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 1.2 or
19 any later version published by the Free Software Foundation; with the
20 Invariant Sections being ``GNU General Public License'' and ``Funding
21 Free Software'', the Front-Cover texts being (a) (see below), and with
22 the Back-Cover Texts being (b) (see below). A copy of the license is
23 included in the gfdl(7) man page.
25 (a) The FSF's Front-Cover Text is:
29 (b) The FSF's Back-Cover Text is:
31 You have freedom to copy and modify this GNU Manual, like GNU
32 software. Copies published by the Free Software Foundation raise
33 funds for GNU development.
35 @c Set file name and title for the man page.
37 @settitle GNU project C and C++ compiler
39 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
40 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
41 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
42 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
43 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
44 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
45 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
47 Only the most useful options are listed here; see below for the
48 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
51 gpl(7), gfdl(7), fsf-funding(7),
52 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
53 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
54 @file{ld}, @file{binutils} and @file{gdb}.
57 For instructions on reporting bugs, see
61 See the Info entry for @command{gcc}, or
62 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
63 for contributors to GCC@.
68 @chapter GCC Command Options
69 @cindex GCC command options
70 @cindex command options
71 @cindex options, GCC command
73 @c man begin DESCRIPTION
74 When you invoke GCC, it normally does preprocessing, compilation,
75 assembly and linking. The ``overall options'' allow you to stop this
76 process at an intermediate stage. For example, the @option{-c} option
77 says not to run the linker. Then the output consists of object files
78 output by the assembler.
80 Other options are passed on to one stage of processing. Some options
81 control the preprocessor and others the compiler itself. Yet other
82 options control the assembler and linker; most of these are not
83 documented here, since you rarely need to use any of them.
85 @cindex C compilation options
86 Most of the command line options that you can use with GCC are useful
87 for C programs; when an option is only useful with another language
88 (usually C++), the explanation says so explicitly. If the description
89 for a particular option does not mention a source language, you can use
90 that option with all supported languages.
92 @cindex C++ compilation options
93 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
94 options for compiling C++ programs.
96 @cindex grouping options
97 @cindex options, grouping
98 The @command{gcc} program accepts options and file names as operands. Many
99 options have multi-letter names; therefore multiple single-letter options
100 may @emph{not} be grouped: @option{-dv} is very different from @w{@samp{-d
103 @cindex order of options
104 @cindex options, order
105 You can mix options and other arguments. For the most part, the order
106 you use doesn't matter. Order does matter when you use several
107 options of the same kind; for example, if you specify @option{-L} more
108 than once, the directories are searched in the order specified. Also,
109 the placement of the @option{-l} option is significant.
111 Many options have long names starting with @samp{-f} or with
112 @samp{-W}---for example,
113 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
114 these have both positive and negative forms; the negative form of
115 @option{-ffoo} would be @option{-fno-foo}. This manual documents
116 only one of these two forms, whichever one is not the default.
120 @xref{Option Index}, for an index to GCC's options.
123 * Option Summary:: Brief list of all options, without explanations.
124 * Overall Options:: Controlling the kind of output:
125 an executable, object files, assembler files,
126 or preprocessed source.
127 * Invoking G++:: Compiling C++ programs.
128 * C Dialect Options:: Controlling the variant of C language compiled.
129 * C++ Dialect Options:: Variations on C++.
130 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
132 * Language Independent Options:: Controlling how diagnostics should be
134 * Warning Options:: How picky should the compiler be?
135 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
136 * Optimize Options:: How much optimization?
137 * Preprocessor Options:: Controlling header files and macro definitions.
138 Also, getting dependency information for Make.
139 * Assembler Options:: Passing options to the assembler.
140 * Link Options:: Specifying libraries and so on.
141 * Directory Options:: Where to find header files and libraries.
142 Where to find the compiler executable files.
143 * Spec Files:: How to pass switches to sub-processes.
144 * Target Options:: Running a cross-compiler, or an old version of GCC.
145 * Submodel Options:: Specifying minor hardware or convention variations,
146 such as 68010 vs 68020.
147 * Code Gen Options:: Specifying conventions for function calls, data layout
149 * Environment Variables:: Env vars that affect GCC.
150 * Precompiled Headers:: Compiling a header once, and using it many times.
156 @section Option Summary
158 Here is a summary of all the options, grouped by type. Explanations are
159 in the following sections.
162 @item Overall Options
163 @xref{Overall Options,,Options Controlling the Kind of Output}.
164 @gccoptlist{-c -S -E -o @var{file} -combine -no-canonical-prefixes @gol
165 -pipe -pass-exit-codes @gol
166 -x @var{language} -v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help @gol
167 --version -wrapper@@@var{file} -fplugin=@var{file} -fplugin-arg-@var{name}=@var{arg}}
169 @item C Language Options
170 @xref{C Dialect Options,,Options Controlling C Dialect}.
171 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
172 -aux-info @var{filename} @gol
173 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
174 -fhosted -ffreestanding -fopenmp -fms-extensions @gol
175 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
176 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
177 -fsigned-bitfields -fsigned-char @gol
178 -funsigned-bitfields -funsigned-char}
180 @item C++ Language Options
181 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
182 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
183 -fconserve-space -ffriend-injection @gol
184 -fno-elide-constructors @gol
185 -fno-enforce-eh-specs @gol
186 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
187 -fno-implicit-templates @gol
188 -fno-implicit-inline-templates @gol
189 -fno-implement-inlines -fms-extensions @gol
190 -fno-nonansi-builtins -fno-operator-names @gol
191 -fno-optional-diags -fpermissive @gol
192 -fno-pretty-templates @gol
193 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
194 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
195 -fno-default-inline -fvisibility-inlines-hidden @gol
196 -fvisibility-ms-compat @gol
197 -Wabi -Wconversion-null -Wctor-dtor-privacy @gol
198 -Wnon-virtual-dtor -Wreorder @gol
199 -Weffc++ -Wstrict-null-sentinel @gol
200 -Wno-non-template-friend -Wold-style-cast @gol
201 -Woverloaded-virtual -Wno-pmf-conversions @gol
204 @item Objective-C and Objective-C++ Language Options
205 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
206 Objective-C and Objective-C++ Dialects}.
207 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
208 -fgnu-runtime -fnext-runtime @gol
209 -fno-nil-receivers @gol
210 -fobjc-call-cxx-cdtors @gol
211 -fobjc-direct-dispatch @gol
212 -fobjc-exceptions @gol
214 -freplace-objc-classes @gol
217 -Wassign-intercept @gol
218 -Wno-protocol -Wselector @gol
219 -Wstrict-selector-match @gol
220 -Wundeclared-selector}
222 @item Language Independent Options
223 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
224 @gccoptlist{-fmessage-length=@var{n} @gol
225 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
226 -fdiagnostics-show-option}
228 @item Warning Options
229 @xref{Warning Options,,Options to Request or Suppress Warnings}.
230 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
231 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
232 -Wno-attributes -Wno-builtin-macro-redefined @gol
233 -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual @gol
234 -Wchar-subscripts -Wclobbered -Wcomment @gol
235 -Wconversion -Wcoverage-mismatch -Wno-deprecated @gol
236 -Wno-deprecated-declarations -Wdisabled-optimization @gol
237 -Wno-div-by-zero -Wempty-body -Wenum-compare -Wno-endif-labels @gol
238 -Werror -Werror=* @gol
239 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
240 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
241 -Wformat-security -Wformat-y2k @gol
242 -Wframe-larger-than=@var{len} -Wjump-misses-init -Wignored-qualifiers @gol
243 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
244 -Winit-self -Winline @gol
245 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
246 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
247 -Wlogical-op -Wlong-long @gol
248 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
249 -Wmissing-format-attribute -Wmissing-include-dirs @gol
250 -Wmissing-noreturn -Wno-mudflap @gol
251 -Wno-multichar -Wnonnull -Wno-overflow @gol
252 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
253 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
254 -Wpointer-arith -Wno-pointer-to-int-cast @gol
255 -Wredundant-decls @gol
256 -Wreturn-type -Wsequence-point -Wshadow @gol
257 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
258 -Wstrict-aliasing -Wstrict-aliasing=n @gol
259 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
260 -Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand @gol
261 -Wsystem-headers -Wtrigraphs -Wtype-limits -Wundef -Wuninitialized @gol
262 -Wunknown-pragmas -Wno-pragmas @gol
263 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
264 -Wunused-label -Wunused-parameter -Wno-unused-result -Wunused-value -Wunused-variable @gol
265 -Wvariadic-macros -Wvla @gol
266 -Wvolatile-register-var -Wwrite-strings}
268 @item C and Objective-C-only Warning Options
269 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
270 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
271 -Wold-style-declaration -Wold-style-definition @gol
272 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
273 -Wdeclaration-after-statement -Wpointer-sign}
275 @item Debugging Options
276 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
277 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
278 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
279 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
280 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
281 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
282 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
283 -fdump-statistics @gol
285 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
286 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
287 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
289 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
290 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
291 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
292 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
293 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
294 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
295 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
296 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
297 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
298 -fdump-tree-nrv -fdump-tree-vect @gol
299 -fdump-tree-sink @gol
300 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
301 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
302 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
303 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
304 -ftree-vectorizer-verbose=@var{n} @gol
305 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
306 -fdump-final-insns=@var{file} @gol
307 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
308 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
309 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
310 -fenable-icf-debug @gol
311 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
312 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
313 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
314 -ftest-coverage -ftime-report -fvar-tracking @gol
315 -fvar-tracking-assigments -fvar-tracking-assignments-toggle @gol
316 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
317 -ggdb -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
318 -gvms -gxcoff -gxcoff+ @gol
319 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
320 -fdebug-prefix-map=@var{old}=@var{new} @gol
321 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
322 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
323 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
324 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
325 -print-prog-name=@var{program} -print-search-dirs -Q @gol
326 -print-sysroot -print-sysroot-headers-suffix @gol
327 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
329 @item Optimization Options
330 @xref{Optimize Options,,Options that Control Optimization}.
332 -falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
333 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
334 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
335 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
336 -fcheck-data-deps -fconserve-stack -fcprop-registers -fcrossjumping @gol
337 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules -fcx-limited-range @gol
338 -fdata-sections -fdce -fdce @gol
339 -fdelayed-branch -fdelete-null-pointer-checks -fdse -fdse @gol
340 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffast-math @gol
341 -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
342 -fforward-propagate -ffunction-sections @gol
343 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm @gol
344 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
345 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
346 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg -fipa-pta @gol
347 -fipa-pure-const -fipa-reference -fipa-struct-reorg @gol
348 -fipa-type-escape -fira-algorithm=@var{algorithm} @gol
349 -fira-region=@var{region} -fira-coalesce @gol
350 -fira-loop-pressure -fno-ira-share-save-slots @gol
351 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
352 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
353 -floop-block -floop-interchange -floop-strip-mine -fgraphite-identity @gol
354 -floop-parallelize-all -flto -flto-compression-level -flto-report -fltrans @gol
355 -fltrans-output-list -fmerge-all-constants -fmerge-constants -fmodulo-sched @gol
356 -fmodulo-sched-allow-regmoves -fmove-loop-invariants -fmudflap @gol
357 -fmudflapir -fmudflapth -fno-branch-count-reg -fno-default-inline @gol
358 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
359 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
360 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
361 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
362 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
363 -fpeel-loops -fpredictive-commoning -fprefetch-loop-arrays @gol
364 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
365 -fprofile-generate=@var{path} @gol
366 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
367 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
368 -freorder-blocks-and-partition -freorder-functions @gol
369 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
370 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
371 -fsched-spec-load -fsched-spec-load-dangerous @gol
372 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
373 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
374 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
375 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
376 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
377 -fselective-scheduling -fselective-scheduling2 @gol
378 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
379 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
380 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
381 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer @gol
382 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
383 -ftree-copyrename -ftree-dce @gol
384 -ftree-dominator-opts -ftree-dse -ftree-forwprop -ftree-fre -ftree-loop-im @gol
385 -ftree-phiprop -ftree-loop-distribution @gol
386 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
387 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-pta -ftree-reassoc @gol
388 -ftree-sink -ftree-sra -ftree-switch-conversion @gol
389 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
390 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
391 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
392 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
393 -fwhole-program -fwhopr -fwpa -fuse-linker-plugin @gol
394 --param @var{name}=@var{value}
395 -O -O0 -O1 -O2 -O3 -Os}
397 @item Preprocessor Options
398 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
399 @gccoptlist{-A@var{question}=@var{answer} @gol
400 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
401 -C -dD -dI -dM -dN @gol
402 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
403 -idirafter @var{dir} @gol
404 -include @var{file} -imacros @var{file} @gol
405 -iprefix @var{file} -iwithprefix @var{dir} @gol
406 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
407 -imultilib @var{dir} -isysroot @var{dir} @gol
408 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
409 -P -fworking-directory -remap @gol
410 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
411 -Xpreprocessor @var{option}}
413 @item Assembler Option
414 @xref{Assembler Options,,Passing Options to the Assembler}.
415 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
418 @xref{Link Options,,Options for Linking}.
419 @gccoptlist{@var{object-file-name} -l@var{library} @gol
420 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
421 -s -static -static-libgcc -static-libstdc++ -shared @gol
422 -shared-libgcc -symbolic @gol
423 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
426 @item Directory Options
427 @xref{Directory Options,,Options for Directory Search}.
428 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir}
429 -specs=@var{file} -I- --sysroot=@var{dir}}
432 @c I wrote this xref this way to avoid overfull hbox. -- rms
433 @xref{Target Options}.
434 @gccoptlist{-V @var{version} -b @var{machine}}
436 @item Machine Dependent Options
437 @xref{Submodel Options,,Hardware Models and Configurations}.
438 @c This list is ordered alphanumerically by subsection name.
439 @c Try and put the significant identifier (CPU or system) first,
440 @c so users have a clue at guessing where the ones they want will be.
443 @gccoptlist{-EB -EL @gol
444 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
445 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
448 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
449 -mabi=@var{name} @gol
450 -mapcs-stack-check -mno-apcs-stack-check @gol
451 -mapcs-float -mno-apcs-float @gol
452 -mapcs-reentrant -mno-apcs-reentrant @gol
453 -msched-prolog -mno-sched-prolog @gol
454 -mlittle-endian -mbig-endian -mwords-little-endian @gol
455 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
456 -mfp16-format=@var{name}
457 -mthumb-interwork -mno-thumb-interwork @gol
458 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
459 -mstructure-size-boundary=@var{n} @gol
460 -mabort-on-noreturn @gol
461 -mlong-calls -mno-long-calls @gol
462 -msingle-pic-base -mno-single-pic-base @gol
463 -mpic-register=@var{reg} @gol
464 -mnop-fun-dllimport @gol
465 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
466 -mpoke-function-name @gol
468 -mtpcs-frame -mtpcs-leaf-frame @gol
469 -mcaller-super-interworking -mcallee-super-interworking @gol
471 -mword-relocations @gol
472 -mfix-cortex-m3-ldrd}
475 @gccoptlist{-mmcu=@var{mcu} -mno-interrupts @gol
476 -mcall-prologues -mtiny-stack -mint8}
478 @emph{Blackfin Options}
479 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
480 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
481 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
482 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
483 -mno-id-shared-library -mshared-library-id=@var{n} @gol
484 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
485 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
486 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
490 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
491 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
492 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
493 -mstack-align -mdata-align -mconst-align @gol
494 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
495 -melf -maout -melinux -mlinux -sim -sim2 @gol
496 -mmul-bug-workaround -mno-mul-bug-workaround}
499 @gccoptlist{-mmac -mpush-args}
501 @emph{Darwin Options}
502 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
503 -arch_only -bind_at_load -bundle -bundle_loader @gol
504 -client_name -compatibility_version -current_version @gol
506 -dependency-file -dylib_file -dylinker_install_name @gol
507 -dynamic -dynamiclib -exported_symbols_list @gol
508 -filelist -flat_namespace -force_cpusubtype_ALL @gol
509 -force_flat_namespace -headerpad_max_install_names @gol
511 -image_base -init -install_name -keep_private_externs @gol
512 -multi_module -multiply_defined -multiply_defined_unused @gol
513 -noall_load -no_dead_strip_inits_and_terms @gol
514 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
515 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
516 -private_bundle -read_only_relocs -sectalign @gol
517 -sectobjectsymbols -whyload -seg1addr @gol
518 -sectcreate -sectobjectsymbols -sectorder @gol
519 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
520 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
521 -segprot -segs_read_only_addr -segs_read_write_addr @gol
522 -single_module -static -sub_library -sub_umbrella @gol
523 -twolevel_namespace -umbrella -undefined @gol
524 -unexported_symbols_list -weak_reference_mismatches @gol
525 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
526 -mkernel -mone-byte-bool}
528 @emph{DEC Alpha Options}
529 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
530 -mieee -mieee-with-inexact -mieee-conformant @gol
531 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
532 -mtrap-precision=@var{mode} -mbuild-constants @gol
533 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
534 -mbwx -mmax -mfix -mcix @gol
535 -mfloat-vax -mfloat-ieee @gol
536 -mexplicit-relocs -msmall-data -mlarge-data @gol
537 -msmall-text -mlarge-text @gol
538 -mmemory-latency=@var{time}}
540 @emph{DEC Alpha/VMS Options}
541 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
544 @gccoptlist{-msmall-model -mno-lsim}
547 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
548 -mhard-float -msoft-float @gol
549 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
550 -mdouble -mno-double @gol
551 -mmedia -mno-media -mmuladd -mno-muladd @gol
552 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
553 -mlinked-fp -mlong-calls -malign-labels @gol
554 -mlibrary-pic -macc-4 -macc-8 @gol
555 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
556 -moptimize-membar -mno-optimize-membar @gol
557 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
558 -mvliw-branch -mno-vliw-branch @gol
559 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
560 -mno-nested-cond-exec -mtomcat-stats @gol
564 @emph{GNU/Linux Options}
565 @gccoptlist{-muclibc}
567 @emph{H8/300 Options}
568 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
571 @gccoptlist{-march=@var{architecture-type} @gol
572 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
573 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
574 -mfixed-range=@var{register-range} @gol
575 -mjump-in-delay -mlinker-opt -mlong-calls @gol
576 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
577 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
578 -mno-jump-in-delay -mno-long-load-store @gol
579 -mno-portable-runtime -mno-soft-float @gol
580 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
581 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
582 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
583 -munix=@var{unix-std} -nolibdld -static -threads}
585 @emph{i386 and x86-64 Options}
586 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
587 -mfpmath=@var{unit} @gol
588 -masm=@var{dialect} -mno-fancy-math-387 @gol
589 -mno-fp-ret-in-387 -msoft-float @gol
590 -mno-wide-multiply -mrtd -malign-double @gol
591 -mpreferred-stack-boundary=@var{num}
592 -mincoming-stack-boundary=@var{num}
593 -mcld -mcx16 -msahf -mmovbe -mcrc32 -mrecip @gol
594 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
595 -maes -mpclmul -mfused-madd @gol
596 -msse4a -m3dnow -mpopcnt -mabm -mfma4 -mxop -mlwp @gol
597 -mthreads -mno-align-stringops -minline-all-stringops @gol
598 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
599 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
600 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
601 -mveclibabi=@var{type} -mpc32 -mpc64 -mpc80 -mstackrealign @gol
602 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
603 -mcmodel=@var{code-model} -mabi=@var{name} @gol
604 -m32 -m64 -mlarge-data-threshold=@var{num} @gol
608 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
609 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
610 -mconstant-gp -mauto-pic -mfused-madd @gol
611 -minline-float-divide-min-latency @gol
612 -minline-float-divide-max-throughput @gol
613 -mno-inline-float-divide @gol
614 -minline-int-divide-min-latency @gol
615 -minline-int-divide-max-throughput @gol
616 -mno-inline-int-divide @gol
617 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
618 -mno-inline-sqrt @gol
619 -mdwarf2-asm -mearly-stop-bits @gol
620 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
621 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
622 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
623 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
624 -msched-spec-ldc -msched-spec-control-ldc @gol
625 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
626 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
627 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
628 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
630 @emph{IA-64/VMS Options}
631 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
634 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
635 -msign-extend-enabled -muser-enabled}
637 @emph{M32R/D Options}
638 @gccoptlist{-m32r2 -m32rx -m32r @gol
640 -malign-loops -mno-align-loops @gol
641 -missue-rate=@var{number} @gol
642 -mbranch-cost=@var{number} @gol
643 -mmodel=@var{code-size-model-type} @gol
644 -msdata=@var{sdata-type} @gol
645 -mno-flush-func -mflush-func=@var{name} @gol
646 -mno-flush-trap -mflush-trap=@var{number} @gol
650 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
652 @emph{M680x0 Options}
653 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
654 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
655 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
656 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
657 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
658 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
659 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
660 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
663 @emph{M68hc1x Options}
664 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
665 -mauto-incdec -minmax -mlong-calls -mshort @gol
666 -msoft-reg-count=@var{count}}
669 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
670 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
671 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
672 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
673 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
676 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
677 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
678 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
679 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
683 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
684 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
685 -mips64 -mips64r2 @gol
686 -mips16 -mno-mips16 -mflip-mips16 @gol
687 -minterlink-mips16 -mno-interlink-mips16 @gol
688 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
689 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
690 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
691 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
692 -mfpu=@var{fpu-type} @gol
693 -msmartmips -mno-smartmips @gol
694 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
695 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
696 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
697 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
698 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
699 -membedded-data -mno-embedded-data @gol
700 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
701 -mcode-readable=@var{setting} @gol
702 -msplit-addresses -mno-split-addresses @gol
703 -mexplicit-relocs -mno-explicit-relocs @gol
704 -mcheck-zero-division -mno-check-zero-division @gol
705 -mdivide-traps -mdivide-breaks @gol
706 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
707 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
708 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
709 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
710 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
711 -mflush-func=@var{func} -mno-flush-func @gol
712 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
713 -mfp-exceptions -mno-fp-exceptions @gol
714 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
715 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
718 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
719 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
720 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
721 -mno-base-addresses -msingle-exit -mno-single-exit}
723 @emph{MN10300 Options}
724 @gccoptlist{-mmult-bug -mno-mult-bug @gol
725 -mam33 -mno-am33 @gol
726 -mam33-2 -mno-am33-2 @gol
727 -mreturn-pointer-on-d0 @gol
730 @emph{PDP-11 Options}
731 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
732 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
733 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
734 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
735 -mbranch-expensive -mbranch-cheap @gol
736 -msplit -mno-split -munix-asm -mdec-asm}
738 @emph{picoChip Options}
739 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N}
740 -msymbol-as-address -mno-inefficient-warnings}
742 @emph{PowerPC Options}
743 See RS/6000 and PowerPC Options.
745 @emph{RS/6000 and PowerPC Options}
746 @gccoptlist{-mcpu=@var{cpu-type} @gol
747 -mtune=@var{cpu-type} @gol
748 -mpower -mno-power -mpower2 -mno-power2 @gol
749 -mpowerpc -mpowerpc64 -mno-powerpc @gol
750 -maltivec -mno-altivec @gol
751 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
752 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
753 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
754 -mfprnd -mno-fprnd @gol
755 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
756 -mnew-mnemonics -mold-mnemonics @gol
757 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
758 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
759 -malign-power -malign-natural @gol
760 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
761 -msingle-float -mdouble-float -msimple-fpu @gol
762 -mstring -mno-string -mupdate -mno-update @gol
763 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
764 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
765 -mstrict-align -mno-strict-align -mrelocatable @gol
766 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
767 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
768 -mdynamic-no-pic -maltivec -mswdiv @gol
769 -mprioritize-restricted-insns=@var{priority} @gol
770 -msched-costly-dep=@var{dependence_type} @gol
771 -minsert-sched-nops=@var{scheme} @gol
772 -mcall-sysv -mcall-netbsd @gol
773 -maix-struct-return -msvr4-struct-return @gol
774 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
775 -misel -mno-isel @gol
776 -misel=yes -misel=no @gol
778 -mspe=yes -mspe=no @gol
780 -mgen-cell-microcode -mwarn-cell-microcode @gol
781 -mvrsave -mno-vrsave @gol
782 -mmulhw -mno-mulhw @gol
783 -mdlmzb -mno-dlmzb @gol
784 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
785 -mprototype -mno-prototype @gol
786 -msim -mmvme -mads -myellowknife -memb -msdata @gol
787 -msdata=@var{opt} -mvxworks -G @var{num} -pthread}
790 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
792 -mbig-endian-data -mlittle-endian-data @gol
795 -mas100-syntax -mno-as100-syntax@gol
797 -mmax-constant-size=@gol
799 -msave-acc-in-interrupts}
801 @emph{S/390 and zSeries Options}
802 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
803 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
804 -mlong-double-64 -mlong-double-128 @gol
805 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
806 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
807 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
808 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
809 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
812 @gccoptlist{-meb -mel @gol
816 -mscore5 -mscore5u -mscore7 -mscore7d}
819 @gccoptlist{-m1 -m2 -m2e @gol
820 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
822 -m4-nofpu -m4-single-only -m4-single -m4 @gol
823 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
824 -m5-64media -m5-64media-nofpu @gol
825 -m5-32media -m5-32media-nofpu @gol
826 -m5-compact -m5-compact-nofpu @gol
827 -mb -ml -mdalign -mrelax @gol
828 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
829 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
830 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
831 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
832 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
836 @gccoptlist{-mcpu=@var{cpu-type} @gol
837 -mtune=@var{cpu-type} @gol
838 -mcmodel=@var{code-model} @gol
839 -m32 -m64 -mapp-regs -mno-app-regs @gol
840 -mfaster-structs -mno-faster-structs @gol
841 -mfpu -mno-fpu -mhard-float -msoft-float @gol
842 -mhard-quad-float -msoft-quad-float @gol
843 -mimpure-text -mno-impure-text -mlittle-endian @gol
844 -mstack-bias -mno-stack-bias @gol
845 -munaligned-doubles -mno-unaligned-doubles @gol
846 -mv8plus -mno-v8plus -mvis -mno-vis
847 -threads -pthreads -pthread}
850 @gccoptlist{-mwarn-reloc -merror-reloc @gol
851 -msafe-dma -munsafe-dma @gol
853 -msmall-mem -mlarge-mem -mstdmain @gol
854 -mfixed-range=@var{register-range} @gol
856 -maddress-space-conversion -mno-address-space-conversion @gol
857 -mcache-size=@var{cache-size} @gol
858 -matomic-updates -mno-atomic-updates}
860 @emph{System V Options}
861 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
864 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
865 -mprolog-function -mno-prolog-function -mspace @gol
866 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
867 -mapp-regs -mno-app-regs @gol
868 -mdisable-callt -mno-disable-callt @gol
874 @gccoptlist{-mg -mgnu -munix}
876 @emph{VxWorks Options}
877 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
878 -Xbind-lazy -Xbind-now}
880 @emph{x86-64 Options}
881 See i386 and x86-64 Options.
883 @emph{i386 and x86-64 Windows Options}
884 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll
885 -mnop-fun-dllimport -mthread -municode -mwin32 -mwindows
886 -fno-set-stack-executable}
888 @emph{Xstormy16 Options}
891 @emph{Xtensa Options}
892 @gccoptlist{-mconst16 -mno-const16 @gol
893 -mfused-madd -mno-fused-madd @gol
894 -mserialize-volatile -mno-serialize-volatile @gol
895 -mtext-section-literals -mno-text-section-literals @gol
896 -mtarget-align -mno-target-align @gol
897 -mlongcalls -mno-longcalls}
899 @emph{zSeries Options}
900 See S/390 and zSeries Options.
902 @item Code Generation Options
903 @xref{Code Gen Options,,Options for Code Generation Conventions}.
904 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
905 -ffixed-@var{reg} -fexceptions @gol
906 -fnon-call-exceptions -funwind-tables @gol
907 -fasynchronous-unwind-tables @gol
908 -finhibit-size-directive -finstrument-functions @gol
909 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
910 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
911 -fno-common -fno-ident @gol
912 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
913 -fno-jump-tables @gol
914 -frecord-gcc-switches @gol
915 -freg-struct-return -fshort-enums @gol
916 -fshort-double -fshort-wchar @gol
917 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
918 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
919 -fno-stack-limit -fargument-alias -fargument-noalias @gol
920 -fargument-noalias-global -fargument-noalias-anything @gol
921 -fleading-underscore -ftls-model=@var{model} @gol
922 -ftrapv -fwrapv -fbounds-check @gol
927 * Overall Options:: Controlling the kind of output:
928 an executable, object files, assembler files,
929 or preprocessed source.
930 * C Dialect Options:: Controlling the variant of C language compiled.
931 * C++ Dialect Options:: Variations on C++.
932 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
934 * Language Independent Options:: Controlling how diagnostics should be
936 * Warning Options:: How picky should the compiler be?
937 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
938 * Optimize Options:: How much optimization?
939 * Preprocessor Options:: Controlling header files and macro definitions.
940 Also, getting dependency information for Make.
941 * Assembler Options:: Passing options to the assembler.
942 * Link Options:: Specifying libraries and so on.
943 * Directory Options:: Where to find header files and libraries.
944 Where to find the compiler executable files.
945 * Spec Files:: How to pass switches to sub-processes.
946 * Target Options:: Running a cross-compiler, or an old version of GCC.
949 @node Overall Options
950 @section Options Controlling the Kind of Output
952 Compilation can involve up to four stages: preprocessing, compilation
953 proper, assembly and linking, always in that order. GCC is capable of
954 preprocessing and compiling several files either into several
955 assembler input files, or into one assembler input file; then each
956 assembler input file produces an object file, and linking combines all
957 the object files (those newly compiled, and those specified as input)
958 into an executable file.
960 @cindex file name suffix
961 For any given input file, the file name suffix determines what kind of
966 C source code which must be preprocessed.
969 C source code which should not be preprocessed.
972 C++ source code which should not be preprocessed.
975 Objective-C source code. Note that you must link with the @file{libobjc}
976 library to make an Objective-C program work.
979 Objective-C source code which should not be preprocessed.
983 Objective-C++ source code. Note that you must link with the @file{libobjc}
984 library to make an Objective-C++ program work. Note that @samp{.M} refers
985 to a literal capital M@.
988 Objective-C++ source code which should not be preprocessed.
991 C, C++, Objective-C or Objective-C++ header file to be turned into a
996 @itemx @var{file}.cxx
997 @itemx @var{file}.cpp
998 @itemx @var{file}.CPP
999 @itemx @var{file}.c++
1001 C++ source code which must be preprocessed. Note that in @samp{.cxx},
1002 the last two letters must both be literally @samp{x}. Likewise,
1003 @samp{.C} refers to a literal capital C@.
1007 Objective-C++ source code which must be preprocessed.
1009 @item @var{file}.mii
1010 Objective-C++ source code which should not be preprocessed.
1014 @itemx @var{file}.hp
1015 @itemx @var{file}.hxx
1016 @itemx @var{file}.hpp
1017 @itemx @var{file}.HPP
1018 @itemx @var{file}.h++
1019 @itemx @var{file}.tcc
1020 C++ header file to be turned into a precompiled header.
1023 @itemx @var{file}.for
1024 @itemx @var{file}.ftn
1025 Fixed form Fortran source code which should not be preprocessed.
1028 @itemx @var{file}.FOR
1029 @itemx @var{file}.fpp
1030 @itemx @var{file}.FPP
1031 @itemx @var{file}.FTN
1032 Fixed form Fortran source code which must be preprocessed (with the traditional
1035 @item @var{file}.f90
1036 @itemx @var{file}.f95
1037 @itemx @var{file}.f03
1038 @itemx @var{file}.f08
1039 Free form Fortran source code which should not be preprocessed.
1041 @item @var{file}.F90
1042 @itemx @var{file}.F95
1043 @itemx @var{file}.F03
1044 @itemx @var{file}.F08
1045 Free form Fortran source code which must be preprocessed (with the
1046 traditional preprocessor).
1048 @c FIXME: Descriptions of Java file types.
1054 @item @var{file}.ads
1055 Ada source code file which contains a library unit declaration (a
1056 declaration of a package, subprogram, or generic, or a generic
1057 instantiation), or a library unit renaming declaration (a package,
1058 generic, or subprogram renaming declaration). Such files are also
1061 @item @var{file}.adb
1062 Ada source code file containing a library unit body (a subprogram or
1063 package body). Such files are also called @dfn{bodies}.
1065 @c GCC also knows about some suffixes for languages not yet included:
1076 @itemx @var{file}.sx
1077 Assembler code which must be preprocessed.
1080 An object file to be fed straight into linking.
1081 Any file name with no recognized suffix is treated this way.
1085 You can specify the input language explicitly with the @option{-x} option:
1088 @item -x @var{language}
1089 Specify explicitly the @var{language} for the following input files
1090 (rather than letting the compiler choose a default based on the file
1091 name suffix). This option applies to all following input files until
1092 the next @option{-x} option. Possible values for @var{language} are:
1094 c c-header c-cpp-output
1095 c++ c++-header c++-cpp-output
1096 objective-c objective-c-header objective-c-cpp-output
1097 objective-c++ objective-c++-header objective-c++-cpp-output
1098 assembler assembler-with-cpp
1100 f77 f77-cpp-input f95 f95-cpp-input
1105 Turn off any specification of a language, so that subsequent files are
1106 handled according to their file name suffixes (as they are if @option{-x}
1107 has not been used at all).
1109 @item -pass-exit-codes
1110 @opindex pass-exit-codes
1111 Normally the @command{gcc} program will exit with the code of 1 if any
1112 phase of the compiler returns a non-success return code. If you specify
1113 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1114 numerically highest error produced by any phase that returned an error
1115 indication. The C, C++, and Fortran frontends return 4, if an internal
1116 compiler error is encountered.
1119 If you only want some of the stages of compilation, you can use
1120 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1121 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1122 @command{gcc} is to stop. Note that some combinations (for example,
1123 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1128 Compile or assemble the source files, but do not link. The linking
1129 stage simply is not done. The ultimate output is in the form of an
1130 object file for each source file.
1132 By default, the object file name for a source file is made by replacing
1133 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1135 Unrecognized input files, not requiring compilation or assembly, are
1140 Stop after the stage of compilation proper; do not assemble. The output
1141 is in the form of an assembler code file for each non-assembler input
1144 By default, the assembler file name for a source file is made by
1145 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1147 Input files that don't require compilation are ignored.
1151 Stop after the preprocessing stage; do not run the compiler proper. The
1152 output is in the form of preprocessed source code, which is sent to the
1155 Input files which don't require preprocessing are ignored.
1157 @cindex output file option
1160 Place output in file @var{file}. This applies regardless to whatever
1161 sort of output is being produced, whether it be an executable file,
1162 an object file, an assembler file or preprocessed C code.
1164 If @option{-o} is not specified, the default is to put an executable
1165 file in @file{a.out}, the object file for
1166 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1167 assembler file in @file{@var{source}.s}, a precompiled header file in
1168 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1173 Print (on standard error output) the commands executed to run the stages
1174 of compilation. Also print the version number of the compiler driver
1175 program and of the preprocessor and the compiler proper.
1179 Like @option{-v} except the commands are not executed and all command
1180 arguments are quoted. This is useful for shell scripts to capture the
1181 driver-generated command lines.
1185 Use pipes rather than temporary files for communication between the
1186 various stages of compilation. This fails to work on some systems where
1187 the assembler is unable to read from a pipe; but the GNU assembler has
1192 If you are compiling multiple source files, this option tells the driver
1193 to pass all the source files to the compiler at once (for those
1194 languages for which the compiler can handle this). This will allow
1195 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1196 language for which this is supported is C@. If you pass source files for
1197 multiple languages to the driver, using this option, the driver will invoke
1198 the compiler(s) that support IMA once each, passing each compiler all the
1199 source files appropriate for it. For those languages that do not support
1200 IMA this option will be ignored, and the compiler will be invoked once for
1201 each source file in that language. If you use this option in conjunction
1202 with @option{-save-temps}, the compiler will generate multiple
1204 (one for each source file), but only one (combined) @file{.o} or
1209 Print (on the standard output) a description of the command line options
1210 understood by @command{gcc}. If the @option{-v} option is also specified
1211 then @option{--help} will also be passed on to the various processes
1212 invoked by @command{gcc}, so that they can display the command line options
1213 they accept. If the @option{-Wextra} option has also been specified
1214 (prior to the @option{--help} option), then command line options which
1215 have no documentation associated with them will also be displayed.
1218 @opindex target-help
1219 Print (on the standard output) a description of target-specific command
1220 line options for each tool. For some targets extra target-specific
1221 information may also be printed.
1223 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1224 Print (on the standard output) a description of the command line
1225 options understood by the compiler that fit into all specified classes
1226 and qualifiers. These are the supported classes:
1229 @item @samp{optimizers}
1230 This will display all of the optimization options supported by the
1233 @item @samp{warnings}
1234 This will display all of the options controlling warning messages
1235 produced by the compiler.
1238 This will display target-specific options. Unlike the
1239 @option{--target-help} option however, target-specific options of the
1240 linker and assembler will not be displayed. This is because those
1241 tools do not currently support the extended @option{--help=} syntax.
1244 This will display the values recognized by the @option{--param}
1247 @item @var{language}
1248 This will display the options supported for @var{language}, where
1249 @var{language} is the name of one of the languages supported in this
1253 This will display the options that are common to all languages.
1256 These are the supported qualifiers:
1259 @item @samp{undocumented}
1260 Display only those options which are undocumented.
1263 Display options which take an argument that appears after an equal
1264 sign in the same continuous piece of text, such as:
1265 @samp{--help=target}.
1267 @item @samp{separate}
1268 Display options which take an argument that appears as a separate word
1269 following the original option, such as: @samp{-o output-file}.
1272 Thus for example to display all the undocumented target-specific
1273 switches supported by the compiler the following can be used:
1276 --help=target,undocumented
1279 The sense of a qualifier can be inverted by prefixing it with the
1280 @samp{^} character, so for example to display all binary warning
1281 options (i.e., ones that are either on or off and that do not take an
1282 argument), which have a description the following can be used:
1285 --help=warnings,^joined,^undocumented
1288 The argument to @option{--help=} should not consist solely of inverted
1291 Combining several classes is possible, although this usually
1292 restricts the output by so much that there is nothing to display. One
1293 case where it does work however is when one of the classes is
1294 @var{target}. So for example to display all the target-specific
1295 optimization options the following can be used:
1298 --help=target,optimizers
1301 The @option{--help=} option can be repeated on the command line. Each
1302 successive use will display its requested class of options, skipping
1303 those that have already been displayed.
1305 If the @option{-Q} option appears on the command line before the
1306 @option{--help=} option, then the descriptive text displayed by
1307 @option{--help=} is changed. Instead of describing the displayed
1308 options, an indication is given as to whether the option is enabled,
1309 disabled or set to a specific value (assuming that the compiler
1310 knows this at the point where the @option{--help=} option is used).
1312 Here is a truncated example from the ARM port of @command{gcc}:
1315 % gcc -Q -mabi=2 --help=target -c
1316 The following options are target specific:
1318 -mabort-on-noreturn [disabled]
1322 The output is sensitive to the effects of previous command line
1323 options, so for example it is possible to find out which optimizations
1324 are enabled at @option{-O2} by using:
1327 -Q -O2 --help=optimizers
1330 Alternatively you can discover which binary optimizations are enabled
1331 by @option{-O3} by using:
1334 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1335 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1336 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1339 @item -no-canonical-prefixes
1340 @opindex no-canonical-prefixes
1341 Do not expand any symbolic links, resolve references to @samp{/../}
1342 or @samp{/./}, or make the path absolute when generating a relative
1347 Display the version number and copyrights of the invoked GCC@.
1351 Invoke all subcommands under a wrapper program. It takes a single
1352 comma separated list as an argument, which will be used to invoke
1356 gcc -c t.c -wrapper gdb,--args
1359 This will invoke all subprograms of gcc under "gdb --args",
1360 thus cc1 invocation will be "gdb --args cc1 ...".
1362 @item -fplugin=@var{name}.so
1363 Load the plugin code in file @var{name}.so, assumed to be a
1364 shared object to be dlopen'd by the compiler. The base name of
1365 the shared object file is used to identify the plugin for the
1366 purposes of argument parsing (See
1367 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1368 Each plugin should define the callback functions specified in the
1371 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1372 Define an argument called @var{key} with a value of @var{value}
1373 for the plugin called @var{name}.
1375 @include @value{srcdir}/../libiberty/at-file.texi
1379 @section Compiling C++ Programs
1381 @cindex suffixes for C++ source
1382 @cindex C++ source file suffixes
1383 C++ source files conventionally use one of the suffixes @samp{.C},
1384 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1385 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1386 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1387 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1388 files with these names and compiles them as C++ programs even if you
1389 call the compiler the same way as for compiling C programs (usually
1390 with the name @command{gcc}).
1394 However, the use of @command{gcc} does not add the C++ library.
1395 @command{g++} is a program that calls GCC and treats @samp{.c},
1396 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1397 files unless @option{-x} is used, and automatically specifies linking
1398 against the C++ library. This program is also useful when
1399 precompiling a C header file with a @samp{.h} extension for use in C++
1400 compilations. On many systems, @command{g++} is also installed with
1401 the name @command{c++}.
1403 @cindex invoking @command{g++}
1404 When you compile C++ programs, you may specify many of the same
1405 command-line options that you use for compiling programs in any
1406 language; or command-line options meaningful for C and related
1407 languages; or options that are meaningful only for C++ programs.
1408 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1409 explanations of options for languages related to C@.
1410 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1411 explanations of options that are meaningful only for C++ programs.
1413 @node C Dialect Options
1414 @section Options Controlling C Dialect
1415 @cindex dialect options
1416 @cindex language dialect options
1417 @cindex options, dialect
1419 The following options control the dialect of C (or languages derived
1420 from C, such as C++, Objective-C and Objective-C++) that the compiler
1424 @cindex ANSI support
1428 In C mode, this is equivalent to @samp{-std=c89}. In C++ mode, it is
1429 equivalent to @samp{-std=c++98}.
1431 This turns off certain features of GCC that are incompatible with ISO
1432 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1433 such as the @code{asm} and @code{typeof} keywords, and
1434 predefined macros such as @code{unix} and @code{vax} that identify the
1435 type of system you are using. It also enables the undesirable and
1436 rarely used ISO trigraph feature. For the C compiler,
1437 it disables recognition of C++ style @samp{//} comments as well as
1438 the @code{inline} keyword.
1440 The alternate keywords @code{__asm__}, @code{__extension__},
1441 @code{__inline__} and @code{__typeof__} continue to work despite
1442 @option{-ansi}. You would not want to use them in an ISO C program, of
1443 course, but it is useful to put them in header files that might be included
1444 in compilations done with @option{-ansi}. Alternate predefined macros
1445 such as @code{__unix__} and @code{__vax__} are also available, with or
1446 without @option{-ansi}.
1448 The @option{-ansi} option does not cause non-ISO programs to be
1449 rejected gratuitously. For that, @option{-pedantic} is required in
1450 addition to @option{-ansi}. @xref{Warning Options}.
1452 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1453 option is used. Some header files may notice this macro and refrain
1454 from declaring certain functions or defining certain macros that the
1455 ISO standard doesn't call for; this is to avoid interfering with any
1456 programs that might use these names for other things.
1458 Functions that would normally be built in but do not have semantics
1459 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1460 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1461 built-in functions provided by GCC}, for details of the functions
1466 Determine the language standard. @xref{Standards,,Language Standards
1467 Supported by GCC}, for details of these standard versions. This option
1468 is currently only supported when compiling C or C++.
1470 The compiler can accept several base standards, such as @samp{c89} or
1471 @samp{c++98}, and GNU dialects of those standards, such as
1472 @samp{gnu89} or @samp{gnu++98}. By specifying a base standard, the
1473 compiler will accept all programs following that standard and those
1474 using GNU extensions that do not contradict it. For example,
1475 @samp{-std=c89} turns off certain features of GCC that are
1476 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1477 keywords, but not other GNU extensions that do not have a meaning in
1478 ISO C90, such as omitting the middle term of a @code{?:}
1479 expression. On the other hand, by specifying a GNU dialect of a
1480 standard, all features the compiler support are enabled, even when
1481 those features change the meaning of the base standard and some
1482 strict-conforming programs may be rejected. The particular standard
1483 is used by @option{-pedantic} to identify which features are GNU
1484 extensions given that version of the standard. For example
1485 @samp{-std=gnu89 -pedantic} would warn about C++ style @samp{//}
1486 comments, while @samp{-std=gnu99 -pedantic} would not.
1488 A value for this option must be provided; possible values are
1493 Support all ISO C90 programs (certain GNU extensions that conflict
1494 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1496 @item iso9899:199409
1497 ISO C90 as modified in amendment 1.
1503 ISO C99. Note that this standard is not yet fully supported; see
1504 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1505 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1508 GNU dialect of ISO C90 (including some C99 features). This
1509 is the default for C code.
1513 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1514 this will become the default. The name @samp{gnu9x} is deprecated.
1517 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1521 GNU dialect of @option{-std=c++98}. This is the default for
1525 The working draft of the upcoming ISO C++0x standard. This option
1526 enables experimental features that are likely to be included in
1527 C++0x. The working draft is constantly changing, and any feature that is
1528 enabled by this flag may be removed from future versions of GCC if it is
1529 not part of the C++0x standard.
1532 GNU dialect of @option{-std=c++0x}. This option enables
1533 experimental features that may be removed in future versions of GCC.
1536 @item -fgnu89-inline
1537 @opindex fgnu89-inline
1538 The option @option{-fgnu89-inline} tells GCC to use the traditional
1539 GNU semantics for @code{inline} functions when in C99 mode.
1540 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1541 is accepted and ignored by GCC versions 4.1.3 up to but not including
1542 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1543 C99 mode. Using this option is roughly equivalent to adding the
1544 @code{gnu_inline} function attribute to all inline functions
1545 (@pxref{Function Attributes}).
1547 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1548 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1549 specifies the default behavior). This option was first supported in
1550 GCC 4.3. This option is not supported in C89 or gnu89 mode.
1552 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1553 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1554 in effect for @code{inline} functions. @xref{Common Predefined
1555 Macros,,,cpp,The C Preprocessor}.
1557 @item -aux-info @var{filename}
1559 Output to the given filename prototyped declarations for all functions
1560 declared and/or defined in a translation unit, including those in header
1561 files. This option is silently ignored in any language other than C@.
1563 Besides declarations, the file indicates, in comments, the origin of
1564 each declaration (source file and line), whether the declaration was
1565 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1566 @samp{O} for old, respectively, in the first character after the line
1567 number and the colon), and whether it came from a declaration or a
1568 definition (@samp{C} or @samp{F}, respectively, in the following
1569 character). In the case of function definitions, a K&R-style list of
1570 arguments followed by their declarations is also provided, inside
1571 comments, after the declaration.
1575 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1576 keyword, so that code can use these words as identifiers. You can use
1577 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1578 instead. @option{-ansi} implies @option{-fno-asm}.
1580 In C++, this switch only affects the @code{typeof} keyword, since
1581 @code{asm} and @code{inline} are standard keywords. You may want to
1582 use the @option{-fno-gnu-keywords} flag instead, which has the same
1583 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1584 switch only affects the @code{asm} and @code{typeof} keywords, since
1585 @code{inline} is a standard keyword in ISO C99.
1588 @itemx -fno-builtin-@var{function}
1589 @opindex fno-builtin
1590 @cindex built-in functions
1591 Don't recognize built-in functions that do not begin with
1592 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1593 functions provided by GCC}, for details of the functions affected,
1594 including those which are not built-in functions when @option{-ansi} or
1595 @option{-std} options for strict ISO C conformance are used because they
1596 do not have an ISO standard meaning.
1598 GCC normally generates special code to handle certain built-in functions
1599 more efficiently; for instance, calls to @code{alloca} may become single
1600 instructions that adjust the stack directly, and calls to @code{memcpy}
1601 may become inline copy loops. The resulting code is often both smaller
1602 and faster, but since the function calls no longer appear as such, you
1603 cannot set a breakpoint on those calls, nor can you change the behavior
1604 of the functions by linking with a different library. In addition,
1605 when a function is recognized as a built-in function, GCC may use
1606 information about that function to warn about problems with calls to
1607 that function, or to generate more efficient code, even if the
1608 resulting code still contains calls to that function. For example,
1609 warnings are given with @option{-Wformat} for bad calls to
1610 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1611 known not to modify global memory.
1613 With the @option{-fno-builtin-@var{function}} option
1614 only the built-in function @var{function} is
1615 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1616 function is named that is not built-in in this version of GCC, this
1617 option is ignored. There is no corresponding
1618 @option{-fbuiltin-@var{function}} option; if you wish to enable
1619 built-in functions selectively when using @option{-fno-builtin} or
1620 @option{-ffreestanding}, you may define macros such as:
1623 #define abs(n) __builtin_abs ((n))
1624 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1629 @cindex hosted environment
1631 Assert that compilation takes place in a hosted environment. This implies
1632 @option{-fbuiltin}. A hosted environment is one in which the
1633 entire standard library is available, and in which @code{main} has a return
1634 type of @code{int}. Examples are nearly everything except a kernel.
1635 This is equivalent to @option{-fno-freestanding}.
1637 @item -ffreestanding
1638 @opindex ffreestanding
1639 @cindex hosted environment
1641 Assert that compilation takes place in a freestanding environment. This
1642 implies @option{-fno-builtin}. A freestanding environment
1643 is one in which the standard library may not exist, and program startup may
1644 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1645 This is equivalent to @option{-fno-hosted}.
1647 @xref{Standards,,Language Standards Supported by GCC}, for details of
1648 freestanding and hosted environments.
1652 @cindex openmp parallel
1653 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1654 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1655 compiler generates parallel code according to the OpenMP Application
1656 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1657 implies @option{-pthread}, and thus is only supported on targets that
1658 have support for @option{-pthread}.
1660 @item -fms-extensions
1661 @opindex fms-extensions
1662 Accept some non-standard constructs used in Microsoft header files.
1664 Some cases of unnamed fields in structures and unions are only
1665 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1666 fields within structs/unions}, for details.
1670 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1671 options for strict ISO C conformance) implies @option{-trigraphs}.
1673 @item -no-integrated-cpp
1674 @opindex no-integrated-cpp
1675 Performs a compilation in two passes: preprocessing and compiling. This
1676 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1677 @option{-B} option. The user supplied compilation step can then add in
1678 an additional preprocessing step after normal preprocessing but before
1679 compiling. The default is to use the integrated cpp (internal cpp)
1681 The semantics of this option will change if "cc1", "cc1plus", and
1682 "cc1obj" are merged.
1684 @cindex traditional C language
1685 @cindex C language, traditional
1687 @itemx -traditional-cpp
1688 @opindex traditional-cpp
1689 @opindex traditional
1690 Formerly, these options caused GCC to attempt to emulate a pre-standard
1691 C compiler. They are now only supported with the @option{-E} switch.
1692 The preprocessor continues to support a pre-standard mode. See the GNU
1693 CPP manual for details.
1695 @item -fcond-mismatch
1696 @opindex fcond-mismatch
1697 Allow conditional expressions with mismatched types in the second and
1698 third arguments. The value of such an expression is void. This option
1699 is not supported for C++.
1701 @item -flax-vector-conversions
1702 @opindex flax-vector-conversions
1703 Allow implicit conversions between vectors with differing numbers of
1704 elements and/or incompatible element types. This option should not be
1707 @item -funsigned-char
1708 @opindex funsigned-char
1709 Let the type @code{char} be unsigned, like @code{unsigned char}.
1711 Each kind of machine has a default for what @code{char} should
1712 be. It is either like @code{unsigned char} by default or like
1713 @code{signed char} by default.
1715 Ideally, a portable program should always use @code{signed char} or
1716 @code{unsigned char} when it depends on the signedness of an object.
1717 But many programs have been written to use plain @code{char} and
1718 expect it to be signed, or expect it to be unsigned, depending on the
1719 machines they were written for. This option, and its inverse, let you
1720 make such a program work with the opposite default.
1722 The type @code{char} is always a distinct type from each of
1723 @code{signed char} or @code{unsigned char}, even though its behavior
1724 is always just like one of those two.
1727 @opindex fsigned-char
1728 Let the type @code{char} be signed, like @code{signed char}.
1730 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1731 the negative form of @option{-funsigned-char}. Likewise, the option
1732 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1734 @item -fsigned-bitfields
1735 @itemx -funsigned-bitfields
1736 @itemx -fno-signed-bitfields
1737 @itemx -fno-unsigned-bitfields
1738 @opindex fsigned-bitfields
1739 @opindex funsigned-bitfields
1740 @opindex fno-signed-bitfields
1741 @opindex fno-unsigned-bitfields
1742 These options control whether a bit-field is signed or unsigned, when the
1743 declaration does not use either @code{signed} or @code{unsigned}. By
1744 default, such a bit-field is signed, because this is consistent: the
1745 basic integer types such as @code{int} are signed types.
1748 @node C++ Dialect Options
1749 @section Options Controlling C++ Dialect
1751 @cindex compiler options, C++
1752 @cindex C++ options, command line
1753 @cindex options, C++
1754 This section describes the command-line options that are only meaningful
1755 for C++ programs; but you can also use most of the GNU compiler options
1756 regardless of what language your program is in. For example, you
1757 might compile a file @code{firstClass.C} like this:
1760 g++ -g -frepo -O -c firstClass.C
1764 In this example, only @option{-frepo} is an option meant
1765 only for C++ programs; you can use the other options with any
1766 language supported by GCC@.
1768 Here is a list of options that are @emph{only} for compiling C++ programs:
1772 @item -fabi-version=@var{n}
1773 @opindex fabi-version
1774 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1775 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1776 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1777 the version that conforms most closely to the C++ ABI specification.
1778 Therefore, the ABI obtained using version 0 will change as ABI bugs
1781 The default is version 2.
1783 @item -fno-access-control
1784 @opindex fno-access-control
1785 Turn off all access checking. This switch is mainly useful for working
1786 around bugs in the access control code.
1790 Check that the pointer returned by @code{operator new} is non-null
1791 before attempting to modify the storage allocated. This check is
1792 normally unnecessary because the C++ standard specifies that
1793 @code{operator new} will only return @code{0} if it is declared
1794 @samp{throw()}, in which case the compiler will always check the
1795 return value even without this option. In all other cases, when
1796 @code{operator new} has a non-empty exception specification, memory
1797 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1798 @samp{new (nothrow)}.
1800 @item -fconserve-space
1801 @opindex fconserve-space
1802 Put uninitialized or runtime-initialized global variables into the
1803 common segment, as C does. This saves space in the executable at the
1804 cost of not diagnosing duplicate definitions. If you compile with this
1805 flag and your program mysteriously crashes after @code{main()} has
1806 completed, you may have an object that is being destroyed twice because
1807 two definitions were merged.
1809 This option is no longer useful on most targets, now that support has
1810 been added for putting variables into BSS without making them common.
1812 @item -fno-deduce-init-list
1813 @opindex fno-deduce-init-list
1814 Disable deduction of a template type parameter as
1815 std::initializer_list from a brace-enclosed initializer list, i.e.
1818 template <class T> auto forward(T t) -> decltype (realfn (t))
1825 forward(@{1,2@}); // call forward<std::initializer_list<int>>
1829 This option is present because this deduction is an extension to the
1830 current specification in the C++0x working draft, and there was
1831 some concern about potential overload resolution problems.
1833 @item -ffriend-injection
1834 @opindex ffriend-injection
1835 Inject friend functions into the enclosing namespace, so that they are
1836 visible outside the scope of the class in which they are declared.
1837 Friend functions were documented to work this way in the old Annotated
1838 C++ Reference Manual, and versions of G++ before 4.1 always worked
1839 that way. However, in ISO C++ a friend function which is not declared
1840 in an enclosing scope can only be found using argument dependent
1841 lookup. This option causes friends to be injected as they were in
1844 This option is for compatibility, and may be removed in a future
1847 @item -fno-elide-constructors
1848 @opindex fno-elide-constructors
1849 The C++ standard allows an implementation to omit creating a temporary
1850 which is only used to initialize another object of the same type.
1851 Specifying this option disables that optimization, and forces G++ to
1852 call the copy constructor in all cases.
1854 @item -fno-enforce-eh-specs
1855 @opindex fno-enforce-eh-specs
1856 Don't generate code to check for violation of exception specifications
1857 at runtime. This option violates the C++ standard, but may be useful
1858 for reducing code size in production builds, much like defining
1859 @samp{NDEBUG}. This does not give user code permission to throw
1860 exceptions in violation of the exception specifications; the compiler
1861 will still optimize based on the specifications, so throwing an
1862 unexpected exception will result in undefined behavior.
1865 @itemx -fno-for-scope
1867 @opindex fno-for-scope
1868 If @option{-ffor-scope} is specified, the scope of variables declared in
1869 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1870 as specified by the C++ standard.
1871 If @option{-fno-for-scope} is specified, the scope of variables declared in
1872 a @i{for-init-statement} extends to the end of the enclosing scope,
1873 as was the case in old versions of G++, and other (traditional)
1874 implementations of C++.
1876 The default if neither flag is given to follow the standard,
1877 but to allow and give a warning for old-style code that would
1878 otherwise be invalid, or have different behavior.
1880 @item -fno-gnu-keywords
1881 @opindex fno-gnu-keywords
1882 Do not recognize @code{typeof} as a keyword, so that code can use this
1883 word as an identifier. You can use the keyword @code{__typeof__} instead.
1884 @option{-ansi} implies @option{-fno-gnu-keywords}.
1886 @item -fno-implicit-templates
1887 @opindex fno-implicit-templates
1888 Never emit code for non-inline templates which are instantiated
1889 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1890 @xref{Template Instantiation}, for more information.
1892 @item -fno-implicit-inline-templates
1893 @opindex fno-implicit-inline-templates
1894 Don't emit code for implicit instantiations of inline templates, either.
1895 The default is to handle inlines differently so that compiles with and
1896 without optimization will need the same set of explicit instantiations.
1898 @item -fno-implement-inlines
1899 @opindex fno-implement-inlines
1900 To save space, do not emit out-of-line copies of inline functions
1901 controlled by @samp{#pragma implementation}. This will cause linker
1902 errors if these functions are not inlined everywhere they are called.
1904 @item -fms-extensions
1905 @opindex fms-extensions
1906 Disable pedantic warnings about constructs used in MFC, such as implicit
1907 int and getting a pointer to member function via non-standard syntax.
1909 @item -fno-nonansi-builtins
1910 @opindex fno-nonansi-builtins
1911 Disable built-in declarations of functions that are not mandated by
1912 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1913 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1915 @item -fno-operator-names
1916 @opindex fno-operator-names
1917 Do not treat the operator name keywords @code{and}, @code{bitand},
1918 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1919 synonyms as keywords.
1921 @item -fno-optional-diags
1922 @opindex fno-optional-diags
1923 Disable diagnostics that the standard says a compiler does not need to
1924 issue. Currently, the only such diagnostic issued by G++ is the one for
1925 a name having multiple meanings within a class.
1928 @opindex fpermissive
1929 Downgrade some diagnostics about nonconformant code from errors to
1930 warnings. Thus, using @option{-fpermissive} will allow some
1931 nonconforming code to compile.
1933 @item -fno-pretty-templates
1934 @opindex fno-pretty-templates
1935 When an error message refers to a specialization of a function
1936 template, the compiler will normally print the signature of the
1937 template followed by the template arguments and any typedefs or
1938 typenames in the signature (e.g. @code{void f(T) [with T = int]}
1939 rather than @code{void f(int)}) so that it's clear which template is
1940 involved. When an error message refers to a specialization of a class
1941 template, the compiler will omit any template arguments which match
1942 the default template arguments for that template. If either of these
1943 behaviors make it harder to understand the error message rather than
1944 easier, using @option{-fno-pretty-templates} will disable them.
