file{.vect}.
5558 For @var{n}=0 no diagnostic information is reported.
5559 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5560 and the total number of loops that got vectorized.
5561 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5562 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5563 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5564 level that @option{-fdump-tree-vect-stats} uses.
5565 Higher verbosity levels mean either more information dumped for each
5566 reported loop, or same amount of information reported for more loops:
5567 if @var{n}=3, vectorizer cost model information is reported.
5568 If @var{n}=4, alignment related information is added to the reports.
5569 If @var{n}=5, data-references related information (e.g.@: memory dependences,
5570 memory access-patterns) is added to the reports.
5571 If @var{n}=6, the vectorizer reports also non-vectorized inner-most loops
5572 that did not pass the first analysis phase (i.e., may not be countable, or
5573 may have complicated control-flow).
5574 If @var{n}=7, the vectorizer reports also non-vectorized nested loops.
5575 If @var{n}=8, SLP related information is added to the reports.
5576 For @var{n}=9, all the information the vectorizer generates during its
5577 analysis and transformation is reported. This is the same verbosity level
5578 that @option{-fdump-tree-vect-details} uses.
5580 @item -frandom-seed=@var{string}
5581 @opindex frandom-seed
5582 This option provides a seed that GCC uses when it would otherwise use
5583 random numbers. It is used to generate certain symbol names
5584 that have to be different in every compiled file. It is also used to
5585 place unique stamps in coverage data files and the object files that
5586 produce them. You can use the @option{-frandom-seed} option to produce
5587 reproducibly identical object files.
5589 The @var{string} should be different for every file you compile.
5591 @item -fsched-verbose=@var{n}
5592 @opindex fsched-verbose
5593 On targets that use instruction scheduling, this option controls the
5594 amount of debugging output the scheduler prints. This information is
5595 written to standard error, unless @option{-fdump-rtl-sched1} or
5596 @option{-fdump-rtl-sched2} is specified, in which case it is output
5597 to the usual dump listing file, @file{.sched1} or @file{.sched2}
5598 respectively. However for @var{n} greater than nine, the output is
5599 always printed to standard error.
5601 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5602 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5603 For @var{n} greater than one, it also output basic block probabilities,
5604 detailed ready list information and unit/insn info. For @var{n} greater
5605 than two, it includes RTL at abort point, control-flow and regions info.
5606 And for @var{n} over four, @option{-fsched-verbose} also includes
5610 @itemx -save-temps=cwd
5612 Store the usual ``temporary'' intermediate files permanently; place them
5613 in the current directory and name them based on the source file. Thus,
5614 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5615 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5616 preprocessed @file{foo.i} output file even though the compiler now
5617 normally uses an integrated preprocessor.
5619 When used in combination with the @option{-x} command line option,
5620 @option{-save-temps} is sensible enough to avoid over writing an
5621 input source file with the same extension as an intermediate file.
5622 The corresponding intermediate file may be obtained by renaming the
5623 source file before using @option{-save-temps}.
5625 If you invoke GCC in parallel, compiling several different source
5626 files that share a common base name in different subdirectories or the
5627 same source file compiled for multiple output destinations, it is
5628 likely that the different parallel compilers will interfere with each
5629 other, and overwrite the temporary files. For instance:
5632 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5633 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5636 may result in @file{foo.i} and @file{foo.o} being written to
5637 simultaneously by both compilers.
5639 @item -save-temps=obj
5640 @opindex save-temps=obj
5641 Store the usual ``temporary'' intermediate files permanently. If the
5642 @option{-o} option is used, the temporary files are based on the
5643 object file. If the @option{-o} option is not used, the
5644 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5649 gcc -save-temps=obj -c foo.c
5650 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5651 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5654 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5655 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5656 @file{dir2/yfoobar.o}.
5658 @item -time@r{[}=@var{file}@r{]}
5660 Report the CPU time taken by each subprocess in the compilation
5661 sequence. For C source files, this is the compiler proper and assembler
5662 (plus the linker if linking is done).
5664 Without the specification of an output file, the output looks like this:
5671 The first number on each line is the ``user time'', that is time spent
5672 executing the program itself. The second number is ``system time'',
5673 time spent executing operating system routines on behalf of the program.
5674 Both numbers are in seconds.
5676 With the specification of an output file, the output is appended to the
5677 named file, and it looks like this:
5680 0.12 0.01 cc1 @var{options}
5681 0.00 0.01 as @var{options}
5684 The ``user time'' and the ``system time'' are moved before the program
5685 name, and the options passed to the program are displayed, so that one
5686 can later tell what file was being compiled, and with which options.
5688 @item -fvar-tracking
5689 @opindex fvar-tracking
5690 Run variable tracking pass. It computes where variables are stored at each
5691 position in code. Better debugging information is then generated
5692 (if the debugging information format supports this information).
5694 It is enabled by default when compiling with optimization (@option{-Os},
5695 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5696 the debug info format supports it.
5698 @item -fvar-tracking-assignments
5699 @opindex fvar-tracking-assignments
5700 @opindex fno-var-tracking-assignments
5701 Annotate assignments to user variables early in the compilation and
5702 attempt to carry the annotations over throughout the compilation all the
5703 way to the end, in an attempt to improve debug information while
5704 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
5706 It can be enabled even if var-tracking is disabled, in which case
5707 annotations will be created and maintained, but discarded at the end.
5709 @item -fvar-tracking-assignments-toggle
5710 @opindex fvar-tracking-assignments-toggle
5711 @opindex fno-var-tracking-assignments-toggle
5712 Toggle @option{-fvar-tracking-assignments}, in the same way that
5713 @option{-gtoggle} toggles @option{-g}.
5715 @item -print-file-name=@var{library}
5716 @opindex print-file-name
5717 Print the full absolute name of the library file @var{library} that
5718 would be used when linking---and don't do anything else. With this
5719 option, GCC does not compile or link anything; it just prints the
5722 @item -print-multi-directory
5723 @opindex print-multi-directory
5724 Print the directory name corresponding to the multilib selected by any
5725 other switches present in the command line. This directory is supposed
5726 to exist in @env{GCC_EXEC_PREFIX}.
5728 @item -print-multi-lib
5729 @opindex print-multi-lib
5730 Print the mapping from multilib directory names to compiler switches
5731 that enable them. The directory name is separated from the switches by
5732 @samp{;}, and each switch starts with an @samp{@@} instead of the
5733 @samp{-}, without spaces between multiple switches. This is supposed to
5734 ease shell-processing.
5736 @item -print-multi-os-directory
5737 @opindex print-multi-os-directory
5738 Print the path to OS libraries for the selected
5739 multilib, relative to some @file{lib} subdirectory. If OS libraries are
5740 present in the @file{lib} subdirectory and no multilibs are used, this is
5741 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
5742 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
5743 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
5744 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
5746 @item -print-prog-name=@var{program}
5747 @opindex print-prog-name
5748 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5750 @item -print-libgcc-file-name
5751 @opindex print-libgcc-file-name
5752 Same as @option{-print-file-name=libgcc.a}.
5754 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5755 but you do want to link with @file{libgcc.a}. You can do
5758 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5761 @item -print-search-dirs
5762 @opindex print-search-dirs
5763 Print the name of the configured installation directory and a list of
5764 program and library directories @command{gcc} will search---and don't do anything else.
5766 This is useful when @command{gcc} prints the error message
5767 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5768 To resolve this you either need to put @file{cpp0} and the other compiler
5769 components where @command{gcc} expects to find them, or you can set the environment
5770 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5771 Don't forget the trailing @samp{/}.
5772 @xref{Environment Variables}.
5774 @item -print-sysroot
5775 @opindex print-sysroot
5776 Print the target sysroot directory that will be used during
5777 compilation. This is the target sysroot specified either at configure
5778 time or using the @option{--sysroot} option, possibly with an extra
5779 suffix that depends on compilation options. If no target sysroot is
5780 specified, the option prints nothing.
5782 @item -print-sysroot-headers-suffix
5783 @opindex print-sysroot-headers-suffix
5784 Print the suffix added to the target sysroot when searching for
5785 headers, or give an error if the compiler is not configured with such
5786 a suffix---and don't do anything else.
5789 @opindex dumpmachine
5790 Print the compiler's target machine (for example,
5791 @samp{i686-pc-linux-gnu})---and don't do anything else.
5794 @opindex dumpversion
5795 Print the compiler version (for example, @samp{3.0})---and don't do
5800 Print the compiler's built-in specs---and don't do anything else. (This
5801 is used when GCC itself is being built.) @xref{Spec Files}.
5803 @item -feliminate-unused-debug-types
5804 @opindex feliminate-unused-debug-types
5805 Normally, when producing DWARF2 output, GCC will emit debugging
5806 information for all types declared in a compilation
5807 unit, regardless of whether or not they are actually used
5808 in that compilation unit. Sometimes this is useful, such as
5809 if, in the debugger, you want to cast a value to a type that is
5810 not actually used in your program (but is declared). More often,
5811 however, this results in a significant amount of wasted space.
5812 With this option, GCC will avoid producing debug symbol output
5813 for types that are nowhere used in the source file being compiled.
5816 @node Optimize Options
5817 @section Options That Control Optimization
5818 @cindex optimize options
5819 @cindex options, optimization
5821 These options control various sorts of optimizations.
5823 Without any optimization option, the compiler's goal is to reduce the
5824 cost of compilation and to make debugging produce the expected
5825 results. Statements are independent: if you stop the program with a
5826 breakpoint between statements, you can then assign a new value to any
5827 variable or change the program counter to any other statement in the
5828 function and get exactly the results you would expect from the source
5831 Turning on optimization flags makes the compiler attempt to improve
5832 the performance and/or code size at the expense of compilation time
5833 and possibly the ability to debug the program.
5835 The compiler performs optimization based on the knowledge it has of the
5836 program. Compiling multiple files at once to a single output file mode allows
5837 the compiler to use information gained from all of the files when compiling
5840 Not all optimizations are controlled directly by a flag. Only
5841 optimizations that have a flag are listed in this section.
5843 Most optimizations are only enabled if an @option{-O} level is set on
5844 the command line. Otherwise they are disabled, even if individual
5845 optimization flags are specified.
5847 Depending on the target and how GCC was configured, a slightly different
5848 set of optimizations may be enabled at each @option{-O} level than
5849 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
5850 to find out the exact set of optimizations that are enabled at each level.
5851 @xref{Overall Options}, for examples.
5858 Optimize. Optimizing compilation takes somewhat more time, and a lot
5859 more memory for a large function.
5861 With @option{-O}, the compiler tries to reduce code size and execution
5862 time, without performing any optimizations that take a great deal of
5865 @option{-O} turns on the following optimization flags:
5868 -fcprop-registers @gol
5871 -fdelayed-branch @gol
5873 -fguess-branch-probability @gol
5874 -fif-conversion2 @gol
5875 -fif-conversion @gol
5876 -fipa-pure-const @gol
5878 -fipa-reference @gol
5880 -fsplit-wide-types @gol
5882 -ftree-builtin-call-dce @gol
5885 -ftree-copyrename @gol
5887 -ftree-dominator-opts @gol
5889 -ftree-forwprop @gol
5897 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5898 where doing so does not interfere with debugging.
5902 Optimize even more. GCC performs nearly all supported optimizations
5903 that do not involve a space-speed tradeoff.
5904 As compared to @option{-O}, this option increases both compilation time
5905 and the performance of the generated code.
5907 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5908 also turns on the following optimization flags:
5909 @gccoptlist{-fthread-jumps @gol
5910 -falign-functions -falign-jumps @gol
5911 -falign-loops -falign-labels @gol
5914 -fcse-follow-jumps -fcse-skip-blocks @gol
5915 -fdelete-null-pointer-checks @gol
5916 -fexpensive-optimizations @gol
5917 -fgcse -fgcse-lm @gol
5918 -finline-small-functions @gol
5919 -findirect-inlining @gol
5921 -foptimize-sibling-calls @gol
5922 -fpartial-inlining @gol
5925 -freorder-blocks -freorder-functions @gol
5926 -frerun-cse-after-loop @gol
5927 -fsched-interblock -fsched-spec @gol
5928 -fschedule-insns -fschedule-insns2 @gol
5929 -fstrict-aliasing -fstrict-overflow @gol
5930 -ftree-switch-conversion @gol
5934 Please note the warning under @option{-fgcse} about
5935 invoking @option{-O2} on programs that use computed gotos.
5939 Optimize yet more. @option{-O3} turns on all optimizations specified
5940 by @option{-O2} and also turns on the @option{-finline-functions},
5941 @option{-funswitch-loops}, @option{-fpredictive-commoning},
5942 @option{-fgcse-after-reload}, @option{-ftree-vectorize} and
5943 @option{-fipa-cp-clone} options.
5947 Reduce compilation time and make debugging produce the expected
5948 results. This is the default.
5952 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5953 do not typically increase code size. It also performs further
5954 optimizations designed to reduce code size.
5956 @option{-Os} disables the following optimization flags:
5957 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5958 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5959 -fprefetch-loop-arrays -ftree-vect-loop-version}
5963 Disregard strict standards compliance. @option{-Ofast} enables all
5964 @option{-O3} optimizations. It also enables optimizations that are not
5965 valid for all standard compliant programs.
5966 It turns on @option{-ffast-math}.
5968 If you use multiple @option{-O} options, with or without level numbers,
5969 the last such option is the one that is effective.
5972 Options of the form @option{-f@var{flag}} specify machine-independent
5973 flags. Most flags have both positive and negative forms; the negative
5974 form of @option{-ffoo} would be @option{-fno-foo}. In the table
5975 below, only one of the forms is listed---the one you typically will
5976 use. You can figure out the other form by either removing @samp{no-}
5979 The following options control specific optimizations. They are either
5980 activated by @option{-O} options or are related to ones that are. You
5981 can use the following flags in the rare cases when ``fine-tuning'' of
5982 optimizations to be performed is desired.
5985 @item -fno-default-inline
5986 @opindex fno-default-inline
5987 Do not make member functions inline by default merely because they are
5988 defined inside the class scope (C++ only). Otherwise, when you specify
5989 @w{@option{-O}}, member functions defined inside class scope are compiled
5990 inline by default; i.e., you don't need to add @samp{inline} in front of
5991 the member function name.
5993 @item -fno-defer-pop
5994 @opindex fno-defer-pop
5995 Always pop the arguments to each function call as soon as that function
5996 returns. For machines which must pop arguments after a function call,
5997 the compiler normally lets arguments accumulate on the stack for several
5998 function calls and pops them all at once.
6000 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6002 @item -fforward-propagate
6003 @opindex fforward-propagate
6004 Perform a forward propagation pass on RTL@. The pass tries to combine two
6005 instructions and checks if the result can be simplified. If loop unrolling
6006 is active, two passes are performed and the second is scheduled after
6009 This option is enabled by default at optimization levels @option{-O},
6010 @option{-O2}, @option{-O3}, @option{-Os}.
6012 @item -ffp-contract=@var{style}
6013 @opindex ffp-contract
6014 @option{-ffp-contract=off} disables floating-point expression contraction.
6015 @option{-ffp-contract=fast} enables floating-point expression contraction
6016 such as forming of fused multiply-add operations if the target has
6017 native support for them.
6018 @option{-ffp-contract=on} enables floating-point expression contraction
6019 if allowed by the language standard. This is currently not implemented
6020 and treated equal to @option{-ffp-contract=off}.
6022 The default is @option{-ffp-contract=fast}.
6024 @item -fomit-frame-pointer
6025 @opindex fomit-frame-pointer
6026 Don't keep the frame pointer in a register for functions that
6027 don't need one. This avoids the instructions to save, set up and
6028 restore frame pointers; it also makes an extra register available
6029 in many functions. @strong{It also makes debugging impossible on
6032 On some machines, such as the VAX, this flag has no effect, because
6033 the standard calling sequence automatically handles the frame pointer
6034 and nothing is saved by pretending it doesn't exist. The
6035 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
6036 whether a target machine supports this flag. @xref{Registers,,Register
6037 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
6039 Starting with GCC version 4.6, the default setting (when not optimizing for
6040 size) for 32-bit Linux x86 and 32-bit Darwin x86 targets has been changed to
6041 @option{-fomit-frame-pointer}. The default can be reverted to
6042 @option{-fno-omit-frame-pointer} by configuring GCC with the
6043 @option{--enable-frame-pointer} configure option.
6045 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6047 @item -foptimize-sibling-calls
6048 @opindex foptimize-sibling-calls
6049 Optimize sibling and tail recursive calls.
6051 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6055 Don't pay attention to the @code{inline} keyword. Normally this option
6056 is used to keep the compiler from expanding any functions inline.
6057 Note that if you are not optimizing, no functions can be expanded inline.
6059 @item -finline-small-functions
6060 @opindex finline-small-functions
6061 Integrate functions into their callers when their body is smaller than expected
6062 function call code (so overall size of program gets smaller). The compiler
6063 heuristically decides which functions are simple enough to be worth integrating
6066 Enabled at level @option{-O2}.
6068 @item -findirect-inlining
6069 @opindex findirect-inlining
6070 Inline also indirect calls that are discovered to be known at compile
6071 time thanks to previous inlining. This option has any effect only
6072 when inlining itself is turned on by the @option{-finline-functions}
6073 or @option{-finline-small-functions} options.
6075 Enabled at level @option{-O2}.
6077 @item -finline-functions
6078 @opindex finline-functions
6079 Integrate all simple functions into their callers. The compiler
6080 heuristically decides which functions are simple enough to be worth
6081 integrating in this way.
6083 If all calls to a given function are integrated, and the function is
6084 declared @code{static}, then the function is normally not output as
6085 assembler code in its own right.
6087 Enabled at level @option{-O3}.
6089 @item -finline-functions-called-once
6090 @opindex finline-functions-called-once
6091 Consider all @code{static} functions called once for inlining into their
6092 caller even if they are not marked @code{inline}. If a call to a given
6093 function is integrated, then the function is not output as assembler code
6096 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
6098 @item -fearly-inlining
6099 @opindex fearly-inlining
6100 Inline functions marked by @code{always_inline} and functions whose body seems
6101 smaller than the function call overhead early before doing
6102 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
6103 makes profiling significantly cheaper and usually inlining faster on programs
6104 having large chains of nested wrapper functions.
6110 Perform interprocedural scalar replacement of aggregates, removal of
6111 unused parameters and replacement of parameters passed by reference
6112 by parameters passed by value.
6114 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
6116 @item -finline-limit=@var{n}
6117 @opindex finline-limit
6118 By default, GCC limits the size of functions that can be inlined. This flag
6119 allows coarse control of this limit. @var{n} is the size of functions that
6120 can be inlined in number of pseudo instructions.
6122 Inlining is actually controlled by a number of parameters, which may be
6123 specified individually by using @option{--param @var{name}=@var{value}}.
6124 The @option{-finline-limit=@var{n}} option sets some of these parameters
6128 @item max-inline-insns-single
6129 is set to @var{n}/2.
6130 @item max-inline-insns-auto
6131 is set to @var{n}/2.
6134 See below for a documentation of the individual
6135 parameters controlling inlining and for the defaults of these parameters.
6137 @emph{Note:} there may be no value to @option{-finline-limit} that results
6138 in default behavior.
6140 @emph{Note:} pseudo instruction represents, in this particular context, an
6141 abstract measurement of function's size. In no way does it represent a count
6142 of assembly instructions and as such its exact meaning might change from one
6143 release to an another.
6145 @item -fkeep-inline-functions
6146 @opindex fkeep-inline-functions
6147 In C, emit @code{static} functions that are declared @code{inline}
6148 into the object file, even if the function has been inlined into all
6149 of its callers. This switch does not affect functions using the
6150 @code{extern inline} extension in GNU C90@. In C++, emit any and all
6151 inline functions into the object file.
6153 @item -fkeep-static-consts
6154 @opindex fkeep-static-consts
6155 Emit variables declared @code{static const} when optimization isn't turned
6156 on, even if the variables aren't referenced.
6158 GCC enables this option by default. If you want to force the compiler to
6159 check if the variable was referenced, regardless of whether or not
6160 optimization is turned on, use the @option{-fno-keep-static-consts} option.
6162 @item -fmerge-constants
6163 @opindex fmerge-constants
6164 Attempt to merge identical constants (string constants and floating point
6165 constants) across compilation units.
6167 This option is the default for optimized compilation if the assembler and
6168 linker support it. Use @option{-fno-merge-constants} to inhibit this
6171 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6173 @item -fmerge-all-constants
6174 @opindex fmerge-all-constants
6175 Attempt to merge identical constants and identical variables.
6177 This option implies @option{-fmerge-constants}. In addition to
6178 @option{-fmerge-constants} this considers e.g.@: even constant initialized
6179 arrays or initialized constant variables with integral or floating point
6180 types. Languages like C or C++ require each variable, including multiple
6181 instances of the same variable in recursive calls, to have distinct locations,
6182 so using this option will result in non-conforming
6185 @item -fmodulo-sched
6186 @opindex fmodulo-sched
6187 Perform swing modulo scheduling immediately before the first scheduling
6188 pass. This pass looks at innermost loops and reorders their
6189 instructions by overlapping different iterations.
6191 @item -fmodulo-sched-allow-regmoves
6192 @opindex fmodulo-sched-allow-regmoves
6193 Perform more aggressive SMS based modulo scheduling with register moves
6194 allowed. By setting this flag certain anti-dependences edges will be
6195 deleted which will trigger the generation of reg-moves based on the
6196 life-range analysis. This option is effective only with
6197 @option{-fmodulo-sched} enabled.
6199 @item -fno-branch-count-reg
6200 @opindex fno-branch-count-reg
6201 Do not use ``decrement and branch'' instructions on a count register,
6202 but instead generate a sequence of instructions that decrement a
6203 register, compare it against zero, then branch based upon the result.
6204 This option is only meaningful on architectures that support such
6205 instructions, which include x86, PowerPC, IA-64 and S/390.
6207 The default is @option{-fbranch-count-reg}.
6209 @item -fno-function-cse
6210 @opindex fno-function-cse
6211 Do not put function addresses in registers; make each instruction that
6212 calls a constant function contain the function's address explicitly.
6214 This option results in less efficient code, but some strange hacks
6215 that alter the assembler output may be confused by the optimizations
6216 performed when this option is not used.
6218 The default is @option{-ffunction-cse}
6220 @item -fno-zero-initialized-in-bss
6221 @opindex fno-zero-initialized-in-bss
6222 If the target supports a BSS section, GCC by default puts variables that
6223 are initialized to zero into BSS@. This can save space in the resulting
6226 This option turns off this behavior because some programs explicitly
6227 rely on variables going to the data section. E.g., so that the
6228 resulting executable can find the beginning of that section and/or make
6229 assumptions based on that.
6231 The default is @option{-fzero-initialized-in-bss}.
6233 @item -fmudflap -fmudflapth -fmudflapir
6237 @cindex bounds checking
6239 For front-ends that support it (C and C++), instrument all risky
6240 pointer/array dereferencing operations, some standard library
6241 string/heap functions, and some other associated constructs with
6242 range/validity tests. Modules so instrumented should be immune to
6243 buffer overflows, invalid heap use, and some other classes of C/C++
6244 programming errors. The instrumentation relies on a separate runtime
6245 library (@file{libmudflap}), which will be linked into a program if
6246 @option{-fmudflap} is given at link time. Run-time behavior of the
6247 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
6248 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
6251 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
6252 link if your program is multi-threaded. Use @option{-fmudflapir}, in
6253 addition to @option{-fmudflap} or @option{-fmudflapth}, if
6254 instrumentation should ignore pointer reads. This produces less
6255 instrumentation (and therefore faster execution) and still provides
6256 some protection against outright memory corrupting writes, but allows
6257 erroneously read data to propagate within a program.
6259 @item -fthread-jumps
6260 @opindex fthread-jumps
6261 Perform optimizations where we check to see if a jump branches to a
6262 location where another comparison subsumed by the first is found. If
6263 so, the first branch is redirected to either the destination of the
6264 second branch or a point immediately following it, depending on whether
6265 the condition is known to be true or false.
6267 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6269 @item -fsplit-wide-types
6270 @opindex fsplit-wide-types
6271 When using a type that occupies multiple registers, such as @code{long
6272 long} on a 32-bit system, split the registers apart and allocate them
6273 independently. This normally generates better code for those types,
6274 but may make debugging more difficult.
6276 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
6279 @item -fcse-follow-jumps
6280 @opindex fcse-follow-jumps
6281 In common subexpression elimination (CSE), scan through jump instructions
6282 when the target of the jump is not reached by any other path. For
6283 example, when CSE encounters an @code{if} statement with an
6284 @code{else} clause, CSE will follow the jump when the condition
6287 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6289 @item -fcse-skip-blocks
6290 @opindex fcse-skip-blocks
6291 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
6292 follow jumps which conditionally skip over blocks. When CSE
6293 encounters a simple @code{if} statement with no else clause,
6294 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6295 body of the @code{if}.
6297 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6299 @item -frerun-cse-after-loop
6300 @opindex frerun-cse-after-loop
6301 Re-run common subexpression elimination after loop optimizations has been
6304 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6308 Perform a global common subexpression elimination pass.
6309 This pass also performs global constant and copy propagation.
6311 @emph{Note:} When compiling a program using computed gotos, a GCC
6312 extension, you may get better runtime performance if you disable
6313 the global common subexpression elimination pass by adding
6314 @option{-fno-gcse} to the command line.
6316 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6320 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
6321 attempt to move loads which are only killed by stores into themselves. This
6322 allows a loop containing a load/store sequence to be changed to a load outside
6323 the loop, and a copy/store within the loop.
6325 Enabled by default when gcse is enabled.
6329 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6330 global common subexpression elimination. This pass will attempt to move
6331 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6332 loops containing a load/store sequence can be changed to a load before
6333 the loop and a store after the loop.
6335 Not enabled at any optimization level.
6339 When @option{-fgcse-las} is enabled, the global common subexpression
6340 elimination pass eliminates redundant loads that come after stores to the
6341 same memory location (both partial and full redundancies).
6343 Not enabled at any optimization level.
6345 @item -fgcse-after-reload
6346 @opindex fgcse-after-reload
6347 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6348 pass is performed after reload. The purpose of this pass is to cleanup
6351 @item -funsafe-loop-optimizations
6352 @opindex funsafe-loop-optimizations
6353 If given, the loop optimizer will assume that loop indices do not
6354 overflow, and that the loops with nontrivial exit condition are not
6355 infinite. This enables a wider range of loop optimizations even if
6356 the loop optimizer itself cannot prove that these assumptions are valid.
6357 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
6358 if it finds this kind of loop.
6360 @item -fcrossjumping
6361 @opindex fcrossjumping
6362 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6363 resulting code may or may not perform better than without cross-jumping.
6365 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6367 @item -fauto-inc-dec
6368 @opindex fauto-inc-dec
6369 Combine increments or decrements of addresses with memory accesses.
6370 This pass is always skipped on architectures that do not have
6371 instructions to support this. Enabled by default at @option{-O} and
6372 higher on architectures that support this.
6376 Perform dead code elimination (DCE) on RTL@.
6377 Enabled by default at @option{-O} and higher.
6381 Perform dead store elimination (DSE) on RTL@.
6382 Enabled by default at @option{-O} and higher.
6384 @item -fif-conversion
6385 @opindex fif-conversion
6386 Attempt to transform conditional jumps into branch-less equivalents. This
6387 include use of conditional moves, min, max, set flags and abs instructions, and
6388 some tricks doable by standard arithmetics. The use of conditional execution
6389 on chips where it is available is controlled by @code{if-conversion2}.
6391 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6393 @item -fif-conversion2
6394 @opindex fif-conversion2
6395 Use conditional execution (where available) to transform conditional jumps into
6396 branch-less equivalents.
6398 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6400 @item -fdelete-null-pointer-checks
6401 @opindex fdelete-null-pointer-checks
6402 Assume that programs cannot safely dereference null pointers, and that
6403 no code or data element resides there. This enables simple constant
6404 folding optimizations at all optimization levels. In addition, other
6405 optimization passes in GCC use this flag to control global dataflow
6406 analyses that eliminate useless checks for null pointers; these assume
6407 that if a pointer is checked after it has already been dereferenced,
6410 Note however that in some environments this assumption is not true.
6411 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6412 for programs which depend on that behavior.
6414 Some targets, especially embedded ones, disable this option at all levels.
6415 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6416 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6417 are enabled independently at different optimization levels.
6419 @item -fexpensive-optimizations
6420 @opindex fexpensive-optimizations
6421 Perform a number of minor optimizations that are relatively expensive.
6423 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6425 @item -foptimize-register-move
6427 @opindex foptimize-register-move
6429 Attempt to reassign register numbers in move instructions and as
6430 operands of other simple instructions in order to maximize the amount of
6431 register tying. This is especially helpful on machines with two-operand
6434 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6437 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6439 @item -fira-algorithm=@var{algorithm}
6440 Use specified coloring algorithm for the integrated register
6441 allocator. The @var{algorithm} argument should be @code{priority} or
6442 @code{CB}. The first algorithm specifies Chow's priority coloring,
6443 the second one specifies Chaitin-Briggs coloring. The second
6444 algorithm can be unimplemented for some architectures. If it is
6445 implemented, it is the default because Chaitin-Briggs coloring as a
6446 rule generates a better code.
6448 @item -fira-region=@var{region}
6449 Use specified regions for the integrated register allocator. The
6450 @var{region} argument should be one of @code{all}, @code{mixed}, or
6451 @code{one}. The first value means using all loops as register
6452 allocation regions, the second value which is the default means using
6453 all loops except for loops with small register pressure as the
6454 regions, and third one means using all function as a single region.
6455 The first value can give best result for machines with small size and
6456 irregular register set, the third one results in faster and generates
6457 decent code and the smallest size code, and the default value usually
6458 give the best results in most cases and for most architectures.
6460 @item -fira-loop-pressure
6461 @opindex fira-loop-pressure
6462 Use IRA to evaluate register pressure in loops for decision to move
6463 loop invariants. Usage of this option usually results in generation
6464 of faster and smaller code on machines with big register files (>= 32
6465 registers) but it can slow compiler down.
6467 This option is enabled at level @option{-O3} for some targets.
6469 @item -fno-ira-share-save-slots
6470 @opindex fno-ira-share-save-slots
6471 Switch off sharing stack slots used for saving call used hard
6472 registers living through a call. Each hard register will get a
6473 separate stack slot and as a result function stack frame will be
6476 @item -fno-ira-share-spill-slots
6477 @opindex fno-ira-share-spill-slots
6478 Switch off sharing stack slots allocated for pseudo-registers. Each
6479 pseudo-register which did not get a hard register will get a separate
6480 stack slot and as a result function stack frame will be bigger.
6482 @item -fira-verbose=@var{n}
6483 @opindex fira-verbose
6484 Set up how verbose dump file for the integrated register allocator
6485 will be. Default value is 5. If the value is greater or equal to 10,
6486 the dump file will be stderr as if the value were @var{n} minus 10.
6488 @item -fdelayed-branch
6489 @opindex fdelayed-branch
6490 If supported for the target machine, attempt to reorder instructions
6491 to exploit instruction slots available after delayed branch
6494 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6496 @item -fschedule-insns
6497 @opindex fschedule-insns
6498 If supported for the target machine, attempt to reorder instructions to
6499 eliminate execution stalls due to required data being unavailable. This
6500 helps machines that have slow floating point or memory load instructions
6501 by allowing other instructions to be issued until the result of the load
6502 or floating point instruction is required.
6504 Enabled at levels @option{-O2}, @option{-O3}.
6506 @item -fschedule-insns2
6507 @opindex fschedule-insns2
6508 Similar to @option{-fschedule-insns}, but requests an additional pass of
6509 instruction scheduling after register allocation has been done. This is
6510 especially useful on machines with a relatively small number of
6511 registers and where memory load instructions take more than one cycle.
6513 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6515 @item -fno-sched-interblock
6516 @opindex fno-sched-interblock
6517 Don't schedule instructions across basic blocks. This is normally
6518 enabled by default when scheduling before register allocation, i.e.@:
6519 with @option{-fschedule-insns} or at @option{-O2} or higher.
6521 @item -fno-sched-spec
6522 @opindex fno-sched-spec
6523 Don't allow speculative motion of non-load instructions. This is normally
6524 enabled by default when scheduling before register allocation, i.e.@:
6525 with @option{-fschedule-insns} or at @option{-O2} or higher.
6527 @item -fsched-pressure
6528 @opindex fsched-pressure
6529 Enable register pressure sensitive insn scheduling before the register
6530 allocation. This only makes sense when scheduling before register
6531 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
6532 @option{-O2} or higher. Usage of this option can improve the
6533 generated code and decrease its size by preventing register pressure
6534 increase above the number of available hard registers and as a
6535 consequence register spills in the register allocation.
6537 @item -fsched-spec-load
6538 @opindex fsched-spec-load
6539 Allow speculative motion of some load instructions. This only makes
6540 sense when scheduling before register allocation, i.e.@: with
6541 @option{-fschedule-insns} or at @option{-O2} or higher.
6543 @item -fsched-spec-load-dangerous
6544 @opindex fsched-spec-load-dangerous
6545 Allow speculative motion of more load instructions. This only makes
6546 sense when scheduling before register allocation, i.e.@: with
6547 @option{-fschedule-insns} or at @option{-O2} or higher.
6549 @item -fsched-stalled-insns
6550 @itemx -fsched-stalled-insns=@var{n}
6551 @opindex fsched-stalled-insns
6552 Define how many insns (if any) can be moved prematurely from the queue
6553 of stalled insns into the ready list, during the second scheduling pass.
6554 @option{-fno-sched-stalled-insns} means that no insns will be moved
6555 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6556 on how many queued insns can be moved prematurely.
6557 @option{-fsched-stalled-insns} without a value is equivalent to
6558 @option{-fsched-stalled-insns=1}.
6560 @item -fsched-stalled-insns-dep
6561 @itemx -fsched-stalled-insns-dep=@var{n}
6562 @opindex fsched-stalled-insns-dep
6563 Define how many insn groups (cycles) will be examined for a dependency
6564 on a stalled insn that is candidate for premature removal from the queue
6565 of stalled insns. This has an effect only during the second scheduling pass,
6566 and only if @option{-fsched-stalled-insns} is used.
6567 @option{-fno-sched-stalled-insns-dep} is equivalent to
6568 @option{-fsched-stalled-insns-dep=0}.
6569 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6570 @option{-fsched-stalled-insns-dep=1}.
6572 @item -fsched2-use-superblocks
6573 @opindex fsched2-use-superblocks
6574 When scheduling after register allocation, do use superblock scheduling
6575 algorithm. Superblock scheduling allows motion across basic block boundaries
6576 resulting on faster schedules. This option is experimental, as not all machine
6577 descriptions used by GCC model the CPU closely enough to avoid unreliable
6578 results from the algorithm.
6580 This only makes sense when scheduling after register allocation, i.e.@: with
6581 @option{-fschedule-insns2} or at @option{-O2} or higher.
6583 @item -fsched-group-heuristic
6584 @opindex fsched-group-heuristic
6585 Enable the group heuristic in the scheduler. This heuristic favors
6586 the instruction that belongs to a schedule group. This is enabled
6587 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6588 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6590 @item -fsched-critical-path-heuristic
6591 @opindex fsched-critical-path-heuristic
6592 Enable the critical-path heuristic in the scheduler. This heuristic favors
6593 instructions on the critical path. This is enabled by default when
6594 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6595 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6597 @item -fsched-spec-insn-heuristic
6598 @opindex fsched-spec-insn-heuristic
6599 Enable the speculative instruction heuristic in the scheduler. This
6600 heuristic favors speculative instructions with greater dependency weakness.
6601 This is enabled by default when scheduling is enabled, i.e.@:
6602 with @option{-fschedule-insns} or @option{-fschedule-insns2}
6603 or at @option{-O2} or higher.
6605 @item -fsched-rank-heuristic
6606 @opindex fsched-rank-heuristic
6607 Enable the rank heuristic in the scheduler. This heuristic favors
6608 the instruction belonging to a basic block with greater size or frequency.
6609 This is enabled by default when scheduling is enabled, i.e.@:
6610 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6611 at @option{-O2} or higher.
6613 @item -fsched-last-insn-heuristic
6614 @opindex fsched-last-insn-heuristic
6615 Enable the last-instruction heuristic in the scheduler. This heuristic
6616 favors the instruction that is less dependent on the last instruction
6617 scheduled. This is enabled by default when scheduling is enabled,
6618 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6619 at @option{-O2} or higher.
6621 @item -fsched-dep-count-heuristic
6622 @opindex fsched-dep-count-heuristic
6623 Enable the dependent-count heuristic in the scheduler. This heuristic
6624 favors the instruction that has more instructions depending on it.
6625 This is enabled by default when scheduling is enabled, i.e.@:
6626 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6627 at @option{-O2} or higher.
6629 @item -freschedule-modulo-scheduled-loops
6630 @opindex freschedule-modulo-scheduled-loops
6631 The modulo scheduling comes before the traditional scheduling, if a loop
6632 was modulo scheduled we may want to prevent the later scheduling passes
6633 from changing its schedule, we use this option to control that.
6635 @item -fselective-scheduling
6636 @opindex fselective-scheduling
6637 Schedule instructions using selective scheduling algorithm. Selective
6638 scheduling runs instead of the first scheduler pass.
6640 @item -fselective-scheduling2
6641 @opindex fselective-scheduling2
6642 Schedule instructions using selective scheduling algorithm. Selective
6643 scheduling runs instead of the second scheduler pass.
6645 @item -fsel-sched-pipelining
6646 @opindex fsel-sched-pipelining
6647 Enable software pipelining of innermost loops during selective scheduling.
6648 This option has no effect until one of @option{-fselective-scheduling} or
6649 @option{-fselective-scheduling2} is turned on.
6651 @item -fsel-sched-pipelining-outer-loops
6652 @opindex fsel-sched-pipelining-outer-loops
6653 When pipelining loops during selective scheduling, also pipeline outer loops.
6654 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6656 @item -fcaller-saves
6657 @opindex fcaller-saves
6658 Enable values to be allocated in registers that will be clobbered by
6659 function calls, by emitting extra instructions to save and restore the
6660 registers around such calls. Such allocation is done only when it
6661 seems to result in better code than would otherwise be produced.
6663 This option is always enabled by default on certain machines, usually
6664 those which have no call-preserved registers to use instead.
6666 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6668 @item -fcombine-stack-adjustments
6669 @opindex fcombine-stack-adjustments
6670 Tracks stack adjustments (pushes and pops) and stack memory references
6671 and then tries to find ways to combine them.
6673 Enabled by default at @option{-O1} and higher.
6675 @item -fconserve-stack
6676 @opindex fconserve-stack
6677 Attempt to minimize stack usage. The compiler will attempt to use less
6678 stack space, even if that makes the program slower. This option
6679 implies setting the @option{large-stack-frame} parameter to 100
6680 and the @option{large-stack-frame-growth} parameter to 400.
6682 @item -ftree-reassoc
6683 @opindex ftree-reassoc
6684 Perform reassociation on trees. This flag is enabled by default
6685 at @option{-O} and higher.
6689 Perform partial redundancy elimination (PRE) on trees. This flag is
6690 enabled by default at @option{-O2} and @option{-O3}.
6692 @item -ftree-forwprop
6693 @opindex ftree-forwprop
6694 Perform forward propagation on trees. This flag is enabled by default
6695 at @option{-O} and higher.
6699 Perform full redundancy elimination (FRE) on trees. The difference
6700 between FRE and PRE is that FRE only considers expressions
6701 that are computed on all paths leading to the redundant computation.
6702 This analysis is faster than PRE, though it exposes fewer redundancies.
6703 This flag is enabled by default at @option{-O} and higher.
6705 @item -ftree-phiprop
6706 @opindex ftree-phiprop
6707 Perform hoisting of loads from conditional pointers on trees. This
6708 pass is enabled by default at @option{-O} and higher.
6710 @item -ftree-copy-prop
6711 @opindex ftree-copy-prop
6712 Perform copy propagation on trees. This pass eliminates unnecessary
6713 copy operations. This flag is enabled by default at @option{-O} and
6716 @item -fipa-pure-const
6717 @opindex fipa-pure-const
6718 Discover which functions are pure or constant.
6719 Enabled by default at @option{-O} and higher.
6721 @item -fipa-reference
6722 @opindex fipa-reference
6723 Discover which static variables do not escape cannot escape the
6725 Enabled by default at @option{-O} and higher.
6727 @item -fipa-struct-reorg
6728 @opindex fipa-struct-reorg
6729 Perform structure reorganization optimization, that change C-like structures
6730 layout in order to better utilize spatial locality. This transformation is
6731 affective for programs containing arrays of structures. Available in two
6732 compilation modes: profile-based (enabled with @option{-fprofile-generate})
6733 or static (which uses built-in heuristics). It works only in whole program
6734 mode, so it requires @option{-fwhole-program} to be
6735 enabled. Structures considered @samp{cold} by this transformation are not
6736 affected (see @option{--param struct-reorg-cold-struct-ratio=@var{value}}).
6738 With this flag, the program debug info reflects a new structure layout.
6742 Perform interprocedural pointer analysis and interprocedural modification
6743 and reference analysis. This option can cause excessive memory and
6744 compile-time usage on large compilation units. It is not enabled by
6745 default at any optimization level.
6748 @opindex fipa-profile
6749 Perform interprocedural profile propagation. The functions called only from
6750 cold functions are marked as cold. Also functions executed once (such as
6751 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
6752 functions and loop less parts of functions executed once are then optimized for
6754 Enabled by default at @option{-O} and higher.
6758 Perform interprocedural constant propagation.
6759 This optimization analyzes the program to determine when values passed
6760 to functions are constants and then optimizes accordingly.
6761 This optimization can substantially increase performance
6762 if the application has constants passed to functions.
6763 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
6765 @item -fipa-cp-clone
6766 @opindex fipa-cp-clone
6767 Perform function cloning to make interprocedural constant propagation stronger.
6768 When enabled, interprocedural constant propagation will perform function cloning
6769 when externally visible function can be called with constant arguments.
6770 Because this optimization can create multiple copies of functions,
6771 it may significantly increase code size
6772 (see @option{--param ipcp-unit-growth=@var{value}}).
6773 This flag is enabled by default at @option{-O3}.
6775 @item -fipa-matrix-reorg
6776 @opindex fipa-matrix-reorg
6777 Perform matrix flattening and transposing.
6778 Matrix flattening tries to replace an @math{m}-dimensional matrix
6779 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
6780 This reduces the level of indirection needed for accessing the elements
6781 of the matrix. The second optimization is matrix transposing that
6782 attempts to change the order of the matrix's dimensions in order to
6783 improve cache locality.
6784 Both optimizations need the @option{-fwhole-program} flag.
6785 Transposing is enabled only if profiling information is available.
6789 Perform forward store motion on trees. This flag is
6790 enabled by default at @option{-O} and higher.
6792 @item -ftree-bit-ccp
6793 @opindex ftree-bit-ccp
6794 Perform sparse conditional bit constant propagation on trees and propagate
6795 pointer alignment information.
6796 This pass only operates on local scalar variables and is enabled by default
6797 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
6801 Perform sparse conditional constant propagation (CCP) on trees. This
6802 pass only operates on local scalar variables and is enabled by default
6803 at @option{-O} and higher.
6805 @item -ftree-switch-conversion
6806 Perform conversion of simple initializations in a switch to
6807 initializations from a scalar array. This flag is enabled by default
6808 at @option{-O2} and higher.
6812 Perform dead code elimination (DCE) on trees. This flag is enabled by
6813 default at @option{-O} and higher.
6815 @item -ftree-builtin-call-dce
6816 @opindex ftree-builtin-call-dce
6817 Perform conditional dead code elimination (DCE) for calls to builtin functions
6818 that may set @code{errno} but are otherwise side-effect free. This flag is
6819 enabled by default at @option{-O2} and higher if @option{-Os} is not also
6822 @item -ftree-dominator-opts
6823 @opindex ftree-dominator-opts
6824 Perform a variety of simple scalar cleanups (constant/copy
6825 propagation, redundancy elimination, range propagation and expression
6826 simplification) based on a dominator tree traversal. This also
6827 performs jump threading (to reduce jumps to jumps). This flag is
6828 enabled by default at @option{-O} and higher.
6832 Perform dead store elimination (DSE) on trees. A dead store is a store into
6833 a memory location which will later be overwritten by another store without
6834 any intervening loads. In this case the earlier store can be deleted. This
6835 flag is enabled by default at @option{-O} and higher.
6839 Perform loop header copying on trees. This is beneficial since it increases
6840 effectiveness of code motion optimizations. It also saves one jump. This flag
6841 is enabled by default at @option{-O} and higher. It is not enabled
6842 for @option{-Os}, since it usually increases code size.
6844 @item -ftree-loop-optimize
6845 @opindex ftree-loop-optimize
6846 Perform loop optimizations on trees. This flag is enabled by default
6847 at @option{-O} and higher.
6849 @item -ftree-loop-linear
6850 @opindex ftree-loop-linear
6851 Perform linear loop transformations on tree. This flag can improve cache
6852 performance and allow further loop optimizations to take place.
6854 @item -floop-interchange
6855 @opindex floop-interchange
6856 Perform loop interchange transformations on loops. Interchanging two
6857 nested loops switches the inner and outer loops. For example, given a
6862 A(J, I) = A(J, I) * C
6866 loop interchange will transform the loop as if the user had written:
6870 A(J, I) = A(J, I) * C
6874 which can be beneficial when @code{N} is larger than the caches,
6875 because in Fortran, the elements of an array are stored in memory
6876 contiguously by column, and the original loop iterates over rows,
6877 potentially creating at each access a cache miss. This optimization
6878 applies to all the languages supported by GCC and is not limited to
6879 Fortran. To use this code transformation, GCC has to be configured
6880 with @option{--with-ppl} and @option{--with-cloog} to enable the
6881 Graphite loop transformation infrastructure.
6883 @item -floop-strip-mine
6884 @opindex floop-strip-mine
6885 Perform loop strip mining transformations on loops. Strip mining
6886 splits a loop into two nested loops. The outer loop has strides
6887 equal to the strip size and the inner loop has strides of the
6888 original loop within a strip. The strip length can be changed
6889 using the @option{loop-block-tile-size} parameter. For example,
6896 loop strip mining will transform the loop as if the user had written:
6899 DO I = II, min (II + 50, N)
6904 This optimization applies to all the languages supported by GCC and is
6905 not limited to Fortran. To use this code transformation, GCC has to
6906 be configured with @option{--with-ppl} and @option{--with-cloog} to
6907 enable the Graphite loop transformation infrastructure.
6910 @opindex floop-block
6911 Perform loop blocking transformations on loops. Blocking strip mines
6912 each loop in the loop nest such that the memory accesses of the
6913 element loops fit inside caches. The strip length can be changed
6914 using the @option{loop-block-tile-size} parameter. For example, given
6919 A(J, I) = B(I) + C(J)
6923 loop blocking will transform the loop as if the user had written:
6927 DO I = II, min (II + 50, N)
6928 DO J = JJ, min (JJ + 50, M)
6929 A(J, I) = B(I) + C(J)
6935 which can be beneficial when @code{M} is larger than the caches,
6936 because the innermost loop will iterate over a smaller amount of data
6937 that can be kept in the caches. This optimization applies to all the
6938 languages supported by GCC and is not limited to Fortran. To use this
6939 code transformation, GCC has to be configured with @option{--with-ppl}
6940 and @option{--with-cloog} to enable the Graphite loop transformation
6943 @item -fgraphite-identity
6944 @opindex fgraphite-identity
6945 Enable the identity transformation for graphite. For every SCoP we generate
6946 the polyhedral representation and transform it back to gimple. Using
6947 @option{-fgraphite-identity} we can check the costs or benefits of the
6948 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
6949 are also performed by the code generator CLooG, like index splitting and
6950 dead code elimination in loops.
6952 @item -floop-flatten
6953 @opindex floop-flatten
6954 Removes the loop nesting structure: transforms the loop nest into a
6955 single loop. This transformation can be useful to vectorize all the
6956 levels of the loop nest.
6958 @item -floop-parallelize-all
6959 @opindex floop-parallelize-all
6960 Use the Graphite data dependence analysis to identify loops that can
6961 be parallelized. Parallelize all the loops that can be analyzed to
6962 not contain loop carried dependences without checking that it is
6963 profitable to parallelize the loops.
6965 @item -fcheck-data-deps
6966 @opindex fcheck-data-deps
6967 Compare the results of several data dependence analyzers. This option
6968 is used for debugging the data dependence analyzers.
6970 @item -ftree-loop-if-convert
6971 Attempt to transform conditional jumps in the innermost loops to
6972 branch-less equivalents. The intent is to remove control-flow from
6973 the innermost loops in order to improve the ability of the
6974 vectorization pass to handle these loops. This is enabled by default
6975 if vectorization is enabled.
6977 @item -ftree-loop-if-convert-stores
6978 Attempt to also if-convert conditional jumps containing memory writes.
6979 This transformation can be unsafe for multi-threaded programs as it
6980 transforms conditional memory writes into unconditional memory writes.
6983 for (i = 0; i < N; i++)
6987 would be transformed to
6989 for (i = 0; i < N; i++)
6990 A[i] = cond ? expr : A[i];
6992 potentially producing data races.
6994 @item -ftree-loop-distribution
6995 Perform loop distribution. This flag can improve cache performance on
6996 big loop bodies and allow further loop optimizations, like
6997 parallelization or vectorization, to take place. For example, the loop
7014 @item -ftree-loop-distribute-patterns
7015 Perform loop distribution of patterns that can be code generated with
7016 calls to a library. This flag is enabled by default at @option{-O3}.
7018 This pass distributes the initialization loops and generates a call to
7019 memset zero. For example, the loop
7035 and the initialization loop is transformed into a call to memset zero.
7037 @item -ftree-loop-im
7038 @opindex ftree-loop-im
7039 Perform loop invariant motion on trees. This pass moves only invariants that
7040 would be hard to handle at RTL level (function calls, operations that expand to
7041 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
7042 operands of conditions that are invariant out of the loop, so that we can use
7043 just trivial invariantness analysis in loop unswitching. The pass also includes
7046 @item -ftree-loop-ivcanon
7047 @opindex ftree-loop-ivcanon
7048 Create a canonical counter for number of iterations in the loop for that
7049 determining number of iterations requires complicated analysis. Later
7050 optimizations then may determine the number easily. Useful especially
7051 in connection with unrolling.
7055 Perform induction variable optimizations (strength reduction, induction
7056 variable merging and induction variable elimination) on trees.
7058 @item -ftree-parallelize-loops=n
7059 @opindex ftree-parallelize-loops
7060 Parallelize loops, i.e., split their iteration space to run in n threads.
7061 This is only possible for loops whose iterations are independent
7062 and can be arbitrarily reordered. The optimization is only
7063 profitable on multiprocessor machines, for loops that are CPU-intensive,
7064 rather than constrained e.g.@: by memory bandwidth. This option
7065 implies @option{-pthread}, and thus is only supported on targets
7066 that have support for @option{-pthread}.
7070 Perform function-local points-to analysis on trees. This flag is
7071 enabled by default at @option{-O} and higher.
7075 Perform scalar replacement of aggregates. This pass replaces structure
7076 references with scalars to prevent committing structures to memory too
7077 early. This flag is enabled by default at @option{-O} and higher.