1948 Enable automatic template instantiation at link time. This option also
1949 implies @option{-fno-implicit-templates}. @xref{Template
1950 Instantiation}, for more information.
1954 Disable generation of information about every class with virtual
1955 functions for use by the C++ runtime type identification features
1956 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1957 of the language, you can save some space by using this flag. Note that
1958 exception handling uses the same information, but it will generate it as
1959 needed. The @samp{dynamic_cast} operator can still be used for casts that
1960 do not require runtime type information, i.e.@: casts to @code{void *} or to
1961 unambiguous base classes.
1965 Emit statistics about front-end processing at the end of the compilation.
1966 This information is generally only useful to the G++ development team.
1968 @item -ftemplate-depth-@var{n}
1969 @opindex ftemplate-depth
1970 Set the maximum instantiation depth for template classes to @var{n}.
1971 A limit on the template instantiation depth is needed to detect
1972 endless recursions during template class instantiation. ANSI/ISO C++
1973 conforming programs must not rely on a maximum depth greater than 17
1974 (changed to 1024 in C++0x).
1976 @item -fno-threadsafe-statics
1977 @opindex fno-threadsafe-statics
1978 Do not emit the extra code to use the routines specified in the C++
1979 ABI for thread-safe initialization of local statics. You can use this
1980 option to reduce code size slightly in code that doesn't need to be
1983 @item -fuse-cxa-atexit
1984 @opindex fuse-cxa-atexit
1985 Register destructors for objects with static storage duration with the
1986 @code{__cxa_atexit} function rather than the @code{atexit} function.
1987 This option is required for fully standards-compliant handling of static
1988 destructors, but will only work if your C library supports
1989 @code{__cxa_atexit}.
1991 @item -fno-use-cxa-get-exception-ptr
1992 @opindex fno-use-cxa-get-exception-ptr
1993 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
1994 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
1995 if the runtime routine is not available.
1997 @item -fvisibility-inlines-hidden
1998 @opindex fvisibility-inlines-hidden
1999 This switch declares that the user does not attempt to compare
2000 pointers to inline methods where the addresses of the two functions
2001 were taken in different shared objects.
2003 The effect of this is that GCC may, effectively, mark inline methods with
2004 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2005 appear in the export table of a DSO and do not require a PLT indirection
2006 when used within the DSO@. Enabling this option can have a dramatic effect
2007 on load and link times of a DSO as it massively reduces the size of the
2008 dynamic export table when the library makes heavy use of templates.
2010 The behavior of this switch is not quite the same as marking the
2011 methods as hidden directly, because it does not affect static variables
2012 local to the function or cause the compiler to deduce that
2013 the function is defined in only one shared object.
2015 You may mark a method as having a visibility explicitly to negate the
2016 effect of the switch for that method. For example, if you do want to
2017 compare pointers to a particular inline method, you might mark it as
2018 having default visibility. Marking the enclosing class with explicit
2019 visibility will have no effect.
2021 Explicitly instantiated inline methods are unaffected by this option
2022 as their linkage might otherwise cross a shared library boundary.
2023 @xref{Template Instantiation}.
2025 @item -fvisibility-ms-compat
2026 @opindex fvisibility-ms-compat
2027 This flag attempts to use visibility settings to make GCC's C++
2028 linkage model compatible with that of Microsoft Visual Studio.
2030 The flag makes these changes to GCC's linkage model:
2034 It sets the default visibility to @code{hidden}, like
2035 @option{-fvisibility=hidden}.
2038 Types, but not their members, are not hidden by default.
2041 The One Definition Rule is relaxed for types without explicit
2042 visibility specifications which are defined in more than one different
2043 shared object: those declarations are permitted if they would have
2044 been permitted when this option was not used.
2047 In new code it is better to use @option{-fvisibility=hidden} and
2048 export those classes which are intended to be externally visible.
2049 Unfortunately it is possible for code to rely, perhaps accidentally,
2050 on the Visual Studio behavior.
2052 Among the consequences of these changes are that static data members
2053 of the same type with the same name but defined in different shared
2054 objects will be different, so changing one will not change the other;
2055 and that pointers to function members defined in different shared
2056 objects may not compare equal. When this flag is given, it is a
2057 violation of the ODR to define types with the same name differently.
2061 Do not use weak symbol support, even if it is provided by the linker.
2062 By default, G++ will use weak symbols if they are available. This
2063 option exists only for testing, and should not be used by end-users;
2064 it will result in inferior code and has no benefits. This option may
2065 be removed in a future release of G++.
2069 Do not search for header files in the standard directories specific to
2070 C++, but do still search the other standard directories. (This option
2071 is used when building the C++ library.)
2074 In addition, these optimization, warning, and code generation options
2075 have meanings only for C++ programs:
2078 @item -fno-default-inline
2079 @opindex fno-default-inline
2080 Do not assume @samp{inline} for functions defined inside a class scope.
2081 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2082 functions will have linkage like inline functions; they just won't be
2085 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2088 Warn when G++ generates code that is probably not compatible with the
2089 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2090 all such cases, there are probably some cases that are not warned about,
2091 even though G++ is generating incompatible code. There may also be
2092 cases where warnings are emitted even though the code that is generated
2095 You should rewrite your code to avoid these warnings if you are
2096 concerned about the fact that code generated by G++ may not be binary
2097 compatible with code generated by other compilers.
2099 The known incompatibilities at this point include:
2104 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2105 pack data into the same byte as a base class. For example:
2108 struct A @{ virtual void f(); int f1 : 1; @};
2109 struct B : public A @{ int f2 : 1; @};
2113 In this case, G++ will place @code{B::f2} into the same byte
2114 as@code{A::f1}; other compilers will not. You can avoid this problem
2115 by explicitly padding @code{A} so that its size is a multiple of the
2116 byte size on your platform; that will cause G++ and other compilers to
2117 layout @code{B} identically.
2120 Incorrect handling of tail-padding for virtual bases. G++ does not use
2121 tail padding when laying out virtual bases. For example:
2124 struct A @{ virtual void f(); char c1; @};
2125 struct B @{ B(); char c2; @};
2126 struct C : public A, public virtual B @{@};
2130 In this case, G++ will not place @code{B} into the tail-padding for
2131 @code{A}; other compilers will. You can avoid this problem by
2132 explicitly padding @code{A} so that its size is a multiple of its
2133 alignment (ignoring virtual base classes); that will cause G++ and other
2134 compilers to layout @code{C} identically.
2137 Incorrect handling of bit-fields with declared widths greater than that
2138 of their underlying types, when the bit-fields appear in a union. For
2142 union U @{ int i : 4096; @};
2146 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2147 union too small by the number of bits in an @code{int}.
2150 Empty classes can be placed at incorrect offsets. For example:
2160 struct C : public B, public A @{@};
2164 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2165 it should be placed at offset zero. G++ mistakenly believes that the
2166 @code{A} data member of @code{B} is already at offset zero.
2169 Names of template functions whose types involve @code{typename} or
2170 template template parameters can be mangled incorrectly.
2173 template <typename Q>
2174 void f(typename Q::X) @{@}
2176 template <template <typename> class Q>
2177 void f(typename Q<int>::X) @{@}
2181 Instantiations of these templates may be mangled incorrectly.
2185 It also warns psABI related changes. The known psABI changes at this
2191 For SYSV/x86-64, when passing union with long double, it is changed to
2192 pass in memory as specified in psABI. For example:
2202 @code{union U} will always be passed in memory.
2206 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2207 @opindex Wctor-dtor-privacy
2208 @opindex Wno-ctor-dtor-privacy
2209 Warn when a class seems unusable because all the constructors or
2210 destructors in that class are private, and it has neither friends nor
2211 public static member functions.
2213 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2214 @opindex Wnon-virtual-dtor
2215 @opindex Wno-non-virtual-dtor
2216 Warn when a class has virtual functions and accessible non-virtual
2217 destructor, in which case it would be possible but unsafe to delete
2218 an instance of a derived class through a pointer to the base class.
2219 This warning is also enabled if -Weffc++ is specified.
2221 @item -Wreorder @r{(C++ and Objective-C++ only)}
2223 @opindex Wno-reorder
2224 @cindex reordering, warning
2225 @cindex warning for reordering of member initializers
2226 Warn when the order of member initializers given in the code does not
2227 match the order in which they must be executed. For instance:
2233 A(): j (0), i (1) @{ @}
2237 The compiler will rearrange the member initializers for @samp{i}
2238 and @samp{j} to match the declaration order of the members, emitting
2239 a warning to that effect. This warning is enabled by @option{-Wall}.
2242 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2245 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2248 Warn about violations of the following style guidelines from Scott Meyers'
2249 @cite{Effective C++} book:
2253 Item 11: Define a copy constructor and an assignment operator for classes
2254 with dynamically allocated memory.
2257 Item 12: Prefer initialization to assignment in constructors.
2260 Item 14: Make destructors virtual in base classes.
2263 Item 15: Have @code{operator=} return a reference to @code{*this}.
2266 Item 23: Don't try to return a reference when you must return an object.
2270 Also warn about violations of the following style guidelines from
2271 Scott Meyers' @cite{More Effective C++} book:
2275 Item 6: Distinguish between prefix and postfix forms of increment and
2276 decrement operators.
2279 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2283 When selecting this option, be aware that the standard library
2284 headers do not obey all of these guidelines; use @samp{grep -v}
2285 to filter out those warnings.
2287 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2288 @opindex Wstrict-null-sentinel
2289 @opindex Wno-strict-null-sentinel
2290 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2291 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2292 to @code{__null}. Although it is a null pointer constant not a null pointer,
2293 it is guaranteed to be of the same size as a pointer. But this use is
2294 not portable across different compilers.
2296 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2297 @opindex Wno-non-template-friend
2298 @opindex Wnon-template-friend
2299 Disable warnings when non-templatized friend functions are declared
2300 within a template. Since the advent of explicit template specification
2301 support in G++, if the name of the friend is an unqualified-id (i.e.,
2302 @samp{friend foo(int)}), the C++ language specification demands that the
2303 friend declare or define an ordinary, nontemplate function. (Section
2304 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2305 could be interpreted as a particular specialization of a templatized
2306 function. Because this non-conforming behavior is no longer the default
2307 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2308 check existing code for potential trouble spots and is on by default.
2309 This new compiler behavior can be turned off with
2310 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2311 but disables the helpful warning.
2313 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2314 @opindex Wold-style-cast
2315 @opindex Wno-old-style-cast
2316 Warn if an old-style (C-style) cast to a non-void type is used within
2317 a C++ program. The new-style casts (@samp{dynamic_cast},
2318 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2319 less vulnerable to unintended effects and much easier to search for.
2321 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2322 @opindex Woverloaded-virtual
2323 @opindex Wno-overloaded-virtual
2324 @cindex overloaded virtual fn, warning
2325 @cindex warning for overloaded virtual fn
2326 Warn when a function declaration hides virtual functions from a
2327 base class. For example, in:
2334 struct B: public A @{
2339 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2347 will fail to compile.
2349 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2350 @opindex Wno-pmf-conversions
2351 @opindex Wpmf-conversions
2352 Disable the diagnostic for converting a bound pointer to member function
2355 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2356 @opindex Wsign-promo
2357 @opindex Wno-sign-promo
2358 Warn when overload resolution chooses a promotion from unsigned or
2359 enumerated type to a signed type, over a conversion to an unsigned type of
2360 the same size. Previous versions of G++ would try to preserve
2361 unsignedness, but the standard mandates the current behavior.
2366 A& operator = (int);
2376 In this example, G++ will synthesize a default @samp{A& operator =
2377 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2380 @node Objective-C and Objective-C++ Dialect Options
2381 @section Options Controlling Objective-C and Objective-C++ Dialects
2383 @cindex compiler options, Objective-C and Objective-C++
2384 @cindex Objective-C and Objective-C++ options, command line
2385 @cindex options, Objective-C and Objective-C++
2386 (NOTE: This manual does not describe the Objective-C and Objective-C++
2387 languages themselves. See @xref{Standards,,Language Standards
2388 Supported by GCC}, for references.)
2390 This section describes the command-line options that are only meaningful
2391 for Objective-C and Objective-C++ programs, but you can also use most of
2392 the language-independent GNU compiler options.
2393 For example, you might compile a file @code{some_class.m} like this:
2396 gcc -g -fgnu-runtime -O -c some_class.m
2400 In this example, @option{-fgnu-runtime} is an option meant only for
2401 Objective-C and Objective-C++ programs; you can use the other options with
2402 any language supported by GCC@.
2404 Note that since Objective-C is an extension of the C language, Objective-C
2405 compilations may also use options specific to the C front-end (e.g.,
2406 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2407 C++-specific options (e.g., @option{-Wabi}).
2409 Here is a list of options that are @emph{only} for compiling Objective-C
2410 and Objective-C++ programs:
2413 @item -fconstant-string-class=@var{class-name}
2414 @opindex fconstant-string-class
2415 Use @var{class-name} as the name of the class to instantiate for each
2416 literal string specified with the syntax @code{@@"@dots{}"}. The default
2417 class name is @code{NXConstantString} if the GNU runtime is being used, and
2418 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2419 @option{-fconstant-cfstrings} option, if also present, will override the
2420 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2421 to be laid out as constant CoreFoundation strings.
2424 @opindex fgnu-runtime
2425 Generate object code compatible with the standard GNU Objective-C
2426 runtime. This is the default for most types of systems.
2428 @item -fnext-runtime
2429 @opindex fnext-runtime
2430 Generate output compatible with the NeXT runtime. This is the default
2431 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2432 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2435 @item -fno-nil-receivers
2436 @opindex fno-nil-receivers
2437 Assume that all Objective-C message dispatches (e.g.,
2438 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
2439 is not @code{nil}. This allows for more efficient entry points in the runtime
2440 to be used. Currently, this option is only available in conjunction with
2441 the NeXT runtime on Mac OS X 10.3 and later.
2443 @item -fobjc-call-cxx-cdtors
2444 @opindex fobjc-call-cxx-cdtors
2445 For each Objective-C class, check if any of its instance variables is a
2446 C++ object with a non-trivial default constructor. If so, synthesize a
2447 special @code{- (id) .cxx_construct} instance method that will run
2448 non-trivial default constructors on any such instance variables, in order,
2449 and then return @code{self}. Similarly, check if any instance variable
2450 is a C++ object with a non-trivial destructor, and if so, synthesize a
2451 special @code{- (void) .cxx_destruct} method that will run
2452 all such default destructors, in reverse order.
2454 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
2455 thusly generated will only operate on instance variables declared in the
2456 current Objective-C class, and not those inherited from superclasses. It
2457 is the responsibility of the Objective-C runtime to invoke all such methods
2458 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
2459 will be invoked by the runtime immediately after a new object
2460 instance is allocated; the @code{- (void) .cxx_destruct} methods will
2461 be invoked immediately before the runtime deallocates an object instance.
2463 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2464 support for invoking the @code{- (id) .cxx_construct} and
2465 @code{- (void) .cxx_destruct} methods.
2467 @item -fobjc-direct-dispatch
2468 @opindex fobjc-direct-dispatch
2469 Allow fast jumps to the message dispatcher. On Darwin this is
2470 accomplished via the comm page.
2472 @item -fobjc-exceptions
2473 @opindex fobjc-exceptions
2474 Enable syntactic support for structured exception handling in Objective-C,
2475 similar to what is offered by C++ and Java. This option is
2476 unavailable in conjunction with the NeXT runtime on Mac OS X 10.2 and
2485 @@catch (AnObjCClass *exc) @{
2492 @@catch (AnotherClass *exc) @{
2495 @@catch (id allOthers) @{
2505 The @code{@@throw} statement may appear anywhere in an Objective-C or
2506 Objective-C++ program; when used inside of a @code{@@catch} block, the
2507 @code{@@throw} may appear without an argument (as shown above), in which case
2508 the object caught by the @code{@@catch} will be rethrown.
2510 Note that only (pointers to) Objective-C objects may be thrown and
2511 caught using this scheme. When an object is thrown, it will be caught
2512 by the nearest @code{@@catch} clause capable of handling objects of that type,
2513 analogously to how @code{catch} blocks work in C++ and Java. A
2514 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2515 any and all Objective-C exceptions not caught by previous @code{@@catch}
2518 The @code{@@finally} clause, if present, will be executed upon exit from the
2519 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2520 regardless of whether any exceptions are thrown, caught or rethrown
2521 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2522 of the @code{finally} clause in Java.
2524 There are several caveats to using the new exception mechanism:
2528 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2529 idioms provided by the @code{NSException} class, the new
2530 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2531 systems, due to additional functionality needed in the (NeXT) Objective-C
2535 As mentioned above, the new exceptions do not support handling
2536 types other than Objective-C objects. Furthermore, when used from
2537 Objective-C++, the Objective-C exception model does not interoperate with C++
2538 exceptions at this time. This means you cannot @code{@@throw} an exception
2539 from Objective-C and @code{catch} it in C++, or vice versa
2540 (i.e., @code{throw @dots{} @@catch}).
2543 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2544 blocks for thread-safe execution:
2547 @@synchronized (ObjCClass *guard) @{
2552 Upon entering the @code{@@synchronized} block, a thread of execution shall
2553 first check whether a lock has been placed on the corresponding @code{guard}
2554 object by another thread. If it has, the current thread shall wait until
2555 the other thread relinquishes its lock. Once @code{guard} becomes available,
2556 the current thread will place its own lock on it, execute the code contained in
2557 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2558 making @code{guard} available to other threads).
2560 Unlike Java, Objective-C does not allow for entire methods to be marked
2561 @code{@@synchronized}. Note that throwing exceptions out of
2562 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2563 to be unlocked properly.
2567 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2569 @item -freplace-objc-classes
2570 @opindex freplace-objc-classes
2571 Emit a special marker instructing @command{ld(1)} not to statically link in
2572 the resulting object file, and allow @command{dyld(1)} to load it in at
2573 run time instead. This is used in conjunction with the Fix-and-Continue
2574 debugging mode, where the object file in question may be recompiled and
2575 dynamically reloaded in the course of program execution, without the need
2576 to restart the program itself. Currently, Fix-and-Continue functionality
2577 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2582 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2583 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2584 compile time) with static class references that get initialized at load time,
2585 which improves run-time performance. Specifying the @option{-fzero-link} flag
2586 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2587 to be retained. This is useful in Zero-Link debugging mode, since it allows
2588 for individual class implementations to be modified during program execution.
2592 Dump interface declarations for all classes seen in the source file to a
2593 file named @file{@var{sourcename}.decl}.
2595 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2596 @opindex Wassign-intercept
2597 @opindex Wno-assign-intercept
2598 Warn whenever an Objective-C assignment is being intercepted by the
2601 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2602 @opindex Wno-protocol
2604 If a class is declared to implement a protocol, a warning is issued for
2605 every method in the protocol that is not implemented by the class. The
2606 default behavior is to issue a warning for every method not explicitly
2607 implemented in the class, even if a method implementation is inherited
2608 from the superclass. If you use the @option{-Wno-protocol} option, then
2609 methods inherited from the superclass are considered to be implemented,
2610 and no warning is issued for them.
2612 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2614 @opindex Wno-selector
2615 Warn if multiple methods of different types for the same selector are
2616 found during compilation. The check is performed on the list of methods
2617 in the final stage of compilation. Additionally, a check is performed
2618 for each selector appearing in a @code{@@selector(@dots{})}
2619 expression, and a corresponding method for that selector has been found
2620 during compilation. Because these checks scan the method table only at
2621 the end of compilation, these warnings are not produced if the final
2622 stage of compilation is not reached, for example because an error is
2623 found during compilation, or because the @option{-fsyntax-only} option is
2626 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2627 @opindex Wstrict-selector-match
2628 @opindex Wno-strict-selector-match
2629 Warn if multiple methods with differing argument and/or return types are
2630 found for a given selector when attempting to send a message using this
2631 selector to a receiver of type @code{id} or @code{Class}. When this flag
2632 is off (which is the default behavior), the compiler will omit such warnings
2633 if any differences found are confined to types which share the same size
2636 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2637 @opindex Wundeclared-selector
2638 @opindex Wno-undeclared-selector
2639 Warn if a @code{@@selector(@dots{})} expression referring to an
2640 undeclared selector is found. A selector is considered undeclared if no
2641 method with that name has been declared before the
2642 @code{@@selector(@dots{})} expression, either explicitly in an
2643 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2644 an @code{@@implementation} section. This option always performs its
2645 checks as soon as a @code{@@selector(@dots{})} expression is found,
2646 while @option{-Wselector} only performs its checks in the final stage of
2647 compilation. This also enforces the coding style convention
2648 that methods and selectors must be declared before being used.
2650 @item -print-objc-runtime-info
2651 @opindex print-objc-runtime-info
2652 Generate C header describing the largest structure that is passed by
2657 @node Language Independent Options
2658 @section Options to Control Diagnostic Messages Formatting
2659 @cindex options to control diagnostics formatting
2660 @cindex diagnostic messages
2661 @cindex message formatting
2663 Traditionally, diagnostic messages have been formatted irrespective of
2664 the output device's aspect (e.g.@: its width, @dots{}). The options described
2665 below can be used to control the diagnostic messages formatting
2666 algorithm, e.g.@: how many characters per line, how often source location
2667 information should be reported. Right now, only the C++ front end can
2668 honor these options. However it is expected, in the near future, that
2669 the remaining front ends would be able to digest them correctly.
2672 @item -fmessage-length=@var{n}
2673 @opindex fmessage-length
2674 Try to format error messages so that they fit on lines of about @var{n}
2675 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2676 the front ends supported by GCC@. If @var{n} is zero, then no
2677 line-wrapping will be done; each error message will appear on a single
2680 @opindex fdiagnostics-show-location
2681 @item -fdiagnostics-show-location=once
2682 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2683 reporter to emit @emph{once} source location information; that is, in
2684 case the message is too long to fit on a single physical line and has to
2685 be wrapped, the source location won't be emitted (as prefix) again,
2686 over and over, in subsequent continuation lines. This is the default
2689 @item -fdiagnostics-show-location=every-line
2690 Only meaningful in line-wrapping mode. Instructs the diagnostic
2691 messages reporter to emit the same source location information (as
2692 prefix) for physical lines that result from the process of breaking
2693 a message which is too long to fit on a single line.
2695 @item -fdiagnostics-show-option
2696 @opindex fdiagnostics-show-option
2697 This option instructs the diagnostic machinery to add text to each
2698 diagnostic emitted, which indicates which command line option directly
2699 controls that diagnostic, when such an option is known to the
2700 diagnostic machinery.
2702 @item -Wcoverage-mismatch
2703 @opindex Wcoverage-mismatch
2704 Warn if feedback profiles do not match when using the
2705 @option{-fprofile-use} option.
2706 If a source file was changed between @option{-fprofile-gen} and
2707 @option{-fprofile-use}, the files with the profile feedback can fail
2708 to match the source file and GCC can not use the profile feedback
2709 information. By default, GCC emits an error message in this case.
2710 The option @option{-Wcoverage-mismatch} emits a warning instead of an
2711 error. GCC does not use appropriate feedback profiles, so using this
2712 option can result in poorly optimized code. This option is useful
2713 only in the case of very minor changes such as bug fixes to an
2718 @node Warning Options
2719 @section Options to Request or Suppress Warnings
2720 @cindex options to control warnings
2721 @cindex warning messages
2722 @cindex messages, warning
2723 @cindex suppressing warnings
2725 Warnings are diagnostic messages that report constructions which
2726 are not inherently erroneous but which are risky or suggest there
2727 may have been an error.
2729 The following language-independent options do not enable specific
2730 warnings but control the kinds of diagnostics produced by GCC.
2733 @cindex syntax checking
2735 @opindex fsyntax-only
2736 Check the code for syntax errors, but don't do anything beyond that.
2740 Inhibit all warning messages.
2745 Make all warnings into errors.
2750 Make the specified warning into an error. The specifier for a warning
2751 is appended, for example @option{-Werror=switch} turns the warnings
2752 controlled by @option{-Wswitch} into errors. This switch takes a
2753 negative form, to be used to negate @option{-Werror} for specific
2754 warnings, for example @option{-Wno-error=switch} makes
2755 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2756 is in effect. You can use the @option{-fdiagnostics-show-option}
2757 option to have each controllable warning amended with the option which
2758 controls it, to determine what to use with this option.
2760 Note that specifying @option{-Werror=}@var{foo} automatically implies
2761 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2764 @item -Wfatal-errors
2765 @opindex Wfatal-errors
2766 @opindex Wno-fatal-errors
2767 This option causes the compiler to abort compilation on the first error
2768 occurred rather than trying to keep going and printing further error
2773 You can request many specific warnings with options beginning
2774 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2775 implicit declarations. Each of these specific warning options also
2776 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2777 example, @option{-Wno-implicit}. This manual lists only one of the
2778 two forms, whichever is not the default. For further,
2779 language-specific options also refer to @ref{C++ Dialect Options} and
2780 @ref{Objective-C and Objective-C++ Dialect Options}.
2785 Issue all the warnings demanded by strict ISO C and ISO C++;
2786 reject all programs that use forbidden extensions, and some other
2787 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2788 version of the ISO C standard specified by any @option{-std} option used.
2790 Valid ISO C and ISO C++ programs should compile properly with or without
2791 this option (though a rare few will require @option{-ansi} or a
2792 @option{-std} option specifying the required version of ISO C)@. However,
2793 without this option, certain GNU extensions and traditional C and C++
2794 features are supported as well. With this option, they are rejected.
2796 @option{-pedantic} does not cause warning messages for use of the
2797 alternate keywords whose names begin and end with @samp{__}. Pedantic
2798 warnings are also disabled in the expression that follows
2799 @code{__extension__}. However, only system header files should use
2800 these escape routes; application programs should avoid them.
2801 @xref{Alternate Keywords}.
2803 Some users try to use @option{-pedantic} to check programs for strict ISO
2804 C conformance. They soon find that it does not do quite what they want:
2805 it finds some non-ISO practices, but not all---only those for which
2806 ISO C @emph{requires} a diagnostic, and some others for which
2807 diagnostics have been added.
2809 A feature to report any failure to conform to ISO C might be useful in
2810 some instances, but would require considerable additional work and would
2811 be quite different from @option{-pedantic}. We don't have plans to
2812 support such a feature in the near future.
2814 Where the standard specified with @option{-std} represents a GNU
2815 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2816 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2817 extended dialect is based. Warnings from @option{-pedantic} are given
2818 where they are required by the base standard. (It would not make sense
2819 for such warnings to be given only for features not in the specified GNU
2820 C dialect, since by definition the GNU dialects of C include all
2821 features the compiler supports with the given option, and there would be
2822 nothing to warn about.)
2824 @item -pedantic-errors
2825 @opindex pedantic-errors
2826 Like @option{-pedantic}, except that errors are produced rather than
2832 This enables all the warnings about constructions that some users
2833 consider questionable, and that are easy to avoid (or modify to
2834 prevent the warning), even in conjunction with macros. This also
2835 enables some language-specific warnings described in @ref{C++ Dialect
2836 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2838 @option{-Wall} turns on the following warning flags:
2840 @gccoptlist{-Waddress @gol
2841 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2843 -Wchar-subscripts @gol
2844 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
2846 -Wimplicit-function-declaration @gol
2849 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2850 -Wmissing-braces @gol
2856 -Wsequence-point @gol
2857 -Wsign-compare @r{(only in C++)} @gol
2858 -Wstrict-aliasing @gol
2859 -Wstrict-overflow=1 @gol
2862 -Wuninitialized @gol
2863 -Wunknown-pragmas @gol
2864 -Wunused-function @gol
2867 -Wunused-variable @gol
2868 -Wvolatile-register-var @gol
2871 Note that some warning flags are not implied by @option{-Wall}. Some of
2872 them warn about constructions that users generally do not consider
2873 questionable, but which occasionally you might wish to check for;
2874 others warn about constructions that are necessary or hard to avoid in
2875 some cases, and there is no simple way to modify the code to suppress
2876 the warning. Some of them are enabled by @option{-Wextra} but many of
2877 them must be enabled individually.
2883 This enables some extra warning flags that are not enabled by
2884 @option{-Wall}. (This option used to be called @option{-W}. The older
2885 name is still supported, but the newer name is more descriptive.)
2887 @gccoptlist{-Wclobbered @gol
2889 -Wignored-qualifiers @gol
2890 -Wmissing-field-initializers @gol
2891 -Wmissing-parameter-type @r{(C only)} @gol
2892 -Wold-style-declaration @r{(C only)} @gol
2893 -Woverride-init @gol
2896 -Wuninitialized @gol
2897 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2900 The option @option{-Wextra} also prints warning messages for the
2906 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2907 @samp{>}, or @samp{>=}.
2910 (C++ only) An enumerator and a non-enumerator both appear in a
2911 conditional expression.
2914 (C++ only) Ambiguous virtual bases.
2917 (C++ only) Subscripting an array which has been declared @samp{register}.
2920 (C++ only) Taking the address of a variable which has been declared
2924 (C++ only) A base class is not initialized in a derived class' copy
2929 @item -Wchar-subscripts
2930 @opindex Wchar-subscripts
2931 @opindex Wno-char-subscripts
2932 Warn if an array subscript has type @code{char}. This is a common cause
2933 of error, as programmers often forget that this type is signed on some
2935 This warning is enabled by @option{-Wall}.
2939 @opindex Wno-comment
2940 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2941 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2942 This warning is enabled by @option{-Wall}.
2947 @opindex ffreestanding
2948 @opindex fno-builtin
2949 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2950 the arguments supplied have types appropriate to the format string
2951 specified, and that the conversions specified in the format string make
2952 sense. This includes standard functions, and others specified by format
2953 attributes (@pxref{Function Attributes}), in the @code{printf},
2954 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2955 not in the C standard) families (or other target-specific families).
2956 Which functions are checked without format attributes having been
2957 specified depends on the standard version selected, and such checks of
2958 functions without the attribute specified are disabled by
2959 @option{-ffreestanding} or @option{-fno-builtin}.
2961 The formats are checked against the format features supported by GNU
2962 libc version 2.2. These include all ISO C90 and C99 features, as well
2963 as features from the Single Unix Specification and some BSD and GNU
2964 extensions. Other library implementations may not support all these
2965 features; GCC does not support warning about features that go beyond a
2966 particular library's limitations. However, if @option{-pedantic} is used
2967 with @option{-Wformat}, warnings will be given about format features not
2968 in the selected standard version (but not for @code{strfmon} formats,
2969 since those are not in any version of the C standard). @xref{C Dialect
2970 Options,,Options Controlling C Dialect}.
2972 Since @option{-Wformat} also checks for null format arguments for
2973 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2975 @option{-Wformat} is included in @option{-Wall}. For more control over some
2976 aspects of format checking, the options @option{-Wformat-y2k},
2977 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2978 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2979 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2982 @opindex Wformat-y2k
2983 @opindex Wno-format-y2k
2984 If @option{-Wformat} is specified, also warn about @code{strftime}
2985 formats which may yield only a two-digit year.
2987 @item -Wno-format-contains-nul
2988 @opindex Wno-format-contains-nul
2989 @opindex Wformat-contains-nul
2990 If @option{-Wformat} is specified, do not warn about format strings that
2993 @item -Wno-format-extra-args
2994 @opindex Wno-format-extra-args
2995 @opindex Wformat-extra-args
2996 If @option{-Wformat} is specified, do not warn about excess arguments to a
2997 @code{printf} or @code{scanf} format function. The C standard specifies
2998 that such arguments are ignored.
3000 Where the unused arguments lie between used arguments that are
3001 specified with @samp{$} operand number specifications, normally
3002 warnings are still given, since the implementation could not know what
3003 type to pass to @code{va_arg} to skip the unused arguments. However,
3004 in the case of @code{scanf} formats, this option will suppress the
3005 warning if the unused arguments are all pointers, since the Single
3006 Unix Specification says that such unused arguments are allowed.
3008 @item -Wno-format-zero-length @r{(C and Objective-C only)}
3009 @opindex Wno-format-zero-length
3010 @opindex Wformat-zero-length
3011 If @option{-Wformat} is specified, do not warn about zero-length formats.
3012 The C standard specifies that zero-length formats are allowed.
3014 @item -Wformat-nonliteral
3015 @opindex Wformat-nonliteral
3016 @opindex Wno-format-nonliteral
3017 If @option{-Wformat} is specified, also warn if the format string is not a
3018 string literal and so cannot be checked, unless the format function
3019 takes its format arguments as a @code{va_list}.
3021 @item -Wformat-security
3022 @opindex Wformat-security
3023 @opindex Wno-format-security
3024 If @option{-Wformat} is specified, also warn about uses of format
3025 functions that represent possible security problems. At present, this
3026 warns about calls to @code{printf} and @code{scanf} functions where the
3027 format string is not a string literal and there are no format arguments,
3028 as in @code{printf (foo);}. This may be a security hole if the format
3029 string came from untrusted input and contains @samp{%n}. (This is
3030 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3031 in future warnings may be added to @option{-Wformat-security} that are not
3032 included in @option{-Wformat-nonliteral}.)
3036 @opindex Wno-format=2
3037 Enable @option{-Wformat} plus format checks not included in
3038 @option{-Wformat}. Currently equivalent to @samp{-Wformat
3039 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
3041 @item -Wnonnull @r{(C and Objective-C only)}
3043 @opindex Wno-nonnull
3044 Warn about passing a null pointer for arguments marked as
3045 requiring a non-null value by the @code{nonnull} function attribute.
3047 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3048 can be disabled with the @option{-Wno-nonnull} option.
3050 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3052 @opindex Wno-init-self
3053 Warn about uninitialized variables which are initialized with themselves.
3054 Note this option can only be used with the @option{-Wuninitialized} option.
3056 For example, GCC will warn about @code{i} being uninitialized in the
3057 following snippet only when @option{-Winit-self} has been specified:
3068 @item -Wimplicit-int @r{(C and Objective-C only)}
3069 @opindex Wimplicit-int
3070 @opindex Wno-implicit-int
3071 Warn when a declaration does not specify a type.
3072 This warning is enabled by @option{-Wall}.
3074 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3075 @opindex Wimplicit-function-declaration
3076 @opindex Wno-implicit-function-declaration
3077 Give a warning whenever a function is used before being declared. In
3078 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3079 enabled by default and it is made into an error by
3080 @option{-pedantic-errors}. This warning is also enabled by
3085 @opindex Wno-implicit
3086 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3087 This warning is enabled by @option{-Wall}.
3089 @item -Wignored-qualifiers @r{(C and C++ only)}
3090 @opindex Wignored-qualifiers
3091 @opindex Wno-ignored-qualifiers
3092 Warn if the return type of a function has a type qualifier
3093 such as @code{const}. For ISO C such a type qualifier has no effect,
3094 since the value returned by a function is not an lvalue.
3095 For C++, the warning is only emitted for scalar types or @code{void}.
3096 ISO C prohibits qualified @code{void} return types on function
3097 definitions, so such return types always receive a warning
3098 even without this option.
3100 This warning is also enabled by @option{-Wextra}.
3105 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3106 a function with external linkage, returning int, taking either zero
3107 arguments, two, or three arguments of appropriate types. This warning
3108 is enabled by default in C++ and is enabled by either @option{-Wall}
3109 or @option{-pedantic}.
3111 @item -Wmissing-braces
3112 @opindex Wmissing-braces
3113 @opindex Wno-missing-braces
3114 Warn if an aggregate or union initializer is not fully bracketed. In
3115 the following example, the initializer for @samp{a} is not fully
3116 bracketed, but that for @samp{b} is fully bracketed.
3119 int a[2][2] = @{ 0, 1, 2, 3 @};
3120 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3123 This warning is enabled by @option{-Wall}.
3125 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3126 @opindex Wmissing-include-dirs
3127 @opindex Wno-missing-include-dirs
3128 Warn if a user-supplied include directory does not exist.
3131 @opindex Wparentheses
3132 @opindex Wno-parentheses
3133 Warn if parentheses are omitted in certain contexts, such
3134 as when there is an assignment in a context where a truth value
3135 is expected, or when operators are nested whose precedence people
3136 often get confused about.
3138 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3139 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3140 interpretation from that of ordinary mathematical notation.
3142 Also warn about constructions where there may be confusion to which
3143 @code{if} statement an @code{else} branch belongs. Here is an example of
3158 In C/C++, every @code{else} branch belongs to the innermost possible
3159 @code{if} statement, which in this example is @code{if (b)}. This is
3160 often not what the programmer expected, as illustrated in the above
3161 example by indentation the programmer chose. When there is the
3162 potential for this confusion, GCC will issue a warning when this flag
3163 is specified. To eliminate the warning, add explicit braces around
3164 the innermost @code{if} statement so there is no way the @code{else}
3165 could belong to the enclosing @code{if}. The resulting code would
3182 This warning is enabled by @option{-Wall}.
3184 @item -Wsequence-point
3185 @opindex Wsequence-point
3186 @opindex Wno-sequence-point
3187 Warn about code that may have undefined semantics because of violations
3188 of sequence point rules in the C and C++ standards.
3190 The C and C++ standards defines the order in which expressions in a C/C++
3191 program are evaluated in terms of @dfn{sequence points}, which represent
3192 a partial ordering between the execution of parts of the program: those
3193 executed before the sequence point, and those executed after it. These
3194 occur after the evaluation of a full expression (one which is not part
3195 of a larger expression), after the evaluation of the first operand of a
3196 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3197 function is called (but after the evaluation of its arguments and the
3198 expression denoting the called function), and in certain other places.
3199 Other than as expressed by the sequence point rules, the order of
3200 evaluation of subexpressions of an expression is not specified. All
3201 these rules describe only a partial order rather than a total order,
3202 since, for example, if two functions are called within one expression
3203 with no sequence point between them, the order in which the functions
3204 are called is not specified. However, the standards committee have
3205 ruled that function calls do not overlap.
3207 It is not specified when between sequence points modifications to the
3208 values of objects take effect. Programs whose behavior depends on this
3209 have undefined behavior; the C and C++ standards specify that ``Between
3210 the previous and next sequence point an object shall have its stored
3211 value modified at most once by the evaluation of an expression.
3212 Furthermore, the prior value shall be read only to determine the value
3213 to be stored.''. If a program breaks these rules, the results on any
3214 particular implementation are entirely unpredictable.
3216 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3217 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3218 diagnosed by this option, and it may give an occasional false positive
3219 result, but in general it has been found fairly effective at detecting
3220 this sort of problem in programs.
3222 The standard is worded confusingly, therefore there is some debate
3223 over the precise meaning of the sequence point rules in subtle cases.
3224 Links to discussions of the problem, including proposed formal
3225 definitions, may be found on the GCC readings page, at
3226 @w{@uref{http://gcc.gnu.org/readings.html}}.
3228 This warning is enabled by @option{-Wall} for C and C++.
3231 @opindex Wreturn-type
3232 @opindex Wno-return-type
3233 Warn whenever a function is defined with a return-type that defaults
3234 to @code{int}. Also warn about any @code{return} statement with no
3235 return-value in a function whose return-type is not @code{void}
3236 (falling off the end of the function body is considered returning
3237 without a value), and about a @code{return} statement with an
3238 expression in a function whose return-type is @code{void}.
3240 For C++, a function without return type always produces a diagnostic
3241 message, even when @option{-Wno-return-type} is specified. The only
3242 exceptions are @samp{main} and functions defined in system headers.
3244 This warning is enabled by @option{-Wall}.
3249 Warn whenever a @code{switch} statement has an index of enumerated type
3250 and lacks a @code{case} for one or more of the named codes of that
3251 enumeration. (The presence of a @code{default} label prevents this
3252 warning.) @code{case} labels outside the enumeration range also
3253 provoke warnings when this option is used (even if there is a
3254 @code{default} label).
3255 This warning is enabled by @option{-Wall}.
3257 @item -Wswitch-default
3258 @opindex Wswitch-default
3259 @opindex Wno-switch-default
3260 Warn whenever a @code{switch} statement does not have a @code{default}
3264 @opindex Wswitch-enum
3265 @opindex Wno-switch-enum
3266 Warn whenever a @code{switch} statement has an index of enumerated type
3267 and lacks a @code{case} for one or more of the named codes of that
3268 enumeration. @code{case} labels outside the enumeration range also
3269 provoke warnings when this option is used. The only difference
3270 between @option{-Wswitch} and this option is that this option gives a
3271 warning about an omitted enumeration code even if there is a
3272 @code{default} label.
3274 @item -Wsync-nand @r{(C and C++ only)}
3276 @opindex Wno-sync-nand
3277 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3278 built-in functions are used. These functions changed semantics in GCC 4.4.
3282 @opindex Wno-trigraphs
3283 Warn if any trigraphs are encountered that might change the meaning of
3284 the program (trigraphs within comments are not warned about).
3285 This warning is enabled by @option{-Wall}.
3287 @item -Wunused-function
3288 @opindex Wunused-function
3289 @opindex Wno-unused-function
3290 Warn whenever a static function is declared but not defined or a
3291 non-inline static function is unused.
3292 This warning is enabled by @option{-Wall}.
3294 @item -Wunused-label
3295 @opindex Wunused-label
3296 @opindex Wno-unused-label
3297 Warn whenever a label is declared but not used.
3298 This warning is enabled by @option{-Wall}.
3300 To suppress this warning use the @samp{unused} attribute
3301 (@pxref{Variable Attributes}).
3303 @item -Wunused-parameter
3304 @opindex Wunused-parameter
3305 @opindex Wno-unused-parameter
3306 Warn whenever a function parameter is unused aside from its declaration.
3308 To suppress this warning use the @samp{unused} attribute
3309 (@pxref{Variable Attributes}).
3311 @item -Wno-unused-result
3312 @opindex Wunused-result
3313 @opindex Wno-unused-result
3314 Do not warn if a caller of a function marked with attribute
3315 @code{warn_unused_result} (@pxref{Variable Attributes}) does not use
3316 its return value. The default is @option{-Wunused-result}.
3318 @item -Wunused-variable
3319 @opindex Wunused-variable
3320 @opindex Wno-unused-variable
3321 Warn whenever a local variable or non-constant static variable is unused
3322 aside from its declaration.
3323 This warning is enabled by @option{-Wall}.
3325 To suppress this warning use the @samp{unused} attribute
3326 (@pxref{Variable Attributes}).
3328 @item -Wunused-value
3329 @opindex Wunused-value
3330 @opindex Wno-unused-value
3331 Warn whenever a statement computes a result that is explicitly not
3332 used. To suppress this warning cast the unused expression to
3333 @samp{void}. This includes an expression-statement or the left-hand
3334 side of a comma expression that contains no side effects. For example,
3335 an expression such as @samp{x[i,j]} will cause a warning, while
3336 @samp{x[(void)i,j]} will not.
3338 This warning is enabled by @option{-Wall}.
3343 All the above @option{-Wunused} options combined.
3345 In order to get a warning about an unused function parameter, you must
3346 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3347 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3349 @item -Wuninitialized
3350 @opindex Wuninitialized
3351 @opindex Wno-uninitialized
3352 Warn if an automatic variable is used without first being initialized
3353 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3354 warn if a non-static reference or non-static @samp{const} member
3355 appears in a class without constructors.
3357 If you want to warn about code which uses the uninitialized value of the
3358 variable in its own initializer, use the @option{-Winit-self} option.
3360 These warnings occur for individual uninitialized or clobbered
3361 elements of structure, union or array variables as well as for
3362 variables which are uninitialized or clobbered as a whole. They do
3363 not occur for variables or elements declared @code{volatile}. Because
3364 these warnings depend on optimization, the exact variables or elements
3365 for which there are warnings will depend on the precise optimization
3366 options and version of GCC used.
3368 Note that there may be no warning about a variable that is used only
3369 to compute a value that itself is never used, because such
3370 computations may be deleted by data flow analysis before the warnings
3373 These warnings are made optional because GCC is not smart
3374 enough to see all the reasons why the code might be correct
3375 despite appearing to have an error. Here is one example of how
3396 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3397 always initialized, but GCC doesn't know this. Here is
3398 another common case:
3403 if (change_y) save_y = y, y = new_y;
3405 if (change_y) y = save_y;
3410 This has no bug because @code{save_y} is used only if it is set.
3412 @cindex @code{longjmp} warnings
3413 This option also warns when a non-volatile automatic variable might be
3414 changed by a call to @code{longjmp}. These warnings as well are possible
3415 only in optimizing compilation.
3417 The compiler sees only the calls to @code{setjmp}. It cannot know
3418 where @code{longjmp} will be called; in fact, a signal handler could
3419 call it at any point in the code. As a result, you may get a warning
3420 even when there is in fact no problem because @code{longjmp} cannot
3421 in fact be called at the place which would cause a problem.
3423 Some spurious warnings can be avoided if you declare all the functions
3424 you use that never return as @code{noreturn}. @xref{Function
3427 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3429 @item -Wunknown-pragmas
3430 @opindex Wunknown-pragmas
3431 @opindex Wno-unknown-pragmas
3432 @cindex warning for unknown pragmas
3433 @cindex unknown pragmas, warning
3434 @cindex pragmas, warning of unknown
3435 Warn when a #pragma directive is encountered which is not understood by
3436 GCC@. If this command line option is used, warnings will even be issued
3437 for unknown pragmas in system header files. This is not the case if
3438 the warnings were only enabled by the @option{-Wall} command line option.
3441 @opindex Wno-pragmas
3443 Do not warn about misuses of pragmas, such as incorrect parameters,
3444 invalid syntax, or conflicts between pragmas. See also
3445 @samp{-Wunknown-pragmas}.
3447 @item -Wstrict-aliasing
3448 @opindex Wstrict-aliasing
3449 @opindex Wno-strict-aliasing
3450 This option is only active when @option{-fstrict-aliasing} is active.
3451 It warns about code which might break the strict aliasing rules that the
3452 compiler is using for optimization. The warning does not catch all
3453 cases, but does attempt to catch the more common pitfalls. It is
3454 included in @option{-Wall}.
3455 It is equivalent to @option{-Wstrict-aliasing=3}
3457 @item -Wstrict-aliasing=n
3458 @opindex Wstrict-aliasing=n
3459 @opindex Wno-strict-aliasing=n
3460 This option is only active when @option{-fstrict-aliasing} is active.
3461 It warns about code which might break the strict aliasing rules that the
3462 compiler is using for optimization.
3463 Higher levels correspond to higher accuracy (fewer false positives).
3464 Higher levels also correspond to more effort, similar to the way -O works.
3465 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3468 Level 1: Most aggressive, quick, least accurate.
3469 Possibly useful when higher levels
3470 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3471 false negatives. However, it has many false positives.
3472 Warns for all pointer conversions between possibly incompatible types,
3473 even if never dereferenced. Runs in the frontend only.
3475 Level 2: Aggressive, quick, not too precise.
3476 May still have many false positives (not as many as level 1 though),
3477 and few false negatives (but possibly more than level 1).
3478 Unlike level 1, it only warns when an address is taken. Warns about
3479 incomplete types. Runs in the frontend only.
3481 Level 3 (default for @option{-Wstrict-aliasing}):
3482 Should have very few false positives and few false
3483 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3484 Takes care of the common punn+dereference pattern in the frontend:
3485 @code{*(int*)&some_float}.
3486 If optimization is enabled, it also runs in the backend, where it deals
3487 with multiple statement cases using flow-sensitive points-to information.
3488 Only warns when the converted pointer is dereferenced.
3489 Does not warn about incomplete types.
3491 @item -Wstrict-overflow
3492 @itemx -Wstrict-overflow=@var{n}
3493 @opindex Wstrict-overflow
3494 @opindex Wno-strict-overflow
3495 This option is only active when @option{-fstrict-overflow} is active.
3496 It warns about cases where the compiler optimizes based on the
3497 assumption that signed overflow does not occur. Note that it does not
3498 warn about all cases where the code might overflow: it only warns
3499 about cases where the compiler implements some optimization. Thus
3500 this warning depends on the optimization level.
3502 An optimization which assumes that signed overflow does not occur is
3503 perfectly safe if the values of the variables involved are such that
3504 overflow never does, in fact, occur. Therefore this warning can
3505 easily give a false positive: a warning about code which is not
3506 actually a problem. To help focus on important issues, several
3507 warning levels are defined. No warnings are issued for the use of
3508 undefined signed overflow when estimating how many iterations a loop
3509 will require, in particular when determining whether a loop will be
3513 @item -Wstrict-overflow=1
3514 Warn about cases which are both questionable and easy to avoid. For
3515 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3516 compiler will simplify this to @code{1}. This level of
3517 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3518 are not, and must be explicitly requested.
3520 @item -Wstrict-overflow=2
3521 Also warn about other cases where a comparison is simplified to a
3522 constant. For example: @code{abs (x) >= 0}. This can only be
3523 simplified when @option{-fstrict-overflow} is in effect, because
3524 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3525 zero. @option{-Wstrict-overflow} (with no level) is the same as
3526 @option{-Wstrict-overflow=2}.
3528 @item -Wstrict-overflow=3
3529 Also warn about other cases where a comparison is simplified. For
3530 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3532 @item -Wstrict-overflow=4
3533 Also warn about other simplifications not covered by the above cases.
3534 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3536 @item -Wstrict-overflow=5
3537 Also warn about cases where the compiler reduces the magnitude of a
3538 constant involved in a comparison. For example: @code{x + 2 > y} will
3539 be simplified to @code{x + 1 >= y}. This is reported only at the
3540 highest warning level because this simplification applies to many
3541 comparisons, so this warning level will give a very large number of
3545 @item -Warray-bounds
3546 @opindex Wno-array-bounds
3547 @opindex Warray-bounds
3548 This option is only active when @option{-ftree-vrp} is active
3549 (default for -O2 and above). It warns about subscripts to arrays
3550 that are always out of bounds. This warning is enabled by @option{-Wall}.
3552 @item -Wno-div-by-zero
3553 @opindex Wno-div-by-zero
3554 @opindex Wdiv-by-zero
3555 Do not warn about compile-time integer division by zero. Floating point
3556 division by zero is not warned about, as it can be a legitimate way of
3557 obtaining infinities and NaNs.
3559 @item -Wsystem-headers
3560 @opindex Wsystem-headers
3561 @opindex Wno-system-headers
3562 @cindex warnings from system headers
3563 @cindex system headers, warnings from
3564 Print warning messages for constructs found in system header files.
3565 Warnings from system headers are normally suppressed, on the assumption
3566 that they usually do not indicate real problems and would only make the
3567 compiler output harder to read. Using this command line option tells
3568 GCC to emit warnings from system headers as if they occurred in user
3569 code. However, note that using @option{-Wall} in conjunction with this
3570 option will @emph{not} warn about unknown pragmas in system
3571 headers---for that, @option{-Wunknown-pragmas} must also be used.
3574 @opindex Wfloat-equal
3575 @opindex Wno-float-equal
3576 Warn if floating point values are used in equality comparisons.
3578 The idea behind this is that sometimes it is convenient (for the
3579 programmer) to consider floating-point values as approximations to
3580 infinitely precise real numbers. If you are doing this, then you need
3581 to compute (by analyzing the code, or in some other way) the maximum or
3582 likely maximum error that the computation introduces, and allow for it
3583 when performing comparisons (and when producing output, but that's a
3584 different problem). In particular, instead of testing for equality, you
3585 would check to see whether the two values have ranges that overlap; and
3586 this is done with the relational operators, so equality comparisons are
3589 @item -Wtraditional @r{(C and Objective-C only)}
3590 @opindex Wtraditional
3591 @opindex Wno-traditional
3592 Warn about certain constructs that behave differently in traditional and
3593 ISO C@. Also warn about ISO C constructs that have no traditional C
3594 equivalent, and/or problematic constructs which should be avoided.
3598 Macro parameters that appear within string literals in the macro body.
3599 In traditional C macro replacement takes place within string literals,
3600 but does not in ISO C@.
3603 In traditional C, some preprocessor directives did not exist.
3604 Traditional preprocessors would only consider a line to be a directive
3605 if the @samp{#} appeared in column 1 on the line. Therefore
3606 @option{-Wtraditional} warns about directives that traditional C
3607 understands but would ignore because the @samp{#} does not appear as the
3608 first character on the line. It also suggests you hide directives like
3609 @samp{#pragma} not understood by traditional C by indenting them. Some
3610 traditional implementations would not recognize @samp{#elif}, so it
3611 suggests avoiding it altogether.
3614 A function-like macro that appears without arguments.
3617 The unary plus operator.
3620 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3621 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3622 constants.) Note, these suffixes appear in macros defined in the system
3623 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3624 Use of these macros in user code might normally lead to spurious
3625 warnings, however GCC's integrated preprocessor has enough context to
3626 avoid warning in these cases.
3629 A function declared external in one block and then used after the end of
3633 A @code{switch} statement has an operand of type @code{long}.
3636 A non-@code{static} function declaration follows a @code{static} one.
3637 This construct is not accepted by some traditional C compilers.
3640 The ISO type of an integer constant has a different width or
3641 signedness from its traditional type. This warning is only issued if
3642 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3643 typically represent bit patterns, are not warned about.
3646 Usage of ISO string concatenation is detected.
3649 Initialization of automatic aggregates.
3652 Identifier conflicts with labels. Traditional C lacks a separate
3653 namespace for labels.
3656 Initialization of unions. If the initializer is zero, the warning is
3657 omitted. This is done under the assumption that the zero initializer in
3658 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3659 initializer warnings and relies on default initialization to zero in the
3663 Conversions by prototypes between fixed/floating point values and vice
3664 versa. The absence of these prototypes when compiling with traditional
3665 C would cause serious problems. This is a subset of the possible
3666 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3669 Use of ISO C style function definitions. This warning intentionally is
3670 @emph{not} issued for prototype declarations or variadic functions
3671 because these ISO C features will appear in your code when using
3672 libiberty's traditional C compatibility macros, @code{PARAMS} and
3673 @code{VPARAMS}. This warning is also bypassed for nested functions
3674 because that feature is already a GCC extension and thus not relevant to
3675 traditional C compatibility.
3678 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3679 @opindex Wtraditional-conversion
3680 @opindex Wno-traditional-conversion
3681 Warn if a prototype causes a type conversion that is different from what
3682 would happen to the same argument in the absence of a prototype. This
3683 includes conversions of fixed point to floating and vice versa, and
3684 conversions changing the width or signedness of a fixed point argument
3685 except when the same as the default promotion.
3687 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3688 @opindex Wdeclaration-after-statement
3689 @opindex Wno-declaration-after-statement
3690 Warn when a declaration is found after a statement in a block. This
3691 construct, known from C++, was introduced with ISO C99 and is by default
3692 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3693 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3698 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3700 @item -Wno-endif-labels
3701 @opindex Wno-endif-labels
3702 @opindex Wendif-labels
3703 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3708 Warn whenever a local variable shadows another local variable, parameter or
3709 global variable or whenever a built-in function is shadowed.
3711 @item -Wlarger-than=@var{len}
3712 @opindex Wlarger-than=@var{len}
3713 @opindex Wlarger-than-@var{len}
3714 Warn whenever an object of larger than @var{len} bytes is defined.
3716 @item -Wframe-larger-than=@var{len}
3717 @opindex Wframe-larger-than
3718 Warn if the size of a function frame is larger than @var{len} bytes.
3719 The computation done to determine the stack frame size is approximate
3720 and not conservative.
3721 The actual requirements may be somewhat greater than @var{len}
3722 even if you do not get a warning. In addition, any space allocated
3723 via @code{alloca}, variable-length arrays, or related constructs
3724 is not included by the compiler when determining
3725 whether or not to issue a warning.
3727 @item -Wunsafe-loop-optimizations
3728 @opindex Wunsafe-loop-optimizations
3729 @opindex Wno-unsafe-loop-optimizations
3730 Warn if the loop cannot be optimized because the compiler could not
3731 assume anything on the bounds of the loop indices. With
3732 @option{-funsafe-loop-optimizations} warn if the compiler made
3735 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
3736 @opindex Wno-pedantic-ms-format
3737 @opindex Wpedantic-ms-format
3738 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
3739 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
3740 depending on the MS runtime, when you are using the options @option{-Wformat}
3741 and @option{-pedantic} without gnu-extensions.
3743 @item -Wpointer-arith
3744 @opindex Wpointer-arith
3745 @opindex Wno-pointer-arith
3746 Warn about anything that depends on the ``size of'' a function type or
3747 of @code{void}. GNU C assigns these types a size of 1, for
3748 convenience in calculations with @code{void *} pointers and pointers
3749 to functions. In C++, warn also when an arithmetic operation involves
3750 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3753 @opindex Wtype-limits
3754 @opindex Wno-type-limits
3755 Warn if a comparison is always true or always false due to the limited
3756 range of the data type, but do not warn for constant expressions. For
3757 example, warn if an unsigned variable is compared against zero with
3758 @samp{<} or @samp{>=}. This warning is also enabled by
3761 @item -Wbad-function-cast @r{(C and Objective-C only)}
3762 @opindex Wbad-function-cast
3763 @opindex Wno-bad-function-cast
3764 Warn whenever a function call is cast to a non-matching type.
3765 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3767 @item -Wc++-compat @r{(C and Objective-C only)}
3768 Warn about ISO C constructs that are outside of the common subset of
3769 ISO C and ISO C++, e.g.@: request for implicit conversion from
3770 @code{void *} to a pointer to non-@code{void} type.
3772 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3773 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3774 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3775 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3779 @opindex Wno-cast-qual
3780 Warn whenever a pointer is cast so as to remove a type qualifier from
3781 the target type. For example, warn if a @code{const char *} is cast
3782 to an ordinary @code{char *}.
3784 Also warn when making a cast which introduces a type qualifier in an
3785 unsafe way. For example, casting @code{char **} to @code{const char **}
3786 is unsafe, as in this example:
3789 /* p is char ** value. */
3790 const char **q = (const char **) p;
3791 /* Assignment of readonly string to const char * is OK. */
3793 /* Now char** pointer points to read-only memory. */
3798 @opindex Wcast-align
3799 @opindex Wno-cast-align
3800 Warn whenever a pointer is cast such that the required alignment of the
3801 target is increased. For example, warn if a @code{char *} is cast to
3802 an @code{int *} on machines where integers can only be accessed at
3803 two- or four-byte boundaries.
3805 @item -Wwrite-strings
3806 @opindex Wwrite-strings
3807 @opindex Wno-write-strings
3808 When compiling C, give string constants the type @code{const
3809 char[@var{length}]} so that copying the address of one into a
3810 non-@code{const} @code{char *} pointer will get a warning. These
3811 warnings will help you find at compile time code that can try to write
3812 into a string constant, but only if you have been very careful about
3813 using @code{const} in declarations and prototypes. Otherwise, it will
3814 just be a nuisance. This is why we did not make @option{-Wall} request
3817 When compiling C++, warn about the deprecated conversion from string
3818 literals to @code{char *}. This warning is enabled by default for C++
3823 @opindex Wno-clobbered
3824 Warn for variables that might be changed by @samp{longjmp} or
3825 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
3828 @opindex Wconversion
3829 @opindex Wno-conversion
3830 Warn for implicit conversions that may alter a value. This includes
3831 conversions between real and integer, like @code{abs (x)} when
3832 @code{x} is @code{double}; conversions between signed and unsigned,
3833 like @code{unsigned ui = -1}; and conversions to smaller types, like
3834 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
3835 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
3836 changed by the conversion like in @code{abs (2.0)}. Warnings about
3837 conversions between signed and unsigned integers can be disabled by
3838 using @option{-Wno-sign-conversion}.
3840 For C++, also warn for confusing overload resolution for user-defined
3841 conversions; and conversions that will never use a type conversion
3842 operator: conversions to @code{void}, the same type, a base class or a
3843 reference to them. Warnings about conversions between signed and
3844 unsigned integers are disabled by default in C++ unless
3845 @option{-Wsign-conversion} is explicitly enabled.
3847 @item -Wno-conversion-null @r{(C++)}
3848 @opindex Wconversion-null
3849 @opindex Wno-conversion-null
3850 Do not warn for conversions between @code{NULL} and non-pointer
3851 types. @option{-Wconversion-null} is enabled by default.
3854 @opindex Wempty-body
3855 @opindex Wno-empty-body
3856 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
3857 while} statement. This warning is also enabled by @option{-Wextra}.
3859 @item -Wenum-compare
3860 @opindex Wenum-compare
3861 @opindex Wno-enum-compare
3862 Warn about a comparison between values of different enum types. In C++
3863 this warning is enabled by default. In C this warning is enabled by
3866 @item -Wjump-misses-init @r{(C, Objective-C only)}
3867 @opindex Wjump-misses-init
3868 @opindex Wno-jump-misses-init
3869 Warn if a @code{goto} statement or a @code{switch} statement jumps
3870 forward across the initialization of a variable, or jumps backward to a
3871 label after the variable has been initialized. This only warns about
3872 variables which are initialized when they are declared. This warning is
3873 only supported for C and Objective C; in C++ this sort of branch is an
3876 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
3877 can be disabled with the @option{-Wno-jump-misses-init} option.
3879 @item -Wsign-compare
3880 @opindex Wsign-compare
3881 @opindex Wno-sign-compare
3882 @cindex warning for comparison of signed and unsigned values
3883 @cindex comparison of signed and unsigned values, warning
3884 @cindex signed and unsigned values, comparison warning
3885 Warn when a comparison between signed and unsigned values could produce
3886 an incorrect result when the signed value is converted to unsigned.
3887 This warning is also enabled by @option{-Wextra}; to get the other warnings
3888 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3890 @item -Wsign-conversion
3891 @opindex Wsign-conversion
3892 @opindex Wno-sign-conversion
3893 Warn for implicit conversions that may change the sign of an integer
3894 value, like assigning a signed integer expression to an unsigned
3895 integer variable. An explicit cast silences the warning. In C, this
3896 option is enabled also by @option{-Wconversion}.
3900 @opindex Wno-address
3901 Warn about suspicious uses of memory addresses. These include using
3902 the address of a function in a conditional expression, such as
3903 @code{void func(void); if (func)}, and comparisons against the memory
3904 address of a string literal, such as @code{if (x == "abc")}. Such
3905 uses typically indicate a programmer error: the address of a function
3906 always evaluates to true, so their use in a conditional usually
3907 indicate that the programmer forgot the parentheses in a function
3908 call; and comparisons against string literals result in unspecified
3909 behavior and are not portable in C, so they usually indicate that the
3910 programmer intended to use @code{strcmp}. This warning is enabled by
3914 @opindex Wlogical-op
3915 @opindex Wno-logical-op
3916 Warn about suspicious uses of logical operators in expressions.
3917 This includes using logical operators in contexts where a
3918 bit-wise operator is likely to be expected.
3920 @item -Waggregate-return
3921 @opindex Waggregate-return
3922 @opindex Wno-aggregate-return
3923 Warn if any functions that return structures or unions are defined or
3924 called. (In languages where you can return an array, this also elicits
3927 @item -Wno-attributes
3928 @opindex Wno-attributes
3929 @opindex Wattributes
3930 Do not warn if an unexpected @code{__attribute__} is used, such as
3931 unrecognized attributes, function attributes applied to variables,
3932 etc. This will not stop errors for incorrect use of supported
3935 @item -Wno-builtin-macro-redefined
3936 @opindex Wno-builtin-macro-redefined
3937 @opindex Wbuiltin-macro-redefined
3938 Do not warn if certain built-in macros are redefined. This suppresses
3939 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
3940 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
3942 @item -Wstrict-prototypes @r{(C and Objective-C only)}
3943 @opindex Wstrict-prototypes
3944 @opindex Wno-strict-prototypes
3945 Warn if a function is declared or defined without specifying the
3946 argument types. (An old-style function definition is permitted without
3947 a warning if preceded by a declaration which specifies the argument
3950 @item -Wold-style-declaration @r{(C and Objective-C only)}
3951 @opindex Wold-style-declaration
3952 @opindex Wno-old-style-declaration
3953 Warn for obsolescent usages, according to the C Standard, in a
3954 declaration. For example, warn if storage-class specifiers like
3955 @code{static} are not the first things in a declaration. This warning
3956 is also enabled by @option{-Wextra}.
3958 @item -Wold-style-definition @r{(C and Objective-C only)}
3959 @opindex Wold-style-definition
3960 @opindex Wno-old-style-definition
3961 Warn if an old-style function definition is used. A warning is given
3962 even if there is a previous prototype.
3964 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
3965 @opindex Wmissing-parameter-type
3966 @opindex Wno-missing-parameter-type
3967 A function parameter is declared without a type specifier in K&R-style
3974 This warning is also enabled by @option{-Wextra}.
3976 @item -Wmissing-prototypes @r{(C and Objective-C only)}
3977 @opindex Wmissing-prototypes
3978 @opindex Wno-missing-prototypes
3979 Warn if a global function is defined without a previous prototype
3980 declaration. This warning is issued even if the definition itself
3981 provides a prototype. The aim is to detect global functions that fail
3982 to be declared in header files.
3984 @item -Wmissing-declarations
3985 @opindex Wmissing-declarations
3986 @opindex Wno-missing-declarations
3987 Warn if a global function is defined without a previous declaration.
3988 Do so even if the definition itself provides a prototype.
3989 Use this option to detect global functions that are not declared in
3990 header files. In C++, no warnings are issued for function templates,
3991 or for inline functions, or for functions in anonymous namespaces.
3993 @item -Wmissing-field-initializers
3994 @opindex Wmissing-field-initializers
3995 @opindex Wno-missing-field-initializers
3999 Warn if a structure's initializer has some fields missing. For
4000 example, the following code would cause such a warning, because
4001 @code{x.h} is implicitly zero:
4004 struct s @{ int f, g, h; @};
4005 struct s x = @{ 3, 4 @};
4008 This option does not warn about designated initializers, so the following
4009 modification would not trigger a warning:
4012 struct s @{ int f, g, h; @};
4013 struct s x = @{ .f = 3, .g = 4 @};
4016 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
4017 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
4019 @item -Wmissing-noreturn
4020 @opindex Wmissing-noreturn
4021 @opindex Wno-missing-noreturn
4022 Warn about functions which might be candidates for attribute @code{noreturn}.
4023 Note these are only possible candidates, not absolute ones. Care should
4024 be taken to manually verify functions actually do not ever return before
4025 adding the @code{noreturn} attribute, otherwise subtle code generation
4026 bugs could be introduced. You will not get a warning for @code{main} in
4027 hosted C environments.
4029 @item -Wmissing-format-attribute
4030 @opindex Wmissing-format-attribute
4031 @opindex Wno-missing-format-attribute
4034 Warn about function pointers which might be candidates for @code{format}
4035 attributes. Note these are only possible candidates, not absolute ones.
4036 GCC will guess that function pointers with @code{format} attributes that
4037 are used in assignment, initialization, parameter passing or return
4038 statements should have a corresponding @code{format} attribute in the
4039 resulting type. I.e.@: the left-hand side of the assignment or
4040 initialization, the type of the parameter variable, or the return type
4041 of the containing function respectively should also have a @code{format}
4042 attribute to avoid the warning.
4044 GCC will also warn about function definitions which might be
4045 candidates for @code{format} attributes. Again, these are only
4046 possible candidates. GCC will guess that @code{format} attributes
4047 might be appropriate for any function that calls a function like
4048 @code{vprintf} or @code{vscanf}, but this might not always be the
4049 case, and some functions for which @code{format} attributes are
4050 appropriate may not be detected.
4052 @item -Wno-multichar
4053 @opindex Wno-multichar
4055 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4056 Usually they indicate a typo in the user's code, as they have
4057 implementation-defined values, and should not be used in portable code.
4059 @item -Wnormalized=<none|id|nfc|nfkc>
4060 @opindex Wnormalized=
4063 @cindex character set, input normalization
4064 In ISO C and ISO C++, two identifiers are different if they are
4065 different sequences of characters. However, sometimes when characters
4066 outside the basic ASCII character set are used, you can have two
4067 different character sequences that look the same. To avoid confusion,
4068 the ISO 10646 standard sets out some @dfn{normalization rules} which
4069 when applied ensure that two sequences that look the same are turned into
4070 the same sequence. GCC can warn you if you are using identifiers which
4071 have not been normalized; this option controls that warning.
4073 There are four levels of warning that GCC supports. The default is
4074 @option{-Wnormalized=nfc}, which warns about any identifier which is
4075 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4076 recommended form for most uses.
4078 Unfortunately, there are some characters which ISO C and ISO C++ allow
4079 in identifiers that when turned into NFC aren't allowable as
4080 identifiers. That is, there's no way to use these symbols in portable
4081 ISO C or C++ and have all your identifiers in NFC@.
4082 @option{-Wnormalized=id} suppresses the warning for these characters.
4083 It is hoped that future versions of the standards involved will correct
4084 this, which is why this option is not the default.
4086 You can switch the warning off for all characters by writing
4087 @option{-Wnormalized=none}. You would only want to do this if you
4088 were using some other normalization scheme (like ``D''), because
4089 otherwise you can easily create bugs that are literally impossible to see.
4091 Some characters in ISO 10646 have distinct meanings but look identical
4092 in some fonts or display methodologies, especially once formatting has
4093 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4094 LETTER N'', will display just like a regular @code{n} which has been
4095 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4096 normalization scheme to convert all these into a standard form as
4097 well, and GCC will warn if your code is not in NFKC if you use
4098 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4099 about every identifier that contains the letter O because it might be
4100 confused with the digit 0, and so is not the default, but may be
4101 useful as a local coding convention if the programming environment is
4102 unable to be fixed to display these characters distinctly.
4104 @item -Wno-deprecated
4105 @opindex Wno-deprecated
4106 @opindex Wdeprecated
4107 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4109 @item -Wno-deprecated-declarations
4110 @opindex Wno-deprecated-declarations
4111 @opindex Wdeprecated-declarations
4112 Do not warn about uses of functions (@pxref{Function Attributes}),
4113 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4114 Attributes}) marked as deprecated by using the @code{deprecated}
4118 @opindex Wno-overflow
4120 Do not warn about compile-time overflow in constant expressions.
4122 @item -Woverride-init @r{(C and Objective-C only)}
4123 @opindex Woverride-init
4124 @opindex Wno-override-init
4128 Warn if an initialized field without side effects is overridden when
4129 using designated initializers (@pxref{Designated Inits, , Designated
4132 This warning is included in @option{-Wextra}. To get other
4133 @option{-Wextra} warnings without this one, use @samp{-Wextra
4134 -Wno-override-init}.
4139 Warn if a structure is given the packed attribute, but the packed
4140 attribute has no effect on the layout or size of the structure.
4141 Such structures may be mis-aligned for little benefit. For
4142 instance, in this code, the variable @code{f.x} in @code{struct bar}
4143 will be misaligned even though @code{struct bar} does not itself
4144 have the packed attribute:
4151 @} __attribute__((packed));
4159 @item -Wpacked-bitfield-compat
4160 @opindex Wpacked-bitfield-compat
4161 @opindex Wno-packed-bitfield-compat
4162 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4163 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4164 the change can lead to differences in the structure layout. GCC
4165 informs you when the offset of such a field has changed in GCC 4.4.
4166 For example there is no longer a 4-bit padding between field @code{a}
4167 and @code{b} in this structure:
4174 @} __attribute__ ((packed));
4177 This warning is enabled by default. Use
4178 @option{-Wno-packed-bitfield-compat} to disable this warning.
4183 Warn if padding is included in a structure, either to align an element
4184 of the structure or to align the whole structure. Sometimes when this
4185 happens it is possible to rearrange the fields of the structure to
4186 reduce the padding and so make the structure smaller.
4188 @item -Wredundant-decls
4189 @opindex Wredundant-decls
4190 @opindex Wno-redundant-decls
4191 Warn if anything is declared more than once in the same scope, even in
4192 cases where multiple declaration is valid and changes nothing.
4194 @item -Wnested-externs @r{(C and Objective-C only)}
4195 @opindex Wnested-externs
4196 @opindex Wno-nested-externs
4197 Warn if an @code{extern} declaration is encountered within a function.
4202 Warn if a function can not be inlined and it was declared as inline.
4203 Even with this option, the compiler will not warn about failures to
4204 inline functions declared in system headers.
4206 The compiler uses a variety of heuristics to determine whether or not
4207 to inline a function. For example, the compiler takes into account
4208 the size of the function being inlined and the amount of inlining
4209 that has already been done in the current function. Therefore,
4210 seemingly insignificant changes in the source program can cause the
4211 warnings produced by @option{-Winline} to appear or disappear.
4213 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4214 @opindex Wno-invalid-offsetof
4215 @opindex Winvalid-offsetof
4216 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4217 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4218 to a non-POD type is undefined. In existing C++ implementations,
4219 however, @samp{offsetof} typically gives meaningful results even when
4220 applied to certain kinds of non-POD types. (Such as a simple
4221 @samp{struct} that fails to be a POD type only by virtue of having a
4222 constructor.) This flag is for users who are aware that they are
4223 writing nonportable code and who have deliberately chosen to ignore the
4226 The restrictions on @samp{offsetof} may be relaxed in a future version
4227 of the C++ standard.
4229 @item -Wno-int-to-pointer-cast @r{(C and Objective-C only)}
4230 @opindex Wno-int-to-pointer-cast
4231 @opindex Wint-to-pointer-cast
4232 Suppress warnings from casts to pointer type of an integer of a
4235 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4236 @opindex Wno-pointer-to-int-cast
4237 @opindex Wpointer-to-int-cast
4238 Suppress warnings from casts from a pointer to an integer type of a
4242 @opindex Winvalid-pch
4243 @opindex Wno-invalid-pch
4244 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4245 the search path but can't be used.
4249 @opindex Wno-long-long
4250 Warn if @samp{long long} type is used. This is enabled by either
4251 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4252 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4254 @item -Wvariadic-macros
4255 @opindex Wvariadic-macros
4256 @opindex Wno-variadic-macros
4257 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4258 alternate syntax when in pedantic ISO C99 mode. This is default.
4259 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4264 Warn if variable length array is used in the code.
4265 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4266 the variable length array.
4268 @item -Wvolatile-register-var
4269 @opindex Wvolatile-register-var
4270 @opindex Wno-volatile-register-var
4271 Warn if a register variable is declared volatile. The volatile
4272 modifier does not inhibit all optimizations that may eliminate reads
4273 and/or writes to register variables. This warning is enabled by
4276 @item -Wdisabled-optimization
4277 @opindex Wdisabled-optimization
4278 @opindex Wno-disabled-optimization
4279 Warn if a requested optimization pass is disabled. This warning does
4280 not generally indicate that there is anything wrong with your code; it
4281 merely indicates that GCC's optimizers were unable to handle the code
4282 effectively. Often, the problem is that your code is too big or too
4283 complex; GCC will refuse to optimize programs when the optimization
4284 itself is likely to take inordinate amounts of time.
4286 @item -Wpointer-sign @r{(C and Objective-C only)}
4287 @opindex Wpointer-sign
4288 @opindex Wno-pointer-sign
4289 Warn for pointer argument passing or assignment with different signedness.
4290 This option is only supported for C and Objective-C@. It is implied by
4291 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4292 @option{-Wno-pointer-sign}.
4294 @item -Wstack-protector
4295 @opindex Wstack-protector
4296 @opindex Wno-stack-protector
4297 This option is only active when @option{-fstack-protector} is active. It
4298 warns about functions that will not be protected against stack smashing.
4301 @opindex Wno-mudflap
4302 Suppress warnings about constructs that cannot be instrumented by
4305 @item -Woverlength-strings
4306 @opindex Woverlength-strings
4307 @opindex Wno-overlength-strings
4308 Warn about string constants which are longer than the ``minimum
4309 maximum'' length specified in the C standard. Modern compilers
4310 generally allow string constants which are much longer than the
4311 standard's minimum limit, but very portable programs should avoid
4312 using longer strings.
4314 The limit applies @emph{after} string constant concatenation, and does
4315 not count the trailing NUL@. In C89, the limit was 509 characters; in
4316 C99, it was raised to 4095. C++98 does not specify a normative
4317 minimum maximum, so we do not diagnose overlength strings in C++@.
4319 This option is implied by @option{-pedantic}, and can be disabled with
4320 @option{-Wno-overlength-strings}.
4322 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
4323 @opindex Wunsuffixed-float-constants
4325 GCC will issue a warning for any floating constant that does not have
4326 a suffix. When used together with @option{-Wsystem-headers} it will
4327 warn about such constants in system header files. This can be useful
4328 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4329 from the decimal floating-point extension to C99.
4332 @node Debugging Options
4333 @section Options for Debugging Your Program or GCC
4334 @cindex options, debugging
4335 @cindex debugging information options
4337 GCC has various special options that are used for debugging
4338 either your program or GCC:
4343 Produce debugging information in the operating system's native format
4344 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4347 On most systems that use stabs format, @option{-g} enables use of extra
4348 debugging information that only GDB can use; this extra information
4349 makes debugging work better in GDB but will probably make other debuggers
4351 refuse to read the program. If you want to control for certain whether
4352 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4353 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4355 GCC allows you to use @option{-g} with
4356 @option{-O}. The shortcuts taken by optimized code may occasionally
4357 produce surprising results: some variables you declared may not exist
4358 at all; flow of control may briefly move where you did not expect it;
4359 some statements may not be executed because they compute constant
4360 results or their values were already at hand; some statements may
4361 execute in different places because they were moved out of loops.
4363 Nevertheless it proves possible to debug optimized output. This makes
4364 it reasonable to use the optimizer for programs that might have bugs.
4366 The following options are useful when GCC is generated with the
4367 capability for more than one debugging format.
4371 Produce debugging information for use by GDB@. This means to use the
4372 most expressive format available (DWARF 2, stabs, or the native format
4373 if neither of those are supported), including GDB extensions if at all
4378 Produce debugging information in stabs format (if that is supported),
4379 without GDB extensions. This is the format used by DBX on most BSD
4380 systems. On MIPS, Alpha and System V Release 4 systems this option
4381 produces stabs debugging output which is not understood by DBX or SDB@.
4382 On System V Release 4 systems this option requires the GNU assembler.
4384 @item -feliminate-unused-debug-symbols
4385 @opindex feliminate-unused-debug-symbols
4386 Produce debugging information in stabs format (if that is supported),
4387 for only symbols that are actually used.
4389 @item -femit-class-debug-always
4390 Instead of emitting debugging information for a C++ class in only one
4391 object file, emit it in all object files using the class. This option
4392 should be used only with debuggers that are unable to handle the way GCC
4393 normally emits debugging information for classes because using this
4394 option will increase the size of debugging information by as much as a
4399 Produce debugging information in stabs format (if that is supported),
4400 using GNU extensions understood only by the GNU debugger (GDB)@. The
4401 use of these extensions is likely to make other debuggers crash or
4402 refuse to read the program.
4406 Produce debugging information in COFF format (if that is supported).
4407 This is the format used by SDB on most System V systems prior to
4412 Produce debugging information in XCOFF format (if that is supported).
4413 This is the format used by the DBX debugger on IBM RS/6000 systems.
4417 Produce debugging information in XCOFF format (if that is supported),
4418 using GNU extensions understood only by the GNU debugger (GDB)@. The
4419 use of these extensions is likely to make other debuggers crash or
4420 refuse to read the program, and may cause assemblers other than the GNU
4421 assembler (GAS) to fail with an error.
4423 @item -gdwarf-@var{version}
4424 @opindex gdwarf-@var{version}
4425 Produce debugging information in DWARF format (if that is
4426 supported). This is the format used by DBX on IRIX 6. The value
4427 of @var{version} may be either 2, 3 or 4; the default version is 2.
4429 Note that with DWARF version 2 some ports require, and will always
4430 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4432 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
4433 for maximum benefit.
4435 @item -gstrict-dwarf
4436 @opindex gstrict-dwarf
4437 Disallow using extensions of later DWARF standard version than selected
4438 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
4439 DWARF extensions from later standard versions is allowed.
4441 @item -gno-strict-dwarf
4442 @opindex gno-strict-dwarf
4443 Allow using extensions of later DWARF standard version than selected with
4444 @option{-gdwarf-@var{version}}.
4448 Produce debugging information in VMS debug format (if that is
4449 supported). This is the format used by DEBUG on VMS systems.
4452 @itemx -ggdb@var{level}
4453 @itemx -gstabs@var{level}
4454 @itemx -gcoff@var{level}
4455 @itemx -gxcoff@var{level}
4456 @itemx -gvms@var{level}
4457 Request debugging information and also use @var{level} to specify how
4458 much information. The default level is 2.
4460 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4463 Level 1 produces minimal information, enough for making backtraces in
4464 parts of the program that you don't plan to debug. This includes
4465 descriptions of functions and external variables, but no information
4466 about local variables and no line numbers.
4468 Level 3 includes extra information, such as all the macro definitions
4469 present in the program. Some debuggers support macro expansion when
4470 you use @option{-g3}.
4472 @option{-gdwarf-2} does not accept a concatenated debug level, because
4473 GCC used to support an option @option{-gdwarf} that meant to generate
4474 debug information in version 1 of the DWARF format (which is very
4475 different from version 2), and it would have been too confusing. That
4476 debug format is long obsolete, but the option cannot be changed now.
4477 Instead use an additional @option{-g@var{level}} option to change the
4478 debug level for DWARF.
4482 Turn off generation of debug info, if leaving out this option would have
4483 generated it, or turn it on at level 2 otherwise. The position of this
4484 argument in the command line does not matter, it takes effect after all
4485 other options are processed, and it does so only once, no matter how
4486 many times it is given. This is mainly intended to be used with
4487 @option{-fcompare-debug}.
4489 @item -fdump-final-insns@r{[}=@var{file}@r{]}
4490 @opindex fdump-final-insns
4491 Dump the final internal representation (RTL) to @var{file}. If the
4492 optional argument is omitted (or if @var{file} is @code{.}), the name
4493 of the dump file will be determined by appending @code{.gkd} to the
4494 compilation output file name.
4496 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4497 @opindex fcompare-debug
4498 @opindex fno-compare-debug
4499 If no error occurs during compilation, run the compiler a second time,
4500 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4501 passed to the second compilation. Dump the final internal
4502 representation in both compilations, and print an error if they differ.
4504 If the equal sign is omitted, the default @option{-gtoggle} is used.
4506 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4507 and nonzero, implicitly enables @option{-fcompare-debug}. If
4508 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4509 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4512 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4513 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4514 of the final representation and the second compilation, preventing even
4515 @env{GCC_COMPARE_DEBUG} from taking effect.
4517 To verify full coverage during @option{-fcompare-debug} testing, set
4518 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4519 which GCC will reject as an invalid option in any actual compilation
4520 (rather than preprocessing, assembly or linking). To get just a
4521 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4522 not overridden} will do.
4524 @item -fcompare-debug-second
4525 @opindex fcompare-debug-second
4526 This option is implicitly passed to the compiler for the second
4527 compilation requested by @option{-fcompare-debug}, along with options to
4528 silence warnings, and omitting other options that would cause
4529 side-effect compiler outputs to files or to the standard output. Dump
4530 files and preserved temporary files are renamed so as to contain the
4531 @code{.gk} additional extension during the second compilation, to avoid
4532 overwriting those generated by the first.
4534 When this option is passed to the compiler driver, it causes the
4535 @emph{first} compilation to be skipped, which makes it useful for little
4536 other than debugging the compiler proper.
4538 @item -feliminate-dwarf2-dups
4539 @opindex feliminate-dwarf2-dups
4540 Compress DWARF2 debugging information by eliminating duplicated
4541 information about each symbol. This option only makes sense when
4542 generating DWARF2 debugging information with @option{-gdwarf-2}.
4544 @item -femit-struct-debug-baseonly
4545 Emit debug information for struct-like types
4546 only when the base name of the compilation source file
4547 matches the base name of file in which the struct was defined.
4549 This option substantially reduces the size of debugging information,
4550 but at significant potential loss in type information to the debugger.
4551 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4552 See @option{-femit-struct-debug-detailed} for more detailed control.
4554 This option works only with DWARF 2.
4556 @item -femit-struct-debug-reduced
4557 Emit debug information for struct-like types
4558 only when the base name of the compilation source file
4559 matches the base name of file in which the type was defined,
4560 unless the struct is a template or defined in a system header.
4562 This option significantly reduces the size of debugging information,
4563 with some potential loss in type information to the debugger.
4564 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4565 See @option{-femit-struct-debug-detailed} for more detailed control.
4567 This option works only with DWARF 2.
4569 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4570 Specify the struct-like types
4571 for which the compiler will generate debug information.
4572 The intent is to reduce duplicate struct debug information
4573 between different object files within the same program.
4575 This option is a detailed version of
4576 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4577 which will serve for most needs.
4579 A specification has the syntax
4580 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4582 The optional first word limits the specification to
4583 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4584 A struct type is used directly when it is the type of a variable, member.
4585 Indirect uses arise through pointers to structs.
4586 That is, when use of an incomplete struct would be legal, the use is indirect.
4588 @samp{struct one direct; struct two * indirect;}.
4590 The optional second word limits the specification to
4591 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4592 Generic structs are a bit complicated to explain.
4593 For C++, these are non-explicit specializations of template classes,
4594 or non-template classes within the above.
4595 Other programming languages have generics,
4596 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4598 The third word specifies the source files for those
4599 structs for which the compiler will emit debug information.
4600 The values @samp{none} and @samp{any} have the normal meaning.
4601 The value @samp{base} means that
4602 the base of name of the file in which the type declaration appears
4603 must match the base of the name of the main compilation file.
4604 In practice, this means that
4605 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4606 but types declared in other header will not.
4607 The value @samp{sys} means those types satisfying @samp{base}
4608 or declared in system or compiler headers.
4610 You may need to experiment to determine the best settings for your application.
4612 The default is @samp{-femit-struct-debug-detailed=all}.
4614 This option works only with DWARF 2.
4616 @item -fenable-icf-debug
4617 @opindex fenable-icf-debug
4618 Generate additional debug information to support identical code folding (ICF).
4619 This option only works with DWARF version 2 or higher.
4621 @item -fno-merge-debug-strings
4622 @opindex fmerge-debug-strings
4623 @opindex fno-merge-debug-strings
4624 Direct the linker to not merge together strings in the debugging
4625 information which are identical in different object files. Merging is
4626 not supported by all assemblers or linkers. Merging decreases the size
4627 of the debug information in the output file at the cost of increasing
4628 link processing time. Merging is enabled by default.
4630 @item -fdebug-prefix-map=@var{old}=@var{new}
4631 @opindex fdebug-prefix-map
4632 When compiling files in directory @file{@var{old}}, record debugging
4633 information describing them as in @file{@var{new}} instead.
4635 @item -fno-dwarf2-cfi-asm
4636 @opindex fdwarf2-cfi-asm
4637 @opindex fno-dwarf2-cfi-asm
4638 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4639 instead of using GAS @code{.cfi_*} directives.
4641 @cindex @command{prof}
4644 Generate extra code to write profile information suitable for the
4645 analysis program @command{prof}. You must use this option when compiling
4646 the source files you want data about, and you must also use it when
4649 @cindex @command{gprof}
4652 Generate extra code to write profile information suitable for the
4653 analysis program @command{gprof}. You must use this option when compiling
4654 the source files you want data about, and you must also use it when
4659 Makes the compiler print out each function name as it is compiled, and
4660 print some statistics about each pass when it finishes.
4663 @opindex ftime-report
4664 Makes the compiler print some statistics about the time consumed by each
4665 pass when it finishes.
4668 @opindex fmem-report
4669 Makes the compiler print some statistics about permanent memory
4670 allocation when it finishes.
4672 @item -fpre-ipa-mem-report
4673 @opindex fpre-ipa-mem-report
4674 @item -fpost-ipa-mem-report
4675 @opindex fpost-ipa-mem-report
4676 Makes the compiler print some statistics about permanent memory
4677 allocation before or after interprocedural optimization.
4679 @item -fprofile-arcs
4680 @opindex fprofile-arcs
4681 Add code so that program flow @dfn{arcs} are instrumented. During
4682 execution the program records how many times each branch and call is
4683 executed and how many times it is taken or returns. When the compiled
4684 program exits it saves this data to a file called
4685 @file{@var{auxname}.gcda} for each source file. The data may be used for
4686 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4687 test coverage analysis (@option{-ftest-coverage}). Each object file's
4688 @var{auxname} is generated from the name of the output file, if
4689 explicitly specified and it is not the final executable, otherwise it is
4690 the basename of the source file. In both cases any suffix is removed
4691 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4692 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4693 @xref{Cross-profiling}.
4695 @cindex @command{gcov}
4699 This option is used to compile and link code instrumented for coverage
4700 analysis. The option is a synonym for @option{-fprofile-arcs}
4701 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4702 linking). See the documentation for those options for more details.
4707 Compile the source files with @option{-fprofile-arcs} plus optimization
4708 and code generation options. For test coverage analysis, use the
4709 additional @option{-ftest-coverage} option. You do not need to profile
4710 every source file in a program.
4713 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4714 (the latter implies the former).
4717 Run the program on a representative workload to generate the arc profile
4718 information. This may be repeated any number of times. You can run
4719 concurrent instances of your program, and provided that the file system
4720 supports locking, the data files will be correctly updated. Also
4721 @code{fork} calls are detected and correctly handled (double counting
4725 For profile-directed optimizations, compile the source files again with
4726 the same optimization and code generation options plus
4727 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4728 Control Optimization}).
4731 For test coverage analysis, use @command{gcov} to produce human readable
4732 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4733 @command{gcov} documentation for further information.
4737 With @option{-fprofile-arcs}, for each function of your program GCC
4738 creates a program flow graph, then finds a spanning tree for the graph.
4739 Only arcs that are not on the spanning tree have to be instrumented: the
4740 compiler adds code to count the number of times that these arcs are
4741 executed. When an arc is the only exit or only entrance to a block, the
4742 instrumentation code can be added to the block; otherwise, a new basic
4743 block must be created to hold the instrumentation code.
4746 @item -ftest-coverage
4747 @opindex ftest-coverage
4748 Produce a notes file that the @command{gcov} code-coverage utility
4749 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4750 show program coverage. Each source file's note file is called
4751 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4752 above for a description of @var{auxname} and instructions on how to
4753 generate test coverage data. Coverage data will match the source files
4754 more closely, if you do not optimize.
4756 @item -fdbg-cnt-list
4757 @opindex fdbg-cnt-list
4758 Print the name and the counter upperbound for all debug counters.
4760 @item -fdbg-cnt=@var{counter-value-list}
4762 Set the internal debug counter upperbound. @var{counter-value-list}
4763 is a comma-separated list of @var{name}:@var{value} pairs
4764 which sets the upperbound of each debug counter @var{name} to @var{value}.
4765 All debug counters have the initial upperbound of @var{UINT_MAX},
4766 thus dbg_cnt() returns true always unless the upperbound is set by this option.
4767 e.g. With -fdbg-cnt=dce:10,tail_call:0
4768 dbg_cnt(dce) will return true only for first 10 invocations
4769 and dbg_cnt(tail_call) will return false always.
4771 @item -d@var{letters}
4772 @itemx -fdump-rtl-@var{pass}
4774 Says to make debugging dumps during compilation at times specified by
4775 @var{letters}. This is used for debugging the RTL-based passes of the
4776 compiler. The file names for most of the dumps are made by appending
4777 a pass number and a word to the @var{dumpname}, and the files are
4778 created in the directory of the output file. @var{dumpname} is
4779 generated from the name of the output file, if explicitly specified
4780 and it is not an executable, otherwise it is the basename of the
4781 source file. These switches may have different effects when
4782 @option{-E} is used for preprocessing.
4784 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
4785 @option{-d} option @var{letters}. Here are the possible
4786 letters for use in @var{pass} and @var{letters}, and their meanings:
4790 @item -fdump-rtl-alignments
4791 @opindex fdump-rtl-alignments
4792 Dump after branch alignments have been computed.
4794 @item -fdump-rtl-asmcons
4795 @opindex fdump-rtl-asmcons
4796 Dump after fixing rtl statements that have unsatisfied in/out constraints.
4798 @item -fdump-rtl-auto_inc_dec
4799 @opindex fdump-rtl-auto_inc_dec
4800 Dump after auto-inc-dec discovery. This pass is only run on
4801 architectures that have auto inc or auto dec instructions.
4803 @item -fdump-rtl-barriers
4804 @opindex fdump-rtl-barriers
4805 Dump after cleaning up the barrier instructions.
4807 @item -fdump-rtl-bbpart
4808 @opindex fdump-rtl-bbpart
4809 Dump after partitioning hot and cold basic blocks.
4811 @item -fdump-rtl-bbro
4812 @opindex fdump-rtl-bbro
4813 Dump after block reordering.
4815 @item -fdump-rtl-btl1
4816 @itemx -fdump-rtl-btl2
4817 @opindex fdump-rtl-btl2
4818 @opindex fdump-rtl-btl2
4819 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
4820 after the two branch
4821 target load optimization passes.
4823 @item -fdump-rtl-bypass
4824 @opindex fdump-rtl-bypass
4825 Dump after jump bypassing and control flow optimizations.
4827 @item -fdump-rtl-combine
4828 @opindex fdump-rtl-combine
4829 Dump after the RTL instruction combination pass.
4831 @item -fdump-rtl-compgotos
4832 @opindex fdump-rtl-compgotos
4833 Dump after duplicating the computed gotos.
4835 @item -fdump-rtl-ce1
4836 @itemx -fdump-rtl-ce2
4837 @itemx -fdump-rtl-ce3
4838 @opindex fdump-rtl-ce1
4839 @opindex fdump-rtl-ce2
4840 @opindex fdump-rtl-ce3
4841 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
4842 @option{-fdump-rtl-ce3} enable dumping after the three
4843 if conversion passes.
4845 @itemx -fdump-rtl-cprop_hardreg
4846 @opindex fdump-rtl-cprop_hardreg
4847 Dump after hard register copy propagation.
4849 @itemx -fdump-rtl-csa
4850 @opindex fdump-rtl-csa
4851 Dump after combining stack adjustments.
4853 @item -fdump-rtl-cse1
4854 @itemx -fdump-rtl-cse2
4855 @opindex fdump-rtl-cse1
4856 @opindex fdump-rtl-cse2
4857 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
4858 the two common sub-expression elimination passes.
4860 @itemx -fdump-rtl-dce
4861 @opindex fdump-rtl-dce
4862 Dump after the standalone dead code elimination passes.
4864 @itemx -fdump-rtl-dbr
4865 @opindex fdump-rtl-dbr
4866 Dump after delayed branch scheduling.
4868 @item -fdump-rtl-dce1
4869 @itemx -fdump-rtl-dce2
4870 @opindex fdump-rtl-dce1
4871 @opindex fdump-rtl-dce2
4872 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
4873 the two dead store elimination passes.
4876 @opindex fdump-rtl-eh
4877 Dump after finalization of EH handling code.
4879 @item -fdump-rtl-eh_ranges
4880 @opindex fdump-rtl-eh_ranges
4881 Dump after conversion of EH handling range regions.
4883 @item -fdump-rtl-expand
4884 @opindex fdump-rtl-expand
4885 Dump after RTL generation.
4887 @item -fdump-rtl-fwprop1
4888 @itemx -fdump-rtl-fwprop2
4889 @opindex fdump-rtl-fwprop1
4890 @opindex fdump-rtl-fwprop2
4891 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
4892 dumping after the two forward propagation passes.
4894 @item -fdump-rtl-gcse1
4895 @itemx -fdump-rtl-gcse2
4896 @opindex fdump-rtl-gcse1
4897 @opindex fdump-rtl-gcse2
4898 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
4899 after global common subexpression elimination.
4901 @item -fdump-rtl-init-regs
4902 @opindex fdump-rtl-init-regs
4903 Dump after the initialization of the registers.
4905 @item -fdump-rtl-initvals
4906 @opindex fdump-rtl-initvals
4907 Dump after the computation of the initial value sets.
4909 @itemx -fdump-rtl-into_cfglayout
4910 @opindex fdump-rtl-into_cfglayout
4911 Dump after converting to cfglayout mode.
4913 @item -fdump-rtl-ira
4914 @opindex fdump-rtl-ira
4915 Dump after iterated register allocation.
4917 @item -fdump-rtl-jump
4918 @opindex fdump-rtl-jump
4919 Dump after the second jump optimization.
4921 @item -fdump-rtl-loop2
4922 @opindex fdump-rtl-loop2
4923 @option{-fdump-rtl-loop2} enables dumping after the rtl
4924 loop optimization passes.
4926 @item -fdump-rtl-mach
4927 @opindex fdump-rtl-mach
4928 Dump after performing the machine dependent reorganization pass, if that
4931 @item -fdump-rtl-mode_sw
4932 @opindex fdump-rtl-mode_sw
4933 Dump after removing redundant mode switches.
4935 @item -fdump-rtl-rnreg
4936 @opindex fdump-rtl-rnreg
4937 Dump after register renumbering.
4939 @itemx -fdump-rtl-outof_cfglayout
4940 @opindex fdump-rtl-outof_cfglayout
4941 Dump after converting from cfglayout mode.
4943 @item -fdump-rtl-peephole2
4944 @opindex fdump-rtl-peephole2
4945 Dump after the peephole pass.
4947 @item -fdump-rtl-postreload
4948 @opindex fdump-rtl-postreload
4949 Dump after post-reload optimizations.
4951 @itemx -fdump-rtl-pro_and_epilogue
4952 @opindex fdump-rtl-pro_and_epilogue
4953 Dump after generating the function pro and epilogues.
4955 @item -fdump-rtl-regmove
4956 @opindex fdump-rtl-regmove
4957 Dump after the register move pass.
4959 @item -fdump-rtl-sched1
4960 @itemx -fdump-rtl-sched2
4961 @opindex fdump-rtl-sched1
4962 @opindex fdump-rtl-sched2
4963 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
4964 after the basic block scheduling passes.
4966 @item -fdump-rtl-see
4967 @opindex fdump-rtl-see
4968 Dump after sign extension elimination.
4970 @item -fdump-rtl-seqabstr
4971 @opindex fdump-rtl-seqabstr
4972 Dump after common sequence discovery.
4974 @item -fdump-rtl-shorten
4975 @opindex fdump-rtl-shorten
4976 Dump after shortening branches.
4978 @item -fdump-rtl-sibling
4979 @opindex fdump-rtl-sibling
4980 Dump after sibling call optimizations.
4982 @item -fdump-rtl-split1
4983 @itemx -fdump-rtl-split2
4984 @itemx -fdump-rtl-split3
4985 @itemx -fdump-rtl-split4
4986 @itemx -fdump-rtl-split5
4987 @opindex fdump-rtl-split1
4988 @opindex fdump-rtl-split2
4989 @opindex fdump-rtl-split3
4990 @opindex fdump-rtl-split4
4991 @opindex fdump-rtl-split5
4992 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
4993 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
4994 @option{-fdump-rtl-split5} enable dumping after five rounds of
4995 instruction splitting.
4997 @item -fdump-rtl-sms
4998 @opindex fdump-rtl-sms
4999 Dump after modulo scheduling. This pass is only run on some
5002 @item -fdump-rtl-stack
5003 @opindex fdump-rtl-stack
5004 Dump after conversion from GCC's "flat register file" registers to the
5005 x87's stack-like registers. This pass is only run on x86 variants.
5007 @item -fdump-rtl-subreg1
5008 @itemx -fdump-rtl-subreg2
5009 @opindex fdump-rtl-subreg1
5010 @opindex fdump-rtl-subreg2
5011 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
5012 the two subreg expansion passes.
5014 @item -fdump-rtl-unshare
5015 @opindex fdump-rtl-unshare
5016 Dump after all rtl has been unshared.
5018 @item -fdump-rtl-vartrack
5019 @opindex fdump-rtl-vartrack
5020 Dump after variable tracking.
5022 @item -fdump-rtl-vregs
5023 @opindex fdump-rtl-vregs
5024 Dump after converting virtual registers to hard registers.
5026 @item -fdump-rtl-web
5027 @opindex fdump-rtl-web
5028 Dump after live range splitting.
5030 @item -fdump-rtl-regclass
5031 @itemx -fdump-rtl-subregs_of_mode_init
5032 @itemx -fdump-rtl-subregs_of_mode_finish
5033 @itemx -fdump-rtl-dfinit
5034 @itemx -fdump-rtl-dfinish
5035 @opindex fdump-rtl-regclass
5036 @opindex fdump-rtl-subregs_of_mode_init
5037 @opindex fdump-rtl-subregs_of_mode_finish
5038 @opindex fdump-rtl-dfinit
5039 @opindex fdump-rtl-dfinish
5040 These dumps are defined but always produce empty files.
5042 @item -fdump-rtl-all
5043 @opindex fdump-rtl-all
5044 Produce all the dumps listed above.
5048 Annotate the assembler output with miscellaneous debugging information.
5052 Dump all macro definitions, at the end of preprocessing, in addition to
5057 Produce a core dump whenever an error occurs.
5061 Print statistics on memory usage, at the end of the run, to
5066 Annotate the assembler output with a comment indicating which
5067 pattern and alternative was used. The length of each instruction is
5072 Dump the RTL in the assembler output as a comment before each instruction.
5073 Also turns on @option{-dp} annotation.
5077 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
5078 dump a representation of the control flow graph suitable for viewing with VCG
5079 to @file{@var{file}.@var{pass}.vcg}.
5083 Just generate RTL for a function instead of compiling it. Usually used
5084 with @option{-fdump-rtl-expand}.
5088 Dump debugging information during parsing, to standard error.
5092 @opindex fdump-noaddr
5093 When doing debugging dumps, suppress address output. This makes it more
5094 feasible to use diff on debugging dumps for compiler invocations with
5095 different compiler binaries and/or different
5096 text / bss / data / heap / stack / dso start locations.
5098 @item -fdump-unnumbered
5099 @opindex fdump-unnumbered
5100 When doing debugging dumps, suppress instruction numbers and address output.
5101 This makes it more feasible to use diff on debugging dumps for compiler
5102 invocations with different options, in particular with and without
5105 @item -fdump-unnumbered-links
5106 @opindex fdump-unnumbered-links
5107 When doing debugging dumps (see @option{-d} option above), suppress
5108 instruction numbers for the links to the previous and next instructions
5111 @item -fdump-translation-unit @r{(C++ only)}
5112 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5113 @opindex fdump-translation-unit
5114 Dump a representation of the tree structure for the entire translation
5115 unit to a file. The file name is made by appending @file{.tu} to the
5116 source file name, and the file is created in the same directory as the
5117 output file. If the @samp{-@var{options}} form is used, @var{options}
5118 controls the details of the dump as described for the
5119 @option{-fdump-tree} options.
5121 @item -fdump-class-hierarchy @r{(C++ only)}
5122 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5123 @opindex fdump-class-hierarchy
5124 Dump a representation of each class's hierarchy and virtual function
5125 table layout to a file. The file name is made by appending
5126 @file{.class} to the source file name, and the file is created in the
5127 same directory as the output file. If the @samp{-@var{options}} form
5128 is used, @var{options} controls the details of the dump as described
5129 for the @option{-fdump-tree} options.
5131 @item -fdump-ipa-@var{switch}
5133 Control the dumping at various stages of inter-procedural analysis
5134 language tree to a file. The file name is generated by appending a
5135 switch specific suffix to the source file name, and the file is created
5136 in the same directory as the output file. The following dumps are
5141 Enables all inter-procedural analysis dumps.
5144 Dumps information about call-graph optimization, unused function removal,
5145 and inlining decisions.
5148 Dump after function inlining.
5152 @item -fdump-statistics-@var{option}
5153 @opindex fdump-statistics
5154 Enable and control dumping of pass statistics in a separate file. The
5155 file name is generated by appending a suffix ending in
5156 @samp{.statistics} to the source file name, and the file is created in
5157 the same directory as the output file. If the @samp{-@var{option}}
5158 form is used, @samp{-stats} will cause counters to be summed over the
5159 whole compilation unit while @samp{-details} will dump every event as
5160 the passes generate them. The default with no option is to sum
5161 counters for each function compiled.
5163 @item -fdump-tree-@var{switch}
5164 @itemx -fdump-tree-@var{switch}-@var{options}
5166 Control the dumping at various stages of processing the intermediate
5167 language tree to a file. The file name is generated by appending a
5168 switch specific suffix to the source file name, and the file is
5169 created in the same directory as the output file. If the
5170 @samp{-@var{options}} form is used, @var{options} is a list of
5171 @samp{-} separated options that control the details of the dump. Not
5172 all options are applicable to all dumps, those which are not
5173 meaningful will be ignored. The following options are available
5177 Print the address of each node. Usually this is not meaningful as it
5178 changes according to the environment and source file. Its primary use
5179 is for tying up a dump file with a debug environment.
5181 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
5182 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
5183 use working backward from mangled names in the assembly file.
5185 Inhibit dumping of members of a scope or body of a function merely
5186 because that scope has been reached. Only dump such items when they
5187 are directly reachable by some other path. When dumping pretty-printed
5188 trees, this option inhibits dumping the bodies of control structures.
5190 Print a raw representation of the tree. By default, trees are
5191 pretty-printed into a C-like representation.
5193 Enable more detailed dumps (not honored by every dump option).
5195 Enable dumping various statistics about the pass (not honored by every dump
5198 Enable showing basic block boundaries (disabled in raw dumps).
5200 Enable showing virtual operands for every statement.
5202 Enable showing line numbers for statements.
5204 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5206 Enable showing the tree dump for each statement.
5208 Enable showing the EH region number holding each statement.
5210 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5211 and @option{lineno}.
5214 The following tree dumps are possible:
5218 @opindex fdump-tree-original
5219 Dump before any tree based optimization, to @file{@var{file}.original}.
5222 @opindex fdump-tree-optimized
5223 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5226 @opindex fdump-tree-gimple
5227 Dump each function before and after the gimplification pass to a file. The
5228 file name is made by appending @file{.gimple} to the source file name.
5231 @opindex fdump-tree-cfg
5232 Dump the control flow graph of each function to a file. The file name is
5233 made by appending @file{.cfg} to the source file name.
5236 @opindex fdump-tree-vcg
5237 Dump the control flow graph of each function to a file in VCG format. The
5238 file name is made by appending @file{.vcg} to the source file name. Note
5239 that if the file contains more than one function, the generated file cannot
5240 be used directly by VCG@. You will need to cut and paste each function's
5241 graph into its own separate file first.
5244 @opindex fdump-tree-ch
5245 Dump each function after copying loop headers. The file name is made by
5246 appending @file{.ch} to the source file name.
5249 @opindex fdump-tree-ssa
5250 Dump SSA related information to a file. The file name is made by appending
5251 @file{.ssa} to the source file name.
5254 @opindex fdump-tree-alias
5255 Dump aliasing information for each function. The file name is made by
5256 appending @file{.alias} to the source file name.
5259 @opindex fdump-tree-ccp
5260 Dump each function after CCP@. The file name is made by appending
5261 @file{.ccp} to the source file name.
5264 @opindex fdump-tree-storeccp
5265 Dump each function after STORE-CCP@. The file name is made by appending
5266 @file{.storeccp} to the source file name.
5269 @opindex fdump-tree-pre
5270 Dump trees after partial redundancy elimination. The file name is made
5271 by appending @file{.pre} to the source file name.
5274 @opindex fdump-tree-fre
5275 Dump trees after full redundancy elimination. The file name is made
5276 by appending @file{.fre} to the source file name.
5279 @opindex fdump-tree-copyprop
5280 Dump trees after copy propagation. The file name is made
5281 by appending @file{.copyprop} to the source file name.
5283 @item store_copyprop
5284 @opindex fdump-tree-store_copyprop
5285 Dump trees after store copy-propagation. The file name is made
5286 by appending @file{.store_copyprop} to the source file name.
5289 @opindex fdump-tree-dce
5290 Dump each function after dead code elimination. The file name is made by
5291 appending @file{.dce} to the source file name.
5294 @opindex fdump-tree-mudflap
5295 Dump each function after adding mudflap instrumentation. The file name is
5296 made by appending @file{.mudflap} to the source file name.
5299 @opindex fdump-tree-sra
5300 Dump each function after performing scalar replacement of aggregates. The
5301 file name is made by appending @file{.sra} to the source file name.
5304 @opindex fdump-tree-sink
5305 Dump each function after performing code sinking. The file name is made
5306 by appending @file{.sink} to the source file name.
5309 @opindex fdump-tree-dom
5310 Dump each function after applying dominator tree optimizations. The file
5311 name is made by appending @file{.dom} to the source file name.
5314 @opindex fdump-tree-dse
5315 Dump each function after applying dead store elimination. The file
5316 name is made by appending @file{.dse} to the source file name.
5319 @opindex fdump-tree-phiopt
5320 Dump each function after optimizing PHI nodes into straightline code. The file
5321 name is made by appending @file{.phiopt} to the source file name.
5324 @opindex fdump-tree-forwprop
5325 Dump each function after forward propagating single use variables. The file
5326 name is made by appending @file{.forwprop} to the source file name.
5329 @opindex fdump-tree-copyrename
5330 Dump each function after applying the copy rename optimization. The file
5331 name is made by appending @file{.copyrename} to the source file name.
5334 @opindex fdump-tree-nrv
5335 Dump each function after applying the named return value optimization on
5336 generic trees. The file name is made by appending @file{.nrv} to the source
5340 @opindex fdump-tree-vect
5341 Dump each function after applying vectorization of loops. The file name is
5342 made by appending @file{.vect} to the source file name.
5345 @opindex fdump-tree-slp
5346 Dump each function after applying vectorization of basic blocks. The file name
5347 is made by appending @file{.slp} to the source file name.
5350 @opindex fdump-tree-vrp
5351 Dump each function after Value Range Propagation (VRP). The file name
5352 is made by appending @file{.vrp} to the source file name.
5355 @opindex fdump-tree-all
5356 Enable all the available tree dumps with the flags provided in this option.
5359 @item -ftree-vectorizer-verbose=@var{n}
5360 @opindex ftree-vectorizer-verbose
5361 This option controls the amount of debugging output the vectorizer prints.
5362 This information is written to standard error, unless
5363 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5364 in which case it is output to the usual dump listing file, @file{.vect}.
5365 For @var{n}=0 no diagnostic information is reported.
5366 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5367 and the total number of loops that got vectorized.
5368 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5369 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5370 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5371 level that @option{-fdump-tree-vect-stats} uses.
5372 Higher verbosity levels mean either more information dumped for each
5373 reported loop, or same amount of information reported for more loops:
5374 if @var{n}=3, vectorizer cost model information is reported.
5375 If @var{n}=4, alignment related information is added to the reports.
5376 If @var{n}=5, data-references related information (e.g.@: memory dependences,
5377 memory access-patterns) is added to the reports.
5378 If @var{n}=6, the vectorizer reports also non-vectorized inner-most loops
5379 that did not pass the first analysis phase (i.e., may not be countable, or
5380 may have complicated control-flow).
5381 If @var{n}=7, the vectorizer reports also non-vectorized nested loops.
5382 If @var{n}=8, SLP related information is added to the reports.
5383 For @var{n}=9, all the information the vectorizer generates during its
5384 analysis and transformation is reported. This is the same verbosity level
5385 that @option{-fdump-tree-vect-details} uses.
5387 @item -frandom-seed=@var{string}
5388 @opindex frandom-seed
5389 This option provides a seed that GCC uses when it would otherwise use
5390 random numbers. It is used to generate certain symbol names
5391 that have to be different in every compiled file. It is also used to
5392 place unique stamps in coverage data files and the object files that
5393 produce them. You can use the @option{-frandom-seed} option to produce
5394 reproducibly identical object files.
5396 The @var{string} should be different for every file you compile.
5398 @item -fsched-verbose=@var{n}
5399 @opindex fsched-verbose
5400 On targets that use instruction scheduling, this option controls the
5401 amount of debugging output the scheduler prints. This information is
5402 written to standard error, unless @option{-fdump-rtl-sched1} or
5403 @option{-fdump-rtl-sched2} is specified, in which case it is output
5404 to the usual dump listing file, @file{.sched1} or @file{.sched2}
5405 respectively. However for @var{n} greater than nine, the output is
5406 always printed to standard error.
5408 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5409 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5410 For @var{n} greater than one, it also output basic block probabilities,
5411 detailed ready list information and unit/insn info. For @var{n} greater
5412 than two, it includes RTL at abort point, control-flow and regions info.
5413 And for @var{n} over four, @option{-fsched-verbose} also includes
5417 @itemx -save-temps=cwd
5419 Store the usual ``temporary'' intermediate files permanently; place them
5420 in the current directory and name them based on the source file. Thus,
5421 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5422 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5423 preprocessed @file{foo.i} output file even though the compiler now
5424 normally uses an integrated preprocessor.
5426 When used in combination with the @option{-x} command line option,
5427 @option{-save-temps} is sensible enough to avoid over writing an
5428 input source file with the same extension as an intermediate file.
5429 The corresponding intermediate file may be obtained by renaming the
5430 source file before using @option{-save-temps}.
5432 If you invoke GCC in parallel, compiling several different source
5433 files that share a common base name in different subdirectories or the
5434 same source file compiled for multiple output destinations, it is
5435 likely that the different parallel compilers will interfere with each
5436 other, and overwrite the temporary files. For instance:
5439 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5440 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5443 may result in @file{foo.i} and @file{foo.o} being written to
5444 simultaneously by both compilers.
5446 @item -save-temps=obj
5447 @opindex save-temps=obj
5448 Store the usual ``temporary'' intermediate files permanently. If the
5449 @option{-o} option is used, the temporary files are based on the
5450 object file. If the @option{-o} option is not used, the
5451 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5456 gcc -save-temps=obj -c foo.c
5457 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5458 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5461 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5462 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5463 @file{dir2/yfoobar.o}.
5465 @item -time@r{[}=@var{file}@r{]}
5467 Report the CPU time taken by each subprocess in the compilation
5468 sequence. For C source files, this is the compiler proper and assembler
5469 (plus the linker if linking is done).
5471 Without the specification of an output file, the output looks like this:
5478 The first number on each line is the ``user time'', that is time spent
5479 executing the program itself. The second number is ``system time'',
5480 time spent executing operating system routines on behalf of the program.
5481 Both numbers are in seconds.
5483 With the specification of an output file, the output is appended to the
5484 named file, and it looks like this:
5487 0.12 0.01 cc1 @var{options}
5488 0.00 0.01 as @var{options}
5491 The ``user time'' and the ``system time'' are moved before the program
5492 name, and the options passed to the program are displayed, so that one
5493 can later tell what file was being compiled, and with which options.
5495 @item -fvar-tracking
5496 @opindex fvar-tracking
5497 Run variable tracking pass. It computes where variables are stored at each
5498 position in code. Better debugging information is then generated
5499 (if the debugging information format supports this information).
5501 It is enabled by default when compiling with optimization (@option{-Os},
5502 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5503 the debug info format supports it.
5505 @item -fvar-tracking-assignments
5506 @opindex fvar-tracking-assignments
5507 @opindex fno-var-tracking-assignments
5508 Annotate assignments to user variables early in the compilation and
5509 attempt to carry the annotations over throughout the compilation all the
5510 way to the end, in an attempt to improve debug information while
5511 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
5513 It can be enabled even if var-tracking is disabled, in which case
5514 annotations will be created and maintained, but discarded at the end.
5516 @item -fvar-tracking-assignments-toggle
5517 @opindex fvar-tracking-assignments-toggle
5518 @opindex fno-var-tracking-assignments-toggle
5519 Toggle @option{-fvar-tracking-assignments}, in the same way that
5520 @option{-gtoggle} toggles @option{-g}.
5522 @item -print-file-name=@var{library}
5523 @opindex print-file-name
5524 Print the full absolute name of the library file @var{library} that
5525 would be used when linking---and don't do anything else. With this
5526 option, GCC does not compile or link anything; it just prints the
5529 @item -print-multi-directory
5530 @opindex print-multi-directory
5531 Print the directory name corresponding to the multilib selected by any
5532 other switches present in the command line. This directory is supposed
5533 to exist in @env{GCC_EXEC_PREFIX}.
5535 @item -print-multi-lib
5536 @opindex print-multi-lib
5537 Print the mapping from multilib directory names to compiler switches
5538 that enable them. The directory name is separated from the switches by
5539 @samp{;}, and each switch starts with an @samp{@@} instead of the
5540 @samp{-}, without spaces between multiple switches. This is supposed to
5541 ease shell-processing.
5543 @item -print-multi-os-directory
5544 @opindex print-multi-os-directory
5545 Print the path to OS libraries for the selected
5546 multilib, relative to some @file{lib} subdirectory. If OS libraries are
5547 present in the @file{lib} subdirectory and no multilibs are used, this is
5548 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
5549 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
5550 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
5551 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
5553 @item -print-prog-name=@var{program}
5554 @opindex print-prog-name
5555 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5557 @item -print-libgcc-file-name
5558 @opindex print-libgcc-file-name
5559 Same as @option{-print-file-name=libgcc.a}.
5561 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5562 but you do want to link with @file{libgcc.a}. You can do
5565 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5568 @item -print-search-dirs
5569 @opindex print-search-dirs
5570 Print the name of the configured installation directory and a list of
5571 program and library directories @command{gcc} will search---and don't do anything else.
5573 This is useful when @command{gcc} prints the error message
5574 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5575 To resolve this you either need to put @file{cpp0} and the other compiler
5576 components where @command{gcc} expects to find them, or you can set the environment
5577 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5578 Don't forget the trailing @samp{/}.
5579 @xref{Environment Variables}.
5581 @item -print-sysroot
5582 @opindex print-sysroot
5583 Print the target sysroot directory that will be used during
5584 compilation. This is the target sysroot specified either at configure
5585 time or using the @option{--sysroot} option, possibly with an extra
5586 suffix that depends on compilation options. If no target sysroot is
5587 specified, the option prints nothing.
5589 @item -print-sysroot-headers-suffix
5590 @opindex print-sysroot-headers-suffix
5591 Print the suffix added to the target sysroot when searching for
5592 headers, or give an error if the compiler is not configured with such
5593 a suffix---and don't do anything else.
5596 @opindex dumpmachine
5597 Print the compiler's target machine (for example,
5598 @samp{i686-pc-linux-gnu})---and don't do anything else.
5601 @opindex dumpversion
5602 Print the compiler version (for example, @samp{3.0})---and don't do
5607 Print the compiler's built-in specs---and don't do anything else. (This
5608 is used when GCC itself is being built.) @xref{Spec Files}.
5610 @item -feliminate-unused-debug-types
5611 @opindex feliminate-unused-debug-types
5612 Normally, when producing DWARF2 output, GCC will emit debugging
5613 information for all types declared in a compilation
5614 unit, regardless of whether or not they are actually used
5615 in that compilation unit. Sometimes this is useful, such as
5616 if, in the debugger, you want to cast a value to a type that is
5617 not actually used in your program (but is declared). More often,
5618 however, this results in a significant amount of wasted space.
5619 With this option, GCC will avoid producing debug symbol output
5620 for types that are nowhere used in the source file being compiled.
5623 @node Optimize Options
5624 @section Options That Control Optimization
5625 @cindex optimize options
5626 @cindex options, optimization
5628 These options control various sorts of optimizations.
5630 Without any optimization option, the compiler's goal is to reduce the
5631 cost of compilation and to make debugging produce the expected
5632 results. Statements are independent: if you stop the program with a
5633 breakpoint between statements, you can then assign a new value to any
5634 variable or change the program counter to any other statement in the
5635 function and get exactly the results you would expect from the source
5638 Turning on optimization flags makes the compiler attempt to improve
5639 the performance and/or code size at the expense of compilation time
5640 and possibly the ability to debug the program.
5642 The compiler performs optimization based on the knowledge it has of the
5643 program. Compiling multiple files at once to a single output file mode allows
5644 the compiler to use information gained from all of the files when compiling
5647 Not all optimizations are controlled directly by a flag. Only
5648 optimizations that have a flag are listed in this section.