7079 @item -ftree-copyrename
7080 @opindex ftree-copyrename
7081 Perform copy renaming on trees. This pass attempts to rename compiler
7082 temporaries to other variables at copy locations, usually resulting in
7083 variable names which more closely resemble the original variables. This flag
7084 is enabled by default at @option{-O} and higher.
7088 Perform temporary expression replacement during the SSA->normal phase. Single
7089 use/single def temporaries are replaced at their use location with their
7090 defining expression. This results in non-GIMPLE code, but gives the expanders
7091 much more complex trees to work on resulting in better RTL generation. This is
7092 enabled by default at @option{-O} and higher.
7094 @item -ftree-vectorize
7095 @opindex ftree-vectorize
7096 Perform loop vectorization on trees. This flag is enabled by default at
7099 @item -ftree-slp-vectorize
7100 @opindex ftree-slp-vectorize
7101 Perform basic block vectorization on trees. This flag is enabled by default at
7102 @option{-O3} and when @option{-ftree-vectorize} is enabled.
7104 @item -ftree-vect-loop-version
7105 @opindex ftree-vect-loop-version
7106 Perform loop versioning when doing loop vectorization on trees. When a loop
7107 appears to be vectorizable except that data alignment or data dependence cannot
7108 be determined at compile time then vectorized and non-vectorized versions of
7109 the loop are generated along with runtime checks for alignment or dependence
7110 to control which version is executed. This option is enabled by default
7111 except at level @option{-Os} where it is disabled.
7113 @item -fvect-cost-model
7114 @opindex fvect-cost-model
7115 Enable cost model for vectorization.
7119 Perform Value Range Propagation on trees. This is similar to the
7120 constant propagation pass, but instead of values, ranges of values are
7121 propagated. This allows the optimizers to remove unnecessary range
7122 checks like array bound checks and null pointer checks. This is
7123 enabled by default at @option{-O2} and higher. Null pointer check
7124 elimination is only done if @option{-fdelete-null-pointer-checks} is
7129 Perform tail duplication to enlarge superblock size. This transformation
7130 simplifies the control flow of the function allowing other optimizations to do
7133 @item -funroll-loops
7134 @opindex funroll-loops
7135 Unroll loops whose number of iterations can be determined at compile
7136 time or upon entry to the loop. @option{-funroll-loops} implies
7137 @option{-frerun-cse-after-loop}. This option makes code larger,
7138 and may or may not make it run faster.
7140 @item -funroll-all-loops
7141 @opindex funroll-all-loops
7142 Unroll all loops, even if their number of iterations is uncertain when
7143 the loop is entered. This usually makes programs run more slowly.
7144 @option{-funroll-all-loops} implies the same options as
7145 @option{-funroll-loops},
7147 @item -fsplit-ivs-in-unroller
7148 @opindex fsplit-ivs-in-unroller
7149 Enables expressing of values of induction variables in later iterations
7150 of the unrolled loop using the value in the first iteration. This breaks
7151 long dependency chains, thus improving efficiency of the scheduling passes.
7153 Combination of @option{-fweb} and CSE is often sufficient to obtain the
7154 same effect. However in cases the loop body is more complicated than
7155 a single basic block, this is not reliable. It also does not work at all
7156 on some of the architectures due to restrictions in the CSE pass.
7158 This optimization is enabled by default.
7160 @item -fvariable-expansion-in-unroller
7161 @opindex fvariable-expansion-in-unroller
7162 With this option, the compiler will create multiple copies of some
7163 local variables when unrolling a loop which can result in superior code.
7165 @item -fpartial-inlining
7166 @opindex fpartial-inlining
7167 Inline parts of functions. This option has any effect only
7168 when inlining itself is turned on by the @option{-finline-functions}
7169 or @option{-finline-small-functions} options.
7171 Enabled at level @option{-O2}.
7173 @item -fpredictive-commoning
7174 @opindex fpredictive-commoning
7175 Perform predictive commoning optimization, i.e., reusing computations
7176 (especially memory loads and stores) performed in previous
7177 iterations of loops.
7179 This option is enabled at level @option{-O3}.
7181 @item -fprefetch-loop-arrays
7182 @opindex fprefetch-loop-arrays
7183 If supported by the target machine, generate instructions to prefetch
7184 memory to improve the performance of loops that access large arrays.
7186 This option may generate better or worse code; results are highly
7187 dependent on the structure of loops within the source code.
7189 Disabled at level @option{-Os}.
7192 @itemx -fno-peephole2
7193 @opindex fno-peephole
7194 @opindex fno-peephole2
7195 Disable any machine-specific peephole optimizations. The difference
7196 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
7197 are implemented in the compiler; some targets use one, some use the
7198 other, a few use both.
7200 @option{-fpeephole} is enabled by default.
7201 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7203 @item -fno-guess-branch-probability
7204 @opindex fno-guess-branch-probability
7205 Do not guess branch probabilities using heuristics.
7207 GCC will use heuristics to guess branch probabilities if they are
7208 not provided by profiling feedback (@option{-fprofile-arcs}). These
7209 heuristics are based on the control flow graph. If some branch probabilities
7210 are specified by @samp{__builtin_expect}, then the heuristics will be
7211 used to guess branch probabilities for the rest of the control flow graph,
7212 taking the @samp{__builtin_expect} info into account. The interactions
7213 between the heuristics and @samp{__builtin_expect} can be complex, and in
7214 some cases, it may be useful to disable the heuristics so that the effects
7215 of @samp{__builtin_expect} are easier to understand.
7217 The default is @option{-fguess-branch-probability} at levels
7218 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7220 @item -freorder-blocks
7221 @opindex freorder-blocks
7222 Reorder basic blocks in the compiled function in order to reduce number of
7223 taken branches and improve code locality.
7225 Enabled at levels @option{-O2}, @option{-O3}.
7227 @item -freorder-blocks-and-partition
7228 @opindex freorder-blocks-and-partition
7229 In addition to reordering basic blocks in the compiled function, in order
7230 to reduce number of taken branches, partitions hot and cold basic blocks
7231 into separate sections of the assembly and .o files, to improve
7232 paging and cache locality performance.
7234 This optimization is automatically turned off in the presence of
7235 exception handling, for linkonce sections, for functions with a user-defined
7236 section attribute and on any architecture that does not support named
7239 @item -freorder-functions
7240 @opindex freorder-functions
7241 Reorder functions in the object file in order to
7242 improve code locality. This is implemented by using special
7243 subsections @code{.text.hot} for most frequently executed functions and
7244 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
7245 the linker so object file format must support named sections and linker must
7246 place them in a reasonable way.
7248 Also profile feedback must be available in to make this option effective. See
7249 @option{-fprofile-arcs} for details.
7251 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7253 @item -fstrict-aliasing
7254 @opindex fstrict-aliasing
7255 Allow the compiler to assume the strictest aliasing rules applicable to
7256 the language being compiled. For C (and C++), this activates
7257 optimizations based on the type of expressions. In particular, an
7258 object of one type is assumed never to reside at the same address as an
7259 object of a different type, unless the types are almost the same. For
7260 example, an @code{unsigned int} can alias an @code{int}, but not a
7261 @code{void*} or a @code{double}. A character type may alias any other
7264 @anchor{Type-punning}Pay special attention to code like this:
7277 The practice of reading from a different union member than the one most
7278 recently written to (called ``type-punning'') is common. Even with
7279 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
7280 is accessed through the union type. So, the code above will work as
7281 expected. @xref{Structures unions enumerations and bit-fields
7282 implementation}. However, this code might not:
7293 Similarly, access by taking the address, casting the resulting pointer
7294 and dereferencing the result has undefined behavior, even if the cast
7295 uses a union type, e.g.:
7299 return ((union a_union *) &d)->i;
7303 The @option{-fstrict-aliasing} option is enabled at levels
7304 @option{-O2}, @option{-O3}, @option{-Os}.
7306 @item -fstrict-overflow
7307 @opindex fstrict-overflow
7308 Allow the compiler to assume strict signed overflow rules, depending
7309 on the language being compiled. For C (and C++) this means that
7310 overflow when doing arithmetic with signed numbers is undefined, which
7311 means that the compiler may assume that it will not happen. This
7312 permits various optimizations. For example, the compiler will assume
7313 that an expression like @code{i + 10 > i} will always be true for
7314 signed @code{i}. This assumption is only valid if signed overflow is
7315 undefined, as the expression is false if @code{i + 10} overflows when
7316 using twos complement arithmetic. When this option is in effect any
7317 attempt to determine whether an operation on signed numbers will
7318 overflow must be written carefully to not actually involve overflow.
7320 This option also allows the compiler to assume strict pointer
7321 semantics: given a pointer to an object, if adding an offset to that
7322 pointer does not produce a pointer to the same object, the addition is
7323 undefined. This permits the compiler to conclude that @code{p + u >
7324 p} is always true for a pointer @code{p} and unsigned integer
7325 @code{u}. This assumption is only valid because pointer wraparound is
7326 undefined, as the expression is false if @code{p + u} overflows using
7327 twos complement arithmetic.
7329 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
7330 that integer signed overflow is fully defined: it wraps. When
7331 @option{-fwrapv} is used, there is no difference between
7332 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
7333 integers. With @option{-fwrapv} certain types of overflow are
7334 permitted. For example, if the compiler gets an overflow when doing
7335 arithmetic on constants, the overflowed value can still be used with
7336 @option{-fwrapv}, but not otherwise.
7338 The @option{-fstrict-overflow} option is enabled at levels
7339 @option{-O2}, @option{-O3}, @option{-Os}.
7341 @item -falign-functions
7342 @itemx -falign-functions=@var{n}
7343 @opindex falign-functions
7344 Align the start of functions to the next power-of-two greater than
7345 @var{n}, skipping up to @var{n} bytes. For instance,
7346 @option{-falign-functions=32} aligns functions to the next 32-byte
7347 boundary, but @option{-falign-functions=24} would align to the next
7348 32-byte boundary only if this can be done by skipping 23 bytes or less.
7350 @option{-fno-align-functions} and @option{-falign-functions=1} are
7351 equivalent and mean that functions will not be aligned.
7353 Some assemblers only support this flag when @var{n} is a power of two;
7354 in that case, it is rounded up.
7356 If @var{n} is not specified or is zero, use a machine-dependent default.
7358 Enabled at levels @option{-O2}, @option{-O3}.
7360 @item -falign-labels
7361 @itemx -falign-labels=@var{n}
7362 @opindex falign-labels
7363 Align all branch targets to a power-of-two boundary, skipping up to
7364 @var{n} bytes like @option{-falign-functions}. This option can easily
7365 make code slower, because it must insert dummy operations for when the
7366 branch target is reached in the usual flow of the code.
7368 @option{-fno-align-labels} and @option{-falign-labels=1} are
7369 equivalent and mean that labels will not be aligned.
7371 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
7372 are greater than this value, then their values are used instead.
7374 If @var{n} is not specified or is zero, use a machine-dependent default
7375 which is very likely to be @samp{1}, meaning no alignment.
7377 Enabled at levels @option{-O2}, @option{-O3}.
7380 @itemx -falign-loops=@var{n}
7381 @opindex falign-loops
7382 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7383 like @option{-falign-functions}. The hope is that the loop will be
7384 executed many times, which will make up for any execution of the dummy
7387 @option{-fno-align-loops} and @option{-falign-loops=1} are
7388 equivalent and mean that loops will not be aligned.
7390 If @var{n} is not specified or is zero, use a machine-dependent default.
7392 Enabled at levels @option{-O2}, @option{-O3}.
7395 @itemx -falign-jumps=@var{n}
7396 @opindex falign-jumps
7397 Align branch targets to a power-of-two boundary, for branch targets
7398 where the targets can only be reached by jumping, skipping up to @var{n}
7399 bytes like @option{-falign-functions}. In this case, no dummy operations
7402 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7403 equivalent and mean that loops will not be aligned.
7405 If @var{n} is not specified or is zero, use a machine-dependent default.
7407 Enabled at levels @option{-O2}, @option{-O3}.
7409 @item -funit-at-a-time
7410 @opindex funit-at-a-time
7411 This option is left for compatibility reasons. @option{-funit-at-a-time}
7412 has no effect, while @option{-fno-unit-at-a-time} implies
7413 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7417 @item -fno-toplevel-reorder
7418 @opindex fno-toplevel-reorder
7419 Do not reorder top-level functions, variables, and @code{asm}
7420 statements. Output them in the same order that they appear in the
7421 input file. When this option is used, unreferenced static variables
7422 will not be removed. This option is intended to support existing code
7423 which relies on a particular ordering. For new code, it is better to
7426 Enabled at level @option{-O0}. When disabled explicitly, it also imply
7427 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
7432 Constructs webs as commonly used for register allocation purposes and assign
7433 each web individual pseudo register. This allows the register allocation pass
7434 to operate on pseudos directly, but also strengthens several other optimization
7435 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7436 however, make debugging impossible, since variables will no longer stay in a
7439 Enabled by default with @option{-funroll-loops}.
7441 @item -fwhole-program
7442 @opindex fwhole-program
7443 Assume that the current compilation unit represents the whole program being
7444 compiled. All public functions and variables with the exception of @code{main}
7445 and those merged by attribute @code{externally_visible} become static functions
7446 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.
7447 While this option is equivalent to proper use of the @code{static} keyword for
7448 programs consisting of a single file, in combination with option
7449 @option{-flto} this flag can be used to
7450 compile many smaller scale programs since the functions and variables become
7451 local for the whole combined compilation unit, not for the single source file
7454 This option implies @option{-fwhole-file} for Fortran programs.
7456 @item -flto[=@var{n}]
7458 This option runs the standard link-time optimizer. When invoked
7459 with source code, it generates GIMPLE (one of GCC's internal
7460 representations) and writes it to special ELF sections in the object
7461 file. When the object files are linked together, all the function
7462 bodies are read from these ELF sections and instantiated as if they
7463 had been part of the same translation unit.
7465 To use the link-timer optimizer, @option{-flto} needs to be specified at
7466 compile time and during the final link. For example,
7469 gcc -c -O2 -flto foo.c
7470 gcc -c -O2 -flto bar.c
7471 gcc -o myprog -flto -O2 foo.o bar.o
7474 The first two invocations to GCC will save a bytecode representation
7475 of GIMPLE into special ELF sections inside @file{foo.o} and
7476 @file{bar.o}. The final invocation will read the GIMPLE bytecode from
7477 @file{foo.o} and @file{bar.o}, merge the two files into a single
7478 internal image, and compile the result as usual. Since both
7479 @file{foo.o} and @file{bar.o} are merged into a single image, this
7480 causes all the inter-procedural analyses and optimizations in GCC to
7481 work across the two files as if they were a single one. This means,
7482 for example, that the inliner will be able to inline functions in
7483 @file{bar.o} into functions in @file{foo.o} and vice-versa.
7485 Another (simpler) way to enable link-time optimization is,
7488 gcc -o myprog -flto -O2 foo.c bar.c
7491 The above will generate bytecode for @file{foo.c} and @file{bar.c},
7492 merge them together into a single GIMPLE representation and optimize
7493 them as usual to produce @file{myprog}.
7495 The only important thing to keep in mind is that to enable link-time
7496 optimizations the @option{-flto} flag needs to be passed to both the
7497 compile and the link commands.
7499 To make whole program optimization effective, it is necesary to make
7500 certain whole program assumptions. The compiler needs to know
7501 what functions and variables can be accessed by libraries and runtime
7502 outside of the link time optimized unit. When supported by the linker,
7503 the linker plugin (see @option{-fuse-linker-plugin}) passes to the
7504 compiler information about used and externally visible symbols. When
7505 the linker plugin is not available, @option{-fwhole-program} should be
7506 used to allow the compiler to make these assumptions, which will lead
7507 to more aggressive optimization decisions.
7509 Note that when a file is compiled with @option{-flto}, the generated
7510 object file will be larger than a regular object file because it will
7511 contain GIMPLE bytecodes and the usual final code. This means that
7512 object files with LTO information can be linked as a normal object
7513 file. So, in the previous example, if the final link is done with
7516 gcc -o myprog foo.o bar.o
7519 The only difference will be that no inter-procedural optimizations
7520 will be applied to produce @file{myprog}. The two object files
7521 @file{foo.o} and @file{bar.o} will be simply sent to the regular
7524 Additionally, the optimization flags used to compile individual files
7525 are not necessarily related to those used at link-time. For instance,
7528 gcc -c -O0 -flto foo.c
7529 gcc -c -O0 -flto bar.c
7530 gcc -o myprog -flto -O3 foo.o bar.o
7533 This will produce individual object files with unoptimized assembler
7534 code, but the resulting binary @file{myprog} will be optimized at
7535 @option{-O3}. Now, if the final binary is generated without
7536 @option{-flto}, then @file{myprog} will not be optimized.
7538 When producing the final binary with @option{-flto}, GCC will only
7539 apply link-time optimizations to those files that contain bytecode.
7540 Therefore, you can mix and match object files and libraries with
7541 GIMPLE bytecodes and final object code. GCC will automatically select
7542 which files to optimize in LTO mode and which files to link without
7545 There are some code generation flags that GCC will preserve when
7546 generating bytecodes, as they need to be used during the final link
7547 stage. Currently, the following options are saved into the GIMPLE
7548 bytecode files: @option{-fPIC}, @option{-fcommon} and all the
7549 @option{-m} target flags.
7551 At link time, these options are read-in and reapplied. Note that the
7552 current implementation makes no attempt at recognizing conflicting
7553 values for these options. If two or more files have a conflicting
7554 value (e.g., one file is compiled with @option{-fPIC} and another
7555 isn't), the compiler will simply use the last value read from the
7556 bytecode files. It is recommended, then, that all the files
7557 participating in the same link be compiled with the same options.
7559 Another feature of LTO is that it is possible to apply interprocedural
7560 optimizations on files written in different languages. This requires
7561 some support in the language front end. Currently, the C, C++ and
7562 Fortran front ends are capable of emitting GIMPLE bytecodes, so
7563 something like this should work
7568 gfortran -c -flto baz.f90
7569 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
7572 Notice that the final link is done with @command{g++} to get the C++
7573 runtime libraries and @option{-lgfortran} is added to get the Fortran
7574 runtime libraries. In general, when mixing languages in LTO mode, you
7575 should use the same link command used when mixing languages in a
7576 regular (non-LTO) compilation. This means that if your build process
7577 was mixing languages before, all you need to add is @option{-flto} to
7578 all the compile and link commands.
7580 If LTO encounters objects with C linkage declared with incompatible
7581 types in separate translation units to be linked together (undefined
7582 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
7583 issued. The behavior is still undefined at runtime.
7585 If object files containing GIMPLE bytecode are stored in a library
7586 archive, say @file{libfoo.a}, it is possible to extract and use them
7587 in an LTO link if you are using @command{gold} as the linker (which,
7588 in turn requires GCC to be configured with @option{--enable-gold}).
7589 To enable this feature, use the flag @option{-fuse-linker-plugin} at
7593 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
7596 With the linker plugin enabled, @command{gold} will extract the needed
7597 GIMPLE files from @file{libfoo.a} and pass them on to the running GCC
7598 to make them part of the aggregated GIMPLE image to be optimized.
7600 If you are not using @command{gold} and/or do not specify
7601 @option{-fuse-linker-plugin} then the objects inside @file{libfoo.a}
7602 will be extracted and linked as usual, but they will not participate
7603 in the LTO optimization process.
7605 Link time optimizations do not require the presence of the whole program to
7606 operate. If the program does not require any symbols to be exported, it is
7607 possible to combine @option{-flto} and with @option{-fwhole-program} to allow
7608 the interprocedural optimizers to use more aggressive assumptions which may
7609 lead to improved optimization opportunities.
7610 Use of @option{-fwhole-program} is not needed when linker plugin is
7611 active (see @option{-fuse-linker-plugin}).
7613 Regarding portability: the current implementation of LTO makes no
7614 attempt at generating bytecode that can be ported between different
7615 types of hosts. The bytecode files are versioned and there is a
7616 strict version check, so bytecode files generated in one version of
7617 GCC will not work with an older/newer version of GCC.
7619 Link time optimization does not play well with generating debugging
7620 information. Combining @option{-flto} with
7621 @option{-g} is currently experimental and expected to produce wrong
7624 If you specify the optional @var{n}, the optimization and code
7625 generation done at link time is executed in parallel using @var{n}
7626 parallel jobs by utilizing an installed @command{make} program. The
7627 environment variable @env{MAKE} may be used to override the program
7628 used. The default value for @var{n} is 1.
7630 You can also specify @option{-flto=jobserver} to use GNU make's
7631 job server mode to determine the number of parallel jobs. This
7632 is useful when the Makefile calling GCC is already executing in parallel.
7633 The parent Makefile will need a @samp{+} prepended to the command recipe
7634 for this to work. This will likely only work if @env{MAKE} is
7637 This option is disabled by default.
7639 @item -flto-partition=@var{alg}
7640 @opindex flto-partition
7641 Specify the partitioning algorithm used by the link time optimizer.
7642 The value is either @code{1to1} to specify a partitioning mirroring
7643 the original source files or @code{balanced} to specify partitioning
7644 into equally sized chunks (whenever possible). Specifying @code{none}
7645 as an algorithm disables partitioning and streaming completely. The
7646 default value is @code{balanced}.
7648 @item -flto-compression-level=@var{n}
7649 This option specifies the level of compression used for intermediate
7650 language written to LTO object files, and is only meaningful in
7651 conjunction with LTO mode (@option{-flto}). Valid
7652 values are 0 (no compression) to 9 (maximum compression). Values
7653 outside this range are clamped to either 0 or 9. If the option is not
7654 given, a default balanced compression setting is used.
7657 Prints a report with internal details on the workings of the link-time
7658 optimizer. The contents of this report vary from version to version,
7659 it is meant to be useful to GCC developers when processing object
7660 files in LTO mode (via @option{-flto}).
7662 Disabled by default.
7664 @item -fuse-linker-plugin
7665 Enables the extraction of objects with GIMPLE bytecode information
7666 from library archives. This option relies on features available only
7667 in @command{gold}, so to use this you must configure GCC with
7668 @option{--enable-gold}. See @option{-flto} for a description on the
7669 effect of this flag and how to use it.
7671 Disabled by default.
7673 @item -fcprop-registers
7674 @opindex fcprop-registers
7675 After register allocation and post-register allocation instruction splitting,
7676 we perform a copy-propagation pass to try to reduce scheduling dependencies
7677 and occasionally eliminate the copy.
7679 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7681 @item -fprofile-correction
7682 @opindex fprofile-correction
7683 Profiles collected using an instrumented binary for multi-threaded programs may
7684 be inconsistent due to missed counter updates. When this option is specified,
7685 GCC will use heuristics to correct or smooth out such inconsistencies. By
7686 default, GCC will emit an error message when an inconsistent profile is detected.
7688 @item -fprofile-dir=@var{path}
7689 @opindex fprofile-dir
7691 Set the directory to search the profile data files in to @var{path}.
7692 This option affects only the profile data generated by
7693 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
7694 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
7695 and its related options.
7696 By default, GCC will use the current directory as @var{path}
7697 thus the profile data file will appear in the same directory as the object file.
7699 @item -fprofile-generate
7700 @itemx -fprofile-generate=@var{path}
7701 @opindex fprofile-generate
7703 Enable options usually used for instrumenting application to produce
7704 profile useful for later recompilation with profile feedback based
7705 optimization. You must use @option{-fprofile-generate} both when
7706 compiling and when linking your program.
7708 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
7710 If @var{path} is specified, GCC will look at the @var{path} to find
7711 the profile feedback data files. See @option{-fprofile-dir}.
7714 @itemx -fprofile-use=@var{path}
7715 @opindex fprofile-use
7716 Enable profile feedback directed optimizations, and optimizations
7717 generally profitable only with profile feedback available.
7719 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
7720 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
7722 By default, GCC emits an error message if the feedback profiles do not
7723 match the source code. This error can be turned into a warning by using
7724 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
7727 If @var{path} is specified, GCC will look at the @var{path} to find
7728 the profile feedback data files. See @option{-fprofile-dir}.
7731 The following options control compiler behavior regarding floating
7732 point arithmetic. These options trade off between speed and
7733 correctness. All must be specifically enabled.
7737 @opindex ffloat-store
7738 Do not store floating point variables in registers, and inhibit other
7739 options that might change whether a floating point value is taken from a
7742 @cindex floating point precision
7743 This option prevents undesirable excess precision on machines such as
7744 the 68000 where the floating registers (of the 68881) keep more
7745 precision than a @code{double} is supposed to have. Similarly for the
7746 x86 architecture. For most programs, the excess precision does only
7747 good, but a few programs rely on the precise definition of IEEE floating
7748 point. Use @option{-ffloat-store} for such programs, after modifying
7749 them to store all pertinent intermediate computations into variables.
7751 @item -fexcess-precision=@var{style}
7752 @opindex fexcess-precision
7753 This option allows further control over excess precision on machines
7754 where floating-point registers have more precision than the IEEE
7755 @code{float} and @code{double} types and the processor does not
7756 support operations rounding to those types. By default,
7757 @option{-fexcess-precision=fast} is in effect; this means that
7758 operations are carried out in the precision of the registers and that
7759 it is unpredictable when rounding to the types specified in the source
7760 code takes place. When compiling C, if
7761 @option{-fexcess-precision=standard} is specified then excess
7762 precision will follow the rules specified in ISO C99; in particular,
7763 both casts and assignments cause values to be rounded to their
7764 semantic types (whereas @option{-ffloat-store} only affects
7765 assignments). This option is enabled by default for C if a strict
7766 conformance option such as @option{-std=c99} is used.
7769 @option{-fexcess-precision=standard} is not implemented for languages
7770 other than C, and has no effect if
7771 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
7772 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
7773 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
7774 semantics apply without excess precision, and in the latter, rounding
7779 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
7780 @option{-ffinite-math-only}, @option{-fno-rounding-math},
7781 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
7783 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
7785 This option is not turned on by any @option{-O} option since
7786 it can result in incorrect output for programs which depend on
7787 an exact implementation of IEEE or ISO rules/specifications for
7788 math functions. It may, however, yield faster code for programs
7789 that do not require the guarantees of these specifications.
7791 @item -fno-math-errno
7792 @opindex fno-math-errno
7793 Do not set ERRNO after calling math functions that are executed
7794 with a single instruction, e.g., sqrt. A program that relies on
7795 IEEE exceptions for math error handling may want to use this flag
7796 for speed while maintaining IEEE arithmetic compatibility.
7798 This option is not turned on by any @option{-O} option since
7799 it can result in incorrect output for programs which depend on
7800 an exact implementation of IEEE or ISO rules/specifications for
7801 math functions. It may, however, yield faster code for programs
7802 that do not require the guarantees of these specifications.
7804 The default is @option{-fmath-errno}.
7806 On Darwin systems, the math library never sets @code{errno}. There is
7807 therefore no reason for the compiler to consider the possibility that
7808 it might, and @option{-fno-math-errno} is the default.
7810 @item -funsafe-math-optimizations
7811 @opindex funsafe-math-optimizations
7813 Allow optimizations for floating-point arithmetic that (a) assume
7814 that arguments and results are valid and (b) may violate IEEE or
7815 ANSI standards. When used at link-time, it may include libraries
7816 or startup files that change the default FPU control word or other
7817 similar optimizations.
7819 This option is not turned on by any @option{-O} option since
7820 it can result in incorrect output for programs which depend on
7821 an exact implementation of IEEE or ISO rules/specifications for
7822 math functions. It may, however, yield faster code for programs
7823 that do not require the guarantees of these specifications.
7824 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
7825 @option{-fassociative-math} and @option{-freciprocal-math}.
7827 The default is @option{-fno-unsafe-math-optimizations}.
7829 @item -fassociative-math
7830 @opindex fassociative-math
7832 Allow re-association of operands in series of floating-point operations.
7833 This violates the ISO C and C++ language standard by possibly changing
7834 computation result. NOTE: re-ordering may change the sign of zero as
7835 well as ignore NaNs and inhibit or create underflow or overflow (and
7836 thus cannot be used on a code which relies on rounding behavior like
7837 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
7838 and thus may not be used when ordered comparisons are required.
7839 This option requires that both @option{-fno-signed-zeros} and
7840 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
7841 much sense with @option{-frounding-math}. For Fortran the option
7842 is automatically enabled when both @option{-fno-signed-zeros} and
7843 @option{-fno-trapping-math} are in effect.
7845 The default is @option{-fno-associative-math}.
7847 @item -freciprocal-math
7848 @opindex freciprocal-math
7850 Allow the reciprocal of a value to be used instead of dividing by
7851 the value if this enables optimizations. For example @code{x / y}
7852 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
7853 is subject to common subexpression elimination. Note that this loses
7854 precision and increases the number of flops operating on the value.
7856 The default is @option{-fno-reciprocal-math}.
7858 @item -ffinite-math-only
7859 @opindex ffinite-math-only
7860 Allow optimizations for floating-point arithmetic that assume
7861 that arguments and results are not NaNs or +-Infs.
7863 This option is not turned on by any @option{-O} option since
7864 it can result in incorrect output for programs which depend on
7865 an exact implementation of IEEE or ISO rules/specifications for
7866 math functions. It may, however, yield faster code for programs
7867 that do not require the guarantees of these specifications.
7869 The default is @option{-fno-finite-math-only}.
7871 @item -fno-signed-zeros
7872 @opindex fno-signed-zeros
7873 Allow optimizations for floating point arithmetic that ignore the
7874 signedness of zero. IEEE arithmetic specifies the behavior of
7875 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
7876 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
7877 This option implies that the sign of a zero result isn't significant.
7879 The default is @option{-fsigned-zeros}.
7881 @item -fno-trapping-math
7882 @opindex fno-trapping-math
7883 Compile code assuming that floating-point operations cannot generate
7884 user-visible traps. These traps include division by zero, overflow,
7885 underflow, inexact result and invalid operation. This option requires
7886 that @option{-fno-signaling-nans} be in effect. Setting this option may
7887 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
7889 This option should never be turned on by any @option{-O} option since
7890 it can result in incorrect output for programs which depend on
7891 an exact implementation of IEEE or ISO rules/specifications for
7894 The default is @option{-ftrapping-math}.
7896 @item -frounding-math
7897 @opindex frounding-math
7898 Disable transformations and optimizations that assume default floating
7899 point rounding behavior. This is round-to-zero for all floating point
7900 to integer conversions, and round-to-nearest for all other arithmetic
7901 truncations. This option should be specified for programs that change
7902 the FP rounding mode dynamically, or that may be executed with a
7903 non-default rounding mode. This option disables constant folding of
7904 floating point expressions at compile-time (which may be affected by
7905 rounding mode) and arithmetic transformations that are unsafe in the
7906 presence of sign-dependent rounding modes.
7908 The default is @option{-fno-rounding-math}.
7910 This option is experimental and does not currently guarantee to
7911 disable all GCC optimizations that are affected by rounding mode.
7912 Future versions of GCC may provide finer control of this setting
7913 using C99's @code{FENV_ACCESS} pragma. This command line option
7914 will be used to specify the default state for @code{FENV_ACCESS}.
7916 @item -fsignaling-nans
7917 @opindex fsignaling-nans
7918 Compile code assuming that IEEE signaling NaNs may generate user-visible
7919 traps during floating-point operations. Setting this option disables
7920 optimizations that may change the number of exceptions visible with
7921 signaling NaNs. This option implies @option{-ftrapping-math}.
7923 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
7926 The default is @option{-fno-signaling-nans}.
7928 This option is experimental and does not currently guarantee to
7929 disable all GCC optimizations that affect signaling NaN behavior.
7931 @item -fsingle-precision-constant
7932 @opindex fsingle-precision-constant
7933 Treat floating point constant as single precision constant instead of
7934 implicitly converting it to double precision constant.
7936 @item -fcx-limited-range
7937 @opindex fcx-limited-range
7938 When enabled, this option states that a range reduction step is not
7939 needed when performing complex division. Also, there is no checking
7940 whether the result of a complex multiplication or division is @code{NaN
7941 + I*NaN}, with an attempt to rescue the situation in that case. The
7942 default is @option{-fno-cx-limited-range}, but is enabled by
7943 @option{-ffast-math}.
7945 This option controls the default setting of the ISO C99
7946 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
7949 @item -fcx-fortran-rules
7950 @opindex fcx-fortran-rules
7951 Complex multiplication and division follow Fortran rules. Range
7952 reduction is done as part of complex division, but there is no checking
7953 whether the result of a complex multiplication or division is @code{NaN
7954 + I*NaN}, with an attempt to rescue the situation in that case.
7956 The default is @option{-fno-cx-fortran-rules}.
7960 The following options control optimizations that may improve
7961 performance, but are not enabled by any @option{-O} options. This
7962 section includes experimental options that may produce broken code.
7965 @item -fbranch-probabilities
7966 @opindex fbranch-probabilities
7967 After running a program compiled with @option{-fprofile-arcs}
7968 (@pxref{Debugging Options,, Options for Debugging Your Program or
7969 @command{gcc}}), you can compile it a second time using
7970 @option{-fbranch-probabilities}, to improve optimizations based on
7971 the number of times each branch was taken. When the program
7972 compiled with @option{-fprofile-arcs} exits it saves arc execution
7973 counts to a file called @file{@var{sourcename}.gcda} for each source
7974 file. The information in this data file is very dependent on the
7975 structure of the generated code, so you must use the same source code
7976 and the same optimization options for both compilations.
7978 With @option{-fbranch-probabilities}, GCC puts a
7979 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
7980 These can be used to improve optimization. Currently, they are only
7981 used in one place: in @file{reorg.c}, instead of guessing which path a
7982 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
7983 exactly determine which path is taken more often.
7985 @item -fprofile-values
7986 @opindex fprofile-values
7987 If combined with @option{-fprofile-arcs}, it adds code so that some
7988 data about values of expressions in the program is gathered.
7990 With @option{-fbranch-probabilities}, it reads back the data gathered
7991 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
7992 notes to instructions for their later usage in optimizations.
7994 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
7998 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
7999 a code to gather information about values of expressions.
8001 With @option{-fbranch-probabilities}, it reads back the data gathered
8002 and actually performs the optimizations based on them.
8003 Currently the optimizations include specialization of division operation
8004 using the knowledge about the value of the denominator.
8006 @item -frename-registers
8007 @opindex frename-registers
8008 Attempt to avoid false dependencies in scheduled code by making use
8009 of registers left over after register allocation. This optimization
8010 will most benefit processors with lots of registers. Depending on the
8011 debug information format adopted by the target, however, it can
8012 make debugging impossible, since variables will no longer stay in
8013 a ``home register''.
8015 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
8019 Perform tail duplication to enlarge superblock size. This transformation
8020 simplifies the control flow of the function allowing other optimizations to do
8023 Enabled with @option{-fprofile-use}.
8025 @item -funroll-loops
8026 @opindex funroll-loops
8027 Unroll loops whose number of iterations can be determined at compile time or
8028 upon entry to the loop. @option{-funroll-loops} implies
8029 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
8030 It also turns on complete loop peeling (i.e.@: complete removal of loops with
8031 small constant number of iterations). This option makes code larger, and may
8032 or may not make it run faster.
8034 Enabled with @option{-fprofile-use}.
8036 @item -funroll-all-loops
8037 @opindex funroll-all-loops
8038 Unroll all loops, even if their number of iterations is uncertain when
8039 the loop is entered. This usually makes programs run more slowly.
8040 @option{-funroll-all-loops} implies the same options as
8041 @option{-funroll-loops}.
8044 @opindex fpeel-loops
8045 Peels the loops for that there is enough information that they do not
8046 roll much (from profile feedback). It also turns on complete loop peeling
8047 (i.e.@: complete removal of loops with small constant number of iterations).
8049 Enabled with @option{-fprofile-use}.
8051 @item -fmove-loop-invariants
8052 @opindex fmove-loop-invariants
8053 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
8054 at level @option{-O1}
8056 @item -funswitch-loops
8057 @opindex funswitch-loops
8058 Move branches with loop invariant conditions out of the loop, with duplicates
8059 of the loop on both branches (modified according to result of the condition).
8061 @item -ffunction-sections
8062 @itemx -fdata-sections
8063 @opindex ffunction-sections
8064 @opindex fdata-sections
8065 Place each function or data item into its own section in the output
8066 file if the target supports arbitrary sections. The name of the
8067 function or the name of the data item determines the section's name
8070 Use these options on systems where the linker can perform optimizations
8071 to improve locality of reference in the instruction space. Most systems
8072 using the ELF object format and SPARC processors running Solaris 2 have
8073 linkers with such optimizations. AIX may have these optimizations in
8076 Only use these options when there are significant benefits from doing
8077 so. When you specify these options, the assembler and linker will
8078 create larger object and executable files and will also be slower.
8079 You will not be able to use @code{gprof} on all systems if you
8080 specify this option and you may have problems with debugging if
8081 you specify both this option and @option{-g}.
8083 @item -fbranch-target-load-optimize
8084 @opindex fbranch-target-load-optimize
8085 Perform branch target register load optimization before prologue / epilogue
8087 The use of target registers can typically be exposed only during reload,
8088 thus hoisting loads out of loops and doing inter-block scheduling needs
8089 a separate optimization pass.
8091 @item -fbranch-target-load-optimize2
8092 @opindex fbranch-target-load-optimize2
8093 Perform branch target register load optimization after prologue / epilogue
8096 @item -fbtr-bb-exclusive
8097 @opindex fbtr-bb-exclusive
8098 When performing branch target register load optimization, don't reuse
8099 branch target registers in within any basic block.
8101 @item -fstack-protector
8102 @opindex fstack-protector
8103 Emit extra code to check for buffer overflows, such as stack smashing
8104 attacks. This is done by adding a guard variable to functions with
8105 vulnerable objects. This includes functions that call alloca, and
8106 functions with buffers larger than 8 bytes. The guards are initialized
8107 when a function is entered and then checked when the function exits.
8108 If a guard check fails, an error message is printed and the program exits.
8110 @item -fstack-protector-all
8111 @opindex fstack-protector-all
8112 Like @option{-fstack-protector} except that all functions are protected.
8114 @item -fsection-anchors
8115 @opindex fsection-anchors
8116 Try to reduce the number of symbolic address calculations by using
8117 shared ``anchor'' symbols to address nearby objects. This transformation
8118 can help to reduce the number of GOT entries and GOT accesses on some
8121 For example, the implementation of the following function @code{foo}:
8125 int foo (void) @{ return a + b + c; @}
8128 would usually calculate the addresses of all three variables, but if you
8129 compile it with @option{-fsection-anchors}, it will access the variables
8130 from a common anchor point instead. The effect is similar to the
8131 following pseudocode (which isn't valid C):
8136 register int *xr = &x;
8137 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
8141 Not all targets support this option.
8143 @item --param @var{name}=@var{value}
8145 In some places, GCC uses various constants to control the amount of
8146 optimization that is done. For example, GCC will not inline functions
8147 that contain more that a certain number of instructions. You can
8148 control some of these constants on the command-line using the
8149 @option{--param} option.
8151 The names of specific parameters, and the meaning of the values, are
8152 tied to the internals of the compiler, and are subject to change
8153 without notice in future releases.
8155 In each case, the @var{value} is an integer. The allowable choices for
8156 @var{name} are given in the following table:
8159 @item struct-reorg-cold-struct-ratio
8160 The threshold ratio (as a percentage) between a structure frequency
8161 and the frequency of the hottest structure in the program. This parameter
8162 is used by struct-reorg optimization enabled by @option{-fipa-struct-reorg}.
8163 We say that if the ratio of a structure frequency, calculated by profiling,
8164 to the hottest structure frequency in the program is less than this
8165 parameter, then structure reorganization is not applied to this structure.
8168 @item predictable-branch-outcome
8169 When branch is predicted to be taken with probability lower than this threshold
8170 (in percent), then it is considered well predictable. The default is 10.
8172 @item max-crossjump-edges
8173 The maximum number of incoming edges to consider for crossjumping.
8174 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
8175 the number of edges incoming to each block. Increasing values mean
8176 more aggressive optimization, making the compile time increase with
8177 probably small improvement in executable size.
8179 @item min-crossjump-insns
8180 The minimum number of instructions which must be matched at the end
8181 of two blocks before crossjumping will be performed on them. This
8182 value is ignored in the case where all instructions in the block being
8183 crossjumped from are matched. The default value is 5.
8185 @item max-grow-copy-bb-insns
8186 The maximum code size expansion factor when copying basic blocks
8187 instead of jumping. The expansion is relative to a jump instruction.
8188 The default value is 8.
8190 @item max-goto-duplication-insns
8191 The maximum number of instructions to duplicate to a block that jumps
8192 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
8193 passes, GCC factors computed gotos early in the compilation process,
8194 and unfactors them as late as possible. Only computed jumps at the
8195 end of a basic blocks with no more than max-goto-duplication-insns are
8196 unfactored. The default value is 8.
8198 @item max-delay-slot-insn-search
8199 The maximum number of instructions to consider when looking for an
8200 instruction to fill a delay slot. If more than this arbitrary number of
8201 instructions is searched, the time savings from filling the delay slot
8202 will be minimal so stop searching. Increasing values mean more
8203 aggressive optimization, making the compile time increase with probably
8204 small improvement in executable run time.
8206 @item max-delay-slot-live-search
8207 When trying to fill delay slots, the maximum number of instructions to
8208 consider when searching for a block with valid live register
8209 information. Increasing this arbitrarily chosen value means more
8210 aggressive optimization, increasing the compile time. This parameter
8211 should be removed when the delay slot code is rewritten to maintain the
8214 @item max-gcse-memory
8215 The approximate maximum amount of memory that will be allocated in
8216 order to perform the global common subexpression elimination
8217 optimization. If more memory than specified is required, the
8218 optimization will not be done.
8220 @item max-pending-list-length
8221 The maximum number of pending dependencies scheduling will allow
8222 before flushing the current state and starting over. Large functions
8223 with few branches or calls can create excessively large lists which
8224 needlessly consume memory and resources.
8226 @item max-inline-insns-single
8227 Several parameters control the tree inliner used in gcc.
8228 This number sets the maximum number of instructions (counted in GCC's
8229 internal representation) in a single function that the tree inliner
8230 will consider for inlining. This only affects functions declared
8231 inline and methods implemented in a class declaration (C++).
8232 The default value is 300.
8234 @item max-inline-insns-auto
8235 When you use @option{-finline-functions} (included in @option{-O3}),
8236 a lot of functions that would otherwise not be considered for inlining
8237 by the compiler will be investigated. To those functions, a different
8238 (more restrictive) limit compared to functions declared inline can
8240 The default value is 40.
8242 @item large-function-insns
8243 The limit specifying really large functions. For functions larger than this
8244 limit after inlining, inlining is constrained by
8245 @option{--param large-function-growth}. This parameter is useful primarily
8246 to avoid extreme compilation time caused by non-linear algorithms used by the
8248 The default value is 2700.
8250 @item large-function-growth
8251 Specifies maximal growth of large function caused by inlining in percents.
8252 The default value is 100 which limits large function growth to 2.0 times
8255 @item large-unit-insns
8256 The limit specifying large translation unit. Growth caused by inlining of
8257 units larger than this limit is limited by @option{--param inline-unit-growth}.
8258 For small units this might be too tight (consider unit consisting of function A
8259 that is inline and B that just calls A three time. If B is small relative to
8260 A, the growth of unit is 300\% and yet such inlining is very sane. For very
8261 large units consisting of small inlineable functions however the overall unit
8262 growth limit is needed to avoid exponential explosion of code size. Thus for
8263 smaller units, the size is increased to @option{--param large-unit-insns}
8264 before applying @option{--param inline-unit-growth}. The default is 10000
8266 @item inline-unit-growth
8267 Specifies maximal overall growth of the compilation unit caused by inlining.
8268 The default value is 30 which limits unit growth to 1.3 times the original
8271 @item ipcp-unit-growth
8272 Specifies maximal overall growth of the compilation unit caused by
8273 interprocedural constant propagation. The default value is 10 which limits
8274 unit growth to 1.1 times the original size.
8276 @item large-stack-frame
8277 The limit specifying large stack frames. While inlining the algorithm is trying
8278 to not grow past this limit too much. Default value is 256 bytes.
8280 @item large-stack-frame-growth
8281 Specifies maximal growth of large stack frames caused by inlining in percents.
8282 The default value is 1000 which limits large stack frame growth to 11 times
8285 @item max-inline-insns-recursive
8286 @itemx max-inline-insns-recursive-auto
8287 Specifies maximum number of instructions out-of-line copy of self recursive inline
8288 function can grow into by performing recursive inlining.
8290 For functions declared inline @option{--param max-inline-insns-recursive} is
8291 taken into account. For function not declared inline, recursive inlining
8292 happens only when @option{-finline-functions} (included in @option{-O3}) is
8293 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
8294 default value is 450.
8296 @item max-inline-recursive-depth
8297 @itemx max-inline-recursive-depth-auto
8298 Specifies maximum recursion depth used by the recursive inlining.
8300 For functions declared inline @option{--param max-inline-recursive-depth} is
8301 taken into account. For function not declared inline, recursive inlining
8302 happens only when @option{-finline-functions} (included in @option{-O3}) is
8303 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
8306 @item min-inline-recursive-probability
8307 Recursive inlining is profitable only for function having deep recursion
8308 in average and can hurt for function having little recursion depth by
8309 increasing the prologue size or complexity of function body to other
8312 When profile feedback is available (see @option{-fprofile-generate}) the actual
8313 recursion depth can be guessed from probability that function will recurse via
8314 given call expression. This parameter limits inlining only to call expression
8315 whose probability exceeds given threshold (in percents). The default value is
8318 @item early-inlining-insns
8319 Specify growth that early inliner can make. In effect it increases amount of
8320 inlining for code having large abstraction penalty. The default value is 10.
8322 @item max-early-inliner-iterations
8323 @itemx max-early-inliner-iterations
8324 Limit of iterations of early inliner. This basically bounds number of nested
8325 indirect calls early inliner can resolve. Deeper chains are still handled by
8328 @item comdat-sharing-probability
8329 @itemx comdat-sharing-probability
8330 Probability (in percent) that C++ inline function with comdat visibility
8331 will be shared acroess multiple compilation units. The default value is 20.
8333 @item min-vect-loop-bound
8334 The minimum number of iterations under which a loop will not get vectorized
8335 when @option{-ftree-vectorize} is used. The number of iterations after
8336 vectorization needs to be greater than the value specified by this option
8337 to allow vectorization. The default value is 0.
8339 @item gcse-cost-distance-ratio
8340 Scaling factor in calculation of maximum distance an expression
8341 can be moved by GCSE optimizations. This is currently supported only in
8342 code hoisting pass. The bigger the ratio, the more agressive code hoisting
8343 will be with simple expressions, i.e., the expressions which have cost
8344 less than @option{gcse-unrestricted-cost}. Specifying 0 will disable
8345 hoisting of simple expressions. The default value is 10.
8347 @item gcse-unrestricted-cost
8348 Cost, roughly measured as the cost of a single typical machine
8349 instruction, at which GCSE optimizations will not constrain
8350 the distance an expression can travel. This is currently
8351 supported only in code hoisting pass. The lesser the cost,
8352 the more aggressive code hoisting will be. Specifying 0 will
8353 allow all expressions to travel unrestricted distances.
8354 The default value is 3.
8356 @item max-hoist-depth
8357 The depth of search in the dominator tree for expressions to hoist.
8358 This is used to avoid quadratic behavior in hoisting algorithm.
8359 The value of 0 will avoid limiting the search, but may slow down compilation
8360 of huge functions. The default value is 30.
8362 @item max-unrolled-insns
8363 The maximum number of instructions that a loop should have if that loop
8364 is unrolled, and if the loop is unrolled, it determines how many times
8365 the loop code is unrolled.
8367 @item max-average-unrolled-insns
8368 The maximum number of instructions biased by probabilities of their execution
8369 that a loop should have if that loop is unrolled, and if the loop is unrolled,
8370 it determines how many times the loop code is unrolled.
8372 @item max-unroll-times
8373 The maximum number of unrollings of a single loop.
8375 @item max-peeled-insns
8376 The maximum number of instructions that a loop should have if that loop
8377 is peeled, and if the loop is peeled, it determines how many times
8378 the loop code is peeled.
8380 @item max-peel-times
8381 The maximum number of peelings of a single loop.
8383 @item max-completely-peeled-insns
8384 The maximum number of insns of a completely peeled loop.
8386 @item max-completely-peel-times
8387 The maximum number of iterations of a loop to be suitable for complete peeling.
8389 @item max-completely-peel-loop-nest-depth
8390 The maximum depth of a loop nest suitable for complete peeling.
8392 @item max-unswitch-insns
8393 The maximum number of insns of an unswitched loop.
8395 @item max-unswitch-level
8396 The maximum number of branches unswitched in a single loop.
8399 The minimum cost of an expensive expression in the loop invariant motion.
8401 @item iv-consider-all-candidates-bound
8402 Bound on number of candidates for induction variables below that
8403 all candidates are considered for each use in induction variable
8404 optimizations. Only the most relevant candidates are considered
8405 if there are more candidates, to avoid quadratic time complexity.
8407 @item iv-max-considered-uses
8408 The induction variable optimizations give up on loops that contain more
8409 induction variable uses.
8411 @item iv-always-prune-cand-set-bound
8412 If number of candidates in the set is smaller than this value,
8413 we always try to remove unnecessary ivs from the set during its
8414 optimization when a new iv is added to the set.
8416 @item scev-max-expr-size
8417 Bound on size of expressions used in the scalar evolutions analyzer.
8418 Large expressions slow the analyzer.
8420 @item omega-max-vars
8421 The maximum number of variables in an Omega constraint system.
8422 The default value is 128.
8424 @item omega-max-geqs
8425 The maximum number of inequalities in an Omega constraint system.
8426 The default value is 256.
8429 The maximum number of equalities in an Omega constraint system.
8430 The default value is 128.
8432 @item omega-max-wild-cards
8433 The maximum number of wildcard variables that the Omega solver will
8434 be able to insert. The default value is 18.
8436 @item omega-hash-table-size
8437 The size of the hash table in the Omega solver. The default value is
8440 @item omega-max-keys
8441 The maximal number of keys used by the Omega solver. The default
8444 @item omega-eliminate-redundant-constraints
8445 When set to 1, use expensive methods to eliminate all redundant
8446 constraints. The default value is 0.
8448 @item vect-max-version-for-alignment-checks
8449 The maximum number of runtime checks that can be performed when
8450 doing loop versioning for alignment in the vectorizer. See option
8451 ftree-vect-loop-version for more information.
8453 @item vect-max-version-for-alias-checks
8454 The maximum number of runtime checks that can be performed when
8455 doing loop versioning for alias in the vectorizer. See option
8456 ftree-vect-loop-version for more information.
8458 @item max-iterations-to-track
8460 The maximum number of iterations of a loop the brute force algorithm
8461 for analysis of # of iterations of the loop tries to evaluate.
8463 @item hot-bb-count-fraction
8464 Select fraction of the maximal count of repetitions of basic block in program
8465 given basic block needs to have to be considered hot.
8467 @item hot-bb-frequency-fraction
8468 Select fraction of the maximal frequency of executions of basic block in
8469 function given basic block needs to have to be considered hot
8471 @item max-predicted-iterations
8472 The maximum number of loop iterations we predict statically. This is useful
8473 in cases where function contain single loop with known bound and other loop
8474 with unknown. We predict the known number of iterations correctly, while
8475 the unknown number of iterations average to roughly 10. This means that the
8476 loop without bounds would appear artificially cold relative to the other one.
8478 @item align-threshold
8480 Select fraction of the maximal frequency of executions of basic block in
8481 function given basic block will get aligned.