5650 Most optimizations are only enabled if an @option{-O} level is set on
5651 the command line. Otherwise they are disabled, even if individual
5652 optimization flags are specified.
5654 Depending on the target and how GCC was configured, a slightly different
5655 set of optimizations may be enabled at each @option{-O} level than
5656 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
5657 to find out the exact set of optimizations that are enabled at each level.
5658 @xref{Overall Options}, for examples.
5665 Optimize. Optimizing compilation takes somewhat more time, and a lot
5666 more memory for a large function.
5668 With @option{-O}, the compiler tries to reduce code size and execution
5669 time, without performing any optimizations that take a great deal of
5672 @option{-O} turns on the following optimization flags:
5675 -fcprop-registers @gol
5678 -fdelayed-branch @gol
5680 -fguess-branch-probability @gol
5681 -fif-conversion2 @gol
5682 -fif-conversion @gol
5683 -fipa-pure-const @gol
5684 -fipa-reference @gol
5686 -fsplit-wide-types @gol
5687 -ftree-builtin-call-dce @gol
5690 -ftree-copyrename @gol
5692 -ftree-dominator-opts @gol
5694 -ftree-forwprop @gol
5702 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5703 where doing so does not interfere with debugging.
5707 Optimize even more. GCC performs nearly all supported optimizations
5708 that do not involve a space-speed tradeoff.
5709 As compared to @option{-O}, this option increases both compilation time
5710 and the performance of the generated code.
5712 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5713 also turns on the following optimization flags:
5714 @gccoptlist{-fthread-jumps @gol
5715 -falign-functions -falign-jumps @gol
5716 -falign-loops -falign-labels @gol
5719 -fcse-follow-jumps -fcse-skip-blocks @gol
5720 -fdelete-null-pointer-checks @gol
5721 -fexpensive-optimizations @gol
5722 -fgcse -fgcse-lm @gol
5723 -finline-small-functions @gol
5724 -findirect-inlining @gol
5726 -foptimize-sibling-calls @gol
5729 -freorder-blocks -freorder-functions @gol
5730 -frerun-cse-after-loop @gol
5731 -fsched-interblock -fsched-spec @gol
5732 -fschedule-insns -fschedule-insns2 @gol
5733 -fstrict-aliasing -fstrict-overflow @gol
5734 -ftree-switch-conversion @gol
5738 Please note the warning under @option{-fgcse} about
5739 invoking @option{-O2} on programs that use computed gotos.
5743 Optimize yet more. @option{-O3} turns on all optimizations specified
5744 by @option{-O2} and also turns on the @option{-finline-functions},
5745 @option{-funswitch-loops}, @option{-fpredictive-commoning},
5746 @option{-fgcse-after-reload} and @option{-ftree-vectorize} options.
5750 Reduce compilation time and make debugging produce the expected
5751 results. This is the default.
5755 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5756 do not typically increase code size. It also performs further
5757 optimizations designed to reduce code size.
5759 @option{-Os} disables the following optimization flags:
5760 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5761 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5762 -fprefetch-loop-arrays -ftree-vect-loop-version}
5764 If you use multiple @option{-O} options, with or without level numbers,
5765 the last such option is the one that is effective.
5768 Options of the form @option{-f@var{flag}} specify machine-independent
5769 flags. Most flags have both positive and negative forms; the negative
5770 form of @option{-ffoo} would be @option{-fno-foo}. In the table
5771 below, only one of the forms is listed---the one you typically will
5772 use. You can figure out the other form by either removing @samp{no-}
5775 The following options control specific optimizations. They are either
5776 activated by @option{-O} options or are related to ones that are. You
5777 can use the following flags in the rare cases when ``fine-tuning'' of
5778 optimizations to be performed is desired.
5781 @item -fno-default-inline
5782 @opindex fno-default-inline
5783 Do not make member functions inline by default merely because they are
5784 defined inside the class scope (C++ only). Otherwise, when you specify
5785 @w{@option{-O}}, member functions defined inside class scope are compiled
5786 inline by default; i.e., you don't need to add @samp{inline} in front of
5787 the member function name.
5789 @item -fno-defer-pop
5790 @opindex fno-defer-pop
5791 Always pop the arguments to each function call as soon as that function
5792 returns. For machines which must pop arguments after a function call,
5793 the compiler normally lets arguments accumulate on the stack for several
5794 function calls and pops them all at once.
5796 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5798 @item -fforward-propagate
5799 @opindex fforward-propagate
5800 Perform a forward propagation pass on RTL@. The pass tries to combine two
5801 instructions and checks if the result can be simplified. If loop unrolling
5802 is active, two passes are performed and the second is scheduled after
5805 This option is enabled by default at optimization levels @option{-O},
5806 @option{-O2}, @option{-O3}, @option{-Os}.
5808 @item -fomit-frame-pointer
5809 @opindex fomit-frame-pointer
5810 Don't keep the frame pointer in a register for functions that
5811 don't need one. This avoids the instructions to save, set up and
5812 restore frame pointers; it also makes an extra register available
5813 in many functions. @strong{It also makes debugging impossible on
5816 On some machines, such as the VAX, this flag has no effect, because
5817 the standard calling sequence automatically handles the frame pointer
5818 and nothing is saved by pretending it doesn't exist. The
5819 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
5820 whether a target machine supports this flag. @xref{Registers,,Register
5821 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
5823 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5825 @item -foptimize-sibling-calls
5826 @opindex foptimize-sibling-calls
5827 Optimize sibling and tail recursive calls.
5829 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5833 Don't pay attention to the @code{inline} keyword. Normally this option
5834 is used to keep the compiler from expanding any functions inline.
5835 Note that if you are not optimizing, no functions can be expanded inline.
5837 @item -finline-small-functions
5838 @opindex finline-small-functions
5839 Integrate functions into their callers when their body is smaller than expected
5840 function call code (so overall size of program gets smaller). The compiler
5841 heuristically decides which functions are simple enough to be worth integrating
5844 Enabled at level @option{-O2}.
5846 @item -findirect-inlining
5847 @opindex findirect-inlining
5848 Inline also indirect calls that are discovered to be known at compile
5849 time thanks to previous inlining. This option has any effect only
5850 when inlining itself is turned on by the @option{-finline-functions}
5851 or @option{-finline-small-functions} options.
5853 Enabled at level @option{-O2}.
5855 @item -finline-functions
5856 @opindex finline-functions
5857 Integrate all simple functions into their callers. The compiler
5858 heuristically decides which functions are simple enough to be worth
5859 integrating in this way.
5861 If all calls to a given function are integrated, and the function is
5862 declared @code{static}, then the function is normally not output as
5863 assembler code in its own right.
5865 Enabled at level @option{-O3}.
5867 @item -finline-functions-called-once
5868 @opindex finline-functions-called-once
5869 Consider all @code{static} functions called once for inlining into their
5870 caller even if they are not marked @code{inline}. If a call to a given
5871 function is integrated, then the function is not output as assembler code
5874 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
5876 @item -fearly-inlining
5877 @opindex fearly-inlining
5878 Inline functions marked by @code{always_inline} and functions whose body seems
5879 smaller than the function call overhead early before doing
5880 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
5881 makes profiling significantly cheaper and usually inlining faster on programs
5882 having large chains of nested wrapper functions.
5888 Perform interprocedural scalar replacement of aggregates, removal of
5889 unused parameters and replacement of parameters passed by reference
5890 by parameters passed by value.
5892 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
5894 @item -finline-limit=@var{n}
5895 @opindex finline-limit
5896 By default, GCC limits the size of functions that can be inlined. This flag
5897 allows coarse control of this limit. @var{n} is the size of functions that
5898 can be inlined in number of pseudo instructions.
5900 Inlining is actually controlled by a number of parameters, which may be
5901 specified individually by using @option{--param @var{name}=@var{value}}.
5902 The @option{-finline-limit=@var{n}} option sets some of these parameters
5906 @item max-inline-insns-single
5907 is set to @var{n}/2.
5908 @item max-inline-insns-auto
5909 is set to @var{n}/2.
5912 See below for a documentation of the individual
5913 parameters controlling inlining and for the defaults of these parameters.
5915 @emph{Note:} there may be no value to @option{-finline-limit} that results
5916 in default behavior.
5918 @emph{Note:} pseudo instruction represents, in this particular context, an
5919 abstract measurement of function's size. In no way does it represent a count
5920 of assembly instructions and as such its exact meaning might change from one
5921 release to an another.
5923 @item -fkeep-inline-functions
5924 @opindex fkeep-inline-functions
5925 In C, emit @code{static} functions that are declared @code{inline}
5926 into the object file, even if the function has been inlined into all
5927 of its callers. This switch does not affect functions using the
5928 @code{extern inline} extension in GNU C89@. In C++, emit any and all
5929 inline functions into the object file.
5931 @item -fkeep-static-consts
5932 @opindex fkeep-static-consts
5933 Emit variables declared @code{static const} when optimization isn't turned
5934 on, even if the variables aren't referenced.
5936 GCC enables this option by default. If you want to force the compiler to
5937 check if the variable was referenced, regardless of whether or not
5938 optimization is turned on, use the @option{-fno-keep-static-consts} option.
5940 @item -fmerge-constants
5941 @opindex fmerge-constants
5942 Attempt to merge identical constants (string constants and floating point
5943 constants) across compilation units.
5945 This option is the default for optimized compilation if the assembler and
5946 linker support it. Use @option{-fno-merge-constants} to inhibit this
5949 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5951 @item -fmerge-all-constants
5952 @opindex fmerge-all-constants
5953 Attempt to merge identical constants and identical variables.
5955 This option implies @option{-fmerge-constants}. In addition to
5956 @option{-fmerge-constants} this considers e.g.@: even constant initialized
5957 arrays or initialized constant variables with integral or floating point
5958 types. Languages like C or C++ require each variable, including multiple
5959 instances of the same variable in recursive calls, to have distinct locations,
5960 so using this option will result in non-conforming
5963 @item -fmodulo-sched
5964 @opindex fmodulo-sched
5965 Perform swing modulo scheduling immediately before the first scheduling
5966 pass. This pass looks at innermost loops and reorders their
5967 instructions by overlapping different iterations.
5969 @item -fmodulo-sched-allow-regmoves
5970 @opindex fmodulo-sched-allow-regmoves
5971 Perform more aggressive SMS based modulo scheduling with register moves
5972 allowed. By setting this flag certain anti-dependences edges will be
5973 deleted which will trigger the generation of reg-moves based on the
5974 life-range analysis. This option is effective only with
5975 @option{-fmodulo-sched} enabled.
5977 @item -fno-branch-count-reg
5978 @opindex fno-branch-count-reg
5979 Do not use ``decrement and branch'' instructions on a count register,
5980 but instead generate a sequence of instructions that decrement a
5981 register, compare it against zero, then branch based upon the result.
5982 This option is only meaningful on architectures that support such
5983 instructions, which include x86, PowerPC, IA-64 and S/390.
5985 The default is @option{-fbranch-count-reg}.
5987 @item -fno-function-cse
5988 @opindex fno-function-cse
5989 Do not put function addresses in registers; make each instruction that
5990 calls a constant function contain the function's address explicitly.
5992 This option results in less efficient code, but some strange hacks
5993 that alter the assembler output may be confused by the optimizations
5994 performed when this option is not used.
5996 The default is @option{-ffunction-cse}
5998 @item -fno-zero-initialized-in-bss
5999 @opindex fno-zero-initialized-in-bss
6000 If the target supports a BSS section, GCC by default puts variables that
6001 are initialized to zero into BSS@. This can save space in the resulting
6004 This option turns off this behavior because some programs explicitly
6005 rely on variables going to the data section. E.g., so that the
6006 resulting executable can find the beginning of that section and/or make
6007 assumptions based on that.
6009 The default is @option{-fzero-initialized-in-bss}.
6011 @item -fmudflap -fmudflapth -fmudflapir
6015 @cindex bounds checking
6017 For front-ends that support it (C and C++), instrument all risky
6018 pointer/array dereferencing operations, some standard library
6019 string/heap functions, and some other associated constructs with
6020 range/validity tests. Modules so instrumented should be immune to
6021 buffer overflows, invalid heap use, and some other classes of C/C++
6022 programming errors. The instrumentation relies on a separate runtime
6023 library (@file{libmudflap}), which will be linked into a program if
6024 @option{-fmudflap} is given at link time. Run-time behavior of the
6025 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
6026 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
6029 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
6030 link if your program is multi-threaded. Use @option{-fmudflapir}, in
6031 addition to @option{-fmudflap} or @option{-fmudflapth}, if
6032 instrumentation should ignore pointer reads. This produces less
6033 instrumentation (and therefore faster execution) and still provides
6034 some protection against outright memory corrupting writes, but allows
6035 erroneously read data to propagate within a program.
6037 @item -fthread-jumps
6038 @opindex fthread-jumps
6039 Perform optimizations where we check to see if a jump branches to a
6040 location where another comparison subsumed by the first is found. If
6041 so, the first branch is redirected to either the destination of the
6042 second branch or a point immediately following it, depending on whether
6043 the condition is known to be true or false.
6045 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6047 @item -fsplit-wide-types
6048 @opindex fsplit-wide-types
6049 When using a type that occupies multiple registers, such as @code{long
6050 long} on a 32-bit system, split the registers apart and allocate them
6051 independently. This normally generates better code for those types,
6052 but may make debugging more difficult.
6054 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
6057 @item -fcse-follow-jumps
6058 @opindex fcse-follow-jumps
6059 In common subexpression elimination (CSE), scan through jump instructions
6060 when the target of the jump is not reached by any other path. For
6061 example, when CSE encounters an @code{if} statement with an
6062 @code{else} clause, CSE will follow the jump when the condition
6065 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6067 @item -fcse-skip-blocks
6068 @opindex fcse-skip-blocks
6069 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
6070 follow jumps which conditionally skip over blocks. When CSE
6071 encounters a simple @code{if} statement with no else clause,
6072 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6073 body of the @code{if}.
6075 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6077 @item -frerun-cse-after-loop
6078 @opindex frerun-cse-after-loop
6079 Re-run common subexpression elimination after loop optimizations has been
6082 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6086 Perform a global common subexpression elimination pass.
6087 This pass also performs global constant and copy propagation.
6089 @emph{Note:} When compiling a program using computed gotos, a GCC
6090 extension, you may get better runtime performance if you disable
6091 the global common subexpression elimination pass by adding
6092 @option{-fno-gcse} to the command line.
6094 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6098 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
6099 attempt to move loads which are only killed by stores into themselves. This
6100 allows a loop containing a load/store sequence to be changed to a load outside
6101 the loop, and a copy/store within the loop.
6103 Enabled by default when gcse is enabled.
6107 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6108 global common subexpression elimination. This pass will attempt to move
6109 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6110 loops containing a load/store sequence can be changed to a load before
6111 the loop and a store after the loop.
6113 Not enabled at any optimization level.
6117 When @option{-fgcse-las} is enabled, the global common subexpression
6118 elimination pass eliminates redundant loads that come after stores to the
6119 same memory location (both partial and full redundancies).
6121 Not enabled at any optimization level.
6123 @item -fgcse-after-reload
6124 @opindex fgcse-after-reload
6125 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6126 pass is performed after reload. The purpose of this pass is to cleanup
6129 @item -funsafe-loop-optimizations
6130 @opindex funsafe-loop-optimizations
6131 If given, the loop optimizer will assume that loop indices do not
6132 overflow, and that the loops with nontrivial exit condition are not
6133 infinite. This enables a wider range of loop optimizations even if
6134 the loop optimizer itself cannot prove that these assumptions are valid.
6135 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
6136 if it finds this kind of loop.
6138 @item -fcrossjumping
6139 @opindex fcrossjumping
6140 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6141 resulting code may or may not perform better than without cross-jumping.
6143 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6145 @item -fauto-inc-dec
6146 @opindex fauto-inc-dec
6147 Combine increments or decrements of addresses with memory accesses.
6148 This pass is always skipped on architectures that do not have
6149 instructions to support this. Enabled by default at @option{-O} and
6150 higher on architectures that support this.
6154 Perform dead code elimination (DCE) on RTL@.
6155 Enabled by default at @option{-O} and higher.
6159 Perform dead store elimination (DSE) on RTL@.
6160 Enabled by default at @option{-O} and higher.
6162 @item -fif-conversion
6163 @opindex fif-conversion
6164 Attempt to transform conditional jumps into branch-less equivalents. This
6165 include use of conditional moves, min, max, set flags and abs instructions, and
6166 some tricks doable by standard arithmetics. The use of conditional execution
6167 on chips where it is available is controlled by @code{if-conversion2}.
6169 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6171 @item -fif-conversion2
6172 @opindex fif-conversion2
6173 Use conditional execution (where available) to transform conditional jumps into
6174 branch-less equivalents.
6176 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6178 @item -fdelete-null-pointer-checks
6179 @opindex fdelete-null-pointer-checks
6180 Assume that programs cannot safely dereference null pointers, and that
6181 no code or data element resides there. This enables simple constant
6182 folding optimizations at all optimization levels. In addition, other
6183 optimization passes in GCC use this flag to control global dataflow
6184 analyses that eliminate useless checks for null pointers; these assume
6185 that if a pointer is checked after it has already been dereferenced,
6188 Note however that in some environments this assumption is not true.
6189 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6190 for programs which depend on that behavior.
6192 Some targets, especially embedded ones, disable this option at all levels.
6193 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6194 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6195 are enabled independently at different optimization levels.
6197 @item -fexpensive-optimizations
6198 @opindex fexpensive-optimizations
6199 Perform a number of minor optimizations that are relatively expensive.
6201 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6203 @item -foptimize-register-move
6205 @opindex foptimize-register-move
6207 Attempt to reassign register numbers in move instructions and as
6208 operands of other simple instructions in order to maximize the amount of
6209 register tying. This is especially helpful on machines with two-operand
6212 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6215 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6217 @item -fira-algorithm=@var{algorithm}
6218 Use specified coloring algorithm for the integrated register
6219 allocator. The @var{algorithm} argument should be @code{priority} or
6220 @code{CB}. The first algorithm specifies Chow's priority coloring,
6221 the second one specifies Chaitin-Briggs coloring. The second
6222 algorithm can be unimplemented for some architectures. If it is
6223 implemented, it is the default because Chaitin-Briggs coloring as a
6224 rule generates a better code.
6226 @item -fira-region=@var{region}
6227 Use specified regions for the integrated register allocator. The
6228 @var{region} argument should be one of @code{all}, @code{mixed}, or
6229 @code{one}. The first value means using all loops as register
6230 allocation regions, the second value which is the default means using
6231 all loops except for loops with small register pressure as the
6232 regions, and third one means using all function as a single region.
6233 The first value can give best result for machines with small size and
6234 irregular register set, the third one results in faster and generates
6235 decent code and the smallest size code, and the default value usually
6236 give the best results in most cases and for most architectures.
6238 @item -fira-coalesce
6239 @opindex fira-coalesce
6240 Do optimistic register coalescing. This option might be profitable for
6241 architectures with big regular register files.
6243 @item -fira-loop-pressure
6244 @opindex fira-loop-pressure
6245 Use IRA to evaluate register pressure in loops for decision to move
6246 loop invariants. Usage of this option usually results in generation
6247 of faster and smaller code on machines with big register files (>= 32
6248 registers) but it can slow compiler down.
6250 This option is enabled at level @option{-O3} for some targets.
6252 @item -fno-ira-share-save-slots
6253 @opindex fno-ira-share-save-slots
6254 Switch off sharing stack slots used for saving call used hard
6255 registers living through a call. Each hard register will get a
6256 separate stack slot and as a result function stack frame will be
6259 @item -fno-ira-share-spill-slots
6260 @opindex fno-ira-share-spill-slots
6261 Switch off sharing stack slots allocated for pseudo-registers. Each
6262 pseudo-register which did not get a hard register will get a separate
6263 stack slot and as a result function stack frame will be bigger.
6265 @item -fira-verbose=@var{n}
6266 @opindex fira-verbose
6267 Set up how verbose dump file for the integrated register allocator
6268 will be. Default value is 5. If the value is greater or equal to 10,
6269 the dump file will be stderr as if the value were @var{n} minus 10.
6271 @item -fdelayed-branch
6272 @opindex fdelayed-branch
6273 If supported for the target machine, attempt to reorder instructions
6274 to exploit instruction slots available after delayed branch
6277 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6279 @item -fschedule-insns
6280 @opindex fschedule-insns
6281 If supported for the target machine, attempt to reorder instructions to
6282 eliminate execution stalls due to required data being unavailable. This
6283 helps machines that have slow floating point or memory load instructions
6284 by allowing other instructions to be issued until the result of the load
6285 or floating point instruction is required.
6287 Enabled at levels @option{-O2}, @option{-O3}.
6289 @item -fschedule-insns2
6290 @opindex fschedule-insns2
6291 Similar to @option{-fschedule-insns}, but requests an additional pass of
6292 instruction scheduling after register allocation has been done. This is
6293 especially useful on machines with a relatively small number of
6294 registers and where memory load instructions take more than one cycle.
6296 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6298 @item -fno-sched-interblock
6299 @opindex fno-sched-interblock
6300 Don't schedule instructions across basic blocks. This is normally
6301 enabled by default when scheduling before register allocation, i.e.@:
6302 with @option{-fschedule-insns} or at @option{-O2} or higher.
6304 @item -fno-sched-spec
6305 @opindex fno-sched-spec
6306 Don't allow speculative motion of non-load instructions. This is normally
6307 enabled by default when scheduling before register allocation, i.e.@:
6308 with @option{-fschedule-insns} or at @option{-O2} or higher.
6310 @item -fsched-pressure
6311 @opindex fsched-pressure
6312 Enable register pressure sensitive insn scheduling before the register
6313 allocation. This only makes sense when scheduling before register
6314 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
6315 @option{-O2} or higher. Usage of this option can improve the
6316 generated code and decrease its size by preventing register pressure
6317 increase above the number of available hard registers and as a
6318 consequence register spills in the register allocation.
6320 @item -fsched-spec-load
6321 @opindex fsched-spec-load
6322 Allow speculative motion of some load instructions. This only makes
6323 sense when scheduling before register allocation, i.e.@: with
6324 @option{-fschedule-insns} or at @option{-O2} or higher.
6326 @item -fsched-spec-load-dangerous
6327 @opindex fsched-spec-load-dangerous
6328 Allow speculative motion of more load instructions. This only makes
6329 sense when scheduling before register allocation, i.e.@: with
6330 @option{-fschedule-insns} or at @option{-O2} or higher.
6332 @item -fsched-stalled-insns
6333 @itemx -fsched-stalled-insns=@var{n}
6334 @opindex fsched-stalled-insns
6335 Define how many insns (if any) can be moved prematurely from the queue
6336 of stalled insns into the ready list, during the second scheduling pass.
6337 @option{-fno-sched-stalled-insns} means that no insns will be moved
6338 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6339 on how many queued insns can be moved prematurely.
6340 @option{-fsched-stalled-insns} without a value is equivalent to
6341 @option{-fsched-stalled-insns=1}.
6343 @item -fsched-stalled-insns-dep
6344 @itemx -fsched-stalled-insns-dep=@var{n}
6345 @opindex fsched-stalled-insns-dep
6346 Define how many insn groups (cycles) will be examined for a dependency
6347 on a stalled insn that is candidate for premature removal from the queue
6348 of stalled insns. This has an effect only during the second scheduling pass,
6349 and only if @option{-fsched-stalled-insns} is used.
6350 @option{-fno-sched-stalled-insns-dep} is equivalent to
6351 @option{-fsched-stalled-insns-dep=0}.
6352 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6353 @option{-fsched-stalled-insns-dep=1}.
6355 @item -fsched2-use-superblocks
6356 @opindex fsched2-use-superblocks
6357 When scheduling after register allocation, do use superblock scheduling
6358 algorithm. Superblock scheduling allows motion across basic block boundaries
6359 resulting on faster schedules. This option is experimental, as not all machine
6360 descriptions used by GCC model the CPU closely enough to avoid unreliable
6361 results from the algorithm.
6363 This only makes sense when scheduling after register allocation, i.e.@: with
6364 @option{-fschedule-insns2} or at @option{-O2} or higher.
6366 @item -fsched-group-heuristic
6367 @opindex fsched-group-heuristic
6368 Enable the group heuristic in the scheduler. This heuristic favors
6369 the instruction that belongs to a schedule group. This is enabled
6370 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6371 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6373 @item -fsched-critical-path-heuristic
6374 @opindex fsched-critical-path-heuristic
6375 Enable the critical-path heuristic in the scheduler. This heuristic favors
6376 instructions on the critical path. This is enabled by default when
6377 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6378 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6380 @item -fsched-spec-insn-heuristic
6381 @opindex fsched-spec-insn-heuristic
6382 Enable the speculative instruction heuristic in the scheduler. This
6383 heuristic favors speculative instructions with greater dependency weakness.
6384 This is enabled by default when scheduling is enabled, i.e.@:
6385 with @option{-fschedule-insns} or @option{-fschedule-insns2}
6386 or at @option{-O2} or higher.
6388 @item -fsched-rank-heuristic
6389 @opindex fsched-rank-heuristic
6390 Enable the rank heuristic in the scheduler. This heuristic favors
6391 the instruction belonging to a basic block with greater size or frequency.
6392 This is enabled by default when scheduling is enabled, i.e.@:
6393 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6394 at @option{-O2} or higher.
6396 @item -fsched-last-insn-heuristic
6397 @opindex fsched-last-insn-heuristic
6398 Enable the last-instruction heuristic in the scheduler. This heuristic
6399 favors the instruction that is less dependent on the last instruction
6400 scheduled. This is enabled by default when scheduling is enabled,
6401 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6402 at @option{-O2} or higher.
6404 @item -fsched-dep-count-heuristic
6405 @opindex fsched-dep-count-heuristic
6406 Enable the dependent-count heuristic in the scheduler. This heuristic
6407 favors the instruction that has more instructions depending on it.
6408 This is enabled by default when scheduling is enabled, i.e.@:
6409 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6410 at @option{-O2} or higher.
6412 @item -freschedule-modulo-scheduled-loops
6413 @opindex freschedule-modulo-scheduled-loops
6414 The modulo scheduling comes before the traditional scheduling, if a loop
6415 was modulo scheduled we may want to prevent the later scheduling passes
6416 from changing its schedule, we use this option to control that.
6418 @item -fselective-scheduling
6419 @opindex fselective-scheduling
6420 Schedule instructions using selective scheduling algorithm. Selective
6421 scheduling runs instead of the first scheduler pass.
6423 @item -fselective-scheduling2
6424 @opindex fselective-scheduling2
6425 Schedule instructions using selective scheduling algorithm. Selective
6426 scheduling runs instead of the second scheduler pass.
6428 @item -fsel-sched-pipelining
6429 @opindex fsel-sched-pipelining
6430 Enable software pipelining of innermost loops during selective scheduling.
6431 This option has no effect until one of @option{-fselective-scheduling} or
6432 @option{-fselective-scheduling2} is turned on.
6434 @item -fsel-sched-pipelining-outer-loops
6435 @opindex fsel-sched-pipelining-outer-loops
6436 When pipelining loops during selective scheduling, also pipeline outer loops.
6437 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6439 @item -fcaller-saves
6440 @opindex fcaller-saves
6441 Enable values to be allocated in registers that will be clobbered by
6442 function calls, by emitting extra instructions to save and restore the
6443 registers around such calls. Such allocation is done only when it
6444 seems to result in better code than would otherwise be produced.
6446 This option is always enabled by default on certain machines, usually
6447 those which have no call-preserved registers to use instead.
6449 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6451 @item -fconserve-stack
6452 @opindex fconserve-stack
6453 Attempt to minimize stack usage. The compiler will attempt to use less
6454 stack space, even if that makes the program slower. This option
6455 implies setting the @option{large-stack-frame} parameter to 100
6456 and the @option{large-stack-frame-growth} parameter to 400.
6458 @item -ftree-reassoc
6459 @opindex ftree-reassoc
6460 Perform reassociation on trees. This flag is enabled by default
6461 at @option{-O} and higher.
6465 Perform partial redundancy elimination (PRE) on trees. This flag is
6466 enabled by default at @option{-O2} and @option{-O3}.
6468 @item -ftree-forwprop
6469 @opindex ftree-forwprop
6470 Perform forward propagation on trees. This flag is enabled by default
6471 at @option{-O} and higher.
6475 Perform full redundancy elimination (FRE) on trees. The difference
6476 between FRE and PRE is that FRE only considers expressions
6477 that are computed on all paths leading to the redundant computation.
6478 This analysis is faster than PRE, though it exposes fewer redundancies.
6479 This flag is enabled by default at @option{-O} and higher.
6481 @item -ftree-phiprop
6482 @opindex ftree-phiprop
6483 Perform hoisting of loads from conditional pointers on trees. This
6484 pass is enabled by default at @option{-O} and higher.
6486 @item -ftree-copy-prop
6487 @opindex ftree-copy-prop
6488 Perform copy propagation on trees. This pass eliminates unnecessary
6489 copy operations. This flag is enabled by default at @option{-O} and
6492 @item -fipa-pure-const
6493 @opindex fipa-pure-const
6494 Discover which functions are pure or constant.
6495 Enabled by default at @option{-O} and higher.
6497 @item -fipa-reference
6498 @opindex fipa-reference
6499 Discover which static variables do not escape cannot escape the
6501 Enabled by default at @option{-O} and higher.
6503 @item -fipa-struct-reorg
6504 @opindex fipa-struct-reorg
6505 Perform structure reorganization optimization, that change C-like structures
6506 layout in order to better utilize spatial locality. This transformation is
6507 affective for programs containing arrays of structures. Available in two
6508 compilation modes: profile-based (enabled with @option{-fprofile-generate})
6509 or static (which uses built-in heuristics). Require @option{-fipa-type-escape}
6510 to provide the safety of this transformation. It works only in whole program
6511 mode, so it requires @option{-fwhole-program} and @option{-combine} to be
6512 enabled. Structures considered @samp{cold} by this transformation are not
6513 affected (see @option{--param struct-reorg-cold-struct-ratio=@var{value}}).
6515 With this flag, the program debug info reflects a new structure layout.
6519 Perform interprocedural pointer analysis. This option is experimental
6520 and does not affect generated code.
6524 Perform interprocedural constant propagation.
6525 This optimization analyzes the program to determine when values passed
6526 to functions are constants and then optimizes accordingly.
6527 This optimization can substantially increase performance
6528 if the application has constants passed to functions.
6529 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
6531 @item -fipa-cp-clone
6532 @opindex fipa-cp-clone
6533 Perform function cloning to make interprocedural constant propagation stronger.
6534 When enabled, interprocedural constant propagation will perform function cloning
6535 when externally visible function can be called with constant arguments.
6536 Because this optimization can create multiple copies of functions,
6537 it may significantly increase code size
6538 (see @option{--param ipcp-unit-growth=@var{value}}).
6539 This flag is enabled by default at @option{-O3}.
6541 @item -fipa-matrix-reorg
6542 @opindex fipa-matrix-reorg
6543 Perform matrix flattening and transposing.
6544 Matrix flattening tries to replace an @math{m}-dimensional matrix
6545 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
6546 This reduces the level of indirection needed for accessing the elements
6547 of the matrix. The second optimization is matrix transposing that
6548 attempts to change the order of the matrix's dimensions in order to
6549 improve cache locality.
6550 Both optimizations need the @option{-fwhole-program} flag.
6551 Transposing is enabled only if profiling information is available.
6555 Perform forward store motion on trees. This flag is
6556 enabled by default at @option{-O} and higher.
6560 Perform sparse conditional constant propagation (CCP) on trees. This
6561 pass only operates on local scalar variables and is enabled by default
6562 at @option{-O} and higher.
6564 @item -ftree-switch-conversion
6565 Perform conversion of simple initializations in a switch to
6566 initializations from a scalar array. This flag is enabled by default
6567 at @option{-O2} and higher.
6571 Perform dead code elimination (DCE) on trees. This flag is enabled by
6572 default at @option{-O} and higher.
6574 @item -ftree-builtin-call-dce
6575 @opindex ftree-builtin-call-dce
6576 Perform conditional dead code elimination (DCE) for calls to builtin functions
6577 that may set @code{errno} but are otherwise side-effect free. This flag is
6578 enabled by default at @option{-O2} and higher if @option{-Os} is not also
6581 @item -ftree-dominator-opts
6582 @opindex ftree-dominator-opts
6583 Perform a variety of simple scalar cleanups (constant/copy
6584 propagation, redundancy elimination, range propagation and expression
6585 simplification) based on a dominator tree traversal. This also
6586 performs jump threading (to reduce jumps to jumps). This flag is
6587 enabled by default at @option{-O} and higher.
6591 Perform dead store elimination (DSE) on trees. A dead store is a store into
6592 a memory location which will later be overwritten by another store without
6593 any intervening loads. In this case the earlier store can be deleted. This
6594 flag is enabled by default at @option{-O} and higher.
6598 Perform loop header copying on trees. This is beneficial since it increases
6599 effectiveness of code motion optimizations. It also saves one jump. This flag
6600 is enabled by default at @option{-O} and higher. It is not enabled
6601 for @option{-Os}, since it usually increases code size.
6603 @item -ftree-loop-optimize
6604 @opindex ftree-loop-optimize
6605 Perform loop optimizations on trees. This flag is enabled by default
6606 at @option{-O} and higher.
6608 @item -ftree-loop-linear
6609 @opindex ftree-loop-linear
6610 Perform linear loop transformations on tree. This flag can improve cache
6611 performance and allow further loop optimizations to take place.
6613 @item -floop-interchange
6614 Perform loop interchange transformations on loops. Interchanging two
6615 nested loops switches the inner and outer loops. For example, given a
6620 A(J, I) = A(J, I) * C
6624 loop interchange will transform the loop as if the user had written:
6628 A(J, I) = A(J, I) * C
6632 which can be beneficial when @code{N} is larger than the caches,
6633 because in Fortran, the elements of an array are stored in memory
6634 contiguously by column, and the original loop iterates over rows,
6635 potentially creating at each access a cache miss. This optimization
6636 applies to all the languages supported by GCC and is not limited to
6637 Fortran. To use this code transformation, GCC has to be configured
6638 with @option{--with-ppl} and @option{--with-cloog} to enable the
6639 Graphite loop transformation infrastructure.
6641 @item -floop-strip-mine
6642 Perform loop strip mining transformations on loops. Strip mining
6643 splits a loop into two nested loops. The outer loop has strides
6644 equal to the strip size and the inner loop has strides of the
6645 original loop within a strip. For example, given a loop like:
6651 loop strip mining will transform the loop as if the user had written:
6654 DO I = II, min (II + 3, N)
6659 This optimization applies to all the languages supported by GCC and is
6660 not limited to Fortran. To use this code transformation, GCC has to
6661 be configured with @option{--with-ppl} and @option{--with-cloog} to
6662 enable the Graphite loop transformation infrastructure.
6665 Perform loop blocking transformations on loops. Blocking strip mines
6666 each loop in the loop nest such that the memory accesses of the
6667 element loops fit inside caches. For example, given a loop like:
6671 A(J, I) = B(I) + C(J)
6675 loop blocking will transform the loop as if the user had written:
6679 DO I = II, min (II + 63, N)
6680 DO J = JJ, min (JJ + 63, M)
6681 A(J, I) = B(I) + C(J)
6687 which can be beneficial when @code{M} is larger than the caches,
6688 because the innermost loop will iterate over a smaller amount of data
6689 that can be kept in the caches. This optimization applies to all the
6690 languages supported by GCC and is not limited to Fortran. To use this
6691 code transformation, GCC has to be configured with @option{--with-ppl}
6692 and @option{--with-cloog} to enable the Graphite loop transformation
6695 @item -fgraphite-identity
6696 @opindex fgraphite-identity
6697 Enable the identity transformation for graphite. For every SCoP we generate
6698 the polyhedral representation and transform it back to gimple. Using
6699 @option{-fgraphite-identity} we can check the costs or benefits of the
6700 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
6701 are also performed by the code generator CLooG, like index splitting and
6702 dead code elimination in loops.
6704 @item -floop-parallelize-all
6705 Use the Graphite data dependence analysis to identify loops that can
6706 be parallelized. Parallelize all the loops that can be analyzed to
6707 not contain loop carried dependences without checking that it is
6708 profitable to parallelize the loops.
6710 @item -fcheck-data-deps
6711 @opindex fcheck-data-deps
6712 Compare the results of several data dependence analyzers. This option
6713 is used for debugging the data dependence analyzers.
6715 @item -ftree-loop-distribution
6716 Perform loop distribution. This flag can improve cache performance on
6717 big loop bodies and allow further loop optimizations, like
6718 parallelization or vectorization, to take place. For example, the loop
6735 @item -ftree-loop-im
6736 @opindex ftree-loop-im
6737 Perform loop invariant motion on trees. This pass moves only invariants that
6738 would be hard to handle at RTL level (function calls, operations that expand to
6739 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
6740 operands of conditions that are invariant out of the loop, so that we can use
6741 just trivial invariantness analysis in loop unswitching. The pass also includes
6744 @item -ftree-loop-ivcanon
6745 @opindex ftree-loop-ivcanon
6746 Create a canonical counter for number of iterations in the loop for that
6747 determining number of iterations requires complicated analysis. Later
6748 optimizations then may determine the number easily. Useful especially
6749 in connection with unrolling.
6753 Perform induction variable optimizations (strength reduction, induction
6754 variable merging and induction variable elimination) on trees.
6756 @item -ftree-parallelize-loops=n
6757 @opindex ftree-parallelize-loops
6758 Parallelize loops, i.e., split their iteration space to run in n threads.
6759 This is only possible for loops whose iterations are independent
6760 and can be arbitrarily reordered. The optimization is only
6761 profitable on multiprocessor machines, for loops that are CPU-intensive,
6762 rather than constrained e.g.@: by memory bandwidth. This option
6763 implies @option{-pthread}, and thus is only supported on targets
6764 that have support for @option{-pthread}.
6768 Perform function-local points-to analysis on trees. This flag is
6769 enabled by default at @option{-O} and higher.
6773 Perform scalar replacement of aggregates. This pass replaces structure
6774 references with scalars to prevent committing structures to memory too
6775 early. This flag is enabled by default at @option{-O} and higher.
6777 @item -ftree-copyrename
6778 @opindex ftree-copyrename
6779 Perform copy renaming on trees. This pass attempts to rename compiler
6780 temporaries to other variables at copy locations, usually resulting in
6781 variable names which more closely resemble the original variables. This flag
6782 is enabled by default at @option{-O} and higher.
6786 Perform temporary expression replacement during the SSA->normal phase. Single
6787 use/single def temporaries are replaced at their use location with their
6788 defining expression. This results in non-GIMPLE code, but gives the expanders
6789 much more complex trees to work on resulting in better RTL generation. This is
6790 enabled by default at @option{-O} and higher.
6792 @item -ftree-vectorize
6793 @opindex ftree-vectorize
6794 Perform loop vectorization on trees. This flag is enabled by default at
6797 @item -ftree-slp-vectorize
6798 @opindex ftree-slp-vectorize
6799 Perform basic block vectorization on trees. This flag is enabled by default at
6800 @option{-O3} and when @option{-ftree-vectorize} is enabled.
6802 @item -ftree-vect-loop-version
6803 @opindex ftree-vect-loop-version
6804 Perform loop versioning when doing loop vectorization on trees. When a loop
6805 appears to be vectorizable except that data alignment or data dependence cannot
6806 be determined at compile time then vectorized and non-vectorized versions of
6807 the loop are generated along with runtime checks for alignment or dependence
6808 to control which version is executed. This option is enabled by default
6809 except at level @option{-Os} where it is disabled.
6811 @item -fvect-cost-model
6812 @opindex fvect-cost-model
6813 Enable cost model for vectorization.
6817 Perform Value Range Propagation on trees. This is similar to the
6818 constant propagation pass, but instead of values, ranges of values are
6819 propagated. This allows the optimizers to remove unnecessary range
6820 checks like array bound checks and null pointer checks. This is
6821 enabled by default at @option{-O2} and higher. Null pointer check
6822 elimination is only done if @option{-fdelete-null-pointer-checks} is
6827 Perform tail duplication to enlarge superblock size. This transformation
6828 simplifies the control flow of the function allowing other optimizations to do
6831 @item -funroll-loops
6832 @opindex funroll-loops
6833 Unroll loops whose number of iterations can be determined at compile
6834 time or upon entry to the loop. @option{-funroll-loops} implies
6835 @option{-frerun-cse-after-loop}. This option makes code larger,
6836 and may or may not make it run faster.
6838 @item -funroll-all-loops
6839 @opindex funroll-all-loops
6840 Unroll all loops, even if their number of iterations is uncertain when
6841 the loop is entered. This usually makes programs run more slowly.
6842 @option{-funroll-all-loops} implies the same options as
6843 @option{-funroll-loops},
6845 @item -fsplit-ivs-in-unroller
6846 @opindex fsplit-ivs-in-unroller
6847 Enables expressing of values of induction variables in later iterations
6848 of the unrolled loop using the value in the first iteration. This breaks
6849 long dependency chains, thus improving efficiency of the scheduling passes.
6851 Combination of @option{-fweb} and CSE is often sufficient to obtain the
6852 same effect. However in cases the loop body is more complicated than
6853 a single basic block, this is not reliable. It also does not work at all
6854 on some of the architectures due to restrictions in the CSE pass.
6856 This optimization is enabled by default.
6858 @item -fvariable-expansion-in-unroller
6859 @opindex fvariable-expansion-in-unroller
6860 With this option, the compiler will create multiple copies of some
6861 local variables when unrolling a loop which can result in superior code.
6863 @item -fpredictive-commoning
6864 @opindex fpredictive-commoning
6865 Perform predictive commoning optimization, i.e., reusing computations
6866 (especially memory loads and stores) performed in previous
6867 iterations of loops.
6869 This option is enabled at level @option{-O3}.
6871 @item -fprefetch-loop-arrays
6872 @opindex fprefetch-loop-arrays
6873 If supported by the target machine, generate instructions to prefetch
6874 memory to improve the performance of loops that access large arrays.
6876 This option may generate better or worse code; results are highly
6877 dependent on the structure of loops within the source code.
6879 Disabled at level @option{-Os}.
6882 @itemx -fno-peephole2
6883 @opindex fno-peephole
6884 @opindex fno-peephole2
6885 Disable any machine-specific peephole optimizations. The difference
6886 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
6887 are implemented in the compiler; some targets use one, some use the
6888 other, a few use both.
6890 @option{-fpeephole} is enabled by default.
6891 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6893 @item -fno-guess-branch-probability
6894 @opindex fno-guess-branch-probability
6895 Do not guess branch probabilities using heuristics.
6897 GCC will use heuristics to guess branch probabilities if they are
6898 not provided by profiling feedback (@option{-fprofile-arcs}). These
6899 heuristics are based on the control flow graph. If some branch probabilities
6900 are specified by @samp{__builtin_expect}, then the heuristics will be
6901 used to guess branch probabilities for the rest of the control flow graph,
6902 taking the @samp{__builtin_expect} info into account. The interactions
6903 between the heuristics and @samp{__builtin_expect} can be complex, and in
6904 some cases, it may be useful to disable the heuristics so that the effects
6905 of @samp{__builtin_expect} are easier to understand.
6907 The default is @option{-fguess-branch-probability} at levels
6908 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6910 @item -freorder-blocks
6911 @opindex freorder-blocks
6912 Reorder basic blocks in the compiled function in order to reduce number of
6913 taken branches and improve code locality.
6915 Enabled at levels @option{-O2}, @option{-O3}.
6917 @item -freorder-blocks-and-partition
6918 @opindex freorder-blocks-and-partition
6919 In addition to reordering basic blocks in the compiled function, in order
6920 to reduce number of taken branches, partitions hot and cold basic blocks
6921 into separate sections of the assembly and .o files, to improve
6922 paging and cache locality performance.
6924 This optimization is automatically turned off in the presence of
6925 exception handling, for linkonce sections, for functions with a user-defined
6926 section attribute and on any architecture that does not support named
6929 @item -freorder-functions
6930 @opindex freorder-functions
6931 Reorder functions in the object file in order to
6932 improve code locality. This is implemented by using special
6933 subsections @code{.text.hot} for most frequently executed functions and
6934 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
6935 the linker so object file format must support named sections and linker must
6936 place them in a reasonable way.
6938 Also profile feedback must be available in to make this option effective. See
6939 @option{-fprofile-arcs} for details.
6941 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6943 @item -fstrict-aliasing
6944 @opindex fstrict-aliasing
6945 Allow the compiler to assume the strictest aliasing rules applicable to
6946 the language being compiled. For C (and C++), this activates
6947 optimizations based on the type of expressions. In particular, an
6948 object of one type is assumed never to reside at the same address as an
6949 object of a different type, unless the types are almost the same. For
6950 example, an @code{unsigned int} can alias an @code{int}, but not a
6951 @code{void*} or a @code{double}. A character type may alias any other
6954 @anchor{Type-punning}Pay special attention to code like this:
6967 The practice of reading from a different union member than the one most
6968 recently written to (called ``type-punning'') is common. Even with
6969 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
6970 is accessed through the union type. So, the code above will work as
6971 expected. @xref{Structures unions enumerations and bit-fields
6972 implementation}. However, this code might not:
6983 Similarly, access by taking the address, casting the resulting pointer
6984 and dereferencing the result has undefined behavior, even if the cast
6985 uses a union type, e.g.:
6989 return ((union a_union *) &d)->i;
6993 The @option{-fstrict-aliasing} option is enabled at levels
6994 @option{-O2}, @option{-O3}, @option{-Os}.
6996 @item -fstrict-overflow
6997 @opindex fstrict-overflow
6998 Allow the compiler to assume strict signed overflow rules, depending
6999 on the language being compiled. For C (and C++) this means that
7000 overflow when doing arithmetic with signed numbers is undefined, which
7001 means that the compiler may assume that it will not happen. This
7002 permits various optimizations. For example, the compiler will assume
7003 that an expression like @code{i + 10 > i} will always be true for
7004 signed @code{i}. This assumption is only valid if signed overflow is
7005 undefined, as the expression is false if @code{i + 10} overflows when
7006 using twos complement arithmetic. When this option is in effect any
7007 attempt to determine whether an operation on signed numbers will
7008 overflow must be written carefully to not actually involve overflow.
7010 This option also allows the compiler to assume strict pointer
7011 semantics: given a pointer to an object, if adding an offset to that
7012 pointer does not produce a pointer to the same object, the addition is
7013 undefined. This permits the compiler to conclude that @code{p + u >
7014 p} is always true for a pointer @code{p} and unsigned integer
7015 @code{u}. This assumption is only valid because pointer wraparound is
7016 undefined, as the expression is false if @code{p + u} overflows using
7017 twos complement arithmetic.
7019 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
7020 that integer signed overflow is fully defined: it wraps. When
7021 @option{-fwrapv} is used, there is no difference between
7022 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
7023 integers. With @option{-fwrapv} certain types of overflow are
7024 permitted. For example, if the compiler gets an overflow when doing
7025 arithmetic on constants, the overflowed value can still be used with
7026 @option{-fwrapv}, but not otherwise.
7028 The @option{-fstrict-overflow} option is enabled at levels
7029 @option{-O2}, @option{-O3}, @option{-Os}.
7031 @item -falign-functions
7032 @itemx -falign-functions=@var{n}
7033 @opindex falign-functions
7034 Align the start of functions to the next power-of-two greater than
7035 @var{n}, skipping up to @var{n} bytes. For instance,
7036 @option{-falign-functions=32} aligns functions to the next 32-byte
7037 boundary, but @option{-falign-functions=24} would align to the next
7038 32-byte boundary only if this can be done by skipping 23 bytes or less.
7040 @option{-fno-align-functions} and @option{-falign-functions=1} are
7041 equivalent and mean that functions will not be aligned.
7043 Some assemblers only support this flag when @var{n} is a power of two;
7044 in that case, it is rounded up.
7046 If @var{n} is not specified or is zero, use a machine-dependent default.
7048 Enabled at levels @option{-O2}, @option{-O3}.
7050 @item -falign-labels
7051 @itemx -falign-labels=@var{n}
7052 @opindex falign-labels
7053 Align all branch targets to a power-of-two boundary, skipping up to
7054 @var{n} bytes like @option{-falign-functions}. This option can easily
7055 make code slower, because it must insert dummy operations for when the
7056 branch target is reached in the usual flow of the code.
7058 @option{-fno-align-labels} and @option{-falign-labels=1} are
7059 equivalent and mean that labels will not be aligned.
7061 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
7062 are greater than this value, then their values are used instead.
7064 If @var{n} is not specified or is zero, use a machine-dependent default
7065 which is very likely to be @samp{1}, meaning no alignment.
7067 Enabled at levels @option{-O2}, @option{-O3}.
7070 @itemx -falign-loops=@var{n}
7071 @opindex falign-loops
7072 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7073 like @option{-falign-functions}. The hope is that the loop will be
7074 executed many times, which will make up for any execution of the dummy
7077 @option{-fno-align-loops} and @option{-falign-loops=1} are
7078 equivalent and mean that loops will not be aligned.
7080 If @var{n} is not specified or is zero, use a machine-dependent default.
7082 Enabled at levels @option{-O2}, @option{-O3}.
7085 @itemx -falign-jumps=@var{n}
7086 @opindex falign-jumps
7087 Align branch targets to a power-of-two boundary, for branch targets
7088 where the targets can only be reached by jumping, skipping up to @var{n}
7089 bytes like @option{-falign-functions}. In this case, no dummy operations
7092 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7093 equivalent and mean that loops will not be aligned.
7095 If @var{n} is not specified or is zero, use a machine-dependent default.
7097 Enabled at levels @option{-O2}, @option{-O3}.
7099 @item -funit-at-a-time
7100 @opindex funit-at-a-time
7101 This option is left for compatibility reasons. @option{-funit-at-a-time}
7102 has no effect, while @option{-fno-unit-at-a-time} implies
7103 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7107 @item -fno-toplevel-reorder
7108 @opindex fno-toplevel-reorder
7109 Do not reorder top-level functions, variables, and @code{asm}
7110 statements. Output them in the same order that they appear in the
7111 input file. When this option is used, unreferenced static variables
7112 will not be removed. This option is intended to support existing code
7113 which relies on a particular ordering. For new code, it is better to
7116 Enabled at level @option{-O0}. When disabled explicitly, it also imply
7117 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
7122 Constructs webs as commonly used for register allocation purposes and assign
7123 each web individual pseudo register. This allows the register allocation pass
7124 to operate on pseudos directly, but also strengthens several other optimization
7125 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7126 however, make debugging impossible, since variables will no longer stay in a
7129 Enabled by default with @option{-funroll-loops}.
7131 @item -fwhole-program
7132 @opindex fwhole-program
7133 Assume that the current compilation unit represents the whole program being
7134 compiled. All public functions and variables with the exception of @code{main}
7135 and those merged by attribute @code{externally_visible} become static functions
7136 and in effect are optimized more aggressively by interprocedural optimizers.
7137 While this option is equivalent to proper use of the @code{static} keyword for
7138 programs consisting of a single file, in combination with option
7139 @option{-combine}, @option{-flto} or @option{-fwhopr} this flag can be used to
7140 compile many smaller scale programs since the functions and variables become
7141 local for the whole combined compilation unit, not for the single source file
7144 This option implies @option{-fwhole-file} for Fortran programs.
7148 This option runs the standard link-time optimizer. When invoked
7149 with source code, it generates GIMPLE (one of GCC's internal
7150 representations) and writes it to special ELF sections in the object
7151 file. When the object files are linked together, all the function
7152 bodies are read from these ELF sections and instantiated as if they
7153 had been part of the same translation unit.
7155 To use the link-timer optimizer, @option{-flto} needs to be specified at
7156 compile time and during the final link. For example,
7159 gcc -c -O2 -flto foo.c
7160 gcc -c -O2 -flto bar.c
7161 gcc -o myprog -flto -O2 foo.o bar.o
7164 The first two invocations to GCC will save a bytecode representation
7165 of GIMPLE into special ELF sections inside @file{foo.o} and
7166 @file{bar.o}. The final invocation will read the GIMPLE bytecode from
7167 @file{foo.o} and @file{bar.o}, merge the two files into a single
7168 internal image, and compile the result as usual. Since both
7169 @file{foo.o} and @file{bar.o} are merged into a single image, this
7170 causes all the inter-procedural analyses and optimizations in GCC to
7171 work across the two files as if they were a single one. This means,
7172 for example, that the inliner will be able to inline functions in
7173 @file{bar.o} into functions in @file{foo.o} and vice-versa.
7175 Another (simpler) way to enable link-time optimization is,
7178 gcc -o myprog -flto -O2 foo.c bar.c
7181 The above will generate bytecode for @file{foo.c} and @file{bar.c},
7182 merge them together into a single GIMPLE representation and optimize
7183 them as usual to produce @file{myprog}.
7185 The only important thing to keep in mind is that to enable link-time
7186 optimizations the @option{-flto} flag needs to be passed to both the
7187 compile and the link commands.
7189 Note that when a file is compiled with @option{-flto}, the generated
7190 object file will be larger than a regular object file because it will
7191 contain GIMPLE bytecodes and the usual final code. This means that
7192 object files with LTO information can be linked as a normal object
7193 file. So, in the previous example, if the final link is done with
7196 gcc -o myprog foo.o bar.o
7199 The only difference will be that no inter-procedural optimizations
7200 will be applied to produce @file{myprog}. The two object files
7201 @file{foo.o} and @file{bar.o} will be simply sent to the regular
7204 Additionally, the optimization flags used to compile individual files
7205 are not necessarily related to those used at link-time. For instance,
7208 gcc -c -O0 -flto foo.c
7209 gcc -c -O0 -flto bar.c
7210 gcc -o myprog -flto -O3 foo.o bar.o
7213 This will produce individual object files with unoptimized assembler
7214 code, but the resulting binary @file{myprog} will be optimized at
7215 @option{-O3}. Now, if the final binary is generated without
7216 @option{-flto}, then @file{myprog} will not be optimized.
7218 When producing the final binary with @option{-flto}, GCC will only
7219 apply link-time optimizations to those files that contain bytecode.
7220 Therefore, you can mix and match object files and libraries with
7221 GIMPLE bytecodes and final object code. GCC will automatically select
7222 which files to optimize in LTO mode and which files to link without
7225 There are some code generation flags that GCC will preserve when
7226 generating bytecodes, as they need to be used during the final link
7227 stage. Currently, the following options are saved into the GIMPLE
7228 bytecode files: @option{-fPIC}, @option{-fcommon} and all the
7229 @option{-m} target flags.
7231 At link time, these options are read-in and reapplied. Note that the
7232 current implementation makes no attempt at recognizing conflicting
7233 values for these options. If two or more files have a conflicting
7234 value (e.g., one file is compiled with @option{-fPIC} and another
7235 isn't), the compiler will simply use the last value read from the
7236 bytecode files. It is recommended, then, that all the files
7237 participating in the same link be compiled with the same options.
7239 Another feature of LTO is that it is possible to apply interprocedural
7240 optimizations on files written in different languages. This requires
7241 some support in the language front end. Currently, the C, C++ and
7242 Fortran front ends are capable of emitting GIMPLE bytecodes, so
7243 something like this should work
7248 gfortran -c -flto baz.f90
7249 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
7252 Notice that the final link is done with @command{g++} to get the C++
7253 runtime libraries and @option{-lgfortran} is added to get the Fortran
7254 runtime libraries. In general, when mixing languages in LTO mode, you
7255 should use the same link command used when mixing languages in a
7256 regular (non-LTO) compilation. This means that if your build process
7257 was mixing languages before, all you need to add is @option{-flto} to
7258 all the compile and link commands.
7260 If object files containing GIMPLE bytecode are stored in a library
7261 archive, say @file{libfoo.a}, it is possible to extract and use them
7262 in an LTO link if you are using @command{gold} as the linker (which,
7263 in turn requires GCC to be configured with @option{--enable-gold}).
7264 To enable this feature, use the flag @option{-fuse-linker-plugin} at
7268 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
7271 With the linker plugin enabled, @command{gold} will extract the needed
7272 GIMPLE files from @file{libfoo.a} and pass them on to the running GCC
7273 to make them part of the aggregated GIMPLE image to be optimized.
7275 If you are not using @command{gold} and/or do not specify
7276 @option{-fuse-linker-plugin} then the objects inside @file{libfoo.a}
7277 will be extracted and linked as usual, but they will not participate
7278 in the LTO optimization process.
7280 Link time optimizations do not require the presence of the whole
7281 program to operate. If the program does not require any symbols to
7282 be exported, it is possible to combine @option{-flto} and
7283 @option{-fwhopr} with @option{-fwhole-program} to allow the
7284 interprocedural optimizers to use more aggressive assumptions which
7285 may lead to improved optimization opportunities.
7287 Regarding portability: the current implementation of LTO makes no
7288 attempt at generating bytecode that can be ported between different
7289 types of hosts. The bytecode files are versioned and there is a
7290 strict version check, so bytecode files generated in one version of
7291 GCC will not work with an older/newer version of GCC.
7293 Link time optimization does not play well with generating debugging
7294 information. Combining @option{-flto} or @option{-fwhopr} with
7295 @option{-g} is experimental.
7297 This option is disabled by default.
7301 This option is identical in functionality to @option{-flto} but it
7302 differs in how the final link stage is executed. Instead of loading
7303 all the function bodies in memory, the callgraph is analyzed and
7304 optimization decisions are made (whole program analysis or WPA). Once
7305 optimization decisions are made, the callgraph is partitioned and the
7306 different sections are compiled separately (local transformations or
7307 LTRANS)@. This process allows optimizations on very large programs
7308 that otherwise would not fit in memory. This option enables
7309 @option{-fwpa} and @option{-fltrans} automatically.
7311 Disabled by default.
7313 This option is experimental.
7317 This is an internal option used by GCC when compiling with
7318 @option{-fwhopr}. You should never need to use it.
7320 This option runs the link-time optimizer in the whole-program-analysis
7321 (WPA) mode, which reads in summary information from all inputs and
7322 performs a whole-program analysis based on summary information only.
7323 It generates object files for subsequent runs of the link-time
7324 optimizer where individual object files are optimized using both
7325 summary information from the WPA mode and the actual function bodies.
7326 It then drives the LTRANS phase.
7328 Disabled by default.
7332 This is an internal option used by GCC when compiling with
7333 @option{-fwhopr}. You should never need to use it.
7335 This option runs the link-time optimizer in the local-transformation (LTRANS)
7336 mode, which reads in output from a previous run of the LTO in WPA mode.
7337 In the LTRANS mode, LTO optimizes an object and produces the final assembly.
7339 Disabled by default.
7341 @item -fltrans-output-list=@var{file}
7342 @opindex fltrans-output-list
7343 This is an internal option used by GCC when compiling with
7344 @option{-fwhopr}. You should never need to use it.
7346 This option specifies a file to which the names of LTRANS output files are
7347 written. This option is only meaningful in conjunction with @option{-fwpa}.
7349 Disabled by default.
7351 @item -flto-compression-level=@var{n}
7352 This option specifies the level of compression used for intermediate
7353 language written to LTO object files, and is only meaningful in
7354 conjunction with LTO mode (@option{-fwhopr}, @option{-flto}). Valid
7355 values are 0 (no compression) to 9 (maximum compression). Values
7356 outside this range are clamped to either 0 or 9. If the option is not
7357 given, a default balanced compression setting is used.
7360 Prints a report with internal details on the workings of the link-time
7361 optimizer. The contents of this report vary from version to version,
7362 it is meant to be useful to GCC developers when processing object
7363 files in LTO mode (via @option{-fwhopr} or @option{-flto}).
7365 Disabled by default.
7367 @item -fuse-linker-plugin
7368 Enables the extraction of objects with GIMPLE bytecode information
7369 from library archives. This option relies on features available only
7370 in @command{gold}, so to use this you must configure GCC with
7371 @option{--enable-gold}. See @option{-flto} for a description on the
7372 effect of this flag and how to use it.
7374 Disabled by default.
7376 @item -fcprop-registers
7377 @opindex fcprop-registers
7378 After register allocation and post-register allocation instruction splitting,
7379 we perform a copy-propagation pass to try to reduce scheduling dependencies
7380 and occasionally eliminate the copy.
7382 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7384 @item -fprofile-correction
7385 @opindex fprofile-correction
7386 Profiles collected using an instrumented binary for multi-threaded programs may
7387 be inconsistent due to missed counter updates. When this option is specified,
7388 GCC will use heuristics to correct or smooth out such inconsistencies. By
7389 default, GCC will emit an error message when an inconsistent profile is detected.
7391 @item -fprofile-dir=@var{path}
7392 @opindex fprofile-dir
7394 Set the directory to search the profile data files in to @var{path}.
7395 This option affects only the profile data generated by
7396 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
7397 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
7398 and its related options.
7399 By default, GCC will use the current directory as @var{path}
7400 thus the profile data file will appear in the same directory as the object file.
7402 @item -fprofile-generate
7403 @itemx -fprofile-generate=@var{path}
7404 @opindex fprofile-generate
7406 Enable options usually used for instrumenting application to produce
7407 profile useful for later recompilation with profile feedback based
7408 optimization. You must use @option{-fprofile-generate} both when
7409 compiling and when linking your program.
7411 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
7413 If @var{path} is specified, GCC will look at the @var{path} to find
7414 the profile feedback data files. See @option{-fprofile-dir}.
7417 @itemx -fprofile-use=@var{path}
7418 @opindex fprofile-use
7419 Enable profile feedback directed optimizations, and optimizations
7420 generally profitable only with profile feedback available.
7422 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
7423 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
7425 By default, GCC emits an error message if the feedback profiles do not
7426 match the source code. This error can be turned into a warning by using
7427 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
7430 If @var{path} is specified, GCC will look at the @var{path} to find
7431 the profile feedback data files. See @option{-fprofile-dir}.
7434 The following options control compiler behavior regarding floating
7435 point arithmetic. These options trade off between speed and
7436 correctness. All must be specifically enabled.
7440 @opindex ffloat-store
7441 Do not store floating point variables in registers, and inhibit other
7442 options that might change whether a floating point value is taken from a
7445 @cindex floating point precision
7446 This option prevents undesirable excess precision on machines such as
7447 the 68000 where the floating registers (of the 68881) keep more
7448 precision than a @code{double} is supposed to have. Similarly for the
7449 x86 architecture. For most programs, the excess precision does only
7450 good, but a few programs rely on the precise definition of IEEE floating
7451 point. Use @option{-ffloat-store} for such programs, after modifying
7452 them to store all pertinent intermediate computations into variables.
7454 @item -fexcess-precision=@var{style}
7455 @opindex fexcess-precision
7456 This option allows further control over excess precision on machines
7457 where floating-point registers have more precision than the IEEE
7458 @code{float} and @code{double} types and the processor does not
7459 support operations rounding to those types. By default,
7460 @option{-fexcess-precision=fast} is in effect; this means that
7461 operations are carried out in the precision of the registers and that
7462 it is unpredictable when rounding to the types specified in the source
7463 code takes place. When compiling C, if
7464 @option{-fexcess-precision=standard} is specified then excess
7465 precision will follow the rules specified in ISO C99; in particular,
7466 both casts and assignments cause values to be rounded to their
7467 semantic types (whereas @option{-ffloat-store} only affects
7468 assignments). This option is enabled by default for C if a strict
7469 conformance option such as @option{-std=c99} is used.
7472 @option{-fexcess-precision=standard} is not implemented for languages
7473 other than C, and has no effect if
7474 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
7475 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
7476 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
7477 semantics apply without excess precision, and in the latter, rounding
7482 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
7483 @option{-ffinite-math-only}, @option{-fno-rounding-math},
7484 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
7486 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
7488 This option is not turned on by any @option{-O} option since
7489 it can result in incorrect output for programs which depend on
7490 an exact implementation of IEEE or ISO rules/specifications for
7491 math functions. It may, however, yield faster code for programs
7492 that do not require the guarantees of these specifications.
7494 @item -fno-math-errno
7495 @opindex fno-math-errno
7496 Do not set ERRNO after calling math functions that are executed
7497 with a single instruction, e.g., sqrt. A program that relies on
7498 IEEE exceptions for math error handling may want to use this flag
7499 for speed while maintaining IEEE arithmetic compatibility.
7501 This option is not turned on by any @option{-O} option since
7502 it can result in incorrect output for programs which depend on
7503 an exact implementation of IEEE or ISO rules/specifications for
7504 math functions. It may, however, yield faster code for programs
7505 that do not require the guarantees of these specifications.
7507 The default is @option{-fmath-errno}.
7509 On Darwin systems, the math library never sets @code{errno}. There is
7510 therefore no reason for the compiler to consider the possibility that
7511 it might, and @option{-fno-math-errno} is the default.
7513 @item -funsafe-math-optimizations
7514 @opindex funsafe-math-optimizations
7516 Allow optimizations for floating-point arithmetic that (a) assume
7517 that arguments and results are valid and (b) may violate IEEE or
7518 ANSI standards. When used at link-time, it may include libraries
7519 or startup files that change the default FPU control word or other
7520 similar optimizations.
7522 This option is not turned on by any @option{-O} option since
7523 it can result in incorrect output for programs which depend on
7524 an exact implementation of IEEE or ISO rules/specifications for
7525 math functions. It may, however, yield faster code for programs
7526 that do not require the guarantees of these specifications.
7527 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
7528 @option{-fassociative-math} and @option{-freciprocal-math}.
7530 The default is @option{-fno-unsafe-math-optimizations}.
7532 @item -fassociative-math
7533 @opindex fassociative-math
7535 Allow re-association of operands in series of floating-point operations.
7536 This violates the ISO C and C++ language standard by possibly changing
7537 computation result. NOTE: re-ordering may change the sign of zero as
7538 well as ignore NaNs and inhibit or create underflow or overflow (and
7539 thus cannot be used on a code which relies on rounding behavior like
7540 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
7541 and thus may not be used when ordered comparisons are required.
7542 This option requires that both @option{-fno-signed-zeros} and
7543 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
7544 much sense with @option{-frounding-math}. For Fortran the option
7545 is automatically enabled when both @option{-fno-signed-zeros} and
7546 @option{-fno-trapping-math} are in effect.
7548 The default is @option{-fno-associative-math}.
7550 @item -freciprocal-math
7551 @opindex freciprocal-math
7553 Allow the reciprocal of a value to be used instead of dividing by
7554 the value if this enables optimizations. For example @code{x / y}
7555 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
7556 is subject to common subexpression elimination. Note that this loses
7557 precision and increases the number of flops operating on the value.
7559 The default is @option{-fno-reciprocal-math}.
7561 @item -ffinite-math-only
7562 @opindex ffinite-math-only
7563 Allow optimizations for floating-point arithmetic that assume
7564 that arguments and results are not NaNs or +-Infs.
7566 This option is not turned on by any @option{-O} option since
7567 it can result in incorrect output for programs which depend on
7568 an exact implementation of IEEE or ISO rules/specifications for
7569 math functions. It may, however, yield faster code for programs
7570 that do not require the guarantees of these specifications.
7572 The default is @option{-fno-finite-math-only}.
7574 @item -fno-signed-zeros
7575 @opindex fno-signed-zeros
7576 Allow optimizations for floating point arithmetic that ignore the
7577 signedness of zero. IEEE arithmetic specifies the behavior of
7578 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
7579 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
7580 This option implies that the sign of a zero result isn't significant.
7582 The default is @option{-fsigned-zeros}.
7584 @item -fno-trapping-math
7585 @opindex fno-trapping-math
7586 Compile code assuming that floating-point operations cannot generate
7587 user-visible traps. These traps include division by zero, overflow,
7588 underflow, inexact result and invalid operation. This option requires
7589 that @option{-fno-signaling-nans} be in effect. Setting this option may
7590 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
7592 This option should never be turned on by any @option{-O} option since
7593 it can result in incorrect output for programs which depend on
7594 an exact implementation of IEEE or ISO rules/specifications for
7597 The default is @option{-ftrapping-math}.
7599 @item -frounding-math
7600 @opindex frounding-math
7601 Disable transformations and optimizations that assume default floating
7602 point rounding behavior. This is round-to-zero for all floating point
7603 to integer conversions, and round-to-nearest for all other arithmetic
7604 truncations. This option should be specified for programs that change
7605 the FP rounding mode dynamically, or that may be executed with a
7606 non-default rounding mode. This option disables constant folding of
7607 floating point expressions at compile-time (which may be affected by
7608 rounding mode) and arithmetic transformations that are unsafe in the
7609 presence of sign-dependent rounding modes.
7611 The default is @option{-fno-rounding-math}.
7613 This option is experimental and does not currently guarantee to
7614 disable all GCC optimizations that are affected by rounding mode.
7615 Future versions of GCC may provide finer control of this setting
7616 using C99's @code{FENV_ACCESS} pragma. This command line option
7617 will be used to specify the default state for @code{FENV_ACCESS}.
7619 @item -fsignaling-nans
7620 @opindex fsignaling-nans
7621 Compile code assuming that IEEE signaling NaNs may generate user-visible
7622 traps during floating-point operations. Setting this option disables
7623 optimizations that may change the number of exceptions visible with
7624 signaling NaNs. This option implies @option{-ftrapping-math}.
7626 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
7629 The default is @option{-fno-signaling-nans}.
7631 This option is experimental and does not currently guarantee to
7632 disable all GCC optimizations that affect signaling NaN behavior.
7634 @item -fsingle-precision-constant
7635 @opindex fsingle-precision-constant
7636 Treat floating point constant as single precision constant instead of
7637 implicitly converting it to double precision constant.
7639 @item -fcx-limited-range
7640 @opindex fcx-limited-range
7641 When enabled, this option states that a range reduction step is not
7642 needed when performing complex division. Also, there is no checking
7643 whether the result of a complex multiplication or division is @code{NaN
7644 + I*NaN}, with an attempt to rescue the situation in that case. The
7645 default is @option{-fno-cx-limited-range}, but is enabled by
7646 @option{-ffast-math}.
7648 This option controls the default setting of the ISO C99
7649 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
7652 @item -fcx-fortran-rules
7653 @opindex fcx-fortran-rules
7654 Complex multiplication and division follow Fortran rules. Range
7655 reduction is done as part of complex division, but there is no checking
7656 whether the result of a complex multiplication or division is @code{NaN
7657 + I*NaN}, with an attempt to rescue the situation in that case.
7659 The default is @option{-fno-cx-fortran-rules}.
7663 The following options control optimizations that may improve
7664 performance, but are not enabled by any @option{-O} options. This
7665 section includes experimental options that may produce broken code.
7668 @item -fbranch-probabilities
7669 @opindex fbranch-probabilities
7670 After running a program compiled with @option{-fprofile-arcs}
7671 (@pxref{Debugging Options,, Options for Debugging Your Program or
7672 @command{gcc}}), you can compile it a second time using
7673 @option{-fbranch-probabilities}, to improve optimizations based on
7674 the number of times each branch was taken. When the program
7675 compiled with @option{-fprofile-arcs} exits it saves arc execution
7676 counts to a file called @file{@var{sourcename}.gcda} for each source
7677 file. The information in this data file is very dependent on the
7678 structure of the generated code, so you must use the same source code
7679 and the same optimization options for both compilations.
7681 With @option{-fbranch-probabilities}, GCC puts a
7682 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
7683 These can be used to improve optimization. Currently, they are only
7684 used in one place: in @file{reorg.c}, instead of guessing which path a
7685 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
7686 exactly determine which path is taken more often.
7688 @item -fprofile-values
7689 @opindex fprofile-values
7690 If combined with @option{-fprofile-arcs}, it adds code so that some
7691 data about values of expressions in the program is gathered.
7693 With @option{-fbranch-probabilities}, it reads back the data gathered
7694 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
7695 notes to instructions for their later usage in optimizations.
7697 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
7701 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
7702 a code to gather information about values of expressions.
7704 With @option{-fbranch-probabilities}, it reads back the data gathered
7705 and actually performs the optimizations based on them.
7706 Currently the optimizations include specialization of division operation
7707 using the knowledge about the value of the denominator.
7709 @item -frename-registers
7710 @opindex frename-registers
7711 Attempt to avoid false dependencies in scheduled code by making use
7712 of registers left over after register allocation. This optimization
7713 will most benefit processors with lots of registers. Depending on the
7714 debug information format adopted by the target, however, it can
7715 make debugging impossible, since variables will no longer stay in
7716 a ``home register''.
7718 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
7722 Perform tail duplication to enlarge superblock size. This transformation
7723 simplifies the control flow of the function allowing other optimizations to do
7726 Enabled with @option{-fprofile-use}.
7728 @item -funroll-loops
7729 @opindex funroll-loops
7730 Unroll loops whose number of iterations can be determined at compile time or
7731 upon entry to the loop. @option{-funroll-loops} implies
7732 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
7733 It also turns on complete loop peeling (i.e.@: complete removal of loops with
7734 small constant number of iterations). This option makes code larger, and may
7735 or may not make it run faster.
7737 Enabled with @option{-fprofile-use}.
7739 @item -funroll-all-loops
7740 @opindex funroll-all-loops
7741 Unroll all loops, even if their number of iterations is uncertain when
7742 the loop is entered. This usually makes programs run more slowly.
7743 @option{-funroll-all-loops} implies the same options as
7744 @option{-funroll-loops}.
7747 @opindex fpeel-loops
7748 Peels the loops for that there is enough information that they do not
7749 roll much (from profile feedback). It also turns on complete loop peeling
7750 (i.e.@: complete removal of loops with small constant number of iterations).
7752 Enabled with @option{-fprofile-use}.
7754 @item -fmove-loop-invariants
7755 @opindex fmove-loop-invariants
7756 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
7757 at level @option{-O1}
7759 @item -funswitch-loops
7760 @opindex funswitch-loops
7761 Move branches with loop invariant conditions out of the loop, with duplicates
7762 of the loop on both branches (modified according to result of the condition).
7764 @item -ffunction-sections
7765 @itemx -fdata-sections
7766 @opindex ffunction-sections
7767 @opindex fdata-sections
7768 Place each function or data item into its own section in the output
7769 file if the target supports arbitrary sections. The name of the
7770 function or the name of the data item determines the section's name
7773 Use these options on systems where the linker can perform optimizations
7774 to improve locality of reference in the instruction space. Most systems
7775 using the ELF object format and SPARC processors running Solaris 2 have
7776 linkers with such optimizations. AIX may have these optimizations in
7779 Only use these options when there are significant benefits from doing
7780 so. When you specify these options, the assembler and linker will
7781 create larger object and executable files and will also be slower.
7782 You will not be able to use @code{gprof} on all systems if you
7783 specify this option and you may have problems with debugging if
7784 you specify both this option and @option{-g}.
7786 @item -fbranch-target-load-optimize
7787 @opindex fbranch-target-load-optimize
7788 Perform branch target register load optimization before prologue / epilogue
7790 The use of target registers can typically be exposed only during reload,
7791 thus hoisting loads out of loops and doing inter-block scheduling needs
7792 a separate optimization pass.
7794 @item -fbranch-target-load-optimize2
7795 @opindex fbranch-target-load-optimize2
7796 Perform branch target register load optimization after prologue / epilogue
7799 @item -fbtr-bb-exclusive
7800 @opindex fbtr-bb-exclusive
7801 When performing branch target register load optimization, don't reuse
7802 branch target registers in within any basic block.
7804 @item -fstack-protector
7805 @opindex fstack-protector
7806 Emit extra code to check for buffer overflows, such as stack smashing
7807 attacks. This is done by adding a guard variable to functions with
7808 vulnerable objects. This includes functions that call alloca, and
7809 functions with buffers larger than 8 bytes. The guards are initialized
7810 when a function is entered and then checked when the function exits.
7811 If a guard check fails, an error message is printed and the program exits.
7813 @item -fstack-protector-all
7814 @opindex fstack-protector-all
7815 Like @option{-fstack-protector} except that all functions are protected.
7817 @item -fsection-anchors
7818 @opindex fsection-anchors
7819 Try to reduce the number of symbolic address calculations by using
7820 shared ``anchor'' symbols to address nearby objects. This transformation
7821 can help to reduce the number of GOT entries and GOT accesses on some
7824 For example, the implementation of the following function @code{foo}:
7828 int foo (void) @{ return a + b + c; @}
7831 would usually calculate the addresses of all three variables, but if you
7832 compile it with @option{-fsection-anchors}, it will access the variables
7833 from a common anchor point instead. The effect is similar to the
7834 following pseudocode (which isn't valid C):
7839 register int *xr = &x;
7840 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
7844 Not all targets support this option.
7846 @item --param @var{name}=@var{value}
7848 In some places, GCC uses various constants to control the amount of
7849 optimization that is done. For example, GCC will not inline functions
7850 that contain more that a certain number of instructions. You can
7851 control some of these constants on the command-line using the
7852 @option{--param} option.
7854 The names of specific parameters, and the meaning of the values, are
7855 tied to the internals of the compiler, and are subject to change
7856 without notice in future releases.
7858 In each case, the @var{value} is an integer. The allowable choices for
7859 @var{name} are given in the following table:
7862 @item struct-reorg-cold-struct-ratio
7863 The threshold ratio (as a percentage) between a structure frequency
7864 and the frequency of the hottest structure in the program. This parameter
7865 is used by struct-reorg optimization enabled by @option{-fipa-struct-reorg}.
7866 We say that if the ratio of a structure frequency, calculated by profiling,
7867 to the hottest structure frequency in the program is less than this
7868 parameter, then structure reorganization is not applied to this structure.
7871 @item predictable-branch-outcome
7872 When branch is predicted to be taken with probability lower than this threshold
7873 (in percent), then it is considered well predictable. The default is 10.
7875 @item max-crossjump-edges
7876 The maximum number of incoming edges to consider for crossjumping.
7877 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
7878 the number of edges incoming to each block. Increasing values mean
7879 more aggressive optimization, making the compile time increase with
7880 probably small improvement in executable size.
7882 @item min-crossjump-insns
7883 The minimum number of instructions which must be matched at the end
7884 of two blocks before crossjumping will be performed on them. This
7885 value is ignored in the case where all instructions in the block being
7886 crossjumped from are matched. The default value is 5.
7888 @item max-grow-copy-bb-insns
7889 The maximum code size expansion factor when copying basic blocks
7890 instead of jumping. The expansion is relative to a jump instruction.
7891 The default value is 8.
7893 @item max-goto-duplication-insns
7894 The maximum number of instructions to duplicate to a block that jumps
7895 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
7896 passes, GCC factors computed gotos early in the compilation process,
7897 and unfactors them as late as possible. Only computed jumps at the
7898 end of a basic blocks with no more than max-goto-duplication-insns are
7899 unfactored. The default value is 8.
7901 @item max-delay-slot-insn-search
7902 The maximum number of instructions to consider when looking for an
7903 instruction to fill a delay slot. If more than this arbitrary number of
7904 instructions is searched, the time savings from filling the delay slot
7905 will be minimal so stop searching. Increasing values mean more
7906 aggressive optimization, making the compile time increase with probably
7907 small improvement in executable run time.
7909 @item max-delay-slot-live-search
7910 When trying to fill delay slots, the maximum number of instructions to
7911 consider when searching for a block with valid live register
7912 information. Increasing this arbitrarily chosen value means more
7913 aggressive optimization, increasing the compile time. This parameter
7914 should be removed when the delay slot code is rewritten to maintain the
7917 @item max-gcse-memory
7918 The approximate maximum amount of memory that will be allocated in
7919 order to perform the global common subexpression elimination
7920 optimization. If more memory than specified is required, the
7921 optimization will not be done.
7923 @item max-pending-list-length
7924 The maximum number of pending dependencies scheduling will allow
7925 before flushing the current state and starting over. Large functions
7926 with few branches or calls can create excessively large lists which
7927 needlessly consume memory and resources.
7929 @item max-inline-insns-single
7930 Several parameters control the tree inliner used in gcc.
7931 This number sets the maximum number of instructions (counted in GCC's
7932 internal representation) in a single function that the tree inliner
7933 will consider for inlining. This only affects functions declared
7934 inline and methods implemented in a class declaration (C++).
7935 The default value is 300.
7937 @item max-inline-insns-auto
7938 When you use @option{-finline-functions} (included in @option{-O3}),
7939 a lot of functions that would otherwise not be considered for inlining
7940 by the compiler will be investigated. To those functions, a different
7941 (more restrictive) limit compared to functions declared inline can
7943 The default value is 50.
7945 @item large-function-insns
7946 The limit specifying really large functions. For functions larger than this
7947 limit after inlining, inlining is constrained by
7948 @option{--param large-function-growth}. This parameter is useful primarily
7949 to avoid extreme compilation time caused by non-linear algorithms used by the
7951 The default value is 2700.
7953 @item large-function-growth
7954 Specifies maximal growth of large function caused by inlining in percents.
7955 The default value is 100 which limits large function growth to 2.0 times
7958 @item large-unit-insns
7959 The limit specifying large translation unit. Growth caused by inlining of
7960 units larger than this limit is limited by @option{--param inline-unit-growth}.
7961 For small units this might be too tight (consider unit consisting of function A
7962 that is inline and B that just calls A three time. If B is small relative to
7963 A, the growth of unit is 300\% and yet such inlining is very sane. For very
7964 large units consisting of small inlineable functions however the overall unit
7965 growth limit is needed to avoid exponential explosion of code size. Thus for
7966 smaller units, the size is increased to @option{--param large-unit-insns}
7967 before applying @option{--param inline-unit-growth}. The default is 10000
7969 @item inline-unit-growth
7970 Specifies maximal overall growth of the compilation unit caused by inlining.
7971 The default value is 30 which limits unit growth to 1.3 times the original
7974 @item ipcp-unit-growth
7975 Specifies maximal overall growth of the compilation unit caused by
7976 interprocedural constant propagation. The default value is 10 which limits
7977 unit growth to 1.1 times the original size.
7979 @item large-stack-frame
7980 The limit specifying large stack frames. While inlining the algorithm is trying
7981 to not grow past this limit too much. Default value is 256 bytes.
7983 @item large-stack-frame-growth
7984 Specifies maximal growth of large stack frames caused by inlining in percents.
7985 The default value is 1000 which limits large stack frame growth to 11 times
7988 @item max-inline-insns-recursive
7989 @itemx max-inline-insns-recursive-auto
7990 Specifies maximum number of instructions out-of-line copy of self recursive inline
7991 function can grow into by performing recursive inlining.
7993 For functions declared inline @option{--param max-inline-insns-recursive} is
7994 taken into account. For function not declared inline, recursive inlining
7995 happens only when @option{-finline-functions} (included in @option{-O3}) is
7996 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
7997 default value is 450.
7999 @item max-inline-recursive-depth
8000 @itemx max-inline-recursive-depth-auto
8001 Specifies maximum recursion depth used by the recursive inlining.
8003 For functions declared inline @option{--param max-inline-recursive-depth} is
8004 taken into account. For function not declared inline, recursive inlining
8005 happens only when @option{-finline-functions} (included in @option{-O3}) is
8006 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
8009 @item min-inline-recursive-probability
8010 Recursive inlining is profitable only for function having deep recursion
8011 in average and can hurt for function having little recursion depth by
8012 increasing the prologue size or complexity of function body to other
8015 When profile feedback is available (see @option{-fprofile-generate}) the actual
8016 recursion depth can be guessed from probability that function will recurse via
8017 given call expression. This parameter limits inlining only to call expression
8018 whose probability exceeds given threshold (in percents). The default value is
8021 @item early-inlining-insns
8022 Specify growth that early inliner can make. In effect it increases amount of
8023 inlining for code having large abstraction penalty. The default value is 8.
8025 @item max-early-inliner-iterations
8026 @itemx max-early-inliner-iterations
8027 Limit of iterations of early inliner. This basically bounds number of nested
8028 indirect calls early inliner can resolve. Deeper chains are still handled by
8031 @item min-vect-loop-bound
8032 The minimum number of iterations under which a loop will not get vectorized
8033 when @option{-ftree-vectorize} is used. The number of iterations after
8034 vectorization needs to be greater than the value specified by this option
8035 to allow vectorization. The default value is 0.
8037 @item max-unrolled-insns
8038 The maximum number of instructions that a loop should have if that loop
8039 is unrolled, and if the loop is unrolled, it determines how many times
8040 the loop code is unrolled.
8042 @item max-average-unrolled-insns
8043 The maximum number of instructions biased by probabilities of their execution
8044 that a loop should have if that loop is unrolled, and if the loop is unrolled,
8045 it determines how many times the loop code is unrolled.
8047 @item max-unroll-times
8048 The maximum number of unrollings of a single loop.
8050 @item max-peeled-insns
8051 The maximum number of instructions that a loop should have if that loop
8052 is peeled, and if the loop is peeled, it determines how many times
8053 the loop code is peeled.
8055 @item max-peel-times
8056 The maximum number of peelings of a single loop.
8058 @item max-completely-peeled-insns
8059 The maximum number of insns of a completely peeled loop.
8061 @item max-completely-peel-times
8062 The maximum number of iterations of a loop to be suitable for complete peeling.
8064 @item max-unswitch-insns
8065 The maximum number of insns of an unswitched loop.
8067 @item max-unswitch-level
8068 The maximum number of branches unswitched in a single loop.
8071 The minimum cost of an expensive expression in the loop invariant motion.
8073 @item iv-consider-all-candidates-bound
8074 Bound on number of candidates for induction variables below that
8075 all candidates are considered for each use in induction variable
8076 optimizations. Only the most relevant candidates are considered
8077 if there are more candidates, to avoid quadratic time complexity.
8079 @item iv-max-considered-uses
8080 The induction variable optimizations give up on loops that contain more
8081 induction variable uses.
8083 @item iv-always-prune-cand-set-bound
8084 If number of candidates in the set is smaller than this value,
8085 we always try to remove unnecessary ivs from the set during its
8086 optimization when a new iv is added to the set.
8088 @item scev-max-expr-size
8089 Bound on size of expressions used in the scalar evolutions analyzer.
8090 Large expressions slow the analyzer.
8092 @item omega-max-vars
8093 The maximum number of variables in an Omega constraint system.
8094 The default value is 128.
8096 @item omega-max-geqs
8097 The maximum number of inequalities in an Omega constraint system.
8098 The default value is 256.
8101 The maximum number of equalities in an Omega constraint system.
8102 The default value is 128.
8104 @item omega-max-wild-cards
8105 The maximum number of wildcard variables that the Omega solver will
8106 be able to insert. The default value is 18.
8108 @item omega-hash-table-size
8109 The size of the hash table in the Omega solver. The default value is
8112 @item omega-max-keys
8113 The maximal number of keys used by the Omega solver. The default
8116 @item omega-eliminate-redundant-constraints
8117 When set to 1, use expensive methods to eliminate all redundant
8118 constraints. The default value is 0.
8120 @item vect-max-version-for-alignment-checks
8121 The maximum number of runtime checks that can be performed when
8122 doing loop versioning for alignment in the vectorizer. See option
8123 ftree-vect-loop-version for more information.
8125 @item vect-max-version-for-alias-checks
8126 The maximum number of runtime checks that can be performed when
8127 doing loop versioning for alias in the vectorizer. See option
8128 ftree-vect-loop-version for more information.
8130 @item max-iterations-to-track
8132 The maximum number of iterations of a loop the brute force algorithm
8133 for analysis of # of iterations of the loop tries to evaluate.
8135 @item hot-bb-count-fraction
8136 Select fraction of the maximal count of repetitions of basic block in program
8137 given basic block needs to have to be considered hot.
8139 @item hot-bb-frequency-fraction
8140 Select fraction of the maximal frequency of executions of basic block in
8141 function given basic block needs to have to be considered hot
8143 @item max-predicted-iterations
8144 The maximum number of loop iterations we predict statically. This is useful
8145 in cases where function contain single loop with known bound and other loop
8146 with unknown. We predict the known number of iterations correctly, while
8147 the unknown number of iterations average to roughly 10. This means that the
8148 loop without bounds would appear artificially cold relative to the other one.
8150 @item align-threshold
8152 Select fraction of the maximal frequency of executions of basic block in
8153 function given basic block will get aligned.
8155 @item align-loop-iterations
8157 A loop expected to iterate at lest the selected number of iterations will get
8160 @item tracer-dynamic-coverage
8161 @itemx tracer-dynamic-coverage-feedback
8163 This value is used to limit superblock formation once the given percentage of
8164 executed instructions is covered. This limits unnecessary code size
8167 The @option{tracer-dynamic-coverage-feedback} is used only when profile
8168 feedback is available. The real profiles (as opposed to statically estimated
8169 ones) are much less balanced allowing the threshold to be larger value.
8171 @item tracer-max-code-growth
8172 Stop tail duplication once code growth has reached given percentage. This is
8173 rather hokey argument, as most of the duplicates will be eliminated later in
8174 cross jumping, so it may be set to much higher values than is the desired code
8177 @item tracer-min-branch-ratio
8179 Stop reverse growth when the reverse probability of best edge is less than this
8180 threshold (in percent).
8182 @item tracer-min-branch-ratio
8183 @itemx tracer-min-branch-ratio-feedback
8185 Stop forward growth if the best edge do have probability lower than this
8188 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
8189 compilation for profile feedback and one for compilation without. The value
8190 for compilation with profile feedback needs to be more conservative (higher) in
8191 order to make tracer effective.
8193 @item max-cse-path-length
8195 Maximum number of basic blocks on path that cse considers. The default is 10.
8198 The maximum instructions CSE process before flushing. The default is 1000.
8200 @item ggc-min-expand
8202 GCC uses a garbage collector to manage its own memory allocation. This
8203 parameter specifies the minimum percentage by which the garbage
8204 collector's heap should be allowed to expand between collections.
8205 Tuning this may improve compilation speed; it has no effect on code
8208 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
8209 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
8210 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
8211 GCC is not able to calculate RAM on a particular platform, the lower
8212 bound of 30% is used. Setting this parameter and
8213 @option{ggc-min-heapsize} to zero causes a full collection to occur at
8214 every opportunity. This is extremely slow, but can be useful for
8217 @item ggc-min-heapsize
8219 Minimum size of the garbage collector's heap before it begins bothering
8220 to collect garbage. The first collection occurs after the heap expands
8221 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
8222 tuning this may improve compilation speed, and has no effect on code
8225 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
8226 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
8227 with a lower bound of 4096 (four megabytes) and an upper bound of
8228 131072 (128 megabytes). If GCC is not able to calculate RAM on a
8229 particular platform, the lower bound is used. Setting this parameter
8230 very large effectively disables garbage collection. Setting this
8231 parameter and @option{ggc-min-expand} to zero causes a full collection
8232 to occur at every opportunity.
8234 @item max-reload-search-insns
8235 The maximum number of instruction reload should look backward for equivalent
8236 register. Increasing values mean more aggressive optimization, making the
8237 compile time increase with probably slightly better performance. The default
8240 @item max-cselib-memory-locations
8241 The maximum number of memory locations cselib should take into account.
8242 Increasing values mean more aggressive optimization, making the compile time
8243 increase with probably slightly better performance. The default value is 500.
8245 @item reorder-blocks-duplicate
8246 @itemx reorder-blocks-duplicate-feedback
8248 Used by basic block reordering pass to decide whether to use unconditional
8249 branch or duplicate the code on its destination. Code is duplicated when its
8250 estimated size is smaller than this value multiplied by the estimated size of
8251 unconditional jump in the hot spots of the program.
8253 The @option{reorder-block-duplicate-feedback} is used only when profile
8254 feedback is available and may be set to higher values than
8255 @option{reorder-block-duplicate} since information about the hot spots is more
8258 @item max-sched-ready-insns
8259 The maximum number of instructions ready to be issued the scheduler should
8260 consider at any given time during the first scheduling pass. Increasing
8261 values mean more thorough searches, making the compilation time increase
8262 with probably little benefit. The default value is 100.
8264 @item max-sched-region-blocks
8265 The maximum number of blocks in a region to be considered for
8266 interblock scheduling. The default value is 10.
8268 @item max-pipeline-region-blocks
8269 The maximum number of blocks in a region to be considered for
8270 pipelining in the selective scheduler. The default value is 15.
8272 @item max-sched-region-insns
8273 The maximum number of insns in a region to be considered for
8274 interblock scheduling. The default value is 100.
8276 @item max-pipeline-region-insns
8277 The maximum number of insns in a region to be considered for
8278 pipelining in the selective scheduler. The default value is 200.
8281 The minimum probability (in percents) of reaching a source block
8282 for interblock speculative scheduling. The default value is 40.
8284 @item max-sched-extend-regions-iters
8285 The maximum number of iterations through CFG to extend regions.
8286 0 - disable region extension,
8287 N - do at most N iterations.
8288 The default value is 0.
8290 @item max-sched-insn-conflict-delay
8291 The maximum conflict delay for an insn to be considered for speculative motion.
8292 The default value is 3.
8294 @item sched-spec-prob-cutoff
8295 The minimal probability of speculation success (in percents), so that
8296 speculative insn will be scheduled.
8297 The default value is 40.
8299 @item sched-mem-true-dep-cost
8300 Minimal distance (in CPU cycles) between store and load targeting same
8301 memory locations. The default value is 1.
8303 @item selsched-max-lookahead
8304 The maximum size of the lookahead window of selective scheduling. It is a
8305 depth of search for available instructions.
8306 The default value is 50.
8308 @item selsched-max-sched-times
8309 The maximum number of times that an instruction will be scheduled during
8310 selective scheduling. This is the limit on the number of iterations
8311 through which the instruction may be pipelined. The default value is 2.
8313 @item selsched-max-insns-to-rename
8314 The maximum number of best instructions in the ready list that are considered
8315 for renaming in the selective scheduler. The default value is 2.
8317 @item max-last-value-rtl
8318 The maximum size measured as number of RTLs that can be recorded in an expression
8319 in combiner for a pseudo register as last known value of that register. The default
8322 @item integer-share-limit
8323 Small integer constants can use a shared data structure, reducing the
8324 compiler's memory usage and increasing its speed. This sets the maximum
8325 value of a shared integer constant. The default value is 256.
8327 @item min-virtual-mappings
8328 Specifies the minimum number of virtual mappings in the incremental
8329 SSA updater that should be registered to trigger the virtual mappings
8330 heuristic defined by virtual-mappings-ratio. The default value is
8333 @item virtual-mappings-ratio
8334 If the number of virtual mappings is virtual-mappings-ratio bigger
8335 than the number of virtual symbols to be updated, then the incremental
8336 SSA updater switches to a full update for those symbols. The default
8339 @item ssp-buffer-size
8340 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
8341 protection when @option{-fstack-protection} is used.
8343 @item max-jump-thread-duplication-stmts
8344 Maximum number of statements allowed in a block that needs to be
8345 duplicated when threading jumps.
8347 @item max-fields-for-field-sensitive
8348 Maximum number of fields in a structure we will treat in
8349 a field sensitive manner during pointer analysis. The default is zero
8350 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
8352 @item prefetch-latency
8353 Estimate on average number of instructions that are executed before
8354 prefetch finishes. The distance we prefetch ahead is proportional
8355 to this constant. Increasing this number may also lead to less
8356 streams being prefetched (see @option{simultaneous-prefetches}).
8358 @item simultaneous-prefetches
8359 Maximum number of prefetches that can run at the same time.
8361 @item l1-cache-line-size
8362 The size of cache line in L1 cache, in bytes.
8365 The size of L1 cache, in kilobytes.
8368 The size of L2 cache, in kilobytes.
8370 @item min-insn-to-prefetch-ratio
8371 The minimum ratio between the number of instructions and the
8372 number of prefetches to enable prefetching in a loop with an
8375 @item prefetch-min-insn-to-mem-ratio
8376 The minimum ratio between the number of instructions and the
8377 number of memory references to enable prefetching in a loop.
8379 @item use-canonical-types
8380 Whether the compiler should use the ``canonical'' type system. By
8381 default, this should always be 1, which uses a more efficient internal
8382 mechanism for comparing types in C++ and Objective-C++. However, if
8383 bugs in the canonical type system are causing compilation failures,
8384 set this value to 0 to disable canonical types.
8386 @item switch-conversion-max-branch-ratio
8387 Switch initialization conversion will refuse to create arrays that are
8388 bigger than @option{switch-conversion-max-branch-ratio} times the number of
8389 branches in the switch.
8391 @item max-partial-antic-length
8392 Maximum length of the partial antic set computed during the tree
8393 partial redundancy elimination optimization (@option{-ftree-pre}) when
8394 optimizing at @option{-O3} and above. For some sorts of source code
8395 the enhanced partial redundancy elimination optimization can run away,
8396 consuming all of the memory available on the host machine. This
8397 parameter sets a limit on the length of the sets that are computed,
8398 which prevents the runaway behavior. Setting a value of 0 for
8399 this parameter will allow an unlimited set length.
8401 @item sccvn-max-scc-size
8402 Maximum size of a strongly connected component (SCC) during SCCVN
8403 processing. If this limit is hit, SCCVN processing for the whole
8404 function will not be done and optimizations depending on it will
8405 be disabled. The default maximum SCC size is 10000.
8407 @item ira-max-loops-num
8408 IRA uses a regional register allocation by default. If a function
8409 contains loops more than number given by the parameter, only at most
8410 given number of the most frequently executed loops will form regions
8411 for the regional register allocation. The default value of the
8414 @item ira-max-conflict-table-size
8415 Although IRA uses a sophisticated algorithm of compression conflict
8416 table, the table can be still big for huge functions. If the conflict
8417 table for a function could be more than size in MB given by the
8418 parameter, the conflict table is not built and faster, simpler, and
8419 lower quality register allocation algorithm will be used. The
8420 algorithm do not use pseudo-register conflicts. The default value of
8421 the parameter is 2000.
8423 @item ira-loop-reserved-regs
8424 IRA can be used to evaluate more accurate register pressure in loops
8425 for decision to move loop invariants (see @option{-O3}). The number
8426 of available registers reserved for some other purposes is described
8427 by this parameter. The default value of the parameter is 2 which is
8428 minimal number of registers needed for execution of typical
8429 instruction. This value is the best found from numerous experiments.
8431 @item loop-invariant-max-bbs-in-loop
8432 Loop invariant motion can be very expensive, both in compile time and
8433 in amount of needed compile time memory, with very large loops. Loops
8434 with more basic blocks than this parameter won't have loop invariant
8435 motion optimization performed on them. The default value of the
8436 parameter is 1000 for -O1 and 10000 for -O2 and above.
8438 @item max-vartrack-size
8439 Sets a maximum number of hash table slots to use during variable
8440 tracking dataflow analysis of any function. If this limit is exceeded
8441 with variable tracking at assignments enabled, analysis for that
8442 function is retried without it, after removing all debug insns from
8443 the function. If the limit is exceeded even without debug insns, var
8444 tracking analysis is completely disabled for the function. Setting
8445 the parameter to zero makes it unlimited.
8447 @item min-nondebug-insn-uid
8448 Use uids starting at this parameter for nondebug insns. The range below
8449 the parameter is reserved exclusively for debug insns created by
8450 @option{-fvar-tracking-assignments}, but debug insns may get
8451 (non-overlapping) uids above it if the reserved range is exhausted.
8453 @item ipa-sra-ptr-growth-factor
8454 IPA-SRA will replace a pointer to an aggregate with one or more new
8455 parameters only when their cumulative size is less or equal to
8456 @option{ipa-sra-ptr-growth-factor} times the size of the original
8462 @node Preprocessor Options
8463 @section Options Controlling the Preprocessor
8464 @cindex preprocessor options
8465 @cindex options, preprocessor
8467 These options control the C preprocessor, which is run on each C source
8468 file before actual compilation.
8470 If you use the @option{-E} option, nothing is done except preprocessing.
8471 Some of these options make sense only together with @option{-E} because
8472 they cause the preprocessor output to be unsuitable for actual
8476 @item -Wp,@var{option}
8478 You can use @option{-Wp,@var{option}} to bypass the compiler driver
8479 and pass @var{option} directly through to the preprocessor. If
8480 @var{option} contains commas, it is split into multiple options at the
8481 commas. However, many options are modified, translated or interpreted
8482 by the compiler driver before being passed to the preprocessor, and
8483 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
8484 interface is undocumented and subject to change, so whenever possible
8485 you should avoid using @option{-Wp} and let the driver handle the
8488 @item -Xpreprocessor @var{option}
8489 @opindex Xpreprocessor
8490 Pass @var{option} as an option to the preprocessor. You can use this to
8491 supply system-specific preprocessor options which GCC does not know how to
8494 If you want to pass an option that takes an argument, you must use
8495 @option{-Xpreprocessor} twice, once for the option and once for the argument.
8498 @include cppopts.texi
8500 @node Assembler Options
8501 @section Passing Options to the Assembler
8503 @c prevent bad page break with this line
8504 You can pass options to the assembler.
8507 @item -Wa,@var{option}
8509 Pass @var{option} as an option to the assembler. If @var{option}
8510 contains commas, it is split into multiple options at the commas.
8512 @item -Xassembler @var{option}
8514 Pass @var{option} as an option to the assembler. You can use this to
8515 supply system-specific assembler options which GCC does not know how to
8518 If you want to pass an option that takes an argument, you must use
8519 @option{-Xassembler} twice, once for the option and once for the argument.
8524 @section Options for Linking
8525 @cindex link options
8526 @cindex options, linking
8528 These options come into play when the compiler links object files into
8529 an executable output file. They are meaningless if the compiler is
8530 not doing a link step.
8534 @item @var{object-file-name}
8535 A file name that does not end in a special recognized suffix is
8536 considered to name an object file or library. (Object files are
8537 distinguished from libraries by the linker according to the file
8538 contents.) If linking is done, these object files are used as input
8547 If any of these options is used, then the linker is not run, and
8548 object file names should not be used as arguments. @xref{Overall
8552 @item -l@var{library}
8553 @itemx -l @var{library}
8555 Search the library named @var{library} when linking. (The second
8556 alternative with the library as a separate argument is only for
8557 POSIX compliance and is not recommended.)
8559 It makes a difference where in the command you write this option; the
8560 linker searches and processes libraries and object files in the order they
8561 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
8562 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
8563 to functions in @samp{z}, those functions may not be loaded.
8565 The linker searches a standard list of directories for the library,
8566 which is actually a file named @file{lib@var{library}.a}. The linker
8567 then uses this file as if it had been specified precisely by name.
8569 The directories searched include several standard system directories
8570 plus any that you specify with @option{-L}.
8572 Normally the files found this way are library files---archive files
8573 whose members are object files. The linker handles an archive file by
8574 scanning through it for members which define symbols that have so far
8575 been referenced but not defined. But if the file that is found is an
8576 ordinary object file, it is linked in the usual fashion. The only
8577 difference between using an @option{-l} option and specifying a file name
8578 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
8579 and searches several directories.
8583 You need this special case of the @option{-l} option in order to
8584 link an Objective-C or Objective-C++ program.
8587 @opindex nostartfiles
8588 Do not use the standard system startup files when linking.
8589 The standard system libraries are used normally, unless @option{-nostdlib}
8590 or @option{-nodefaultlibs} is used.
8592 @item -nodefaultlibs
8593 @opindex nodefaultlibs
8594 Do not use the standard system libraries when linking.
8595 Only the libraries you specify will be passed to the linker, options
8596 specifying linkage of the system libraries, such as @code{-static-libgcc}
8597 or @code{-shared-libgcc}, will be ignored.
8598 The standard startup files are used normally, unless @option{-nostartfiles}
8599 is used. The compiler may generate calls to @code{memcmp},
8600 @code{memset}, @code{memcpy} and @code{memmove}.
8601 These entries are usually resolved by entries in
8602 libc. These entry points should be supplied through some other
8603 mechanism when this option is specified.
8607 Do not use the standard system startup files or libraries when linking.
8608 No startup files and only the libraries you specify will be passed to
8609 the linker, options specifying linkage of the system libraries, such as
8610 @code{-static-libgcc} or @code{-shared-libgcc}, will be ignored.
8611 The compiler may generate calls to @code{memcmp}, @code{memset},
8612 @code{memcpy} and @code{memmove}.
8613 These entries are usually resolved by entries in
8614 libc. These entry points should be supplied through some other
8615 mechanism when this option is specified.
8617 @cindex @option{-lgcc}, use with @option{-nostdlib}
8618 @cindex @option{-nostdlib} and unresolved references
8619 @cindex unresolved references and @option{-nostdlib}
8620 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
8621 @cindex @option{-nodefaultlibs} and unresolved references
8622 @cindex unresolved references and @option{-nodefaultlibs}
8623 One of the standard libraries bypassed by @option{-nostdlib} and
8624 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
8625 that GCC uses to overcome shortcomings of particular machines, or special
8626 needs for some languages.
8627 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
8628 Collection (GCC) Internals},
8629 for more discussion of @file{libgcc.a}.)
8630 In most cases, you need @file{libgcc.a} even when you want to avoid
8631 other standard libraries. In other words, when you specify @option{-nostdlib}
8632 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
8633 This ensures that you have no unresolved references to internal GCC
8634 library subroutines. (For example, @samp{__main}, used to ensure C++
8635 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
8636 GNU Compiler Collection (GCC) Internals}.)
8640 Produce a position independent executable on targets which support it.
8641 For predictable results, you must also specify the same set of options
8642 that were used to generate code (@option{-fpie}, @option{-fPIE},
8643 or model suboptions) when you specify this option.
8647 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
8648 that support it. This instructs the linker to add all symbols, not
8649 only used ones, to the dynamic symbol table. This option is needed
8650 for some uses of @code{dlopen} or to allow obtaining backtraces
8651 from within a program.
8655 Remove all symbol table and relocation information from the executable.
8659 On systems that support dynamic linking, this prevents linking with the shared
8660 libraries. On other systems, this option has no effect.
8664 Produce a shared object which can then be linked with other objects to
8665 form an executable. Not all systems support this option. For predictable
8666 results, you must also specify the same set of options that were used to
8667 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
8668 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
8669 needs to build supplementary stub code for constructors to work. On
8670 multi-libbed systems, @samp{gcc -shared} must select the correct support
8671 libraries to link against. Failing to supply the correct flags may lead
8672 to subtle defects. Supplying them in cases where they are not necessary
8675 @item -shared-libgcc
8676 @itemx -static-libgcc
8677 @opindex shared-libgcc
8678 @opindex static-libgcc
8679 On systems that provide @file{libgcc} as a shared library, these options
8680 force the use of either the shared or static version respectively.
8681 If no shared version of @file{libgcc} was built when the compiler was
8682 configured, these options have no effect.
8684 There are several situations in which an application should use the
8685 shared @file{libgcc} instead of the static version. The most common
8686 of these is when the application wishes to throw and catch exceptions
8687 across different shared libraries. In that case, each of the libraries
8688 as well as the application itself should use the shared @file{libgcc}.
8690 Therefore, the G++ and GCJ drivers automatically add
8691 @option{-shared-libgcc} whenever you build a shared library or a main
8692 executable, because C++ and Java programs typically use exceptions, so
8693 this is the right thing to do.
8695 If, instead, you use the GCC driver to create shared libraries, you may
8696 find that they will not always be linked with the shared @file{libgcc}.
8697 If GCC finds, at its configuration time, that you have a non-GNU linker
8698 or a GNU linker that does not support option @option{--eh-frame-hdr},
8699 it will link the shared version of @file{libgcc} into shared libraries
8700 by default. Otherwise, it will take advantage of the linker and optimize
8701 away the linking with the shared version of @file{libgcc}, linking with
8702 the static version of libgcc by default. This allows exceptions to
8703 propagate through such shared libraries, without incurring relocation
8704 costs at library load time.
8706 However, if a library or main executable is supposed to throw or catch
8707 exceptions, you must link it using the G++ or GCJ driver, as appropriate
8708 for the languages used in the program, or using the option
8709 @option{-shared-libgcc}, such that it is linked with the shared
8712 @item -static-libstdc++
8713 When the @command{g++} program is used to link a C++ program, it will
8714 normally automatically link against @option{libstdc++}. If
8715 @file{libstdc++} is available as a shared library, and the
8716 @option{-static} option is not used, then this will link against the
8717 shared version of @file{libstdc++}. That is normally fine. However, it
8718 is sometimes useful to freeze the version of @file{libstdc++} used by
8719 the program without going all the way to a fully static link. The
8720 @option{-static-libstdc++} option directs the @command{g++} driver to
8721 link @file{libstdc++} statically, without necessarily linking other
8722 libraries statically.
8726 Bind references to global symbols when building a shared object. Warn
8727 about any unresolved references (unless overridden by the link editor
8728 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
8731 @item -T @var{script}
8733 @cindex linker script
8734 Use @var{script} as the linker script. This option is supported by most
8735 systems using the GNU linker. On some targets, such as bare-board
8736 targets without an operating system, the @option{-T} option may be required
8737 when linking to avoid references to undefined symbols.
8739 @item -Xlinker @var{option}
8741 Pass @var{option} as an option to the linker. You can use this to
8742 supply system-specific linker options which GCC does not know how to
8745 If you want to pass an option that takes a separate argument, you must use
8746 @option{-Xlinker} twice, once for the option and once for the argument.
8747 For example, to pass @option{-assert definitions}, you must write
8748 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
8749 @option{-Xlinker "-assert definitions"}, because this passes the entire
8750 string as a single argument, which is not what the linker expects.
8752 When using the GNU linker, it is usually more convenient to pass
8753 arguments to linker options using the @option{@var{option}=@var{value}}
8754 syntax than as separate arguments. For example, you can specify
8755 @samp{-Xlinker -Map=output.map} rather than
8756 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
8757 this syntax for command-line options.
8759 @item -Wl,@var{option}
8761 Pass @var{option} as an option to the linker. If @var{option} contains
8762 commas, it is split into multiple options at the commas. You can use this
8763 syntax to pass an argument to the option.
8764 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
8765 linker. When using the GNU linker, you can also get the same effect with
8766 @samp{-Wl,-Map=output.map}.
8768 @item -u @var{symbol}
8770 Pretend the symbol @var{symbol} is undefined, to force linking of
8771 library modules to define it. You can use @option{-u} multiple times with
8772 different symbols to force loading of additional library modules.
8775 @node Directory Options
8776 @section Options for Directory Search
8777 @cindex directory options
8778 @cindex options, directory search
8781 These options specify directories to search for header files, for
8782 libraries and for parts of the compiler:
8787 Add the directory @var{dir} to the head of the list of directories to be
8788 searched for header files. This can be used to override a system header
8789 file, substituting your own version, since these directories are
8790 searched before the system header file directories. However, you should
8791 not use this option to add directories that contain vendor-supplied
8792 system header files (use @option{-isystem} for that). If you use more than
8793 one @option{-I} option, the directories are scanned in left-to-right
8794 order; the standard system directories come after.
8796 If a standard system include directory, or a directory specified with
8797 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
8798 option will be ignored. The directory will still be searched but as a
8799 system directory at its normal position in the system include chain.
8800 This is to ensure that GCC's procedure to fix buggy system headers and
8801 the ordering for the include_next directive are not inadvertently changed.
8802 If you really need to change the search order for system directories,
8803 use the @option{-nostdinc} and/or @option{-isystem} options.
8805 @item -iquote@var{dir}
8807 Add the directory @var{dir} to the head of the list of directories to
8808 be searched for header files only for the case of @samp{#include
8809 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
8810 otherwise just like @option{-I}.
8814 Add directory @var{dir} to the list of directories to be searched
8817 @item -B@var{prefix}
8819 This option specifies where to find the executables, libraries,
8820 include files, and data files of the compiler itself.
8822 The compiler driver program runs one or more of the subprograms
8823 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
8824 @var{prefix} as a prefix for each program it tries to run, both with and
8825 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
8827 For each subprogram to be run, the compiler driver first tries the
8828 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
8829 was not specified, the driver tries two standard prefixes, which are
8830 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
8831 those results in a file name that is found, the unmodified program
8832 name is searched for using the directories specified in your
8833 @env{PATH} environment variable.
8835 The compiler will check to see if the path provided by the @option{-B}
8836 refers to a directory, and if necessary it will add a directory
8837 separator character at the end of the path.
8839 @option{-B} prefixes that effectively specify directory names also apply
8840 to libraries in the linker, because the compiler translates these
8841 options into @option{-L} options for the linker. They also apply to
8842 includes files in the preprocessor, because the compiler translates these
8843 options into @option{-isystem} options for the preprocessor. In this case,
8844 the compiler appends @samp{include} to the prefix.
8846 The run-time support file @file{libgcc.a} can also be searched for using
8847 the @option{-B} prefix, if needed. If it is not found there, the two
8848 standard prefixes above are tried, and that is all. The file is left
8849 out of the link if it is not found by those means.
8851 Another way to specify a prefix much like the @option{-B} prefix is to use
8852 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
8855 As a special kludge, if the path provided by @option{-B} is
8856 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
8857 9, then it will be replaced by @file{[dir/]include}. This is to help
8858 with boot-strapping the compiler.
8860 @item -specs=@var{file}
8862 Process @var{file} after the compiler reads in the standard @file{specs}
8863 file, in order to override the defaults that the @file{gcc} driver
8864 program uses when determining what switches to pass to @file{cc1},
8865 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
8866 @option{-specs=@var{file}} can be specified on the command line, and they
8867 are processed in order, from left to right.
8869 @item --sysroot=@var{dir}
8871 Use @var{dir} as the logical root directory for headers and libraries.
8872 For example, if the compiler would normally search for headers in
8873 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
8874 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
8876 If you use both this option and the @option{-isysroot} option, then
8877 the @option{--sysroot} option will apply to libraries, but the
8878 @option{-isysroot} option will apply to header files.
8880 The GNU linker (beginning with version 2.16) has the necessary support
8881 for this option. If your linker does not support this option, the
8882 header file aspect of @option{--sysroot} will still work, but the
8883 library aspect will not.
8887 This option has been deprecated. Please use @option{-iquote} instead for
8888 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
8889 Any directories you specify with @option{-I} options before the @option{-I-}
8890 option are searched only for the case of @samp{#include "@var{file}"};
8891 they are not searched for @samp{#include <@var{file}>}.
8893 If additional directories are specified with @option{-I} options after
8894 the @option{-I-}, these directories are searched for all @samp{#include}
8895 directives. (Ordinarily @emph{all} @option{-I} directories are used
8898 In addition, the @option{-I-} option inhibits the use of the current
8899 directory (where the current input file came from) as the first search
8900 directory for @samp{#include "@var{file}"}. There is no way to
8901 override this effect of @option{-I-}. With @option{-I.} you can specify
8902 searching the directory which was current when the compiler was
8903 invoked. That is not exactly the same as what the preprocessor does
8904 by default, but it is often satisfactory.
8906 @option{-I-} does not inhibit the use of the standard system directories
8907 for header files. Thus, @option{-I-} and @option{-nostdinc} are
8914 @section Specifying subprocesses and the switches to pass to them
8917 @command{gcc} is a driver program. It performs its job by invoking a
8918 sequence of other programs to do the work of compiling, assembling and
8919 linking. GCC interprets its command-line parameters and uses these to
8920 deduce which programs it should invoke, and which command-line options
8921 it ought to place on their command lines. This behavior is controlled
8922 by @dfn{spec strings}. In most cases there is one spec string for each
8923 program that GCC can invoke, but a few programs have multiple spec
8924 strings to control their behavior. The spec strings built into GCC can
8925 be overridden by using the @option{-specs=} command-line switch to specify
8928 @dfn{Spec files} are plaintext files that are used to construct spec
8929 strings. They consist of a sequence of directives separated by blank
8930 lines. The type of directive is determined by the first non-whitespace
8931 character on the line and it can be one of the following:
8934 @item %@var{command}
8935 Issues a @var{command} to the spec file processor. The commands that can
8939 @item %include <@var{file}>
8941 Search for @var{file} and insert its text at the current point in the
8944 @item %include_noerr <@var{file}>
8945 @cindex %include_noerr
8946 Just like @samp{%include}, but do not generate an error message if the include
8947 file cannot be found.
8949 @item %rename @var{old_name} @var{new_name}
8951 Rename the spec string @var{old_name} to @var{new_name}.
8955 @item *[@var{spec_name}]:
8956 This tells the compiler to create, override or delete the named spec
8957 string. All lines after this directive up to the next directive or
8958 blank line are considered to be the text for the spec string. If this
8959 results in an empty string then the spec will be deleted. (Or, if the
8960 spec did not exist, then nothing will happened.) Otherwise, if the spec
8961 does not currently exist a new spec will be created. If the spec does
8962 exist then its contents will be overridden by the text of this
8963 directive, unless the first character of that text is the @samp{+}
8964 character, in which case the text will be appended to the spec.
8966 @item [@var{suffix}]:
8967 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
8968 and up to the next directive or blank line are considered to make up the
8969 spec string for the indicated suffix. When the compiler encounters an
8970 input file with the named suffix, it will processes the spec string in
8971 order to work out how to compile that file. For example:
8978 This says that any input file whose name ends in @samp{.ZZ} should be
8979 passed to the program @samp{z-compile}, which should be invoked with the
8980 command-line switch @option{-input} and with the result of performing the
8981 @samp{%i} substitution. (See below.)
8983 As an alternative to providing a spec string, the text that follows a
8984 suffix directive can be one of the following:
8987 @item @@@var{language}
8988 This says that the suffix is an alias for a known @var{language}. This is
8989 similar to using the @option{-x} command-line switch to GCC to specify a
8990 language explicitly. For example:
8997 Says that .ZZ files are, in fact, C++ source files.
9000 This causes an error messages saying:
9003 @var{name} compiler not installed on this system.
9007 GCC already has an extensive list of suffixes built into it.
9008 This directive will add an entry to the end of the list of suffixes, but
9009 since the list is searched from the end backwards, it is effectively
9010 possible to override earlier entries using this technique.
9014 GCC has the following spec strings built into it. Spec files can
9015 override these strings or create their own. Note that individual
9016 targets can also add their own spec strings to this list.
9019 asm Options to pass to the assembler
9020 asm_final Options to pass to the assembler post-processor
9021 cpp Options to pass to the C preprocessor
9022 cc1 Options to pass to the C compiler
9023 cc1plus Options to pass to the C++ compiler
9024 endfile Object files to include at the end of the link
9025 link Options to pass to the linker
9026 lib Libraries to include on the command line to the linker
9027 libgcc Decides which GCC support library to pass to the linker
9028 linker Sets the name of the linker
9029 predefines Defines to be passed to the C preprocessor
9030 signed_char Defines to pass to CPP to say whether @code{char} is signed
9032 startfile Object files to include at the start of the link
9035 Here is a small example of a spec file:
9041 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
9044 This example renames the spec called @samp{lib} to @samp{old_lib} and
9045 then overrides the previous definition of @samp{lib} with a new one.
9046 The new definition adds in some extra command-line options before
9047 including the text of the old definition.
9049 @dfn{Spec strings} are a list of command-line options to be passed to their
9050 corresponding program. In addition, the spec strings can contain
9051 @samp{%}-prefixed sequences to substitute variable text or to
9052 conditionally insert text into the command line. Using these constructs
9053 it is possible to generate quite complex command lines.
9055 Here is a table of all defined @samp{%}-sequences for spec
9056 strings. Note that spaces are not generated automatically around the
9057 results of expanding these sequences. Therefore you can concatenate them
9058 together or combine them with constant text in a single argument.
9062 Substitute one @samp{%} into the program name or argument.
9065 Substitute the name of the input file being processed.
9068 Substitute the basename of the input file being processed.
9069 This is the substring up to (and not including) the last period
9070 and not including the directory.