8483 @item align-loop-iterations
8485 A loop expected to iterate at lest the selected number of iterations will get
8488 @item tracer-dynamic-coverage
8489 @itemx tracer-dynamic-coverage-feedback
8491 This value is used to limit superblock formation once the given percentage of
8492 executed instructions is covered. This limits unnecessary code size
8495 The @option{tracer-dynamic-coverage-feedback} is used only when profile
8496 feedback is available. The real profiles (as opposed to statically estimated
8497 ones) are much less balanced allowing the threshold to be larger value.
8499 @item tracer-max-code-growth
8500 Stop tail duplication once code growth has reached given percentage. This is
8501 rather hokey argument, as most of the duplicates will be eliminated later in
8502 cross jumping, so it may be set to much higher values than is the desired code
8505 @item tracer-min-branch-ratio
8507 Stop reverse growth when the reverse probability of best edge is less than this
8508 threshold (in percent).
8510 @item tracer-min-branch-ratio
8511 @itemx tracer-min-branch-ratio-feedback
8513 Stop forward growth if the best edge do have probability lower than this
8516 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
8517 compilation for profile feedback and one for compilation without. The value
8518 for compilation with profile feedback needs to be more conservative (higher) in
8519 order to make tracer effective.
8521 @item max-cse-path-length
8523 Maximum number of basic blocks on path that cse considers. The default is 10.
8526 The maximum instructions CSE process before flushing. The default is 1000.
8528 @item ggc-min-expand
8530 GCC uses a garbage collector to manage its own memory allocation. This
8531 parameter specifies the minimum percentage by which the garbage
8532 collector's heap should be allowed to expand between collections.
8533 Tuning this may improve compilation speed; it has no effect on code
8536 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
8537 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
8538 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
8539 GCC is not able to calculate RAM on a particular platform, the lower
8540 bound of 30% is used. Setting this parameter and
8541 @option{ggc-min-heapsize} to zero causes a full collection to occur at
8542 every opportunity. This is extremely slow, but can be useful for
8545 @item ggc-min-heapsize
8547 Minimum size of the garbage collector's heap before it begins bothering
8548 to collect garbage. The first collection occurs after the heap expands
8549 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
8550 tuning this may improve compilation speed, and has no effect on code
8553 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
8554 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
8555 with a lower bound of 4096 (four megabytes) and an upper bound of
8556 131072 (128 megabytes). If GCC is not able to calculate RAM on a
8557 particular platform, the lower bound is used. Setting this parameter
8558 very large effectively disables garbage collection. Setting this
8559 parameter and @option{ggc-min-expand} to zero causes a full collection
8560 to occur at every opportunity.
8562 @item max-reload-search-insns
8563 The maximum number of instruction reload should look backward for equivalent
8564 register. Increasing values mean more aggressive optimization, making the
8565 compile time increase with probably slightly better performance. The default
8568 @item max-cselib-memory-locations
8569 The maximum number of memory locations cselib should take into account.
8570 Increasing values mean more aggressive optimization, making the compile time
8571 increase with probably slightly better performance. The default value is 500.
8573 @item reorder-blocks-duplicate
8574 @itemx reorder-blocks-duplicate-feedback
8576 Used by basic block reordering pass to decide whether to use unconditional
8577 branch or duplicate the code on its destination. Code is duplicated when its
8578 estimated size is smaller than this value multiplied by the estimated size of
8579 unconditional jump in the hot spots of the program.
8581 The @option{reorder-block-duplicate-feedback} is used only when profile
8582 feedback is available and may be set to higher values than
8583 @option{reorder-block-duplicate} since information about the hot spots is more
8586 @item max-sched-ready-insns
8587 The maximum number of instructions ready to be issued the scheduler should
8588 consider at any given time during the first scheduling pass. Increasing
8589 values mean more thorough searches, making the compilation time increase
8590 with probably little benefit. The default value is 100.
8592 @item max-sched-region-blocks
8593 The maximum number of blocks in a region to be considered for
8594 interblock scheduling. The default value is 10.
8596 @item max-pipeline-region-blocks
8597 The maximum number of blocks in a region to be considered for
8598 pipelining in the selective scheduler. The default value is 15.
8600 @item max-sched-region-insns
8601 The maximum number of insns in a region to be considered for
8602 interblock scheduling. The default value is 100.
8604 @item max-pipeline-region-insns
8605 The maximum number of insns in a region to be considered for
8606 pipelining in the selective scheduler. The default value is 200.
8609 The minimum probability (in percents) of reaching a source block
8610 for interblock speculative scheduling. The default value is 40.
8612 @item max-sched-extend-regions-iters
8613 The maximum number of iterations through CFG to extend regions.
8614 0 - disable region extension,
8615 N - do at most N iterations.
8616 The default value is 0.
8618 @item max-sched-insn-conflict-delay
8619 The maximum conflict delay for an insn to be considered for speculative motion.
8620 The default value is 3.
8622 @item sched-spec-prob-cutoff
8623 The minimal probability of speculation success (in percents), so that
8624 speculative insn will be scheduled.
8625 The default value is 40.
8627 @item sched-mem-true-dep-cost
8628 Minimal distance (in CPU cycles) between store and load targeting same
8629 memory locations. The default value is 1.
8631 @item selsched-max-lookahead
8632 The maximum size of the lookahead window of selective scheduling. It is a
8633 depth of search for available instructions.
8634 The default value is 50.
8636 @item selsched-max-sched-times
8637 The maximum number of times that an instruction will be scheduled during
8638 selective scheduling. This is the limit on the number of iterations
8639 through which the instruction may be pipelined. The default value is 2.
8641 @item selsched-max-insns-to-rename
8642 The maximum number of best instructions in the ready list that are considered
8643 for renaming in the selective scheduler. The default value is 2.
8645 @item max-last-value-rtl
8646 The maximum size measured as number of RTLs that can be recorded in an expression
8647 in combiner for a pseudo register as last known value of that register. The default
8650 @item integer-share-limit
8651 Small integer constants can use a shared data structure, reducing the
8652 compiler's memory usage and increasing its speed. This sets the maximum
8653 value of a shared integer constant. The default value is 256.
8655 @item min-virtual-mappings
8656 Specifies the minimum number of virtual mappings in the incremental
8657 SSA updater that should be registered to trigger the virtual mappings
8658 heuristic defined by virtual-mappings-ratio. The default value is
8661 @item virtual-mappings-ratio
8662 If the number of virtual mappings is virtual-mappings-ratio bigger
8663 than the number of virtual symbols to be updated, then the incremental
8664 SSA updater switches to a full update for those symbols. The default
8667 @item ssp-buffer-size
8668 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
8669 protection when @option{-fstack-protection} is used.
8671 @item max-jump-thread-duplication-stmts
8672 Maximum number of statements allowed in a block that needs to be
8673 duplicated when threading jumps.
8675 @item max-fields-for-field-sensitive
8676 Maximum number of fields in a structure we will treat in
8677 a field sensitive manner during pointer analysis. The default is zero
8678 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
8680 @item prefetch-latency
8681 Estimate on average number of instructions that are executed before
8682 prefetch finishes. The distance we prefetch ahead is proportional
8683 to this constant. Increasing this number may also lead to less
8684 streams being prefetched (see @option{simultaneous-prefetches}).
8686 @item simultaneous-prefetches
8687 Maximum number of prefetches that can run at the same time.
8689 @item l1-cache-line-size
8690 The size of cache line in L1 cache, in bytes.
8693 The size of L1 cache, in kilobytes.
8696 The size of L2 cache, in kilobytes.
8698 @item min-insn-to-prefetch-ratio
8699 The minimum ratio between the number of instructions and the
8700 number of prefetches to enable prefetching in a loop.
8702 @item prefetch-min-insn-to-mem-ratio
8703 The minimum ratio between the number of instructions and the
8704 number of memory references to enable prefetching in a loop.
8706 @item use-canonical-types
8707 Whether the compiler should use the ``canonical'' type system. By
8708 default, this should always be 1, which uses a more efficient internal
8709 mechanism for comparing types in C++ and Objective-C++. However, if
8710 bugs in the canonical type system are causing compilation failures,
8711 set this value to 0 to disable canonical types.
8713 @item switch-conversion-max-branch-ratio
8714 Switch initialization conversion will refuse to create arrays that are
8715 bigger than @option{switch-conversion-max-branch-ratio} times the number of
8716 branches in the switch.
8718 @item max-partial-antic-length
8719 Maximum length of the partial antic set computed during the tree
8720 partial redundancy elimination optimization (@option{-ftree-pre}) when
8721 optimizing at @option{-O3} and above. For some sorts of source code
8722 the enhanced partial redundancy elimination optimization can run away,
8723 consuming all of the memory available on the host machine. This
8724 parameter sets a limit on the length of the sets that are computed,
8725 which prevents the runaway behavior. Setting a value of 0 for
8726 this parameter will allow an unlimited set length.
8728 @item sccvn-max-scc-size
8729 Maximum size of a strongly connected component (SCC) during SCCVN
8730 processing. If this limit is hit, SCCVN processing for the whole
8731 function will not be done and optimizations depending on it will
8732 be disabled. The default maximum SCC size is 10000.
8734 @item ira-max-loops-num
8735 IRA uses a regional register allocation by default. If a function
8736 contains loops more than number given by the parameter, only at most
8737 given number of the most frequently executed loops will form regions
8738 for the regional register allocation. The default value of the
8741 @item ira-max-conflict-table-size
8742 Although IRA uses a sophisticated algorithm of compression conflict
8743 table, the table can be still big for huge functions. If the conflict
8744 table for a function could be more than size in MB given by the
8745 parameter, the conflict table is not built and faster, simpler, and
8746 lower quality register allocation algorithm will be used. The
8747 algorithm do not use pseudo-register conflicts. The default value of
8748 the parameter is 2000.
8750 @item ira-loop-reserved-regs
8751 IRA can be used to evaluate more accurate register pressure in loops
8752 for decision to move loop invariants (see @option{-O3}). The number
8753 of available registers reserved for some other purposes is described
8754 by this parameter. The default value of the parameter is 2 which is
8755 minimal number of registers needed for execution of typical
8756 instruction. This value is the best found from numerous experiments.
8758 @item loop-invariant-max-bbs-in-loop
8759 Loop invariant motion can be very expensive, both in compile time and
8760 in amount of needed compile time memory, with very large loops. Loops
8761 with more basic blocks than this parameter won't have loop invariant
8762 motion optimization performed on them. The default value of the
8763 parameter is 1000 for -O1 and 10000 for -O2 and above.
8765 @item max-vartrack-size
8766 Sets a maximum number of hash table slots to use during variable
8767 tracking dataflow analysis of any function. If this limit is exceeded
8768 with variable tracking at assignments enabled, analysis for that
8769 function is retried without it, after removing all debug insns from
8770 the function. If the limit is exceeded even without debug insns, var
8771 tracking analysis is completely disabled for the function. Setting
8772 the parameter to zero makes it unlimited.
8774 @item min-nondebug-insn-uid
8775 Use uids starting at this parameter for nondebug insns. The range below
8776 the parameter is reserved exclusively for debug insns created by
8777 @option{-fvar-tracking-assignments}, but debug insns may get
8778 (non-overlapping) uids above it if the reserved range is exhausted.
8780 @item ipa-sra-ptr-growth-factor
8781 IPA-SRA will replace a pointer to an aggregate with one or more new
8782 parameters only when their cumulative size is less or equal to
8783 @option{ipa-sra-ptr-growth-factor} times the size of the original
8786 @item graphite-max-nb-scop-params
8787 To avoid exponential effects in the Graphite loop transforms, the
8788 number of parameters in a Static Control Part (SCoP) is bounded. The
8789 default value is 10 parameters. A variable whose value is unknown at
8790 compile time and defined outside a SCoP is a parameter of the SCoP.
8792 @item graphite-max-bbs-per-function
8793 To avoid exponential effects in the detection of SCoPs, the size of
8794 the functions analyzed by Graphite is bounded. The default value is
8797 @item loop-block-tile-size
8798 Loop blocking or strip mining transforms, enabled with
8799 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
8800 loop in the loop nest by a given number of iterations. The strip
8801 length can be changed using the @option{loop-block-tile-size}
8802 parameter. The default value is 51 iterations.
8804 @item devirt-type-list-size
8805 IPA-CP attempts to track all possible types passed to a function's
8806 parameter in order to perform devirtualization.
8807 @option{devirt-type-list-size} is the maximum number of types it
8808 stores per a single formal parameter of a function.
8810 @item lto-partitions
8811 Specify desired nuber of partitions produced during WHOPR copmilation.
8812 Number of partitions should exceed number of CPUs used for compilatoin.
8813 Default value is 32.
8815 @item lto-minpartition
8816 Size of minimal paritition for WHOPR (in estimated instructions).
8817 This prevents expenses of splitting very small programs into too many
8823 @node Preprocessor Options
8824 @section Options Controlling the Preprocessor
8825 @cindex preprocessor options
8826 @cindex options, preprocessor
8828 These options control the C preprocessor, which is run on each C source
8829 file before actual compilation.
8831 If you use the @option{-E} option, nothing is done except preprocessing.
8832 Some of these options make sense only together with @option{-E} because
8833 they cause the preprocessor output to be unsuitable for actual
8837 @item -Wp,@var{option}
8839 You can use @option{-Wp,@var{option}} to bypass the compiler driver
8840 and pass @var{option} directly through to the preprocessor. If
8841 @var{option} contains commas, it is split into multiple options at the
8842 commas. However, many options are modified, translated or interpreted
8843 by the compiler driver before being passed to the preprocessor, and
8844 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
8845 interface is undocumented and subject to change, so whenever possible
8846 you should avoid using @option{-Wp} and let the driver handle the
8849 @item -Xpreprocessor @var{option}
8850 @opindex Xpreprocessor
8851 Pass @var{option} as an option to the preprocessor. You can use this to
8852 supply system-specific preprocessor options which GCC does not know how to
8855 If you want to pass an option that takes an argument, you must use
8856 @option{-Xpreprocessor} twice, once for the option and once for the argument.
8859 @include cppopts.texi
8861 @node Assembler Options
8862 @section Passing Options to the Assembler
8864 @c prevent bad page break with this line
8865 You can pass options to the assembler.
8868 @item -Wa,@var{option}
8870 Pass @var{option} as an option to the assembler. If @var{option}
8871 contains commas, it is split into multiple options at the commas.
8873 @item -Xassembler @var{option}
8875 Pass @var{option} as an option to the assembler. You can use this to
8876 supply system-specific assembler options which GCC does not know how to
8879 If you want to pass an option that takes an argument, you must use
8880 @option{-Xassembler} twice, once for the option and once for the argument.
8885 @section Options for Linking
8886 @cindex link options
8887 @cindex options, linking
8889 These options come into play when the compiler links object files into
8890 an executable output file. They are meaningless if the compiler is
8891 not doing a link step.
8895 @item @var{object-file-name}
8896 A file name that does not end in a special recognized suffix is
8897 considered to name an object file or library. (Object files are
8898 distinguished from libraries by the linker according to the file
8899 contents.) If linking is done, these object files are used as input
8908 If any of these options is used, then the linker is not run, and
8909 object file names should not be used as arguments. @xref{Overall
8913 @item -l@var{library}
8914 @itemx -l @var{library}
8916 Search the library named @var{library} when linking. (The second
8917 alternative with the library as a separate argument is only for
8918 POSIX compliance and is not recommended.)
8920 It makes a difference where in the command you write this option; the
8921 linker searches and processes libraries and object files in the order they
8922 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
8923 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
8924 to functions in @samp{z}, those functions may not be loaded.
8926 The linker searches a standard list of directories for the library,
8927 which is actually a file named @file{lib@var{library}.a}. The linker
8928 then uses this file as if it had been specified precisely by name.
8930 The directories searched include several standard system directories
8931 plus any that you specify with @option{-L}.
8933 Normally the files found this way are library files---archive files
8934 whose members are object files. The linker handles an archive file by
8935 scanning through it for members which define symbols that have so far
8936 been referenced but not defined. But if the file that is found is an
8937 ordinary object file, it is linked in the usual fashion. The only
8938 difference between using an @option{-l} option and specifying a file name
8939 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
8940 and searches several directories.
8944 You need this special case of the @option{-l} option in order to
8945 link an Objective-C or Objective-C++ program.
8948 @opindex nostartfiles
8949 Do not use the standard system startup files when linking.
8950 The standard system libraries are used normally, unless @option{-nostdlib}
8951 or @option{-nodefaultlibs} is used.
8953 @item -nodefaultlibs
8954 @opindex nodefaultlibs
8955 Do not use the standard system libraries when linking.
8956 Only the libraries you specify will be passed to the linker, options
8957 specifying linkage of the system libraries, such as @code{-static-libgcc}
8958 or @code{-shared-libgcc}, will be ignored.
8959 The standard startup files are used normally, unless @option{-nostartfiles}
8960 is used. The compiler may generate calls to @code{memcmp},
8961 @code{memset}, @code{memcpy} and @code{memmove}.
8962 These entries are usually resolved by entries in
8963 libc. These entry points should be supplied through some other
8964 mechanism when this option is specified.
8968 Do not use the standard system startup files or libraries when linking.
8969 No startup files and only the libraries you specify will be passed to
8970 the linker, options specifying linkage of the system libraries, such as
8971 @code{-static-libgcc} or @code{-shared-libgcc}, will be ignored.
8972 The compiler may generate calls to @code{memcmp}, @code{memset},
8973 @code{memcpy} and @code{memmove}.
8974 These entries are usually resolved by entries in
8975 libc. These entry points should be supplied through some other
8976 mechanism when this option is specified.
8978 @cindex @option{-lgcc}, use with @option{-nostdlib}
8979 @cindex @option{-nostdlib} and unresolved references
8980 @cindex unresolved references and @option{-nostdlib}
8981 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
8982 @cindex @option{-nodefaultlibs} and unresolved references
8983 @cindex unresolved references and @option{-nodefaultlibs}
8984 One of the standard libraries bypassed by @option{-nostdlib} and
8985 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
8986 that GCC uses to overcome shortcomings of particular machines, or special
8987 needs for some languages.
8988 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
8989 Collection (GCC) Internals},
8990 for more discussion of @file{libgcc.a}.)
8991 In most cases, you need @file{libgcc.a} even when you want to avoid
8992 other standard libraries. In other words, when you specify @option{-nostdlib}
8993 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
8994 This ensures that you have no unresolved references to internal GCC
8995 library subroutines. (For example, @samp{__main}, used to ensure C++
8996 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
8997 GNU Compiler Collection (GCC) Internals}.)
9001 Produce a position independent executable on targets which support it.
9002 For predictable results, you must also specify the same set of options
9003 that were used to generate code (@option{-fpie}, @option{-fPIE},
9004 or model suboptions) when you specify this option.
9008 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
9009 that support it. This instructs the linker to add all symbols, not
9010 only used ones, to the dynamic symbol table. This option is needed
9011 for some uses of @code{dlopen} or to allow obtaining backtraces
9012 from within a program.
9016 Remove all symbol table and relocation information from the executable.
9020 On systems that support dynamic linking, this prevents linking with the shared
9021 libraries. On other systems, this option has no effect.
9025 Produce a shared object which can then be linked with other objects to
9026 form an executable. Not all systems support this option. For predictable
9027 results, you must also specify the same set of options that were used to
9028 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
9029 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
9030 needs to build supplementary stub code for constructors to work. On
9031 multi-libbed systems, @samp{gcc -shared} must select the correct support
9032 libraries to link against. Failing to supply the correct flags may lead
9033 to subtle defects. Supplying them in cases where they are not necessary
9036 @item -shared-libgcc
9037 @itemx -static-libgcc
9038 @opindex shared-libgcc
9039 @opindex static-libgcc
9040 On systems that provide @file{libgcc} as a shared library, these options
9041 force the use of either the shared or static version respectively.
9042 If no shared version of @file{libgcc} was built when the compiler was
9043 configured, these options have no effect.
9045 There are several situations in which an application should use the
9046 shared @file{libgcc} instead of the static version. The most common
9047 of these is when the application wishes to throw and catch exceptions
9048 across different shared libraries. In that case, each of the libraries
9049 as well as the application itself should use the shared @file{libgcc}.
9051 Therefore, the G++ and GCJ drivers automatically add
9052 @option{-shared-libgcc} whenever you build a shared library or a main
9053 executable, because C++ and Java programs typically use exceptions, so
9054 this is the right thing to do.
9056 If, instead, you use the GCC driver to create shared libraries, you may
9057 find that they will not always be linked with the shared @file{libgcc}.
9058 If GCC finds, at its configuration time, that you have a non-GNU linker
9059 or a GNU linker that does not support option @option{--eh-frame-hdr},
9060 it will link the shared version of @file{libgcc} into shared libraries
9061 by default. Otherwise, it will take advantage of the linker and optimize
9062 away the linking with the shared version of @file{libgcc}, linking with
9063 the static version of libgcc by default. This allows exceptions to
9064 propagate through such shared libraries, without incurring relocation
9065 costs at library load time.
9067 However, if a library or main executable is supposed to throw or catch
9068 exceptions, you must link it using the G++ or GCJ driver, as appropriate
9069 for the languages used in the program, or using the option
9070 @option{-shared-libgcc}, such that it is linked with the shared
9073 @item -static-libstdc++
9074 When the @command{g++} program is used to link a C++ program, it will
9075 normally automatically link against @option{libstdc++}. If
9076 @file{libstdc++} is available as a shared library, and the
9077 @option{-static} option is not used, then this will link against the
9078 shared version of @file{libstdc++}. That is normally fine. However, it
9079 is sometimes useful to freeze the version of @file{libstdc++} used by
9080 the program without going all the way to a fully static link. The
9081 @option{-static-libstdc++} option directs the @command{g++} driver to
9082 link @file{libstdc++} statically, without necessarily linking other
9083 libraries statically.
9087 Bind references to global symbols when building a shared object. Warn
9088 about any unresolved references (unless overridden by the link editor
9089 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
9092 @item -T @var{script}
9094 @cindex linker script
9095 Use @var{script} as the linker script. This option is supported by most
9096 systems using the GNU linker. On some targets, such as bare-board
9097 targets without an operating system, the @option{-T} option may be required
9098 when linking to avoid references to undefined symbols.
9100 @item -Xlinker @var{option}
9102 Pass @var{option} as an option to the linker. You can use this to
9103 supply system-specific linker options which GCC does not know how to
9106 If you want to pass an option that takes a separate argument, you must use
9107 @option{-Xlinker} twice, once for the option and once for the argument.
9108 For example, to pass @option{-assert definitions}, you must write
9109 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
9110 @option{-Xlinker "-assert definitions"}, because this passes the entire
9111 string as a single argument, which is not what the linker expects.
9113 When using the GNU linker, it is usually more convenient to pass
9114 arguments to linker options using the @option{@var{option}=@var{value}}
9115 syntax than as separate arguments. For example, you can specify
9116 @samp{-Xlinker -Map=output.map} rather than
9117 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
9118 this syntax for command-line options.
9120 @item -Wl,@var{option}
9122 Pass @var{option} as an option to the linker. If @var{option} contains
9123 commas, it is split into multiple options at the commas. You can use this
9124 syntax to pass an argument to the option.
9125 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
9126 linker. When using the GNU linker, you can also get the same effect with
9127 @samp{-Wl,-Map=output.map}.
9129 @item -u @var{symbol}
9131 Pretend the symbol @var{symbol} is undefined, to force linking of
9132 library modules to define it. You can use @option{-u} multiple times with
9133 different symbols to force loading of additional library modules.
9136 @node Directory Options
9137 @section Options for Directory Search
9138 @cindex directory options
9139 @cindex options, directory search
9142 These options specify directories to search for header files, for
9143 libraries and for parts of the compiler:
9148 Add the directory @var{dir} to the head of the list of directories to be
9149 searched for header files. This can be used to override a system header
9150 file, substituting your own version, since these directories are
9151 searched before the system header file directories. However, you should
9152 not use this option to add directories that contain vendor-supplied
9153 system header files (use @option{-isystem} for that). If you use more than
9154 one @option{-I} option, the directories are scanned in left-to-right
9155 order; the standard system directories come after.
9157 If a standard system include directory, or a directory specified with
9158 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
9159 option will be ignored. The directory will still be searched but as a
9160 system directory at its normal position in the system include chain.
9161 This is to ensure that GCC's procedure to fix buggy system headers and
9162 the ordering for the include_next directive are not inadvertently changed.
9163 If you really need to change the search order for system directories,
9164 use the @option{-nostdinc} and/or @option{-isystem} options.
9166 @item -iplugindir=@var{dir}
9167 Set the directory to search for plugins which are passed
9168 by @option{-fplugin=@var{name}} instead of
9169 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
9170 to be used by the user, but only passed by the driver.
9172 @item -iquote@var{dir}
9174 Add the directory @var{dir} to the head of the list of directories to
9175 be searched for header files only for the case of @samp{#include
9176 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
9177 otherwise just like @option{-I}.
9181 Add directory @var{dir} to the list of directories to be searched
9184 @item -B@var{prefix}
9186 This option specifies where to find the executables, libraries,
9187 include files, and data files of the compiler itself.
9189 The compiler driver program runs one or more of the subprograms
9190 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
9191 @var{prefix} as a prefix for each program it tries to run, both with and
9192 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
9194 For each subprogram to be run, the compiler driver first tries the
9195 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
9196 was not specified, the driver tries two standard prefixes, which are
9197 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
9198 those results in a file name that is found, the unmodified program
9199 name is searched for using the directories specified in your
9200 @env{PATH} environment variable.
9202 The compiler will check to see if the path provided by the @option{-B}
9203 refers to a directory, and if necessary it will add a directory
9204 separator character at the end of the path.
9206 @option{-B} prefixes that effectively specify directory names also apply
9207 to libraries in the linker, because the compiler translates these
9208 options into @option{-L} options for the linker. They also apply to
9209 includes files in the preprocessor, because the compiler translates these
9210 options into @option{-isystem} options for the preprocessor. In this case,
9211 the compiler appends @samp{include} to the prefix.
9213 The run-time support file @file{libgcc.a} can also be searched for using
9214 the @option{-B} prefix, if needed. If it is not found there, the two
9215 standard prefixes above are tried, and that is all. The file is left
9216 out of the link if it is not found by those means.
9218 Another way to specify a prefix much like the @option{-B} prefix is to use
9219 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
9222 As a special kludge, if the path provided by @option{-B} is
9223 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
9224 9, then it will be replaced by @file{[dir/]include}. This is to help
9225 with boot-strapping the compiler.
9227 @item -specs=@var{file}
9229 Process @var{file} after the compiler reads in the standard @file{specs}
9230 file, in order to override the defaults that the @file{gcc} driver
9231 program uses when determining what switches to pass to @file{cc1},
9232 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
9233 @option{-specs=@var{file}} can be specified on the command line, and they
9234 are processed in order, from left to right.
9236 @item --sysroot=@var{dir}
9238 Use @var{dir} as the logical root directory for headers and libraries.
9239 For example, if the compiler would normally search for headers in
9240 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
9241 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
9243 If you use both this option and the @option{-isysroot} option, then
9244 the @option{--sysroot} option will apply to libraries, but the
9245 @option{-isysroot} option will apply to header files.
9247 The GNU linker (beginning with version 2.16) has the necessary support
9248 for this option. If your linker does not support this option, the
9249 header file aspect of @option{--sysroot} will still work, but the
9250 library aspect will not.
9254 This option has been deprecated. Please use @option{-iquote} instead for
9255 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
9256 Any directories you specify with @option{-I} options before the @option{-I-}
9257 option are searched only for the case of @samp{#include "@var{file}"};
9258 they are not searched for @samp{#include <@var{file}>}.
9260 If additional directories are specified with @option{-I} options after
9261 the @option{-I-}, these directories are searched for all @samp{#include}
9262 directives. (Ordinarily @emph{all} @option{-I} directories are used
9265 In addition, the @option{-I-} option inhibits the use of the current
9266 directory (where the current input file came from) as the first search
9267 directory for @samp{#include "@var{file}"}. There is no way to
9268 override this effect of @option{-I-}. With @option{-I.} you can specify
9269 searching the directory which was current when the compiler was
9270 invoked. That is not exactly the same as what the preprocessor does
9271 by default, but it is often satisfactory.
9273 @option{-I-} does not inhibit the use of the standard system directories
9274 for header files. Thus, @option{-I-} and @option{-nostdinc} are
9281 @section Specifying subprocesses and the switches to pass to them
9284 @command{gcc} is a driver program. It performs its job by invoking a
9285 sequence of other programs to do the work of compiling, assembling and
9286 linking. GCC interprets its command-line parameters and uses these to
9287 deduce which programs it should invoke, and which command-line options
9288 it ought to place on their command lines. This behavior is controlled
9289 by @dfn{spec strings}. In most cases there is one spec string for each
9290 program that GCC can invoke, but a few programs have multiple spec
9291 strings to control their behavior. The spec strings built into GCC can
9292 be overridden by using the @option{-specs=} command-line switch to specify
9295 @dfn{Spec files} are plaintext files that are used to construct spec
9296 strings. They consist of a sequence of directives separated by blank
9297 lines. The type of directive is determined by the first non-whitespace
9298 character on the line and it can be one of the following:
9301 @item %@var{command}
9302 Issues a @var{command} to the spec file processor. The commands that can
9306 @item %include <@var{file}>
9307 @cindex @code{%include}
9308 Search for @var{file} and insert its text at the current point in the
9311 @item %include_noerr <@var{file}>
9312 @cindex @code{%include_noerr}
9313 Just like @samp{%include}, but do not generate an error message if the include
9314 file cannot be found.
9316 @item %rename @var{old_name} @var{new_name}
9317 @cindex @code{%rename}
9318 Rename the spec string @var{old_name} to @var{new_name}.
9322 @item *[@var{spec_name}]:
9323 This tells the compiler to create, override or delete the named spec
9324 string. All lines after this directive up to the next directive or
9325 blank line are considered to be the text for the spec string. If this
9326 results in an empty string then the spec will be deleted. (Or, if the
9327 spec did not exist, then nothing will happened.) Otherwise, if the spec
9328 does not currently exist a new spec will be created. If the spec does
9329 exist then its contents will be overridden by the text of this
9330 directive, unless the first character of that text is the @samp{+}
9331 character, in which case the text will be appended to the spec.
9333 @item [@var{suffix}]:
9334 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
9335 and up to the next directive or blank line are considered to make up the
9336 spec string for the indicated suffix. When the compiler encounters an
9337 input file with the named suffix, it will processes the spec string in
9338 order to work out how to compile that file. For example:
9345 This says that any input file whose name ends in @samp{.ZZ} should be
9346 passed to the program @samp{z-compile}, which should be invoked with the
9347 command-line switch @option{-input} and with the result of performing the
9348 @samp{%i} substitution. (See below.)
9350 As an alternative to providing a spec string, the text that follows a
9351 suffix directive can be one of the following:
9354 @item @@@var{language}
9355 This says that the suffix is an alias for a known @var{language}. This is
9356 similar to using the @option{-x} command-line switch to GCC to specify a
9357 language explicitly. For example:
9364 Says that .ZZ files are, in fact, C++ source files.
9367 This causes an error messages saying:
9370 @var{name} compiler not installed on this system.
9374 GCC already has an extensive list of suffixes built into it.
9375 This directive will add an entry to the end of the list of suffixes, but
9376 since the list is searched from the end backwards, it is effectively
9377 possible to override earlier entries using this technique.
9381 GCC has the following spec strings built into it. Spec files can
9382 override these strings or create their own. Note that individual
9383 targets can also add their own spec strings to this list.
9386 asm Options to pass to the assembler
9387 asm_final Options to pass to the assembler post-processor
9388 cpp Options to pass to the C preprocessor
9389 cc1 Options to pass to the C compiler
9390 cc1plus Options to pass to the C++ compiler
9391 endfile Object files to include at the end of the link
9392 link Options to pass to the linker
9393 lib Libraries to include on the command line to the linker
9394 libgcc Decides which GCC support library to pass to the linker
9395 linker Sets the name of the linker
9396 predefines Defines to be passed to the C preprocessor
9397 signed_char Defines to pass to CPP to say whether @code{char} is signed
9399 startfile Object files to include at the start of the link
9402 Here is a small example of a spec file:
9408 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
9411 This example renames the spec called @samp{lib} to @samp{old_lib} and
9412 then overrides the previous definition of @samp{lib} with a new one.
9413 The new definition adds in some extra command-line options before
9414 including the text of the old definition.
9416 @dfn{Spec strings} are a list of command-line options to be passed to their
9417 corresponding program. In addition, the spec strings can contain
9418 @samp{%}-prefixed sequences to substitute variable text or to
9419 conditionally insert text into the command line. Using these constructs
9420 it is possible to generate quite complex command lines.
9422 Here is a table of all defined @samp{%}-sequences for spec
9423 strings. Note that spaces are not generated automatically around the
9424 results of expanding these sequences. Therefore you can concatenate them
9425 together or combine them with constant text in a single argument.
9429 Substitute one @samp{%} into the program name or argument.
9432 Substitute the name of the input file being processed.
9435 Substitute the basename of the input file being processed.
9436 This is the substring up to (and not including) the last period
9437 and not including the directory.
9440 This is the same as @samp{%b}, but include the file suffix (text after
9444 Marks the argument containing or following the @samp{%d} as a
9445 temporary file name, so that that file will be deleted if GCC exits
9446 successfully. Unlike @samp{%g}, this contributes no text to the
9449 @item %g@var{suffix}
9450 Substitute a file name that has suffix @var{suffix} and is chosen
9451 once per compilation, and mark the argument in the same way as
9452 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
9453 name is now chosen in a way that is hard to predict even when previously
9454 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
9455 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
9456 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
9457 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
9458 was simply substituted with a file name chosen once per compilation,
9459 without regard to any appended suffix (which was therefore treated
9460 just like ordinary text), making such attacks more likely to succeed.
9462 @item %u@var{suffix}
9463 Like @samp{%g}, but generates a new temporary file name even if
9464 @samp{%u@var{suffix}} was already seen.
9466 @item %U@var{suffix}
9467 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
9468 new one if there is no such last file name. In the absence of any
9469 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
9470 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
9471 would involve the generation of two distinct file names, one
9472 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
9473 simply substituted with a file name chosen for the previous @samp{%u},
9474 without regard to any appended suffix.
9476 @item %j@var{suffix}
9477 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
9478 writable, and if save-temps is off; otherwise, substitute the name
9479 of a temporary file, just like @samp{%u}. This temporary file is not
9480 meant for communication between processes, but rather as a junk
9483 @item %|@var{suffix}
9484 @itemx %m@var{suffix}
9485 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
9486 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
9487 all. These are the two most common ways to instruct a program that it
9488 should read from standard input or write to standard output. If you
9489 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
9490 construct: see for example @file{f/lang-specs.h}.
9492 @item %.@var{SUFFIX}
9493 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
9494 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
9495 terminated by the next space or %.
9498 Marks the argument containing or following the @samp{%w} as the
9499 designated output file of this compilation. This puts the argument
9500 into the sequence of arguments that @samp{%o} will substitute later.
9503 Substitutes the names of all the output files, with spaces
9504 automatically placed around them. You should write spaces
9505 around the @samp{%o} as well or the results are undefined.
9506 @samp{%o} is for use in the specs for running the linker.
9507 Input files whose names have no recognized suffix are not compiled
9508 at all, but they are included among the output files, so they will
9512 Substitutes the suffix for object files. Note that this is
9513 handled specially when it immediately follows @samp{%g, %u, or %U},
9514 because of the need for those to form complete file names. The
9515 handling is such that @samp{%O} is treated exactly as if it had already
9516 been substituted, except that @samp{%g, %u, and %U} do not currently
9517 support additional @var{suffix} characters following @samp{%O} as they would
9518 following, for example, @samp{.o}.
9521 Substitutes the standard macro predefinitions for the
9522 current target machine. Use this when running @code{cpp}.
9525 Like @samp{%p}, but puts @samp{__} before and after the name of each
9526 predefined macro, except for macros that start with @samp{__} or with
9527 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
9531 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
9532 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
9533 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
9534 and @option{-imultilib} as necessary.
9537 Current argument is the name of a library or startup file of some sort.
9538 Search for that file in a standard list of directories and substitute
9539 the full name found. The current working directory is included in the
9540 list of directories scanned.
9543 Current argument is the name of a linker script. Search for that file
9544 in the current list of directories to scan for libraries. If the file
9545 is located insert a @option{--script} option into the command line
9546 followed by the full path name found. If the file is not found then
9547 generate an error message. Note: the current working directory is not
9551 Print @var{str} as an error message. @var{str} is terminated by a newline.
9552 Use this when inconsistent options are detected.
9555 Substitute the contents of spec string @var{name} at this point.
9558 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
9560 @item %x@{@var{option}@}
9561 Accumulate an option for @samp{%X}.
9564 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
9568 Output the accumulated assembler options specified by @option{-Wa}.
9571 Output the accumulated preprocessor options specified by @option{-Wp}.
9574 Process the @code{asm} spec. This is used to compute the
9575 switches to be passed to the assembler.
9578 Process the @code{asm_final} spec. This is a spec string for
9579 passing switches to an assembler post-processor, if such a program is
9583 Process the @code{link} spec. This is the spec for computing the
9584 command line passed to the linker. Typically it will make use of the
9585 @samp{%L %G %S %D and %E} sequences.
9588 Dump out a @option{-L} option for each directory that GCC believes might
9589 contain startup files. If the target supports multilibs then the
9590 current multilib directory will be prepended to each of these paths.
9593 Process the @code{lib} spec. This is a spec string for deciding which
9594 libraries should be included on the command line to the linker.
9597 Process the @code{libgcc} spec. This is a spec string for deciding
9598 which GCC support library should be included on the command line to the linker.
9601 Process the @code{startfile} spec. This is a spec for deciding which
9602 object files should be the first ones passed to the linker. Typically
9603 this might be a file named @file{crt0.o}.
9606 Process the @code{endfile} spec. This is a spec string that specifies
9607 the last object files that will be passed to the linker.
9610 Process the @code{cpp} spec. This is used to construct the arguments
9611 to be passed to the C preprocessor.
9614 Process the @code{cc1} spec. This is used to construct the options to be
9615 passed to the actual C compiler (@samp{cc1}).
9618 Process the @code{cc1plus} spec. This is used to construct the options to be
9619 passed to the actual C++ compiler (@samp{cc1plus}).
9622 Substitute the variable part of a matched option. See below.
9623 Note that each comma in the substituted string is replaced by
9627 Remove all occurrences of @code{-S} from the command line. Note---this
9628 command is position dependent. @samp{%} commands in the spec string
9629 before this one will see @code{-S}, @samp{%} commands in the spec string
9630 after this one will not.
9632 @item %:@var{function}(@var{args})
9633 Call the named function @var{function}, passing it @var{args}.
9634 @var{args} is first processed as a nested spec string, then split
9635 into an argument vector in the usual fashion. The function returns
9636 a string which is processed as if it had appeared literally as part
9637 of the current spec.
9639 The following built-in spec functions are provided:
9643 The @code{getenv} spec function takes two arguments: an environment
9644 variable name and a string. If the environment variable is not
9645 defined, a fatal error is issued. Otherwise, the return value is the
9646 value of the environment variable concatenated with the string. For
9647 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
9650 %:getenv(TOPDIR /include)
9653 expands to @file{/path/to/top/include}.
9655 @item @code{if-exists}
9656 The @code{if-exists} spec function takes one argument, an absolute
9657 pathname to a file. If the file exists, @code{if-exists} returns the
9658 pathname. Here is a small example of its usage:
9662 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
9665 @item @code{if-exists-else}
9666 The @code{if-exists-else} spec function is similar to the @code{if-exists}
9667 spec function, except that it takes two arguments. The first argument is
9668 an absolute pathname to a file. If the file exists, @code{if-exists-else}
9669 returns the pathname. If it does not exist, it returns the second argument.
9670 This way, @code{if-exists-else} can be used to select one file or another,
9671 based on the existence of the first. Here is a small example of its usage:
9675 crt0%O%s %:if-exists(crti%O%s) \
9676 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
9679 @item @code{replace-outfile}
9680 The @code{replace-outfile} spec function takes two arguments. It looks for the
9681 first argument in the outfiles array and replaces it with the second argument. Here
9682 is a small example of its usage:
9685 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
9688 @item @code{remove-outfile}
9689 The @code{remove-outfile} spec function takes one argument. It looks for the
9690 first argument in the outfiles array and removes it. Here is a small example
9694 %:remove-outfile(-lm)
9697 @item @code{pass-through-libs}
9698 The @code{pass-through-libs} spec function takes any number of arguments. It
9699 finds any @option{-l} options and any non-options ending in ".a" (which it
9700 assumes are the names of linker input library archive files) and returns a
9701 result containing all the found arguments each prepended by
9702 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
9703 intended to be passed to the LTO linker plugin.
9706 %:pass-through-libs(%G %L %G)
9709 @item @code{print-asm-header}
9710 The @code{print-asm-header} function takes no arguments and simply
9711 prints a banner like:
9717 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
9720 It is used to separate compiler options from assembler options
9721 in the @option{--target-help} output.
9725 Substitutes the @code{-S} switch, if that switch was given to GCC@.
9726 If that switch was not specified, this substitutes nothing. Note that
9727 the leading dash is omitted when specifying this option, and it is
9728 automatically inserted if the substitution is performed. Thus the spec
9729 string @samp{%@{foo@}} would match the command-line option @option{-foo}
9730 and would output the command line option @option{-foo}.
9732 @item %W@{@code{S}@}
9733 Like %@{@code{S}@} but mark last argument supplied within as a file to be
9736 @item %@{@code{S}*@}
9737 Substitutes all the switches specified to GCC whose names start
9738 with @code{-S}, but which also take an argument. This is used for
9739 switches like @option{-o}, @option{-D}, @option{-I}, etc.
9740 GCC considers @option{-o foo} as being
9741 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
9742 text, including the space. Thus two arguments would be generated.
9744 @item %@{@code{S}*&@code{T}*@}
9745 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
9746 (the order of @code{S} and @code{T} in the spec is not significant).
9747 There can be any number of ampersand-separated variables; for each the
9748 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
9750 @item %@{@code{S}:@code{X}@}
9751 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
9753 @item %@{!@code{S}:@code{X}@}
9754 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
9756 @item %@{@code{S}*:@code{X}@}
9757 Substitutes @code{X} if one or more switches whose names start with
9758 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
9759 once, no matter how many such switches appeared. However, if @code{%*}
9760 appears somewhere in @code{X}, then @code{X} will be substituted once
9761 for each matching switch, with the @code{%*} replaced by the part of
9762 that switch that matched the @code{*}.
9764 @item %@{.@code{S}:@code{X}@}
9765 Substitutes @code{X}, if processing a file with suffix @code{S}.
9767 @item %@{!.@code{S}:@code{X}@}
9768 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
9770 @item %@{,@code{S}:@code{X}@}
9771 Substitutes @code{X}, if processing a file for language @code{S}.
9773 @item %@{!,@code{S}:@code{X}@}
9774 Substitutes @code{X}, if not processing a file for language @code{S}.
9776 @item %@{@code{S}|@code{P}:@code{X}@}
9777 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
9778 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
9779 @code{*} sequences as well, although they have a stronger binding than
9780 the @samp{|}. If @code{%*} appears in @code{X}, all of the
9781 alternatives must be starred, and only the first matching alternative
9784 For example, a spec string like this:
9787 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
9790 will output the following command-line options from the following input
9791 command-line options:
9796 -d fred.c -foo -baz -boggle
9797 -d jim.d -bar -baz -boggle
9800 @item %@{S:X; T:Y; :D@}
9802 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
9803 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
9804 be as many clauses as you need. This may be combined with @code{.},
9805 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
9810 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
9811 construct may contain other nested @samp{%} constructs or spaces, or
9812 even newlines. They are processed as usual, as described above.
9813 Trailing white space in @code{X} is ignored. White space may also
9814 appear anywhere on the left side of the colon in these constructs,
9815 except between @code{.} or @code{*} and the corresponding word.
9817 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
9818 handled specifically in these constructs. If another value of
9819 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
9820 @option{-W} switch is found later in the command line, the earlier
9821 switch value is ignored, except with @{@code{S}*@} where @code{S} is
9822 just one letter, which passes all matching options.
9824 The character @samp{|} at the beginning of the predicate text is used to
9825 indicate that a command should be piped to the following command, but
9826 only if @option{-pipe} is specified.
9828 It is built into GCC which switches take arguments and which do not.
9829 (You might think it would be useful to generalize this to allow each
9830 compiler's spec to say which switches take arguments. But this cannot
9831 be done in a consistent fashion. GCC cannot even decide which input
9832 files have been specified without knowing which switches take arguments,
9833 and it must know which input files to compile in order to tell which
9836 GCC also knows implicitly that arguments starting in @option{-l} are to be
9837 treated as compiler output files, and passed to the linker in their
9838 proper position among the other output files.
9840 @c man begin OPTIONS
9842 @node Target Options
9843 @section Specifying Target Machine and Compiler Version
9844 @cindex target options
9845 @cindex cross compiling
9846 @cindex specifying machine version
9847 @cindex specifying compiler version and target machine
9848 @cindex compiler version, specifying
9849 @cindex target machine, specifying
9851 The usual way to run GCC is to run the executable called @file{gcc}, or
9852 @file{<machine>-gcc} when cross-compiling, or
9853 @file{<machine>-gcc-<version>} to run a version other than the one that
9856 @node Submodel Options
9857 @section Hardware Models and Configurations
9858 @cindex submodel options
9859 @cindex specifying hardware config
9860 @cindex hardware models and configurations, specifying
9861 @cindex machine dependent options
9863 Each target machine types can have its own
9864 special options, starting with @samp{-m}, to choose among various
9865 hardware models or configurations---for example, 68010 vs 68020,
9866 floating coprocessor or none. A single installed version of the
9867 compiler can compile for any model or configuration, according to the
9870 Some configurations of the compiler also support additional special
9871 options, usually for compatibility with other compilers on the same
9874 @c This list is ordered alphanumerically by subsection name.
9875 @c It should be the same order and spelling as these options are listed
9876 @c in Machine Dependent Options
9882 * Blackfin Options::
9886 * DEC Alpha Options::
9887 * DEC Alpha/VMS Options::
9890 * GNU/Linux Options::
9893 * i386 and x86-64 Options::
9894 * i386 and x86-64 Windows Options::
9896 * IA-64/VMS Options::
9904 * MicroBlaze Options::
9909 * picoChip Options::
9911 * RS/6000 and PowerPC Options::
9913 * S/390 and zSeries Options::
9916 * Solaris 2 Options::
9919 * System V Options::
9924 * Xstormy16 Options::
9930 @subsection ARC Options
9933 These options are defined for ARC implementations:
9938 Compile code for little endian mode. This is the default.
9942 Compile code for big endian mode.
9945 @opindex mmangle-cpu
9946 Prepend the name of the cpu to all public symbol names.
9947 In multiple-processor systems, there are many ARC variants with different
9948 instruction and register set characteristics. This flag prevents code
9949 compiled for one cpu to be linked with code compiled for another.
9950 No facility exists for handling variants that are ``almost identical''.
9951 This is an all or nothing option.
9953 @item -mcpu=@var{cpu}
9955 Compile code for ARC variant @var{cpu}.
9956 Which variants are supported depend on the configuration.
9957 All variants support @option{-mcpu=base}, this is the default.
9959 @item -mtext=@var{text-section}
9960 @itemx -mdata=@var{data-section}
9961 @itemx -mrodata=@var{readonly-data-section}
9965 Put functions, data, and readonly data in @var{text-section},
9966 @var{data-section}, and @var{readonly-data-section} respectively
9967 by default. This can be overridden with the @code{section} attribute.
9968 @xref{Variable Attributes}.
9973 @subsection ARM Options
9976 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
9980 @item -mabi=@var{name}
9982 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
9983 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
9986 @opindex mapcs-frame
9987 Generate a stack frame that is compliant with the ARM Procedure Call
9988 Standard for all functions, even if this is not strictly necessary for
9989 correct execution of the code. Specifying @option{-fomit-frame-pointer}
9990 with this option will cause the stack frames not to be generated for
9991 leaf functions. The default is @option{-mno-apcs-frame}.
9995 This is a synonym for @option{-mapcs-frame}.
9998 @c not currently implemented
9999 @item -mapcs-stack-check
10000 @opindex mapcs-stack-check
10001 Generate code to check the amount of stack space available upon entry to
10002 every function (that actually uses some stack space). If there is
10003 insufficient space available then either the function
10004 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
10005 called, depending upon the amount of stack space required. The run time
10006 system is required to provide these functions. The default is
10007 @option{-mno-apcs-stack-check}, since this produces smaller code.
10009 @c not currently implemented
10011 @opindex mapcs-float
10012 Pass floating point arguments using the float point registers. This is
10013 one of the variants of the APCS@. This option is recommended if the
10014 target hardware has a floating point unit or if a lot of floating point
10015 arithmetic is going to be performed by the code. The default is
10016 @option{-mno-apcs-float}, since integer only code is slightly increased in
10017 size if @option{-mapcs-float} is used.
10019 @c not currently implemented
10020 @item -mapcs-reentrant
10021 @opindex mapcs-reentrant
10022 Generate reentrant, position independent code. The default is
10023 @option{-mno-apcs-reentrant}.
10026 @item -mthumb-interwork
10027 @opindex mthumb-interwork
10028 Generate code which supports calling between the ARM and Thumb
10029 instruction sets. Without this option the two instruction sets cannot
10030 be reliably used inside one program. The default is
10031 @option{-mno-thumb-interwork}, since slightly larger code is generated
10032 when @option{-mthumb-interwork} is specified.
10034 @item -mno-sched-prolog
10035 @opindex mno-sched-prolog
10036 Prevent the reordering of instructions in the function prolog, or the
10037 merging of those instruction with the instructions in the function's
10038 body. This means that all functions will start with a recognizable set
10039 of instructions (or in fact one of a choice from a small set of
10040 different function prologues), and this information can be used to
10041 locate the start if functions inside an executable piece of code. The
10042 default is @option{-msched-prolog}.
10044 @item -mfloat-abi=@var{name}
10045 @opindex mfloat-abi
10046 Specifies which floating-point ABI to use. Permissible values
10047 are: @samp{soft}, @samp{softfp} and @samp{hard}.
10049 Specifying @samp{soft} causes GCC to generate output containing
10050 library calls for floating-point operations.
10051 @samp{softfp} allows the generation of code using hardware floating-point
10052 instructions, but still uses the soft-float calling conventions.
10053 @samp{hard} allows generation of floating-point instructions
10054 and uses FPU-specific calling conventions.
10056 The default depends on the specific target configuration. Note that
10057 the hard-float and soft-float ABIs are not link-compatible; you must
10058 compile your entire program with the same ABI, and link with a
10059 compatible set of libraries.
10062 @opindex mhard-float
10063 Equivalent to @option{-mfloat-abi=hard}.
10066 @opindex msoft-float
10067 Equivalent to @option{-mfloat-abi=soft}.
10069 @item -mlittle-endian
10070 @opindex mlittle-endian
10071 Generate code for a processor running in little-endian mode. This is
10072 the default for all standard configurations.
10075 @opindex mbig-endian
10076 Generate code for a processor running in big-endian mode; the default is
10077 to compile code for a little-endian processor.