9073 This is the same as @samp{%b}, but include the file suffix (text after
9077 Marks the argument containing or following the @samp{%d} as a
9078 temporary file name, so that that file will be deleted if GCC exits
9079 successfully. Unlike @samp{%g}, this contributes no text to the
9082 @item %g@var{suffix}
9083 Substitute a file name that has suffix @var{suffix} and is chosen
9084 once per compilation, and mark the argument in the same way as
9085 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
9086 name is now chosen in a way that is hard to predict even when previously
9087 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
9088 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
9089 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
9090 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
9091 was simply substituted with a file name chosen once per compilation,
9092 without regard to any appended suffix (which was therefore treated
9093 just like ordinary text), making such attacks more likely to succeed.
9095 @item %u@var{suffix}
9096 Like @samp{%g}, but generates a new temporary file name even if
9097 @samp{%u@var{suffix}} was already seen.
9099 @item %U@var{suffix}
9100 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
9101 new one if there is no such last file name. In the absence of any
9102 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
9103 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
9104 would involve the generation of two distinct file names, one
9105 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
9106 simply substituted with a file name chosen for the previous @samp{%u},
9107 without regard to any appended suffix.
9109 @item %j@var{suffix}
9110 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
9111 writable, and if save-temps is off; otherwise, substitute the name
9112 of a temporary file, just like @samp{%u}. This temporary file is not
9113 meant for communication between processes, but rather as a junk
9116 @item %|@var{suffix}
9117 @itemx %m@var{suffix}
9118 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
9119 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
9120 all. These are the two most common ways to instruct a program that it
9121 should read from standard input or write to standard output. If you
9122 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
9123 construct: see for example @file{f/lang-specs.h}.
9125 @item %.@var{SUFFIX}
9126 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
9127 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
9128 terminated by the next space or %.
9131 Marks the argument containing or following the @samp{%w} as the
9132 designated output file of this compilation. This puts the argument
9133 into the sequence of arguments that @samp{%o} will substitute later.
9136 Substitutes the names of all the output files, with spaces
9137 automatically placed around them. You should write spaces
9138 around the @samp{%o} as well or the results are undefined.
9139 @samp{%o} is for use in the specs for running the linker.
9140 Input files whose names have no recognized suffix are not compiled
9141 at all, but they are included among the output files, so they will
9145 Substitutes the suffix for object files. Note that this is
9146 handled specially when it immediately follows @samp{%g, %u, or %U},
9147 because of the need for those to form complete file names. The
9148 handling is such that @samp{%O} is treated exactly as if it had already
9149 been substituted, except that @samp{%g, %u, and %U} do not currently
9150 support additional @var{suffix} characters following @samp{%O} as they would
9151 following, for example, @samp{.o}.
9154 Substitutes the standard macro predefinitions for the
9155 current target machine. Use this when running @code{cpp}.
9158 Like @samp{%p}, but puts @samp{__} before and after the name of each
9159 predefined macro, except for macros that start with @samp{__} or with
9160 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
9164 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
9165 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
9166 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
9167 and @option{-imultilib} as necessary.
9170 Current argument is the name of a library or startup file of some sort.
9171 Search for that file in a standard list of directories and substitute
9172 the full name found. The current working directory is included in the
9173 list of directories scanned.
9176 Current argument is the name of a linker script. Search for that file
9177 in the current list of directories to scan for libraries. If the file
9178 is located insert a @option{--script} option into the command line
9179 followed by the full path name found. If the file is not found then
9180 generate an error message. Note: the current working directory is not
9184 Print @var{str} as an error message. @var{str} is terminated by a newline.
9185 Use this when inconsistent options are detected.
9188 Substitute the contents of spec string @var{name} at this point.
9191 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
9193 @item %x@{@var{option}@}
9194 Accumulate an option for @samp{%X}.
9197 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
9201 Output the accumulated assembler options specified by @option{-Wa}.
9204 Output the accumulated preprocessor options specified by @option{-Wp}.
9207 Process the @code{asm} spec. This is used to compute the
9208 switches to be passed to the assembler.
9211 Process the @code{asm_final} spec. This is a spec string for
9212 passing switches to an assembler post-processor, if such a program is
9216 Process the @code{link} spec. This is the spec for computing the
9217 command line passed to the linker. Typically it will make use of the
9218 @samp{%L %G %S %D and %E} sequences.
9221 Dump out a @option{-L} option for each directory that GCC believes might
9222 contain startup files. If the target supports multilibs then the
9223 current multilib directory will be prepended to each of these paths.
9226 Process the @code{lib} spec. This is a spec string for deciding which
9227 libraries should be included on the command line to the linker.
9230 Process the @code{libgcc} spec. This is a spec string for deciding
9231 which GCC support library should be included on the command line to the linker.
9234 Process the @code{startfile} spec. This is a spec for deciding which
9235 object files should be the first ones passed to the linker. Typically
9236 this might be a file named @file{crt0.o}.
9239 Process the @code{endfile} spec. This is a spec string that specifies
9240 the last object files that will be passed to the linker.
9243 Process the @code{cpp} spec. This is used to construct the arguments
9244 to be passed to the C preprocessor.
9247 Process the @code{cc1} spec. This is used to construct the options to be
9248 passed to the actual C compiler (@samp{cc1}).
9251 Process the @code{cc1plus} spec. This is used to construct the options to be
9252 passed to the actual C++ compiler (@samp{cc1plus}).
9255 Substitute the variable part of a matched option. See below.
9256 Note that each comma in the substituted string is replaced by
9260 Remove all occurrences of @code{-S} from the command line. Note---this
9261 command is position dependent. @samp{%} commands in the spec string
9262 before this one will see @code{-S}, @samp{%} commands in the spec string
9263 after this one will not.
9265 @item %:@var{function}(@var{args})
9266 Call the named function @var{function}, passing it @var{args}.
9267 @var{args} is first processed as a nested spec string, then split
9268 into an argument vector in the usual fashion. The function returns
9269 a string which is processed as if it had appeared literally as part
9270 of the current spec.
9272 The following built-in spec functions are provided:
9276 The @code{getenv} spec function takes two arguments: an environment
9277 variable name and a string. If the environment variable is not
9278 defined, a fatal error is issued. Otherwise, the return value is the
9279 value of the environment variable concatenated with the string. For
9280 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
9283 %:getenv(TOPDIR /include)
9286 expands to @file{/path/to/top/include}.
9288 @item @code{if-exists}
9289 The @code{if-exists} spec function takes one argument, an absolute
9290 pathname to a file. If the file exists, @code{if-exists} returns the
9291 pathname. Here is a small example of its usage:
9295 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
9298 @item @code{if-exists-else}
9299 The @code{if-exists-else} spec function is similar to the @code{if-exists}
9300 spec function, except that it takes two arguments. The first argument is
9301 an absolute pathname to a file. If the file exists, @code{if-exists-else}
9302 returns the pathname. If it does not exist, it returns the second argument.
9303 This way, @code{if-exists-else} can be used to select one file or another,
9304 based on the existence of the first. Here is a small example of its usage:
9308 crt0%O%s %:if-exists(crti%O%s) \
9309 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
9312 @item @code{replace-outfile}
9313 The @code{replace-outfile} spec function takes two arguments. It looks for the
9314 first argument in the outfiles array and replaces it with the second argument. Here
9315 is a small example of its usage:
9318 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
9321 @item @code{print-asm-header}
9322 The @code{print-asm-header} function takes no arguments and simply
9323 prints a banner like:
9329 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
9332 It is used to separate compiler options from assembler options
9333 in the @option{--target-help} output.
9337 Substitutes the @code{-S} switch, if that switch was given to GCC@.
9338 If that switch was not specified, this substitutes nothing. Note that
9339 the leading dash is omitted when specifying this option, and it is
9340 automatically inserted if the substitution is performed. Thus the spec
9341 string @samp{%@{foo@}} would match the command-line option @option{-foo}
9342 and would output the command line option @option{-foo}.
9344 @item %W@{@code{S}@}
9345 Like %@{@code{S}@} but mark last argument supplied within as a file to be
9348 @item %@{@code{S}*@}
9349 Substitutes all the switches specified to GCC whose names start
9350 with @code{-S}, but which also take an argument. This is used for
9351 switches like @option{-o}, @option{-D}, @option{-I}, etc.
9352 GCC considers @option{-o foo} as being
9353 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
9354 text, including the space. Thus two arguments would be generated.
9356 @item %@{@code{S}*&@code{T}*@}
9357 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
9358 (the order of @code{S} and @code{T} in the spec is not significant).
9359 There can be any number of ampersand-separated variables; for each the
9360 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
9362 @item %@{@code{S}:@code{X}@}
9363 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
9365 @item %@{!@code{S}:@code{X}@}
9366 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
9368 @item %@{@code{S}*:@code{X}@}
9369 Substitutes @code{X} if one or more switches whose names start with
9370 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
9371 once, no matter how many such switches appeared. However, if @code{%*}
9372 appears somewhere in @code{X}, then @code{X} will be substituted once
9373 for each matching switch, with the @code{%*} replaced by the part of
9374 that switch that matched the @code{*}.
9376 @item %@{.@code{S}:@code{X}@}
9377 Substitutes @code{X}, if processing a file with suffix @code{S}.
9379 @item %@{!.@code{S}:@code{X}@}
9380 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
9382 @item %@{,@code{S}:@code{X}@}
9383 Substitutes @code{X}, if processing a file for language @code{S}.
9385 @item %@{!,@code{S}:@code{X}@}
9386 Substitutes @code{X}, if not processing a file for language @code{S}.
9388 @item %@{@code{S}|@code{P}:@code{X}@}
9389 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
9390 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
9391 @code{*} sequences as well, although they have a stronger binding than
9392 the @samp{|}. If @code{%*} appears in @code{X}, all of the
9393 alternatives must be starred, and only the first matching alternative
9396 For example, a spec string like this:
9399 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
9402 will output the following command-line options from the following input
9403 command-line options:
9408 -d fred.c -foo -baz -boggle
9409 -d jim.d -bar -baz -boggle
9412 @item %@{S:X; T:Y; :D@}
9414 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
9415 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
9416 be as many clauses as you need. This may be combined with @code{.},
9417 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
9422 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
9423 construct may contain other nested @samp{%} constructs or spaces, or
9424 even newlines. They are processed as usual, as described above.
9425 Trailing white space in @code{X} is ignored. White space may also
9426 appear anywhere on the left side of the colon in these constructs,
9427 except between @code{.} or @code{*} and the corresponding word.
9429 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
9430 handled specifically in these constructs. If another value of
9431 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
9432 @option{-W} switch is found later in the command line, the earlier
9433 switch value is ignored, except with @{@code{S}*@} where @code{S} is
9434 just one letter, which passes all matching options.
9436 The character @samp{|} at the beginning of the predicate text is used to
9437 indicate that a command should be piped to the following command, but
9438 only if @option{-pipe} is specified.
9440 It is built into GCC which switches take arguments and which do not.
9441 (You might think it would be useful to generalize this to allow each
9442 compiler's spec to say which switches take arguments. But this cannot
9443 be done in a consistent fashion. GCC cannot even decide which input
9444 files have been specified without knowing which switches take arguments,
9445 and it must know which input files to compile in order to tell which
9448 GCC also knows implicitly that arguments starting in @option{-l} are to be
9449 treated as compiler output files, and passed to the linker in their
9450 proper position among the other output files.
9452 @c man begin OPTIONS
9454 @node Target Options
9455 @section Specifying Target Machine and Compiler Version
9456 @cindex target options
9457 @cindex cross compiling
9458 @cindex specifying machine version
9459 @cindex specifying compiler version and target machine
9460 @cindex compiler version, specifying
9461 @cindex target machine, specifying
9463 The usual way to run GCC is to run the executable called @file{gcc}, or
9464 @file{<machine>-gcc} when cross-compiling, or
9465 @file{<machine>-gcc-<version>} to run a version other than the one that
9466 was installed last. Sometimes this is inconvenient, so GCC provides
9467 options that will switch to another cross-compiler or version.
9470 @item -b @var{machine}
9472 The argument @var{machine} specifies the target machine for compilation.
9474 The value to use for @var{machine} is the same as was specified as the
9475 machine type when configuring GCC as a cross-compiler. For
9476 example, if a cross-compiler was configured with @samp{configure
9477 arm-elf}, meaning to compile for an arm processor with elf binaries,
9478 then you would specify @option{-b arm-elf} to run that cross compiler.
9479 Because there are other options beginning with @option{-b}, the
9480 configuration must contain a hyphen, or @option{-b} alone should be one
9481 argument followed by the configuration in the next argument.
9483 @item -V @var{version}
9485 The argument @var{version} specifies which version of GCC to run.
9486 This is useful when multiple versions are installed. For example,
9487 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
9490 The @option{-V} and @option{-b} options work by running the
9491 @file{<machine>-gcc-<version>} executable, so there's no real reason to
9492 use them if you can just run that directly.
9494 @node Submodel Options
9495 @section Hardware Models and Configurations
9496 @cindex submodel options
9497 @cindex specifying hardware config
9498 @cindex hardware models and configurations, specifying
9499 @cindex machine dependent options
9501 Earlier we discussed the standard option @option{-b} which chooses among
9502 different installed compilers for completely different target
9503 machines, such as VAX vs.@: 68000 vs.@: 80386.
9505 In addition, each of these target machine types can have its own
9506 special options, starting with @samp{-m}, to choose among various
9507 hardware models or configurations---for example, 68010 vs 68020,
9508 floating coprocessor or none. A single installed version of the
9509 compiler can compile for any model or configuration, according to the
9512 Some configurations of the compiler also support additional special
9513 options, usually for compatibility with other compilers on the same
9516 @c This list is ordered alphanumerically by subsection name.
9517 @c It should be the same order and spelling as these options are listed
9518 @c in Machine Dependent Options
9524 * Blackfin Options::
9528 * DEC Alpha Options::
9529 * DEC Alpha/VMS Options::
9532 * GNU/Linux Options::
9535 * i386 and x86-64 Options::
9536 * i386 and x86-64 Windows Options::
9538 * IA-64/VMS Options::
9550 * picoChip Options::
9552 * RS/6000 and PowerPC Options::
9554 * S/390 and zSeries Options::
9559 * System V Options::
9564 * Xstormy16 Options::
9570 @subsection ARC Options
9573 These options are defined for ARC implementations:
9578 Compile code for little endian mode. This is the default.
9582 Compile code for big endian mode.
9585 @opindex mmangle-cpu
9586 Prepend the name of the cpu to all public symbol names.
9587 In multiple-processor systems, there are many ARC variants with different
9588 instruction and register set characteristics. This flag prevents code
9589 compiled for one cpu to be linked with code compiled for another.
9590 No facility exists for handling variants that are ``almost identical''.
9591 This is an all or nothing option.
9593 @item -mcpu=@var{cpu}
9595 Compile code for ARC variant @var{cpu}.
9596 Which variants are supported depend on the configuration.
9597 All variants support @option{-mcpu=base}, this is the default.
9599 @item -mtext=@var{text-section}
9600 @itemx -mdata=@var{data-section}
9601 @itemx -mrodata=@var{readonly-data-section}
9605 Put functions, data, and readonly data in @var{text-section},
9606 @var{data-section}, and @var{readonly-data-section} respectively
9607 by default. This can be overridden with the @code{section} attribute.
9608 @xref{Variable Attributes}.
9610 @item -mfix-cortex-m3-ldrd
9611 @opindex mfix-cortex-m3-ldrd
9612 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
9613 with overlapping destination and base registers are used. This option avoids
9614 generating these instructions. This option is enabled by default when
9615 @option{-mcpu=cortex-m3} is specified.
9620 @subsection ARM Options
9623 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
9627 @item -mabi=@var{name}
9629 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
9630 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
9633 @opindex mapcs-frame
9634 Generate a stack frame that is compliant with the ARM Procedure Call
9635 Standard for all functions, even if this is not strictly necessary for
9636 correct execution of the code. Specifying @option{-fomit-frame-pointer}
9637 with this option will cause the stack frames not to be generated for
9638 leaf functions. The default is @option{-mno-apcs-frame}.
9642 This is a synonym for @option{-mapcs-frame}.
9645 @c not currently implemented
9646 @item -mapcs-stack-check
9647 @opindex mapcs-stack-check
9648 Generate code to check the amount of stack space available upon entry to
9649 every function (that actually uses some stack space). If there is
9650 insufficient space available then either the function
9651 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
9652 called, depending upon the amount of stack space required. The run time
9653 system is required to provide these functions. The default is
9654 @option{-mno-apcs-stack-check}, since this produces smaller code.
9656 @c not currently implemented
9658 @opindex mapcs-float
9659 Pass floating point arguments using the float point registers. This is
9660 one of the variants of the APCS@. This option is recommended if the
9661 target hardware has a floating point unit or if a lot of floating point
9662 arithmetic is going to be performed by the code. The default is
9663 @option{-mno-apcs-float}, since integer only code is slightly increased in
9664 size if @option{-mapcs-float} is used.
9666 @c not currently implemented
9667 @item -mapcs-reentrant
9668 @opindex mapcs-reentrant
9669 Generate reentrant, position independent code. The default is
9670 @option{-mno-apcs-reentrant}.
9673 @item -mthumb-interwork
9674 @opindex mthumb-interwork
9675 Generate code which supports calling between the ARM and Thumb
9676 instruction sets. Without this option the two instruction sets cannot
9677 be reliably used inside one program. The default is
9678 @option{-mno-thumb-interwork}, since slightly larger code is generated
9679 when @option{-mthumb-interwork} is specified.
9681 @item -mno-sched-prolog
9682 @opindex mno-sched-prolog
9683 Prevent the reordering of instructions in the function prolog, or the
9684 merging of those instruction with the instructions in the function's
9685 body. This means that all functions will start with a recognizable set
9686 of instructions (or in fact one of a choice from a small set of
9687 different function prologues), and this information can be used to
9688 locate the start if functions inside an executable piece of code. The
9689 default is @option{-msched-prolog}.
9691 @item -mfloat-abi=@var{name}
9693 Specifies which floating-point ABI to use. Permissible values
9694 are: @samp{soft}, @samp{softfp} and @samp{hard}.
9696 Specifying @samp{soft} causes GCC to generate output containing
9697 library calls for floating-point operations.
9698 @samp{softfp} allows the generation of code using hardware floating-point
9699 instructions, but still uses the soft-float calling conventions.
9700 @samp{hard} allows generation of floating-point instructions
9701 and uses FPU-specific calling conventions.
9703 The default depends on the specific target configuration. Note that
9704 the hard-float and soft-float ABIs are not link-compatible; you must
9705 compile your entire program with the same ABI, and link with a
9706 compatible set of libraries.
9709 @opindex mhard-float
9710 Equivalent to @option{-mfloat-abi=hard}.
9713 @opindex msoft-float
9714 Equivalent to @option{-mfloat-abi=soft}.
9716 @item -mlittle-endian
9717 @opindex mlittle-endian
9718 Generate code for a processor running in little-endian mode. This is
9719 the default for all standard configurations.
9722 @opindex mbig-endian
9723 Generate code for a processor running in big-endian mode; the default is
9724 to compile code for a little-endian processor.
9726 @item -mwords-little-endian
9727 @opindex mwords-little-endian
9728 This option only applies when generating code for big-endian processors.
9729 Generate code for a little-endian word order but a big-endian byte
9730 order. That is, a byte order of the form @samp{32107654}. Note: this
9731 option should only be used if you require compatibility with code for
9732 big-endian ARM processors generated by versions of the compiler prior to
9735 @item -mcpu=@var{name}
9737 This specifies the name of the target ARM processor. GCC uses this name
9738 to determine what kind of instructions it can emit when generating
9739 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
9740 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
9741 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
9742 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
9743 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
9745 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
9746 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
9747 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
9748 @samp{strongarm1110},
9749 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
9750 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
9751 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
9752 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
9753 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
9754 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
9755 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
9756 @samp{cortex-a5}, @samp{cortex-a8}, @samp{cortex-a9},
9757 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-m3},
9760 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9762 @item -mtune=@var{name}
9764 This option is very similar to the @option{-mcpu=} option, except that
9765 instead of specifying the actual target processor type, and hence
9766 restricting which instructions can be used, it specifies that GCC should
9767 tune the performance of the code as if the target were of the type
9768 specified in this option, but still choosing the instructions that it
9769 will generate based on the cpu specified by a @option{-mcpu=} option.
9770 For some ARM implementations better performance can be obtained by using
9773 @item -march=@var{name}
9775 This specifies the name of the target ARM architecture. GCC uses this
9776 name to determine what kind of instructions it can emit when generating
9777 assembly code. This option can be used in conjunction with or instead
9778 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
9779 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
9780 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
9781 @samp{armv6}, @samp{armv6j},
9782 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
9783 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
9784 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9786 @item -mfpu=@var{name}
9787 @itemx -mfpe=@var{number}
9788 @itemx -mfp=@var{number}
9792 This specifies what floating point hardware (or hardware emulation) is
9793 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
9794 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-fp16},
9795 @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd}, @samp{vfpv3xd-fp16},
9796 @samp{neon}, @samp{neon-fp16}, @samp{vfpv4}, @samp{vfpv4-d16},
9797 @samp{fpv4-sp-d16} and @samp{neon-vfpv4}.
9798 @option{-mfp} and @option{-mfpe} are synonyms for
9799 @option{-mfpu}=@samp{fpe}@var{number}, for compatibility with older versions
9802 If @option{-msoft-float} is specified this specifies the format of
9803 floating point values.
9805 @item -mfp16-format=@var{name}
9806 @opindex mfp16-format
9807 Specify the format of the @code{__fp16} half-precision floating-point type.
9808 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
9809 the default is @samp{none}, in which case the @code{__fp16} type is not
9810 defined. @xref{Half-Precision}, for more information.
9812 @item -mstructure-size-boundary=@var{n}
9813 @opindex mstructure-size-boundary
9814 The size of all structures and unions will be rounded up to a multiple
9815 of the number of bits set by this option. Permissible values are 8, 32
9816 and 64. The default value varies for different toolchains. For the COFF
9817 targeted toolchain the default value is 8. A value of 64 is only allowed
9818 if the underlying ABI supports it.
9820 Specifying the larger number can produce faster, more efficient code, but
9821 can also increase the size of the program. Different values are potentially
9822 incompatible. Code compiled with one value cannot necessarily expect to
9823 work with code or libraries compiled with another value, if they exchange
9824 information using structures or unions.
9826 @item -mabort-on-noreturn
9827 @opindex mabort-on-noreturn
9828 Generate a call to the function @code{abort} at the end of a
9829 @code{noreturn} function. It will be executed if the function tries to
9833 @itemx -mno-long-calls
9834 @opindex mlong-calls
9835 @opindex mno-long-calls
9836 Tells the compiler to perform function calls by first loading the
9837 address of the function into a register and then performing a subroutine
9838 call on this register. This switch is needed if the target function
9839 will lie outside of the 64 megabyte addressing range of the offset based
9840 version of subroutine call instruction.
9842 Even if this switch is enabled, not all function calls will be turned
9843 into long calls. The heuristic is that static functions, functions
9844 which have the @samp{short-call} attribute, functions that are inside
9845 the scope of a @samp{#pragma no_long_calls} directive and functions whose
9846 definitions have already been compiled within the current compilation
9847 unit, will not be turned into long calls. The exception to this rule is
9848 that weak function definitions, functions with the @samp{long-call}
9849 attribute or the @samp{section} attribute, and functions that are within
9850 the scope of a @samp{#pragma long_calls} directive, will always be
9851 turned into long calls.
9853 This feature is not enabled by default. Specifying
9854 @option{-mno-long-calls} will restore the default behavior, as will
9855 placing the function calls within the scope of a @samp{#pragma
9856 long_calls_off} directive. Note these switches have no effect on how
9857 the compiler generates code to handle function calls via function
9860 @item -msingle-pic-base
9861 @opindex msingle-pic-base
9862 Treat the register used for PIC addressing as read-only, rather than
9863 loading it in the prologue for each function. The run-time system is
9864 responsible for initializing this register with an appropriate value
9865 before execution begins.
9867 @item -mpic-register=@var{reg}
9868 @opindex mpic-register
9869 Specify the register to be used for PIC addressing. The default is R10
9870 unless stack-checking is enabled, when R9 is used.
9872 @item -mcirrus-fix-invalid-insns
9873 @opindex mcirrus-fix-invalid-insns
9874 @opindex mno-cirrus-fix-invalid-insns
9875 Insert NOPs into the instruction stream to in order to work around
9876 problems with invalid Maverick instruction combinations. This option
9877 is only valid if the @option{-mcpu=ep9312} option has been used to
9878 enable generation of instructions for the Cirrus Maverick floating
9879 point co-processor. This option is not enabled by default, since the
9880 problem is only present in older Maverick implementations. The default
9881 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
9884 @item -mpoke-function-name
9885 @opindex mpoke-function-name
9886 Write the name of each function into the text section, directly
9887 preceding the function prologue. The generated code is similar to this:
9891 .ascii "arm_poke_function_name", 0
9894 .word 0xff000000 + (t1 - t0)
9895 arm_poke_function_name
9897 stmfd sp!, @{fp, ip, lr, pc@}
9901 When performing a stack backtrace, code can inspect the value of
9902 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
9903 location @code{pc - 12} and the top 8 bits are set, then we know that
9904 there is a function name embedded immediately preceding this location
9905 and has length @code{((pc[-3]) & 0xff000000)}.
9909 Generate code for the Thumb instruction set. The default is to
9910 use the 32-bit ARM instruction set.
9911 This option automatically enables either 16-bit Thumb-1 or
9912 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
9913 and @option{-march=@var{name}} options. This option is not passed to the
9914 assembler. If you want to force assembler files to be interpreted as Thumb code,
9915 either add a @samp{.thumb} directive to the source or pass the @option{-mthumb}
9916 option directly to the assembler by prefixing it with @option{-Wa}.
9919 @opindex mtpcs-frame
9920 Generate a stack frame that is compliant with the Thumb Procedure Call
9921 Standard for all non-leaf functions. (A leaf function is one that does
9922 not call any other functions.) The default is @option{-mno-tpcs-frame}.
9924 @item -mtpcs-leaf-frame
9925 @opindex mtpcs-leaf-frame
9926 Generate a stack frame that is compliant with the Thumb Procedure Call
9927 Standard for all leaf functions. (A leaf function is one that does
9928 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
9930 @item -mcallee-super-interworking
9931 @opindex mcallee-super-interworking
9932 Gives all externally visible functions in the file being compiled an ARM
9933 instruction set header which switches to Thumb mode before executing the
9934 rest of the function. This allows these functions to be called from
9935 non-interworking code. This option is not valid in AAPCS configurations
9936 because interworking is enabled by default.
9938 @item -mcaller-super-interworking
9939 @opindex mcaller-super-interworking
9940 Allows calls via function pointers (including virtual functions) to
9941 execute correctly regardless of whether the target code has been
9942 compiled for interworking or not. There is a small overhead in the cost
9943 of executing a function pointer if this option is enabled. This option
9944 is not valid in AAPCS configurations because interworking is enabled
9947 @item -mtp=@var{name}
9949 Specify the access model for the thread local storage pointer. The valid
9950 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
9951 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
9952 (supported in the arm6k architecture), and @option{auto}, which uses the
9953 best available method for the selected processor. The default setting is
9956 @item -mword-relocations
9957 @opindex mword-relocations
9958 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
9959 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
9960 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
9966 @subsection AVR Options
9969 These options are defined for AVR implementations:
9972 @item -mmcu=@var{mcu}
9974 Specify ATMEL AVR instruction set or MCU type.
9976 Instruction set avr1 is for the minimal AVR core, not supported by the C
9977 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
9978 attiny11, attiny12, attiny15, attiny28).
9980 Instruction set avr2 (default) is for the classic AVR core with up to
9981 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
9982 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
9983 at90c8534, at90s8535).
9985 Instruction set avr3 is for the classic AVR core with up to 128K program
9986 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
9988 Instruction set avr4 is for the enhanced AVR core with up to 8K program
9989 memory space (MCU types: atmega8, atmega83, atmega85).
9991 Instruction set avr5 is for the enhanced AVR core with up to 128K program
9992 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
9993 atmega64, atmega128, at43usb355, at94k).
9995 @item -mno-interrupts
9996 @opindex mno-interrupts
9997 Generated code is not compatible with hardware interrupts.
9998 Code size will be smaller.
10000 @item -mcall-prologues
10001 @opindex mcall-prologues
10002 Functions prologues/epilogues expanded as call to appropriate
10003 subroutines. Code size will be smaller.
10006 @opindex mtiny-stack
10007 Change only the low 8 bits of the stack pointer.
10011 Assume int to be 8 bit integer. This affects the sizes of all types: A
10012 char will be 1 byte, an int will be 1 byte, a long will be 2 bytes
10013 and long long will be 4 bytes. Please note that this option does not
10014 comply to the C standards, but it will provide you with smaller code
10018 @node Blackfin Options
10019 @subsection Blackfin Options
10020 @cindex Blackfin Options
10023 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
10025 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
10026 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
10027 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
10028 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
10029 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
10030 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
10031 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
10033 The optional @var{sirevision} specifies the silicon revision of the target
10034 Blackfin processor. Any workarounds available for the targeted silicon revision
10035 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
10036 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
10037 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
10038 hexadecimal digits representing the major and minor numbers in the silicon
10039 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
10040 is not defined. If @var{sirevision} is @samp{any}, the
10041 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
10042 If this optional @var{sirevision} is not used, GCC assumes the latest known
10043 silicon revision of the targeted Blackfin processor.
10045 Support for @samp{bf561} is incomplete. For @samp{bf561},
10046 Only the processor macro is defined.
10047 Without this option, @samp{bf532} is used as the processor by default.
10048 The corresponding predefined processor macros for @var{cpu} is to
10049 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
10050 provided by libgloss to be linked in if @option{-msim} is not given.
10054 Specifies that the program will be run on the simulator. This causes
10055 the simulator BSP provided by libgloss to be linked in. This option
10056 has effect only for @samp{bfin-elf} toolchain.
10057 Certain other options, such as @option{-mid-shared-library} and
10058 @option{-mfdpic}, imply @option{-msim}.
10060 @item -momit-leaf-frame-pointer
10061 @opindex momit-leaf-frame-pointer
10062 Don't keep the frame pointer in a register for leaf functions. This
10063 avoids the instructions to save, set up and restore frame pointers and
10064 makes an extra register available in leaf functions. The option
10065 @option{-fomit-frame-pointer} removes the frame pointer for all functions
10066 which might make debugging harder.
10068 @item -mspecld-anomaly
10069 @opindex mspecld-anomaly
10070 When enabled, the compiler will ensure that the generated code does not
10071 contain speculative loads after jump instructions. If this option is used,
10072 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
10074 @item -mno-specld-anomaly
10075 @opindex mno-specld-anomaly
10076 Don't generate extra code to prevent speculative loads from occurring.
10078 @item -mcsync-anomaly
10079 @opindex mcsync-anomaly
10080 When enabled, the compiler will ensure that the generated code does not
10081 contain CSYNC or SSYNC instructions too soon after conditional branches.
10082 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
10084 @item -mno-csync-anomaly
10085 @opindex mno-csync-anomaly
10086 Don't generate extra code to prevent CSYNC or SSYNC instructions from
10087 occurring too soon after a conditional branch.
10091 When enabled, the compiler is free to take advantage of the knowledge that
10092 the entire program fits into the low 64k of memory.
10095 @opindex mno-low-64k
10096 Assume that the program is arbitrarily large. This is the default.
10098 @item -mstack-check-l1
10099 @opindex mstack-check-l1
10100 Do stack checking using information placed into L1 scratchpad memory by the
10103 @item -mid-shared-library
10104 @opindex mid-shared-library
10105 Generate code that supports shared libraries via the library ID method.
10106 This allows for execute in place and shared libraries in an environment
10107 without virtual memory management. This option implies @option{-fPIC}.
10108 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10110 @item -mno-id-shared-library
10111 @opindex mno-id-shared-library
10112 Generate code that doesn't assume ID based shared libraries are being used.
10113 This is the default.
10115 @item -mleaf-id-shared-library
10116 @opindex mleaf-id-shared-library
10117 Generate code that supports shared libraries via the library ID method,
10118 but assumes that this library or executable won't link against any other
10119 ID shared libraries. That allows the compiler to use faster code for jumps
10122 @item -mno-leaf-id-shared-library
10123 @opindex mno-leaf-id-shared-library
10124 Do not assume that the code being compiled won't link against any ID shared
10125 libraries. Slower code will be generated for jump and call insns.
10127 @item -mshared-library-id=n
10128 @opindex mshared-library-id
10129 Specified the identification number of the ID based shared library being
10130 compiled. Specifying a value of 0 will generate more compact code, specifying
10131 other values will force the allocation of that number to the current
10132 library but is no more space or time efficient than omitting this option.
10136 Generate code that allows the data segment to be located in a different
10137 area of memory from the text segment. This allows for execute in place in
10138 an environment without virtual memory management by eliminating relocations
10139 against the text section.
10141 @item -mno-sep-data
10142 @opindex mno-sep-data
10143 Generate code that assumes that the data segment follows the text segment.
10144 This is the default.
10147 @itemx -mno-long-calls
10148 @opindex mlong-calls
10149 @opindex mno-long-calls
10150 Tells the compiler to perform function calls by first loading the
10151 address of the function into a register and then performing a subroutine
10152 call on this register. This switch is needed if the target function
10153 will lie outside of the 24 bit addressing range of the offset based
10154 version of subroutine call instruction.
10156 This feature is not enabled by default. Specifying
10157 @option{-mno-long-calls} will restore the default behavior. Note these
10158 switches have no effect on how the compiler generates code to handle
10159 function calls via function pointers.
10163 Link with the fast floating-point library. This library relaxes some of
10164 the IEEE floating-point standard's rules for checking inputs against
10165 Not-a-Number (NAN), in the interest of performance.
10168 @opindex minline-plt
10169 Enable inlining of PLT entries in function calls to functions that are
10170 not known to bind locally. It has no effect without @option{-mfdpic}.
10173 @opindex mmulticore
10174 Build standalone application for multicore Blackfin processor. Proper
10175 start files and link scripts will be used to support multicore.
10176 This option defines @code{__BFIN_MULTICORE}. It can only be used with
10177 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
10178 @option{-mcorea} or @option{-mcoreb}. If it's used without
10179 @option{-mcorea} or @option{-mcoreb}, single application/dual core
10180 programming model is used. In this model, the main function of Core B
10181 should be named as coreb_main. If it's used with @option{-mcorea} or
10182 @option{-mcoreb}, one application per core programming model is used.
10183 If this option is not used, single core application programming
10188 Build standalone application for Core A of BF561 when using
10189 one application per core programming model. Proper start files
10190 and link scripts will be used to support Core A. This option
10191 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
10195 Build standalone application for Core B of BF561 when using
10196 one application per core programming model. Proper start files
10197 and link scripts will be used to support Core B. This option
10198 defines @code{__BFIN_COREB}. When this option is used, coreb_main
10199 should be used instead of main. It must be used with
10200 @option{-mmulticore}.
10204 Build standalone application for SDRAM. Proper start files and
10205 link scripts will be used to put the application into SDRAM.
10206 Loader should initialize SDRAM before loading the application
10207 into SDRAM. This option defines @code{__BFIN_SDRAM}.
10211 Assume that ICPLBs are enabled at runtime. This has an effect on certain
10212 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
10213 are enabled; for standalone applications the default is off.
10217 @subsection CRIS Options
10218 @cindex CRIS Options
10220 These options are defined specifically for the CRIS ports.
10223 @item -march=@var{architecture-type}
10224 @itemx -mcpu=@var{architecture-type}
10227 Generate code for the specified architecture. The choices for
10228 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
10229 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
10230 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
10233 @item -mtune=@var{architecture-type}
10235 Tune to @var{architecture-type} everything applicable about the generated
10236 code, except for the ABI and the set of available instructions. The
10237 choices for @var{architecture-type} are the same as for
10238 @option{-march=@var{architecture-type}}.
10240 @item -mmax-stack-frame=@var{n}
10241 @opindex mmax-stack-frame
10242 Warn when the stack frame of a function exceeds @var{n} bytes.
10248 The options @option{-metrax4} and @option{-metrax100} are synonyms for
10249 @option{-march=v3} and @option{-march=v8} respectively.
10251 @item -mmul-bug-workaround
10252 @itemx -mno-mul-bug-workaround
10253 @opindex mmul-bug-workaround
10254 @opindex mno-mul-bug-workaround
10255 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
10256 models where it applies. This option is active by default.
10260 Enable CRIS-specific verbose debug-related information in the assembly
10261 code. This option also has the effect to turn off the @samp{#NO_APP}
10262 formatted-code indicator to the assembler at the beginning of the
10267 Do not use condition-code results from previous instruction; always emit
10268 compare and test instructions before use of condition codes.
10270 @item -mno-side-effects
10271 @opindex mno-side-effects
10272 Do not emit instructions with side-effects in addressing modes other than
10275 @item -mstack-align
10276 @itemx -mno-stack-align
10277 @itemx -mdata-align
10278 @itemx -mno-data-align
10279 @itemx -mconst-align
10280 @itemx -mno-const-align
10281 @opindex mstack-align
10282 @opindex mno-stack-align
10283 @opindex mdata-align
10284 @opindex mno-data-align
10285 @opindex mconst-align
10286 @opindex mno-const-align
10287 These options (no-options) arranges (eliminate arrangements) for the
10288 stack-frame, individual data and constants to be aligned for the maximum
10289 single data access size for the chosen CPU model. The default is to
10290 arrange for 32-bit alignment. ABI details such as structure layout are
10291 not affected by these options.
10299 Similar to the stack- data- and const-align options above, these options
10300 arrange for stack-frame, writable data and constants to all be 32-bit,
10301 16-bit or 8-bit aligned. The default is 32-bit alignment.
10303 @item -mno-prologue-epilogue
10304 @itemx -mprologue-epilogue
10305 @opindex mno-prologue-epilogue
10306 @opindex mprologue-epilogue
10307 With @option{-mno-prologue-epilogue}, the normal function prologue and
10308 epilogue that sets up the stack-frame are omitted and no return
10309 instructions or return sequences are generated in the code. Use this
10310 option only together with visual inspection of the compiled code: no
10311 warnings or errors are generated when call-saved registers must be saved,
10312 or storage for local variable needs to be allocated.
10316 @opindex mno-gotplt
10318 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
10319 instruction sequences that load addresses for functions from the PLT part
10320 of the GOT rather than (traditional on other architectures) calls to the
10321 PLT@. The default is @option{-mgotplt}.
10325 Legacy no-op option only recognized with the cris-axis-elf and
10326 cris-axis-linux-gnu targets.
10330 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
10334 This option, recognized for the cris-axis-elf arranges
10335 to link with input-output functions from a simulator library. Code,
10336 initialized data and zero-initialized data are allocated consecutively.
10340 Like @option{-sim}, but pass linker options to locate initialized data at
10341 0x40000000 and zero-initialized data at 0x80000000.
10345 @subsection CRX Options
10346 @cindex CRX Options
10348 These options are defined specifically for the CRX ports.
10354 Enable the use of multiply-accumulate instructions. Disabled by default.
10357 @opindex mpush-args
10358 Push instructions will be used to pass outgoing arguments when functions
10359 are called. Enabled by default.
10362 @node Darwin Options
10363 @subsection Darwin Options
10364 @cindex Darwin options
10366 These options are defined for all architectures running the Darwin operating
10369 FSF GCC on Darwin does not create ``fat'' object files; it will create
10370 an object file for the single architecture that it was built to
10371 target. Apple's GCC on Darwin does create ``fat'' files if multiple
10372 @option{-arch} options are used; it does so by running the compiler or
10373 linker multiple times and joining the results together with
10376 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
10377 @samp{i686}) is determined by the flags that specify the ISA
10378 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
10379 @option{-force_cpusubtype_ALL} option can be used to override this.
10381 The Darwin tools vary in their behavior when presented with an ISA
10382 mismatch. The assembler, @file{as}, will only permit instructions to
10383 be used that are valid for the subtype of the file it is generating,
10384 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
10385 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
10386 and print an error if asked to create a shared library with a less
10387 restrictive subtype than its input files (for instance, trying to put
10388 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
10389 for executables, @file{ld}, will quietly give the executable the most
10390 restrictive subtype of any of its input files.
10395 Add the framework directory @var{dir} to the head of the list of
10396 directories to be searched for header files. These directories are
10397 interleaved with those specified by @option{-I} options and are
10398 scanned in a left-to-right order.
10400 A framework directory is a directory with frameworks in it. A
10401 framework is a directory with a @samp{"Headers"} and/or
10402 @samp{"PrivateHeaders"} directory contained directly in it that ends
10403 in @samp{".framework"}. The name of a framework is the name of this
10404 directory excluding the @samp{".framework"}. Headers associated with
10405 the framework are found in one of those two directories, with
10406 @samp{"Headers"} being searched first. A subframework is a framework
10407 directory that is in a framework's @samp{"Frameworks"} directory.
10408 Includes of subframework headers can only appear in a header of a
10409 framework that contains the subframework, or in a sibling subframework
10410 header. Two subframeworks are siblings if they occur in the same
10411 framework. A subframework should not have the same name as a
10412 framework, a warning will be issued if this is violated. Currently a
10413 subframework cannot have subframeworks, in the future, the mechanism
10414 may be extended to support this. The standard frameworks can be found
10415 in @samp{"/System/Library/Frameworks"} and
10416 @samp{"/Library/Frameworks"}. An example include looks like
10417 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
10418 the name of the framework and header.h is found in the
10419 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
10421 @item -iframework@var{dir}
10422 @opindex iframework
10423 Like @option{-F} except the directory is a treated as a system
10424 directory. The main difference between this @option{-iframework} and
10425 @option{-F} is that with @option{-iframework} the compiler does not
10426 warn about constructs contained within header files found via
10427 @var{dir}. This option is valid only for the C family of languages.
10431 Emit debugging information for symbols that are used. For STABS
10432 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
10433 This is by default ON@.
10437 Emit debugging information for all symbols and types.
10439 @item -mmacosx-version-min=@var{version}
10440 The earliest version of MacOS X that this executable will run on
10441 is @var{version}. Typical values of @var{version} include @code{10.1},
10442 @code{10.2}, and @code{10.3.9}.
10444 If the compiler was built to use the system's headers by default,
10445 then the default for this option is the system version on which the
10446 compiler is running, otherwise the default is to make choices which
10447 are compatible with as many systems and code bases as possible.
10451 Enable kernel development mode. The @option{-mkernel} option sets
10452 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
10453 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
10454 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
10455 applicable. This mode also sets @option{-mno-altivec},
10456 @option{-msoft-float}, @option{-fno-builtin} and
10457 @option{-mlong-branch} for PowerPC targets.
10459 @item -mone-byte-bool
10460 @opindex mone-byte-bool
10461 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
10462 By default @samp{sizeof(bool)} is @samp{4} when compiling for
10463 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
10464 option has no effect on x86.
10466 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
10467 to generate code that is not binary compatible with code generated
10468 without that switch. Using this switch may require recompiling all
10469 other modules in a program, including system libraries. Use this
10470 switch to conform to a non-default data model.
10472 @item -mfix-and-continue
10473 @itemx -ffix-and-continue
10474 @itemx -findirect-data
10475 @opindex mfix-and-continue
10476 @opindex ffix-and-continue
10477 @opindex findirect-data
10478 Generate code suitable for fast turn around development. Needed to
10479 enable gdb to dynamically load @code{.o} files into already running
10480 programs. @option{-findirect-data} and @option{-ffix-and-continue}
10481 are provided for backwards compatibility.
10485 Loads all members of static archive libraries.
10486 See man ld(1) for more information.
10488 @item -arch_errors_fatal
10489 @opindex arch_errors_fatal
10490 Cause the errors having to do with files that have the wrong architecture
10493 @item -bind_at_load
10494 @opindex bind_at_load
10495 Causes the output file to be marked such that the dynamic linker will
10496 bind all undefined references when the file is loaded or launched.
10500 Produce a Mach-o bundle format file.
10501 See man ld(1) for more information.
10503 @item -bundle_loader @var{executable}
10504 @opindex bundle_loader
10505 This option specifies the @var{executable} that will be loading the build
10506 output file being linked. See man ld(1) for more information.
10509 @opindex dynamiclib
10510 When passed this option, GCC will produce a dynamic library instead of
10511 an executable when linking, using the Darwin @file{libtool} command.
10513 @item -force_cpusubtype_ALL
10514 @opindex force_cpusubtype_ALL
10515 This causes GCC's output file to have the @var{ALL} subtype, instead of
10516 one controlled by the @option{-mcpu} or @option{-march} option.
10518 @item -allowable_client @var{client_name}
10519 @itemx -client_name
10520 @itemx -compatibility_version
10521 @itemx -current_version
10523 @itemx -dependency-file
10525 @itemx -dylinker_install_name
10527 @itemx -exported_symbols_list
10529 @itemx -flat_namespace
10530 @itemx -force_flat_namespace
10531 @itemx -headerpad_max_install_names
10534 @itemx -install_name
10535 @itemx -keep_private_externs
10536 @itemx -multi_module
10537 @itemx -multiply_defined
10538 @itemx -multiply_defined_unused
10540 @itemx -no_dead_strip_inits_and_terms
10541 @itemx -nofixprebinding
10542 @itemx -nomultidefs
10544 @itemx -noseglinkedit
10545 @itemx -pagezero_size
10547 @itemx -prebind_all_twolevel_modules
10548 @itemx -private_bundle
10549 @itemx -read_only_relocs
10551 @itemx -sectobjectsymbols
10555 @itemx -sectobjectsymbols
10558 @itemx -segs_read_only_addr
10559 @itemx -segs_read_write_addr
10560 @itemx -seg_addr_table
10561 @itemx -seg_addr_table_filename
10562 @itemx -seglinkedit
10564 @itemx -segs_read_only_addr
10565 @itemx -segs_read_write_addr
10566 @itemx -single_module
10568 @itemx -sub_library
10569 @itemx -sub_umbrella
10570 @itemx -twolevel_namespace
10573 @itemx -unexported_symbols_list
10574 @itemx -weak_reference_mismatches
10575 @itemx -whatsloaded
10576 @opindex allowable_client
10577 @opindex client_name
10578 @opindex compatibility_version
10579 @opindex current_version
10580 @opindex dead_strip
10581 @opindex dependency-file
10582 @opindex dylib_file
10583 @opindex dylinker_install_name
10585 @opindex exported_symbols_list
10587 @opindex flat_namespace
10588 @opindex force_flat_namespace
10589 @opindex headerpad_max_install_names
10590 @opindex image_base
10592 @opindex install_name
10593 @opindex keep_private_externs
10594 @opindex multi_module
10595 @opindex multiply_defined
10596 @opindex multiply_defined_unused
10597 @opindex noall_load
10598 @opindex no_dead_strip_inits_and_terms
10599 @opindex nofixprebinding
10600 @opindex nomultidefs
10602 @opindex noseglinkedit
10603 @opindex pagezero_size
10605 @opindex prebind_all_twolevel_modules
10606 @opindex private_bundle
10607 @opindex read_only_relocs
10609 @opindex sectobjectsymbols
10612 @opindex sectcreate
10613 @opindex sectobjectsymbols
10616 @opindex segs_read_only_addr
10617 @opindex segs_read_write_addr
10618 @opindex seg_addr_table
10619 @opindex seg_addr_table_filename
10620 @opindex seglinkedit
10622 @opindex segs_read_only_addr
10623 @opindex segs_read_write_addr
10624 @opindex single_module
10626 @opindex sub_library
10627 @opindex sub_umbrella
10628 @opindex twolevel_namespace
10631 @opindex unexported_symbols_list
10632 @opindex weak_reference_mismatches
10633 @opindex whatsloaded
10634 These options are passed to the Darwin linker. The Darwin linker man page
10635 describes them in detail.
10638 @node DEC Alpha Options
10639 @subsection DEC Alpha Options
10641 These @samp{-m} options are defined for the DEC Alpha implementations:
10644 @item -mno-soft-float
10645 @itemx -msoft-float
10646 @opindex mno-soft-float
10647 @opindex msoft-float
10648 Use (do not use) the hardware floating-point instructions for
10649 floating-point operations. When @option{-msoft-float} is specified,
10650 functions in @file{libgcc.a} will be used to perform floating-point
10651 operations. Unless they are replaced by routines that emulate the
10652 floating-point operations, or compiled in such a way as to call such
10653 emulations routines, these routines will issue floating-point
10654 operations. If you are compiling for an Alpha without floating-point
10655 operations, you must ensure that the library is built so as not to call
10658 Note that Alpha implementations without floating-point operations are
10659 required to have floating-point registers.
10662 @itemx -mno-fp-regs
10664 @opindex mno-fp-regs
10665 Generate code that uses (does not use) the floating-point register set.
10666 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
10667 register set is not used, floating point operands are passed in integer
10668 registers as if they were integers and floating-point results are passed
10669 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
10670 so any function with a floating-point argument or return value called by code
10671 compiled with @option{-mno-fp-regs} must also be compiled with that
10674 A typical use of this option is building a kernel that does not use,
10675 and hence need not save and restore, any floating-point registers.
10679 The Alpha architecture implements floating-point hardware optimized for
10680 maximum performance. It is mostly compliant with the IEEE floating
10681 point standard. However, for full compliance, software assistance is
10682 required. This option generates code fully IEEE compliant code
10683 @emph{except} that the @var{inexact-flag} is not maintained (see below).
10684 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
10685 defined during compilation. The resulting code is less efficient but is
10686 able to correctly support denormalized numbers and exceptional IEEE
10687 values such as not-a-number and plus/minus infinity. Other Alpha
10688 compilers call this option @option{-ieee_with_no_inexact}.
10690 @item -mieee-with-inexact
10691 @opindex mieee-with-inexact
10692 This is like @option{-mieee} except the generated code also maintains
10693 the IEEE @var{inexact-flag}. Turning on this option causes the
10694 generated code to implement fully-compliant IEEE math. In addition to
10695 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
10696 macro. On some Alpha implementations the resulting code may execute
10697 significantly slower than the code generated by default. Since there is
10698 very little code that depends on the @var{inexact-flag}, you should
10699 normally not specify this option. Other Alpha compilers call this
10700 option @option{-ieee_with_inexact}.
10702 @item -mfp-trap-mode=@var{trap-mode}
10703 @opindex mfp-trap-mode
10704 This option controls what floating-point related traps are enabled.
10705 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
10706 The trap mode can be set to one of four values:
10710 This is the default (normal) setting. The only traps that are enabled
10711 are the ones that cannot be disabled in software (e.g., division by zero
10715 In addition to the traps enabled by @samp{n}, underflow traps are enabled
10719 Like @samp{u}, but the instructions are marked to be safe for software
10720 completion (see Alpha architecture manual for details).
10723 Like @samp{su}, but inexact traps are enabled as well.
10726 @item -mfp-rounding-mode=@var{rounding-mode}
10727 @opindex mfp-rounding-mode
10728 Selects the IEEE rounding mode. Other Alpha compilers call this option
10729 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
10734 Normal IEEE rounding mode. Floating point numbers are rounded towards
10735 the nearest machine number or towards the even machine number in case
10739 Round towards minus infinity.
10742 Chopped rounding mode. Floating point numbers are rounded towards zero.
10745 Dynamic rounding mode. A field in the floating point control register
10746 (@var{fpcr}, see Alpha architecture reference manual) controls the
10747 rounding mode in effect. The C library initializes this register for
10748 rounding towards plus infinity. Thus, unless your program modifies the
10749 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
10752 @item -mtrap-precision=@var{trap-precision}
10753 @opindex mtrap-precision
10754 In the Alpha architecture, floating point traps are imprecise. This
10755 means without software assistance it is impossible to recover from a
10756 floating trap and program execution normally needs to be terminated.
10757 GCC can generate code that can assist operating system trap handlers
10758 in determining the exact location that caused a floating point trap.
10759 Depending on the requirements of an application, different levels of
10760 precisions can be selected:
10764 Program precision. This option is the default and means a trap handler
10765 can only identify which program caused a floating point exception.
10768 Function precision. The trap handler can determine the function that
10769 caused a floating point exception.
10772 Instruction precision. The trap handler can determine the exact
10773 instruction that caused a floating point exception.
10776 Other Alpha compilers provide the equivalent options called
10777 @option{-scope_safe} and @option{-resumption_safe}.
10779 @item -mieee-conformant
10780 @opindex mieee-conformant
10781 This option marks the generated code as IEEE conformant. You must not
10782 use this option unless you also specify @option{-mtrap-precision=i} and either
10783 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
10784 is to emit the line @samp{.eflag 48} in the function prologue of the
10785 generated assembly file. Under DEC Unix, this has the effect that
10786 IEEE-conformant math library routines will be linked in.
10788 @item -mbuild-constants
10789 @opindex mbuild-constants
10790 Normally GCC examines a 32- or 64-bit integer constant to
10791 see if it can construct it from smaller constants in two or three
10792 instructions. If it cannot, it will output the constant as a literal and
10793 generate code to load it from the data segment at runtime.
10795 Use this option to require GCC to construct @emph{all} integer constants
10796 using code, even if it takes more instructions (the maximum is six).
10798 You would typically use this option to build a shared library dynamic
10799 loader. Itself a shared library, it must relocate itself in memory
10800 before it can find the variables and constants in its own data segment.
10806 Select whether to generate code to be assembled by the vendor-supplied
10807 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
10825 Indicate whether GCC should generate code to use the optional BWX,
10826 CIX, FIX and MAX instruction sets. The default is to use the instruction
10827 sets supported by the CPU type specified via @option{-mcpu=} option or that
10828 of the CPU on which GCC was built if none was specified.
10831 @itemx -mfloat-ieee
10832 @opindex mfloat-vax
10833 @opindex mfloat-ieee
10834 Generate code that uses (does not use) VAX F and G floating point
10835 arithmetic instead of IEEE single and double precision.
10837 @item -mexplicit-relocs
10838 @itemx -mno-explicit-relocs
10839 @opindex mexplicit-relocs
10840 @opindex mno-explicit-relocs
10841 Older Alpha assemblers provided no way to generate symbol relocations
10842 except via assembler macros. Use of these macros does not allow
10843 optimal instruction scheduling. GNU binutils as of version 2.12
10844 supports a new syntax that allows the compiler to explicitly mark
10845 which relocations should apply to which instructions. This option
10846 is mostly useful for debugging, as GCC detects the capabilities of
10847 the assembler when it is built and sets the default accordingly.
10850 @itemx -mlarge-data
10851 @opindex msmall-data
10852 @opindex mlarge-data
10853 When @option{-mexplicit-relocs} is in effect, static data is
10854 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
10855 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
10856 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
10857 16-bit relocations off of the @code{$gp} register. This limits the
10858 size of the small data area to 64KB, but allows the variables to be
10859 directly accessed via a single instruction.
10861 The default is @option{-mlarge-data}. With this option the data area
10862 is limited to just below 2GB@. Programs that require more than 2GB of
10863 data must use @code{malloc} or @code{mmap} to allocate the data in the
10864 heap instead of in the program's data segment.
10866 When generating code for shared libraries, @option{-fpic} implies
10867 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
10870 @itemx -mlarge-text
10871 @opindex msmall-text
10872 @opindex mlarge-text
10873 When @option{-msmall-text} is used, the compiler assumes that the
10874 code of the entire program (or shared library) fits in 4MB, and is
10875 thus reachable with a branch instruction. When @option{-msmall-data}
10876 is used, the compiler can assume that all local symbols share the
10877 same @code{$gp} value, and thus reduce the number of instructions
10878 required for a function call from 4 to 1.
10880 The default is @option{-mlarge-text}.
10882 @item -mcpu=@var{cpu_type}
10884 Set the instruction set and instruction scheduling parameters for
10885 machine type @var{cpu_type}. You can specify either the @samp{EV}
10886 style name or the corresponding chip number. GCC supports scheduling
10887 parameters for the EV4, EV5 and EV6 family of processors and will
10888 choose the default values for the instruction set from the processor
10889 you specify. If you do not specify a processor type, GCC will default
10890 to the processor on which the compiler was built.
10892 Supported values for @var{cpu_type} are
10898 Schedules as an EV4 and has no instruction set extensions.
10902 Schedules as an EV5 and has no instruction set extensions.
10906 Schedules as an EV5 and supports the BWX extension.
10911 Schedules as an EV5 and supports the BWX and MAX extensions.
10915 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
10919 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
10922 Native Linux/GNU toolchains also support the value @samp{native},
10923 which selects the best architecture option for the host processor.
10924 @option{-mcpu=native} has no effect if GCC does not recognize
10927 @item -mtune=@var{cpu_type}
10929 Set only the instruction scheduling parameters for machine type
10930 @var{cpu_type}. The instruction set is not changed.
10932 Native Linux/GNU toolchains also support the value @samp{native},
10933 which selects the best architecture option for the host processor.
10934 @option{-mtune=native} has no effect if GCC does not recognize
10937 @item -mmemory-latency=@var{time}
10938 @opindex mmemory-latency
10939 Sets the latency the scheduler should assume for typical memory
10940 references as seen by the application. This number is highly
10941 dependent on the memory access patterns used by the application
10942 and the size of the external cache on the machine.
10944 Valid options for @var{time} are
10948 A decimal number representing clock cycles.
10954 The compiler contains estimates of the number of clock cycles for
10955 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
10956 (also called Dcache, Scache, and Bcache), as well as to main memory.
10957 Note that L3 is only valid for EV5.
10962 @node DEC Alpha/VMS Options
10963 @subsection DEC Alpha/VMS Options
10965 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
10968 @item -mvms-return-codes
10969 @opindex mvms-return-codes
10970 Return VMS condition codes from main. The default is to return POSIX
10971 style condition (e.g.@: error) codes.
10973 @item -mdebug-main=@var{prefix}
10974 @opindex mdebug-main=@var{prefix}
10975 Flag the first routine whose name starts with @var{prefix} as the main
10976 routine for the debugger.
10980 Default to 64bit memory allocation routines.
10984 @subsection FR30 Options
10985 @cindex FR30 Options
10987 These options are defined specifically for the FR30 port.
10991 @item -msmall-model
10992 @opindex msmall-model
10993 Use the small address space model. This can produce smaller code, but
10994 it does assume that all symbolic values and addresses will fit into a
10999 Assume that run-time support has been provided and so there is no need
11000 to include the simulator library (@file{libsim.a}) on the linker
11006 @subsection FRV Options
11007 @cindex FRV Options
11013 Only use the first 32 general purpose registers.
11018 Use all 64 general purpose registers.
11023 Use only the first 32 floating point registers.
11028 Use all 64 floating point registers
11031 @opindex mhard-float
11033 Use hardware instructions for floating point operations.
11036 @opindex msoft-float
11038 Use library routines for floating point operations.
11043 Dynamically allocate condition code registers.
11048 Do not try to dynamically allocate condition code registers, only
11049 use @code{icc0} and @code{fcc0}.
11054 Change ABI to use double word insns.
11059 Do not use double word instructions.
11064 Use floating point double instructions.
11067 @opindex mno-double
11069 Do not use floating point double instructions.
11074 Use media instructions.
11079 Do not use media instructions.
11084 Use multiply and add/subtract instructions.
11087 @opindex mno-muladd
11089 Do not use multiply and add/subtract instructions.
11094 Select the FDPIC ABI, that uses function descriptors to represent
11095 pointers to functions. Without any PIC/PIE-related options, it
11096 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
11097 assumes GOT entries and small data are within a 12-bit range from the
11098 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
11099 are computed with 32 bits.
11100 With a @samp{bfin-elf} target, this option implies @option{-msim}.
11103 @opindex minline-plt
11105 Enable inlining of PLT entries in function calls to functions that are
11106 not known to bind locally. It has no effect without @option{-mfdpic}.