10079 @item -mwords-little-endian
10080 @opindex mwords-little-endian
10081 This option only applies when generating code for big-endian processors.
10082 Generate code for a little-endian word order but a big-endian byte
10083 order. That is, a byte order of the form @samp{32107654}. Note: this
10084 option should only be used if you require compatibility with code for
10085 big-endian ARM processors generated by versions of the compiler prior to
10088 @item -mcpu=@var{name}
10090 This specifies the name of the target ARM processor. GCC uses this name
10091 to determine what kind of instructions it can emit when generating
10092 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
10093 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
10094 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
10095 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
10096 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
10098 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
10099 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
10100 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
10101 @samp{strongarm1110},
10102 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
10103 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
10104 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
10105 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
10106 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
10107 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
10108 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
10109 @samp{cortex-a5}, @samp{cortex-a8}, @samp{cortex-a9}, @samp{cortex-a15},
10110 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-m4}, @samp{cortex-m3},
10113 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
10115 @item -mtune=@var{name}
10117 This option is very similar to the @option{-mcpu=} option, except that
10118 instead of specifying the actual target processor type, and hence
10119 restricting which instructions can be used, it specifies that GCC should
10120 tune the performance of the code as if the target were of the type
10121 specified in this option, but still choosing the instructions that it
10122 will generate based on the cpu specified by a @option{-mcpu=} option.
10123 For some ARM implementations better performance can be obtained by using
10126 @item -march=@var{name}
10128 This specifies the name of the target ARM architecture. GCC uses this
10129 name to determine what kind of instructions it can emit when generating
10130 assembly code. This option can be used in conjunction with or instead
10131 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
10132 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
10133 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
10134 @samp{armv6}, @samp{armv6j},
10135 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
10136 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
10137 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
10139 @item -mfpu=@var{name}
10140 @itemx -mfpe=@var{number}
10141 @itemx -mfp=@var{number}
10145 This specifies what floating point hardware (or hardware emulation) is
10146 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
10147 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-fp16},
10148 @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd}, @samp{vfpv3xd-fp16},
10149 @samp{neon}, @samp{neon-fp16}, @samp{vfpv4}, @samp{vfpv4-d16},
10150 @samp{fpv4-sp-d16} and @samp{neon-vfpv4}.
10151 @option{-mfp} and @option{-mfpe} are synonyms for
10152 @option{-mfpu}=@samp{fpe}@var{number}, for compatibility with older versions
10155 If @option{-msoft-float} is specified this specifies the format of
10156 floating point values.
10158 If the selected floating-point hardware includes the NEON extension
10159 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
10160 operations will not be used by GCC's auto-vectorization pass unless
10161 @option{-funsafe-math-optimizations} is also specified. This is
10162 because NEON hardware does not fully implement the IEEE 754 standard for
10163 floating-point arithmetic (in particular denormal values are treated as
10164 zero), so the use of NEON instructions may lead to a loss of precision.
10166 @item -mfp16-format=@var{name}
10167 @opindex mfp16-format
10168 Specify the format of the @code{__fp16} half-precision floating-point type.
10169 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
10170 the default is @samp{none}, in which case the @code{__fp16} type is not
10171 defined. @xref{Half-Precision}, for more information.
10173 @item -mstructure-size-boundary=@var{n}
10174 @opindex mstructure-size-boundary
10175 The size of all structures and unions will be rounded up to a multiple
10176 of the number of bits set by this option. Permissible values are 8, 32
10177 and 64. The default value varies for different toolchains. For the COFF
10178 targeted toolchain the default value is 8. A value of 64 is only allowed
10179 if the underlying ABI supports it.
10181 Specifying the larger number can produce faster, more efficient code, but
10182 can also increase the size of the program. Different values are potentially
10183 incompatible. Code compiled with one value cannot necessarily expect to
10184 work with code or libraries compiled with another value, if they exchange
10185 information using structures or unions.
10187 @item -mabort-on-noreturn
10188 @opindex mabort-on-noreturn
10189 Generate a call to the function @code{abort} at the end of a
10190 @code{noreturn} function. It will be executed if the function tries to
10194 @itemx -mno-long-calls
10195 @opindex mlong-calls
10196 @opindex mno-long-calls
10197 Tells the compiler to perform function calls by first loading the
10198 address of the function into a register and then performing a subroutine
10199 call on this register. This switch is needed if the target function
10200 will lie outside of the 64 megabyte addressing range of the offset based
10201 version of subroutine call instruction.
10203 Even if this switch is enabled, not all function calls will be turned
10204 into long calls. The heuristic is that static functions, functions
10205 which have the @samp{short-call} attribute, functions that are inside
10206 the scope of a @samp{#pragma no_long_calls} directive and functions whose
10207 definitions have already been compiled within the current compilation
10208 unit, will not be turned into long calls. The exception to this rule is
10209 that weak function definitions, functions with the @samp{long-call}
10210 attribute or the @samp{section} attribute, and functions that are within
10211 the scope of a @samp{#pragma long_calls} directive, will always be
10212 turned into long calls.
10214 This feature is not enabled by default. Specifying
10215 @option{-mno-long-calls} will restore the default behavior, as will
10216 placing the function calls within the scope of a @samp{#pragma
10217 long_calls_off} directive. Note these switches have no effect on how
10218 the compiler generates code to handle function calls via function
10221 @item -msingle-pic-base
10222 @opindex msingle-pic-base
10223 Treat the register used for PIC addressing as read-only, rather than
10224 loading it in the prologue for each function. The run-time system is
10225 responsible for initializing this register with an appropriate value
10226 before execution begins.
10228 @item -mpic-register=@var{reg}
10229 @opindex mpic-register
10230 Specify the register to be used for PIC addressing. The default is R10
10231 unless stack-checking is enabled, when R9 is used.
10233 @item -mcirrus-fix-invalid-insns
10234 @opindex mcirrus-fix-invalid-insns
10235 @opindex mno-cirrus-fix-invalid-insns
10236 Insert NOPs into the instruction stream to in order to work around
10237 problems with invalid Maverick instruction combinations. This option
10238 is only valid if the @option{-mcpu=ep9312} option has been used to
10239 enable generation of instructions for the Cirrus Maverick floating
10240 point co-processor. This option is not enabled by default, since the
10241 problem is only present in older Maverick implementations. The default
10242 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
10245 @item -mpoke-function-name
10246 @opindex mpoke-function-name
10247 Write the name of each function into the text section, directly
10248 preceding the function prologue. The generated code is similar to this:
10252 .ascii "arm_poke_function_name", 0
10255 .word 0xff000000 + (t1 - t0)
10256 arm_poke_function_name
10258 stmfd sp!, @{fp, ip, lr, pc@}
10262 When performing a stack backtrace, code can inspect the value of
10263 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
10264 location @code{pc - 12} and the top 8 bits are set, then we know that
10265 there is a function name embedded immediately preceding this location
10266 and has length @code{((pc[-3]) & 0xff000000)}.
10270 Generate code for the Thumb instruction set. The default is to
10271 use the 32-bit ARM instruction set.
10272 This option automatically enables either 16-bit Thumb-1 or
10273 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
10274 and @option{-march=@var{name}} options. This option is not passed to the
10275 assembler. If you want to force assembler files to be interpreted as Thumb code,
10276 either add a @samp{.thumb} directive to the source or pass the @option{-mthumb}
10277 option directly to the assembler by prefixing it with @option{-Wa}.
10280 @opindex mtpcs-frame
10281 Generate a stack frame that is compliant with the Thumb Procedure Call
10282 Standard for all non-leaf functions. (A leaf function is one that does
10283 not call any other functions.) The default is @option{-mno-tpcs-frame}.
10285 @item -mtpcs-leaf-frame
10286 @opindex mtpcs-leaf-frame
10287 Generate a stack frame that is compliant with the Thumb Procedure Call
10288 Standard for all leaf functions. (A leaf function is one that does
10289 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
10291 @item -mcallee-super-interworking
10292 @opindex mcallee-super-interworking
10293 Gives all externally visible functions in the file being compiled an ARM
10294 instruction set header which switches to Thumb mode before executing the
10295 rest of the function. This allows these functions to be called from
10296 non-interworking code. This option is not valid in AAPCS configurations
10297 because interworking is enabled by default.
10299 @item -mcaller-super-interworking
10300 @opindex mcaller-super-interworking
10301 Allows calls via function pointers (including virtual functions) to
10302 execute correctly regardless of whether the target code has been
10303 compiled for interworking or not. There is a small overhead in the cost
10304 of executing a function pointer if this option is enabled. This option
10305 is not valid in AAPCS configurations because interworking is enabled
10308 @item -mtp=@var{name}
10310 Specify the access model for the thread local storage pointer. The valid
10311 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
10312 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
10313 (supported in the arm6k architecture), and @option{auto}, which uses the
10314 best available method for the selected processor. The default setting is
10317 @item -mword-relocations
10318 @opindex mword-relocations
10319 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
10320 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
10321 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
10324 @item -mfix-cortex-m3-ldrd
10325 @opindex mfix-cortex-m3-ldrd
10326 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
10327 with overlapping destination and base registers are used. This option avoids
10328 generating these instructions. This option is enabled by default when
10329 @option{-mcpu=cortex-m3} is specified.
10334 @subsection AVR Options
10335 @cindex AVR Options
10337 These options are defined for AVR implementations:
10340 @item -mmcu=@var{mcu}
10342 Specify ATMEL AVR instruction set or MCU type.
10344 Instruction set avr1 is for the minimal AVR core, not supported by the C
10345 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
10346 attiny11, attiny12, attiny15, attiny28).
10348 Instruction set avr2 (default) is for the classic AVR core with up to
10349 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
10350 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
10351 at90c8534, at90s8535).
10353 Instruction set avr3 is for the classic AVR core with up to 128K program
10354 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
10356 Instruction set avr4 is for the enhanced AVR core with up to 8K program
10357 memory space (MCU types: atmega8, atmega83, atmega85).
10359 Instruction set avr5 is for the enhanced AVR core with up to 128K program
10360 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
10361 atmega64, atmega128, at43usb355, at94k).
10363 @item -mno-interrupts
10364 @opindex mno-interrupts
10365 Generated code is not compatible with hardware interrupts.
10366 Code size will be smaller.
10368 @item -mcall-prologues
10369 @opindex mcall-prologues
10370 Functions prologues/epilogues expanded as call to appropriate
10371 subroutines. Code size will be smaller.
10374 @opindex mtiny-stack
10375 Change only the low 8 bits of the stack pointer.
10379 Assume int to be 8 bit integer. This affects the sizes of all types: A
10380 char will be 1 byte, an int will be 1 byte, a long will be 2 bytes
10381 and long long will be 4 bytes. Please note that this option does not
10382 comply to the C standards, but it will provide you with smaller code
10386 @node Blackfin Options
10387 @subsection Blackfin Options
10388 @cindex Blackfin Options
10391 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
10393 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
10394 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
10395 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
10396 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
10397 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
10398 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
10399 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
10401 The optional @var{sirevision} specifies the silicon revision of the target
10402 Blackfin processor. Any workarounds available for the targeted silicon revision
10403 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
10404 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
10405 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
10406 hexadecimal digits representing the major and minor numbers in the silicon
10407 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
10408 is not defined. If @var{sirevision} is @samp{any}, the
10409 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
10410 If this optional @var{sirevision} is not used, GCC assumes the latest known
10411 silicon revision of the targeted Blackfin processor.
10413 Support for @samp{bf561} is incomplete. For @samp{bf561},
10414 Only the processor macro is defined.
10415 Without this option, @samp{bf532} is used as the processor by default.
10416 The corresponding predefined processor macros for @var{cpu} is to
10417 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
10418 provided by libgloss to be linked in if @option{-msim} is not given.
10422 Specifies that the program will be run on the simulator. This causes
10423 the simulator BSP provided by libgloss to be linked in. This option
10424 has effect only for @samp{bfin-elf} toolchain.
10425 Certain other options, such as @option{-mid-shared-library} and
10426 @option{-mfdpic}, imply @option{-msim}.
10428 @item -momit-leaf-frame-pointer
10429 @opindex momit-leaf-frame-pointer
10430 Don't keep the frame pointer in a register for leaf functions. This
10431 avoids the instructions to save, set up and restore frame pointers and
10432 makes an extra register available in leaf functions. The option
10433 @option{-fomit-frame-pointer} removes the frame pointer for all functions
10434 which might make debugging harder.
10436 @item -mspecld-anomaly
10437 @opindex mspecld-anomaly
10438 When enabled, the compiler will ensure that the generated code does not
10439 contain speculative loads after jump instructions. If this option is used,
10440 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
10442 @item -mno-specld-anomaly
10443 @opindex mno-specld-anomaly
10444 Don't generate extra code to prevent speculative loads from occurring.
10446 @item -mcsync-anomaly
10447 @opindex mcsync-anomaly
10448 When enabled, the compiler will ensure that the generated code does not
10449 contain CSYNC or SSYNC instructions too soon after conditional branches.
10450 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
10452 @item -mno-csync-anomaly
10453 @opindex mno-csync-anomaly
10454 Don't generate extra code to prevent CSYNC or SSYNC instructions from
10455 occurring too soon after a conditional branch.
10459 When enabled, the compiler is free to take advantage of the knowledge that
10460 the entire program fits into the low 64k of memory.
10463 @opindex mno-low-64k
10464 Assume that the program is arbitrarily large. This is the default.
10466 @item -mstack-check-l1
10467 @opindex mstack-check-l1
10468 Do stack checking using information placed into L1 scratchpad memory by the
10471 @item -mid-shared-library
10472 @opindex mid-shared-library
10473 Generate code that supports shared libraries via the library ID method.
10474 This allows for execute in place and shared libraries in an environment
10475 without virtual memory management. This option implies @option{-fPIC}.
10476 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10478 @item -mno-id-shared-library
10479 @opindex mno-id-shared-library
10480 Generate code that doesn't assume ID based shared libraries are being used.
10481 This is the default.
10483 @item -mleaf-id-shared-library
10484 @opindex mleaf-id-shared-library
10485 Generate code that supports shared libraries via the library ID method,
10486 but assumes that this library or executable won't link against any other
10487 ID shared libraries. That allows the compiler to use faster code for jumps
10490 @item -mno-leaf-id-shared-library
10491 @opindex mno-leaf-id-shared-library
10492 Do not assume that the code being compiled won't link against any ID shared
10493 libraries. Slower code will be generated for jump and call insns.
10495 @item -mshared-library-id=n
10496 @opindex mshared-library-id
10497 Specified the identification number of the ID based shared library being
10498 compiled. Specifying a value of 0 will generate more compact code, specifying
10499 other values will force the allocation of that number to the current
10500 library but is no more space or time efficient than omitting this option.
10504 Generate code that allows the data segment to be located in a different
10505 area of memory from the text segment. This allows for execute in place in
10506 an environment without virtual memory management by eliminating relocations
10507 against the text section.
10509 @item -mno-sep-data
10510 @opindex mno-sep-data
10511 Generate code that assumes that the data segment follows the text segment.
10512 This is the default.
10515 @itemx -mno-long-calls
10516 @opindex mlong-calls
10517 @opindex mno-long-calls
10518 Tells the compiler to perform function calls by first loading the
10519 address of the function into a register and then performing a subroutine
10520 call on this register. This switch is needed if the target function
10521 will lie outside of the 24 bit addressing range of the offset based
10522 version of subroutine call instruction.
10524 This feature is not enabled by default. Specifying
10525 @option{-mno-long-calls} will restore the default behavior. Note these
10526 switches have no effect on how the compiler generates code to handle
10527 function calls via function pointers.
10531 Link with the fast floating-point library. This library relaxes some of
10532 the IEEE floating-point standard's rules for checking inputs against
10533 Not-a-Number (NAN), in the interest of performance.
10536 @opindex minline-plt
10537 Enable inlining of PLT entries in function calls to functions that are
10538 not known to bind locally. It has no effect without @option{-mfdpic}.
10541 @opindex mmulticore
10542 Build standalone application for multicore Blackfin processor. Proper
10543 start files and link scripts will be used to support multicore.
10544 This option defines @code{__BFIN_MULTICORE}. It can only be used with
10545 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
10546 @option{-mcorea} or @option{-mcoreb}. If it's used without
10547 @option{-mcorea} or @option{-mcoreb}, single application/dual core
10548 programming model is used. In this model, the main function of Core B
10549 should be named as coreb_main. If it's used with @option{-mcorea} or
10550 @option{-mcoreb}, one application per core programming model is used.
10551 If this option is not used, single core application programming
10556 Build standalone application for Core A of BF561 when using
10557 one application per core programming model. Proper start files
10558 and link scripts will be used to support Core A. This option
10559 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
10563 Build standalone application for Core B of BF561 when using
10564 one application per core programming model. Proper start files
10565 and link scripts will be used to support Core B. This option
10566 defines @code{__BFIN_COREB}. When this option is used, coreb_main
10567 should be used instead of main. It must be used with
10568 @option{-mmulticore}.
10572 Build standalone application for SDRAM. Proper start files and
10573 link scripts will be used to put the application into SDRAM.
10574 Loader should initialize SDRAM before loading the application
10575 into SDRAM. This option defines @code{__BFIN_SDRAM}.
10579 Assume that ICPLBs are enabled at runtime. This has an effect on certain
10580 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
10581 are enabled; for standalone applications the default is off.
10585 @subsection CRIS Options
10586 @cindex CRIS Options
10588 These options are defined specifically for the CRIS ports.
10591 @item -march=@var{architecture-type}
10592 @itemx -mcpu=@var{architecture-type}
10595 Generate code for the specified architecture. The choices for
10596 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
10597 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
10598 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
10601 @item -mtune=@var{architecture-type}
10603 Tune to @var{architecture-type} everything applicable about the generated
10604 code, except for the ABI and the set of available instructions. The
10605 choices for @var{architecture-type} are the same as for
10606 @option{-march=@var{architecture-type}}.
10608 @item -mmax-stack-frame=@var{n}
10609 @opindex mmax-stack-frame
10610 Warn when the stack frame of a function exceeds @var{n} bytes.
10616 The options @option{-metrax4} and @option{-metrax100} are synonyms for
10617 @option{-march=v3} and @option{-march=v8} respectively.
10619 @item -mmul-bug-workaround
10620 @itemx -mno-mul-bug-workaround
10621 @opindex mmul-bug-workaround
10622 @opindex mno-mul-bug-workaround
10623 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
10624 models where it applies. This option is active by default.
10628 Enable CRIS-specific verbose debug-related information in the assembly
10629 code. This option also has the effect to turn off the @samp{#NO_APP}
10630 formatted-code indicator to the assembler at the beginning of the
10635 Do not use condition-code results from previous instruction; always emit
10636 compare and test instructions before use of condition codes.
10638 @item -mno-side-effects
10639 @opindex mno-side-effects
10640 Do not emit instructions with side-effects in addressing modes other than
10643 @item -mstack-align
10644 @itemx -mno-stack-align
10645 @itemx -mdata-align
10646 @itemx -mno-data-align
10647 @itemx -mconst-align
10648 @itemx -mno-const-align
10649 @opindex mstack-align
10650 @opindex mno-stack-align
10651 @opindex mdata-align
10652 @opindex mno-data-align
10653 @opindex mconst-align
10654 @opindex mno-const-align
10655 These options (no-options) arranges (eliminate arrangements) for the
10656 stack-frame, individual data and constants to be aligned for the maximum
10657 single data access size for the chosen CPU model. The default is to
10658 arrange for 32-bit alignment. ABI details such as structure layout are
10659 not affected by these options.
10667 Similar to the stack- data- and const-align options above, these options
10668 arrange for stack-frame, writable data and constants to all be 32-bit,
10669 16-bit or 8-bit aligned. The default is 32-bit alignment.
10671 @item -mno-prologue-epilogue
10672 @itemx -mprologue-epilogue
10673 @opindex mno-prologue-epilogue
10674 @opindex mprologue-epilogue
10675 With @option{-mno-prologue-epilogue}, the normal function prologue and
10676 epilogue that sets up the stack-frame are omitted and no return
10677 instructions or return sequences are generated in the code. Use this
10678 option only together with visual inspection of the compiled code: no
10679 warnings or errors are generated when call-saved registers must be saved,
10680 or storage for local variable needs to be allocated.
10684 @opindex mno-gotplt
10686 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
10687 instruction sequences that load addresses for functions from the PLT part
10688 of the GOT rather than (traditional on other architectures) calls to the
10689 PLT@. The default is @option{-mgotplt}.
10693 Legacy no-op option only recognized with the cris-axis-elf and
10694 cris-axis-linux-gnu targets.
10698 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
10702 This option, recognized for the cris-axis-elf arranges
10703 to link with input-output functions from a simulator library. Code,
10704 initialized data and zero-initialized data are allocated consecutively.
10708 Like @option{-sim}, but pass linker options to locate initialized data at
10709 0x40000000 and zero-initialized data at 0x80000000.
10713 @subsection CRX Options
10714 @cindex CRX Options
10716 These options are defined specifically for the CRX ports.
10722 Enable the use of multiply-accumulate instructions. Disabled by default.
10725 @opindex mpush-args
10726 Push instructions will be used to pass outgoing arguments when functions
10727 are called. Enabled by default.
10730 @node Darwin Options
10731 @subsection Darwin Options
10732 @cindex Darwin options
10734 These options are defined for all architectures running the Darwin operating
10737 FSF GCC on Darwin does not create ``fat'' object files; it will create
10738 an object file for the single architecture that it was built to
10739 target. Apple's GCC on Darwin does create ``fat'' files if multiple
10740 @option{-arch} options are used; it does so by running the compiler or
10741 linker multiple times and joining the results together with
10744 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
10745 @samp{i686}) is determined by the flags that specify the ISA
10746 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
10747 @option{-force_cpusubtype_ALL} option can be used to override this.
10749 The Darwin tools vary in their behavior when presented with an ISA
10750 mismatch. The assembler, @file{as}, will only permit instructions to
10751 be used that are valid for the subtype of the file it is generating,
10752 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
10753 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
10754 and print an error if asked to create a shared library with a less
10755 restrictive subtype than its input files (for instance, trying to put
10756 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
10757 for executables, @file{ld}, will quietly give the executable the most
10758 restrictive subtype of any of its input files.
10763 Add the framework directory @var{dir} to the head of the list of
10764 directories to be searched for header files. These directories are
10765 interleaved with those specified by @option{-I} options and are
10766 scanned in a left-to-right order.
10768 A framework directory is a directory with frameworks in it. A
10769 framework is a directory with a @samp{"Headers"} and/or
10770 @samp{"PrivateHeaders"} directory contained directly in it that ends
10771 in @samp{".framework"}. The name of a framework is the name of this
10772 directory excluding the @samp{".framework"}. Headers associated with
10773 the framework are found in one of those two directories, with
10774 @samp{"Headers"} being searched first. A subframework is a framework
10775 directory that is in a framework's @samp{"Frameworks"} directory.
10776 Includes of subframework headers can only appear in a header of a
10777 framework that contains the subframework, or in a sibling subframework
10778 header. Two subframeworks are siblings if they occur in the same
10779 framework. A subframework should not have the same name as a
10780 framework, a warning will be issued if this is violated. Currently a
10781 subframework cannot have subframeworks, in the future, the mechanism
10782 may be extended to support this. The standard frameworks can be found
10783 in @samp{"/System/Library/Frameworks"} and
10784 @samp{"/Library/Frameworks"}. An example include looks like
10785 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
10786 the name of the framework and header.h is found in the
10787 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
10789 @item -iframework@var{dir}
10790 @opindex iframework
10791 Like @option{-F} except the directory is a treated as a system
10792 directory. The main difference between this @option{-iframework} and
10793 @option{-F} is that with @option{-iframework} the compiler does not
10794 warn about constructs contained within header files found via
10795 @var{dir}. This option is valid only for the C family of languages.
10799 Emit debugging information for symbols that are used. For STABS
10800 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
10801 This is by default ON@.
10805 Emit debugging information for all symbols and types.
10807 @item -mmacosx-version-min=@var{version}
10808 The earliest version of MacOS X that this executable will run on
10809 is @var{version}. Typical values of @var{version} include @code{10.1},
10810 @code{10.2}, and @code{10.3.9}.
10812 If the compiler was built to use the system's headers by default,
10813 then the default for this option is the system version on which the
10814 compiler is running, otherwise the default is to make choices which
10815 are compatible with as many systems and code bases as possible.
10819 Enable kernel development mode. The @option{-mkernel} option sets
10820 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
10821 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
10822 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
10823 applicable. This mode also sets @option{-mno-altivec},
10824 @option{-msoft-float}, @option{-fno-builtin} and
10825 @option{-mlong-branch} for PowerPC targets.
10827 @item -mone-byte-bool
10828 @opindex mone-byte-bool
10829 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
10830 By default @samp{sizeof(bool)} is @samp{4} when compiling for
10831 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
10832 option has no effect on x86.
10834 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
10835 to generate code that is not binary compatible with code generated
10836 without that switch. Using this switch may require recompiling all
10837 other modules in a program, including system libraries. Use this
10838 switch to conform to a non-default data model.
10840 @item -mfix-and-continue
10841 @itemx -ffix-and-continue
10842 @itemx -findirect-data
10843 @opindex mfix-and-continue
10844 @opindex ffix-and-continue
10845 @opindex findirect-data
10846 Generate code suitable for fast turn around development. Needed to
10847 enable gdb to dynamically load @code{.o} files into already running
10848 programs. @option{-findirect-data} and @option{-ffix-and-continue}
10849 are provided for backwards compatibility.
10853 Loads all members of static archive libraries.
10854 See man ld(1) for more information.
10856 @item -arch_errors_fatal
10857 @opindex arch_errors_fatal
10858 Cause the errors having to do with files that have the wrong architecture
10861 @item -bind_at_load
10862 @opindex bind_at_load
10863 Causes the output file to be marked such that the dynamic linker will
10864 bind all undefined references when the file is loaded or launched.
10868 Produce a Mach-o bundle format file.
10869 See man ld(1) for more information.
10871 @item -bundle_loader @var{executable}
10872 @opindex bundle_loader
10873 This option specifies the @var{executable} that will be loading the build
10874 output file being linked. See man ld(1) for more information.
10877 @opindex dynamiclib
10878 When passed this option, GCC will produce a dynamic library instead of
10879 an executable when linking, using the Darwin @file{libtool} command.
10881 @item -force_cpusubtype_ALL
10882 @opindex force_cpusubtype_ALL
10883 This causes GCC's output file to have the @var{ALL} subtype, instead of
10884 one controlled by the @option{-mcpu} or @option{-march} option.
10886 @item -allowable_client @var{client_name}
10887 @itemx -client_name
10888 @itemx -compatibility_version
10889 @itemx -current_version
10891 @itemx -dependency-file
10893 @itemx -dylinker_install_name
10895 @itemx -exported_symbols_list
10898 @itemx -flat_namespace
10899 @itemx -force_flat_namespace
10900 @itemx -headerpad_max_install_names
10903 @itemx -install_name
10904 @itemx -keep_private_externs
10905 @itemx -multi_module
10906 @itemx -multiply_defined
10907 @itemx -multiply_defined_unused
10910 @itemx -no_dead_strip_inits_and_terms
10911 @itemx -nofixprebinding
10912 @itemx -nomultidefs
10914 @itemx -noseglinkedit
10915 @itemx -pagezero_size
10917 @itemx -prebind_all_twolevel_modules
10918 @itemx -private_bundle
10920 @itemx -read_only_relocs
10922 @itemx -sectobjectsymbols
10926 @itemx -sectobjectsymbols
10929 @itemx -segs_read_only_addr
10931 @itemx -segs_read_write_addr
10932 @itemx -seg_addr_table
10933 @itemx -seg_addr_table_filename
10934 @itemx -seglinkedit
10936 @itemx -segs_read_only_addr
10937 @itemx -segs_read_write_addr
10938 @itemx -single_module
10940 @itemx -sub_library
10942 @itemx -sub_umbrella
10943 @itemx -twolevel_namespace
10946 @itemx -unexported_symbols_list
10947 @itemx -weak_reference_mismatches
10948 @itemx -whatsloaded
10949 @opindex allowable_client
10950 @opindex client_name
10951 @opindex compatibility_version
10952 @opindex current_version
10953 @opindex dead_strip
10954 @opindex dependency-file
10955 @opindex dylib_file
10956 @opindex dylinker_install_name
10958 @opindex exported_symbols_list
10960 @opindex flat_namespace
10961 @opindex force_flat_namespace
10962 @opindex headerpad_max_install_names
10963 @opindex image_base
10965 @opindex install_name
10966 @opindex keep_private_externs
10967 @opindex multi_module
10968 @opindex multiply_defined
10969 @opindex multiply_defined_unused
10970 @opindex noall_load
10971 @opindex no_dead_strip_inits_and_terms
10972 @opindex nofixprebinding
10973 @opindex nomultidefs
10975 @opindex noseglinkedit
10976 @opindex pagezero_size
10978 @opindex prebind_all_twolevel_modules
10979 @opindex private_bundle
10980 @opindex read_only_relocs
10982 @opindex sectobjectsymbols
10985 @opindex sectcreate
10986 @opindex sectobjectsymbols
10989 @opindex segs_read_only_addr
10990 @opindex segs_read_write_addr
10991 @opindex seg_addr_table
10992 @opindex seg_addr_table_filename
10993 @opindex seglinkedit
10995 @opindex segs_read_only_addr
10996 @opindex segs_read_write_addr
10997 @opindex single_module
10999 @opindex sub_library
11000 @opindex sub_umbrella
11001 @opindex twolevel_namespace
11004 @opindex unexported_symbols_list
11005 @opindex weak_reference_mismatches
11006 @opindex whatsloaded
11007 These options are passed to the Darwin linker. The Darwin linker man page
11008 describes them in detail.
11011 @node DEC Alpha Options
11012 @subsection DEC Alpha Options
11014 These @samp{-m} options are defined for the DEC Alpha implementations:
11017 @item -mno-soft-float
11018 @itemx -msoft-float
11019 @opindex mno-soft-float
11020 @opindex msoft-float
11021 Use (do not use) the hardware floating-point instructions for
11022 floating-point operations. When @option{-msoft-float} is specified,
11023 functions in @file{libgcc.a} will be used to perform floating-point
11024 operations. Unless they are replaced by routines that emulate the
11025 floating-point operations, or compiled in such a way as to call such
11026 emulations routines, these routines will issue floating-point
11027 operations. If you are compiling for an Alpha without floating-point
11028 operations, you must ensure that the library is built so as not to call
11031 Note that Alpha implementations without floating-point operations are
11032 required to have floating-point registers.
11035 @itemx -mno-fp-regs
11037 @opindex mno-fp-regs
11038 Generate code that uses (does not use) the floating-point register set.
11039 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
11040 register set is not used, floating point operands are passed in integer
11041 registers as if they were integers and floating-point results are passed
11042 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
11043 so any function with a floating-point argument or return value called by code
11044 compiled with @option{-mno-fp-regs} must also be compiled with that
11047 A typical use of this option is building a kernel that does not use,
11048 and hence need not save and restore, any floating-point registers.
11052 The Alpha architecture implements floating-point hardware optimized for
11053 maximum performance. It is mostly compliant with the IEEE floating
11054 point standard. However, for full compliance, software assistance is
11055 required. This option generates code fully IEEE compliant code
11056 @emph{except} that the @var{inexact-flag} is not maintained (see below).
11057 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
11058 defined during compilation. The resulting code is less efficient but is
11059 able to correctly support denormalized numbers and exceptional IEEE
11060 values such as not-a-number and plus/minus infinity. Other Alpha
11061 compilers call this option @option{-ieee_with_no_inexact}.
11063 @item -mieee-with-inexact
11064 @opindex mieee-with-inexact
11065 This is like @option{-mieee} except the generated code also maintains
11066 the IEEE @var{inexact-flag}. Turning on this option causes the
11067 generated code to implement fully-compliant IEEE math. In addition to
11068 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
11069 macro. On some Alpha implementations the resulting code may execute
11070 significantly slower than the code generated by default. Since there is
11071 very little code that depends on the @var{inexact-flag}, you should
11072 normally not specify this option. Other Alpha compilers call this
11073 option @option{-ieee_with_inexact}.
11075 @item -mfp-trap-mode=@var{trap-mode}
11076 @opindex mfp-trap-mode
11077 This option controls what floating-point related traps are enabled.
11078 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
11079 The trap mode can be set to one of four values:
11083 This is the default (normal) setting. The only traps that are enabled
11084 are the ones that cannot be disabled in software (e.g., division by zero
11088 In addition to the traps enabled by @samp{n}, underflow traps are enabled
11092 Like @samp{u}, but the instructions are marked to be safe for software
11093 completion (see Alpha architecture manual for details).
11096 Like @samp{su}, but inexact traps are enabled as well.
11099 @item -mfp-rounding-mode=@var{rounding-mode}
11100 @opindex mfp-rounding-mode
11101 Selects the IEEE rounding mode. Other Alpha compilers call this option
11102 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
11107 Normal IEEE rounding mode. Floating point numbers are rounded towards
11108 the nearest machine number or towards the even machine number in case
11112 Round towards minus infinity.
11115 Chopped rounding mode. Floating point numbers are rounded towards zero.
11118 Dynamic rounding mode. A field in the floating point control register
11119 (@var{fpcr}, see Alpha architecture reference manual) controls the
11120 rounding mode in effect. The C library initializes this register for
11121 rounding towards plus infinity. Thus, unless your program modifies the
11122 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
11125 @item -mtrap-precision=@var{trap-precision}
11126 @opindex mtrap-precision
11127 In the Alpha architecture, floating point traps are imprecise. This
11128 means without software assistance it is impossible to recover from a
11129 floating trap and program execution normally needs to be terminated.
11130 GCC can generate code that can assist operating system trap handlers
11131 in determining the exact location that caused a floating point trap.
11132 Depending on the requirements of an application, different levels of
11133 precisions can be selected:
11137 Program precision. This option is the default and means a trap handler
11138 can only identify which program caused a floating point exception.
11141 Function precision. The trap handler can determine the function that
11142 caused a floating point exception.
11145 Instruction precision. The trap handler can determine the exact
11146 instruction that caused a floating point exception.
11149 Other Alpha compilers provide the equivalent options called
11150 @option{-scope_safe} and @option{-resumption_safe}.
11152 @item -mieee-conformant
11153 @opindex mieee-conformant
11154 This option marks the generated code as IEEE conformant. You must not
11155 use this option unless you also specify @option{-mtrap-precision=i} and either
11156 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
11157 is to emit the line @samp{.eflag 48} in the function prologue of the
11158 generated assembly file. Under DEC Unix, this has the effect that
11159 IEEE-conformant math library routines will be linked in.
11161 @item -mbuild-constants
11162 @opindex mbuild-constants
11163 Normally GCC examines a 32- or 64-bit integer constant to
11164 see if it can construct it from smaller constants in two or three
11165 instructions. If it cannot, it will output the constant as a literal and
11166 generate code to load it from the data segment at runtime.
11168 Use this option to require GCC to construct @emph{all} integer constants
11169 using code, even if it takes more instructions (the maximum is six).
11171 You would typically use this option to build a shared library dynamic
11172 loader. Itself a shared library, it must relocate itself in memory
11173 before it can find the variables and constants in its own data segment.
11179 Select whether to generate code to be assembled by the vendor-supplied
11180 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
11198 Indicate whether GCC should generate code to use the optional BWX,
11199 CIX, FIX and MAX instruction sets. The default is to use the instruction
11200 sets supported by the CPU type specified via @option{-mcpu=} option or that
11201 of the CPU on which GCC was built if none was specified.
11204 @itemx -mfloat-ieee
11205 @opindex mfloat-vax
11206 @opindex mfloat-ieee
11207 Generate code that uses (does not use) VAX F and G floating point
11208 arithmetic instead of IEEE single and double precision.
11210 @item -mexplicit-relocs
11211 @itemx -mno-explicit-relocs
11212 @opindex mexplicit-relocs
11213 @opindex mno-explicit-relocs
11214 Older Alpha assemblers provided no way to generate symbol relocations
11215 except via assembler macros. Use of these macros does not allow
11216 optimal instruction scheduling. GNU binutils as of version 2.12
11217 supports a new syntax that allows the compiler to explicitly mark
11218 which relocations should apply to which instructions. This option
11219 is mostly useful for debugging, as GCC detects the capabilities of
11220 the assembler when it is built and sets the default accordingly.
11223 @itemx -mlarge-data
11224 @opindex msmall-data
11225 @opindex mlarge-data
11226 When @option{-mexplicit-relocs} is in effect, static data is
11227 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
11228 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
11229 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
11230 16-bit relocations off of the @code{$gp} register. This limits the
11231 size of the small data area to 64KB, but allows the variables to be
11232 directly accessed via a single instruction.
11234 The default is @option{-mlarge-data}. With this option the data area
11235 is limited to just below 2GB@. Programs that require more than 2GB of
11236 data must use @code{malloc} or @code{mmap} to allocate the data in the
11237 heap instead of in the program's data segment.
11239 When generating code for shared libraries, @option{-fpic} implies
11240 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
11243 @itemx -mlarge-text
11244 @opindex msmall-text
11245 @opindex mlarge-text
11246 When @option{-msmall-text} is used, the compiler assumes that the
11247 code of the entire program (or shared library) fits in 4MB, and is
11248 thus reachable with a branch instruction. When @option{-msmall-data}
11249 is used, the compiler can assume that all local symbols share the
11250 same @code{$gp} value, and thus reduce the number of instructions
11251 required for a function call from 4 to 1.
11253 The default is @option{-mlarge-text}.
11255 @item -mcpu=@var{cpu_type}
11257 Set the instruction set and instruction scheduling parameters for
11258 machine type @var{cpu_type}. You can specify either the @samp{EV}
11259 style name or the corresponding chip number. GCC supports scheduling
11260 parameters for the EV4, EV5 and EV6 family of processors and will
11261 choose the default values for the instruction set from the processor
11262 you specify. If you do not specify a processor type, GCC will default
11263 to the processor on which the compiler was built.
11265 Supported values for @var{cpu_type} are
11271 Schedules as an EV4 and has no instruction set extensions.
11275 Schedules as an EV5 and has no instruction set extensions.
11279 Schedules as an EV5 and supports the BWX extension.
11284 Schedules as an EV5 and supports the BWX and MAX extensions.
11288 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
11292 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
11295 Native Linux/GNU toolchains also support the value @samp{native},
11296 which selects the best architecture option for the host processor.
11297 @option{-mcpu=native} has no effect if GCC does not recognize
11300 @item -mtune=@var{cpu_type}
11302 Set only the instruction scheduling parameters for machine type
11303 @var{cpu_type}. The instruction set is not changed.
11305 Native Linux/GNU toolchains also support the value @samp{native},
11306 which selects the best architecture option for the host processor.
11307 @option{-mtune=native} has no effect if GCC does not recognize
11310 @item -mmemory-latency=@var{time}
11311 @opindex mmemory-latency
11312 Sets the latency the scheduler should assume for typical memory
11313 references as seen by the application. This number is highly
11314 dependent on the memory access patterns used by the application
11315 and the size of the external cache on the machine.
11317 Valid options for @var{time} are
11321 A decimal number representing clock cycles.
11327 The compiler contains estimates of the number of clock cycles for
11328 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
11329 (also called Dcache, Scache, and Bcache), as well as to main memory.
11330 Note that L3 is only valid for EV5.
11335 @node DEC Alpha/VMS Options
11336 @subsection DEC Alpha/VMS Options
11338 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
11341 @item -mvms-return-codes
11342 @opindex mvms-return-codes
11343 Return VMS condition codes from main. The default is to return POSIX
11344 style condition (e.g.@: error) codes.
11346 @item -mdebug-main=@var{prefix}
11347 @opindex mdebug-main=@var{prefix}
11348 Flag the first routine whose name starts with @var{prefix} as the main
11349 routine for the debugger.
11353 Default to 64bit memory allocation routines.
11357 @subsection FR30 Options
11358 @cindex FR30 Options
11360 These options are defined specifically for the FR30 port.
11364 @item -msmall-model
11365 @opindex msmall-model
11366 Use the small address space model. This can produce smaller code, but
11367 it does assume that all symbolic values and addresses will fit into a
11372 Assume that run-time support has been provided and so there is no need
11373 to include the simulator library (@file{libsim.a}) on the linker
11379 @subsection FRV Options
11380 @cindex FRV Options
11386 Only use the first 32 general purpose registers.
11391 Use all 64 general purpose registers.
11396 Use only the first 32 floating point registers.
11401 Use all 64 floating point registers
11404 @opindex mhard-float
11406 Use hardware instructions for floating point operations.
11409 @opindex msoft-float
11411 Use library routines for floating point operations.
11416 Dynamically allocate condition code registers.
11421 Do not try to dynamically allocate condition code registers, only
11422 use @code{icc0} and @code{fcc0}.
11427 Change ABI to use double word insns.
11432 Do not use double word instructions.
11437 Use floating point double instructions.
11440 @opindex mno-double
11442 Do not use floating point double instructions.
11447 Use media instructions.
11452 Do not use media instructions.
11457 Use multiply and add/subtract instructions.
11460 @opindex mno-muladd
11462 Do not use multiply and add/subtract instructions.
11467 Select the FDPIC ABI, that uses function descriptors to represent
11468 pointers to functions. Without any PIC/PIE-related options, it
11469 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
11470 assumes GOT entries and small data are within a 12-bit range from the
11471 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
11472 are computed with 32 bits.
11473 With a @samp{bfin-elf} target, this option implies @option{-msim}.
11476 @opindex minline-plt
11478 Enable inlining of PLT entries in function calls to functions that are
11479 not known to bind locally. It has no effect without @option{-mfdpic}.
11480 It's enabled by default if optimizing for speed and compiling for
11481 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
11482 optimization option such as @option{-O3} or above is present in the
11488 Assume a large TLS segment when generating thread-local code.
11493 Do not assume a large TLS segment when generating thread-local code.
11498 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
11499 that is known to be in read-only sections. It's enabled by default,
11500 except for @option{-fpic} or @option{-fpie}: even though it may help
11501 make the global offset table smaller, it trades 1 instruction for 4.
11502 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
11503 one of which may be shared by multiple symbols, and it avoids the need
11504 for a GOT entry for the referenced symbol, so it's more likely to be a
11505 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
11507 @item -multilib-library-pic
11508 @opindex multilib-library-pic
11510 Link with the (library, not FD) pic libraries. It's implied by
11511 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
11512 @option{-fpic} without @option{-mfdpic}. You should never have to use
11516 @opindex mlinked-fp
11518 Follow the EABI requirement of always creating a frame pointer whenever
11519 a stack frame is allocated. This option is enabled by default and can
11520 be disabled with @option{-mno-linked-fp}.
11523 @opindex mlong-calls
11525 Use indirect addressing to call functions outside the current
11526 compilation unit. This allows the functions to be placed anywhere
11527 within the 32-bit address space.
11529 @item -malign-labels
11530 @opindex malign-labels
11532 Try to align labels to an 8-byte boundary by inserting nops into the
11533 previous packet. This option only has an effect when VLIW packing
11534 is enabled. It doesn't create new packets; it merely adds nops to
11537 @item -mlibrary-pic
11538 @opindex mlibrary-pic
11540 Generate position-independent EABI code.
11545 Use only the first four media accumulator registers.
11550 Use all eight media accumulator registers.
11555 Pack VLIW instructions.
11560 Do not pack VLIW instructions.
11563 @opindex mno-eflags
11565 Do not mark ABI switches in e_flags.
11568 @opindex mcond-move
11570 Enable the use of conditional-move instructions (default).
11572 This switch is mainly for debugging the compiler and will likely be removed
11573 in a future version.
11575 @item -mno-cond-move
11576 @opindex mno-cond-move
11578 Disable the use of conditional-move instructions.
11580 This switch is mainly for debugging the compiler and will likely be removed
11581 in a future version.
11586 Enable the use of conditional set instructions (default).
11588 This switch is mainly for debugging the compiler and will likely be removed
11589 in a future version.
11594 Disable the use of conditional set instructions.
11596 This switch is mainly for debugging the compiler and will likely be removed
11597 in a future version.
11600 @opindex mcond-exec
11602 Enable the use of conditional execution (default).
11604 This switch is mainly for debugging the compiler and will likely be removed
11605 in a future version.
11607 @item -mno-cond-exec
11608 @opindex mno-cond-exec
11610 Disable the use of conditional execution.
11612 This switch is mainly for debugging the compiler and will likely be removed
11613 in a future version.
11615 @item -mvliw-branch
11616 @opindex mvliw-branch
11618 Run a pass to pack branches into VLIW instructions (default).
11620 This switch is mainly for debugging the compiler and will likely be removed
11621 in a future version.
11623 @item -mno-vliw-branch
11624 @opindex mno-vliw-branch
11626 Do not run a pass to pack branches into VLIW instructions.
11628 This switch is mainly for debugging the compiler and will likely be removed
11629 in a future version.
11631 @item -mmulti-cond-exec
11632 @opindex mmulti-cond-exec
11634 Enable optimization of @code{&&} and @code{||} in conditional execution
11637 This switch is mainly for debugging the compiler and will likely be removed
11638 in a future version.
11640 @item -mno-multi-cond-exec
11641 @opindex mno-multi-cond-exec
11643 Disable optimization of @code{&&} and @code{||} in conditional execution.
11645 This switch is mainly for debugging the compiler and will likely be removed
11646 in a future version.
11648 @item -mnested-cond-exec
11649 @opindex mnested-cond-exec
11651 Enable nested conditional execution optimizations (default).
11653 This switch is mainly for debugging the compiler and will likely be removed
11654 in a future version.
11656 @item -mno-nested-cond-exec
11657 @opindex mno-nested-cond-exec
11659 Disable nested conditional execution optimizations.
11661 This switch is mainly for debugging the compiler and will likely be removed
11662 in a future version.
11664 @item -moptimize-membar
11665 @opindex moptimize-membar
11667 This switch removes redundant @code{membar} instructions from the
11668 compiler generated code. It is enabled by default.
11670 @item -mno-optimize-membar
11671 @opindex mno-optimize-membar
11673 This switch disables the automatic removal of redundant @code{membar}
11674 instructions from the generated code.
11676 @item -mtomcat-stats
11677 @opindex mtomcat-stats
11679 Cause gas to print out tomcat statistics.
11681 @item -mcpu=@var{cpu}
11684 Select the processor type for which to generate code. Possible values are
11685 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
11686 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
11690 @node GNU/Linux Options
11691 @subsection GNU/Linux Options
11693 These @samp{-m} options are defined for GNU/Linux targets:
11698 Use the GNU C library. This is the default except
11699 on @samp{*-*-linux-*uclibc*} and @samp{*-*-linux-*android*} targets.
11703 Use uClibc C library. This is the default on
11704 @samp{*-*-linux-*uclibc*} targets.
11708 Use Bionic C library. This is the default on
11709 @samp{*-*-linux-*android*} targets.
11713 Compile code compatible with Android platform. This is the default on
11714 @samp{*-*-linux-*android*} targets.
11716 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
11717 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
11718 this option makes the GCC driver pass Android-specific options to the linker.
11719 Finally, this option causes the preprocessor macro @code{__ANDROID__}
11722 @item -tno-android-cc
11723 @opindex tno-android-cc
11724 Disable compilation effects of @option{-mandroid}, i.e., do not enable
11725 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
11726 @option{-fno-rtti} by default.
11728 @item -tno-android-ld
11729 @opindex tno-android-ld
11730 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
11731 linking options to the linker.
11735 @node H8/300 Options
11736 @subsection H8/300 Options
11738 These @samp{-m} options are defined for the H8/300 implementations:
11743 Shorten some address references at link time, when possible; uses the
11744 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
11745 ld, Using ld}, for a fuller description.
11749 Generate code for the H8/300H@.
11753 Generate code for the H8S@.
11757 Generate code for the H8S and H8/300H in the normal mode. This switch
11758 must be used either with @option{-mh} or @option{-ms}.
11762 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
11766 Make @code{int} data 32 bits by default.
11769 @opindex malign-300
11770 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
11771 The default for the H8/300H and H8S is to align longs and floats on 4
11773 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
11774 This option has no effect on the H8/300.
11778 @subsection HPPA Options
11779 @cindex HPPA Options
11781 These @samp{-m} options are defined for the HPPA family of computers:
11784 @item -march=@var{architecture-type}
11786 Generate code for the specified architecture. The choices for
11787 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
11788 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
11789 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
11790 architecture option for your machine. Code compiled for lower numbered
11791 architectures will run on higher numbered architectures, but not the
11794 @item -mpa-risc-1-0
11795 @itemx -mpa-risc-1-1
11796 @itemx -mpa-risc-2-0
11797 @opindex mpa-risc-1-0
11798 @opindex mpa-risc-1-1
11799 @opindex mpa-risc-2-0
11800 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
11803 @opindex mbig-switch
11804 Generate code suitable for big switch tables. Use this option only if
11805 the assembler/linker complain about out of range branches within a switch
11808 @item -mjump-in-delay
11809 @opindex mjump-in-delay
11810 Fill delay slots of function calls with unconditional jump instructions
11811 by modifying the return pointer for the function call to be the target
11812 of the conditional jump.
11814 @item -mdisable-fpregs
11815 @opindex mdisable-fpregs
11816 Prevent floating point registers from being used in any manner. This is
11817 necessary for compiling kernels which perform lazy context switching of
11818 floating point registers. If you use this option and attempt to perform
11819 floating point operations, the compiler will abort.
11821 @item -mdisable-indexing
11822 @opindex mdisable-indexing
11823 Prevent the compiler from using indexing address modes. This avoids some
11824 rather obscure problems when compiling MIG generated code under MACH@.
11826 @item -mno-space-regs
11827 @opindex mno-space-regs
11828 Generate code that assumes the target has no space registers. This allows
11829 GCC to generate faster indirect calls and use unscaled index address modes.
11831 Such code is suitable for level 0 PA systems and kernels.
11833 @item -mfast-indirect-calls
11834 @opindex mfast-indirect-calls
11835 Generate code that assumes calls never cross space boundaries. This
11836 allows GCC to emit code which performs faster indirect calls.
11838 This option will not work in the presence of shared libraries or nested
11841 @item -mfixed-range=@var{register-range}
11842 @opindex mfixed-range
11843 Generate code treating the given register range as fixed registers.
11844 A fixed register is one that the register allocator can not use. This is
11845 useful when compiling kernel code. A register range is specified as
11846 two registers separated by a dash. Multiple register ranges can be
11847 specified separated by a comma.
11849 @item -mlong-load-store
11850 @opindex mlong-load-store
11851 Generate 3-instruction load and store sequences as sometimes required by
11852 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
11855 @item -mportable-runtime
11856 @opindex mportable-runtime
11857 Use the portable calling conventions proposed by HP for ELF systems.
11861 Enable the use of assembler directives only GAS understands.
11863 @item -mschedule=@var{cpu-type}
11865 Schedule code according to the constraints for the machine type
11866 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
11867 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
11868 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
11869 proper scheduling option for your machine. The default scheduling is
11873 @opindex mlinker-opt
11874 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
11875 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
11876 linkers in which they give bogus error messages when linking some programs.
11879 @opindex msoft-float
11880 Generate output containing library calls for floating point.
11881 @strong{Warning:} the requisite libraries are not available for all HPPA
11882 targets. Normally the facilities of the machine's usual C compiler are
11883 used, but this cannot be done directly in cross-compilation. You must make
11884 your own arrangements to provide suitable library functions for
11887 @option{-msoft-float} changes the calling convention in the output file;
11888 therefore, it is only useful if you compile @emph{all} of a program with
11889 this option. In particular, you need to compile @file{libgcc.a}, the
11890 library that comes with GCC, with @option{-msoft-float} in order for
11895 Generate the predefine, @code{_SIO}, for server IO@. The default is
11896 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
11897 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
11898 options are available under HP-UX and HI-UX@.