11107 It's enabled by default if optimizing for speed and compiling for
11108 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
11109 optimization option such as @option{-O3} or above is present in the
11115 Assume a large TLS segment when generating thread-local code.
11120 Do not assume a large TLS segment when generating thread-local code.
11125 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
11126 that is known to be in read-only sections. It's enabled by default,
11127 except for @option{-fpic} or @option{-fpie}: even though it may help
11128 make the global offset table smaller, it trades 1 instruction for 4.
11129 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
11130 one of which may be shared by multiple symbols, and it avoids the need
11131 for a GOT entry for the referenced symbol, so it's more likely to be a
11132 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
11134 @item -multilib-library-pic
11135 @opindex multilib-library-pic
11137 Link with the (library, not FD) pic libraries. It's implied by
11138 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
11139 @option{-fpic} without @option{-mfdpic}. You should never have to use
11143 @opindex mlinked-fp
11145 Follow the EABI requirement of always creating a frame pointer whenever
11146 a stack frame is allocated. This option is enabled by default and can
11147 be disabled with @option{-mno-linked-fp}.
11150 @opindex mlong-calls
11152 Use indirect addressing to call functions outside the current
11153 compilation unit. This allows the functions to be placed anywhere
11154 within the 32-bit address space.
11156 @item -malign-labels
11157 @opindex malign-labels
11159 Try to align labels to an 8-byte boundary by inserting nops into the
11160 previous packet. This option only has an effect when VLIW packing
11161 is enabled. It doesn't create new packets; it merely adds nops to
11164 @item -mlibrary-pic
11165 @opindex mlibrary-pic
11167 Generate position-independent EABI code.
11172 Use only the first four media accumulator registers.
11177 Use all eight media accumulator registers.
11182 Pack VLIW instructions.
11187 Do not pack VLIW instructions.
11190 @opindex mno-eflags
11192 Do not mark ABI switches in e_flags.
11195 @opindex mcond-move
11197 Enable the use of conditional-move instructions (default).
11199 This switch is mainly for debugging the compiler and will likely be removed
11200 in a future version.
11202 @item -mno-cond-move
11203 @opindex mno-cond-move
11205 Disable the use of conditional-move instructions.
11207 This switch is mainly for debugging the compiler and will likely be removed
11208 in a future version.
11213 Enable the use of conditional set instructions (default).
11215 This switch is mainly for debugging the compiler and will likely be removed
11216 in a future version.
11221 Disable the use of conditional set instructions.
11223 This switch is mainly for debugging the compiler and will likely be removed
11224 in a future version.
11227 @opindex mcond-exec
11229 Enable the use of conditional execution (default).
11231 This switch is mainly for debugging the compiler and will likely be removed
11232 in a future version.
11234 @item -mno-cond-exec
11235 @opindex mno-cond-exec
11237 Disable the use of conditional execution.
11239 This switch is mainly for debugging the compiler and will likely be removed
11240 in a future version.
11242 @item -mvliw-branch
11243 @opindex mvliw-branch
11245 Run a pass to pack branches into VLIW instructions (default).
11247 This switch is mainly for debugging the compiler and will likely be removed
11248 in a future version.
11250 @item -mno-vliw-branch
11251 @opindex mno-vliw-branch
11253 Do not run a pass to pack branches into VLIW instructions.
11255 This switch is mainly for debugging the compiler and will likely be removed
11256 in a future version.
11258 @item -mmulti-cond-exec
11259 @opindex mmulti-cond-exec
11261 Enable optimization of @code{&&} and @code{||} in conditional execution
11264 This switch is mainly for debugging the compiler and will likely be removed
11265 in a future version.
11267 @item -mno-multi-cond-exec
11268 @opindex mno-multi-cond-exec
11270 Disable optimization of @code{&&} and @code{||} in conditional execution.
11272 This switch is mainly for debugging the compiler and will likely be removed
11273 in a future version.
11275 @item -mnested-cond-exec
11276 @opindex mnested-cond-exec
11278 Enable nested conditional execution optimizations (default).
11280 This switch is mainly for debugging the compiler and will likely be removed
11281 in a future version.
11283 @item -mno-nested-cond-exec
11284 @opindex mno-nested-cond-exec
11286 Disable nested conditional execution optimizations.
11288 This switch is mainly for debugging the compiler and will likely be removed
11289 in a future version.
11291 @item -moptimize-membar
11292 @opindex moptimize-membar
11294 This switch removes redundant @code{membar} instructions from the
11295 compiler generated code. It is enabled by default.
11297 @item -mno-optimize-membar
11298 @opindex mno-optimize-membar
11300 This switch disables the automatic removal of redundant @code{membar}
11301 instructions from the generated code.
11303 @item -mtomcat-stats
11304 @opindex mtomcat-stats
11306 Cause gas to print out tomcat statistics.
11308 @item -mcpu=@var{cpu}
11311 Select the processor type for which to generate code. Possible values are
11312 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
11313 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
11317 @node GNU/Linux Options
11318 @subsection GNU/Linux Options
11320 These @samp{-m} options are defined for GNU/Linux targets:
11325 Use the GNU C library instead of uClibc. This is the default except
11326 on @samp{*-*-linux-*uclibc*} targets.
11330 Use uClibc instead of the GNU C library. This is the default on
11331 @samp{*-*-linux-*uclibc*} targets.
11334 @node H8/300 Options
11335 @subsection H8/300 Options
11337 These @samp{-m} options are defined for the H8/300 implementations:
11342 Shorten some address references at link time, when possible; uses the
11343 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
11344 ld, Using ld}, for a fuller description.
11348 Generate code for the H8/300H@.
11352 Generate code for the H8S@.
11356 Generate code for the H8S and H8/300H in the normal mode. This switch
11357 must be used either with @option{-mh} or @option{-ms}.
11361 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
11365 Make @code{int} data 32 bits by default.
11368 @opindex malign-300
11369 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
11370 The default for the H8/300H and H8S is to align longs and floats on 4
11372 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
11373 This option has no effect on the H8/300.
11377 @subsection HPPA Options
11378 @cindex HPPA Options
11380 These @samp{-m} options are defined for the HPPA family of computers:
11383 @item -march=@var{architecture-type}
11385 Generate code for the specified architecture. The choices for
11386 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
11387 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
11388 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
11389 architecture option for your machine. Code compiled for lower numbered
11390 architectures will run on higher numbered architectures, but not the
11393 @item -mpa-risc-1-0
11394 @itemx -mpa-risc-1-1
11395 @itemx -mpa-risc-2-0
11396 @opindex mpa-risc-1-0
11397 @opindex mpa-risc-1-1
11398 @opindex mpa-risc-2-0
11399 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
11402 @opindex mbig-switch
11403 Generate code suitable for big switch tables. Use this option only if
11404 the assembler/linker complain about out of range branches within a switch
11407 @item -mjump-in-delay
11408 @opindex mjump-in-delay
11409 Fill delay slots of function calls with unconditional jump instructions
11410 by modifying the return pointer for the function call to be the target
11411 of the conditional jump.
11413 @item -mdisable-fpregs
11414 @opindex mdisable-fpregs
11415 Prevent floating point registers from being used in any manner. This is
11416 necessary for compiling kernels which perform lazy context switching of
11417 floating point registers. If you use this option and attempt to perform
11418 floating point operations, the compiler will abort.
11420 @item -mdisable-indexing
11421 @opindex mdisable-indexing
11422 Prevent the compiler from using indexing address modes. This avoids some
11423 rather obscure problems when compiling MIG generated code under MACH@.
11425 @item -mno-space-regs
11426 @opindex mno-space-regs
11427 Generate code that assumes the target has no space registers. This allows
11428 GCC to generate faster indirect calls and use unscaled index address modes.
11430 Such code is suitable for level 0 PA systems and kernels.
11432 @item -mfast-indirect-calls
11433 @opindex mfast-indirect-calls
11434 Generate code that assumes calls never cross space boundaries. This
11435 allows GCC to emit code which performs faster indirect calls.
11437 This option will not work in the presence of shared libraries or nested
11440 @item -mfixed-range=@var{register-range}
11441 @opindex mfixed-range
11442 Generate code treating the given register range as fixed registers.
11443 A fixed register is one that the register allocator can not use. This is
11444 useful when compiling kernel code. A register range is specified as
11445 two registers separated by a dash. Multiple register ranges can be
11446 specified separated by a comma.
11448 @item -mlong-load-store
11449 @opindex mlong-load-store
11450 Generate 3-instruction load and store sequences as sometimes required by
11451 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
11454 @item -mportable-runtime
11455 @opindex mportable-runtime
11456 Use the portable calling conventions proposed by HP for ELF systems.
11460 Enable the use of assembler directives only GAS understands.
11462 @item -mschedule=@var{cpu-type}
11464 Schedule code according to the constraints for the machine type
11465 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
11466 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
11467 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
11468 proper scheduling option for your machine. The default scheduling is
11472 @opindex mlinker-opt
11473 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
11474 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
11475 linkers in which they give bogus error messages when linking some programs.
11478 @opindex msoft-float
11479 Generate output containing library calls for floating point.
11480 @strong{Warning:} the requisite libraries are not available for all HPPA
11481 targets. Normally the facilities of the machine's usual C compiler are
11482 used, but this cannot be done directly in cross-compilation. You must make
11483 your own arrangements to provide suitable library functions for
11486 @option{-msoft-float} changes the calling convention in the output file;
11487 therefore, it is only useful if you compile @emph{all} of a program with
11488 this option. In particular, you need to compile @file{libgcc.a}, the
11489 library that comes with GCC, with @option{-msoft-float} in order for
11494 Generate the predefine, @code{_SIO}, for server IO@. The default is
11495 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
11496 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
11497 options are available under HP-UX and HI-UX@.
11501 Use GNU ld specific options. This passes @option{-shared} to ld when
11502 building a shared library. It is the default when GCC is configured,
11503 explicitly or implicitly, with the GNU linker. This option does not
11504 have any affect on which ld is called, it only changes what parameters
11505 are passed to that ld. The ld that is called is determined by the
11506 @option{--with-ld} configure option, GCC's program search path, and
11507 finally by the user's @env{PATH}. The linker used by GCC can be printed
11508 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
11509 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11513 Use HP ld specific options. This passes @option{-b} to ld when building
11514 a shared library and passes @option{+Accept TypeMismatch} to ld on all
11515 links. It is the default when GCC is configured, explicitly or
11516 implicitly, with the HP linker. This option does not have any affect on
11517 which ld is called, it only changes what parameters are passed to that
11518 ld. The ld that is called is determined by the @option{--with-ld}
11519 configure option, GCC's program search path, and finally by the user's
11520 @env{PATH}. The linker used by GCC can be printed using @samp{which
11521 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
11522 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11525 @opindex mno-long-calls
11526 Generate code that uses long call sequences. This ensures that a call
11527 is always able to reach linker generated stubs. The default is to generate
11528 long calls only when the distance from the call site to the beginning
11529 of the function or translation unit, as the case may be, exceeds a
11530 predefined limit set by the branch type being used. The limits for
11531 normal calls are 7,600,000 and 240,000 bytes, respectively for the
11532 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
11535 Distances are measured from the beginning of functions when using the
11536 @option{-ffunction-sections} option, or when using the @option{-mgas}
11537 and @option{-mno-portable-runtime} options together under HP-UX with
11540 It is normally not desirable to use this option as it will degrade
11541 performance. However, it may be useful in large applications,
11542 particularly when partial linking is used to build the application.
11544 The types of long calls used depends on the capabilities of the
11545 assembler and linker, and the type of code being generated. The
11546 impact on systems that support long absolute calls, and long pic
11547 symbol-difference or pc-relative calls should be relatively small.
11548 However, an indirect call is used on 32-bit ELF systems in pic code
11549 and it is quite long.
11551 @item -munix=@var{unix-std}
11553 Generate compiler predefines and select a startfile for the specified
11554 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
11555 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
11556 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
11557 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
11558 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
11561 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
11562 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
11563 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
11564 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
11565 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
11566 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
11568 It is @emph{important} to note that this option changes the interfaces
11569 for various library routines. It also affects the operational behavior
11570 of the C library. Thus, @emph{extreme} care is needed in using this
11573 Library code that is intended to operate with more than one UNIX
11574 standard must test, set and restore the variable @var{__xpg4_extended_mask}
11575 as appropriate. Most GNU software doesn't provide this capability.
11579 Suppress the generation of link options to search libdld.sl when the
11580 @option{-static} option is specified on HP-UX 10 and later.
11584 The HP-UX implementation of setlocale in libc has a dependency on
11585 libdld.sl. There isn't an archive version of libdld.sl. Thus,
11586 when the @option{-static} option is specified, special link options
11587 are needed to resolve this dependency.
11589 On HP-UX 10 and later, the GCC driver adds the necessary options to
11590 link with libdld.sl when the @option{-static} option is specified.
11591 This causes the resulting binary to be dynamic. On the 64-bit port,
11592 the linkers generate dynamic binaries by default in any case. The
11593 @option{-nolibdld} option can be used to prevent the GCC driver from
11594 adding these link options.
11598 Add support for multithreading with the @dfn{dce thread} library
11599 under HP-UX@. This option sets flags for both the preprocessor and
11603 @node i386 and x86-64 Options
11604 @subsection Intel 386 and AMD x86-64 Options
11605 @cindex i386 Options
11606 @cindex x86-64 Options
11607 @cindex Intel 386 Options
11608 @cindex AMD x86-64 Options
11610 These @samp{-m} options are defined for the i386 and x86-64 family of
11614 @item -mtune=@var{cpu-type}
11616 Tune to @var{cpu-type} everything applicable about the generated code, except
11617 for the ABI and the set of available instructions. The choices for
11618 @var{cpu-type} are:
11621 Produce code optimized for the most common IA32/AMD64/EM64T processors.
11622 If you know the CPU on which your code will run, then you should use
11623 the corresponding @option{-mtune} option instead of
11624 @option{-mtune=generic}. But, if you do not know exactly what CPU users
11625 of your application will have, then you should use this option.
11627 As new processors are deployed in the marketplace, the behavior of this
11628 option will change. Therefore, if you upgrade to a newer version of
11629 GCC, the code generated option will change to reflect the processors
11630 that were most common when that version of GCC was released.
11632 There is no @option{-march=generic} option because @option{-march}
11633 indicates the instruction set the compiler can use, and there is no
11634 generic instruction set applicable to all processors. In contrast,
11635 @option{-mtune} indicates the processor (or, in this case, collection of
11636 processors) for which the code is optimized.
11638 This selects the CPU to tune for at compilation time by determining
11639 the processor type of the compiling machine. Using @option{-mtune=native}
11640 will produce code optimized for the local machine under the constraints
11641 of the selected instruction set. Using @option{-march=native} will
11642 enable all instruction subsets supported by the local machine (hence
11643 the result might not run on different machines).
11645 Original Intel's i386 CPU@.
11647 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
11648 @item i586, pentium
11649 Intel Pentium CPU with no MMX support.
11651 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
11653 Intel PentiumPro CPU@.
11655 Same as @code{generic}, but when used as @code{march} option, PentiumPro
11656 instruction set will be used, so the code will run on all i686 family chips.
11658 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
11659 @item pentium3, pentium3m
11660 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
11663 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
11664 support. Used by Centrino notebooks.
11665 @item pentium4, pentium4m
11666 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
11668 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
11671 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
11672 SSE2 and SSE3 instruction set support.
11674 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
11675 instruction set support.
11677 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
11678 instruction set support.
11680 AMD K6 CPU with MMX instruction set support.
11682 Improved versions of AMD K6 CPU with MMX and 3dNOW!@: instruction set support.
11683 @item athlon, athlon-tbird
11684 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and SSE prefetch instructions
11686 @item athlon-4, athlon-xp, athlon-mp
11687 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and full SSE
11688 instruction set support.
11689 @item k8, opteron, athlon64, athlon-fx
11690 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
11691 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW!@: and 64-bit instruction set extensions.)
11692 @item k8-sse3, opteron-sse3, athlon64-sse3
11693 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
11694 @item amdfam10, barcelona
11695 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
11696 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3dNOW!, enhanced 3dNOW!, ABM and 64-bit
11697 instruction set extensions.)
11699 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
11702 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!@:
11703 instruction set support.
11705 Via C3 CPU with MMX and 3dNOW!@: instruction set support. (No scheduling is
11706 implemented for this chip.)
11708 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
11709 implemented for this chip.)
11711 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
11714 While picking a specific @var{cpu-type} will schedule things appropriately
11715 for that particular chip, the compiler will not generate any code that
11716 does not run on the i386 without the @option{-march=@var{cpu-type}} option
11719 @item -march=@var{cpu-type}
11721 Generate instructions for the machine type @var{cpu-type}. The choices
11722 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
11723 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
11725 @item -mcpu=@var{cpu-type}
11727 A deprecated synonym for @option{-mtune}.
11729 @item -mfpmath=@var{unit}
11731 Generate floating point arithmetics for selected unit @var{unit}. The choices
11732 for @var{unit} are:
11736 Use the standard 387 floating point coprocessor present majority of chips and
11737 emulated otherwise. Code compiled with this option will run almost everywhere.
11738 The temporary results are computed in 80bit precision instead of precision
11739 specified by the type resulting in slightly different results compared to most
11740 of other chips. See @option{-ffloat-store} for more detailed description.
11742 This is the default choice for i386 compiler.
11745 Use scalar floating point instructions present in the SSE instruction set.
11746 This instruction set is supported by Pentium3 and newer chips, in the AMD line
11747 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
11748 instruction set supports only single precision arithmetics, thus the double and
11749 extended precision arithmetics is still done using 387. Later version, present
11750 only in Pentium4 and the future AMD x86-64 chips supports double precision
11753 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
11754 or @option{-msse2} switches to enable SSE extensions and make this option
11755 effective. For the x86-64 compiler, these extensions are enabled by default.
11757 The resulting code should be considerably faster in the majority of cases and avoid
11758 the numerical instability problems of 387 code, but may break some existing
11759 code that expects temporaries to be 80bit.
11761 This is the default choice for the x86-64 compiler.
11766 Attempt to utilize both instruction sets at once. This effectively double the
11767 amount of available registers and on chips with separate execution units for
11768 387 and SSE the execution resources too. Use this option with care, as it is
11769 still experimental, because the GCC register allocator does not model separate
11770 functional units well resulting in instable performance.
11773 @item -masm=@var{dialect}
11774 @opindex masm=@var{dialect}
11775 Output asm instructions using selected @var{dialect}. Supported
11776 choices are @samp{intel} or @samp{att} (the default one). Darwin does
11777 not support @samp{intel}.
11780 @itemx -mno-ieee-fp
11782 @opindex mno-ieee-fp
11783 Control whether or not the compiler uses IEEE floating point
11784 comparisons. These handle correctly the case where the result of a
11785 comparison is unordered.
11788 @opindex msoft-float
11789 Generate output containing library calls for floating point.
11790 @strong{Warning:} the requisite libraries are not part of GCC@.
11791 Normally the facilities of the machine's usual C compiler are used, but
11792 this can't be done directly in cross-compilation. You must make your
11793 own arrangements to provide suitable library functions for
11796 On machines where a function returns floating point results in the 80387
11797 register stack, some floating point opcodes may be emitted even if
11798 @option{-msoft-float} is used.
11800 @item -mno-fp-ret-in-387
11801 @opindex mno-fp-ret-in-387
11802 Do not use the FPU registers for return values of functions.
11804 The usual calling convention has functions return values of types
11805 @code{float} and @code{double} in an FPU register, even if there
11806 is no FPU@. The idea is that the operating system should emulate
11809 The option @option{-mno-fp-ret-in-387} causes such values to be returned
11810 in ordinary CPU registers instead.
11812 @item -mno-fancy-math-387
11813 @opindex mno-fancy-math-387
11814 Some 387 emulators do not support the @code{sin}, @code{cos} and
11815 @code{sqrt} instructions for the 387. Specify this option to avoid
11816 generating those instructions. This option is the default on FreeBSD,
11817 OpenBSD and NetBSD@. This option is overridden when @option{-march}
11818 indicates that the target cpu will always have an FPU and so the
11819 instruction will not need emulation. As of revision 2.6.1, these
11820 instructions are not generated unless you also use the
11821 @option{-funsafe-math-optimizations} switch.
11823 @item -malign-double
11824 @itemx -mno-align-double
11825 @opindex malign-double
11826 @opindex mno-align-double
11827 Control whether GCC aligns @code{double}, @code{long double}, and
11828 @code{long long} variables on a two word boundary or a one word
11829 boundary. Aligning @code{double} variables on a two word boundary will
11830 produce code that runs somewhat faster on a @samp{Pentium} at the
11831 expense of more memory.
11833 On x86-64, @option{-malign-double} is enabled by default.
11835 @strong{Warning:} if you use the @option{-malign-double} switch,
11836 structures containing the above types will be aligned differently than
11837 the published application binary interface specifications for the 386
11838 and will not be binary compatible with structures in code compiled
11839 without that switch.
11841 @item -m96bit-long-double
11842 @itemx -m128bit-long-double
11843 @opindex m96bit-long-double
11844 @opindex m128bit-long-double
11845 These switches control the size of @code{long double} type. The i386
11846 application binary interface specifies the size to be 96 bits,
11847 so @option{-m96bit-long-double} is the default in 32 bit mode.
11849 Modern architectures (Pentium and newer) would prefer @code{long double}
11850 to be aligned to an 8 or 16 byte boundary. In arrays or structures
11851 conforming to the ABI, this would not be possible. So specifying a
11852 @option{-m128bit-long-double} will align @code{long double}
11853 to a 16 byte boundary by padding the @code{long double} with an additional
11856 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
11857 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
11859 Notice that neither of these options enable any extra precision over the x87
11860 standard of 80 bits for a @code{long double}.
11862 @strong{Warning:} if you override the default value for your target ABI, the
11863 structures and arrays containing @code{long double} variables will change
11864 their size as well as function calling convention for function taking
11865 @code{long double} will be modified. Hence they will not be binary
11866 compatible with arrays or structures in code compiled without that switch.
11868 @item -mlarge-data-threshold=@var{number}
11869 @opindex mlarge-data-threshold=@var{number}
11870 When @option{-mcmodel=medium} is specified, the data greater than
11871 @var{threshold} are placed in large data section. This value must be the
11872 same across all object linked into the binary and defaults to 65535.
11876 Use a different function-calling convention, in which functions that
11877 take a fixed number of arguments return with the @code{ret} @var{num}
11878 instruction, which pops their arguments while returning. This saves one
11879 instruction in the caller since there is no need to pop the arguments
11882 You can specify that an individual function is called with this calling
11883 sequence with the function attribute @samp{stdcall}. You can also
11884 override the @option{-mrtd} option by using the function attribute
11885 @samp{cdecl}. @xref{Function Attributes}.
11887 @strong{Warning:} this calling convention is incompatible with the one
11888 normally used on Unix, so you cannot use it if you need to call
11889 libraries compiled with the Unix compiler.
11891 Also, you must provide function prototypes for all functions that
11892 take variable numbers of arguments (including @code{printf});
11893 otherwise incorrect code will be generated for calls to those
11896 In addition, seriously incorrect code will result if you call a
11897 function with too many arguments. (Normally, extra arguments are
11898 harmlessly ignored.)
11900 @item -mregparm=@var{num}
11902 Control how many registers are used to pass integer arguments. By
11903 default, no registers are used to pass arguments, and at most 3
11904 registers can be used. You can control this behavior for a specific
11905 function by using the function attribute @samp{regparm}.
11906 @xref{Function Attributes}.
11908 @strong{Warning:} if you use this switch, and
11909 @var{num} is nonzero, then you must build all modules with the same
11910 value, including any libraries. This includes the system libraries and
11914 @opindex msseregparm
11915 Use SSE register passing conventions for float and double arguments
11916 and return values. You can control this behavior for a specific
11917 function by using the function attribute @samp{sseregparm}.
11918 @xref{Function Attributes}.
11920 @strong{Warning:} if you use this switch then you must build all
11921 modules with the same value, including any libraries. This includes
11922 the system libraries and startup modules.
11931 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
11932 is specified, the significands of results of floating-point operations are
11933 rounded to 24 bits (single precision); @option{-mpc64} rounds the
11934 significands of results of floating-point operations to 53 bits (double
11935 precision) and @option{-mpc80} rounds the significands of results of
11936 floating-point operations to 64 bits (extended double precision), which is
11937 the default. When this option is used, floating-point operations in higher
11938 precisions are not available to the programmer without setting the FPU
11939 control word explicitly.
11941 Setting the rounding of floating-point operations to less than the default
11942 80 bits can speed some programs by 2% or more. Note that some mathematical
11943 libraries assume that extended precision (80 bit) floating-point operations
11944 are enabled by default; routines in such libraries could suffer significant
11945 loss of accuracy, typically through so-called "catastrophic cancellation",
11946 when this option is used to set the precision to less than extended precision.
11948 @item -mstackrealign
11949 @opindex mstackrealign
11950 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
11951 option will generate an alternate prologue and epilogue that realigns the
11952 runtime stack if necessary. This supports mixing legacy codes that keep
11953 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
11954 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
11955 applicable to individual functions.
11957 @item -mpreferred-stack-boundary=@var{num}
11958 @opindex mpreferred-stack-boundary
11959 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
11960 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
11961 the default is 4 (16 bytes or 128 bits).
11963 @item -mincoming-stack-boundary=@var{num}
11964 @opindex mincoming-stack-boundary
11965 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
11966 boundary. If @option{-mincoming-stack-boundary} is not specified,
11967 the one specified by @option{-mpreferred-stack-boundary} will be used.
11969 On Pentium and PentiumPro, @code{double} and @code{long double} values
11970 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
11971 suffer significant run time performance penalties. On Pentium III, the
11972 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
11973 properly if it is not 16 byte aligned.
11975 To ensure proper alignment of this values on the stack, the stack boundary
11976 must be as aligned as that required by any value stored on the stack.
11977 Further, every function must be generated such that it keeps the stack
11978 aligned. Thus calling a function compiled with a higher preferred
11979 stack boundary from a function compiled with a lower preferred stack
11980 boundary will most likely misalign the stack. It is recommended that
11981 libraries that use callbacks always use the default setting.
11983 This extra alignment does consume extra stack space, and generally
11984 increases code size. Code that is sensitive to stack space usage, such
11985 as embedded systems and operating system kernels, may want to reduce the
11986 preferred alignment to @option{-mpreferred-stack-boundary=2}.
12030 These switches enable or disable the use of instructions in the MMX,
12031 SSE, SSE2, SSE3, SSSE3, SSE4.1, AVX, AES, PCLMUL, SSE4A, FMA4, XOP,
12032 LWP, ABM or 3DNow!@: extended instruction sets.
12033 These extensions are also available as built-in functions: see
12034 @ref{X86 Built-in Functions}, for details of the functions enabled and
12035 disabled by these switches.
12037 To have SSE/SSE2 instructions generated automatically from floating-point
12038 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
12040 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
12041 generates new AVX instructions or AVX equivalence for all SSEx instructions
12044 These options will enable GCC to use these extended instructions in
12045 generated code, even without @option{-mfpmath=sse}. Applications which
12046 perform runtime CPU detection must compile separate files for each
12047 supported architecture, using the appropriate flags. In particular,
12048 the file containing the CPU detection code should be compiled without
12052 @itemx -mno-fused-madd
12053 @opindex mfused-madd
12054 @opindex mno-fused-madd
12055 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
12056 instructions. The default is to use these instructions.
12060 This option instructs GCC to emit a @code{cld} instruction in the prologue
12061 of functions that use string instructions. String instructions depend on
12062 the DF flag to select between autoincrement or autodecrement mode. While the
12063 ABI specifies the DF flag to be cleared on function entry, some operating
12064 systems violate this specification by not clearing the DF flag in their
12065 exception dispatchers. The exception handler can be invoked with the DF flag
12066 set which leads to wrong direction mode, when string instructions are used.
12067 This option can be enabled by default on 32-bit x86 targets by configuring
12068 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
12069 instructions can be suppressed with the @option{-mno-cld} compiler option
12074 This option will enable GCC to use CMPXCHG16B instruction in generated code.
12075 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
12076 data types. This is useful for high resolution counters that could be updated
12077 by multiple processors (or cores). This instruction is generated as part of
12078 atomic built-in functions: see @ref{Atomic Builtins} for details.
12082 This option will enable GCC to use SAHF instruction in generated 64-bit code.
12083 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
12084 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
12085 SAHF are load and store instructions, respectively, for certain status flags.
12086 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
12087 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
12091 This option will enable GCC to use movbe instruction to implement
12092 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
12096 This option will enable built-in functions, @code{__builtin_ia32_crc32qi},
12097 @code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and
12098 @code{__builtin_ia32_crc32di} to generate the crc32 machine instruction.
12102 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
12103 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
12104 to increase precision instead of DIVSS and SQRTSS (and their vectorized
12105 variants) for single precision floating point arguments. These instructions
12106 are generated only when @option{-funsafe-math-optimizations} is enabled
12107 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
12108 Note that while the throughput of the sequence is higher than the throughput
12109 of the non-reciprocal instruction, the precision of the sequence can be
12110 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
12112 Note that GCC implements 1.0f/sqrtf(x) in terms of RSQRTSS (or RSQRTPS)
12113 already with @option{-ffast-math} (or the above option combination), and
12114 doesn't need @option{-mrecip}.
12116 @item -mveclibabi=@var{type}
12117 @opindex mveclibabi
12118 Specifies the ABI type to use for vectorizing intrinsics using an
12119 external library. Supported types are @code{svml} for the Intel short
12120 vector math library and @code{acml} for the AMD math core library style
12121 of interfacing. GCC will currently emit calls to @code{vmldExp2},
12122 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
12123 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
12124 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
12125 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
12126 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
12127 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
12128 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
12129 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
12130 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
12131 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
12132 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
12133 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
12134 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
12135 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
12136 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
12137 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
12138 compatible library will have to be specified at link time.
12140 @item -mabi=@var{name}
12142 Generate code for the specified calling convention. Permissible values
12143 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
12144 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
12145 ABI when targeting Windows. On all other systems, the default is the
12146 SYSV ABI. You can control this behavior for a specific function by
12147 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
12148 @xref{Function Attributes}.
12151 @itemx -mno-push-args
12152 @opindex mpush-args
12153 @opindex mno-push-args
12154 Use PUSH operations to store outgoing parameters. This method is shorter
12155 and usually equally fast as method using SUB/MOV operations and is enabled
12156 by default. In some cases disabling it may improve performance because of
12157 improved scheduling and reduced dependencies.
12159 @item -maccumulate-outgoing-args
12160 @opindex maccumulate-outgoing-args
12161 If enabled, the maximum amount of space required for outgoing arguments will be
12162 computed in the function prologue. This is faster on most modern CPUs
12163 because of reduced dependencies, improved scheduling and reduced stack usage
12164 when preferred stack boundary is not equal to 2. The drawback is a notable
12165 increase in code size. This switch implies @option{-mno-push-args}.
12169 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
12170 on thread-safe exception handling must compile and link all code with the
12171 @option{-mthreads} option. When compiling, @option{-mthreads} defines
12172 @option{-D_MT}; when linking, it links in a special thread helper library
12173 @option{-lmingwthrd} which cleans up per thread exception handling data.
12175 @item -mno-align-stringops
12176 @opindex mno-align-stringops
12177 Do not align destination of inlined string operations. This switch reduces
12178 code size and improves performance in case the destination is already aligned,
12179 but GCC doesn't know about it.
12181 @item -minline-all-stringops
12182 @opindex minline-all-stringops
12183 By default GCC inlines string operations only when destination is known to be
12184 aligned at least to 4 byte boundary. This enables more inlining, increase code
12185 size, but may improve performance of code that depends on fast memcpy, strlen
12186 and memset for short lengths.
12188 @item -minline-stringops-dynamically
12189 @opindex minline-stringops-dynamically
12190 For string operation of unknown size, inline runtime checks so for small
12191 blocks inline code is used, while for large blocks library call is used.
12193 @item -mstringop-strategy=@var{alg}
12194 @opindex mstringop-strategy=@var{alg}
12195 Overwrite internal decision heuristic about particular algorithm to inline
12196 string operation with. The allowed values are @code{rep_byte},
12197 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
12198 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
12199 expanding inline loop, @code{libcall} for always expanding library call.
12201 @item -momit-leaf-frame-pointer
12202 @opindex momit-leaf-frame-pointer
12203 Don't keep the frame pointer in a register for leaf functions. This
12204 avoids the instructions to save, set up and restore frame pointers and
12205 makes an extra register available in leaf functions. The option
12206 @option{-fomit-frame-pointer} removes the frame pointer for all functions
12207 which might make debugging harder.
12209 @item -mtls-direct-seg-refs
12210 @itemx -mno-tls-direct-seg-refs
12211 @opindex mtls-direct-seg-refs
12212 Controls whether TLS variables may be accessed with offsets from the
12213 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
12214 or whether the thread base pointer must be added. Whether or not this
12215 is legal depends on the operating system, and whether it maps the
12216 segment to cover the entire TLS area.
12218 For systems that use GNU libc, the default is on.
12221 @itemx -mno-sse2avx
12223 Specify that the assembler should encode SSE instructions with VEX
12224 prefix. The option @option{-mavx} turns this on by default.
12227 These @samp{-m} switches are supported in addition to the above
12228 on AMD x86-64 processors in 64-bit environments.
12235 Generate code for a 32-bit or 64-bit environment.
12236 The 32-bit environment sets int, long and pointer to 32 bits and
12237 generates code that runs on any i386 system.
12238 The 64-bit environment sets int to 32 bits and long and pointer
12239 to 64 bits and generates code for AMD's x86-64 architecture. For
12240 darwin only the -m64 option turns off the @option{-fno-pic} and
12241 @option{-mdynamic-no-pic} options.
12243 @item -mno-red-zone
12244 @opindex mno-red-zone
12245 Do not use a so called red zone for x86-64 code. The red zone is mandated
12246 by the x86-64 ABI, it is a 128-byte area beyond the location of the
12247 stack pointer that will not be modified by signal or interrupt handlers
12248 and therefore can be used for temporary data without adjusting the stack
12249 pointer. The flag @option{-mno-red-zone} disables this red zone.
12251 @item -mcmodel=small
12252 @opindex mcmodel=small
12253 Generate code for the small code model: the program and its symbols must
12254 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
12255 Programs can be statically or dynamically linked. This is the default
12258 @item -mcmodel=kernel
12259 @opindex mcmodel=kernel
12260 Generate code for the kernel code model. The kernel runs in the
12261 negative 2 GB of the address space.
12262 This model has to be used for Linux kernel code.
12264 @item -mcmodel=medium
12265 @opindex mcmodel=medium
12266 Generate code for the medium model: The program is linked in the lower 2
12267 GB of the address space. Small symbols are also placed there. Symbols
12268 with sizes larger than @option{-mlarge-data-threshold} are put into
12269 large data or bss sections and can be located above 2GB. Programs can
12270 be statically or dynamically linked.
12272 @item -mcmodel=large
12273 @opindex mcmodel=large
12274 Generate code for the large model: This model makes no assumptions
12275 about addresses and sizes of sections.
12278 @node IA-64 Options
12279 @subsection IA-64 Options
12280 @cindex IA-64 Options
12282 These are the @samp{-m} options defined for the Intel IA-64 architecture.
12286 @opindex mbig-endian
12287 Generate code for a big endian target. This is the default for HP-UX@.
12289 @item -mlittle-endian
12290 @opindex mlittle-endian
12291 Generate code for a little endian target. This is the default for AIX5
12297 @opindex mno-gnu-as
12298 Generate (or don't) code for the GNU assembler. This is the default.
12299 @c Also, this is the default if the configure option @option{--with-gnu-as}
12305 @opindex mno-gnu-ld
12306 Generate (or don't) code for the GNU linker. This is the default.
12307 @c Also, this is the default if the configure option @option{--with-gnu-ld}
12312 Generate code that does not use a global pointer register. The result
12313 is not position independent code, and violates the IA-64 ABI@.
12315 @item -mvolatile-asm-stop
12316 @itemx -mno-volatile-asm-stop
12317 @opindex mvolatile-asm-stop
12318 @opindex mno-volatile-asm-stop
12319 Generate (or don't) a stop bit immediately before and after volatile asm
12322 @item -mregister-names
12323 @itemx -mno-register-names
12324 @opindex mregister-names
12325 @opindex mno-register-names
12326 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
12327 the stacked registers. This may make assembler output more readable.
12333 Disable (or enable) optimizations that use the small data section. This may
12334 be useful for working around optimizer bugs.
12336 @item -mconstant-gp
12337 @opindex mconstant-gp
12338 Generate code that uses a single constant global pointer value. This is
12339 useful when compiling kernel code.
12343 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
12344 This is useful when compiling firmware code.
12346 @item -minline-float-divide-min-latency
12347 @opindex minline-float-divide-min-latency
12348 Generate code for inline divides of floating point values
12349 using the minimum latency algorithm.
12351 @item -minline-float-divide-max-throughput
12352 @opindex minline-float-divide-max-throughput
12353 Generate code for inline divides of floating point values
12354 using the maximum throughput algorithm.
12356 @item -mno-inline-float-divide
12357 @opindex mno-inline-float-divide
12358 Do not generate inline code for divides of floating point values.
12360 @item -minline-int-divide-min-latency
12361 @opindex minline-int-divide-min-latency
12362 Generate code for inline divides of integer values
12363 using the minimum latency algorithm.
12365 @item -minline-int-divide-max-throughput
12366 @opindex minline-int-divide-max-throughput
12367 Generate code for inline divides of integer values
12368 using the maximum throughput algorithm.
12370 @item -mno-inline-int-divide
12371 @opindex mno-inline-int-divide
12372 Do not generate inline code for divides of integer values.
12374 @item -minline-sqrt-min-latency
12375 @opindex minline-sqrt-min-latency
12376 Generate code for inline square roots
12377 using the minimum latency algorithm.
12379 @item -minline-sqrt-max-throughput
12380 @opindex minline-sqrt-max-throughput
12381 Generate code for inline square roots
12382 using the maximum throughput algorithm.
12384 @item -mno-inline-sqrt
12385 @opindex mno-inline-sqrt
12386 Do not generate inline code for sqrt.
12389 @itemx -mno-fused-madd
12390 @opindex mfused-madd
12391 @opindex mno-fused-madd
12392 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
12393 instructions. The default is to use these instructions.
12395 @item -mno-dwarf2-asm
12396 @itemx -mdwarf2-asm
12397 @opindex mno-dwarf2-asm
12398 @opindex mdwarf2-asm
12399 Don't (or do) generate assembler code for the DWARF2 line number debugging
12400 info. This may be useful when not using the GNU assembler.
12402 @item -mearly-stop-bits
12403 @itemx -mno-early-stop-bits
12404 @opindex mearly-stop-bits
12405 @opindex mno-early-stop-bits
12406 Allow stop bits to be placed earlier than immediately preceding the
12407 instruction that triggered the stop bit. This can improve instruction
12408 scheduling, but does not always do so.
12410 @item -mfixed-range=@var{register-range}
12411 @opindex mfixed-range
12412 Generate code treating the given register range as fixed registers.
12413 A fixed register is one that the register allocator can not use. This is
12414 useful when compiling kernel code. A register range is specified as
12415 two registers separated by a dash. Multiple register ranges can be
12416 specified separated by a comma.
12418 @item -mtls-size=@var{tls-size}
12420 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
12423 @item -mtune=@var{cpu-type}
12425 Tune the instruction scheduling for a particular CPU, Valid values are
12426 itanium, itanium1, merced, itanium2, and mckinley.
12432 Generate code for a 32-bit or 64-bit environment.
12433 The 32-bit environment sets int, long and pointer to 32 bits.
12434 The 64-bit environment sets int to 32 bits and long and pointer
12435 to 64 bits. These are HP-UX specific flags.
12437 @item -mno-sched-br-data-spec
12438 @itemx -msched-br-data-spec
12439 @opindex mno-sched-br-data-spec
12440 @opindex msched-br-data-spec
12441 (Dis/En)able data speculative scheduling before reload.
12442 This will result in generation of the ld.a instructions and
12443 the corresponding check instructions (ld.c / chk.a).
12444 The default is 'disable'.
12446 @item -msched-ar-data-spec
12447 @itemx -mno-sched-ar-data-spec
12448 @opindex msched-ar-data-spec
12449 @opindex mno-sched-ar-data-spec
12450 (En/Dis)able data speculative scheduling after reload.
12451 This will result in generation of the ld.a instructions and
12452 the corresponding check instructions (ld.c / chk.a).
12453 The default is 'enable'.
12455 @item -mno-sched-control-spec
12456 @itemx -msched-control-spec
12457 @opindex mno-sched-control-spec
12458 @opindex msched-control-spec
12459 (Dis/En)able control speculative scheduling. This feature is
12460 available only during region scheduling (i.e.@: before reload).
12461 This will result in generation of the ld.s instructions and
12462 the corresponding check instructions chk.s .
12463 The default is 'disable'.
12465 @item -msched-br-in-data-spec
12466 @itemx -mno-sched-br-in-data-spec
12467 @opindex msched-br-in-data-spec
12468 @opindex mno-sched-br-in-data-spec
12469 (En/Dis)able speculative scheduling of the instructions that
12470 are dependent on the data speculative loads before reload.
12471 This is effective only with @option{-msched-br-data-spec} enabled.
12472 The default is 'enable'.
12474 @item -msched-ar-in-data-spec
12475 @itemx -mno-sched-ar-in-data-spec
12476 @opindex msched-ar-in-data-spec
12477 @opindex mno-sched-ar-in-data-spec
12478 (En/Dis)able speculative scheduling of the instructions that
12479 are dependent on the data speculative loads after reload.
12480 This is effective only with @option{-msched-ar-data-spec} enabled.
12481 The default is 'enable'.
12483 @item -msched-in-control-spec
12484 @itemx -mno-sched-in-control-spec
12485 @opindex msched-in-control-spec
12486 @opindex mno-sched-in-control-spec
12487 (En/Dis)able speculative scheduling of the instructions that
12488 are dependent on the control speculative loads.
12489 This is effective only with @option{-msched-control-spec} enabled.
12490 The default is 'enable'.
12492 @item -mno-sched-prefer-non-data-spec-insns
12493 @itemx -msched-prefer-non-data-spec-insns
12494 @opindex mno-sched-prefer-non-data-spec-insns
12495 @opindex msched-prefer-non-data-spec-insns
12496 If enabled, data speculative instructions will be chosen for schedule
12497 only if there are no other choices at the moment. This will make
12498 the use of the data speculation much more conservative.
12499 The default is 'disable'.
12501 @item -mno-sched-prefer-non-control-spec-insns
12502 @itemx -msched-prefer-non-control-spec-insns
12503 @opindex mno-sched-prefer-non-control-spec-insns
12504 @opindex msched-prefer-non-control-spec-insns
12505 If enabled, control speculative instructions will be chosen for schedule
12506 only if there are no other choices at the moment. This will make
12507 the use of the control speculation much more conservative.
12508 The default is 'disable'.
12510 @item -mno-sched-count-spec-in-critical-path
12511 @itemx -msched-count-spec-in-critical-path
12512 @opindex mno-sched-count-spec-in-critical-path
12513 @opindex msched-count-spec-in-critical-path
12514 If enabled, speculative dependencies will be considered during
12515 computation of the instructions priorities. This will make the use of the
12516 speculation a bit more conservative.
12517 The default is 'disable'.
12519 @item -msched-spec-ldc
12520 @opindex msched-spec-ldc
12521 Use a simple data speculation check. This option is on by default.
12523 @item -msched-control-spec-ldc
12524 @opindex msched-spec-ldc
12525 Use a simple check for control speculation. This option is on by default.
12527 @item -msched-stop-bits-after-every-cycle
12528 @opindex msched-stop-bits-after-every-cycle
12529 Place a stop bit after every cycle when scheduling. This option is on
12532 @item -msched-fp-mem-deps-zero-cost
12533 @opindex msched-fp-mem-deps-zero-cost
12534 Assume that floating-point stores and loads are not likely to cause a conflict
12535 when placed into the same instruction group. This option is disabled by
12538 @item -msel-sched-dont-check-control-spec
12539 @opindex msel-sched-dont-check-control-spec
12540 Generate checks for control speculation in selective scheduling.
12541 This flag is disabled by default.
12543 @item -msched-max-memory-insns=@var{max-insns}
12544 @opindex msched-max-memory-insns
12545 Limit on the number of memory insns per instruction group, giving lower
12546 priority to subsequent memory insns attempting to schedule in the same
12547 instruction group. Frequently useful to prevent cache bank conflicts.
12548 The default value is 1.
12550 @item -msched-max-memory-insns-hard-limit
12551 @opindex msched-max-memory-insns-hard-limit
12552 Disallow more than `msched-max-memory-insns' in instruction group.
12553 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
12554 when limit is reached but may still schedule memory operations.
12558 @node IA-64/VMS Options
12559 @subsection IA-64/VMS Options
12561 These @samp{-m} options are defined for the IA-64/VMS implementations:
12564 @item -mvms-return-codes
12565 @opindex mvms-return-codes
12566 Return VMS condition codes from main. The default is to return POSIX
12567 style condition (e.g.@ error) codes.
12569 @item -mdebug-main=@var{prefix}
12570 @opindex mdebug-main=@var{prefix}
12571 Flag the first routine whose name starts with @var{prefix} as the main
12572 routine for the debugger.
12576 Default to 64bit memory allocation routines.
12580 @subsection LM32 Options
12581 @cindex LM32 options
12583 These @option{-m} options are defined for the Lattice Mico32 architecture:
12586 @item -mbarrel-shift-enabled
12587 @opindex mbarrel-shift-enabled
12588 Enable barrel-shift instructions.
12590 @item -mdivide-enabled
12591 @opindex mdivide-enabled
12592 Enable divide and modulus instructions.
12594 @item -mmultiply-enabled
12595 @opindex multiply-enabled
12596 Enable multiply instructions.
12598 @item -msign-extend-enabled
12599 @opindex msign-extend-enabled
12600 Enable sign extend instructions.
12602 @item -muser-enabled
12603 @opindex muser-enabled
12604 Enable user-defined instructions.
12609 @subsection M32C Options
12610 @cindex M32C options
12613 @item -mcpu=@var{name}
12615 Select the CPU for which code is generated. @var{name} may be one of
12616 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
12617 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
12618 the M32C/80 series.
12622 Specifies that the program will be run on the simulator. This causes
12623 an alternate runtime library to be linked in which supports, for
12624 example, file I/O@. You must not use this option when generating
12625 programs that will run on real hardware; you must provide your own
12626 runtime library for whatever I/O functions are needed.
12628 @item -memregs=@var{number}
12630 Specifies the number of memory-based pseudo-registers GCC will use
12631 during code generation. These pseudo-registers will be used like real
12632 registers, so there is a tradeoff between GCC's ability to fit the
12633 code into available registers, and the performance penalty of using
12634 memory instead of registers. Note that all modules in a program must
12635 be compiled with the same value for this option. Because of that, you
12636 must not use this option with the default runtime libraries gcc
12641 @node M32R/D Options
12642 @subsection M32R/D Options
12643 @cindex M32R/D options
12645 These @option{-m} options are defined for Renesas M32R/D architectures:
12650 Generate code for the M32R/2@.
12654 Generate code for the M32R/X@.
12658 Generate code for the M32R@. This is the default.
12660 @item -mmodel=small
12661 @opindex mmodel=small
12662 Assume all objects live in the lower 16MB of memory (so that their addresses
12663 can be loaded with the @code{ld24} instruction), and assume all subroutines
12664 are reachable with the @code{bl} instruction.
12665 This is the default.
12667 The addressability of a particular object can be set with the
12668 @code{model} attribute.
12670 @item -mmodel=medium
12671 @opindex mmodel=medium
12672 Assume objects may be anywhere in the 32-bit address space (the compiler
12673 will generate @code{seth/add3} instructions to load their addresses), and
12674 assume all subroutines are reachable with the @code{bl} instruction.
12676 @item -mmodel=large
12677 @opindex mmodel=large
12678 Assume objects may be anywhere in the 32-bit address space (the compiler
12679 will generate @code{seth/add3} instructions to load their addresses), and
12680 assume subroutines may not be reachable with the @code{bl} instruction
12681 (the compiler will generate the much slower @code{seth/add3/jl}
12682 instruction sequence).
12685 @opindex msdata=none
12686 Disable use of the small data area. Variables will be put into
12687 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
12688 @code{section} attribute has been specified).
12689 This is the default.
12691 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
12692 Objects may be explicitly put in the small data area with the
12693 @code{section} attribute using one of these sections.
12695 @item -msdata=sdata
12696 @opindex msdata=sdata
12697 Put small global and static data in the small data area, but do not
12698 generate special code to reference them.
12701 @opindex msdata=use
12702 Put small global and static data in the small data area, and generate
12703 special instructions to reference them.
12707 @cindex smaller data references
12708 Put global and static objects less than or equal to @var{num} bytes
12709 into the small data or bss sections instead of the normal data or bss
12710 sections. The default value of @var{num} is 8.
12711 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
12712 for this option to have any effect.
12714 All modules should be compiled with the same @option{-G @var{num}} value.
12715 Compiling with different values of @var{num} may or may not work; if it
12716 doesn't the linker will give an error message---incorrect code will not be
12721 Makes the M32R specific code in the compiler display some statistics
12722 that might help in debugging programs.
12724 @item -malign-loops
12725 @opindex malign-loops
12726 Align all loops to a 32-byte boundary.
12728 @item -mno-align-loops
12729 @opindex mno-align-loops
12730 Do not enforce a 32-byte alignment for loops. This is the default.
12732 @item -missue-rate=@var{number}
12733 @opindex missue-rate=@var{number}
12734 Issue @var{number} instructions per cycle. @var{number} can only be 1
12737 @item -mbranch-cost=@var{number}
12738 @opindex mbranch-cost=@var{number}
12739 @var{number} can only be 1 or 2. If it is 1 then branches will be
12740 preferred over conditional code, if it is 2, then the opposite will
12743 @item -mflush-trap=@var{number}
12744 @opindex mflush-trap=@var{number}
12745 Specifies the trap number to use to flush the cache. The default is
12746 12. Valid numbers are between 0 and 15 inclusive.
12748 @item -mno-flush-trap
12749 @opindex mno-flush-trap
12750 Specifies that the cache cannot be flushed by using a trap.
12752 @item -mflush-func=@var{name}
12753 @opindex mflush-func=@var{name}
12754 Specifies the name of the operating system function to call to flush
12755 the cache. The default is @emph{_flush_cache}, but a function call
12756 will only be used if a trap is not available.
12758 @item -mno-flush-func
12759 @opindex mno-flush-func
12760 Indicates that there is no OS function for flushing the cache.
12764 @node M680x0 Options
12765 @subsection M680x0 Options
12766 @cindex M680x0 options
12768 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
12769 The default settings depend on which architecture was selected when
12770 the compiler was configured; the defaults for the most common choices
12774 @item -march=@var{arch}
12776 Generate code for a specific M680x0 or ColdFire instruction set
12777 architecture. Permissible values of @var{arch} for M680x0
12778 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
12779 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
12780 architectures are selected according to Freescale's ISA classification
12781 and the permissible values are: @samp{isaa}, @samp{isaaplus},
12782 @samp{isab} and @samp{isac}.
12784 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
12785 code for a ColdFire target. The @var{arch} in this macro is one of the
12786 @option{-march} arguments given above.
12788 When used together, @option{-march} and @option{-mtune} select code
12789 that runs on a family of similar processors but that is optimized
12790 for a particular microarchitecture.
12792 @item -mcpu=@var{cpu}
12794 Generate code for a specific M680x0 or ColdFire processor.
12795 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
12796 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
12797 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
12798 below, which also classifies the CPUs into families:
12800 @multitable @columnfractions 0.20 0.80
12801 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
12802 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
12803 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
12804 @item @samp{5206e} @tab @samp{5206e}
12805 @item @samp{5208} @tab @samp{5207} @samp{5208}
12806 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
12807 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
12808 @item @samp{5216} @tab @samp{5214} @samp{5216}
12809 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
12810 @item @samp{5225} @tab @samp{5224} @samp{5225}
12811 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
12812 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
12813 @item @samp{5249} @tab @samp{5249}
12814 @item @samp{5250} @tab @samp{5250}
12815 @item @samp{5271} @tab @samp{5270} @samp{5271}
12816 @item @samp{5272} @tab @samp{5272}
12817 @item @samp{5275} @tab @samp{5274} @samp{5275}
12818 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
12819 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
12820 @item @samp{5307} @tab @samp{5307}
12821 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
12822 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
12823 @item @samp{5407} @tab @samp{5407}
12824 @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}
12827 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
12828 @var{arch} is compatible with @var{cpu}. Other combinations of
12829 @option{-mcpu} and @option{-march} are rejected.
12831 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
12832 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
12833 where the value of @var{family} is given by the table above.
12835 @item -mtune=@var{tune}
12837 Tune the code for a particular microarchitecture, within the
12838 constraints set by @option{-march} and @option{-mcpu}.
12839 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
12840 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
12841 and @samp{cpu32}. The ColdFire microarchitectures
12842 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
12844 You can also use @option{-mtune=68020-40} for code that needs
12845 to run relatively well on 68020, 68030 and 68040 targets.
12846 @option{-mtune=68020-60} is similar but includes 68060 targets
12847 as well. These two options select the same tuning decisions as
12848 @option{-m68020-40} and @option{-m68020-60} respectively.
12850 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
12851 when tuning for 680x0 architecture @var{arch}. It also defines
12852 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
12853 option is used. If gcc is tuning for a range of architectures,
12854 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
12855 it defines the macros for every architecture in the range.
12857 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
12858 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
12859 of the arguments given above.
12865 Generate output for a 68000. This is the default
12866 when the compiler is configured for 68000-based systems.
12867 It is equivalent to @option{-march=68000}.
12869 Use this option for microcontrollers with a 68000 or EC000 core,
12870 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
12874 Generate output for a 68010. This is the default
12875 when the compiler is configured for 68010-based systems.
12876 It is equivalent to @option{-march=68010}.
12882 Generate output for a 68020. This is the default
12883 when the compiler is configured for 68020-based systems.
12884 It is equivalent to @option{-march=68020}.
12888 Generate output for a 68030. This is the default when the compiler is
12889 configured for 68030-based systems. It is equivalent to
12890 @option{-march=68030}.
12894 Generate output for a 68040. This is the default when the compiler is
12895 configured for 68040-based systems. It is equivalent to
12896 @option{-march=68040}.
12898 This option inhibits the use of 68881/68882 instructions that have to be
12899 emulated by software on the 68040. Use this option if your 68040 does not
12900 have code to emulate those instructions.
12904 Generate output for a 68060. This is the default when the compiler is
12905 configured for 68060-based systems. It is equivalent to
12906 @option{-march=68060}.
12908 This option inhibits the use of 68020 and 68881/68882 instructions that
12909 have to be emulated by software on the 68060. Use this option if your 68060
12910 does not have code to emulate those instructions.
12914 Generate output for a CPU32. This is the default
12915 when the compiler is configured for CPU32-based systems.
12916 It is equivalent to @option{-march=cpu32}.
12918 Use this option for microcontrollers with a
12919 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
12920 68336, 68340, 68341, 68349 and 68360.
12924 Generate output for a 520X ColdFire CPU@. This is the default
12925 when the compiler is configured for 520X-based systems.
12926 It is equivalent to @option{-mcpu=5206}, and is now deprecated
12927 in favor of that option.
12929 Use this option for microcontroller with a 5200 core, including
12930 the MCF5202, MCF5203, MCF5204 and MCF5206.
12934 Generate output for a 5206e ColdFire CPU@. The option is now
12935 deprecated in favor of the equivalent @option{-mcpu=5206e}.
12939 Generate output for a member of the ColdFire 528X family.
12940 The option is now deprecated in favor of the equivalent
12941 @option{-mcpu=528x}.
12945 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
12946 in favor of the equivalent @option{-mcpu=5307}.
12950 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
12951 in favor of the equivalent @option{-mcpu=5407}.
12955 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
12956 This includes use of hardware floating point instructions.
12957 The option is equivalent to @option{-mcpu=547x}, and is now
12958 deprecated in favor of that option.
12962 Generate output for a 68040, without using any of the new instructions.
12963 This results in code which can run relatively efficiently on either a
12964 68020/68881 or a 68030 or a 68040. The generated code does use the
12965 68881 instructions that are emulated on the 68040.
12967 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
12971 Generate output for a 68060, without using any of the new instructions.
12972 This results in code which can run relatively efficiently on either a
12973 68020/68881 or a 68030 or a 68040. The generated code does use the
12974 68881 instructions that are emulated on the 68060.
12976 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
12980 @opindex mhard-float
12982 Generate floating-point instructions. This is the default for 68020
12983 and above, and for ColdFire devices that have an FPU@. It defines the
12984 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
12985 on ColdFire targets.
12988 @opindex msoft-float
12989 Do not generate floating-point instructions; use library calls instead.
12990 This is the default for 68000, 68010, and 68832 targets. It is also
12991 the default for ColdFire devices that have no FPU.
12997 Generate (do not generate) ColdFire hardware divide and remainder
12998 instructions. If @option{-march} is used without @option{-mcpu},
12999 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
13000 architectures. Otherwise, the default is taken from the target CPU
13001 (either the default CPU, or the one specified by @option{-mcpu}). For
13002 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
13003 @option{-mcpu=5206e}.
13005 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
13009 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13010 Additionally, parameters passed on the stack are also aligned to a
13011 16-bit boundary even on targets whose API mandates promotion to 32-bit.
13015 Do not consider type @code{int} to be 16 bits wide. This is the default.
13018 @itemx -mno-bitfield
13019 @opindex mnobitfield
13020 @opindex mno-bitfield
13021 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
13022 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
13026 Do use the bit-field instructions. The @option{-m68020} option implies
13027 @option{-mbitfield}. This is the default if you use a configuration
13028 designed for a 68020.
13032 Use a different function-calling convention, in which functions
13033 that take a fixed number of arguments return with the @code{rtd}
13034 instruction, which pops their arguments while returning. This
13035 saves one instruction in the caller since there is no need to pop
13036 the arguments there.
13038 This calling convention is incompatible with the one normally
13039 used on Unix, so you cannot use it if you need to call libraries
13040 compiled with the Unix compiler.
13042 Also, you must provide function prototypes for all functions that
13043 take variable numbers of arguments (including @code{printf});
13044 otherwise incorrect code will be generated for calls to those
13047 In addition, seriously incorrect code will result if you call a
13048 function with too many arguments. (Normally, extra arguments are
13049 harmlessly ignored.)
13051 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
13052 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
13056 Do not use the calling conventions selected by @option{-mrtd}.
13057 This is the default.
13060 @itemx -mno-align-int
13061 @opindex malign-int
13062 @opindex mno-align-int
13063 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
13064 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
13065 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
13066 Aligning variables on 32-bit boundaries produces code that runs somewhat
13067 faster on processors with 32-bit busses at the expense of more memory.
13069 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
13070 align structures containing the above types differently than
13071 most published application binary interface specifications for the m68k.
13075 Use the pc-relative addressing mode of the 68000 directly, instead of
13076 using a global offset table. At present, this option implies @option{-fpic},
13077 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
13078 not presently supported with @option{-mpcrel}, though this could be supported for
13079 68020 and higher processors.
13081 @item -mno-strict-align
13082 @itemx -mstrict-align
13083 @opindex mno-strict-align
13084 @opindex mstrict-align
13085 Do not (do) assume that unaligned memory references will be handled by
13089 Generate code that allows the data segment to be located in a different
13090 area of memory from the text segment. This allows for execute in place in
13091 an environment without virtual memory management. This option implies
13094 @item -mno-sep-data
13095 Generate code that assumes that the data segment follows the text segment.
13096 This is the default.
13098 @item -mid-shared-library
13099 Generate code that supports shared libraries via the library ID method.
13100 This allows for execute in place and shared libraries in an environment
13101 without virtual memory management. This option implies @option{-fPIC}.