11902 Use GNU ld specific options. This passes @option{-shared} to ld when
11903 building a shared library. It is the default when GCC is configured,
11904 explicitly or implicitly, with the GNU linker. This option does not
11905 have any affect on which ld is called, it only changes what parameters
11906 are passed to that ld. The ld that is called is determined by the
11907 @option{--with-ld} configure option, GCC's program search path, and
11908 finally by the user's @env{PATH}. The linker used by GCC can be printed
11909 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
11910 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11914 Use HP ld specific options. This passes @option{-b} to ld when building
11915 a shared library and passes @option{+Accept TypeMismatch} to ld on all
11916 links. It is the default when GCC is configured, explicitly or
11917 implicitly, with the HP linker. This option does not have any affect on
11918 which ld is called, it only changes what parameters are passed to that
11919 ld. The ld that is called is determined by the @option{--with-ld}
11920 configure option, GCC's program search path, and finally by the user's
11921 @env{PATH}. The linker used by GCC can be printed using @samp{which
11922 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
11923 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11926 @opindex mno-long-calls
11927 Generate code that uses long call sequences. This ensures that a call
11928 is always able to reach linker generated stubs. The default is to generate
11929 long calls only when the distance from the call site to the beginning
11930 of the function or translation unit, as the case may be, exceeds a
11931 predefined limit set by the branch type being used. The limits for
11932 normal calls are 7,600,000 and 240,000 bytes, respectively for the
11933 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
11936 Distances are measured from the beginning of functions when using the
11937 @option{-ffunction-sections} option, or when using the @option{-mgas}
11938 and @option{-mno-portable-runtime} options together under HP-UX with
11941 It is normally not desirable to use this option as it will degrade
11942 performance. However, it may be useful in large applications,
11943 particularly when partial linking is used to build the application.
11945 The types of long calls used depends on the capabilities of the
11946 assembler and linker, and the type of code being generated. The
11947 impact on systems that support long absolute calls, and long pic
11948 symbol-difference or pc-relative calls should be relatively small.
11949 However, an indirect call is used on 32-bit ELF systems in pic code
11950 and it is quite long.
11952 @item -munix=@var{unix-std}
11954 Generate compiler predefines and select a startfile for the specified
11955 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
11956 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
11957 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
11958 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
11959 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
11962 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
11963 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
11964 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
11965 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
11966 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
11967 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
11969 It is @emph{important} to note that this option changes the interfaces
11970 for various library routines. It also affects the operational behavior
11971 of the C library. Thus, @emph{extreme} care is needed in using this
11974 Library code that is intended to operate with more than one UNIX
11975 standard must test, set and restore the variable @var{__xpg4_extended_mask}
11976 as appropriate. Most GNU software doesn't provide this capability.
11980 Suppress the generation of link options to search libdld.sl when the
11981 @option{-static} option is specified on HP-UX 10 and later.
11985 The HP-UX implementation of setlocale in libc has a dependency on
11986 libdld.sl. There isn't an archive version of libdld.sl. Thus,
11987 when the @option{-static} option is specified, special link options
11988 are needed to resolve this dependency.
11990 On HP-UX 10 and later, the GCC driver adds the necessary options to
11991 link with libdld.sl when the @option{-static} option is specified.
11992 This causes the resulting binary to be dynamic. On the 64-bit port,
11993 the linkers generate dynamic binaries by default in any case. The
11994 @option{-nolibdld} option can be used to prevent the GCC driver from
11995 adding these link options.
11999 Add support for multithreading with the @dfn{dce thread} library
12000 under HP-UX@. This option sets flags for both the preprocessor and
12004 @node i386 and x86-64 Options
12005 @subsection Intel 386 and AMD x86-64 Options
12006 @cindex i386 Options
12007 @cindex x86-64 Options
12008 @cindex Intel 386 Options
12009 @cindex AMD x86-64 Options
12011 These @samp{-m} options are defined for the i386 and x86-64 family of
12015 @item -mtune=@var{cpu-type}
12017 Tune to @var{cpu-type} everything applicable about the generated code, except
12018 for the ABI and the set of available instructions. The choices for
12019 @var{cpu-type} are:
12022 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
12023 If you know the CPU on which your code will run, then you should use
12024 the corresponding @option{-mtune} option instead of
12025 @option{-mtune=generic}. But, if you do not know exactly what CPU users
12026 of your application will have, then you should use this option.
12028 As new processors are deployed in the marketplace, the behavior of this
12029 option will change. Therefore, if you upgrade to a newer version of
12030 GCC, the code generated option will change to reflect the processors
12031 that were most common when that version of GCC was released.
12033 There is no @option{-march=generic} option because @option{-march}
12034 indicates the instruction set the compiler can use, and there is no
12035 generic instruction set applicable to all processors. In contrast,
12036 @option{-mtune} indicates the processor (or, in this case, collection of
12037 processors) for which the code is optimized.
12039 This selects the CPU to tune for at compilation time by determining
12040 the processor type of the compiling machine. Using @option{-mtune=native}
12041 will produce code optimized for the local machine under the constraints
12042 of the selected instruction set. Using @option{-march=native} will
12043 enable all instruction subsets supported by the local machine (hence
12044 the result might not run on different machines).
12046 Original Intel's i386 CPU@.
12048 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
12049 @item i586, pentium
12050 Intel Pentium CPU with no MMX support.
12052 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
12054 Intel PentiumPro CPU@.
12056 Same as @code{generic}, but when used as @code{march} option, PentiumPro
12057 instruction set will be used, so the code will run on all i686 family chips.
12059 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
12060 @item pentium3, pentium3m
12061 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
12064 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
12065 support. Used by Centrino notebooks.
12066 @item pentium4, pentium4m
12067 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
12069 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
12072 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
12073 SSE2 and SSE3 instruction set support.
12075 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
12076 instruction set support.
12078 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1
12079 and SSE4.2 instruction set support.
12081 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
12082 instruction set support.
12084 AMD K6 CPU with MMX instruction set support.
12086 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
12087 @item athlon, athlon-tbird
12088 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
12090 @item athlon-4, athlon-xp, athlon-mp
12091 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
12092 instruction set support.
12093 @item k8, opteron, athlon64, athlon-fx
12094 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
12095 MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit instruction set extensions.)
12096 @item k8-sse3, opteron-sse3, athlon64-sse3
12097 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
12098 @item amdfam10, barcelona
12099 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
12100 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
12101 instruction set extensions.)
12103 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
12106 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
12107 instruction set support.
12109 Via C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
12110 implemented for this chip.)
12112 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
12113 implemented for this chip.)
12115 Embedded AMD CPU with MMX and 3DNow!@: instruction set support.
12118 While picking a specific @var{cpu-type} will schedule things appropriately
12119 for that particular chip, the compiler will not generate any code that
12120 does not run on the i386 without the @option{-march=@var{cpu-type}} option
12123 @item -march=@var{cpu-type}
12125 Generate instructions for the machine type @var{cpu-type}. The choices
12126 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
12127 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
12129 @item -mcpu=@var{cpu-type}
12131 A deprecated synonym for @option{-mtune}.
12133 @item -mfpmath=@var{unit}
12135 Generate floating point arithmetics for selected unit @var{unit}. The choices
12136 for @var{unit} are:
12140 Use the standard 387 floating point coprocessor present majority of chips and
12141 emulated otherwise. Code compiled with this option will run almost everywhere.
12142 The temporary results are computed in 80bit precision instead of precision
12143 specified by the type resulting in slightly different results compared to most
12144 of other chips. See @option{-ffloat-store} for more detailed description.
12146 This is the default choice for i386 compiler.
12149 Use scalar floating point instructions present in the SSE instruction set.
12150 This instruction set is supported by Pentium3 and newer chips, in the AMD line
12151 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
12152 instruction set supports only single precision arithmetics, thus the double and
12153 extended precision arithmetics is still done using 387. Later version, present
12154 only in Pentium4 and the future AMD x86-64 chips supports double precision
12157 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
12158 or @option{-msse2} switches to enable SSE extensions and make this option
12159 effective. For the x86-64 compiler, these extensions are enabled by default.
12161 The resulting code should be considerably faster in the majority of cases and avoid
12162 the numerical instability problems of 387 code, but may break some existing
12163 code that expects temporaries to be 80bit.
12165 This is the default choice for the x86-64 compiler.
12170 Attempt to utilize both instruction sets at once. This effectively double the
12171 amount of available registers and on chips with separate execution units for
12172 387 and SSE the execution resources too. Use this option with care, as it is
12173 still experimental, because the GCC register allocator does not model separate
12174 functional units well resulting in instable performance.
12177 @item -masm=@var{dialect}
12178 @opindex masm=@var{dialect}
12179 Output asm instructions using selected @var{dialect}. Supported
12180 choices are @samp{intel} or @samp{att} (the default one). Darwin does
12181 not support @samp{intel}.
12184 @itemx -mno-ieee-fp
12186 @opindex mno-ieee-fp
12187 Control whether or not the compiler uses IEEE floating point
12188 comparisons. These handle correctly the case where the result of a
12189 comparison is unordered.
12192 @opindex msoft-float
12193 Generate output containing library calls for floating point.
12194 @strong{Warning:} the requisite libraries are not part of GCC@.
12195 Normally the facilities of the machine's usual C compiler are used, but
12196 this can't be done directly in cross-compilation. You must make your
12197 own arrangements to provide suitable library functions for
12200 On machines where a function returns floating point results in the 80387
12201 register stack, some floating point opcodes may be emitted even if
12202 @option{-msoft-float} is used.
12204 @item -mno-fp-ret-in-387
12205 @opindex mno-fp-ret-in-387
12206 Do not use the FPU registers for return values of functions.
12208 The usual calling convention has functions return values of types
12209 @code{float} and @code{double} in an FPU register, even if there
12210 is no FPU@. The idea is that the operating system should emulate
12213 The option @option{-mno-fp-ret-in-387} causes such values to be returned
12214 in ordinary CPU registers instead.
12216 @item -mno-fancy-math-387
12217 @opindex mno-fancy-math-387
12218 Some 387 emulators do not support the @code{sin}, @code{cos} and
12219 @code{sqrt} instructions for the 387. Specify this option to avoid
12220 generating those instructions. This option is the default on FreeBSD,
12221 OpenBSD and NetBSD@. This option is overridden when @option{-march}
12222 indicates that the target cpu will always have an FPU and so the
12223 instruction will not need emulation. As of revision 2.6.1, these
12224 instructions are not generated unless you also use the
12225 @option{-funsafe-math-optimizations} switch.
12227 @item -malign-double
12228 @itemx -mno-align-double
12229 @opindex malign-double
12230 @opindex mno-align-double
12231 Control whether GCC aligns @code{double}, @code{long double}, and
12232 @code{long long} variables on a two word boundary or a one word
12233 boundary. Aligning @code{double} variables on a two word boundary will
12234 produce code that runs somewhat faster on a @samp{Pentium} at the
12235 expense of more memory.
12237 On x86-64, @option{-malign-double} is enabled by default.
12239 @strong{Warning:} if you use the @option{-malign-double} switch,
12240 structures containing the above types will be aligned differently than
12241 the published application binary interface specifications for the 386
12242 and will not be binary compatible with structures in code compiled
12243 without that switch.
12245 @item -m96bit-long-double
12246 @itemx -m128bit-long-double
12247 @opindex m96bit-long-double
12248 @opindex m128bit-long-double
12249 These switches control the size of @code{long double} type. The i386
12250 application binary interface specifies the size to be 96 bits,
12251 so @option{-m96bit-long-double} is the default in 32 bit mode.
12253 Modern architectures (Pentium and newer) would prefer @code{long double}
12254 to be aligned to an 8 or 16 byte boundary. In arrays or structures
12255 conforming to the ABI, this would not be possible. So specifying a
12256 @option{-m128bit-long-double} will align @code{long double}
12257 to a 16 byte boundary by padding the @code{long double} with an additional
12260 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
12261 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
12263 Notice that neither of these options enable any extra precision over the x87
12264 standard of 80 bits for a @code{long double}.
12266 @strong{Warning:} if you override the default value for your target ABI, the
12267 structures and arrays containing @code{long double} variables will change
12268 their size as well as function calling convention for function taking
12269 @code{long double} will be modified. Hence they will not be binary
12270 compatible with arrays or structures in code compiled without that switch.
12272 @item -mlarge-data-threshold=@var{number}
12273 @opindex mlarge-data-threshold=@var{number}
12274 When @option{-mcmodel=medium} is specified, the data greater than
12275 @var{threshold} are placed in large data section. This value must be the
12276 same across all object linked into the binary and defaults to 65535.
12280 Use a different function-calling convention, in which functions that
12281 take a fixed number of arguments return with the @code{ret} @var{num}
12282 instruction, which pops their arguments while returning. This saves one
12283 instruction in the caller since there is no need to pop the arguments
12286 You can specify that an individual function is called with this calling
12287 sequence with the function attribute @samp{stdcall}. You can also
12288 override the @option{-mrtd} option by using the function attribute
12289 @samp{cdecl}. @xref{Function Attributes}.
12291 @strong{Warning:} this calling convention is incompatible with the one
12292 normally used on Unix, so you cannot use it if you need to call
12293 libraries compiled with the Unix compiler.
12295 Also, you must provide function prototypes for all functions that
12296 take variable numbers of arguments (including @code{printf});
12297 otherwise incorrect code will be generated for calls to those
12300 In addition, seriously incorrect code will result if you call a
12301 function with too many arguments. (Normally, extra arguments are
12302 harmlessly ignored.)
12304 @item -mregparm=@var{num}
12306 Control how many registers are used to pass integer arguments. By
12307 default, no registers are used to pass arguments, and at most 3
12308 registers can be used. You can control this behavior for a specific
12309 function by using the function attribute @samp{regparm}.
12310 @xref{Function Attributes}.
12312 @strong{Warning:} if you use this switch, and
12313 @var{num} is nonzero, then you must build all modules with the same
12314 value, including any libraries. This includes the system libraries and
12318 @opindex msseregparm
12319 Use SSE register passing conventions for float and double arguments
12320 and return values. You can control this behavior for a specific
12321 function by using the function attribute @samp{sseregparm}.
12322 @xref{Function Attributes}.
12324 @strong{Warning:} if you use this switch then you must build all
12325 modules with the same value, including any libraries. This includes
12326 the system libraries and startup modules.
12328 @item -mvect8-ret-in-mem
12329 @opindex mvect8-ret-in-mem
12330 Return 8-byte vectors in memory instead of MMX registers. This is the
12331 default on Solaris~8 and 9 and VxWorks to match the ABI of the Sun
12332 Studio compilers until version 12. Later compiler versions (starting
12333 with Studio 12 Update~1) follow the ABI used by other x86 targets, which
12334 is the default on Solaris~10 and later. @emph{Only} use this option if
12335 you need to remain compatible with existing code produced by those
12336 previous compiler versions or older versions of GCC.
12345 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
12346 is specified, the significands of results of floating-point operations are
12347 rounded to 24 bits (single precision); @option{-mpc64} rounds the
12348 significands of results of floating-point operations to 53 bits (double
12349 precision) and @option{-mpc80} rounds the significands of results of
12350 floating-point operations to 64 bits (extended double precision), which is
12351 the default. When this option is used, floating-point operations in higher
12352 precisions are not available to the programmer without setting the FPU
12353 control word explicitly.
12355 Setting the rounding of floating-point operations to less than the default
12356 80 bits can speed some programs by 2% or more. Note that some mathematical
12357 libraries assume that extended precision (80 bit) floating-point operations
12358 are enabled by default; routines in such libraries could suffer significant
12359 loss of accuracy, typically through so-called "catastrophic cancellation",
12360 when this option is used to set the precision to less than extended precision.
12362 @item -mstackrealign
12363 @opindex mstackrealign
12364 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
12365 option will generate an alternate prologue and epilogue that realigns the
12366 runtime stack if necessary. This supports mixing legacy codes that keep
12367 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
12368 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
12369 applicable to individual functions.
12371 @item -mpreferred-stack-boundary=@var{num}
12372 @opindex mpreferred-stack-boundary
12373 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
12374 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
12375 the default is 4 (16 bytes or 128 bits).
12377 @item -mincoming-stack-boundary=@var{num}
12378 @opindex mincoming-stack-boundary
12379 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
12380 boundary. If @option{-mincoming-stack-boundary} is not specified,
12381 the one specified by @option{-mpreferred-stack-boundary} will be used.
12383 On Pentium and PentiumPro, @code{double} and @code{long double} values
12384 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
12385 suffer significant run time performance penalties. On Pentium III, the
12386 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
12387 properly if it is not 16 byte aligned.
12389 To ensure proper alignment of this values on the stack, the stack boundary
12390 must be as aligned as that required by any value stored on the stack.
12391 Further, every function must be generated such that it keeps the stack
12392 aligned. Thus calling a function compiled with a higher preferred
12393 stack boundary from a function compiled with a lower preferred stack
12394 boundary will most likely misalign the stack. It is recommended that
12395 libraries that use callbacks always use the default setting.
12397 This extra alignment does consume extra stack space, and generally
12398 increases code size. Code that is sensitive to stack space usage, such
12399 as embedded systems and operating system kernels, may want to reduce the
12400 preferred alignment to @option{-mpreferred-stack-boundary=2}.
12427 @itemx -mno-fsgsbase
12457 These switches enable or disable the use of instructions in the MMX,
12458 SSE, SSE2, SSE3, SSSE3, SSE4.1, AVX, AES, PCLMUL, FSGSBASE, RDRND,
12459 F16C, SSE4A, FMA4, XOP, LWP, ABM, BMI, or 3DNow!@: extended instruction sets.
12460 These extensions are also available as built-in functions: see
12461 @ref{X86 Built-in Functions}, for details of the functions enabled and
12462 disabled by these switches.
12464 To have SSE/SSE2 instructions generated automatically from floating-point
12465 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
12467 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
12468 generates new AVX instructions or AVX equivalence for all SSEx instructions
12471 These options will enable GCC to use these extended instructions in
12472 generated code, even without @option{-mfpmath=sse}. Applications which
12473 perform runtime CPU detection must compile separate files for each
12474 supported architecture, using the appropriate flags. In particular,
12475 the file containing the CPU detection code should be compiled without
12479 @itemx -mno-fused-madd
12480 @opindex mfused-madd
12481 @opindex mno-fused-madd
12482 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
12483 instructions. The default is to use these instructions.
12487 This option instructs GCC to emit a @code{cld} instruction in the prologue
12488 of functions that use string instructions. String instructions depend on
12489 the DF flag to select between autoincrement or autodecrement mode. While the
12490 ABI specifies the DF flag to be cleared on function entry, some operating
12491 systems violate this specification by not clearing the DF flag in their
12492 exception dispatchers. The exception handler can be invoked with the DF flag
12493 set which leads to wrong direction mode, when string instructions are used.
12494 This option can be enabled by default on 32-bit x86 targets by configuring
12495 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
12496 instructions can be suppressed with the @option{-mno-cld} compiler option
12500 @opindex mvzeroupper
12501 This option instructs GCC to emit a @code{vzeroupper} instruction
12502 before a transfer of control flow out of the function to minimize
12503 AVX to SSE transition penalty as well as remove unnecessary zeroupper
12508 This option will enable GCC to use CMPXCHG16B instruction in generated code.
12509 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
12510 data types. This is useful for high resolution counters that could be updated
12511 by multiple processors (or cores). This instruction is generated as part of
12512 atomic built-in functions: see @ref{Atomic Builtins} for details.
12516 This option will enable GCC to use SAHF instruction in generated 64-bit code.
12517 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
12518 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
12519 SAHF are load and store instructions, respectively, for certain status flags.
12520 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
12521 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
12525 This option will enable GCC to use movbe instruction to implement
12526 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
12530 This option will enable built-in functions, @code{__builtin_ia32_crc32qi},
12531 @code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and
12532 @code{__builtin_ia32_crc32di} to generate the crc32 machine instruction.
12536 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
12537 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
12538 to increase precision instead of DIVSS and SQRTSS (and their vectorized
12539 variants) for single precision floating point arguments. These instructions
12540 are generated only when @option{-funsafe-math-optimizations} is enabled
12541 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
12542 Note that while the throughput of the sequence is higher than the throughput
12543 of the non-reciprocal instruction, the precision of the sequence can be
12544 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
12546 Note that GCC implements 1.0f/sqrtf(x) in terms of RSQRTSS (or RSQRTPS)
12547 already with @option{-ffast-math} (or the above option combination), and
12548 doesn't need @option{-mrecip}.
12550 @item -mveclibabi=@var{type}
12551 @opindex mveclibabi
12552 Specifies the ABI type to use for vectorizing intrinsics using an
12553 external library. Supported types are @code{svml} for the Intel short
12554 vector math library and @code{acml} for the AMD math core library style
12555 of interfacing. GCC will currently emit calls to @code{vmldExp2},
12556 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
12557 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
12558 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
12559 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
12560 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
12561 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
12562 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
12563 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
12564 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
12565 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
12566 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
12567 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
12568 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
12569 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
12570 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
12571 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
12572 compatible library will have to be specified at link time.
12574 @item -mabi=@var{name}
12576 Generate code for the specified calling convention. Permissible values
12577 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
12578 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
12579 ABI when targeting Windows. On all other systems, the default is the
12580 SYSV ABI. You can control this behavior for a specific function by
12581 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
12582 @xref{Function Attributes}.
12585 @itemx -mno-push-args
12586 @opindex mpush-args
12587 @opindex mno-push-args
12588 Use PUSH operations to store outgoing parameters. This method is shorter
12589 and usually equally fast as method using SUB/MOV operations and is enabled
12590 by default. In some cases disabling it may improve performance because of
12591 improved scheduling and reduced dependencies.
12593 @item -maccumulate-outgoing-args
12594 @opindex maccumulate-outgoing-args
12595 If enabled, the maximum amount of space required for outgoing arguments will be
12596 computed in the function prologue. This is faster on most modern CPUs
12597 because of reduced dependencies, improved scheduling and reduced stack usage
12598 when preferred stack boundary is not equal to 2. The drawback is a notable
12599 increase in code size. This switch implies @option{-mno-push-args}.
12603 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
12604 on thread-safe exception handling must compile and link all code with the
12605 @option{-mthreads} option. When compiling, @option{-mthreads} defines
12606 @option{-D_MT}; when linking, it links in a special thread helper library
12607 @option{-lmingwthrd} which cleans up per thread exception handling data.
12609 @item -mno-align-stringops
12610 @opindex mno-align-stringops
12611 Do not align destination of inlined string operations. This switch reduces
12612 code size and improves performance in case the destination is already aligned,
12613 but GCC doesn't know about it.
12615 @item -minline-all-stringops
12616 @opindex minline-all-stringops
12617 By default GCC inlines string operations only when destination is known to be
12618 aligned at least to 4 byte boundary. This enables more inlining, increase code
12619 size, but may improve performance of code that depends on fast memcpy, strlen
12620 and memset for short lengths.
12622 @item -minline-stringops-dynamically
12623 @opindex minline-stringops-dynamically
12624 For string operation of unknown size, inline runtime checks so for small
12625 blocks inline code is used, while for large blocks library call is used.
12627 @item -mstringop-strategy=@var{alg}
12628 @opindex mstringop-strategy=@var{alg}
12629 Overwrite internal decision heuristic about particular algorithm to inline
12630 string operation with. The allowed values are @code{rep_byte},
12631 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
12632 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
12633 expanding inline loop, @code{libcall} for always expanding library call.
12635 @item -momit-leaf-frame-pointer
12636 @opindex momit-leaf-frame-pointer
12637 Don't keep the frame pointer in a register for leaf functions. This
12638 avoids the instructions to save, set up and restore frame pointers and
12639 makes an extra register available in leaf functions. The option
12640 @option{-fomit-frame-pointer} removes the frame pointer for all functions
12641 which might make debugging harder.
12643 @item -mtls-direct-seg-refs
12644 @itemx -mno-tls-direct-seg-refs
12645 @opindex mtls-direct-seg-refs
12646 Controls whether TLS variables may be accessed with offsets from the
12647 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
12648 or whether the thread base pointer must be added. Whether or not this
12649 is legal depends on the operating system, and whether it maps the
12650 segment to cover the entire TLS area.
12652 For systems that use GNU libc, the default is on.
12655 @itemx -mno-sse2avx
12657 Specify that the assembler should encode SSE instructions with VEX
12658 prefix. The option @option{-mavx} turns this on by default.
12663 If profiling is active @option{-pg} put the profiling
12664 counter call before prologue.
12665 Note: On x86 architectures the attribute @code{ms_hook_prologue}
12666 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
12669 @itemx -mno-8bit-idiv
12671 On some processors, like Intel Atom, 8bit unsigned integer divide is
12672 much faster than 32bit/64bit integer divide. This option will generate a
12673 runt-time check. If both dividend and divisor are within range of 0
12674 to 255, 8bit unsigned integer divide will be used instead of
12675 32bit/64bit integer divide.
12679 These @samp{-m} switches are supported in addition to the above
12680 on AMD x86-64 processors in 64-bit environments.
12687 Generate code for a 32-bit or 64-bit environment.
12688 The 32-bit environment sets int, long and pointer to 32 bits and
12689 generates code that runs on any i386 system.
12690 The 64-bit environment sets int to 32 bits and long and pointer
12691 to 64 bits and generates code for AMD's x86-64 architecture. For
12692 darwin only the -m64 option turns off the @option{-fno-pic} and
12693 @option{-mdynamic-no-pic} options.
12695 @item -mno-red-zone
12696 @opindex mno-red-zone
12697 Do not use a so called red zone for x86-64 code. The red zone is mandated
12698 by the x86-64 ABI, it is a 128-byte area beyond the location of the
12699 stack pointer that will not be modified by signal or interrupt handlers
12700 and therefore can be used for temporary data without adjusting the stack
12701 pointer. The flag @option{-mno-red-zone} disables this red zone.
12703 @item -mcmodel=small
12704 @opindex mcmodel=small
12705 Generate code for the small code model: the program and its symbols must
12706 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
12707 Programs can be statically or dynamically linked. This is the default
12710 @item -mcmodel=kernel
12711 @opindex mcmodel=kernel
12712 Generate code for the kernel code model. The kernel runs in the
12713 negative 2 GB of the address space.
12714 This model has to be used for Linux kernel code.
12716 @item -mcmodel=medium
12717 @opindex mcmodel=medium
12718 Generate code for the medium model: The program is linked in the lower 2
12719 GB of the address space. Small symbols are also placed there. Symbols
12720 with sizes larger than @option{-mlarge-data-threshold} are put into
12721 large data or bss sections and can be located above 2GB. Programs can
12722 be statically or dynamically linked.
12724 @item -mcmodel=large
12725 @opindex mcmodel=large
12726 Generate code for the large model: This model makes no assumptions
12727 about addresses and sizes of sections.
12730 @node IA-64 Options
12731 @subsection IA-64 Options
12732 @cindex IA-64 Options
12734 These are the @samp{-m} options defined for the Intel IA-64 architecture.
12738 @opindex mbig-endian
12739 Generate code for a big endian target. This is the default for HP-UX@.
12741 @item -mlittle-endian
12742 @opindex mlittle-endian
12743 Generate code for a little endian target. This is the default for AIX5
12749 @opindex mno-gnu-as
12750 Generate (or don't) code for the GNU assembler. This is the default.
12751 @c Also, this is the default if the configure option @option{--with-gnu-as}
12757 @opindex mno-gnu-ld
12758 Generate (or don't) code for the GNU linker. This is the default.
12759 @c Also, this is the default if the configure option @option{--with-gnu-ld}
12764 Generate code that does not use a global pointer register. The result
12765 is not position independent code, and violates the IA-64 ABI@.
12767 @item -mvolatile-asm-stop
12768 @itemx -mno-volatile-asm-stop
12769 @opindex mvolatile-asm-stop
12770 @opindex mno-volatile-asm-stop
12771 Generate (or don't) a stop bit immediately before and after volatile asm
12774 @item -mregister-names
12775 @itemx -mno-register-names
12776 @opindex mregister-names
12777 @opindex mno-register-names
12778 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
12779 the stacked registers. This may make assembler output more readable.
12785 Disable (or enable) optimizations that use the small data section. This may
12786 be useful for working around optimizer bugs.
12788 @item -mconstant-gp
12789 @opindex mconstant-gp
12790 Generate code that uses a single constant global pointer value. This is
12791 useful when compiling kernel code.
12795 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
12796 This is useful when compiling firmware code.
12798 @item -minline-float-divide-min-latency
12799 @opindex minline-float-divide-min-latency
12800 Generate code for inline divides of floating point values
12801 using the minimum latency algorithm.
12803 @item -minline-float-divide-max-throughput
12804 @opindex minline-float-divide-max-throughput
12805 Generate code for inline divides of floating point values
12806 using the maximum throughput algorithm.
12808 @item -mno-inline-float-divide
12809 @opindex mno-inline-float-divide
12810 Do not generate inline code for divides of floating point values.
12812 @item -minline-int-divide-min-latency
12813 @opindex minline-int-divide-min-latency
12814 Generate code for inline divides of integer values
12815 using the minimum latency algorithm.
12817 @item -minline-int-divide-max-throughput
12818 @opindex minline-int-divide-max-throughput
12819 Generate code for inline divides of integer values
12820 using the maximum throughput algorithm.
12822 @item -mno-inline-int-divide
12823 @opindex mno-inline-int-divide
12824 Do not generate inline code for divides of integer values.
12826 @item -minline-sqrt-min-latency
12827 @opindex minline-sqrt-min-latency
12828 Generate code for inline square roots
12829 using the minimum latency algorithm.
12831 @item -minline-sqrt-max-throughput
12832 @opindex minline-sqrt-max-throughput
12833 Generate code for inline square roots
12834 using the maximum throughput algorithm.
12836 @item -mno-inline-sqrt
12837 @opindex mno-inline-sqrt
12838 Do not generate inline code for sqrt.
12841 @itemx -mno-fused-madd
12842 @opindex mfused-madd
12843 @opindex mno-fused-madd
12844 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
12845 instructions. The default is to use these instructions.
12847 @item -mno-dwarf2-asm
12848 @itemx -mdwarf2-asm
12849 @opindex mno-dwarf2-asm
12850 @opindex mdwarf2-asm
12851 Don't (or do) generate assembler code for the DWARF2 line number debugging
12852 info. This may be useful when not using the GNU assembler.
12854 @item -mearly-stop-bits
12855 @itemx -mno-early-stop-bits
12856 @opindex mearly-stop-bits
12857 @opindex mno-early-stop-bits
12858 Allow stop bits to be placed earlier than immediately preceding the
12859 instruction that triggered the stop bit. This can improve instruction
12860 scheduling, but does not always do so.
12862 @item -mfixed-range=@var{register-range}
12863 @opindex mfixed-range
12864 Generate code treating the given register range as fixed registers.
12865 A fixed register is one that the register allocator can not use. This is
12866 useful when compiling kernel code. A register range is specified as
12867 two registers separated by a dash. Multiple register ranges can be
12868 specified separated by a comma.
12870 @item -mtls-size=@var{tls-size}
12872 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
12875 @item -mtune=@var{cpu-type}
12877 Tune the instruction scheduling for a particular CPU, Valid values are
12878 itanium, itanium1, merced, itanium2, and mckinley.
12884 Generate code for a 32-bit or 64-bit environment.
12885 The 32-bit environment sets int, long and pointer to 32 bits.
12886 The 64-bit environment sets int to 32 bits and long and pointer
12887 to 64 bits. These are HP-UX specific flags.
12889 @item -mno-sched-br-data-spec
12890 @itemx -msched-br-data-spec
12891 @opindex mno-sched-br-data-spec
12892 @opindex msched-br-data-spec
12893 (Dis/En)able data speculative scheduling before reload.
12894 This will result in generation of the ld.a instructions and
12895 the corresponding check instructions (ld.c / chk.a).
12896 The default is 'disable'.
12898 @item -msched-ar-data-spec
12899 @itemx -mno-sched-ar-data-spec
12900 @opindex msched-ar-data-spec
12901 @opindex mno-sched-ar-data-spec
12902 (En/Dis)able data speculative scheduling after reload.
12903 This will result in generation of the ld.a instructions and
12904 the corresponding check instructions (ld.c / chk.a).
12905 The default is 'enable'.
12907 @item -mno-sched-control-spec
12908 @itemx -msched-control-spec
12909 @opindex mno-sched-control-spec
12910 @opindex msched-control-spec
12911 (Dis/En)able control speculative scheduling. This feature is
12912 available only during region scheduling (i.e.@: before reload).
12913 This will result in generation of the ld.s instructions and
12914 the corresponding check instructions chk.s .
12915 The default is 'disable'.
12917 @item -msched-br-in-data-spec
12918 @itemx -mno-sched-br-in-data-spec
12919 @opindex msched-br-in-data-spec
12920 @opindex mno-sched-br-in-data-spec
12921 (En/Dis)able speculative scheduling of the instructions that
12922 are dependent on the data speculative loads before reload.
12923 This is effective only with @option{-msched-br-data-spec} enabled.
12924 The default is 'enable'.
12926 @item -msched-ar-in-data-spec
12927 @itemx -mno-sched-ar-in-data-spec
12928 @opindex msched-ar-in-data-spec
12929 @opindex mno-sched-ar-in-data-spec
12930 (En/Dis)able speculative scheduling of the instructions that
12931 are dependent on the data speculative loads after reload.
12932 This is effective only with @option{-msched-ar-data-spec} enabled.
12933 The default is 'enable'.
12935 @item -msched-in-control-spec
12936 @itemx -mno-sched-in-control-spec
12937 @opindex msched-in-control-spec
12938 @opindex mno-sched-in-control-spec
12939 (En/Dis)able speculative scheduling of the instructions that
12940 are dependent on the control speculative loads.
12941 This is effective only with @option{-msched-control-spec} enabled.
12942 The default is 'enable'.
12944 @item -mno-sched-prefer-non-data-spec-insns
12945 @itemx -msched-prefer-non-data-spec-insns
12946 @opindex mno-sched-prefer-non-data-spec-insns
12947 @opindex msched-prefer-non-data-spec-insns
12948 If enabled, data speculative instructions will be chosen for schedule
12949 only if there are no other choices at the moment. This will make
12950 the use of the data speculation much more conservative.
12951 The default is 'disable'.
12953 @item -mno-sched-prefer-non-control-spec-insns
12954 @itemx -msched-prefer-non-control-spec-insns
12955 @opindex mno-sched-prefer-non-control-spec-insns
12956 @opindex msched-prefer-non-control-spec-insns
12957 If enabled, control speculative instructions will be chosen for schedule
12958 only if there are no other choices at the moment. This will make
12959 the use of the control speculation much more conservative.
12960 The default is 'disable'.
12962 @item -mno-sched-count-spec-in-critical-path
12963 @itemx -msched-count-spec-in-critical-path
12964 @opindex mno-sched-count-spec-in-critical-path
12965 @opindex msched-count-spec-in-critical-path
12966 If enabled, speculative dependencies will be considered during
12967 computation of the instructions priorities. This will make the use of the
12968 speculation a bit more conservative.
12969 The default is 'disable'.
12971 @item -msched-spec-ldc
12972 @opindex msched-spec-ldc
12973 Use a simple data speculation check. This option is on by default.
12975 @item -msched-control-spec-ldc
12976 @opindex msched-spec-ldc
12977 Use a simple check for control speculation. This option is on by default.
12979 @item -msched-stop-bits-after-every-cycle
12980 @opindex msched-stop-bits-after-every-cycle
12981 Place a stop bit after every cycle when scheduling. This option is on
12984 @item -msched-fp-mem-deps-zero-cost
12985 @opindex msched-fp-mem-deps-zero-cost
12986 Assume that floating-point stores and loads are not likely to cause a conflict
12987 when placed into the same instruction group. This option is disabled by
12990 @item -msel-sched-dont-check-control-spec
12991 @opindex msel-sched-dont-check-control-spec
12992 Generate checks for control speculation in selective scheduling.
12993 This flag is disabled by default.
12995 @item -msched-max-memory-insns=@var{max-insns}
12996 @opindex msched-max-memory-insns
12997 Limit on the number of memory insns per instruction group, giving lower
12998 priority to subsequent memory insns attempting to schedule in the same
12999 instruction group. Frequently useful to prevent cache bank conflicts.
13000 The default value is 1.
13002 @item -msched-max-memory-insns-hard-limit
13003 @opindex msched-max-memory-insns-hard-limit
13004 Disallow more than `msched-max-memory-insns' in instruction group.
13005 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
13006 when limit is reached but may still schedule memory operations.
13010 @node IA-64/VMS Options
13011 @subsection IA-64/VMS Options
13013 These @samp{-m} options are defined for the IA-64/VMS implementations:
13016 @item -mvms-return-codes
13017 @opindex mvms-return-codes
13018 Return VMS condition codes from main. The default is to return POSIX
13019 style condition (e.g.@ error) codes.
13021 @item -mdebug-main=@var{prefix}
13022 @opindex mdebug-main=@var{prefix}
13023 Flag the first routine whose name starts with @var{prefix} as the main
13024 routine for the debugger.
13028 Default to 64bit memory allocation routines.
13032 @subsection LM32 Options
13033 @cindex LM32 options
13035 These @option{-m} options are defined for the Lattice Mico32 architecture:
13038 @item -mbarrel-shift-enabled
13039 @opindex mbarrel-shift-enabled
13040 Enable barrel-shift instructions.
13042 @item -mdivide-enabled
13043 @opindex mdivide-enabled
13044 Enable divide and modulus instructions.
13046 @item -mmultiply-enabled
13047 @opindex multiply-enabled
13048 Enable multiply instructions.
13050 @item -msign-extend-enabled
13051 @opindex msign-extend-enabled
13052 Enable sign extend instructions.
13054 @item -muser-enabled
13055 @opindex muser-enabled
13056 Enable user-defined instructions.
13061 @subsection M32C Options
13062 @cindex M32C options
13065 @item -mcpu=@var{name}
13067 Select the CPU for which code is generated. @var{name} may be one of
13068 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
13069 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
13070 the M32C/80 series.
13074 Specifies that the program will be run on the simulator. This causes
13075 an alternate runtime library to be linked in which supports, for
13076 example, file I/O@. You must not use this option when generating
13077 programs that will run on real hardware; you must provide your own
13078 runtime library for whatever I/O functions are needed.
13080 @item -memregs=@var{number}
13082 Specifies the number of memory-based pseudo-registers GCC will use
13083 during code generation. These pseudo-registers will be used like real
13084 registers, so there is a tradeoff between GCC's ability to fit the
13085 code into available registers, and the performance penalty of using
13086 memory instead of registers. Note that all modules in a program must
13087 be compiled with the same value for this option. Because of that, you
13088 must not use this option with the default runtime libraries gcc
13093 @node M32R/D Options
13094 @subsection M32R/D Options
13095 @cindex M32R/D options
13097 These @option{-m} options are defined for Renesas M32R/D architectures:
13102 Generate code for the M32R/2@.
13106 Generate code for the M32R/X@.
13110 Generate code for the M32R@. This is the default.
13112 @item -mmodel=small
13113 @opindex mmodel=small
13114 Assume all objects live in the lower 16MB of memory (so that their addresses
13115 can be loaded with the @code{ld24} instruction), and assume all subroutines
13116 are reachable with the @code{bl} instruction.
13117 This is the default.
13119 The addressability of a particular object can be set with the
13120 @code{model} attribute.
13122 @item -mmodel=medium
13123 @opindex mmodel=medium
13124 Assume objects may be anywhere in the 32-bit address space (the compiler
13125 will generate @code{seth/add3} instructions to load their addresses), and
13126 assume all subroutines are reachable with the @code{bl} instruction.
13128 @item -mmodel=large
13129 @opindex mmodel=large
13130 Assume objects may be anywhere in the 32-bit address space (the compiler
13131 will generate @code{seth/add3} instructions to load their addresses), and
13132 assume subroutines may not be reachable with the @code{bl} instruction
13133 (the compiler will generate the much slower @code{seth/add3/jl}
13134 instruction sequence).
13137 @opindex msdata=none
13138 Disable use of the small data area. Variables will be put into
13139 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
13140 @code{section} attribute has been specified).
13141 This is the default.
13143 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
13144 Objects may be explicitly put in the small data area with the
13145 @code{section} attribute using one of these sections.
13147 @item -msdata=sdata
13148 @opindex msdata=sdata
13149 Put small global and static data in the small data area, but do not
13150 generate special code to reference them.
13153 @opindex msdata=use
13154 Put small global and static data in the small data area, and generate
13155 special instructions to reference them.
13159 @cindex smaller data references
13160 Put global and static objects less than or equal to @var{num} bytes
13161 into the small data or bss sections instead of the normal data or bss
13162 sections. The default value of @var{num} is 8.
13163 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
13164 for this option to have any effect.
13166 All modules should be compiled with the same @option{-G @var{num}} value.
13167 Compiling with different values of @var{num} may or may not work; if it
13168 doesn't the linker will give an error message---incorrect code will not be
13173 Makes the M32R specific code in the compiler display some statistics
13174 that might help in debugging programs.
13176 @item -malign-loops
13177 @opindex malign-loops
13178 Align all loops to a 32-byte boundary.
13180 @item -mno-align-loops
13181 @opindex mno-align-loops
13182 Do not enforce a 32-byte alignment for loops. This is the default.
13184 @item -missue-rate=@var{number}
13185 @opindex missue-rate=@var{number}
13186 Issue @var{number} instructions per cycle. @var{number} can only be 1
13189 @item -mbranch-cost=@var{number}
13190 @opindex mbranch-cost=@var{number}
13191 @var{number} can only be 1 or 2. If it is 1 then branches will be
13192 preferred over conditional code, if it is 2, then the opposite will
13195 @item -mflush-trap=@var{number}
13196 @opindex mflush-trap=@var{number}
13197 Specifies the trap number to use to flush the cache. The default is
13198 12. Valid numbers are between 0 and 15 inclusive.
13200 @item -mno-flush-trap
13201 @opindex mno-flush-trap
13202 Specifies that the cache cannot be flushed by using a trap.
13204 @item -mflush-func=@var{name}
13205 @opindex mflush-func=@var{name}
13206 Specifies the name of the operating system function to call to flush
13207 the cache. The default is @emph{_flush_cache}, but a function call
13208 will only be used if a trap is not available.
13210 @item -mno-flush-func
13211 @opindex mno-flush-func
13212 Indicates that there is no OS function for flushing the cache.
13216 @node M680x0 Options
13217 @subsection M680x0 Options
13218 @cindex M680x0 options
13220 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
13221 The default settings depend on which architecture was selected when
13222 the compiler was configured; the defaults for the most common choices
13226 @item -march=@var{arch}
13228 Generate code for a specific M680x0 or ColdFire instruction set
13229 architecture. Permissible values of @var{arch} for M680x0
13230 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
13231 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
13232 architectures are selected according to Freescale's ISA classification
13233 and the permissible values are: @samp{isaa}, @samp{isaaplus},
13234 @samp{isab} and @samp{isac}.
13236 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
13237 code for a ColdFire target. The @var{arch} in this macro is one of the
13238 @option{-march} arguments given above.
13240 When used together, @option{-march} and @option{-mtune} select code
13241 that runs on a family of similar processors but that is optimized
13242 for a particular microarchitecture.
13244 @item -mcpu=@var{cpu}
13246 Generate code for a specific M680x0 or ColdFire processor.
13247 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
13248 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
13249 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
13250 below, which also classifies the CPUs into families:
13252 @multitable @columnfractions 0.20 0.80
13253 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
13254 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
13255 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
13256 @item @samp{5206e} @tab @samp{5206e}
13257 @item @samp{5208} @tab @samp{5207} @samp{5208}
13258 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
13259 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
13260 @item @samp{5216} @tab @samp{5214} @samp{5216}
13261 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
13262 @item @samp{5225} @tab @samp{5224} @samp{5225}
13263 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
13264 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
13265 @item @samp{5249} @tab @samp{5249}
13266 @item @samp{5250} @tab @samp{5250}
13267 @item @samp{5271} @tab @samp{5270} @samp{5271}
13268 @item @samp{5272} @tab @samp{5272}
13269 @item @samp{5275} @tab @samp{5274} @samp{5275}
13270 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
13271 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
13272 @item @samp{5307} @tab @samp{5307}
13273 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
13274 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
13275 @item @samp{5407} @tab @samp{5407}
13276 @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}
13279 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
13280 @var{arch} is compatible with @var{cpu}. Other combinations of
13281 @option{-mcpu} and @option{-march} are rejected.
13283 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
13284 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
13285 where the value of @var{family} is given by the table above.
13287 @item -mtune=@var{tune}
13289 Tune the code for a particular microarchitecture, within the
13290 constraints set by @option{-march} and @option{-mcpu}.
13291 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
13292 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
13293 and @samp{cpu32}. The ColdFire microarchitectures
13294 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
13296 You can also use @option{-mtune=68020-40} for code that needs
13297 to run relatively well on 68020, 68030 and 68040 targets.
13298 @option{-mtune=68020-60} is similar but includes 68060 targets
13299 as well. These two options select the same tuning decisions as
13300 @option{-m68020-40} and @option{-m68020-60} respectively.
13302 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
13303 when tuning for 680x0 architecture @var{arch}. It also defines
13304 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
13305 option is used. If gcc is tuning for a range of architectures,
13306 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
13307 it defines the macros for every architecture in the range.
13309 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
13310 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
13311 of the arguments given above.
13317 Generate output for a 68000. This is the default
13318 when the compiler is configured for 68000-based systems.
13319 It is equivalent to @option{-march=68000}.
13321 Use this option for microcontrollers with a 68000 or EC000 core,
13322 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
13326 Generate output for a 68010. This is the default
13327 when the compiler is configured for 68010-based systems.
13328 It is equivalent to @option{-march=68010}.
13334 Generate output for a 68020. This is the default
13335 when the compiler is configured for 68020-based systems.
13336 It is equivalent to @option{-march=68020}.
13340 Generate output for a 68030. This is the default when the compiler is
13341 configured for 68030-based systems. It is equivalent to
13342 @option{-march=68030}.
13346 Generate output for a 68040. This is the default when the compiler is
13347 configured for 68040-based systems. It is equivalent to
13348 @option{-march=68040}.
13350 This option inhibits the use of 68881/68882 instructions that have to be
13351 emulated by software on the 68040. Use this option if your 68040 does not
13352 have code to emulate those instructions.
13356 Generate output for a 68060. This is the default when the compiler is
13357 configured for 68060-based systems. It is equivalent to
13358 @option{-march=68060}.
13360 This option inhibits the use of 68020 and 68881/68882 instructions that
13361 have to be emulated by software on the 68060. Use this option if your 68060
13362 does not have code to emulate those instructions.
13366 Generate output for a CPU32. This is the default
13367 when the compiler is configured for CPU32-based systems.
13368 It is equivalent to @option{-march=cpu32}.
13370 Use this option for microcontrollers with a
13371 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
13372 68336, 68340, 68341, 68349 and 68360.
13376 Generate output for a 520X ColdFire CPU@. This is the default
13377 when the compiler is configured for 520X-based systems.
13378 It is equivalent to @option{-mcpu=5206}, and is now deprecated
13379 in favor of that option.
13381 Use this option for microcontroller with a 5200 core, including
13382 the MCF5202, MCF5203, MCF5204 and MCF5206.
13386 Generate output for a 5206e ColdFire CPU@. The option is now
13387 deprecated in favor of the equivalent @option{-mcpu=5206e}.
13391 Generate output for a member of the ColdFire 528X family.
13392 The option is now deprecated in favor of the equivalent
13393 @option{-mcpu=528x}.
13397 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
13398 in favor of the equivalent @option{-mcpu=5307}.
13402 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
13403 in favor of the equivalent @option{-mcpu=5407}.
13407 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
13408 This includes use of hardware floating point instructions.
13409 The option is equivalent to @option{-mcpu=547x}, and is now
13410 deprecated in favor of that option.
13414 Generate output for a 68040, without using any of the new instructions.
13415 This results in code which can run relatively efficiently on either a
13416 68020/68881 or a 68030 or a 68040. The generated code does use the
13417 68881 instructions that are emulated on the 68040.
13419 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
13423 Generate output for a 68060, without using any of the new instructions.
13424 This results in code which can run relatively efficiently on either a
13425 68020/68881 or a 68030 or a 68040. The generated code does use the
13426 68881 instructions that are emulated on the 68060.
13428 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
13432 @opindex mhard-float
13434 Generate floating-point instructions. This is the default for 68020
13435 and above, and for ColdFire devices that have an FPU@. It defines the
13436 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
13437 on ColdFire targets.
13440 @opindex msoft-float
13441 Do not generate floating-point instructions; use library calls instead.
13442 This is the default for 68000, 68010, and 68832 targets. It is also
13443 the default for ColdFire devices that have no FPU.
13449 Generate (do not generate) ColdFire hardware divide and remainder
13450 instructions. If @option{-march} is used without @option{-mcpu},
13451 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
13452 architectures. Otherwise, the default is taken from the target CPU
13453 (either the default CPU, or the one specified by @option{-mcpu}). For
13454 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
13455 @option{-mcpu=5206e}.
13457 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
13461 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13462 Additionally, parameters passed on the stack are also aligned to a
13463 16-bit boundary even on targets whose API mandates promotion to 32-bit.
13467 Do not consider type @code{int} to be 16 bits wide. This is the default.
13470 @itemx -mno-bitfield
13471 @opindex mnobitfield
13472 @opindex mno-bitfield
13473 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
13474 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
13478 Do use the bit-field instructions. The @option{-m68020} option implies
13479 @option{-mbitfield}. This is the default if you use a configuration
13480 designed for a 68020.
13484 Use a different function-calling convention, in which functions
13485 that take a fixed number of arguments return with the @code{rtd}
13486 instruction, which pops their arguments while returning. This
13487 saves one instruction in the caller since there is no need to pop
13488 the arguments there.
13490 This calling convention is incompatible with the one normally
13491 used on Unix, so you cannot use it if you need to call libraries
13492 compiled with the Unix compiler.
13494 Also, you must provide function prototypes for all functions that
13495 take variable numbers of arguments (including @code{printf});
13496 otherwise incorrect code will be generated for calls to those
13499 In addition, seriously incorrect code will result if you call a
13500 function with too many arguments. (Normally, extra arguments are
13501 harmlessly ignored.)
13503 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
13504 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
13508 Do not use the calling conventions selected by @option{-mrtd}.
13509 This is the default.
13512 @itemx -mno-align-int
13513 @opindex malign-int
13514 @opindex mno-align-int
13515 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
13516 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
13517 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
13518 Aligning variables on 32-bit boundaries produces code that runs somewhat
13519 faster on processors with 32-bit busses at the expense of more memory.
13521 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
13522 align structures containing the above types differently than
13523 most published application binary interface specifications for the m68k.
13527 Use the pc-relative addressing mode of the 68000 directly, instead of
13528 using a global offset table. At present, this option implies @option{-fpic},
13529 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
13530 not presently supported with @option{-mpcrel}, though this could be supported for
13531 68020 and higher processors.
13533 @item -mno-strict-align
13534 @itemx -mstrict-align
13535 @opindex mno-strict-align
13536 @opindex mstrict-align
13537 Do not (do) assume that unaligned memory references will be handled by
13541 Generate code that allows the data segment to be located in a different
13542 area of memory from the text segment. This allows for execute in place in
13543 an environment without virtual memory management. This option implies
13546 @item -mno-sep-data
13547 Generate code that assumes that the data segment follows the text segment.
13548 This is the default.
13550 @item -mid-shared-library
13551 Generate code that supports shared libraries via the library ID method.
13552 This allows for execute in place and shared libraries in an environment
13553 without virtual memory management. This option implies @option{-fPIC}.
13555 @item -mno-id-shared-library
13556 Generate code that doesn't assume ID based shared libraries are being used.
13557 This is the default.
13559 @item -mshared-library-id=n
13560 Specified the identification number of the ID based shared library being
13561 compiled. Specifying a value of 0 will generate more compact code, specifying
13562 other values will force the allocation of that number to the current
13563 library but is no more space or time efficient than omitting this option.
13569 When generating position-independent code for ColdFire, generate code
13570 that works if the GOT has more than 8192 entries. This code is
13571 larger and slower than code generated without this option. On M680x0
13572 processors, this option is not needed; @option{-fPIC} suffices.
13574 GCC normally uses a single instruction to load values from the GOT@.
13575 While this is relatively efficient, it only works if the GOT
13576 is smaller than about 64k. Anything larger causes the linker
13577 to report an error such as:
13579 @cindex relocation truncated to fit (ColdFire)
13581 relocation truncated to fit: R_68K_GOT16O foobar
13584 If this happens, you should recompile your code with @option{-mxgot}.
13585 It should then work with very large GOTs. However, code generated with
13586 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
13587 the value of a global symbol.
13589 Note that some linkers, including newer versions of the GNU linker,
13590 can create multiple GOTs and sort GOT entries. If you have such a linker,
13591 you should only need to use @option{-mxgot} when compiling a single
13592 object file that accesses more than 8192 GOT entries. Very few do.
13594 These options have no effect unless GCC is generating
13595 position-independent code.
13599 @node M68hc1x Options
13600 @subsection M68hc1x Options
13601 @cindex M68hc1x options
13603 These are the @samp{-m} options defined for the 68hc11 and 68hc12
13604 microcontrollers. The default values for these options depends on
13605 which style of microcontroller was selected when the compiler was configured;
13606 the defaults for the most common choices are given below.
13613 Generate output for a 68HC11. This is the default
13614 when the compiler is configured for 68HC11-based systems.
13620 Generate output for a 68HC12. This is the default
13621 when the compiler is configured for 68HC12-based systems.
13627 Generate output for a 68HCS12.
13629 @item -mauto-incdec
13630 @opindex mauto-incdec
13631 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
13638 Enable the use of 68HC12 min and max instructions.
13641 @itemx -mno-long-calls
13642 @opindex mlong-calls
13643 @opindex mno-long-calls
13644 Treat all calls as being far away (near). If calls are assumed to be
13645 far away, the compiler will use the @code{call} instruction to
13646 call a function and the @code{rtc} instruction for returning.
13650 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13652 @item -msoft-reg-count=@var{count}
13653 @opindex msoft-reg-count
13654 Specify the number of pseudo-soft registers which are used for the
13655 code generation. The maximum number is 32. Using more pseudo-soft
13656 register may or may not result in better code depending on the program.
13657 The default is 4 for 68HC11 and 2 for 68HC12.
13661 @node MCore Options
13662 @subsection MCore Options
13663 @cindex MCore options
13665 These are the @samp{-m} options defined for the Motorola M*Core
13671 @itemx -mno-hardlit
13673 @opindex mno-hardlit
13674 Inline constants into the code stream if it can be done in two
13675 instructions or less.
13681 Use the divide instruction. (Enabled by default).
13683 @item -mrelax-immediate
13684 @itemx -mno-relax-immediate
13685 @opindex mrelax-immediate
13686 @opindex mno-relax-immediate
13687 Allow arbitrary sized immediates in bit operations.
13689 @item -mwide-bitfields
13690 @itemx -mno-wide-bitfields
13691 @opindex mwide-bitfields
13692 @opindex mno-wide-bitfields
13693 Always treat bit-fields as int-sized.
13695 @item -m4byte-functions
13696 @itemx -mno-4byte-functions
13697 @opindex m4byte-functions
13698 @opindex mno-4byte-functions
13699 Force all functions to be aligned to a four byte boundary.
13701 @item -mcallgraph-data
13702 @itemx -mno-callgraph-data
13703 @opindex mcallgraph-data
13704 @opindex mno-callgraph-data
13705 Emit callgraph information.
13708 @itemx -mno-slow-bytes
13709 @opindex mslow-bytes
13710 @opindex mno-slow-bytes
13711 Prefer word access when reading byte quantities.
13713 @item -mlittle-endian
13714 @itemx -mbig-endian
13715 @opindex mlittle-endian
13716 @opindex mbig-endian
13717 Generate code for a little endian target.
13723 Generate code for the 210 processor.
13727 Assume that run-time support has been provided and so omit the
13728 simulator library (@file{libsim.a)} from the linker command line.
13730 @item -mstack-increment=@var{size}
13731 @opindex mstack-increment
13732 Set the maximum amount for a single stack increment operation. Large
13733 values can increase the speed of programs which contain functions
13734 that need a large amount of stack space, but they can also trigger a
13735 segmentation fault if the stack is extended too much. The default
13741 @subsection MeP Options
13742 @cindex MeP options
13748 Enables the @code{abs} instruction, which is the absolute difference
13749 between two registers.
13753 Enables all the optional instructions - average, multiply, divide, bit
13754 operations, leading zero, absolute difference, min/max, clip, and
13760 Enables the @code{ave} instruction, which computes the average of two
13763 @item -mbased=@var{n}
13765 Variables of size @var{n} bytes or smaller will be placed in the
13766 @code{.based} section by default. Based variables use the @code{$tp}
13767 register as a base register, and there is a 128 byte limit to the
13768 @code{.based} section.
13772 Enables the bit operation instructions - bit test (@code{btstm}), set
13773 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
13774 test-and-set (@code{tas}).
13776 @item -mc=@var{name}
13778 Selects which section constant data will be placed in. @var{name} may
13779 be @code{tiny}, @code{near}, or @code{far}.
13783 Enables the @code{clip} instruction. Note that @code{-mclip} is not
13784 useful unless you also provide @code{-mminmax}.
13786 @item -mconfig=@var{name}
13788 Selects one of the build-in core configurations. Each MeP chip has
13789 one or more modules in it; each module has a core CPU and a variety of
13790 coprocessors, optional instructions, and peripherals. The
13791 @code{MeP-Integrator} tool, not part of GCC, provides these
13792 configurations through this option; using this option is the same as
13793 using all the corresponding command line options. The default
13794 configuration is @code{default}.
13798 Enables the coprocessor instructions. By default, this is a 32-bit
13799 coprocessor. Note that the coprocessor is normally enabled via the
13800 @code{-mconfig=} option.
13804 Enables the 32-bit coprocessor's instructions.
13808 Enables the 64-bit coprocessor's instructions.
13812 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
13816 Causes constant variables to be placed in the @code{.near} section.
13820 Enables the @code{div} and @code{divu} instructions.
13824 Generate big-endian code.
13828 Generate little-endian code.
13830 @item -mio-volatile
13831 @opindex mio-volatile
13832 Tells the compiler that any variable marked with the @code{io}
13833 attribute is to be considered volatile.
13837 Causes variables to be assigned to the @code{.far} section by default.
13841 Enables the @code{leadz} (leading zero) instruction.
13845 Causes variables to be assigned to the @code{.near} section by default.
13849 Enables the @code{min} and @code{max} instructions.
13853 Enables the multiplication and multiply-accumulate instructions.
13857 Disables all the optional instructions enabled by @code{-mall-opts}.
13861 Enables the @code{repeat} and @code{erepeat} instructions, used for
13862 low-overhead looping.
13866 Causes all variables to default to the @code{.tiny} section. Note
13867 that there is a 65536 byte limit to this section. Accesses to these
13868 variables use the @code{%gp} base register.
13872 Enables the saturation instructions. Note that the compiler does not
13873 currently generate these itself, but this option is included for
13874 compatibility with other tools, like @code{as}.
13878 Link the SDRAM-based runtime instead of the default ROM-based runtime.
13882 Link the simulator runtime libraries.
13886 Link the simulator runtime libraries, excluding built-in support
13887 for reset and exception vectors and tables.
13891 Causes all functions to default to the @code{.far} section. Without
13892 this option, functions default to the @code{.near} section.
13894 @item -mtiny=@var{n}
13896 Variables that are @var{n} bytes or smaller will be allocated to the
13897 @code{.tiny} section. These variables use the @code{$gp} base
13898 register. The default for this option is 4, but note that there's a
13899 65536 byte limit to the @code{.tiny} section.
13903 @node MicroBlaze Options
13904 @subsection MicroBlaze Options
13905 @cindex MicroBlaze Options
13910 @opindex msoft-float
13911 Use software emulation for floating point (default).
13914 @opindex mhard-float
13915 Use hardware floating point instructions.
13919 Do not optimize block moves, use @code{memcpy}.
13921 @item -mno-clearbss
13922 @opindex mno-clearbss
13923 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
13925 @item -mcpu=@var{cpu-type}
13927 Use features of and schedule code for given CPU.
13928 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
13929 where @var{X} is a major version, @var{YY} is the minor version, and
13930 @var{Z} is compatiblity code. Example values are @samp{v3.00.a},
13931 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
13933 @item -mxl-soft-mul
13934 @opindex mxl-soft-mul
13935 Use software multiply emulation (default).
13937 @item -mxl-soft-div
13938 @opindex mxl-soft-div
13939 Use software emulation for divides (default).
13941 @item -mxl-barrel-shift
13942 @opindex mxl-barrel-shift
13943 Use the hardware barrel shifter.
13945 @item -mxl-pattern-compare
13946 @opindex mxl-pattern-compare
13947 Use pattern compare instructions.
13949 @item -msmall-divides
13950 @opindex msmall-divides
13951 Use table lookup optimization for small signed integer divisions.
13953 @item -mxl-stack-check
13954 @opindex mxl-stack-check
13955 This option is deprecated. Use -fstack-check instead.
13958 @opindex mxl-gp-opt
13959 Use GP relative sdata/sbss sections.
13961 @item -mxl-multiply-high
13962 @opindex mxl-multiply-high
13963 Use multiply high instructions for high part of 32x32 multiply.
13965 @item -mxl-float-convert
13966 @opindex mxl-float-convert
13967 Use hardware floating point converstion instructions.
13969 @item -mxl-float-sqrt
13970 @opindex mxl-float-sqrt
13971 Use hardware floating point square root instruction.
13973 @item -mxl-mode-@var{app-model}
13974 Select application model @var{app-model}. Valid models are
13977 normal executable (default), uses startup code @file{crt0.o}.
13980 for use with Xilinx Microprocessor Debugger (XMD) based
13981 software intrusive debug agent called xmdstub. This uses startup file
13982 @file{crt1.o} and sets the start address of the program to be 0x800.
13985 for applications that are loaded using a bootloader.
13986 This model uses startup file @file{crt2.o} which does not contain a processor
13987 reset vector handler. This is suitable for transferring control on a
13988 processor reset to the bootloader rather than the application.
13991 for applications that do not require any of the
13992 MicroBlaze vectors. This option may be useful for applications running
13993 within a monitoring application. This model uses @file{crt3.o} as a startup file.
13996 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
13997 @option{-mxl-mode-@var{app-model}}.
14002 @subsection MIPS Options
14003 @cindex MIPS options
14009 Generate big-endian code.
14013 Generate little-endian code. This is the default for @samp{mips*el-*-*}
14016 @item -march=@var{arch}
14018 Generate code that will run on @var{arch}, which can be the name of a
14019 generic MIPS ISA, or the name of a particular processor.
14021 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
14022 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
14023 The processor names are:
14024 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
14025 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
14026 @samp{5kc}, @samp{5kf},
14028 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
14029 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
14030 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
14031 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
14032 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
14033 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
14037 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
14038 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
14039 @samp{rm7000}, @samp{rm9000},
14040 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
14043 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
14044 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
14046 The special value @samp{from-abi} selects the
14047 most compatible architecture for the selected ABI (that is,
14048 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
14050 Native Linux/GNU toolchains also support the value @samp{native},
14051 which selects the best architecture option for the host processor.
14052 @option{-march=native} has no effect if GCC does not recognize
14055 In processor names, a final @samp{000} can be abbreviated as @samp{k}
14056 (for example, @samp{-march=r2k}). Prefixes are optional, and
14057 @samp{vr} may be written @samp{r}.
14059 Names of the form @samp{@var{n}f2_1} refer to processors with
14060 FPUs clocked at half the rate of the core, names of the form
14061 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
14062 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
14063 processors with FPUs clocked a ratio of 3:2 with respect to the core.
14064 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
14065 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
14066 accepted as synonyms for @samp{@var{n}f1_1}.
14068 GCC defines two macros based on the value of this option. The first
14069 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
14070 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
14071 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
14072 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
14073 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
14075 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
14076 above. In other words, it will have the full prefix and will not
14077 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
14078 the macro names the resolved architecture (either @samp{"mips1"} or
14079 @samp{"mips3"}). It names the default architecture when no
14080 @option{-march} option is given.
14082 @item -mtune=@var{arch}
14084 Optimize for @var{arch}. Among other things, this option controls
14085 the way instructions are scheduled, and the perceived cost of arithmetic
14086 operations. The list of @var{arch} values is the same as for
14089 When this option is not used, GCC will optimize for the processor
14090 specified by @option{-march}. By using @option{-march} and
14091 @option{-mtune} together, it is possible to generate code that will
14092 run on a family of processors, but optimize the code for one
14093 particular member of that family.
14095 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
14096 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
14097 @samp{-march} ones described above.
14101 Equivalent to @samp{-march=mips1}.
14105 Equivalent to @samp{-march=mips2}.
14109 Equivalent to @samp{-march=mips3}.
14113 Equivalent to @samp{-march=mips4}.
14117 Equivalent to @samp{-march=mips32}.
14121 Equivalent to @samp{-march=mips32r2}.
14125 Equivalent to @samp{-march=mips64}.
14129 Equivalent to @samp{-march=mips64r2}.
14134 @opindex mno-mips16
14135 Generate (do not generate) MIPS16 code. If GCC is targetting a
14136 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
14138 MIPS16 code generation can also be controlled on a per-function basis
14139 by means of @code{mips16} and @code{nomips16} attributes.
14140 @xref{Function Attributes}, for more information.
14142 @item -mflip-mips16
14143 @opindex mflip-mips16
14144 Generate MIPS16 code on alternating functions. This option is provided
14145 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
14146 not intended for ordinary use in compiling user code.
14148 @item -minterlink-mips16
14149 @itemx -mno-interlink-mips16
14150 @opindex minterlink-mips16
14151 @opindex mno-interlink-mips16
14152 Require (do not require) that non-MIPS16 code be link-compatible with
14155 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
14156 it must either use a call or an indirect jump. @option{-minterlink-mips16}
14157 therefore disables direct jumps unless GCC knows that the target of the
14158 jump is not MIPS16.
14170 Generate code for the given ABI@.
14172 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
14173 generates 64-bit code when you select a 64-bit architecture, but you
14174 can use @option{-mgp32} to get 32-bit code instead.
14176 For information about the O64 ABI, see
14177 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
14179 GCC supports a variant of the o32 ABI in which floating-point registers
14180 are 64 rather than 32 bits wide. You can select this combination with
14181 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
14182 and @samp{mfhc1} instructions and is therefore only supported for
14183 MIPS32R2 processors.
14185 The register assignments for arguments and return values remain the
14186 same, but each scalar value is passed in a single 64-bit register
14187 rather than a pair of 32-bit registers. For example, scalar
14188 floating-point values are returned in @samp{$f0} only, not a
14189 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
14190 remains the same, but all 64 bits are saved.
14193 @itemx -mno-abicalls
14195 @opindex mno-abicalls
14196 Generate (do not generate) code that is suitable for SVR4-style
14197 dynamic objects. @option{-mabicalls} is the default for SVR4-based
14202 Generate (do not generate) code that is fully position-independent,
14203 and that can therefore be linked into shared libraries. This option
14204 only affects @option{-mabicalls}.
14206 All @option{-mabicalls} code has traditionally been position-independent,
14207 regardless of options like @option{-fPIC} and @option{-fpic}. However,
14208 as an extension, the GNU toolchain allows executables to use absolute
14209 accesses for locally-binding symbols. It can also use shorter GP
14210 initialization sequences and generate direct calls to locally-defined
14211 functions. This mode is selected by @option{-mno-shared}.
14213 @option{-mno-shared} depends on binutils 2.16 or higher and generates
14214 objects that can only be linked by the GNU linker. However, the option
14215 does not affect the ABI of the final executable; it only affects the ABI
14216 of relocatable objects. Using @option{-mno-shared} will generally make
14217 executables both smaller and quicker.
14219 @option{-mshared} is the default.
14225 Assume (do not assume) that the static and dynamic linkers
14226 support PLTs and copy relocations. This option only affects
14227 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
14228 has no effect without @samp{-msym32}.
14230 You can make @option{-mplt} the default by configuring
14231 GCC with @option{--with-mips-plt}. The default is
14232 @option{-mno-plt} otherwise.
14238 Lift (do not lift) the usual restrictions on the size of the global
14241 GCC normally uses a single instruction to load values from the GOT@.
14242 While this is relatively efficient, it will only work if the GOT
14243 is smaller than about 64k. Anything larger will cause the linker
14244 to report an error such as:
14246 @cindex relocation truncated to fit (MIPS)
14248 relocation truncated to fit: R_MIPS_GOT16 foobar
14251 If this happens, you should recompile your code with @option{-mxgot}.
14252 It should then work with very large GOTs, although it will also be
14253 less efficient, since it will take three instructions to fetch the
14254 value of a global symbol.
14256 Note that some linkers can create multiple GOTs. If you have such a
14257 linker, you should only need to use @option{-mxgot} when a single object
14258 file accesses more than 64k's worth of GOT entries. Very few do.
14260 These options have no effect unless GCC is generating position
14265 Assume that general-purpose registers are 32 bits wide.
14269 Assume that general-purpose registers are 64 bits wide.
14273 Assume that floating-point registers are 32 bits wide.
14277 Assume that floating-point registers are 64 bits wide.
14280 @opindex mhard-float
14281 Use floating-point coprocessor instructions.
14284 @opindex msoft-float
14285 Do not use floating-point coprocessor instructions. Implement
14286 floating-point calculations using library calls instead.
14288 @item -msingle-float
14289 @opindex msingle-float
14290 Assume that the floating-point coprocessor only supports single-precision
14293 @item -mdouble-float
14294 @opindex mdouble-float
14295 Assume that the floating-point coprocessor supports double-precision
14296 operations. This is the default.
14302 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
14303 implement atomic memory built-in functions. When neither option is
14304 specified, GCC will use the instructions if the target architecture
14307 @option{-mllsc} is useful if the runtime environment can emulate the
14308 instructions and @option{-mno-llsc} can be useful when compiling for
14309 nonstandard ISAs. You can make either option the default by
14310 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
14311 respectively. @option{--with-llsc} is the default for some
14312 configurations; see the installation documentation for details.
14318 Use (do not use) revision 1 of the MIPS DSP ASE@.
14319 @xref{MIPS DSP Built-in Functions}. This option defines the
14320 preprocessor macro @samp{__mips_dsp}. It also defines
14321 @samp{__mips_dsp_rev} to 1.
14327 Use (do not use) revision 2 of the MIPS DSP ASE@.
14328 @xref{MIPS DSP Built-in Functions}. This option defines the
14329 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
14330 It also defines @samp{__mips_dsp_rev} to 2.
14333 @itemx -mno-smartmips
14334 @opindex msmartmips
14335 @opindex mno-smartmips
14336 Use (do not use) the MIPS SmartMIPS ASE.
14338 @item -mpaired-single
14339 @itemx -mno-paired-single
14340 @opindex mpaired-single
14341 @opindex mno-paired-single
14342 Use (do not use) paired-single floating-point instructions.
14343 @xref{MIPS Paired-Single Support}. This option requires
14344 hardware floating-point support to be enabled.
14350 Use (do not use) MIPS Digital Media Extension instructions.
14351 This option can only be used when generating 64-bit code and requires
14352 hardware floating-point support to be enabled.
14357 @opindex mno-mips3d
14358 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
14359 The option @option{-mips3d} implies @option{-mpaired-single}.
14365 Use (do not use) MT Multithreading instructions.
14369 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
14370 an explanation of the default and the way that the pointer size is
14375 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
14377 The default size of @code{int}s, @code{long}s and pointers depends on
14378 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
14379 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
14380 32-bit @code{long}s. Pointers are the same size as @code{long}s,
14381 or the same size as integer registers, whichever is smaller.
14387 Assume (do not assume) that all symbols have 32-bit values, regardless
14388 of the selected ABI@. This option is useful in combination with
14389 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
14390 to generate shorter and faster references to symbolic addresses.
14394 Put definitions of externally-visible data in a small data section
14395 if that data is no bigger than @var{num} bytes. GCC can then access
14396 the data more efficiently; see @option{-mgpopt} for details.
14398 The default @option{-G} option depends on the configuration.
14400 @item -mlocal-sdata
14401 @itemx -mno-local-sdata
14402 @opindex mlocal-sdata
14403 @opindex mno-local-sdata
14404 Extend (do not extend) the @option{-G} behavior to local data too,
14405 such as to static variables in C@. @option{-mlocal-sdata} is the
14406 default for all configurations.
14408 If the linker complains that an application is using too much small data,
14409 you might want to try rebuilding the less performance-critical parts with
14410 @option{-mno-local-sdata}. You might also want to build large
14411 libraries with @option{-mno-local-sdata}, so that the libraries leave
14412 more room for the main program.
14414 @item -mextern-sdata
14415 @itemx -mno-extern-sdata
14416 @opindex mextern-sdata
14417 @opindex mno-extern-sdata
14418 Assume (do not assume) that externally-defined data will be in
14419 a small data section if that data is within the @option{-G} limit.
14420 @option{-mextern-sdata} is the default for all configurations.
14422 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
14423 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
14424 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
14425 is placed in a small data section. If @var{Var} is defined by another
14426 module, you must either compile that module with a high-enough
14427 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
14428 definition. If @var{Var} is common, you must link the application
14429 with a high-enough @option{-G} setting.
14431 The easiest way of satisfying these restrictions is to compile
14432 and link every module with the same @option{-G} option. However,
14433 you may wish to build a library that supports several different
14434 small data limits. You can do this by compiling the library with
14435 the highest supported @option{-G} setting and additionally using
14436 @option{-mno-extern-sdata} to stop the library from making assumptions
14437 about externally-defined data.
14443 Use (do not use) GP-relative accesses for symbols that are known to be
14444 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
14445 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
14448 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
14449 might not hold the value of @code{_gp}. For example, if the code is
14450 part of a library that might be used in a boot monitor, programs that
14451 call boot monitor routines will pass an unknown value in @code{$gp}.
14452 (In such situations, the boot monitor itself would usually be compiled
14453 with @option{-G0}.)
14455 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
14456 @option{-mno-extern-sdata}.
14458 @item -membedded-data
14459 @itemx -mno-embedded-data
14460 @opindex membedded-data
14461 @opindex mno-embedded-data
14462 Allocate variables to the read-only data section first if possible, then
14463 next in the small data section if possible, otherwise in data. This gives
14464 slightly slower code than the default, but reduces the amount of RAM required
14465 when executing, and thus may be preferred for some embedded systems.
14467 @item -muninit-const-in-rodata
14468 @itemx -mno-uninit-const-in-rodata
14469 @opindex muninit-const-in-rodata
14470 @opindex mno-uninit-const-in-rodata
14471 Put uninitialized @code{const} variables in the read-only data section.
14472 This option is only meaningful in conjunction with @option{-membedded-data}.
14474 @item -mcode-readable=@var{setting}
14475 @opindex mcode-readable
14476 Specify whether GCC may generate code that reads from executable sections.
14477 There are three possible settings:
14480 @item -mcode-readable=yes
14481 Instructions may freely access executable sections. This is the
14484 @item -mcode-readable=pcrel
14485 MIPS16 PC-relative load instructions can access executable sections,
14486 but other instructions must not do so. This option is useful on 4KSc
14487 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
14488 It is also useful on processors that can be configured to have a dual
14489 instruction/data SRAM interface and that, like the M4K, automatically
14490 redirect PC-relative loads to the instruction RAM.
14492 @item -mcode-readable=no
14493 Instructions must not access executable sections. This option can be
14494 useful on targets that are configured to have a dual instruction/data
14495 SRAM interface but that (unlike the M4K) do not automatically redirect
14496 PC-relative loads to the instruction RAM.
14499 @item -msplit-addresses
14500 @itemx -mno-split-addresses
14501 @opindex msplit-addresses
14502 @opindex mno-split-addresses
14503 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
14504 relocation operators. This option has been superseded by
14505 @option{-mexplicit-relocs} but is retained for backwards compatibility.
14507 @item -mexplicit-relocs
14508 @itemx -mno-explicit-relocs
14509 @opindex mexplicit-relocs
14510 @opindex mno-explicit-relocs
14511 Use (do not use) assembler relocation operators when dealing with symbolic
14512 addresses. The alternative, selected by @option{-mno-explicit-relocs},
14513 is to use assembler macros instead.
14515 @option{-mexplicit-relocs} is the default if GCC was configured
14516 to use an assembler that supports relocation operators.
14518 @item -mcheck-zero-division
14519 @itemx -mno-check-zero-division
14520 @opindex mcheck-zero-division
14521 @opindex mno-check-zero-division
14522 Trap (do not trap) on integer division by zero.
14524 The default is @option{-mcheck-zero-division}.
14526 @item -mdivide-traps
14527 @itemx -mdivide-breaks
14528 @opindex mdivide-traps
14529 @opindex mdivide-breaks
14530 MIPS systems check for division by zero by generating either a
14531 conditional trap or a break instruction. Using traps results in
14532 smaller code, but is only supported on MIPS II and later. Also, some
14533 versions of the Linux kernel have a bug that prevents trap from
14534 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
14535 allow conditional traps on architectures that support them and
14536 @option{-mdivide-breaks} to force the use of breaks.
14538 The default is usually @option{-mdivide-traps}, but this can be
14539 overridden at configure time using @option{--with-divide=breaks}.
14540 Divide-by-zero checks can be completely disabled using
14541 @option{-mno-check-zero-division}.
14546 @opindex mno-memcpy
14547 Force (do not force) the use of @code{memcpy()} for non-trivial block
14548 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
14549 most constant-sized copies.
14552 @itemx -mno-long-calls
14553 @opindex mlong-calls
14554 @opindex mno-long-calls
14555 Disable (do not disable) use of the @code{jal} instruction. Calling
14556 functions using @code{jal} is more efficient but requires the caller
14557 and callee to be in the same 256 megabyte segment.
14559 This option has no effect on abicalls code. The default is
14560 @option{-mno-long-calls}.
14566 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
14567 instructions, as provided by the R4650 ISA@.
14570 @itemx -mno-fused-madd
14571 @opindex mfused-madd
14572 @opindex mno-fused-madd
14573 Enable (disable) use of the floating point multiply-accumulate
14574 instructions, when they are available. The default is
14575 @option{-mfused-madd}.
14577 When multiply-accumulate instructions are used, the intermediate
14578 product is calculated to infinite precision and is not subject to
14579 the FCSR Flush to Zero bit. This may be undesirable in some
14584 Tell the MIPS assembler to not run its preprocessor over user
14585 assembler files (with a @samp{.s} suffix) when assembling them.
14588 @itemx -mno-fix-r4000
14589 @opindex mfix-r4000
14590 @opindex mno-fix-r4000
14591 Work around certain R4000 CPU errata:
14594 A double-word or a variable shift may give an incorrect result if executed
14595 immediately after starting an integer division.
14597 A double-word or a variable shift may give an incorrect result if executed
14598 while an integer multiplication is in progress.
14600 An integer division may give an incorrect result if started in a delay slot
14601 of a taken branch or a jump.
14605 @itemx -mno-fix-r4400
14606 @opindex mfix-r4400
14607 @opindex mno-fix-r4400
14608 Work around certain R4400 CPU errata:
14611 A double-word or a variable shift may give an incorrect result if executed
14612 immediately after starting an integer division.
14616 @itemx -mno-fix-r10000
14617 @opindex mfix-r10000
14618 @opindex mno-fix-r10000
14619 Work around certain R10000 errata:
14622 @code{ll}/@code{sc} sequences may not behave atomically on revisions
14623 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
14626 This option can only be used if the target architecture supports
14627 branch-likely instructions. @option{-mfix-r10000} is the default when
14628 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
14632 @itemx -mno-fix-vr4120
14633 @opindex mfix-vr4120
14634 Work around certain VR4120 errata:
14637 @code{dmultu} does not always produce the correct result.
14639 @code{div} and @code{ddiv} do not always produce the correct result if one
14640 of the operands is negative.
14642 The workarounds for the division errata rely on special functions in
14643 @file{libgcc.a}. At present, these functions are only provided by
14644 the @code{mips64vr*-elf} configurations.
14646 Other VR4120 errata require a nop to be inserted between certain pairs of
14647 instructions. These errata are handled by the assembler, not by GCC itself.
14650 @opindex mfix-vr4130
14651 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
14652 workarounds are implemented by the assembler rather than by GCC,
14653 although GCC will avoid using @code{mflo} and @code{mfhi} if the
14654 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
14655 instructions are available instead.
14658 @itemx -mno-fix-sb1
14660 Work around certain SB-1 CPU core errata.
14661 (This flag currently works around the SB-1 revision 2
14662 ``F1'' and ``F2'' floating point errata.)
14664 @item -mr10k-cache-barrier=@var{setting}
14665 @opindex mr10k-cache-barrier
14666 Specify whether GCC should insert cache barriers to avoid the
14667 side-effects of speculation on R10K processors.
14669 In common with many processors, the R10K tries to predict the outcome
14670 of a conditional branch and speculatively executes instructions from
14671 the ``taken'' branch. It later aborts these instructions if the
14672 predicted outcome was wrong. However, on the R10K, even aborted
14673 instructions can have side effects.
14675 This problem only affects kernel stores and, depending on the system,
14676 kernel loads. As an example, a speculatively-executed store may load
14677 the target memory into cache and mark the cache line as dirty, even if
14678 the store itself is later aborted. If a DMA operation writes to the
14679 same area of memory before the ``dirty'' line is flushed, the cached
14680 data will overwrite the DMA-ed data. See the R10K processor manual
14681 for a full description, including other potential problems.
14683 One workaround is to insert cache barrier instructions before every memory
14684 access that might be speculatively executed and that might have side
14685 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
14686 controls GCC's implementation of this workaround. It assumes that
14687 aborted accesses to any byte in the following regions will not have
14692 the memory occupied by the current function's stack frame;
14695 the memory occupied by an incoming stack argument;
14698 the memory occupied by an object with a link-time-constant address.
14701 It is the kernel's responsibility to ensure that speculative
14702 accesses to these regions are indeed safe.
14704 If the input program contains a function declaration such as:
14710 then the implementation of @code{foo} must allow @code{j foo} and
14711 @code{jal foo} to be executed speculatively. GCC honors this
14712 restriction for functions it compiles itself. It expects non-GCC
14713 functions (such as hand-written assembly code) to do the same.
14715 The option has three forms:
14718 @item -mr10k-cache-barrier=load-store
14719 Insert a cache barrier before a load or store that might be
14720 speculatively executed and that might have side effects even
14723 @item -mr10k-cache-barrier=store
14724 Insert a cache barrier before a store that might be speculatively
14725 executed and that might have side effects even if aborted.
14727 @item -mr10k-cache-barrier=none
14728 Disable the insertion of cache barriers. This is the default setting.
14731 @item -mflush-func=@var{func}
14732 @itemx -mno-flush-func
14733 @opindex mflush-func
14734 Specifies the function to call to flush the I and D caches, or to not
14735 call any such function. If called, the function must take the same
14736 arguments as the common @code{_flush_func()}, that is, the address of the
14737 memory range for which the cache is being flushed, the size of the
14738 memory range, and the number 3 (to flush both caches). The default
14739 depends on the target GCC was configured for, but commonly is either
14740 @samp{_flush_func} or @samp{__cpu_flush}.
14742 @item mbranch-cost=@var{num}
14743 @opindex mbranch-cost
14744 Set the cost of branches to roughly @var{num} ``simple'' instructions.
14745 This cost is only a heuristic and is not guaranteed to produce
14746 consistent results across releases. A zero cost redundantly selects
14747 the default, which is based on the @option{-mtune} setting.
14749 @item -mbranch-likely
14750 @itemx -mno-branch-likely
14751 @opindex mbranch-likely
14752 @opindex mno-branch-likely
14753 Enable or disable use of Branch Likely instructions, regardless of the
14754 default for the selected architecture. By default, Branch Likely
14755 instructions may be generated if they are supported by the selected
14756 architecture. An exception is for the MIPS32 and MIPS64 architectures
14757 and processors which implement those architectures; for those, Branch
14758 Likely instructions will not be generated by default because the MIPS32
14759 and MIPS64 architectures specifically deprecate their use.
14761 @item -mfp-exceptions
14762 @itemx -mno-fp-exceptions
14763 @opindex mfp-exceptions
14764 Specifies whether FP exceptions are enabled. This affects how we schedule
14765 FP instructions for some processors. The default is that FP exceptions are
14768 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
14769 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
14772 @item -mvr4130-align
14773 @itemx -mno-vr4130-align
14774 @opindex mvr4130-align
14775 The VR4130 pipeline is two-way superscalar, but can only issue two
14776 instructions together if the first one is 8-byte aligned. When this
14777 option is enabled, GCC will align pairs of instructions that it
14778 thinks should execute in parallel.
14780 This option only has an effect when optimizing for the VR4130.
14781 It normally makes code faster, but at the expense of making it bigger.
14782 It is enabled by default at optimization level @option{-O3}.
14787 Enable (disable) generation of @code{synci} instructions on
14788 architectures that support it. The @code{synci} instructions (if
14789 enabled) will be generated when @code{__builtin___clear_cache()} is
14792 This option defaults to @code{-mno-synci}, but the default can be
14793 overridden by configuring with @code{--with-synci}.
14795 When compiling code for single processor systems, it is generally safe
14796 to use @code{synci}. However, on many multi-core (SMP) systems, it
14797 will not invalidate the instruction caches on all cores and may lead
14798 to undefined behavior.
14800 @item -mrelax-pic-calls
14801 @itemx -mno-relax-pic-calls
14802 @opindex mrelax-pic-calls
14803 Try to turn PIC calls that are normally dispatched via register
14804 @code{$25} into direct calls. This is only possible if the linker can
14805 resolve the destination at link-time and if the destination is within
14806 range for a direct call.
14808 @option{-mrelax-pic-calls} is the default if GCC was configured to use
14809 an assembler and a linker that supports the @code{.reloc} assembly
14810 directive and @code{-mexplicit-relocs} is in effect. With
14811 @code{-mno-explicit-relocs}, this optimization can be performed by the
14812 assembler and the linker alone without help from the compiler.
14814 @item -mmcount-ra-address
14815 @itemx -mno-mcount-ra-address
14816 @opindex mmcount-ra-address
14817 @opindex mno-mcount-ra-address
14818 Emit (do not emit) code that allows @code{_mcount} to modify the
14819 calling function's return address. When enabled, this option extends
14820 the usual @code{_mcount} interface with a new @var{ra-address}
14821 parameter, which has type @code{intptr_t *} and is passed in register
14822 @code{$12}. @code{_mcount} can then modify the return address by
14823 doing both of the following:
14826 Returning the new address in register @code{$31}.
14828 Storing the new address in @code{*@var{ra-address}},
14829 if @var{ra-address} is nonnull.
14832 The default is @option{-mno-mcount-ra-address}.
14837 @subsection MMIX Options
14838 @cindex MMIX Options
14840 These options are defined for the MMIX:
14844 @itemx -mno-libfuncs
14846 @opindex mno-libfuncs
14847 Specify that intrinsic library functions are being compiled, passing all
14848 values in registers, no matter the size.
14851 @itemx -mno-epsilon
14853 @opindex mno-epsilon
14854 Generate floating-point comparison instructions that compare with respect
14855 to the @code{rE} epsilon register.
14857 @item -mabi=mmixware
14859 @opindex mabi=mmixware
14861 Generate code that passes function parameters and return values that (in
14862 the called function) are seen as registers @code{$0} and up, as opposed to
14863 the GNU ABI which uses global registers @code{$231} and up.
14865 @item -mzero-extend
14866 @itemx -mno-zero-extend
14867 @opindex mzero-extend
14868 @opindex mno-zero-extend
14869 When reading data from memory in sizes shorter than 64 bits, use (do not
14870 use) zero-extending load instructions by default, rather than
14871 sign-extending ones.
14874 @itemx -mno-knuthdiv
14876 @opindex mno-knuthdiv
14877 Make the result of a division yielding a remainder have the same sign as
14878 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
14879 remainder follows the sign of the dividend. Both methods are
14880 arithmetically valid, the latter being almost exclusively used.
14882 @item -mtoplevel-symbols
14883 @itemx -mno-toplevel-symbols
14884 @opindex mtoplevel-symbols
14885 @opindex mno-toplevel-symbols
14886 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
14887 code can be used with the @code{PREFIX} assembly directive.
14891 Generate an executable in the ELF format, rather than the default
14892 @samp{mmo} format used by the @command{mmix} simulator.
14894 @item -mbranch-predict
14895 @itemx -mno-branch-predict
14896 @opindex mbranch-predict
14897 @opindex mno-branch-predict
14898 Use (do not use) the probable-branch instructions, when static branch
14899 prediction indicates a probable branch.
14901 @item -mbase-addresses
14902 @itemx -mno-base-addresses
14903 @opindex mbase-addresses
14904 @opindex mno-base-addresses
14905 Generate (do not generate) code that uses @emph{base addresses}. Using a
14906 base address automatically generates a request (handled by the assembler
14907 and the linker) for a constant to be set up in a global register. The
14908 register is used for one or more base address requests within the range 0
14909 to 255 from the value held in the register. The generally leads to short
14910 and fast code, but the number of different data items that can be
14911 addressed is limited. This means that a program that uses lots of static
14912 data may require @option{-mno-base-addresses}.
14914 @item -msingle-exit
14915 @itemx -mno-single-exit
14916 @opindex msingle-exit
14917 @opindex mno-single-exit
14918 Force (do not force) generated code to have a single exit point in each
14922 @node MN10300 Options
14923 @subsection MN10300 Options
14924 @cindex MN10300 options
14926 These @option{-m} options are defined for Matsushita MN10300 architectures:
14931 Generate code to avoid bugs in the multiply instructions for the MN10300
14932 processors. This is the default.
14934 @item -mno-mult-bug
14935 @opindex mno-mult-bug
14936 Do not generate code to avoid bugs in the multiply instructions for the
14937 MN10300 processors.
14941 Generate code which uses features specific to the AM33 processor.
14945 Do not generate code which uses features specific to the AM33 processor. This
14950 Generate code which uses features specific to the AM33/2.0 processor.
14954 Generate code which uses features specific to the AM34 processor.
14956 @item -mtune=@var{cpu-type}
14958 Use the timing characteristics of the indicated CPU type when
14959 scheduling instructions. This does not change the targeted processor
14960 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
14961 @samp{am33-2} or @samp{am34}.
14963 @item -mreturn-pointer-on-d0
14964 @opindex mreturn-pointer-on-d0
14965 When generating a function which returns a pointer, return the pointer
14966 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
14967 only in a0, and attempts to call such functions without a prototype
14968 would result in errors. Note that this option is on by default; use
14969 @option{-mno-return-pointer-on-d0} to disable it.
14973 Do not link in the C run-time initialization object file.
14977 Indicate to the linker that it should perform a relaxation optimization pass
14978 to shorten branches, calls and absolute memory addresses. This option only
14979 has an effect when used on the command line for the final link step.
14981 This option makes symbolic debugging impossible.
14984 @node PDP-11 Options
14985 @subsection PDP-11 Options
14986 @cindex PDP-11 Options
14988 These options are defined for the PDP-11:
14993 Use hardware FPP floating point. This is the default. (FIS floating
14994 point on the PDP-11/40 is not supported.)
14997 @opindex msoft-float
14998 Do not use hardware floating point.
15002 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
15006 Return floating-point results in memory. This is the default.
15010 Generate code for a PDP-11/40.
15014 Generate code for a PDP-11/45. This is the default.
15018 Generate code for a PDP-11/10.
15020 @item -mbcopy-builtin
15021 @opindex mbcopy-builtin
15022 Use inline @code{movmemhi} patterns for copying memory. This is the
15027 Do not use inline @code{movmemhi} patterns for copying memory.
15033 Use 16-bit @code{int}. This is the default.
15039 Use 32-bit @code{int}.
15042 @itemx -mno-float32
15044 @opindex mno-float32
15045 Use 64-bit @code{float}. This is the default.
15048 @itemx -mno-float64
15050 @opindex mno-float64
15051 Use 32-bit @code{float}.
15055 Use @code{abshi2} pattern. This is the default.
15059 Do not use @code{abshi2} pattern.
15061 @item -mbranch-expensive
15062 @opindex mbranch-expensive
15063 Pretend that branches are expensive. This is for experimenting with
15064 code generation only.
15066 @item -mbranch-cheap
15067 @opindex mbranch-cheap
15068 Do not pretend that branches are expensive. This is the default.
15072 Use Unix assembler syntax. This is the default when configured for
15073 @samp{pdp11-*-bsd}.
15077 Use DEC assembler syntax. This is the default when configured for any
15078 PDP-11 target other than @samp{pdp11-*-bsd}.
15081 @node picoChip Options
15082 @subsection picoChip Options
15083 @cindex picoChip options
15085 These @samp{-m} options are defined for picoChip implementations:
15089 @item -mae=@var{ae_type}
15091 Set the instruction set, register set, and instruction scheduling
15092 parameters for array element type @var{ae_type}. Supported values
15093 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
15095 @option{-mae=ANY} selects a completely generic AE type. Code
15096 generated with this option will run on any of the other AE types. The
15097 code will not be as efficient as it would be if compiled for a specific
15098 AE type, and some types of operation (e.g., multiplication) will not
15099 work properly on all types of AE.
15101 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
15102 for compiled code, and is the default.
15104 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
15105 option may suffer from poor performance of byte (char) manipulation,
15106 since the DSP AE does not provide hardware support for byte load/stores.
15108 @item -msymbol-as-address
15109 Enable the compiler to directly use a symbol name as an address in a
15110 load/store instruction, without first loading it into a
15111 register. Typically, the use of this option will generate larger
15112 programs, which run faster than when the option isn't used. However, the
15113 results vary from program to program, so it is left as a user option,
15114 rather than being permanently enabled.
15116 @item -mno-inefficient-warnings
15117 Disables warnings about the generation of inefficient code. These
15118 warnings can be generated, for example, when compiling code which
15119 performs byte-level memory operations on the MAC AE type. The MAC AE has
15120 no hardware support for byte-level memory operations, so all byte
15121 load/stores must be synthesized from word load/store operations. This is
15122 inefficient and a warning will be generated indicating to the programmer
15123 that they should rewrite the code to avoid byte operations, or to target
15124 an AE type which has the necessary hardware support. This option enables
15125 the warning to be turned off.
15129 @node PowerPC Options
15130 @subsection PowerPC Options
15131 @cindex PowerPC options
15133 These are listed under @xref{RS/6000 and PowerPC Options}.
15135 @node RS/6000 and PowerPC Options
15136 @subsection IBM RS/6000 and PowerPC Options
15137 @cindex RS/6000 and PowerPC Options
15138 @cindex IBM RS/6000 and PowerPC Options
15140 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
15147 @itemx -mno-powerpc
15148 @itemx -mpowerpc-gpopt
15149 @itemx -mno-powerpc-gpopt
15150 @itemx -mpowerpc-gfxopt
15151 @itemx -mno-powerpc-gfxopt
15154 @itemx -mno-powerpc64
15158 @itemx -mno-popcntb
15160 @itemx -mno-popcntd
15169 @itemx -mno-hard-dfp
15173 @opindex mno-power2
15175 @opindex mno-powerpc
15176 @opindex mpowerpc-gpopt
15177 @opindex mno-powerpc-gpopt
15178 @opindex mpowerpc-gfxopt
15179 @opindex mno-powerpc-gfxopt
15180 @opindex mpowerpc64
15181 @opindex mno-powerpc64
15185 @opindex mno-popcntb
15187 @opindex mno-popcntd
15193 @opindex mno-mfpgpr
15195 @opindex mno-hard-dfp
15196 GCC supports two related instruction set architectures for the
15197 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
15198 instructions supported by the @samp{rios} chip set used in the original
15199 RS/6000 systems and the @dfn{PowerPC} instruction set is the
15200 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
15201 the IBM 4xx, 6xx, and follow-on microprocessors.
15203 Neither architecture is a subset of the other. However there is a
15204 large common subset of instructions supported by both. An MQ
15205 register is included in processors supporting the POWER architecture.
15207 You use these options to specify which instructions are available on the
15208 processor you are using. The default value of these options is
15209 determined when configuring GCC@. Specifying the
15210 @option{-mcpu=@var{cpu_type}} overrides the specification of these
15211 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
15212 rather than the options listed above.
15214 The @option{-mpower} option allows GCC to generate instructions that
15215 are found only in the POWER architecture and to use the MQ register.
15216 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
15217 to generate instructions that are present in the POWER2 architecture but
15218 not the original POWER architecture.
15220 The @option{-mpowerpc} option allows GCC to generate instructions that
15221 are found only in the 32-bit subset of the PowerPC architecture.
15222 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
15223 GCC to use the optional PowerPC architecture instructions in the
15224 General Purpose group, including floating-point square root. Specifying
15225 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
15226 use the optional PowerPC architecture instructions in the Graphics
15227 group, including floating-point select.
15229 The @option{-mmfcrf} option allows GCC to generate the move from
15230 condition register field instruction implemented on the POWER4
15231 processor and other processors that support the PowerPC V2.01
15233 The @option{-mpopcntb} option allows GCC to generate the popcount and
15234 double precision FP reciprocal estimate instruction implemented on the
15235 POWER5 processor and other processors that support the PowerPC V2.02
15237 The @option{-mpopcntd} option allows GCC to generate the popcount
15238 instruction implemented on the POWER7 processor and other processors
15239 that support the PowerPC V2.06 architecture.
15240 The @option{-mfprnd} option allows GCC to generate the FP round to
15241 integer instructions implemented on the POWER5+ processor and other
15242 processors that support the PowerPC V2.03 architecture.
15243 The @option{-mcmpb} option allows GCC to generate the compare bytes
15244 instruction implemented on the POWER6 processor and other processors
15245 that support the PowerPC V2.05 architecture.
15246 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
15247 general purpose register instructions implemented on the POWER6X
15248 processor and other processors that support the extended PowerPC V2.05
15250 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
15251 point instructions implemented on some POWER processors.
15253 The @option{-mpowerpc64} option allows GCC to generate the additional
15254 64-bit instructions that are found in the full PowerPC64 architecture
15255 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
15256 @option{-mno-powerpc64}.
15258 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
15259 will use only the instructions in the common subset of both
15260 architectures plus some special AIX common-mode calls, and will not use
15261 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
15262 permits GCC to use any instruction from either architecture and to
15263 allow use of the MQ register; specify this for the Motorola MPC601.
15265 @item -mnew-mnemonics
15266 @itemx -mold-mnemonics
15267 @opindex mnew-mnemonics
15268 @opindex mold-mnemonics
15269 Select which mnemonics to use in the generated assembler code. With
15270 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
15271 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
15272 assembler mnemonics defined for the POWER architecture. Instructions
15273 defined in only one architecture have only one mnemonic; GCC uses that
15274 mnemonic irrespective of which of these options is specified.
15276 GCC defaults to the mnemonics appropriate for the architecture in
15277 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
15278 value of these option. Unless you are building a cross-compiler, you
15279 should normally not specify either @option{-mnew-mnemonics} or
15280 @option{-mold-mnemonics}, but should instead accept the default.
15282 @item -mcpu=@var{cpu_type}
15284 Set architecture type, register usage, choice of mnemonics, and
15285 instruction scheduling parameters for machine type @var{cpu_type}.
15286 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
15287 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
15288 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
15289 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
15290 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
15291 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
15292 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{ec603e}, @samp{G3},
15293 @samp{G4}, @samp{G5}, @samp{titan}, @samp{power}, @samp{power2}, @samp{power3},
15294 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x},
15295 @samp{power7}, @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
15296 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
15298 @option{-mcpu=common} selects a completely generic processor. Code
15299 generated under this option will run on any POWER or PowerPC processor.
15300 GCC will use only the instructions in the common subset of both
15301 architectures, and will not use the MQ register. GCC assumes a generic
15302 processor model for scheduling purposes.
15304 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
15305 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
15306 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
15307 types, with an appropriate, generic processor model assumed for
15308 scheduling purposes.
15310 The other options specify a specific processor. Code generated under
15311 those options will run best on that processor, and may not run at all on
15314 The @option{-mcpu} options automatically enable or disable the
15317 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
15318 -mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol
15319 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
15320 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx}
15322 The particular options set for any particular CPU will vary between
15323 compiler versions, depending on what setting seems to produce optimal
15324 code for that CPU; it doesn't necessarily reflect the actual hardware's
15325 capabilities. If you wish to set an individual option to a particular
15326 value, you may specify it after the @option{-mcpu} option, like
15327 @samp{-mcpu=970 -mno-altivec}.
15329 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
15330 not enabled or disabled by the @option{-mcpu} option at present because
15331 AIX does not have full support for these options. You may still
15332 enable or disable them individually if you're sure it'll work in your
15335 @item -mtune=@var{cpu_type}
15337 Set the instruction scheduling parameters for machine type
15338 @var{cpu_type}, but do not set the architecture type, register usage, or
15339 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
15340 values for @var{cpu_type} are used for @option{-mtune} as for
15341 @option{-mcpu}. If both are specified, the code generated will use the
15342 architecture, registers, and mnemonics set by @option{-mcpu}, but the
15343 scheduling parameters set by @option{-mtune}.
15345 @item -mcmodel=small
15346 @opindex mcmodel=small
15347 Generate PowerPC64 code for the small model: The TOC is limited to
15350 @item -mcmodel=medium
15351 @opindex mcmodel=medium
15352 Generate PowerPC64 code for the medium model: The TOC and other static
15353 data may be up to a total of 4G in size.
15355 @item -mcmodel=large
15356 @opindex mcmodel=large
15357 Generate PowerPC64 code for the large model: The TOC may be up to 4G
15358 in size. Other data and code is only limited by the 64-bit address
15362 @itemx -mno-altivec
15364 @opindex mno-altivec
15365 Generate code that uses (does not use) AltiVec instructions, and also
15366 enable the use of built-in functions that allow more direct access to
15367 the AltiVec instruction set. You may also need to set
15368 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
15374 @opindex mno-vrsave
15375 Generate VRSAVE instructions when generating AltiVec code.
15377 @item -mgen-cell-microcode
15378 @opindex mgen-cell-microcode
15379 Generate Cell microcode instructions
15381 @item -mwarn-cell-microcode
15382 @opindex mwarn-cell-microcode
15383 Warning when a Cell microcode instruction is going to emitted. An example
15384 of a Cell microcode instruction is a variable shift.
15387 @opindex msecure-plt
15388 Generate code that allows ld and ld.so to build executables and shared
15389 libraries with non-exec .plt and .got sections. This is a PowerPC
15390 32-bit SYSV ABI option.
15394 Generate code that uses a BSS .plt section that ld.so fills in, and
15395 requires .plt and .got sections that are both writable and executable.
15396 This is a PowerPC 32-bit SYSV ABI option.
15402 This switch enables or disables the generation of ISEL instructions.
15404 @item -misel=@var{yes/no}
15405 This switch has been deprecated. Use @option{-misel} and
15406 @option{-mno-isel} instead.
15412 This switch enables or disables the generation of SPE simd
15418 @opindex mno-paired
15419 This switch enables or disables the generation of PAIRED simd
15422 @item -mspe=@var{yes/no}
15423 This option has been deprecated. Use @option{-mspe} and
15424 @option{-mno-spe} instead.
15430 Generate code that uses (does not use) vector/scalar (VSX)
15431 instructions, and also enable the use of built-in functions that allow
15432 more direct access to the VSX instruction set.
15434 @item -mfloat-gprs=@var{yes/single/double/no}
15435 @itemx -mfloat-gprs
15436 @opindex mfloat-gprs
15437 This switch enables or disables the generation of floating point
15438 operations on the general purpose registers for architectures that
15441 The argument @var{yes} or @var{single} enables the use of
15442 single-precision floating point operations.
15444 The argument @var{double} enables the use of single and
15445 double-precision floating point operations.
15447 The argument @var{no} disables floating point operations on the
15448 general purpose registers.
15450 This option is currently only available on the MPC854x.
15456 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
15457 targets (including GNU/Linux). The 32-bit environment sets int, long
15458 and pointer to 32 bits and generates code that runs on any PowerPC
15459 variant. The 64-bit environment sets int to 32 bits and long and
15460 pointer to 64 bits, and generates code for PowerPC64, as for
15461 @option{-mpowerpc64}.
15464 @itemx -mno-fp-in-toc
15465 @itemx -mno-sum-in-toc
15466 @itemx -mminimal-toc
15468 @opindex mno-fp-in-toc
15469 @opindex mno-sum-in-toc
15470 @opindex mminimal-toc
15471 Modify generation of the TOC (Table Of Contents), which is created for
15472 every executable file. The @option{-mfull-toc} option is selected by
15473 default. In that case, GCC will allocate at least one TOC entry for
15474 each unique non-automatic variable reference in your program. GCC
15475 will also place floating-point constants in the TOC@. However, only
15476 16,384 entries are available in the TOC@.
15478 If you receive a linker error message that saying you have overflowed
15479 the available TOC space, you can reduce the amount of TOC space used
15480 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
15481 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
15482 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
15483 generate code to calculate the sum of an address and a constant at
15484 run-time instead of putting that sum into the TOC@. You may specify one
15485 or both of these options. Each causes GCC to produce very slightly
15486 slower and larger code at the expense of conserving TOC space.
15488 If you still run out of space in the TOC even when you specify both of
15489 these options, specify @option{-mminimal-toc} instead. This option causes
15490 GCC to make only one TOC entry for every file. When you specify this
15491 option, GCC will produce code that is slower and larger but which
15492 uses extremely little TOC space. You may wish to use this option
15493 only on files that contain less frequently executed code.
15499 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
15500 @code{long} type, and the infrastructure needed to support them.
15501 Specifying @option{-maix64} implies @option{-mpowerpc64} and
15502 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
15503 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
15506 @itemx -mno-xl-compat
15507 @opindex mxl-compat
15508 @opindex mno-xl-compat
15509 Produce code that conforms more closely to IBM XL compiler semantics
15510 when using AIX-compatible ABI@. Pass floating-point arguments to
15511 prototyped functions beyond the register save area (RSA) on the stack
15512 in addition to argument FPRs. Do not assume that most significant
15513 double in 128-bit long double value is properly rounded when comparing
15514 values and converting to double. Use XL symbol names for long double
15517 The AIX calling convention was extended but not initially documented to
15518 handle an obscure K&R C case of calling a function that takes the
15519 address of its arguments with fewer arguments than declared. IBM XL
15520 compilers access floating point arguments which do not fit in the
15521 RSA from the stack when a subroutine is compiled without
15522 optimization. Because always storing floating-point arguments on the
15523 stack is inefficient and rarely needed, this option is not enabled by
15524 default and only is necessary when calling subroutines compiled by IBM
15525 XL compilers without optimization.
15529 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
15530 application written to use message passing with special startup code to
15531 enable the application to run. The system must have PE installed in the
15532 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
15533 must be overridden with the @option{-specs=} option to specify the
15534 appropriate directory location. The Parallel Environment does not
15535 support threads, so the @option{-mpe} option and the @option{-pthread}
15536 option are incompatible.
15538 @item -malign-natural
15539 @itemx -malign-power
15540 @opindex malign-natural
15541 @opindex malign-power
15542 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
15543 @option{-malign-natural} overrides the ABI-defined alignment of larger
15544 types, such as floating-point doubles, on their natural size-based boundary.
15545 The option @option{-malign-power} instructs GCC to follow the ABI-specified
15546 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
15548 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
15552 @itemx -mhard-float
15553 @opindex msoft-float
15554 @opindex mhard-float
15555 Generate code that does not use (uses) the floating-point register set.
15556 Software floating point emulation is provided if you use the
15557 @option{-msoft-float} option, and pass the option to GCC when linking.
15559 @item -msingle-float
15560 @itemx -mdouble-float
15561 @opindex msingle-float
15562 @opindex mdouble-float
15563 Generate code for single or double-precision floating point operations.
15564 @option{-mdouble-float} implies @option{-msingle-float}.
15567 @opindex msimple-fpu
15568 Do not generate sqrt and div instructions for hardware floating point unit.
15572 Specify type of floating point unit. Valid values are @var{sp_lite}
15573 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
15574 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
15575 and @var{dp_full} (equivalent to -mdouble-float).
15578 @opindex mxilinx-fpu
15579 Perform optimizations for floating point unit on Xilinx PPC 405/440.
15582 @itemx -mno-multiple
15584 @opindex mno-multiple
15585 Generate code that uses (does not use) the load multiple word
15586 instructions and the store multiple word instructions. These
15587 instructions are generated by default on POWER systems, and not
15588 generated on PowerPC systems. Do not use @option{-mmultiple} on little
15589 endian PowerPC systems, since those instructions do not work when the
15590 processor is in little endian mode. The exceptions are PPC740 and
15591 PPC750 which permit the instructions usage in little endian mode.
15596 @opindex mno-string
15597 Generate code that uses (does not use) the load string instructions
15598 and the store string word instructions to save multiple registers and
15599 do small block moves. These instructions are generated by default on
15600 POWER systems, and not generated on PowerPC systems. Do not use
15601 @option{-mstring} on little endian PowerPC systems, since those
15602 instructions do not work when the processor is in little endian mode.
15603 The exceptions are PPC740 and PPC750 which permit the instructions
15604 usage in little endian mode.
15609 @opindex mno-update
15610 Generate code that uses (does not use) the load or store instructions
15611 that update the base register to the address of the calculated memory
15612 location. These instructions are generated by default. If you use
15613 @option{-mno-update}, there is a small window between the time that the
15614 stack pointer is updated and the address of the previous frame is
15615 stored, which means code that walks the stack frame across interrupts or
15616 signals may get corrupted data.
15618 @item -mavoid-indexed-addresses
15619 @itemx -mno-avoid-indexed-addresses
15620 @opindex mavoid-indexed-addresses
15621 @opindex mno-avoid-indexed-addresses
15622 Generate code that tries to avoid (not avoid) the use of indexed load
15623 or store instructions. These instructions can incur a performance
15624 penalty on Power6 processors in certain situations, such as when
15625 stepping through large arrays that cross a 16M boundary. This option
15626 is enabled by default when targetting Power6 and disabled otherwise.
15629 @itemx -mno-fused-madd
15630 @opindex mfused-madd
15631 @opindex mno-fused-madd
15632 Generate code that uses (does not use) the floating point multiply and
15633 accumulate instructions. These instructions are generated by default
15634 if hardware floating point is used. The machine dependent
15635 @option{-mfused-madd} option is now mapped to the machine independent
15636 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
15637 mapped to @option{-ffp-contract=off}.
15643 Generate code that uses (does not use) the half-word multiply and
15644 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
15645 These instructions are generated by default when targetting those
15652 Generate code that uses (does not use) the string-search @samp{dlmzb}
15653 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
15654 generated by default when targetting those processors.
15656 @item -mno-bit-align
15658 @opindex mno-bit-align
15659 @opindex mbit-align
15660 On System V.4 and embedded PowerPC systems do not (do) force structures
15661 and unions that contain bit-fields to be aligned to the base type of the
15664 For example, by default a structure containing nothing but 8
15665 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
15666 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
15667 the structure would be aligned to a 1 byte boundary and be one byte in
15670 @item -mno-strict-align
15671 @itemx -mstrict-align
15672 @opindex mno-strict-align
15673 @opindex mstrict-align
15674 On System V.4 and embedded PowerPC systems do not (do) assume that
15675 unaligned memory references will be handled by the system.
15677 @item -mrelocatable
15678 @itemx -mno-relocatable
15679 @opindex mrelocatable
15680 @opindex mno-relocatable
15681 Generate code that allows (does not allow) a static executable to be
15682 relocated to a different address at runtime. A simple embedded
15683 PowerPC system loader should relocate the entire contents of
15684 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
15685 a table of 32-bit addresses generated by this option. For this to
15686 work, all objects linked together must be compiled with
15687 @option{-mrelocatable} or @option{-mrelocatable-lib}.
15688 @option{-mrelocatable} code aligns the stack to an 8 byte boundary.
15690 @item -mrelocatable-lib
15691 @itemx -mno-relocatable-lib
15692 @opindex mrelocatable-lib
15693 @opindex mno-relocatable-lib
15694 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
15695 @code{.fixup} section to allow static executables to be relocated at
15696 runtime, but @option{-mrelocatable-lib} does not use the smaller stack
15697 alignment of @option{-mrelocatable}. Objects compiled with
15698 @option{-mrelocatable-lib} may be linked with objects compiled with
15699 any combination of the @option{-mrelocatable} options.
15705 On System V.4 and embedded PowerPC systems do not (do) assume that
15706 register 2 contains a pointer to a global area pointing to the addresses
15707 used in the program.
15710 @itemx -mlittle-endian
15712 @opindex mlittle-endian
15713 On System V.4 and embedded PowerPC systems compile code for the
15714 processor in little endian mode. The @option{-mlittle-endian} option is
15715 the same as @option{-mlittle}.
15718 @itemx -mbig-endian
15720 @opindex mbig-endian
15721 On System V.4 and embedded PowerPC systems compile code for the
15722 processor in big endian mode. The @option{-mbig-endian} option is
15723 the same as @option{-mbig}.
15725 @item -mdynamic-no-pic
15726 @opindex mdynamic-no-pic
15727 On Darwin and Mac OS X systems, compile code so that it is not
15728 relocatable, but that its external references are relocatable. The
15729 resulting code is suitable for applications, but not shared
15732 @item -mprioritize-restricted-insns=@var{priority}
15733 @opindex mprioritize-restricted-insns
15734 This option controls the priority that is assigned to
15735 dispatch-slot restricted instructions during the second scheduling
15736 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
15737 @var{no/highest/second-highest} priority to dispatch slot restricted
15740 @item -msched-costly-dep=@var{dependence_type}
15741 @opindex msched-costly-dep
15742 This option controls which dependences are considered costly
15743 by the target during instruction scheduling. The argument
15744 @var{dependence_type} takes one of the following values:
15745 @var{no}: no dependence is costly,
15746 @var{all}: all dependences are costly,
15747 @var{true_store_to_load}: a true dependence from store to load is costly,
15748 @var{store_to_load}: any dependence from store to load is costly,
15749 @var{number}: any dependence which latency >= @var{number} is costly.
15751 @item -minsert-sched-nops=@var{scheme}
15752 @opindex minsert-sched-nops
15753 This option controls which nop insertion scheme will be used during
15754 the second scheduling pass. The argument @var{scheme} takes one of the
15756 @var{no}: Don't insert nops.
15757 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
15758 according to the scheduler's grouping.
15759 @var{regroup_exact}: Insert nops to force costly dependent insns into
15760 separate groups. Insert exactly as many nops as needed to force an insn
15761 to a new group, according to the estimated processor grouping.
15762 @var{number}: Insert nops to force costly dependent insns into
15763 separate groups. Insert @var{number} nops to force an insn to a new group.
15766 @opindex mcall-sysv
15767 On System V.4 and embedded PowerPC systems compile code using calling
15768 conventions that adheres to the March 1995 draft of the System V
15769 Application Binary Interface, PowerPC processor supplement. This is the
15770 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
15772 @item -mcall-sysv-eabi
15774 @opindex mcall-sysv-eabi
15775 @opindex mcall-eabi
15776 Specify both @option{-mcall-sysv} and @option{-meabi} options.
15778 @item -mcall-sysv-noeabi
15779 @opindex mcall-sysv-noeabi
15780 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
15782 @item -mcall-aixdesc
15784 On System V.4 and embedded PowerPC systems compile code for the AIX
15788 @opindex mcall-linux
15789 On System V.4 and embedded PowerPC systems compile code for the
15790 Linux-based GNU system.
15794 On System V.4 and embedded PowerPC systems compile code for the
15795 Hurd-based GNU system.
15797 @item -mcall-freebsd
15798 @opindex mcall-freebsd
15799 On System V.4 and embedded PowerPC systems compile code for the
15800 FreeBSD operating system.
15802 @item -mcall-netbsd
15803 @opindex mcall-netbsd
15804 On System V.4 and embedded PowerPC systems compile code for the
15805 NetBSD operating system.
15807 @item -mcall-openbsd
15808 @opindex mcall-netbsd
15809 On System V.4 and embedded PowerPC systems compile code for the
15810 OpenBSD operating system.
15812 @item -maix-struct-return
15813 @opindex maix-struct-return
15814 Return all structures in memory (as specified by the AIX ABI)@.
15816 @item -msvr4-struct-return
15817 @opindex msvr4-struct-return
15818 Return structures smaller than 8 bytes in registers (as specified by the
15821 @item -mabi=@var{abi-type}
15823 Extend the current ABI with a particular extension, or remove such extension.
15824 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
15825 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
15829 Extend the current ABI with SPE ABI extensions. This does not change
15830 the default ABI, instead it adds the SPE ABI extensions to the current
15834 @opindex mabi=no-spe
15835 Disable Booke SPE ABI extensions for the current ABI@.
15837 @item -mabi=ibmlongdouble
15838 @opindex mabi=ibmlongdouble
15839 Change the current ABI to use IBM extended precision long double.
15840 This is a PowerPC 32-bit SYSV ABI option.
15842 @item -mabi=ieeelongdouble
15843 @opindex mabi=ieeelongdouble
15844 Change the current ABI to use IEEE extended precision long double.
15845 This is a PowerPC 32-bit Linux ABI option.
15848 @itemx -mno-prototype
15849 @opindex mprototype
15850 @opindex mno-prototype
15851 On System V.4 and embedded PowerPC systems assume that all calls to
15852 variable argument functions are properly prototyped. Otherwise, the
15853 compiler must insert an instruction before every non prototyped call to
15854 set or clear bit 6 of the condition code register (@var{CR}) to
15855 indicate whether floating point values were passed in the floating point
15856 registers in case the function takes a variable arguments. With
15857 @option{-mprototype}, only calls to prototyped variable argument functions
15858 will set or clear the bit.
15862 On embedded PowerPC systems, assume that the startup module is called
15863 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
15864 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
15869 On embedded PowerPC systems, assume that the startup module is called
15870 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
15875 On embedded PowerPC systems, assume that the startup module is called
15876 @file{crt0.o} and the standard C libraries are @file{libads.a} and
15879 @item -myellowknife
15880 @opindex myellowknife
15881 On embedded PowerPC systems, assume that the startup module is called
15882 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
15887 On System V.4 and embedded PowerPC systems, specify that you are
15888 compiling for a VxWorks system.
15892 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
15893 header to indicate that @samp{eabi} extended relocations are used.
15899 On System V.4 and embedded PowerPC systems do (do not) adhere to the
15900 Embedded Applications Binary Interface (eabi) which is a set of
15901 modifications to the System V.4 specifications. Selecting @option{-meabi}
15902 means that the stack is aligned to an 8 byte boundary, a function
15903 @code{__eabi} is called to from @code{main} to set up the eabi
15904 environment, and the @option{-msdata} option can use both @code{r2} and
15905 @code{r13} to point to two separate small data areas. Selecting
15906 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
15907 do not call an initialization function from @code{main}, and the
15908 @option{-msdata} option will only use @code{r13} to point to a single
15909 small data area. The @option{-meabi} option is on by default if you
15910 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
15913 @opindex msdata=eabi
15914 On System V.4 and embedded PowerPC systems, put small initialized
15915 @code{const} global and static data in the @samp{.sdata2} section, which
15916 is pointed to by register @code{r2}. Put small initialized
15917 non-@code{const} global and static data in the @samp{.sdata} section,
15918 which is pointed to by register @code{r13}. Put small uninitialized
15919 global and static data in the @samp{.sbss} section, which is adjacent to
15920 the @samp{.sdata} section. The @option{-msdata=eabi} option is
15921 incompatible with the @option{-mrelocatable} option. The
15922 @option{-msdata=eabi} option also sets the @option{-memb} option.
15925 @opindex msdata=sysv
15926 On System V.4 and embedded PowerPC systems, put small global and static
15927 data in the @samp{.sdata} section, which is pointed to by register
15928 @code{r13}. Put small uninitialized global and static data in the
15929 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
15930 The @option{-msdata=sysv} option is incompatible with the
15931 @option{-mrelocatable} option.
15933 @item -msdata=default
15935 @opindex msdata=default
15937 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
15938 compile code the same as @option{-msdata=eabi}, otherwise compile code the
15939 same as @option{-msdata=sysv}.
15942 @opindex msdata=data
15943 On System V.4 and embedded PowerPC systems, put small global
15944 data in the @samp{.sdata} section. Put small uninitialized global
15945 data in the @samp{.sbss} section. Do not use register @code{r13}
15946 to address small data however. This is the default behavior unless
15947 other @option{-msdata} options are used.
15951 @opindex msdata=none
15953 On embedded PowerPC systems, put all initialized global and static data
15954 in the @samp{.data} section, and all uninitialized data in the
15955 @samp{.bss} section.
15957 @item -mblock-move-inline-limit=@var{num}
15958 @opindex mblock-move-inline-limit
15959 Inline all block moves (such as calls to @code{memcpy} or structure
15960 copies) less than or equal to @var{num} bytes. The minimum value for
15961 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
15962 targets. The default value is target-specific.
15966 @cindex smaller data references (PowerPC)
15967 @cindex .sdata/.sdata2 references (PowerPC)
15968 On embedded PowerPC systems, put global and static items less than or
15969 equal to @var{num} bytes into the small data or bss sections instead of
15970 the normal data or bss section. By default, @var{num} is 8. The
15971 @option{-G @var{num}} switch is also passed to the linker.
15972 All modules should be compiled with the same @option{-G @var{num}} value.
15975 @itemx -mno-regnames
15977 @opindex mno-regnames
15978 On System V.4 and embedded PowerPC systems do (do not) emit register
15979 names in the assembly language output using symbolic forms.
15982 @itemx -mno-longcall
15984 @opindex mno-longcall
15985 By default assume that all calls are far away so that a longer more
15986 expensive calling sequence is required. This is required for calls
15987 further than 32 megabytes (33,554,432 bytes) from the current location.
15988 A short call will be generated if the compiler knows
15989 the call cannot be that far away. This setting can be overridden by
15990 the @code{shortcall} function attribute, or by @code{#pragma
15993 Some linkers are capable of detecting out-of-range calls and generating
15994 glue code on the fly. On these systems, long calls are unnecessary and
15995 generate slower code. As of this writing, the AIX linker can do this,
15996 as can the GNU linker for PowerPC/64. It is planned to add this feature
15997 to the GNU linker for 32-bit PowerPC systems as well.
15999 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
16000 callee, L42'', plus a ``branch island'' (glue code). The two target
16001 addresses represent the callee and the ``branch island''. The
16002 Darwin/PPC linker will prefer the first address and generate a ``bl
16003 callee'' if the PPC ``bl'' instruction will reach the callee directly;
16004 otherwise, the linker will generate ``bl L42'' to call the ``branch
16005 island''. The ``branch island'' is appended to the body of the
16006 calling function; it computes the full 32-bit address of the callee
16009 On Mach-O (Darwin) systems, this option directs the compiler emit to
16010 the glue for every direct call, and the Darwin linker decides whether
16011 to use or discard it.
16013 In the future, we may cause GCC to ignore all longcall specifications
16014 when the linker is known to generate glue.
16016 @item -mtls-markers
16017 @itemx -mno-tls-markers
16018 @opindex mtls-markers
16019 @opindex mno-tls-markers
16020 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
16021 specifying the function argument. The relocation allows ld to
16022 reliably associate function call with argument setup instructions for
16023 TLS optimization, which in turn allows gcc to better schedule the
16028 Adds support for multithreading with the @dfn{pthreads} library.
16029 This option sets flags for both the preprocessor and linker.
16034 This option will enable GCC to use the reciprocal estimate and
16035 reciprocal square root estimate instructions with additional
16036 Newton-Raphson steps to increase precision instead of doing a divide or
16037 square root and divide for floating point arguments. You should use
16038 the @option{-ffast-math} option when using @option{-mrecip} (or at
16039 least @option{-funsafe-math-optimizations},
16040 @option{-finite-math-only}, @option{-freciprocal-math} and
16041 @option{-fno-trapping-math}). Note that while the throughput of the
16042 sequence is generally higher than the throughput of the non-reciprocal
16043 instruction, the precision of the sequence can be decreased by up to 2
16044 ulp (i.e. the inverse of 1.0 equals 0.99999994) for reciprocal square
16047 @item -mrecip=@var{opt}
16048 @opindex mrecip=opt
16049 This option allows to control which reciprocal estimate instructions
16050 may be used. @var{opt} is a comma separated list of options, that may
16051 be preceeded by a @code{!} to invert the option:
16052 @code{all}: enable all estimate instructions,
16053 @code{default}: enable the default instructions, equvalent to @option{-mrecip},
16054 @code{none}: disable all estimate instructions, equivalent to @option{-mno-recip};
16055 @code{div}: enable the reciprocal approximation instructions for both single and double precision;
16056 @code{divf}: enable the single precision reciprocal approximation instructions;
16057 @code{divd}: enable the double precision reciprocal approximation instructions;
16058 @code{rsqrt}: enable the reciprocal square root approximation instructions for both single and double precision;
16059 @code{rsqrtf}: enable the single precision reciprocal square root approximation instructions;
16060 @code{rsqrtd}: enable the double precision reciprocal square root approximation instructions;
16062 So for example, @option{-mrecip=all,!rsqrtd} would enable the
16063 all of the reciprocal estimate instructions, except for the
16064 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
16065 which handle the double precision reciprocal square root calculations.
16067 @item -mrecip-precision
16068 @itemx -mno-recip-precision
16069 @opindex mrecip-precision
16070 Assume (do not assume) that the reciprocal estimate instructions
16071 provide higher precision estimates than is mandated by the powerpc
16072 ABI. Selecting @option{-mcpu=power6} or @option{-mcpu=power7}
16073 automatically selects @option{-mrecip-precision}. The double
16074 precision square root estimate instructions are not generated by
16075 default on low precision machines, since they do not provide an
16076 estimate that converges after three steps.
16078 @item -mveclibabi=@var{type}
16079 @opindex mveclibabi
16080 Specifies the ABI type to use for vectorizing intrinsics using an
16081 external library. The only type supported at present is @code{mass},
16082 which specifies to use IBM's Mathematical Acceleration Subsystem
16083 (MASS) libraries for vectorizing intrinsics using external libraries.
16084 GCC will currently emit calls to @code{acosd2}, @code{acosf4},
16085 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
16086 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
16087 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
16088 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
16089 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
16090 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
16091 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
16092 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
16093 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
16094 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
16095 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
16096 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
16097 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
16098 for power7. Both @option{-ftree-vectorize} and
16099 @option{-funsafe-math-optimizations} have to be enabled. The MASS
16100 libraries will have to be specified at link time.
16105 Generate (do not generate) the @code{friz} instruction when the
16106 @option{-funsafe-math-optimizations} option is used to optimize
16107 rounding a floating point value to 64-bit integer and back to floating
16108 point. The @code{friz} instruction does not return the same value if
16109 the floating point number is too large to fit in an integer.
16113 @subsection RX Options
16116 These command line options are defined for RX targets:
16119 @item -m64bit-doubles
16120 @itemx -m32bit-doubles
16121 @opindex m64bit-doubles
16122 @opindex m32bit-doubles
16123 Make the @code{double} data type be 64-bits (@option{-m64bit-doubles})
16124 or 32-bits (@option{-m32bit-doubles}) in size. The default is
16125 @option{-m32bit-doubles}. @emph{Note} RX floating point hardware only
16126 works on 32-bit values, which is why the default is
16127 @option{-m32bit-doubles}.
16133 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
16134 floating point hardware. The default is enabled for the @var{RX600}
16135 series and disabled for the @var{RX200} series.
16137 Floating point instructions will only be generated for 32-bit floating
16138 point values however, so if the @option{-m64bit-doubles} option is in
16139 use then the FPU hardware will not be used for doubles.
16141 @emph{Note} If the @option{-fpu} option is enabled then
16142 @option{-funsafe-math-optimizations} is also enabled automatically.
16143 This is because the RX FPU instructions are themselves unsafe.
16145 @item -mcpu=@var{name}
16147 Selects the type of RX CPU to be targeted. Currently three types are
16148 supported, the generic @var{RX600} and @var{RX200} series hardware and
16149 the specific @var{RX610} cpu. The default is @var{RX600}.
16151 The only difference between @var{RX600} and @var{RX610} is that the
16152 @var{RX610} does not support the @code{MVTIPL} instruction.
16154 The @var{RX200} series does not have a hardware floating point unit
16155 and so @option{-nofpu} is enabled by default when this type is
16158 @item -mbig-endian-data
16159 @itemx -mlittle-endian-data
16160 @opindex mbig-endian-data
16161 @opindex mlittle-endian-data
16162 Store data (but not code) in the big-endian format. The default is
16163 @option{-mlittle-endian-data}, ie to store data in the little endian
16166 @item -msmall-data-limit=@var{N}
16167 @opindex msmall-data-limit
16168 Specifies the maximum size in bytes of global and static variables
16169 which can be placed into the small data area. Using the small data
16170 area can lead to smaller and faster code, but the size of area is
16171 limited and it is up to the programmer to ensure that the area does
16172 not overflow. Also when the small data area is used one of the RX's
16173 registers (@code{r13}) is reserved for use pointing to this area, so
16174 it is no longer available for use by the compiler. This could result
16175 in slower and/or larger code if variables which once could have been
16176 held in @code{r13} are now pushed onto the stack.
16178 Note, common variables (variables which have not been initialised) and
16179 constants are not placed into the small data area as they are assigned
16180 to other sections in the output executable.
16182 The default value is zero, which disables this feature. Note, this
16183 feature is not enabled by default with higher optimization levels
16184 (@option{-O2} etc) because of the potentially detrimental effects of
16185 reserving register @code{r13}. It is up to the programmer to
16186 experiment and discover whether this feature is of benefit to their
16193 Use the simulator runtime. The default is to use the libgloss board
16196 @item -mas100-syntax
16197 @itemx -mno-as100-syntax
16198 @opindex mas100-syntax
16199 @opindex mno-as100-syntax
16200 When generating assembler output use a syntax that is compatible with
16201 Renesas's AS100 assembler. This syntax can also be handled by the GAS
16202 assembler but it has some restrictions so generating it is not the
16205 @item -mmax-constant-size=@var{N}
16206 @opindex mmax-constant-size
16207 Specifies the maximum size, in bytes, of a constant that can be used as
16208 an operand in a RX instruction. Although the RX instruction set does
16209 allow constants of up to 4 bytes in length to be used in instructions,
16210 a longer value equates to a longer instruction. Thus in some
16211 circumstances it can be beneficial to restrict the size of constants
16212 that are used in instructions. Constants that are too big are instead
16213 placed into a constant pool and referenced via register indirection.
16215 The value @var{N} can be between 0 and 4. A value of 0 (the default)
16216 or 4 means that constants of any size are allowed.
16220 Enable linker relaxation. Linker relaxation is a process whereby the
16221 linker will attempt to reduce the size of a program by finding shorter
16222 versions of various instructions. Disabled by default.
16224 @item -mint-register=@var{N}
16225 @opindex mint-register
16226 Specify the number of registers to reserve for fast interrupt handler
16227 functions. The value @var{N} can be between 0 and 4. A value of 1
16228 means that register @code{r13} will be reserved for the exclusive use
16229 of fast interrupt handlers. A value of 2 reserves @code{r13} and
16230 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
16231 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
16232 A value of 0, the default, does not reserve any registers.
16234 @item -msave-acc-in-interrupts
16235 @opindex msave-acc-in-interrupts
16236 Specifies that interrupt handler functions should preserve the
16237 accumulator register. This is only necessary if normal code might use
16238 the accumulator register, for example because it performs 64-bit
16239 multiplications. The default is to ignore the accumulator as this
16240 makes the interrupt handlers faster.
16244 @emph{Note:} The generic GCC command line @option{-ffixed-@var{reg}}
16245 has special significance to the RX port when used with the
16246 @code{interrupt} function attribute. This attribute indicates a
16247 function intended to process fast interrupts. GCC will will ensure
16248 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
16249 and/or @code{r13} and only provided that the normal use of the
16250 corresponding registers have been restricted via the
16251 @option{-ffixed-@var{reg}} or @option{-mint-register} command line
16254 @node S/390 and zSeries Options
16255 @subsection S/390 and zSeries Options
16256 @cindex S/390 and zSeries Options
16258 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
16262 @itemx -msoft-float
16263 @opindex mhard-float
16264 @opindex msoft-float
16265 Use (do not use) the hardware floating-point instructions and registers
16266 for floating-point operations. When @option{-msoft-float} is specified,
16267 functions in @file{libgcc.a} will be used to perform floating-point
16268 operations. When @option{-mhard-float} is specified, the compiler
16269 generates IEEE floating-point instructions. This is the default.
16272 @itemx -mno-hard-dfp
16274 @opindex mno-hard-dfp
16275 Use (do not use) the hardware decimal-floating-point instructions for
16276 decimal-floating-point operations. When @option{-mno-hard-dfp} is
16277 specified, functions in @file{libgcc.a} will be used to perform
16278 decimal-floating-point operations. When @option{-mhard-dfp} is
16279 specified, the compiler generates decimal-floating-point hardware
16280 instructions. This is the default for @option{-march=z9-ec} or higher.
16282 @item -mlong-double-64
16283 @itemx -mlong-double-128
16284 @opindex mlong-double-64
16285 @opindex mlong-double-128
16286 These switches control the size of @code{long double} type. A size
16287 of 64bit makes the @code{long double} type equivalent to the @code{double}
16288 type. This is the default.
16291 @itemx -mno-backchain
16292 @opindex mbackchain
16293 @opindex mno-backchain
16294 Store (do not store) the address of the caller's frame as backchain pointer
16295 into the callee's stack frame.
16296 A backchain may be needed to allow debugging using tools that do not understand
16297 DWARF-2 call frame information.
16298 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
16299 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
16300 the backchain is placed into the topmost word of the 96/160 byte register
16303 In general, code compiled with @option{-mbackchain} is call-compatible with
16304 code compiled with @option{-mmo-backchain}; however, use of the backchain
16305 for debugging purposes usually requires that the whole binary is built with
16306 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
16307 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
16308 to build a linux kernel use @option{-msoft-float}.
16310 The default is to not maintain the backchain.
16312 @item -mpacked-stack
16313 @itemx -mno-packed-stack
16314 @opindex mpacked-stack
16315 @opindex mno-packed-stack
16316 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
16317 specified, the compiler uses the all fields of the 96/160 byte register save
16318 area only for their default purpose; unused fields still take up stack space.
16319 When @option{-mpacked-stack} is specified, register save slots are densely
16320 packed at the top of the register save area; unused space is reused for other
16321 purposes, allowing for more efficient use of the available stack space.
16322 However, when @option{-mbackchain} is also in effect, the topmost word of
16323 the save area is always used to store the backchain, and the return address
16324 register is always saved two words below the backchain.
16326 As long as the stack frame backchain is not used, code generated with
16327 @option{-mpacked-stack} is call-compatible with code generated with
16328 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
16329 S/390 or zSeries generated code that uses the stack frame backchain at run
16330 time, not just for debugging purposes. Such code is not call-compatible
16331 with code compiled with @option{-mpacked-stack}. Also, note that the
16332 combination of @option{-mbackchain},
16333 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
16334 to build a linux kernel use @option{-msoft-float}.
16336 The default is to not use the packed stack layout.
16339 @itemx -mno-small-exec
16340 @opindex msmall-exec
16341 @opindex mno-small-exec
16342 Generate (or do not generate) code using the @code{bras} instruction
16343 to do subroutine calls.
16344 This only works reliably if the total executable size does not
16345 exceed 64k. The default is to use the @code{basr} instruction instead,
16346 which does not have this limitation.
16352 When @option{-m31} is specified, generate code compliant to the
16353 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
16354 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
16355 particular to generate 64-bit instructions. For the @samp{s390}
16356 targets, the default is @option{-m31}, while the @samp{s390x}
16357 targets default to @option{-m64}.
16363 When @option{-mzarch} is specified, generate code using the
16364 instructions available on z/Architecture.
16365 When @option{-mesa} is specified, generate code using the
16366 instructions available on ESA/390. Note that @option{-mesa} is
16367 not possible with @option{-m64}.
16368 When generating code compliant to the GNU/Linux for S/390 ABI,
16369 the default is @option{-mesa}. When generating code compliant
16370 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
16376 Generate (or do not generate) code using the @code{mvcle} instruction
16377 to perform block moves. When @option{-mno-mvcle} is specified,
16378 use a @code{mvc} loop instead. This is the default unless optimizing for
16385 Print (or do not print) additional debug information when compiling.
16386 The default is to not print debug information.
16388 @item -march=@var{cpu-type}
16390 Generate code that will run on @var{cpu-type}, which is the name of a system
16391 representing a certain processor type. Possible values for
16392 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
16393 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
16394 When generating code using the instructions available on z/Architecture,
16395 the default is @option{-march=z900}. Otherwise, the default is
16396 @option{-march=g5}.
16398 @item -mtune=@var{cpu-type}
16400 Tune to @var{cpu-type} everything applicable about the generated code,
16401 except for the ABI and the set of available instructions.
16402 The list of @var{cpu-type} values is the same as for @option{-march}.
16403 The default is the value used for @option{-march}.
16406 @itemx -mno-tpf-trace
16407 @opindex mtpf-trace
16408 @opindex mno-tpf-trace
16409 Generate code that adds (does not add) in TPF OS specific branches to trace
16410 routines in the operating system. This option is off by default, even
16411 when compiling for the TPF OS@.
16414 @itemx -mno-fused-madd
16415 @opindex mfused-madd
16416 @opindex mno-fused-madd
16417 Generate code that uses (does not use) the floating point multiply and
16418 accumulate instructions. These instructions are generated by default if
16419 hardware floating point is used.
16421 @item -mwarn-framesize=@var{framesize}
16422 @opindex mwarn-framesize
16423 Emit a warning if the current function exceeds the given frame size. Because
16424 this is a compile time check it doesn't need to be a real problem when the program
16425 runs. It is intended to identify functions which most probably cause
16426 a stack overflow. It is useful to be used in an environment with limited stack
16427 size e.g.@: the linux kernel.
16429 @item -mwarn-dynamicstack
16430 @opindex mwarn-dynamicstack
16431 Emit a warning if the function calls alloca or uses dynamically
16432 sized arrays. This is generally a bad idea with a limited stack size.
16434 @item -mstack-guard=@var{stack-guard}
16435 @itemx -mstack-size=@var{stack-size}
16436 @opindex mstack-guard
16437 @opindex mstack-size
16438 If these options are provided the s390 back end emits additional instructions in
16439 the function prologue which trigger a trap if the stack size is @var{stack-guard}
16440 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
16441 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
16442 the frame size of the compiled function is chosen.
16443 These options are intended to be used to help debugging stack overflow problems.
16444 The additionally emitted code causes only little overhead and hence can also be
16445 used in production like systems without greater performance degradation. The given
16446 values have to be exact powers of 2 and @var{stack-size} has to be greater than
16447 @var{stack-guard} without exceeding 64k.
16448 In order to be efficient the extra code makes the assumption that the stack starts
16449 at an address aligned to the value given by @var{stack-size}.
16450 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
16453 @node Score Options
16454 @subsection Score Options
16455 @cindex Score Options
16457 These options are defined for Score implementations:
16462 Compile code for big endian mode. This is the default.
16466 Compile code for little endian mode.
16470 Disable generate bcnz instruction.
16474 Enable generate unaligned load and store instruction.
16478 Enable the use of multiply-accumulate instructions. Disabled by default.
16482 Specify the SCORE5 as the target architecture.
16486 Specify the SCORE5U of the target architecture.
16490 Specify the SCORE7 as the target architecture. This is the default.
16494 Specify the SCORE7D as the target architecture.
16498 @subsection SH Options
16500 These @samp{-m} options are defined for the SH implementations:
16505 Generate code for the SH1.
16509 Generate code for the SH2.
16512 Generate code for the SH2e.
16516 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
16517 that the floating-point unit is not used.
16519 @item -m2a-single-only
16520 @opindex m2a-single-only
16521 Generate code for the SH2a-FPU, in such a way that no double-precision
16522 floating point operations are used.
16525 @opindex m2a-single
16526 Generate code for the SH2a-FPU assuming the floating-point unit is in
16527 single-precision mode by default.
16531 Generate code for the SH2a-FPU assuming the floating-point unit is in
16532 double-precision mode by default.
16536 Generate code for the SH3.
16540 Generate code for the SH3e.
16544 Generate code for the SH4 without a floating-point unit.
16546 @item -m4-single-only
16547 @opindex m4-single-only
16548 Generate code for the SH4 with a floating-point unit that only
16549 supports single-precision arithmetic.
16553 Generate code for the SH4 assuming the floating-point unit is in
16554 single-precision mode by default.
16558 Generate code for the SH4.
16562 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
16563 floating-point unit is not used.
16565 @item -m4a-single-only
16566 @opindex m4a-single-only
16567 Generate code for the SH4a, in such a way that no double-precision
16568 floating point operations are used.
16571 @opindex m4a-single
16572 Generate code for the SH4a assuming the floating-point unit is in
16573 single-precision mode by default.
16577 Generate code for the SH4a.
16581 Same as @option{-m4a-nofpu}, except that it implicitly passes
16582 @option{-dsp} to the assembler. GCC doesn't generate any DSP
16583 instructions at the moment.
16587 Compile code for the processor in big endian mode.
16591 Compile code for the processor in little endian mode.
16595 Align doubles at 64-bit boundaries. Note that this changes the calling
16596 conventions, and thus some functions from the standard C library will
16597 not work unless you recompile it first with @option{-mdalign}.
16601 Shorten some address references at link time, when possible; uses the
16602 linker option @option{-relax}.
16606 Use 32-bit offsets in @code{switch} tables. The default is to use
16611 Enable the use of bit manipulation instructions on SH2A.
16615 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
16616 alignment constraints.
16620 Comply with the calling conventions defined by Renesas.
16624 Comply with the calling conventions defined by Renesas.
16628 Comply with the calling conventions defined for GCC before the Renesas
16629 conventions were available. This option is the default for all
16630 targets of the SH toolchain except for @samp{sh-symbianelf}.
16633 @opindex mnomacsave
16634 Mark the @code{MAC} register as call-clobbered, even if
16635 @option{-mhitachi} is given.
16639 Increase IEEE-compliance of floating-point code.
16640 At the moment, this is equivalent to @option{-fno-finite-math-only}.
16641 When generating 16 bit SH opcodes, getting IEEE-conforming results for
16642 comparisons of NANs / infinities incurs extra overhead in every
16643 floating point comparison, therefore the default is set to
16644 @option{-ffinite-math-only}.
16646 @item -minline-ic_invalidate
16647 @opindex minline-ic_invalidate
16648 Inline code to invalidate instruction cache entries after setting up
16649 nested function trampolines.
16650 This option has no effect if -musermode is in effect and the selected
16651 code generation option (e.g. -m4) does not allow the use of the icbi
16653 If the selected code generation option does not allow the use of the icbi
16654 instruction, and -musermode is not in effect, the inlined code will
16655 manipulate the instruction cache address array directly with an associative
16656 write. This not only requires privileged mode, but it will also
16657 fail if the cache line had been mapped via the TLB and has become unmapped.
16661 Dump instruction size and location in the assembly code.
16664 @opindex mpadstruct
16665 This option is deprecated. It pads structures to multiple of 4 bytes,
16666 which is incompatible with the SH ABI@.
16670 Optimize for space instead of speed. Implied by @option{-Os}.
16673 @opindex mprefergot
16674 When generating position-independent code, emit function calls using
16675 the Global Offset Table instead of the Procedure Linkage Table.
16679 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
16680 if the inlined code would not work in user mode.
16681 This is the default when the target is @code{sh-*-linux*}.
16683 @item -multcost=@var{number}
16684 @opindex multcost=@var{number}
16685 Set the cost to assume for a multiply insn.
16687 @item -mdiv=@var{strategy}
16688 @opindex mdiv=@var{strategy}
16689 Set the division strategy to use for SHmedia code. @var{strategy} must be
16690 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
16691 inv:call2, inv:fp .
16692 "fp" performs the operation in floating point. This has a very high latency,
16693 but needs only a few instructions, so it might be a good choice if
16694 your code has enough easily exploitable ILP to allow the compiler to
16695 schedule the floating point instructions together with other instructions.
16696 Division by zero causes a floating point exception.
16697 "inv" uses integer operations to calculate the inverse of the divisor,
16698 and then multiplies the dividend with the inverse. This strategy allows
16699 cse and hoisting of the inverse calculation. Division by zero calculates
16700 an unspecified result, but does not trap.
16701 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
16702 have been found, or if the entire operation has been hoisted to the same
16703 place, the last stages of the inverse calculation are intertwined with the
16704 final multiply to reduce the overall latency, at the expense of using a few
16705 more instructions, and thus offering fewer scheduling opportunities with
16707 "call" calls a library function that usually implements the inv:minlat
16709 This gives high code density for m5-*media-nofpu compilations.
16710 "call2" uses a different entry point of the same library function, where it
16711 assumes that a pointer to a lookup table has already been set up, which
16712 exposes the pointer load to cse / code hoisting optimizations.
16713 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
16714 code generation, but if the code stays unoptimized, revert to the "call",
16715 "call2", or "fp" strategies, respectively. Note that the
16716 potentially-trapping side effect of division by zero is carried by a
16717 separate instruction, so it is possible that all the integer instructions
16718 are hoisted out, but the marker for the side effect stays where it is.
16719 A recombination to fp operations or a call is not possible in that case.
16720 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
16721 that the inverse calculation was nor separated from the multiply, they speed
16722 up division where the dividend fits into 20 bits (plus sign where applicable),
16723 by inserting a test to skip a number of operations in this case; this test
16724 slows down the case of larger dividends. inv20u assumes the case of a such
16725 a small dividend to be unlikely, and inv20l assumes it to be likely.
16727 @item -maccumulate-outgoing-args
16728 @opindex maccumulate-outgoing-args
16729 Reserve space once for outgoing arguments in the function prologue rather
16730 than around each call. Generally beneficial for performance and size. Also
16731 needed for unwinding to avoid changing the stack frame around conditional code.
16733 @item -mdivsi3_libfunc=@var{name}
16734 @opindex mdivsi3_libfunc=@var{name}
16735 Set the name of the library function used for 32 bit signed division to
16736 @var{name}. This only affect the name used in the call and inv:call
16737 division strategies, and the compiler will still expect the same
16738 sets of input/output/clobbered registers as if this option was not present.
16740 @item -mfixed-range=@var{register-range}
16741 @opindex mfixed-range
16742 Generate code treating the given register range as fixed registers.
16743 A fixed register is one that the register allocator can not use. This is
16744 useful when compiling kernel code. A register range is specified as
16745 two registers separated by a dash. Multiple register ranges can be
16746 specified separated by a comma.
16748 @item -madjust-unroll
16749 @opindex madjust-unroll
16750 Throttle unrolling to avoid thrashing target registers.
16751 This option only has an effect if the gcc code base supports the
16752 TARGET_ADJUST_UNROLL_MAX target hook.
16754 @item -mindexed-addressing
16755 @opindex mindexed-addressing
16756 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
16757 This is only safe if the hardware and/or OS implement 32 bit wrap-around
16758 semantics for the indexed addressing mode. The architecture allows the
16759 implementation of processors with 64 bit MMU, which the OS could use to
16760 get 32 bit addressing, but since no current hardware implementation supports
16761 this or any other way to make the indexed addressing mode safe to use in
16762 the 32 bit ABI, the default is -mno-indexed-addressing.
16764 @item -mgettrcost=@var{number}
16765 @opindex mgettrcost=@var{number}
16766 Set the cost assumed for the gettr instruction to @var{number}.
16767 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
16771 Assume pt* instructions won't trap. This will generally generate better
16772 scheduled code, but is unsafe on current hardware. The current architecture
16773 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
16774 This has the unintentional effect of making it unsafe to schedule ptabs /
16775 ptrel before a branch, or hoist it out of a loop. For example,
16776 __do_global_ctors, a part of libgcc that runs constructors at program
16777 startup, calls functions in a list which is delimited by @minus{}1. With the
16778 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
16779 That means that all the constructors will be run a bit quicker, but when
16780 the loop comes to the end of the list, the program crashes because ptabs
16781 loads @minus{}1 into a target register. Since this option is unsafe for any
16782 hardware implementing the current architecture specification, the default
16783 is -mno-pt-fixed. Unless the user specifies a specific cost with
16784 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
16785 this deters register allocation using target registers for storing
16788 @item -minvalid-symbols
16789 @opindex minvalid-symbols
16790 Assume symbols might be invalid. Ordinary function symbols generated by
16791 the compiler will always be valid to load with movi/shori/ptabs or
16792 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
16793 to generate symbols that will cause ptabs / ptrel to trap.
16794 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
16795 It will then prevent cross-basic-block cse, hoisting and most scheduling
16796 of symbol loads. The default is @option{-mno-invalid-symbols}.
16799 @node Solaris 2 Options
16800 @subsection Solaris 2 Options
16801 @cindex Solaris 2 options
16803 These @samp{-m} options are supported on Solaris 2:
16806 @item -mimpure-text
16807 @opindex mimpure-text
16808 @option{-mimpure-text}, used in addition to @option{-shared}, tells
16809 the compiler to not pass @option{-z text} to the linker when linking a
16810 shared object. Using this option, you can link position-dependent
16811 code into a shared object.
16813 @option{-mimpure-text} suppresses the ``relocations remain against
16814 allocatable but non-writable sections'' linker error message.
16815 However, the necessary relocations will trigger copy-on-write, and the
16816 shared object is not actually shared across processes. Instead of
16817 using @option{-mimpure-text}, you should compile all source code with
16818 @option{-fpic} or @option{-fPIC}.
16822 These switches are supported in addition to the above on Solaris 2:
16827 Add support for multithreading using the Solaris threads library. This
16828 option sets flags for both the preprocessor and linker. This option does
16829 not affect the thread safety of object code produced by the compiler or
16830 that of libraries supplied with it.
16834 Add support for multithreading using the POSIX threads library. This
16835 option sets flags for both the preprocessor and linker. This option does
16836 not affect the thread safety of object code produced by the compiler or
16837 that of libraries supplied with it.
16841 This is a synonym for @option{-pthreads}.
16844 @node SPARC Options
16845 @subsection SPARC Options
16846 @cindex SPARC options
16848 These @samp{-m} options are supported on the SPARC:
16851 @item -mno-app-regs
16853 @opindex mno-app-regs
16855 Specify @option{-mapp-regs} to generate output using the global registers
16856 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
16859 To be fully SVR4 ABI compliant at the cost of some performance loss,
16860 specify @option{-mno-app-regs}. You should compile libraries and system
16861 software with this option.
16864 @itemx -mhard-float
16866 @opindex mhard-float
16867 Generate output containing floating point instructions. This is the
16871 @itemx -msoft-float
16873 @opindex msoft-float
16874 Generate output containing library calls for floating point.
16875 @strong{Warning:} the requisite libraries are not available for all SPARC
16876 targets. Normally the facilities of the machine's usual C compiler are
16877 used, but this cannot be done directly in cross-compilation. You must make
16878 your own arrangements to provide suitable library functions for
16879 cross-compilation. The embedded targets @samp{sparc-*-aout} and
16880 @samp{sparclite-*-*} do provide software floating point support.
16882 @option{-msoft-float} changes the calling convention in the output file;
16883 therefore, it is only useful if you compile @emph{all} of a program with
16884 this option. In particular, you need to compile @file{libgcc.a}, the
16885 library that comes with GCC, with @option{-msoft-float} in order for
16888 @item -mhard-quad-float
16889 @opindex mhard-quad-float
16890 Generate output containing quad-word (long double) floating point
16893 @item -msoft-quad-float
16894 @opindex msoft-quad-float
16895 Generate output containing library calls for quad-word (long double)
16896 floating point instructions. The functions called are those specified
16897 in the SPARC ABI@. This is the default.
16899 As of this writing, there are no SPARC implementations that have hardware
16900 support for the quad-word floating point instructions. They all invoke
16901 a trap handler for one of these instructions, and then the trap handler
16902 emulates the effect of the instruction. Because of the trap handler overhead,
16903 this is much slower than calling the ABI library routines. Thus the
16904 @option{-msoft-quad-float} option is the default.
16906 @item -mno-unaligned-doubles
16907 @itemx -munaligned-doubles
16908 @opindex mno-unaligned-doubles
16909 @opindex munaligned-doubles
16910 Assume that doubles have 8 byte alignment. This is the default.
16912 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
16913 alignment only if they are contained in another type, or if they have an
16914 absolute address. Otherwise, it assumes they have 4 byte alignment.
16915 Specifying this option avoids some rare compatibility problems with code
16916 generated by other compilers. It is not the default because it results
16917 in a performance loss, especially for floating point code.
16919 @item -mno-faster-structs
16920 @itemx -mfaster-structs
16921 @opindex mno-faster-structs
16922 @opindex mfaster-structs
16923 With @option{-mfaster-structs}, the compiler assumes that structures
16924 should have 8 byte alignment. This enables the use of pairs of
16925 @code{ldd} and @code{std} instructions for copies in structure
16926 assignment, in place of twice as many @code{ld} and @code{st} pairs.
16927 However, the use of this changed alignment directly violates the SPARC
16928 ABI@. Thus, it's intended only for use on targets where the developer
16929 acknowledges that their resulting code will not be directly in line with
16930 the rules of the ABI@.
16932 @item -mcpu=@var{cpu_type}
16934 Set the instruction set, register set, and instruction scheduling parameters
16935 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
16936 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
16937 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
16938 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
16939 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
16941 Default instruction scheduling parameters are used for values that select
16942 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
16943 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
16945 Here is a list of each supported architecture and their supported
16950 v8: supersparc, hypersparc
16951 sparclite: f930, f934, sparclite86x
16953 v9: ultrasparc, ultrasparc3, niagara, niagara2
16956 By default (unless configured otherwise), GCC generates code for the V7
16957 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
16958 additionally optimizes it for the Cypress CY7C602 chip, as used in the
16959 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
16960 SPARCStation 1, 2, IPX etc.
16962 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
16963 architecture. The only difference from V7 code is that the compiler emits
16964 the integer multiply and integer divide instructions which exist in SPARC-V8
16965 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
16966 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
16969 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
16970 the SPARC architecture. This adds the integer multiply, integer divide step
16971 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
16972 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
16973 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
16974 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
16975 MB86934 chip, which is the more recent SPARClite with FPU@.
16977 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
16978 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
16979 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
16980 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
16981 optimizes it for the TEMIC SPARClet chip.
16983 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
16984 architecture. This adds 64-bit integer and floating-point move instructions,
16985 3 additional floating-point condition code registers and conditional move
16986 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
16987 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
16988 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
16989 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
16990 @option{-mcpu=niagara}, the compiler additionally optimizes it for
16991 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
16992 additionally optimizes it for Sun UltraSPARC T2 chips.
16994 @item -mtune=@var{cpu_type}
16996 Set the instruction scheduling parameters for machine type
16997 @var{cpu_type}, but do not set the instruction set or register set that the
16998 option @option{-mcpu=@var{cpu_type}} would.
17000 The same values for @option{-mcpu=@var{cpu_type}} can be used for
17001 @option{-mtune=@var{cpu_type}}, but the only useful values are those
17002 that select a particular cpu implementation. Those are @samp{cypress},
17003 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
17004 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
17005 @samp{ultrasparc3}, @samp{niagara}, and @samp{niagara2}.
17010 @opindex mno-v8plus
17011 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
17012 difference from the V8 ABI is that the global and out registers are
17013 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
17014 mode for all SPARC-V9 processors.
17020 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
17021 Visual Instruction Set extensions. The default is @option{-mno-vis}.
17024 These @samp{-m} options are supported in addition to the above
17025 on SPARC-V9 processors in 64-bit environments:
17028 @item -mlittle-endian
17029 @opindex mlittle-endian
17030 Generate code for a processor running in little-endian mode. It is only
17031 available for a few configurations and most notably not on Solaris and Linux.
17037 Generate code for a 32-bit or 64-bit environment.
17038 The 32-bit environment sets int, long and pointer to 32 bits.
17039 The 64-bit environment sets int to 32 bits and long and pointer
17042 @item -mcmodel=medlow
17043 @opindex mcmodel=medlow
17044 Generate code for the Medium/Low code model: 64-bit addresses, programs
17045 must be linked in the low 32 bits of memory. Programs can be statically
17046 or dynamically linked.
17048 @item -mcmodel=medmid
17049 @opindex mcmodel=medmid
17050 Generate code for the Medium/Middle code model: 64-bit addresses, programs
17051 must be linked in the low 44 bits of memory, the text and data segments must
17052 be less than 2GB in size and the data segment must be located within 2GB of
17055 @item -mcmodel=medany
17056 @opindex mcmodel=medany
17057 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
17058 may be linked anywhere in memory, the text and data segments must be less
17059 than 2GB in size and the data segment must be located within 2GB of the
17062 @item -mcmodel=embmedany
17063 @opindex mcmodel=embmedany
17064 Generate code for the Medium/Anywhere code model for embedded systems:
17065 64-bit addresses, the text and data segments must be less than 2GB in
17066 size, both starting anywhere in memory (determined at link time). The
17067 global register %g4 points to the base of the data segment. Programs
17068 are statically linked and PIC is not supported.
17071 @itemx -mno-stack-bias
17072 @opindex mstack-bias
17073 @opindex mno-stack-bias
17074 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
17075 frame pointer if present, are offset by @minus{}2047 which must be added back
17076 when making stack frame references. This is the default in 64-bit mode.
17077 Otherwise, assume no such offset is present.
17081 @subsection SPU Options
17082 @cindex SPU options
17084 These @samp{-m} options are supported on the SPU:
17088 @itemx -merror-reloc
17089 @opindex mwarn-reloc
17090 @opindex merror-reloc
17092 The loader for SPU does not handle dynamic relocations. By default, GCC
17093 will give an error when it generates code that requires a dynamic
17094 relocation. @option{-mno-error-reloc} disables the error,
17095 @option{-mwarn-reloc} will generate a warning instead.
17098 @itemx -munsafe-dma
17100 @opindex munsafe-dma
17102 Instructions which initiate or test completion of DMA must not be
17103 reordered with respect to loads and stores of the memory which is being
17104 accessed. Users typically address this problem using the volatile
17105 keyword, but that can lead to inefficient code in places where the
17106 memory is known to not change. Rather than mark the memory as volatile
17107 we treat the DMA instructions as potentially effecting all memory. With
17108 @option{-munsafe-dma} users must use the volatile keyword to protect
17111 @item -mbranch-hints
17112 @opindex mbranch-hints
17114 By default, GCC will generate a branch hint instruction to avoid
17115 pipeline stalls for always taken or probably taken branches. A hint
17116 will not be generated closer than 8 instructions away from its branch.
17117 There is little reason to disable them, except for debugging purposes,
17118 or to make an object a little bit smaller.
17122 @opindex msmall-mem
17123 @opindex mlarge-mem
17125 By default, GCC generates code assuming that addresses are never larger
17126 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
17127 a full 32 bit address.
17132 By default, GCC links against startup code that assumes the SPU-style
17133 main function interface (which has an unconventional parameter list).
17134 With @option{-mstdmain}, GCC will link your program against startup
17135 code that assumes a C99-style interface to @code{main}, including a
17136 local copy of @code{argv} strings.
17138 @item -mfixed-range=@var{register-range}
17139 @opindex mfixed-range
17140 Generate code treating the given register range as fixed registers.
17141 A fixed register is one that the register allocator can not use. This is
17142 useful when compiling kernel code. A register range is specified as
17143 two registers separated by a dash. Multiple register ranges can be
17144 specified separated by a comma.
17150 Compile code assuming that pointers to the PPU address space accessed
17151 via the @code{__ea} named address space qualifier are either 32 or 64
17152 bits wide. The default is 32 bits. As this is an ABI changing option,
17153 all object code in an executable must be compiled with the same setting.
17155 @item -maddress-space-conversion
17156 @itemx -mno-address-space-conversion
17157 @opindex maddress-space-conversion
17158 @opindex mno-address-space-conversion
17159 Allow/disallow treating the @code{__ea} address space as superset
17160 of the generic address space. This enables explicit type casts
17161 between @code{__ea} and generic pointer as well as implicit
17162 conversions of generic pointers to @code{__ea} pointers. The
17163 default is to allow address space pointer conversions.
17165 @item -mcache-size=@var{cache-size}
17166 @opindex mcache-size
17167 This option controls the version of libgcc that the compiler links to an
17168 executable and selects a software-managed cache for accessing variables
17169 in the @code{__ea} address space with a particular cache size. Possible
17170 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
17171 and @samp{128}. The default cache size is 64KB.
17173 @item -matomic-updates
17174 @itemx -mno-atomic-updates
17175 @opindex matomic-updates
17176 @opindex mno-atomic-updates
17177 This option controls the version of libgcc that the compiler links to an
17178 executable and selects whether atomic updates to the software-managed
17179 cache of PPU-side variables are used. If you use atomic updates, changes
17180 to a PPU variable from SPU code using the @code{__ea} named address space
17181 qualifier will not interfere with changes to other PPU variables residing
17182 in the same cache line from PPU code. If you do not use atomic updates,
17183 such interference may occur; however, writing back cache lines will be
17184 more efficient. The default behavior is to use atomic updates.
17187 @itemx -mdual-nops=@var{n}
17188 @opindex mdual-nops
17189 By default, GCC will insert nops to increase dual issue when it expects
17190 it to increase performance. @var{n} can be a value from 0 to 10. A
17191 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
17192 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
17194 @item -mhint-max-nops=@var{n}
17195 @opindex mhint-max-nops
17196 Maximum number of nops to insert for a branch hint. A branch hint must
17197 be at least 8 instructions away from the branch it is effecting. GCC
17198 will insert up to @var{n} nops to enforce this, otherwise it will not
17199 generate the branch hint.
17201 @item -mhint-max-distance=@var{n}
17202 @opindex mhint-max-distance
17203 The encoding of the branch hint instruction limits the hint to be within
17204 256 instructions of the branch it is effecting. By default, GCC makes
17205 sure it is within 125.
17208 @opindex msafe-hints
17209 Work around a hardware bug which causes the SPU to stall indefinitely.
17210 By default, GCC will insert the @code{hbrp} instruction to make sure
17211 this stall won't happen.
17215 @node System V Options
17216 @subsection Options for System V
17218 These additional options are available on System V Release 4 for
17219 compatibility with other compilers on those systems:
17224 Create a shared object.
17225 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
17229 Identify the versions of each tool used by the compiler, in a
17230 @code{.ident} assembler directive in the output.
17234 Refrain from adding @code{.ident} directives to the output file (this is
17237 @item -YP,@var{dirs}
17239 Search the directories @var{dirs}, and no others, for libraries
17240 specified with @option{-l}.
17242 @item -Ym,@var{dir}
17244 Look in the directory @var{dir} to find the M4 preprocessor.
17245 The assembler uses this option.
17246 @c This is supposed to go with a -Yd for predefined M4 macro files, but
17247 @c the generic assembler that comes with Solaris takes just -Ym.
17251 @subsection V850 Options
17252 @cindex V850 Options
17254 These @samp{-m} options are defined for V850 implementations:
17258 @itemx -mno-long-calls
17259 @opindex mlong-calls
17260 @opindex mno-long-calls
17261 Treat all calls as being far away (near). If calls are assumed to be
17262 far away, the compiler will always load the functions address up into a
17263 register, and call indirect through the pointer.
17269 Do not optimize (do optimize) basic blocks that use the same index
17270 pointer 4 or more times to copy pointer into the @code{ep} register, and
17271 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
17272 option is on by default if you optimize.
17274 @item -mno-prolog-function
17275 @itemx -mprolog-function
17276 @opindex mno-prolog-function
17277 @opindex mprolog-function
17278 Do not use (do use) external functions to save and restore registers
17279 at the prologue and epilogue of a function. The external functions
17280 are slower, but use less code space if more than one function saves
17281 the same number of registers. The @option{-mprolog-function} option
17282 is on by default if you optimize.
17286 Try to make the code as small as possible. At present, this just turns
17287 on the @option{-mep} and @option{-mprolog-function} options.
17289 @item -mtda=@var{n}
17291 Put static or global variables whose size is @var{n} bytes or less into
17292 the tiny data area that register @code{ep} points to. The tiny data
17293 area can hold up to 256 bytes in total (128 bytes for byte references).
17295 @item -msda=@var{n}
17297 Put static or global variables whose size is @var{n} bytes or less into
17298 the small data area that register @code{gp} points to. The small data
17299 area can hold up to 64 kilobytes.
17301 @item -mzda=@var{n}
17303 Put static or global variables whose size is @var{n} bytes or less into
17304 the first 32 kilobytes of memory.
17308 Specify that the target processor is the V850.
17311 @opindex mbig-switch
17312 Generate code suitable for big switch tables. Use this option only if
17313 the assembler/linker complain about out of range branches within a switch
17318 This option will cause r2 and r5 to be used in the code generated by
17319 the compiler. This setting is the default.
17321 @item -mno-app-regs
17322 @opindex mno-app-regs
17323 This option will cause r2 and r5 to be treated as fixed registers.
17327 Specify that the target processor is the V850E2V3. The preprocessor
17328 constants @samp{__v850e2v3__} will be defined if
17329 this option is used.
17333 Specify that the target processor is the V850E2. The preprocessor
17334 constants @samp{__v850e2__} will be defined if
17338 Specify that the target processor is the V850E1. The preprocessor
17339 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
17343 Specify that the target processor is the V850E@. The preprocessor
17344 constant @samp{__v850e__} will be defined if this option is used.
17346 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
17347 nor @option{-mv850e2} nor @option{-mv850e2v3}
17348 are defined then a default target processor will be chosen and the
17349 relevant @samp{__v850*__} preprocessor constant will be defined.
17351 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
17352 defined, regardless of which processor variant is the target.
17354 @item -mdisable-callt
17355 @opindex mdisable-callt
17356 This option will suppress generation of the CALLT instruction for the
17357 v850e, v850e1, v850e2 and v850e2v3 flavors of the v850 architecture. The default is
17358 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
17363 @subsection VAX Options
17364 @cindex VAX options
17366 These @samp{-m} options are defined for the VAX:
17371 Do not output certain jump instructions (@code{aobleq} and so on)
17372 that the Unix assembler for the VAX cannot handle across long
17377 Do output those jump instructions, on the assumption that you
17378 will assemble with the GNU assembler.
17382 Output code for g-format floating point numbers instead of d-format.
17385 @node VxWorks Options
17386 @subsection VxWorks Options
17387 @cindex VxWorks Options
17389 The options in this section are defined for all VxWorks targets.
17390 Options specific to the target hardware are listed with the other
17391 options for that target.
17396 GCC can generate code for both VxWorks kernels and real time processes
17397 (RTPs). This option switches from the former to the latter. It also
17398 defines the preprocessor macro @code{__RTP__}.
17401 @opindex non-static
17402 Link an RTP executable against shared libraries rather than static
17403 libraries. The options @option{-static} and @option{-shared} can
17404 also be used for RTPs (@pxref{Link Options}); @option{-static}
17411 These options are passed down to the linker. They are defined for
17412 compatibility with Diab.
17415 @opindex Xbind-lazy
17416 Enable lazy binding of function calls. This option is equivalent to
17417 @option{-Wl,-z,now} and is defined for compatibility with Diab.
17421 Disable lazy binding of function calls. This option is the default and
17422 is defined for compatibility with Diab.
17425 @node x86-64 Options
17426 @subsection x86-64 Options
17427 @cindex x86-64 options
17429 These are listed under @xref{i386 and x86-64 Options}.
17431 @node i386 and x86-64 Windows Options
17432 @subsection i386 and x86-64 Windows Options
17433 @cindex i386 and x86-64 Windows Options
17435 These additional options are available for Windows targets:
17440 This option is available for Cygwin and MinGW targets. It
17441 specifies that a console application is to be generated, by
17442 instructing the linker to set the PE header subsystem type
17443 required for console applications.
17444 This is the default behavior for Cygwin and MinGW targets.
17448 This option is available for Cygwin and MinGW targets. It
17449 specifies that a DLL - a dynamic link library - is to be
17450 generated, enabling the selection of the required runtime
17451 startup object and entry point.
17453 @item -mnop-fun-dllimport
17454 @opindex mnop-fun-dllimport
17455 This option is available for Cygwin and MinGW targets. It
17456 specifies that the dllimport attribute should be ignored.
17460 This option is available for MinGW targets. It specifies
17461 that MinGW-specific thread support is to be used.
17465 This option is available for mingw-w64 targets. It specifies
17466 that the UNICODE macro is getting pre-defined and that the
17467 unicode capable runtime startup code is chosen.
17471 This option is available for Cygwin and MinGW targets. It
17472 specifies that the typical Windows pre-defined macros are to
17473 be set in the pre-processor, but does not influence the choice
17474 of runtime library/startup code.
17478 This option is available for Cygwin and MinGW targets. It
17479 specifies that a GUI application is to be generated by
17480 instructing the linker to set the PE header subsystem type
17483 @item -fno-set-stack-executable
17484 @opindex fno-set-stack-executable
17485 This option is available for MinGW targets. It specifies that
17486 the executable flag for stack used by nested functions isn't
17487 set. This is necessary for binaries running in kernel mode of
17488 Windows, as there the user32 API, which is used to set executable
17489 privileges, isn't available.
17491 @item -mpe-aligned-commons
17492 @opindex mpe-aligned-commons
17493 This option is available for Cygwin and MinGW targets. It
17494 specifies that the GNU extension to the PE file format that
17495 permits the correct alignment of COMMON variables should be
17496 used when generating code. It will be enabled by default if
17497 GCC detects that the target assembler found during configuration
17498 supports the feature.
17501 See also under @ref{i386 and x86-64 Options} for standard options.
17503 @node Xstormy16 Options
17504 @subsection Xstormy16 Options
17505 @cindex Xstormy16 Options
17507 These options are defined for Xstormy16:
17512 Choose startup files and linker script suitable for the simulator.
17515 @node Xtensa Options
17516 @subsection Xtensa Options
17517 @cindex Xtensa Options
17519 These options are supported for Xtensa targets:
17523 @itemx -mno-const16
17525 @opindex mno-const16
17526 Enable or disable use of @code{CONST16} instructions for loading
17527 constant values. The @code{CONST16} instruction is currently not a
17528 standard option from Tensilica. When enabled, @code{CONST16}
17529 instructions are always used in place of the standard @code{L32R}
17530 instructions. The use of @code{CONST16} is enabled by default only if
17531 the @code{L32R} instruction is not available.
17534 @itemx -mno-fused-madd
17535 @opindex mfused-madd
17536 @opindex mno-fused-madd
17537 Enable or disable use of fused multiply/add and multiply/subtract
17538 instructions in the floating-point option. This has no effect if the
17539 floating-point option is not also enabled. Disabling fused multiply/add
17540 and multiply/subtract instructions forces the compiler to use separate
17541 instructions for the multiply and add/subtract operations. This may be
17542 desirable in some cases where strict IEEE 754-compliant results are
17543 required: the fused multiply add/subtract instructions do not round the
17544 intermediate result, thereby producing results with @emph{more} bits of
17545 precision than specified by the IEEE standard. Disabling fused multiply
17546 add/subtract instructions also ensures that the program output is not
17547 sensitive to the compiler's ability to combine multiply and add/subtract
17550 @item -mserialize-volatile
17551 @itemx -mno-serialize-volatile
17552 @opindex mserialize-volatile
17553 @opindex mno-serialize-volatile
17554 When this option is enabled, GCC inserts @code{MEMW} instructions before
17555 @code{volatile} memory references to guarantee sequential consistency.
17556 The default is @option{-mserialize-volatile}. Use
17557 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
17559 @item -mforce-no-pic
17560 @opindex mforce-no-pic
17561 For targets, like GNU/Linux, where all user-mode Xtensa code must be
17562 position-independent code (PIC), this option disables PIC for compiling
17565 @item -mtext-section-literals
17566 @itemx -mno-text-section-literals
17567 @opindex mtext-section-literals
17568 @opindex mno-text-section-literals
17569 Control the treatment of literal pools. The default is
17570 @option{-mno-text-section-literals}, which places literals in a separate
17571 section in the output file. This allows the literal pool to be placed
17572 in a data RAM/ROM, and it also allows the linker to combine literal
17573 pools from separate object files to remove redundant literals and
17574 improve code size. With @option{-mtext-section-literals}, the literals
17575 are interspersed in the text section in order to keep them as close as
17576 possible to their references. This may be necessary for large assembly
17579 @item -mtarget-align
17580 @itemx -mno-target-align
17581 @opindex mtarget-align
17582 @opindex mno-target-align
17583 When this option is enabled, GCC instructs the assembler to
17584 automatically align instructions to reduce branch penalties at the
17585 expense of some code density. The assembler attempts to widen density
17586 instructions to align branch targets and the instructions following call
17587 instructions. If there are not enough preceding safe density
17588 instructions to align a target, no widening will be performed. The
17589 default is @option{-mtarget-align}. These options do not affect the
17590 treatment of auto-aligned instructions like @code{LOOP}, which the
17591 assembler will always align, either by widening density instructions or
17592 by inserting no-op instructions.
17595 @itemx -mno-longcalls
17596 @opindex mlongcalls
17597 @opindex mno-longcalls
17598 When this option is enabled, GCC instructs the assembler to translate
17599 direct calls to indirect calls unless it can determine that the target
17600 of a direct call is in the range allowed by the call instruction. This
17601 translation typically occurs for calls to functions in other source
17602 files. Specifically, the assembler translates a direct @code{CALL}
17603 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
17604 The default is @option{-mno-longcalls}. This option should be used in
17605 programs where the call target can potentially be out of range. This
17606 option is implemented in the assembler, not the compiler, so the
17607 assembly code generated by GCC will still show direct call
17608 instructions---look at the disassembled object code to see the actual
17609 instructions. Note that the assembler will use an indirect call for
17610 every cross-file call, not just those that really will be out of range.
17613 @node zSeries Options
17614 @subsection zSeries Options
17615 @cindex zSeries options
17617 These are listed under @xref{S/390 and zSeries Options}.
17619 @node Code Gen Options
17620 @section Options for Code Generation Conventions
17621 @cindex code generation conventions
17622 @cindex options, code generation
17623 @cindex run-time options
17625 These machine-independent options control the interface conventions
17626 used in code generation.
17628 Most of them have both positive and negative forms; the negative form
17629 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
17630 one of the forms is listed---the one which is not the default. You
17631 can figure out the other form by either removing @samp{no-} or adding
17635 @item -fbounds-check
17636 @opindex fbounds-check
17637 For front-ends that support it, generate additional code to check that
17638 indices used to access arrays are within the declared range. This is
17639 currently only supported by the Java and Fortran front-ends, where
17640 this option defaults to true and false respectively.
17644 This option generates traps for signed overflow on addition, subtraction,
17645 multiplication operations.
17649 This option instructs the compiler to assume that signed arithmetic
17650 overflow of addition, subtraction and multiplication wraps around
17651 using twos-complement representation. This flag enables some optimizations
17652 and disables others. This option is enabled by default for the Java
17653 front-end, as required by the Java language specification.
17656 @opindex fexceptions
17657 Enable exception handling. Generates extra code needed to propagate
17658 exceptions. For some targets, this implies GCC will generate frame
17659 unwind information for all functions, which can produce significant data
17660 size overhead, although it does not affect execution. If you do not
17661 specify this option, GCC will enable it by default for languages like
17662 C++ which normally require exception handling, and disable it for
17663 languages like C that do not normally require it. However, you may need
17664 to enable this option when compiling C code that needs to interoperate
17665 properly with exception handlers written in C++. You may also wish to
17666 disable this option if you are compiling older C++ programs that don't
17667 use exception handling.
17669 @item -fnon-call-exceptions
17670 @opindex fnon-call-exceptions
17671 Generate code that allows trapping instructions to throw exceptions.
17672 Note that this requires platform-specific runtime support that does
17673 not exist everywhere. Moreover, it only allows @emph{trapping}
17674 instructions to throw exceptions, i.e.@: memory references or floating
17675 point instructions. It does not allow exceptions to be thrown from
17676 arbitrary signal handlers such as @code{SIGALRM}.
17678 @item -funwind-tables
17679 @opindex funwind-tables
17680 Similar to @option{-fexceptions}, except that it will just generate any needed
17681 static data, but will not affect the generated code in any other way.
17682 You will normally not enable this option; instead, a language processor
17683 that needs this handling would enable it on your behalf.
17685 @item -fasynchronous-unwind-tables
17686 @opindex fasynchronous-unwind-tables
17687 Generate unwind table in dwarf2 format, if supported by target machine. The
17688 table is exact at each instruction boundary, so it can be used for stack
17689 unwinding from asynchronous events (such as debugger or garbage collector).
17691 @item -fpcc-struct-return
17692 @opindex fpcc-struct-return
17693 Return ``short'' @code{struct} and @code{union} values in memory like
17694 longer ones, rather than in registers. This convention is less
17695 efficient, but it has the advantage of allowing intercallability between
17696 GCC-compiled files and files compiled with other compilers, particularly
17697 the Portable C Compiler (pcc).
17699 The precise convention for returning structures in memory depends
17700 on the target configuration macros.
17702 Short structures and unions are those whose size and alignment match
17703 that of some integer type.
17705 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
17706 switch is not binary compatible with code compiled with the
17707 @option{-freg-struct-return} switch.
17708 Use it to conform to a non-default application binary interface.
17710 @item -freg-struct-return
17711 @opindex freg-struct-return
17712 Return @code{struct} and @code{union} values in registers when possible.
17713 This is more efficient for small structures than
17714 @option{-fpcc-struct-return}.
17716 If you specify neither @option{-fpcc-struct-return} nor
17717 @option{-freg-struct-return}, GCC defaults to whichever convention is
17718 standard for the target. If there is no standard convention, GCC
17719 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
17720 the principal compiler. In those cases, we can choose the standard, and
17721 we chose the more efficient register return alternative.
17723 @strong{Warning:} code compiled with the @option{-freg-struct-return}
17724 switch is not binary compatible with code compiled with the
17725 @option{-fpcc-struct-return} switch.
17726 Use it to conform to a non-default application binary interface.
17728 @item -fshort-enums
17729 @opindex fshort-enums
17730 Allocate to an @code{enum} type only as many bytes as it needs for the
17731 declared range of possible values. Specifically, the @code{enum} type
17732 will be equivalent to the smallest integer type which has enough room.
17734 @strong{Warning:} the @option{-fshort-enums} 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-double
17739 @opindex fshort-double
17740 Use the same size for @code{double} as for @code{float}.
17742 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
17743 code that is not binary compatible with code generated without that switch.
17744 Use it to conform to a non-default application binary interface.
17746 @item -fshort-wchar
17747 @opindex fshort-wchar
17748 Override the underlying type for @samp{wchar_t} to be @samp{short
17749 unsigned int} instead of the default for the target. This option is
17750 useful for building programs to run under WINE@.
17752 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
17753 code that is not binary compatible with code generated without that switch.
17754 Use it to conform to a non-default application binary interface.
17757 @opindex fno-common
17758 In C code, controls the placement of uninitialized global variables.
17759 Unix C compilers have traditionally permitted multiple definitions of
17760 such variables in different compilation units by placing the variables
17762 This is the behavior specified by @option{-fcommon}, and is the default
17763 for GCC on most targets.
17764 On the other hand, this behavior is not required by ISO C, and on some
17765 targets may carry a speed or code size penalty on variable references.
17766 The @option{-fno-common} option specifies that the compiler should place
17767 uninitialized global variables in the data section of the object file,
17768 rather than generating them as common blocks.
17769 This has the effect that if the same variable is declared
17770 (without @code{extern}) in two different compilations,
17771 you will get a multiple-definition error when you link them.
17772 In this case, you must compile with @option{-fcommon} instead.
17773 Compiling with @option{-fno-common} is useful on targets for which
17774 it provides better performance, or if you wish to verify that the
17775 program will work on other systems which always treat uninitialized
17776 variable declarations this way.
17780 Ignore the @samp{#ident} directive.
17782 @item -finhibit-size-directive
17783 @opindex finhibit-size-directive
17784 Don't output a @code{.size} assembler directive, or anything else that
17785 would cause trouble if the function is split in the middle, and the
17786 two halves are placed at locations far apart in memory. This option is
17787 used when compiling @file{crtstuff.c}; you should not need to use it
17790 @item -fverbose-asm
17791 @opindex fverbose-asm
17792 Put extra commentary information in the generated assembly code to
17793 make it more readable. This option is generally only of use to those
17794 who actually need to read the generated assembly code (perhaps while
17795 debugging the compiler itself).
17797 @option{-fno-verbose-asm}, the default, causes the
17798 extra information to be omitted and is useful when comparing two assembler
17801 @item -frecord-gcc-switches
17802 @opindex frecord-gcc-switches
17803 This switch causes the command line that was used to invoke the
17804 compiler to be recorded into the object file that is being created.
17805 This switch is only implemented on some targets and the exact format
17806 of the recording is target and binary file format dependent, but it
17807 usually takes the form of a section containing ASCII text. This
17808 switch is related to the @option{-fverbose-asm} switch, but that
17809 switch only records information in the assembler output file as
17810 comments, so it never reaches the object file.
17814 @cindex global offset table
17816 Generate position-independent code (PIC) suitable for use in a shared
17817 library, if supported for the target machine. Such code accesses all
17818 constant addresses through a global offset table (GOT)@. The dynamic
17819 loader resolves the GOT entries when the program starts (the dynamic
17820 loader is not part of GCC; it is part of the operating system). If
17821 the GOT size for the linked executable exceeds a machine-specific
17822 maximum size, you get an error message from the linker indicating that
17823 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
17824 instead. (These maximums are 8k on the SPARC and 32k
17825 on the m68k and RS/6000. The 386 has no such limit.)
17827 Position-independent code requires special support, and therefore works
17828 only on certain machines. For the 386, GCC supports PIC for System V
17829 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
17830 position-independent.
17832 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
17837 If supported for the target machine, emit position-independent code,
17838 suitable for dynamic linking and avoiding any limit on the size of the
17839 global offset table. This option makes a difference on the m68k,
17840 PowerPC and SPARC@.
17842 Position-independent code requires special support, and therefore works
17843 only on certain machines.
17845 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
17852 These options are similar to @option{-fpic} and @option{-fPIC}, but
17853 generated position independent code can be only linked into executables.
17854 Usually these options are used when @option{-pie} GCC option will be
17855 used during linking.
17857 @option{-fpie} and @option{-fPIE} both define the macros
17858 @code{__pie__} and @code{__PIE__}. The macros have the value 1
17859 for @option{-fpie} and 2 for @option{-fPIE}.
17861 @item -fno-jump-tables
17862 @opindex fno-jump-tables
17863 Do not use jump tables for switch statements even where it would be
17864 more efficient than other code generation strategies. This option is
17865 of use in conjunction with @option{-fpic} or @option{-fPIC} for
17866 building code which forms part of a dynamic linker and cannot
17867 reference the address of a jump table. On some targets, jump tables
17868 do not require a GOT and this option is not needed.
17870 @item -ffixed-@var{reg}
17872 Treat the register named @var{reg} as a fixed register; generated code
17873 should never refer to it (except perhaps as a stack pointer, frame
17874 pointer or in some other fixed role).
17876 @var{reg} must be the name of a register. The register names accepted
17877 are machine-specific and are defined in the @code{REGISTER_NAMES}
17878 macro in the machine description macro file.
17880 This flag does not have a negative form, because it specifies a
17883 @item -fcall-used-@var{reg}
17884 @opindex fcall-used
17885 Treat the register named @var{reg} as an allocable register that is
17886 clobbered by function calls. It may be allocated for temporaries or
17887 variables that do not live across a call. Functions compiled this way
17888 will not save and restore the register @var{reg}.
17890 It is an error to used this flag with the frame pointer or stack pointer.
17891 Use of this flag for other registers that have fixed pervasive roles in
17892 the machine's execution model will produce disastrous results.
17894 This flag does not have a negative form, because it specifies a
17897 @item -fcall-saved-@var{reg}
17898 @opindex fcall-saved
17899 Treat the register named @var{reg} as an allocable register saved by
17900 functions. It may be allocated even for temporaries or variables that
17901 live across a call. Functions compiled this way will save and restore
17902 the register @var{reg} if they use it.
17904 It is an error to used this flag with the frame pointer or stack pointer.
17905 Use of this flag for other registers that have fixed pervasive roles in
17906 the machine's execution model will produce disastrous results.
17908 A different sort of disaster will result from the use of this flag for
17909 a register in which function values may be returned.
17911 This flag does not have a negative form, because it specifies a
17914 @item -fpack-struct[=@var{n}]
17915 @opindex fpack-struct
17916 Without a value specified, pack all structure members together without
17917 holes. When a value is specified (which must be a small power of two), pack
17918 structure members according to this value, representing the maximum
17919 alignment (that is, objects with default alignment requirements larger than
17920 this will be output potentially unaligned at the next fitting location.
17922 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
17923 code that is not binary compatible with code generated without that switch.
17924 Additionally, it makes the code suboptimal.
17925 Use it to conform to a non-default application binary interface.
17927 @item -finstrument-functions
17928 @opindex finstrument-functions
17929 Generate instrumentation calls for entry and exit to functions. Just
17930 after function entry and just before function exit, the following
17931 profiling functions will be called with the address of the current
17932 function and its call site. (On some platforms,
17933 @code{__builtin_return_address} does not work beyond the current
17934 function, so the call site information may not be available to the
17935 profiling functions otherwise.)
17938 void __cyg_profile_func_enter (void *this_fn,
17940 void __cyg_profile_func_exit (void *this_fn,
17944 The first argument is the address of the start of the current function,
17945 which may be looked up exactly in the symbol table.
17947 This instrumentation is also done for functions expanded inline in other
17948 functions. The profiling calls will indicate where, conceptually, the
17949 inline function is entered and exited. This means that addressable
17950 versions of such functions must be available. If all your uses of a
17951 function are expanded inline, this may mean an additional expansion of
17952 code size. If you use @samp{extern inline} in your C code, an
17953 addressable version of such functions must be provided. (This is
17954 normally the case anyways, but if you get lucky and the optimizer always
17955 expands the functions inline, you might have gotten away without
17956 providing static copies.)
17958 A function may be given the attribute @code{no_instrument_function}, in
17959 which case this instrumentation will not be done. This can be used, for
17960 example, for the profiling functions listed above, high-priority
17961 interrupt routines, and any functions from which the profiling functions
17962 cannot safely be called (perhaps signal handlers, if the profiling
17963 routines generate output or allocate memory).
17965 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
17966 @opindex finstrument-functions-exclude-file-list
17968 Set the list of functions that are excluded from instrumentation (see
17969 the description of @code{-finstrument-functions}). If the file that
17970 contains a function definition matches with one of @var{file}, then
17971 that function is not instrumented. The match is done on substrings:
17972 if the @var{file} parameter is a substring of the file name, it is
17973 considered to be a match.
17978 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
17982 will exclude any inline function defined in files whose pathnames
17983 contain @code{/bits/stl} or @code{include/sys}.
17985 If, for some reason, you want to include letter @code{','} in one of
17986 @var{sym}, write @code{'\,'}. For example,
17987 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
17988 (note the single quote surrounding the option).
17990 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
17991 @opindex finstrument-functions-exclude-function-list
17993 This is similar to @code{-finstrument-functions-exclude-file-list},
17994 but this option sets the list of function names to be excluded from
17995 instrumentation. The function name to be matched is its user-visible
17996 name, such as @code{vector<int> blah(const vector<int> &)}, not the
17997 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
17998 match is done on substrings: if the @var{sym} parameter is a substring
17999 of the function name, it is considered to be a match. For C99 and C++
18000 extended identifiers, the function name must be given in UTF-8, not
18001 using universal character names.
18003 @item -fstack-check
18004 @opindex fstack-check
18005 Generate code to verify that you do not go beyond the boundary of the
18006 stack. You should specify this flag if you are running in an
18007 environment with multiple threads, but only rarely need to specify it in
18008 a single-threaded environment since stack overflow is automatically
18009 detected on nearly all systems if there is only one stack.
18011 Note that this switch does not actually cause checking to be done; the
18012 operating system or the language runtime must do that. The switch causes
18013 generation of code to ensure that they see the stack being extended.
18015 You can additionally specify a string parameter: @code{no} means no
18016 checking, @code{generic} means force the use of old-style checking,
18017 @code{specific} means use the best checking method and is equivalent
18018 to bare @option{-fstack-check}.
18020 Old-style checking is a generic mechanism that requires no specific
18021 target support in the compiler but comes with the following drawbacks:
18025 Modified allocation strategy for large objects: they will always be
18026 allocated dynamically if their size exceeds a fixed threshold.
18029 Fixed limit on the size of the static frame of functions: when it is
18030 topped by a particular function, stack checking is not reliable and
18031 a warning is issued by the compiler.
18034 Inefficiency: because of both the modified allocation strategy and the
18035 generic implementation, the performances of the code are hampered.
18038 Note that old-style stack checking is also the fallback method for
18039 @code{specific} if no target support has been added in the compiler.
18041 @item -fstack-limit-register=@var{reg}
18042 @itemx -fstack-limit-symbol=@var{sym}
18043 @itemx -fno-stack-limit
18044 @opindex fstack-limit-register
18045 @opindex fstack-limit-symbol
18046 @opindex fno-stack-limit
18047 Generate code to ensure that the stack does not grow beyond a certain value,
18048 either the value of a register or the address of a symbol. If the stack
18049 would grow beyond the value, a signal is raised. For most targets,
18050 the signal is raised before the stack overruns the boundary, so
18051 it is possible to catch the signal without taking special precautions.
18053 For instance, if the stack starts at absolute address @samp{0x80000000}
18054 and grows downwards, you can use the flags
18055 @option{-fstack-limit-symbol=__stack_limit} and
18056 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
18057 of 128KB@. Note that this may only work with the GNU linker.
18059 @item -fsplit-stack
18060 @opindex fsplit-stack
18061 Generate code to automatically split the stack before it overflows.
18062 The resulting program has a discontiguous stack which can only
18063 overflow if the program is unable to allocate any more memory. This
18064 is most useful when running threaded programs, as it is no longer
18065 necessary to calculate a good stack size to use for each thread. This
18066 is currently only implemented for the i386 and x86_64 backends running
18069 When code compiled with @option{-fsplit-stack} calls code compiled
18070 without @option{-fsplit-stack}, there may not be much stack space
18071 available for the latter code to run. If compiling all code,
18072 including library code, with @option{-fsplit-stack} is not an option,
18073 then the linker can fix up these calls so that the code compiled
18074 without @option{-fsplit-stack} always has a large stack. Support for
18075 this is implemented in the gold linker in GNU binutils release 2.21
18078 @item -fleading-underscore
18079 @opindex fleading-underscore
18080 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
18081 change the way C symbols are represented in the object file. One use
18082 is to help link with legacy assembly code.
18084 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
18085 generate code that is not binary compatible with code generated without that
18086 switch. Use it to conform to a non-default application binary interface.
18087 Not all targets provide complete support for this switch.
18089 @item -ftls-model=@var{model}
18090 @opindex ftls-model
18091 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
18092 The @var{model} argument should be one of @code{global-dynamic},
18093 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
18095 The default without @option{-fpic} is @code{initial-exec}; with
18096 @option{-fpic} the default is @code{global-dynamic}.
18098 @item -fvisibility=@var{default|internal|hidden|protected}
18099 @opindex fvisibility
18100 Set the default ELF image symbol visibility to the specified option---all
18101 symbols will be marked with this unless overridden within the code.
18102 Using this feature can very substantially improve linking and
18103 load times of shared object libraries, produce more optimized
18104 code, provide near-perfect API export and prevent symbol clashes.
18105 It is @strong{strongly} recommended that you use this in any shared objects
18108 Despite the nomenclature, @code{default} always means public ie;
18109 available to be linked against from outside the shared object.
18110 @code{protected} and @code{internal} are pretty useless in real-world
18111 usage so the only other commonly used option will be @code{hidden}.
18112 The default if @option{-fvisibility} isn't specified is
18113 @code{default}, i.e., make every
18114 symbol public---this causes the same behavior as previous versions of
18117 A good explanation of the benefits offered by ensuring ELF
18118 symbols have the correct visibility is given by ``How To Write
18119 Shared Libraries'' by Ulrich Drepper (which can be found at
18120 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
18121 solution made possible by this option to marking things hidden when
18122 the default is public is to make the default hidden and mark things
18123 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
18124 and @code{__attribute__ ((visibility("default")))} instead of
18125 @code{__declspec(dllexport)} you get almost identical semantics with
18126 identical syntax. This is a great boon to those working with
18127 cross-platform projects.
18129 For those adding visibility support to existing code, you may find
18130 @samp{#pragma GCC visibility} of use. This works by you enclosing
18131 the declarations you wish to set visibility for with (for example)
18132 @samp{#pragma GCC visibility push(hidden)} and
18133 @samp{#pragma GCC visibility pop}.
18134 Bear in mind that symbol visibility should be viewed @strong{as
18135 part of the API interface contract} and thus all new code should
18136 always specify visibility when it is not the default ie; declarations
18137 only for use within the local DSO should @strong{always} be marked explicitly
18138 as hidden as so to avoid PLT indirection overheads---making this
18139 abundantly clear also aids readability and self-documentation of the code.
18140 Note that due to ISO C++ specification requirements, operator new and
18141 operator delete must always be of default visibility.
18143 Be aware that headers from outside your project, in particular system
18144 headers and headers from any other library you use, may not be
18145 expecting to be compiled with visibility other than the default. You
18146 may need to explicitly say @samp{#pragma GCC visibility push(default)}
18147 before including any such headers.
18149 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
18150 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
18151 no modifications. However, this means that calls to @samp{extern}
18152 functions with no explicit visibility will use the PLT, so it is more
18153 effective to use @samp{__attribute ((visibility))} and/or
18154 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
18155 declarations should be treated as hidden.
18157 Note that @samp{-fvisibility} does affect C++ vague linkage
18158 entities. This means that, for instance, an exception class that will
18159 be thrown between DSOs must be explicitly marked with default
18160 visibility so that the @samp{type_info} nodes will be unified between
18163 An overview of these techniques, their benefits and how to use them
18164 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
18166 @item -fstrict-volatile-bitfields
18167 @opindex fstrict-volatile-bitfields
18168 This option should be used if accesses to volatile bitfields (or other
18169 structure fields, although the compiler usually honors those types
18170 anyway) should use a single access of the width of the
18171 field's type, aligned to a natural alignment if possible. For
18172 example, targets with memory-mapped peripheral registers might require
18173 all such accesses to be 16 bits wide; with this flag the user could
18174 declare all peripheral bitfields as ``unsigned short'' (assuming short
18175 is 16 bits on these targets) to force GCC to use 16 bit accesses
18176 instead of, perhaps, a more efficient 32 bit access.
18178 If this option is disabled, the compiler will use the most efficient
18179 instruction. In the previous example, that might be a 32-bit load
18180 instruction, even though that will access bytes that do not contain
18181 any portion of the bitfield, or memory-mapped registers unrelated to
18182 the one being updated.
18184 If the target requires strict alignment, and honoring the field
18185 type would require violating this alignment, a warning is issued.
18186 If the field has @code{packed} attribute, the access is done without
18187 honoring the field type. If the field doesn't have @code{packed}
18188 attribute, the access is done honoring the field type. In both cases,
18189 GCC assumes that the user knows something about the target hardware
18190 that it is unaware of.
18192 The default value of this option is determined by the application binary
18193 interface for the target processor.
18199 @node Environment Variables
18200 @section Environment Variables Affecting GCC
18201 @cindex environment variables
18203 @c man begin ENVIRONMENT
18204 This section describes several environment variables that affect how GCC
18205 operates. Some of them work by specifying directories or prefixes to use
18206 when searching for various kinds of files. Some are used to specify other
18207 aspects of the compilation environment.
18209 Note that you can also specify places to search using options such as
18210 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
18211 take precedence over places specified using environment variables, which
18212 in turn take precedence over those specified by the configuration of GCC@.
18213 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
18214 GNU Compiler Collection (GCC) Internals}.
18219 @c @itemx LC_COLLATE
18221 @c @itemx LC_MONETARY
18222 @c @itemx LC_NUMERIC
18227 @c @findex LC_COLLATE
18228 @findex LC_MESSAGES
18229 @c @findex LC_MONETARY
18230 @c @findex LC_NUMERIC
18234 These environment variables control the way that GCC uses
18235 localization information that allow GCC to work with different
18236 national conventions. GCC inspects the locale categories
18237 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
18238 so. These locale categories can be set to any value supported by your
18239 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
18240 Kingdom encoded in UTF-8.
18242 The @env{LC_CTYPE} environment variable specifies character
18243 classification. GCC uses it to determine the character boundaries in
18244 a string; this is needed for some multibyte encodings that contain quote
18245 and escape characters that would otherwise be interpreted as a string
18248 The @env{LC_MESSAGES} environment variable specifies the language to
18249 use in diagnostic messages.
18251 If the @env{LC_ALL} environment variable is set, it overrides the value
18252 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
18253 and @env{LC_MESSAGES} default to the value of the @env{LANG}
18254 environment variable. If none of these variables are set, GCC
18255 defaults to traditional C English behavior.
18259 If @env{TMPDIR} is set, it specifies the directory to use for temporary
18260 files. GCC uses temporary files to hold the output of one stage of
18261 compilation which is to be used as input to the next stage: for example,
18262 the output of the preprocessor, which is the input to the compiler
18265 @item GCC_EXEC_PREFIX
18266 @findex GCC_EXEC_PREFIX
18267 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
18268 names of the subprograms executed by the compiler. No slash is added
18269 when this prefix is combined with the name of a subprogram, but you can
18270 specify a prefix that ends with a slash if you wish.
18272 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
18273 an appropriate prefix to use based on the pathname it was invoked with.
18275 If GCC cannot find the subprogram using the specified prefix, it
18276 tries looking in the usual places for the subprogram.
18278 The default value of @env{GCC_EXEC_PREFIX} is
18279 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
18280 the installed compiler. In many cases @var{prefix} is the value
18281 of @code{prefix} when you ran the @file{configure} script.
18283 Other prefixes specified with @option{-B} take precedence over this prefix.
18285 This prefix is also used for finding files such as @file{crt0.o} that are
18288 In addition, the prefix is used in an unusual way in finding the
18289 directories to search for header files. For each of the standard
18290 directories whose name normally begins with @samp{/usr/local/lib/gcc}
18291 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
18292 replacing that beginning with the specified prefix to produce an
18293 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
18294 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
18295 These alternate directories are searched first; the standard directories
18296 come next. If a standard directory begins with the configured
18297 @var{prefix} then the value of @var{prefix} is replaced by
18298 @env{GCC_EXEC_PREFIX} when looking for header files.
18300 @item COMPILER_PATH
18301 @findex COMPILER_PATH
18302 The value of @env{COMPILER_PATH} is a colon-separated list of
18303 directories, much like @env{PATH}. GCC tries the directories thus
18304 specified when searching for subprograms, if it can't find the
18305 subprograms using @env{GCC_EXEC_PREFIX}.
18308 @findex LIBRARY_PATH
18309 The value of @env{LIBRARY_PATH} is a colon-separated list of
18310 directories, much like @env{PATH}. When configured as a native compiler,
18311 GCC tries the directories thus specified when searching for special
18312 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
18313 using GCC also uses these directories when searching for ordinary
18314 libraries for the @option{-l} option (but directories specified with
18315 @option{-L} come first).
18319 @cindex locale definition
18320 This variable is used to pass locale information to the compiler. One way in
18321 which this information is used is to determine the character set to be used
18322 when character literals, string literals and comments are parsed in C and C++.
18323 When the compiler is configured to allow multibyte characters,
18324 the following values for @env{LANG} are recognized:
18328 Recognize JIS characters.
18330 Recognize SJIS characters.
18332 Recognize EUCJP characters.
18335 If @env{LANG} is not defined, or if it has some other value, then the
18336 compiler will use mblen and mbtowc as defined by the default locale to
18337 recognize and translate multibyte characters.
18341 Some additional environments variables affect the behavior of the
18344 @include cppenv.texi
18348 @node Precompiled Headers
18349 @section Using Precompiled Headers
18350 @cindex precompiled headers
18351 @cindex speed of compilation
18353 Often large projects have many header files that are included in every
18354 source file. The time the compiler takes to process these header files
18355 over and over again can account for nearly all of the time required to
18356 build the project. To make builds faster, GCC allows users to
18357 `precompile' a header file; then, if builds can use the precompiled
18358 header file they will be much faster.
18360 To create a precompiled header file, simply compile it as you would any
18361 other file, if necessary using the @option{-x} option to make the driver
18362 treat it as a C or C++ header file. You will probably want to use a
18363 tool like @command{make} to keep the precompiled header up-to-date when
18364 the headers it contains change.
18366 A precompiled header file will be searched for when @code{#include} is
18367 seen in the compilation. As it searches for the included file
18368 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
18369 compiler looks for a precompiled header in each directory just before it
18370 looks for the include file in that directory. The name searched for is
18371 the name specified in the @code{#include} with @samp{.gch} appended. If
18372 the precompiled header file can't be used, it is ignored.
18374 For instance, if you have @code{#include "all.h"}, and you have
18375 @file{all.h.gch} in the same directory as @file{all.h}, then the
18376 precompiled header file will be used if possible, and the original
18377 header will be used otherwise.
18379 Alternatively, you might decide to put the precompiled header file in a
18380 directory and use @option{-I} to ensure that directory is searched
18381 before (or instead of) the directory containing the original header.
18382 Then, if you want to check that the precompiled header file is always
18383 used, you can put a file of the same name as the original header in this
18384 directory containing an @code{#error} command.
18386 This also works with @option{-include}. So yet another way to use
18387 precompiled headers, good for projects not designed with precompiled
18388 header files in mind, is to simply take most of the header files used by
18389 a project, include them from another header file, precompile that header
18390 file, and @option{-include} the precompiled header. If the header files
18391 have guards against multiple inclusion, they will be skipped because
18392 they've already been included (in the precompiled header).
18394 If you need to precompile the same header file for different
18395 languages, targets, or compiler options, you can instead make a
18396 @emph{directory} named like @file{all.h.gch}, and put each precompiled
18397 header in the directory, perhaps using @option{-o}. It doesn't matter
18398 what you call the files in the directory, every precompiled header in
18399 the directory will be considered. The first precompiled header
18400 encountered in the directory that is valid for this compilation will
18401 be used; they're searched in no particular order.
18403 There are many other possibilities, limited only by your imagination,
18404 good sense, and the constraints of your build system.
18406 A precompiled header file can be used only when these conditions apply:
18410 Only one precompiled header can be used in a particular compilation.
18413 A precompiled header can't be used once the first C token is seen. You
18414 can have preprocessor directives before a precompiled header; you can
18415 even include a precompiled header from inside another header, so long as
18416 there are no C tokens before the @code{#include}.
18419 The precompiled header file must be produced for the same language as
18420 the current compilation. You can't use a C precompiled header for a C++
18424 The precompiled header file must have been produced by the same compiler
18425 binary as the current compilation is using.
18428 Any macros defined before the precompiled header is included must
18429 either be defined in the same way as when the precompiled header was
18430 generated, or must not affect the precompiled header, which usually
18431 means that they don't appear in the precompiled header at all.
18433 The @option{-D} option is one way to define a macro before a
18434 precompiled header is included; using a @code{#define} can also do it.
18435 There are also some options that define macros implicitly, like
18436 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
18439 @item If debugging information is output when using the precompiled
18440 header, using @option{-g} or similar, the same kind of debugging information
18441 must have been output when building the precompiled header. However,
18442 a precompiled header built using @option{-g} can be used in a compilation
18443 when no debugging information is being output.
18445 @item The same @option{-m} options must generally be used when building
18446 and using the precompiled header. @xref{Submodel Options},
18447 for any cases where this rule is relaxed.
18449 @item Each of the following options must be the same when building and using
18450 the precompiled header:
18452 @gccoptlist{-fexceptions}
18455 Some other command-line options starting with @option{-f},
18456 @option{-p}, or @option{-O} must be defined in the same way as when
18457 the precompiled header was generated. At present, it's not clear
18458 which options are safe to change and which are not; the safest choice
18459 is to use exactly the same options when generating and using the
18460 precompiled header. The following are known to be safe:
18462 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
18463 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
18464 -fsched-verbose=<number> -fschedule-insns -fvisibility= @gol
18469 For all of these except the last, the compiler will automatically
18470 ignore the precompiled header if the conditions aren't met. If you
18471 find an option combination that doesn't work and doesn't cause the
18472 precompiled header to be ignored, please consider filing a bug report,
18475 If you do use differing options when generating and using the
18476 precompiled header, the actual behavior will be a mixture of the
18477 behavior for the options. For instance, if you use @option{-g} to
18478 generate the precompiled header but not when using it, you may or may
18479 not get debugging information for routines in the precompiled header.