13103 @item -mno-id-shared-library
13104 Generate code that doesn't assume ID based shared libraries are being used.
13105 This is the default.
13107 @item -mshared-library-id=n
13108 Specified the identification number of the ID based shared library being
13109 compiled. Specifying a value of 0 will generate more compact code, specifying
13110 other values will force the allocation of that number to the current
13111 library but is no more space or time efficient than omitting this option.
13117 When generating position-independent code for ColdFire, generate code
13118 that works if the GOT has more than 8192 entries. This code is
13119 larger and slower than code generated without this option. On M680x0
13120 processors, this option is not needed; @option{-fPIC} suffices.
13122 GCC normally uses a single instruction to load values from the GOT@.
13123 While this is relatively efficient, it only works if the GOT
13124 is smaller than about 64k. Anything larger causes the linker
13125 to report an error such as:
13127 @cindex relocation truncated to fit (ColdFire)
13129 relocation truncated to fit: R_68K_GOT16O foobar
13132 If this happens, you should recompile your code with @option{-mxgot}.
13133 It should then work with very large GOTs. However, code generated with
13134 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
13135 the value of a global symbol.
13137 Note that some linkers, including newer versions of the GNU linker,
13138 can create multiple GOTs and sort GOT entries. If you have such a linker,
13139 you should only need to use @option{-mxgot} when compiling a single
13140 object file that accesses more than 8192 GOT entries. Very few do.
13142 These options have no effect unless GCC is generating
13143 position-independent code.
13147 @node M68hc1x Options
13148 @subsection M68hc1x Options
13149 @cindex M68hc1x options
13151 These are the @samp{-m} options defined for the 68hc11 and 68hc12
13152 microcontrollers. The default values for these options depends on
13153 which style of microcontroller was selected when the compiler was configured;
13154 the defaults for the most common choices are given below.
13161 Generate output for a 68HC11. This is the default
13162 when the compiler is configured for 68HC11-based systems.
13168 Generate output for a 68HC12. This is the default
13169 when the compiler is configured for 68HC12-based systems.
13175 Generate output for a 68HCS12.
13177 @item -mauto-incdec
13178 @opindex mauto-incdec
13179 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
13186 Enable the use of 68HC12 min and max instructions.
13189 @itemx -mno-long-calls
13190 @opindex mlong-calls
13191 @opindex mno-long-calls
13192 Treat all calls as being far away (near). If calls are assumed to be
13193 far away, the compiler will use the @code{call} instruction to
13194 call a function and the @code{rtc} instruction for returning.
13198 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13200 @item -msoft-reg-count=@var{count}
13201 @opindex msoft-reg-count
13202 Specify the number of pseudo-soft registers which are used for the
13203 code generation. The maximum number is 32. Using more pseudo-soft
13204 register may or may not result in better code depending on the program.
13205 The default is 4 for 68HC11 and 2 for 68HC12.
13209 @node MCore Options
13210 @subsection MCore Options
13211 @cindex MCore options
13213 These are the @samp{-m} options defined for the Motorola M*Core
13219 @itemx -mno-hardlit
13221 @opindex mno-hardlit
13222 Inline constants into the code stream if it can be done in two
13223 instructions or less.
13229 Use the divide instruction. (Enabled by default).
13231 @item -mrelax-immediate
13232 @itemx -mno-relax-immediate
13233 @opindex mrelax-immediate
13234 @opindex mno-relax-immediate
13235 Allow arbitrary sized immediates in bit operations.
13237 @item -mwide-bitfields
13238 @itemx -mno-wide-bitfields
13239 @opindex mwide-bitfields
13240 @opindex mno-wide-bitfields
13241 Always treat bit-fields as int-sized.
13243 @item -m4byte-functions
13244 @itemx -mno-4byte-functions
13245 @opindex m4byte-functions
13246 @opindex mno-4byte-functions
13247 Force all functions to be aligned to a four byte boundary.
13249 @item -mcallgraph-data
13250 @itemx -mno-callgraph-data
13251 @opindex mcallgraph-data
13252 @opindex mno-callgraph-data
13253 Emit callgraph information.
13256 @itemx -mno-slow-bytes
13257 @opindex mslow-bytes
13258 @opindex mno-slow-bytes
13259 Prefer word access when reading byte quantities.
13261 @item -mlittle-endian
13262 @itemx -mbig-endian
13263 @opindex mlittle-endian
13264 @opindex mbig-endian
13265 Generate code for a little endian target.
13271 Generate code for the 210 processor.
13275 Assume that run-time support has been provided and so omit the
13276 simulator library (@file{libsim.a)} from the linker command line.
13278 @item -mstack-increment=@var{size}
13279 @opindex mstack-increment
13280 Set the maximum amount for a single stack increment operation. Large
13281 values can increase the speed of programs which contain functions
13282 that need a large amount of stack space, but they can also trigger a
13283 segmentation fault if the stack is extended too much. The default
13289 @subsection MeP Options
13290 @cindex MeP options
13296 Enables the @code{abs} instruction, which is the absolute difference
13297 between two registers.
13301 Enables all the optional instructions - average, multiply, divide, bit
13302 operations, leading zero, absolute difference, min/max, clip, and
13308 Enables the @code{ave} instruction, which computes the average of two
13311 @item -mbased=@var{n}
13313 Variables of size @var{n} bytes or smaller will be placed in the
13314 @code{.based} section by default. Based variables use the @code{$tp}
13315 register as a base register, and there is a 128 byte limit to the
13316 @code{.based} section.
13320 Enables the bit operation instructions - bit test (@code{btstm}), set
13321 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
13322 test-and-set (@code{tas}).
13324 @item -mc=@var{name}
13326 Selects which section constant data will be placed in. @var{name} may
13327 be @code{tiny}, @code{near}, or @code{far}.
13331 Enables the @code{clip} instruction. Note that @code{-mclip} is not
13332 useful unless you also provide @code{-mminmax}.
13334 @item -mconfig=@var{name}
13336 Selects one of the build-in core configurations. Each MeP chip has
13337 one or more modules in it; each module has a core CPU and a variety of
13338 coprocessors, optional instructions, and peripherals. The
13339 @code{MeP-Integrator} tool, not part of GCC, provides these
13340 configurations through this option; using this option is the same as
13341 using all the corresponding command line options. The default
13342 configuration is @code{default}.
13346 Enables the coprocessor instructions. By default, this is a 32-bit
13347 coprocessor. Note that the coprocessor is normally enabled via the
13348 @code{-mconfig=} option.
13352 Enables the 32-bit coprocessor's instructions.
13356 Enables the 64-bit coprocessor's instructions.
13360 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
13364 Causes constant variables to be placed in the @code{.near} section.
13368 Enables the @code{div} and @code{divu} instructions.
13372 Generate big-endian code.
13376 Generate little-endian code.
13378 @item -mio-volatile
13379 @opindex mio-volatile
13380 Tells the compiler that any variable marked with the @code{io}
13381 attribute is to be considered volatile.
13385 Causes variables to be assigned to the @code{.far} section by default.
13389 Enables the @code{leadz} (leading zero) instruction.
13393 Causes variables to be assigned to the @code{.near} section by default.
13397 Enables the @code{min} and @code{max} instructions.
13401 Enables the multiplication and multiply-accumulate instructions.
13405 Disables all the optional instructions enabled by @code{-mall-opts}.
13409 Enables the @code{repeat} and @code{erepeat} instructions, used for
13410 low-overhead looping.
13414 Causes all variables to default to the @code{.tiny} section. Note
13415 that there is a 65536 byte limit to this section. Accesses to these
13416 variables use the @code{%gp} base register.
13420 Enables the saturation instructions. Note that the compiler does not
13421 currently generate these itself, but this option is included for
13422 compatibility with other tools, like @code{as}.
13426 Link the SDRAM-based runtime instead of the default ROM-based runtime.
13430 Link the simulator runtime libraries.
13434 Link the simulator runtime libraries, excluding built-in support
13435 for reset and exception vectors and tables.
13439 Causes all functions to default to the @code{.far} section. Without
13440 this option, functions default to the @code{.near} section.
13442 @item -mtiny=@var{n}
13444 Variables that are @var{n} bytes or smaller will be allocated to the
13445 @code{.tiny} section. These variables use the @code{$gp} base
13446 register. The default for this option is 4, but note that there's a
13447 65536 byte limit to the @code{.tiny} section.
13452 @subsection MIPS Options
13453 @cindex MIPS options
13459 Generate big-endian code.
13463 Generate little-endian code. This is the default for @samp{mips*el-*-*}
13466 @item -march=@var{arch}
13468 Generate code that will run on @var{arch}, which can be the name of a
13469 generic MIPS ISA, or the name of a particular processor.
13471 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
13472 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
13473 The processor names are:
13474 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
13475 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
13476 @samp{5kc}, @samp{5kf},
13478 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
13479 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
13480 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
13481 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
13482 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
13483 @samp{loongson2e}, @samp{loongson2f},
13487 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
13488 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
13489 @samp{rm7000}, @samp{rm9000},
13490 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
13493 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
13494 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
13496 The special value @samp{from-abi} selects the
13497 most compatible architecture for the selected ABI (that is,
13498 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
13500 Native Linux/GNU toolchains also support the value @samp{native},
13501 which selects the best architecture option for the host processor.
13502 @option{-march=native} has no effect if GCC does not recognize
13505 In processor names, a final @samp{000} can be abbreviated as @samp{k}
13506 (for example, @samp{-march=r2k}). Prefixes are optional, and
13507 @samp{vr} may be written @samp{r}.
13509 Names of the form @samp{@var{n}f2_1} refer to processors with
13510 FPUs clocked at half the rate of the core, names of the form
13511 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
13512 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
13513 processors with FPUs clocked a ratio of 3:2 with respect to the core.
13514 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
13515 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
13516 accepted as synonyms for @samp{@var{n}f1_1}.
13518 GCC defines two macros based on the value of this option. The first
13519 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
13520 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
13521 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
13522 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
13523 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
13525 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
13526 above. In other words, it will have the full prefix and will not
13527 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
13528 the macro names the resolved architecture (either @samp{"mips1"} or
13529 @samp{"mips3"}). It names the default architecture when no
13530 @option{-march} option is given.
13532 @item -mtune=@var{arch}
13534 Optimize for @var{arch}. Among other things, this option controls
13535 the way instructions are scheduled, and the perceived cost of arithmetic
13536 operations. The list of @var{arch} values is the same as for
13539 When this option is not used, GCC will optimize for the processor
13540 specified by @option{-march}. By using @option{-march} and
13541 @option{-mtune} together, it is possible to generate code that will
13542 run on a family of processors, but optimize the code for one
13543 particular member of that family.
13545 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
13546 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
13547 @samp{-march} ones described above.
13551 Equivalent to @samp{-march=mips1}.
13555 Equivalent to @samp{-march=mips2}.
13559 Equivalent to @samp{-march=mips3}.
13563 Equivalent to @samp{-march=mips4}.
13567 Equivalent to @samp{-march=mips32}.
13571 Equivalent to @samp{-march=mips32r2}.
13575 Equivalent to @samp{-march=mips64}.
13579 Equivalent to @samp{-march=mips64r2}.
13584 @opindex mno-mips16
13585 Generate (do not generate) MIPS16 code. If GCC is targetting a
13586 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
13588 MIPS16 code generation can also be controlled on a per-function basis
13589 by means of @code{mips16} and @code{nomips16} attributes.
13590 @xref{Function Attributes}, for more information.
13592 @item -mflip-mips16
13593 @opindex mflip-mips16
13594 Generate MIPS16 code on alternating functions. This option is provided
13595 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
13596 not intended for ordinary use in compiling user code.
13598 @item -minterlink-mips16
13599 @itemx -mno-interlink-mips16
13600 @opindex minterlink-mips16
13601 @opindex mno-interlink-mips16
13602 Require (do not require) that non-MIPS16 code be link-compatible with
13605 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
13606 it must either use a call or an indirect jump. @option{-minterlink-mips16}
13607 therefore disables direct jumps unless GCC knows that the target of the
13608 jump is not MIPS16.
13620 Generate code for the given ABI@.
13622 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
13623 generates 64-bit code when you select a 64-bit architecture, but you
13624 can use @option{-mgp32} to get 32-bit code instead.
13626 For information about the O64 ABI, see
13627 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
13629 GCC supports a variant of the o32 ABI in which floating-point registers
13630 are 64 rather than 32 bits wide. You can select this combination with
13631 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
13632 and @samp{mfhc1} instructions and is therefore only supported for
13633 MIPS32R2 processors.
13635 The register assignments for arguments and return values remain the
13636 same, but each scalar value is passed in a single 64-bit register
13637 rather than a pair of 32-bit registers. For example, scalar
13638 floating-point values are returned in @samp{$f0} only, not a
13639 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
13640 remains the same, but all 64 bits are saved.
13643 @itemx -mno-abicalls
13645 @opindex mno-abicalls
13646 Generate (do not generate) code that is suitable for SVR4-style
13647 dynamic objects. @option{-mabicalls} is the default for SVR4-based
13652 Generate (do not generate) code that is fully position-independent,
13653 and that can therefore be linked into shared libraries. This option
13654 only affects @option{-mabicalls}.
13656 All @option{-mabicalls} code has traditionally been position-independent,
13657 regardless of options like @option{-fPIC} and @option{-fpic}. However,
13658 as an extension, the GNU toolchain allows executables to use absolute
13659 accesses for locally-binding symbols. It can also use shorter GP
13660 initialization sequences and generate direct calls to locally-defined
13661 functions. This mode is selected by @option{-mno-shared}.
13663 @option{-mno-shared} depends on binutils 2.16 or higher and generates
13664 objects that can only be linked by the GNU linker. However, the option
13665 does not affect the ABI of the final executable; it only affects the ABI
13666 of relocatable objects. Using @option{-mno-shared} will generally make
13667 executables both smaller and quicker.
13669 @option{-mshared} is the default.
13675 Assume (do not assume) that the static and dynamic linkers
13676 support PLTs and copy relocations. This option only affects
13677 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
13678 has no effect without @samp{-msym32}.
13680 You can make @option{-mplt} the default by configuring
13681 GCC with @option{--with-mips-plt}. The default is
13682 @option{-mno-plt} otherwise.
13688 Lift (do not lift) the usual restrictions on the size of the global
13691 GCC normally uses a single instruction to load values from the GOT@.
13692 While this is relatively efficient, it will only work if the GOT
13693 is smaller than about 64k. Anything larger will cause the linker
13694 to report an error such as:
13696 @cindex relocation truncated to fit (MIPS)
13698 relocation truncated to fit: R_MIPS_GOT16 foobar
13701 If this happens, you should recompile your code with @option{-mxgot}.
13702 It should then work with very large GOTs, although it will also be
13703 less efficient, since it will take three instructions to fetch the
13704 value of a global symbol.
13706 Note that some linkers can create multiple GOTs. If you have such a
13707 linker, you should only need to use @option{-mxgot} when a single object
13708 file accesses more than 64k's worth of GOT entries. Very few do.
13710 These options have no effect unless GCC is generating position
13715 Assume that general-purpose registers are 32 bits wide.
13719 Assume that general-purpose registers are 64 bits wide.
13723 Assume that floating-point registers are 32 bits wide.
13727 Assume that floating-point registers are 64 bits wide.
13730 @opindex mhard-float
13731 Use floating-point coprocessor instructions.
13734 @opindex msoft-float
13735 Do not use floating-point coprocessor instructions. Implement
13736 floating-point calculations using library calls instead.
13738 @item -msingle-float
13739 @opindex msingle-float
13740 Assume that the floating-point coprocessor only supports single-precision
13743 @item -mdouble-float
13744 @opindex mdouble-float
13745 Assume that the floating-point coprocessor supports double-precision
13746 operations. This is the default.
13752 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
13753 implement atomic memory built-in functions. When neither option is
13754 specified, GCC will use the instructions if the target architecture
13757 @option{-mllsc} is useful if the runtime environment can emulate the
13758 instructions and @option{-mno-llsc} can be useful when compiling for
13759 nonstandard ISAs. You can make either option the default by
13760 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
13761 respectively. @option{--with-llsc} is the default for some
13762 configurations; see the installation documentation for details.
13768 Use (do not use) revision 1 of the MIPS DSP ASE@.
13769 @xref{MIPS DSP Built-in Functions}. This option defines the
13770 preprocessor macro @samp{__mips_dsp}. It also defines
13771 @samp{__mips_dsp_rev} to 1.
13777 Use (do not use) revision 2 of the MIPS DSP ASE@.
13778 @xref{MIPS DSP Built-in Functions}. This option defines the
13779 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
13780 It also defines @samp{__mips_dsp_rev} to 2.
13783 @itemx -mno-smartmips
13784 @opindex msmartmips
13785 @opindex mno-smartmips
13786 Use (do not use) the MIPS SmartMIPS ASE.
13788 @item -mpaired-single
13789 @itemx -mno-paired-single
13790 @opindex mpaired-single
13791 @opindex mno-paired-single
13792 Use (do not use) paired-single floating-point instructions.
13793 @xref{MIPS Paired-Single Support}. This option requires
13794 hardware floating-point support to be enabled.
13800 Use (do not use) MIPS Digital Media Extension instructions.
13801 This option can only be used when generating 64-bit code and requires
13802 hardware floating-point support to be enabled.
13807 @opindex mno-mips3d
13808 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
13809 The option @option{-mips3d} implies @option{-mpaired-single}.
13815 Use (do not use) MT Multithreading instructions.
13819 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
13820 an explanation of the default and the way that the pointer size is
13825 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
13827 The default size of @code{int}s, @code{long}s and pointers depends on
13828 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
13829 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
13830 32-bit @code{long}s. Pointers are the same size as @code{long}s,
13831 or the same size as integer registers, whichever is smaller.
13837 Assume (do not assume) that all symbols have 32-bit values, regardless
13838 of the selected ABI@. This option is useful in combination with
13839 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
13840 to generate shorter and faster references to symbolic addresses.
13844 Put definitions of externally-visible data in a small data section
13845 if that data is no bigger than @var{num} bytes. GCC can then access
13846 the data more efficiently; see @option{-mgpopt} for details.
13848 The default @option{-G} option depends on the configuration.
13850 @item -mlocal-sdata
13851 @itemx -mno-local-sdata
13852 @opindex mlocal-sdata
13853 @opindex mno-local-sdata
13854 Extend (do not extend) the @option{-G} behavior to local data too,
13855 such as to static variables in C@. @option{-mlocal-sdata} is the
13856 default for all configurations.
13858 If the linker complains that an application is using too much small data,
13859 you might want to try rebuilding the less performance-critical parts with
13860 @option{-mno-local-sdata}. You might also want to build large
13861 libraries with @option{-mno-local-sdata}, so that the libraries leave
13862 more room for the main program.
13864 @item -mextern-sdata
13865 @itemx -mno-extern-sdata
13866 @opindex mextern-sdata
13867 @opindex mno-extern-sdata
13868 Assume (do not assume) that externally-defined data will be in
13869 a small data section if that data is within the @option{-G} limit.
13870 @option{-mextern-sdata} is the default for all configurations.
13872 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
13873 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
13874 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
13875 is placed in a small data section. If @var{Var} is defined by another
13876 module, you must either compile that module with a high-enough
13877 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
13878 definition. If @var{Var} is common, you must link the application
13879 with a high-enough @option{-G} setting.
13881 The easiest way of satisfying these restrictions is to compile
13882 and link every module with the same @option{-G} option. However,
13883 you may wish to build a library that supports several different
13884 small data limits. You can do this by compiling the library with
13885 the highest supported @option{-G} setting and additionally using
13886 @option{-mno-extern-sdata} to stop the library from making assumptions
13887 about externally-defined data.
13893 Use (do not use) GP-relative accesses for symbols that are known to be
13894 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
13895 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
13898 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
13899 might not hold the value of @code{_gp}. For example, if the code is
13900 part of a library that might be used in a boot monitor, programs that
13901 call boot monitor routines will pass an unknown value in @code{$gp}.
13902 (In such situations, the boot monitor itself would usually be compiled
13903 with @option{-G0}.)
13905 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
13906 @option{-mno-extern-sdata}.
13908 @item -membedded-data
13909 @itemx -mno-embedded-data
13910 @opindex membedded-data
13911 @opindex mno-embedded-data
13912 Allocate variables to the read-only data section first if possible, then
13913 next in the small data section if possible, otherwise in data. This gives
13914 slightly slower code than the default, but reduces the amount of RAM required
13915 when executing, and thus may be preferred for some embedded systems.
13917 @item -muninit-const-in-rodata
13918 @itemx -mno-uninit-const-in-rodata
13919 @opindex muninit-const-in-rodata
13920 @opindex mno-uninit-const-in-rodata
13921 Put uninitialized @code{const} variables in the read-only data section.
13922 This option is only meaningful in conjunction with @option{-membedded-data}.
13924 @item -mcode-readable=@var{setting}
13925 @opindex mcode-readable
13926 Specify whether GCC may generate code that reads from executable sections.
13927 There are three possible settings:
13930 @item -mcode-readable=yes
13931 Instructions may freely access executable sections. This is the
13934 @item -mcode-readable=pcrel
13935 MIPS16 PC-relative load instructions can access executable sections,
13936 but other instructions must not do so. This option is useful on 4KSc
13937 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
13938 It is also useful on processors that can be configured to have a dual
13939 instruction/data SRAM interface and that, like the M4K, automatically
13940 redirect PC-relative loads to the instruction RAM.
13942 @item -mcode-readable=no
13943 Instructions must not access executable sections. This option can be
13944 useful on targets that are configured to have a dual instruction/data
13945 SRAM interface but that (unlike the M4K) do not automatically redirect
13946 PC-relative loads to the instruction RAM.
13949 @item -msplit-addresses
13950 @itemx -mno-split-addresses
13951 @opindex msplit-addresses
13952 @opindex mno-split-addresses
13953 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
13954 relocation operators. This option has been superseded by
13955 @option{-mexplicit-relocs} but is retained for backwards compatibility.
13957 @item -mexplicit-relocs
13958 @itemx -mno-explicit-relocs
13959 @opindex mexplicit-relocs
13960 @opindex mno-explicit-relocs
13961 Use (do not use) assembler relocation operators when dealing with symbolic
13962 addresses. The alternative, selected by @option{-mno-explicit-relocs},
13963 is to use assembler macros instead.
13965 @option{-mexplicit-relocs} is the default if GCC was configured
13966 to use an assembler that supports relocation operators.
13968 @item -mcheck-zero-division
13969 @itemx -mno-check-zero-division
13970 @opindex mcheck-zero-division
13971 @opindex mno-check-zero-division
13972 Trap (do not trap) on integer division by zero.
13974 The default is @option{-mcheck-zero-division}.
13976 @item -mdivide-traps
13977 @itemx -mdivide-breaks
13978 @opindex mdivide-traps
13979 @opindex mdivide-breaks
13980 MIPS systems check for division by zero by generating either a
13981 conditional trap or a break instruction. Using traps results in
13982 smaller code, but is only supported on MIPS II and later. Also, some
13983 versions of the Linux kernel have a bug that prevents trap from
13984 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
13985 allow conditional traps on architectures that support them and
13986 @option{-mdivide-breaks} to force the use of breaks.
13988 The default is usually @option{-mdivide-traps}, but this can be
13989 overridden at configure time using @option{--with-divide=breaks}.
13990 Divide-by-zero checks can be completely disabled using
13991 @option{-mno-check-zero-division}.
13996 @opindex mno-memcpy
13997 Force (do not force) the use of @code{memcpy()} for non-trivial block
13998 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
13999 most constant-sized copies.
14002 @itemx -mno-long-calls
14003 @opindex mlong-calls
14004 @opindex mno-long-calls
14005 Disable (do not disable) use of the @code{jal} instruction. Calling
14006 functions using @code{jal} is more efficient but requires the caller
14007 and callee to be in the same 256 megabyte segment.
14009 This option has no effect on abicalls code. The default is
14010 @option{-mno-long-calls}.
14016 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
14017 instructions, as provided by the R4650 ISA@.
14020 @itemx -mno-fused-madd
14021 @opindex mfused-madd
14022 @opindex mno-fused-madd
14023 Enable (disable) use of the floating point multiply-accumulate
14024 instructions, when they are available. The default is
14025 @option{-mfused-madd}.
14027 When multiply-accumulate instructions are used, the intermediate
14028 product is calculated to infinite precision and is not subject to
14029 the FCSR Flush to Zero bit. This may be undesirable in some
14034 Tell the MIPS assembler to not run its preprocessor over user
14035 assembler files (with a @samp{.s} suffix) when assembling them.
14038 @itemx -mno-fix-r4000
14039 @opindex mfix-r4000
14040 @opindex mno-fix-r4000
14041 Work around certain R4000 CPU errata:
14044 A double-word or a variable shift may give an incorrect result if executed
14045 immediately after starting an integer division.
14047 A double-word or a variable shift may give an incorrect result if executed
14048 while an integer multiplication is in progress.
14050 An integer division may give an incorrect result if started in a delay slot
14051 of a taken branch or a jump.
14055 @itemx -mno-fix-r4400
14056 @opindex mfix-r4400
14057 @opindex mno-fix-r4400
14058 Work around certain R4400 CPU errata:
14061 A double-word or a variable shift may give an incorrect result if executed
14062 immediately after starting an integer division.
14066 @itemx -mno-fix-r10000
14067 @opindex mfix-r10000
14068 @opindex mno-fix-r10000
14069 Work around certain R10000 errata:
14072 @code{ll}/@code{sc} sequences may not behave atomically on revisions
14073 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
14076 This option can only be used if the target architecture supports
14077 branch-likely instructions. @option{-mfix-r10000} is the default when
14078 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
14082 @itemx -mno-fix-vr4120
14083 @opindex mfix-vr4120
14084 Work around certain VR4120 errata:
14087 @code{dmultu} does not always produce the correct result.
14089 @code{div} and @code{ddiv} do not always produce the correct result if one
14090 of the operands is negative.
14092 The workarounds for the division errata rely on special functions in
14093 @file{libgcc.a}. At present, these functions are only provided by
14094 the @code{mips64vr*-elf} configurations.
14096 Other VR4120 errata require a nop to be inserted between certain pairs of
14097 instructions. These errata are handled by the assembler, not by GCC itself.
14100 @opindex mfix-vr4130
14101 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
14102 workarounds are implemented by the assembler rather than by GCC,
14103 although GCC will avoid using @code{mflo} and @code{mfhi} if the
14104 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
14105 instructions are available instead.
14108 @itemx -mno-fix-sb1
14110 Work around certain SB-1 CPU core errata.
14111 (This flag currently works around the SB-1 revision 2
14112 ``F1'' and ``F2'' floating point errata.)
14114 @item -mr10k-cache-barrier=@var{setting}
14115 @opindex mr10k-cache-barrier
14116 Specify whether GCC should insert cache barriers to avoid the
14117 side-effects of speculation on R10K processors.
14119 In common with many processors, the R10K tries to predict the outcome
14120 of a conditional branch and speculatively executes instructions from
14121 the ``taken'' branch. It later aborts these instructions if the
14122 predicted outcome was wrong. However, on the R10K, even aborted
14123 instructions can have side effects.
14125 This problem only affects kernel stores and, depending on the system,
14126 kernel loads. As an example, a speculatively-executed store may load
14127 the target memory into cache and mark the cache line as dirty, even if
14128 the store itself is later aborted. If a DMA operation writes to the
14129 same area of memory before the ``dirty'' line is flushed, the cached
14130 data will overwrite the DMA-ed data. See the R10K processor manual
14131 for a full description, including other potential problems.
14133 One workaround is to insert cache barrier instructions before every memory
14134 access that might be speculatively executed and that might have side
14135 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
14136 controls GCC's implementation of this workaround. It assumes that
14137 aborted accesses to any byte in the following regions will not have
14142 the memory occupied by the current function's stack frame;
14145 the memory occupied by an incoming stack argument;
14148 the memory occupied by an object with a link-time-constant address.
14151 It is the kernel's responsibility to ensure that speculative
14152 accesses to these regions are indeed safe.
14154 If the input program contains a function declaration such as:
14160 then the implementation of @code{foo} must allow @code{j foo} and
14161 @code{jal foo} to be executed speculatively. GCC honors this
14162 restriction for functions it compiles itself. It expects non-GCC
14163 functions (such as hand-written assembly code) to do the same.
14165 The option has three forms:
14168 @item -mr10k-cache-barrier=load-store
14169 Insert a cache barrier before a load or store that might be
14170 speculatively executed and that might have side effects even
14173 @item -mr10k-cache-barrier=store
14174 Insert a cache barrier before a store that might be speculatively
14175 executed and that might have side effects even if aborted.
14177 @item -mr10k-cache-barrier=none
14178 Disable the insertion of cache barriers. This is the default setting.
14181 @item -mflush-func=@var{func}
14182 @itemx -mno-flush-func
14183 @opindex mflush-func
14184 Specifies the function to call to flush the I and D caches, or to not
14185 call any such function. If called, the function must take the same
14186 arguments as the common @code{_flush_func()}, that is, the address of the
14187 memory range for which the cache is being flushed, the size of the
14188 memory range, and the number 3 (to flush both caches). The default
14189 depends on the target GCC was configured for, but commonly is either
14190 @samp{_flush_func} or @samp{__cpu_flush}.
14192 @item mbranch-cost=@var{num}
14193 @opindex mbranch-cost
14194 Set the cost of branches to roughly @var{num} ``simple'' instructions.
14195 This cost is only a heuristic and is not guaranteed to produce
14196 consistent results across releases. A zero cost redundantly selects
14197 the default, which is based on the @option{-mtune} setting.
14199 @item -mbranch-likely
14200 @itemx -mno-branch-likely
14201 @opindex mbranch-likely
14202 @opindex mno-branch-likely
14203 Enable or disable use of Branch Likely instructions, regardless of the
14204 default for the selected architecture. By default, Branch Likely
14205 instructions may be generated if they are supported by the selected
14206 architecture. An exception is for the MIPS32 and MIPS64 architectures
14207 and processors which implement those architectures; for those, Branch
14208 Likely instructions will not be generated by default because the MIPS32
14209 and MIPS64 architectures specifically deprecate their use.
14211 @item -mfp-exceptions
14212 @itemx -mno-fp-exceptions
14213 @opindex mfp-exceptions
14214 Specifies whether FP exceptions are enabled. This affects how we schedule
14215 FP instructions for some processors. The default is that FP exceptions are
14218 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
14219 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
14222 @item -mvr4130-align
14223 @itemx -mno-vr4130-align
14224 @opindex mvr4130-align
14225 The VR4130 pipeline is two-way superscalar, but can only issue two
14226 instructions together if the first one is 8-byte aligned. When this
14227 option is enabled, GCC will align pairs of instructions that it
14228 thinks should execute in parallel.
14230 This option only has an effect when optimizing for the VR4130.
14231 It normally makes code faster, but at the expense of making it bigger.
14232 It is enabled by default at optimization level @option{-O3}.
14237 Enable (disable) generation of @code{synci} instructions on
14238 architectures that support it. The @code{synci} instructions (if
14239 enabled) will be generated when @code{__builtin___clear_cache()} is
14242 This option defaults to @code{-mno-synci}, but the default can be
14243 overridden by configuring with @code{--with-synci}.
14245 When compiling code for single processor systems, it is generally safe
14246 to use @code{synci}. However, on many multi-core (SMP) systems, it
14247 will not invalidate the instruction caches on all cores and may lead
14248 to undefined behavior.
14250 @item -mrelax-pic-calls
14251 @itemx -mno-relax-pic-calls
14252 @opindex mrelax-pic-calls
14253 Try to turn PIC calls that are normally dispatched via register
14254 @code{$25} into direct calls. This is only possible if the linker can
14255 resolve the destination at link-time and if the destination is within
14256 range for a direct call.
14258 @option{-mrelax-pic-calls} is the default if GCC was configured to use
14259 an assembler and a linker that supports the @code{.reloc} assembly
14260 directive and @code{-mexplicit-relocs} is in effect. With
14261 @code{-mno-explicit-relocs}, this optimization can be performed by the
14262 assembler and the linker alone without help from the compiler.
14264 @item -mmcount-ra-address
14265 @itemx -mno-mcount-ra-address
14266 @opindex mmcount-ra-address
14267 @opindex mno-mcount-ra-address
14268 Emit (do not emit) code that allows @code{_mcount} to modify the
14269 calling function's return address. When enabled, this option extends
14270 the usual @code{_mcount} interface with a new @var{ra-address}
14271 parameter, which has type @code{intptr_t *} and is passed in register
14272 @code{$12}. @code{_mcount} can then modify the return address by
14273 doing both of the following:
14276 Returning the new address in register @code{$31}.
14278 Storing the new address in @code{*@var{ra-address}},
14279 if @var{ra-address} is nonnull.
14282 The default is @option{-mno-mcount-ra-address}.
14287 @subsection MMIX Options
14288 @cindex MMIX Options
14290 These options are defined for the MMIX:
14294 @itemx -mno-libfuncs
14296 @opindex mno-libfuncs
14297 Specify that intrinsic library functions are being compiled, passing all
14298 values in registers, no matter the size.
14301 @itemx -mno-epsilon
14303 @opindex mno-epsilon
14304 Generate floating-point comparison instructions that compare with respect
14305 to the @code{rE} epsilon register.
14307 @item -mabi=mmixware
14309 @opindex mabi=mmixware
14311 Generate code that passes function parameters and return values that (in
14312 the called function) are seen as registers @code{$0} and up, as opposed to
14313 the GNU ABI which uses global registers @code{$231} and up.
14315 @item -mzero-extend
14316 @itemx -mno-zero-extend
14317 @opindex mzero-extend
14318 @opindex mno-zero-extend
14319 When reading data from memory in sizes shorter than 64 bits, use (do not
14320 use) zero-extending load instructions by default, rather than
14321 sign-extending ones.
14324 @itemx -mno-knuthdiv
14326 @opindex mno-knuthdiv
14327 Make the result of a division yielding a remainder have the same sign as
14328 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
14329 remainder follows the sign of the dividend. Both methods are
14330 arithmetically valid, the latter being almost exclusively used.
14332 @item -mtoplevel-symbols
14333 @itemx -mno-toplevel-symbols
14334 @opindex mtoplevel-symbols
14335 @opindex mno-toplevel-symbols
14336 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
14337 code can be used with the @code{PREFIX} assembly directive.
14341 Generate an executable in the ELF format, rather than the default
14342 @samp{mmo} format used by the @command{mmix} simulator.
14344 @item -mbranch-predict
14345 @itemx -mno-branch-predict
14346 @opindex mbranch-predict
14347 @opindex mno-branch-predict
14348 Use (do not use) the probable-branch instructions, when static branch
14349 prediction indicates a probable branch.
14351 @item -mbase-addresses
14352 @itemx -mno-base-addresses
14353 @opindex mbase-addresses
14354 @opindex mno-base-addresses
14355 Generate (do not generate) code that uses @emph{base addresses}. Using a
14356 base address automatically generates a request (handled by the assembler
14357 and the linker) for a constant to be set up in a global register. The
14358 register is used for one or more base address requests within the range 0
14359 to 255 from the value held in the register. The generally leads to short
14360 and fast code, but the number of different data items that can be
14361 addressed is limited. This means that a program that uses lots of static
14362 data may require @option{-mno-base-addresses}.
14364 @item -msingle-exit
14365 @itemx -mno-single-exit
14366 @opindex msingle-exit
14367 @opindex mno-single-exit
14368 Force (do not force) generated code to have a single exit point in each
14372 @node MN10300 Options
14373 @subsection MN10300 Options
14374 @cindex MN10300 options
14376 These @option{-m} options are defined for Matsushita MN10300 architectures:
14381 Generate code to avoid bugs in the multiply instructions for the MN10300
14382 processors. This is the default.
14384 @item -mno-mult-bug
14385 @opindex mno-mult-bug
14386 Do not generate code to avoid bugs in the multiply instructions for the
14387 MN10300 processors.
14391 Generate code which uses features specific to the AM33 processor.
14395 Do not generate code which uses features specific to the AM33 processor. This
14398 @item -mreturn-pointer-on-d0
14399 @opindex mreturn-pointer-on-d0
14400 When generating a function which returns a pointer, return the pointer
14401 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
14402 only in a0, and attempts to call such functions without a prototype
14403 would result in errors. Note that this option is on by default; use
14404 @option{-mno-return-pointer-on-d0} to disable it.
14408 Do not link in the C run-time initialization object file.
14412 Indicate to the linker that it should perform a relaxation optimization pass
14413 to shorten branches, calls and absolute memory addresses. This option only
14414 has an effect when used on the command line for the final link step.
14416 This option makes symbolic debugging impossible.
14419 @node PDP-11 Options
14420 @subsection PDP-11 Options
14421 @cindex PDP-11 Options
14423 These options are defined for the PDP-11:
14428 Use hardware FPP floating point. This is the default. (FIS floating
14429 point on the PDP-11/40 is not supported.)
14432 @opindex msoft-float
14433 Do not use hardware floating point.
14437 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
14441 Return floating-point results in memory. This is the default.
14445 Generate code for a PDP-11/40.
14449 Generate code for a PDP-11/45. This is the default.
14453 Generate code for a PDP-11/10.
14455 @item -mbcopy-builtin
14456 @opindex mbcopy-builtin
14457 Use inline @code{movmemhi} patterns for copying memory. This is the
14462 Do not use inline @code{movmemhi} patterns for copying memory.
14468 Use 16-bit @code{int}. This is the default.
14474 Use 32-bit @code{int}.
14477 @itemx -mno-float32
14479 @opindex mno-float32
14480 Use 64-bit @code{float}. This is the default.
14483 @itemx -mno-float64
14485 @opindex mno-float64
14486 Use 32-bit @code{float}.
14490 Use @code{abshi2} pattern. This is the default.
14494 Do not use @code{abshi2} pattern.
14496 @item -mbranch-expensive
14497 @opindex mbranch-expensive
14498 Pretend that branches are expensive. This is for experimenting with
14499 code generation only.
14501 @item -mbranch-cheap
14502 @opindex mbranch-cheap
14503 Do not pretend that branches are expensive. This is the default.
14507 Generate code for a system with split I&D@.
14511 Generate code for a system without split I&D@. This is the default.
14515 Use Unix assembler syntax. This is the default when configured for
14516 @samp{pdp11-*-bsd}.
14520 Use DEC assembler syntax. This is the default when configured for any
14521 PDP-11 target other than @samp{pdp11-*-bsd}.
14524 @node picoChip Options
14525 @subsection picoChip Options
14526 @cindex picoChip options
14528 These @samp{-m} options are defined for picoChip implementations:
14532 @item -mae=@var{ae_type}
14534 Set the instruction set, register set, and instruction scheduling
14535 parameters for array element type @var{ae_type}. Supported values
14536 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
14538 @option{-mae=ANY} selects a completely generic AE type. Code
14539 generated with this option will run on any of the other AE types. The
14540 code will not be as efficient as it would be if compiled for a specific
14541 AE type, and some types of operation (e.g., multiplication) will not
14542 work properly on all types of AE.
14544 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
14545 for compiled code, and is the default.
14547 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
14548 option may suffer from poor performance of byte (char) manipulation,
14549 since the DSP AE does not provide hardware support for byte load/stores.
14551 @item -msymbol-as-address
14552 Enable the compiler to directly use a symbol name as an address in a
14553 load/store instruction, without first loading it into a
14554 register. Typically, the use of this option will generate larger
14555 programs, which run faster than when the option isn't used. However, the
14556 results vary from program to program, so it is left as a user option,
14557 rather than being permanently enabled.
14559 @item -mno-inefficient-warnings
14560 Disables warnings about the generation of inefficient code. These
14561 warnings can be generated, for example, when compiling code which
14562 performs byte-level memory operations on the MAC AE type. The MAC AE has
14563 no hardware support for byte-level memory operations, so all byte
14564 load/stores must be synthesized from word load/store operations. This is
14565 inefficient and a warning will be generated indicating to the programmer
14566 that they should rewrite the code to avoid byte operations, or to target
14567 an AE type which has the necessary hardware support. This option enables
14568 the warning to be turned off.
14572 @node PowerPC Options
14573 @subsection PowerPC Options
14574 @cindex PowerPC options
14576 These are listed under @xref{RS/6000 and PowerPC Options}.
14578 @node RS/6000 and PowerPC Options
14579 @subsection IBM RS/6000 and PowerPC Options
14580 @cindex RS/6000 and PowerPC Options
14581 @cindex IBM RS/6000 and PowerPC Options
14583 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
14590 @itemx -mno-powerpc
14591 @itemx -mpowerpc-gpopt
14592 @itemx -mno-powerpc-gpopt
14593 @itemx -mpowerpc-gfxopt
14594 @itemx -mno-powerpc-gfxopt
14596 @itemx -mno-powerpc64
14600 @itemx -mno-popcntb
14602 @itemx -mno-popcntd
14610 @itemx -mno-hard-dfp
14614 @opindex mno-power2
14616 @opindex mno-powerpc
14617 @opindex mpowerpc-gpopt
14618 @opindex mno-powerpc-gpopt
14619 @opindex mpowerpc-gfxopt
14620 @opindex mno-powerpc-gfxopt
14621 @opindex mpowerpc64
14622 @opindex mno-powerpc64
14626 @opindex mno-popcntb
14628 @opindex mno-popcntd
14634 @opindex mno-mfpgpr
14636 @opindex mno-hard-dfp
14637 GCC supports two related instruction set architectures for the
14638 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
14639 instructions supported by the @samp{rios} chip set used in the original
14640 RS/6000 systems and the @dfn{PowerPC} instruction set is the
14641 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
14642 the IBM 4xx, 6xx, and follow-on microprocessors.
14644 Neither architecture is a subset of the other. However there is a
14645 large common subset of instructions supported by both. An MQ
14646 register is included in processors supporting the POWER architecture.
14648 You use these options to specify which instructions are available on the
14649 processor you are using. The default value of these options is
14650 determined when configuring GCC@. Specifying the
14651 @option{-mcpu=@var{cpu_type}} overrides the specification of these
14652 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
14653 rather than the options listed above.
14655 The @option{-mpower} option allows GCC to generate instructions that
14656 are found only in the POWER architecture and to use the MQ register.
14657 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
14658 to generate instructions that are present in the POWER2 architecture but
14659 not the original POWER architecture.
14661 The @option{-mpowerpc} option allows GCC to generate instructions that
14662 are found only in the 32-bit subset of the PowerPC architecture.
14663 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
14664 GCC to use the optional PowerPC architecture instructions in the
14665 General Purpose group, including floating-point square root. Specifying
14666 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
14667 use the optional PowerPC architecture instructions in the Graphics
14668 group, including floating-point select.
14670 The @option{-mmfcrf} option allows GCC to generate the move from
14671 condition register field instruction implemented on the POWER4
14672 processor and other processors that support the PowerPC V2.01
14674 The @option{-mpopcntb} option allows GCC to generate the popcount and
14675 double precision FP reciprocal estimate instruction implemented on the
14676 POWER5 processor and other processors that support the PowerPC V2.02
14678 The @option{-mpopcntd} option allows GCC to generate the popcount
14679 instruction implemented on the POWER7 processor and other processors
14680 that support the PowerPC V2.06 architecture.
14681 The @option{-mfprnd} option allows GCC to generate the FP round to
14682 integer instructions implemented on the POWER5+ processor and other
14683 processors that support the PowerPC V2.03 architecture.
14684 The @option{-mcmpb} option allows GCC to generate the compare bytes
14685 instruction implemented on the POWER6 processor and other processors
14686 that support the PowerPC V2.05 architecture.
14687 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
14688 general purpose register instructions implemented on the POWER6X
14689 processor and other processors that support the extended PowerPC V2.05
14691 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
14692 point instructions implemented on some POWER processors.
14694 The @option{-mpowerpc64} option allows GCC to generate the additional
14695 64-bit instructions that are found in the full PowerPC64 architecture
14696 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
14697 @option{-mno-powerpc64}.
14699 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
14700 will use only the instructions in the common subset of both
14701 architectures plus some special AIX common-mode calls, and will not use
14702 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
14703 permits GCC to use any instruction from either architecture and to
14704 allow use of the MQ register; specify this for the Motorola MPC601.
14706 @item -mnew-mnemonics
14707 @itemx -mold-mnemonics
14708 @opindex mnew-mnemonics
14709 @opindex mold-mnemonics
14710 Select which mnemonics to use in the generated assembler code. With
14711 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
14712 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
14713 assembler mnemonics defined for the POWER architecture. Instructions
14714 defined in only one architecture have only one mnemonic; GCC uses that
14715 mnemonic irrespective of which of these options is specified.
14717 GCC defaults to the mnemonics appropriate for the architecture in
14718 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
14719 value of these option. Unless you are building a cross-compiler, you
14720 should normally not specify either @option{-mnew-mnemonics} or
14721 @option{-mold-mnemonics}, but should instead accept the default.
14723 @item -mcpu=@var{cpu_type}
14725 Set architecture type, register usage, choice of mnemonics, and
14726 instruction scheduling parameters for machine type @var{cpu_type}.
14727 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
14728 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
14729 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
14730 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
14731 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
14732 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
14733 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{ec603e}, @samp{G3},
14734 @samp{G4}, @samp{G5}, @samp{power}, @samp{power2}, @samp{power3},
14735 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x},
14736 @samp{power7}, @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
14737 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
14739 @option{-mcpu=common} selects a completely generic processor. Code
14740 generated under this option will run on any POWER or PowerPC processor.
14741 GCC will use only the instructions in the common subset of both
14742 architectures, and will not use the MQ register. GCC assumes a generic
14743 processor model for scheduling purposes.
14745 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
14746 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
14747 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
14748 types, with an appropriate, generic processor model assumed for
14749 scheduling purposes.
14751 The other options specify a specific processor. Code generated under
14752 those options will run best on that processor, and may not run at all on
14755 The @option{-mcpu} options automatically enable or disable the
14758 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
14759 -mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol
14760 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
14761 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx}
14763 The particular options set for any particular CPU will vary between
14764 compiler versions, depending on what setting seems to produce optimal
14765 code for that CPU; it doesn't necessarily reflect the actual hardware's
14766 capabilities. If you wish to set an individual option to a particular
14767 value, you may specify it after the @option{-mcpu} option, like
14768 @samp{-mcpu=970 -mno-altivec}.
14770 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
14771 not enabled or disabled by the @option{-mcpu} option at present because
14772 AIX does not have full support for these options. You may still
14773 enable or disable them individually if you're sure it'll work in your
14776 @item -mtune=@var{cpu_type}
14778 Set the instruction scheduling parameters for machine type
14779 @var{cpu_type}, but do not set the architecture type, register usage, or
14780 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
14781 values for @var{cpu_type} are used for @option{-mtune} as for
14782 @option{-mcpu}. If both are specified, the code generated will use the
14783 architecture, registers, and mnemonics set by @option{-mcpu}, but the
14784 scheduling parameters set by @option{-mtune}.
14790 Generate code to compute division as reciprocal estimate and iterative
14791 refinement, creating opportunities for increased throughput. This
14792 feature requires: optional PowerPC Graphics instruction set for single
14793 precision and FRE instruction for double precision, assuming divides
14794 cannot generate user-visible traps, and the domain values not include
14795 Infinities, denormals or zero denominator.
14798 @itemx -mno-altivec
14800 @opindex mno-altivec
14801 Generate code that uses (does not use) AltiVec instructions, and also
14802 enable the use of built-in functions that allow more direct access to
14803 the AltiVec instruction set. You may also need to set
14804 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
14810 @opindex mno-vrsave
14811 Generate VRSAVE instructions when generating AltiVec code.
14813 @item -mgen-cell-microcode
14814 @opindex mgen-cell-microcode
14815 Generate Cell microcode instructions
14817 @item -mwarn-cell-microcode
14818 @opindex mwarn-cell-microcode
14819 Warning when a Cell microcode instruction is going to emitted. An example
14820 of a Cell microcode instruction is a variable shift.
14823 @opindex msecure-plt
14824 Generate code that allows ld and ld.so to build executables and shared
14825 libraries with non-exec .plt and .got sections. This is a PowerPC
14826 32-bit SYSV ABI option.
14830 Generate code that uses a BSS .plt section that ld.so fills in, and
14831 requires .plt and .got sections that are both writable and executable.
14832 This is a PowerPC 32-bit SYSV ABI option.
14838 This switch enables or disables the generation of ISEL instructions.
14840 @item -misel=@var{yes/no}
14841 This switch has been deprecated. Use @option{-misel} and
14842 @option{-mno-isel} instead.
14848 This switch enables or disables the generation of SPE simd
14854 @opindex mno-paired
14855 This switch enables or disables the generation of PAIRED simd
14858 @item -mspe=@var{yes/no}
14859 This option has been deprecated. Use @option{-mspe} and
14860 @option{-mno-spe} instead.
14866 Generate code that uses (does not use) vector/scalar (VSX)
14867 instructions, and also enable the use of built-in functions that allow
14868 more direct access to the VSX instruction set.
14870 @item -mfloat-gprs=@var{yes/single/double/no}
14871 @itemx -mfloat-gprs
14872 @opindex mfloat-gprs
14873 This switch enables or disables the generation of floating point
14874 operations on the general purpose registers for architectures that
14877 The argument @var{yes} or @var{single} enables the use of
14878 single-precision floating point operations.
14880 The argument @var{double} enables the use of single and
14881 double-precision floating point operations.
14883 The argument @var{no} disables floating point operations on the
14884 general purpose registers.
14886 This option is currently only available on the MPC854x.
14892 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
14893 targets (including GNU/Linux). The 32-bit environment sets int, long
14894 and pointer to 32 bits and generates code that runs on any PowerPC
14895 variant. The 64-bit environment sets int to 32 bits and long and
14896 pointer to 64 bits, and generates code for PowerPC64, as for
14897 @option{-mpowerpc64}.
14900 @itemx -mno-fp-in-toc
14901 @itemx -mno-sum-in-toc
14902 @itemx -mminimal-toc
14904 @opindex mno-fp-in-toc
14905 @opindex mno-sum-in-toc
14906 @opindex mminimal-toc
14907 Modify generation of the TOC (Table Of Contents), which is created for
14908 every executable file. The @option{-mfull-toc} option is selected by
14909 default. In that case, GCC will allocate at least one TOC entry for
14910 each unique non-automatic variable reference in your program. GCC
14911 will also place floating-point constants in the TOC@. However, only
14912 16,384 entries are available in the TOC@.
14914 If you receive a linker error message that saying you have overflowed
14915 the available TOC space, you can reduce the amount of TOC space used
14916 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
14917 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
14918 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
14919 generate code to calculate the sum of an address and a constant at
14920 run-time instead of putting that sum into the TOC@. You may specify one
14921 or both of these options. Each causes GCC to produce very slightly
14922 slower and larger code at the expense of conserving TOC space.
14924 If you still run out of space in the TOC even when you specify both of
14925 these options, specify @option{-mminimal-toc} instead. This option causes
14926 GCC to make only one TOC entry for every file. When you specify this
14927 option, GCC will produce code that is slower and larger but which
14928 uses extremely little TOC space. You may wish to use this option
14929 only on files that contain less frequently executed code.
14935 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
14936 @code{long} type, and the infrastructure needed to support them.
14937 Specifying @option{-maix64} implies @option{-mpowerpc64} and
14938 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
14939 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
14942 @itemx -mno-xl-compat
14943 @opindex mxl-compat
14944 @opindex mno-xl-compat
14945 Produce code that conforms more closely to IBM XL compiler semantics
14946 when using AIX-compatible ABI@. Pass floating-point arguments to
14947 prototyped functions beyond the register save area (RSA) on the stack
14948 in addition to argument FPRs. Do not assume that most significant
14949 double in 128-bit long double value is properly rounded when comparing
14950 values and converting to double. Use XL symbol names for long double
14953 The AIX calling convention was extended but not initially documented to
14954 handle an obscure K&R C case of calling a function that takes the
14955 address of its arguments with fewer arguments than declared. IBM XL
14956 compilers access floating point arguments which do not fit in the
14957 RSA from the stack when a subroutine is compiled without
14958 optimization. Because always storing floating-point arguments on the
14959 stack is inefficient and rarely needed, this option is not enabled by
14960 default and only is necessary when calling subroutines compiled by IBM
14961 XL compilers without optimization.
14965 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
14966 application written to use message passing with special startup code to
14967 enable the application to run. The system must have PE installed in the
14968 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
14969 must be overridden with the @option{-specs=} option to specify the
14970 appropriate directory location. The Parallel Environment does not
14971 support threads, so the @option{-mpe} option and the @option{-pthread}
14972 option are incompatible.
14974 @item -malign-natural
14975 @itemx -malign-power
14976 @opindex malign-natural
14977 @opindex malign-power
14978 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
14979 @option{-malign-natural} overrides the ABI-defined alignment of larger
14980 types, such as floating-point doubles, on their natural size-based boundary.
14981 The option @option{-malign-power} instructs GCC to follow the ABI-specified
14982 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
14984 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
14988 @itemx -mhard-float
14989 @opindex msoft-float
14990 @opindex mhard-float
14991 Generate code that does not use (uses) the floating-point register set.
14992 Software floating point emulation is provided if you use the
14993 @option{-msoft-float} option, and pass the option to GCC when linking.
14995 @item -msingle-float
14996 @itemx -mdouble-float
14997 @opindex msingle-float
14998 @opindex mdouble-float
14999 Generate code for single or double-precision floating point operations.
15000 @option{-mdouble-float} implies @option{-msingle-float}.
15003 @opindex msimple-fpu
15004 Do not generate sqrt and div instructions for hardware floating point unit.
15008 Specify type of floating point unit. Valid values are @var{sp_lite}
15009 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
15010 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
15011 and @var{dp_full} (equivalent to -mdouble-float).
15014 @opindex mxilinx-fpu
15015 Perform optimizations for floating point unit on Xilinx PPC 405/440.
15018 @itemx -mno-multiple
15020 @opindex mno-multiple
15021 Generate code that uses (does not use) the load multiple word
15022 instructions and the store multiple word instructions. These
15023 instructions are generated by default on POWER systems, and not
15024 generated on PowerPC systems. Do not use @option{-mmultiple} on little
15025 endian PowerPC systems, since those instructions do not work when the
15026 processor is in little endian mode. The exceptions are PPC740 and
15027 PPC750 which permit the instructions usage in little endian mode.
15032 @opindex mno-string
15033 Generate code that uses (does not use) the load string instructions
15034 and the store string word instructions to save multiple registers and
15035 do small block moves. These instructions are generated by default on
15036 POWER systems, and not generated on PowerPC systems. Do not use
15037 @option{-mstring} on little endian PowerPC systems, since those
15038 instructions do not work when the processor is in little endian mode.
15039 The exceptions are PPC740 and PPC750 which permit the instructions
15040 usage in little endian mode.
15045 @opindex mno-update
15046 Generate code that uses (does not use) the load or store instructions
15047 that update the base register to the address of the calculated memory
15048 location. These instructions are generated by default. If you use
15049 @option{-mno-update}, there is a small window between the time that the
15050 stack pointer is updated and the address of the previous frame is
15051 stored, which means code that walks the stack frame across interrupts or
15052 signals may get corrupted data.
15054 @item -mavoid-indexed-addresses
15055 @itemx -mno-avoid-indexed-addresses
15056 @opindex mavoid-indexed-addresses
15057 @opindex mno-avoid-indexed-addresses
15058 Generate code that tries to avoid (not avoid) the use of indexed load
15059 or store instructions. These instructions can incur a performance
15060 penalty on Power6 processors in certain situations, such as when
15061 stepping through large arrays that cross a 16M boundary. This option
15062 is enabled by default when targetting Power6 and disabled otherwise.
15065 @itemx -mno-fused-madd
15066 @opindex mfused-madd
15067 @opindex mno-fused-madd
15068 Generate code that uses (does not use) the floating point multiply and
15069 accumulate instructions. These instructions are generated by default if
15070 hardware floating is used.
15076 Generate code that uses (does not use) the half-word multiply and
15077 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
15078 These instructions are generated by default when targetting those
15085 Generate code that uses (does not use) the string-search @samp{dlmzb}
15086 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
15087 generated by default when targetting those processors.
15089 @item -mno-bit-align
15091 @opindex mno-bit-align
15092 @opindex mbit-align
15093 On System V.4 and embedded PowerPC systems do not (do) force structures
15094 and unions that contain bit-fields to be aligned to the base type of the
15097 For example, by default a structure containing nothing but 8
15098 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
15099 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
15100 the structure would be aligned to a 1 byte boundary and be one byte in
15103 @item -mno-strict-align
15104 @itemx -mstrict-align
15105 @opindex mno-strict-align
15106 @opindex mstrict-align
15107 On System V.4 and embedded PowerPC systems do not (do) assume that
15108 unaligned memory references will be handled by the system.
15110 @item -mrelocatable
15111 @itemx -mno-relocatable
15112 @opindex mrelocatable
15113 @opindex mno-relocatable
15114 On embedded PowerPC systems generate code that allows (does not allow)
15115 the program to be relocated to a different address at runtime. If you
15116 use @option{-mrelocatable} on any module, all objects linked together must
15117 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
15119 @item -mrelocatable-lib
15120 @itemx -mno-relocatable-lib
15121 @opindex mrelocatable-lib
15122 @opindex mno-relocatable-lib
15123 On embedded PowerPC systems generate code that allows (does not allow)
15124 the program to be relocated to a different address at runtime. Modules
15125 compiled with @option{-mrelocatable-lib} can be linked with either modules
15126 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
15127 with modules compiled with the @option{-mrelocatable} options.
15133 On System V.4 and embedded PowerPC systems do not (do) assume that
15134 register 2 contains a pointer to a global area pointing to the addresses
15135 used in the program.
15138 @itemx -mlittle-endian
15140 @opindex mlittle-endian
15141 On System V.4 and embedded PowerPC systems compile code for the
15142 processor in little endian mode. The @option{-mlittle-endian} option is
15143 the same as @option{-mlittle}.
15146 @itemx -mbig-endian
15148 @opindex mbig-endian
15149 On System V.4 and embedded PowerPC systems compile code for the
15150 processor in big endian mode. The @option{-mbig-endian} option is
15151 the same as @option{-mbig}.
15153 @item -mdynamic-no-pic
15154 @opindex mdynamic-no-pic
15155 On Darwin and Mac OS X systems, compile code so that it is not
15156 relocatable, but that its external references are relocatable. The
15157 resulting code is suitable for applications, but not shared
15160 @item -mprioritize-restricted-insns=@var{priority}
15161 @opindex mprioritize-restricted-insns
15162 This option controls the priority that is assigned to
15163 dispatch-slot restricted instructions during the second scheduling
15164 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
15165 @var{no/highest/second-highest} priority to dispatch slot restricted
15168 @item -msched-costly-dep=@var{dependence_type}
15169 @opindex msched-costly-dep
15170 This option controls which dependences are considered costly
15171 by the target during instruction scheduling. The argument
15172 @var{dependence_type} takes one of the following values:
15173 @var{no}: no dependence is costly,
15174 @var{all}: all dependences are costly,
15175 @var{true_store_to_load}: a true dependence from store to load is costly,
15176 @var{store_to_load}: any dependence from store to load is costly,
15177 @var{number}: any dependence which latency >= @var{number} is costly.
15179 @item -minsert-sched-nops=@var{scheme}
15180 @opindex minsert-sched-nops
15181 This option controls which nop insertion scheme will be used during
15182 the second scheduling pass. The argument @var{scheme} takes one of the
15184 @var{no}: Don't insert nops.
15185 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
15186 according to the scheduler's grouping.
15187 @var{regroup_exact}: Insert nops to force costly dependent insns into
15188 separate groups. Insert exactly as many nops as needed to force an insn
15189 to a new group, according to the estimated processor grouping.
15190 @var{number}: Insert nops to force costly dependent insns into
15191 separate groups. Insert @var{number} nops to force an insn to a new group.
15194 @opindex mcall-sysv
15195 On System V.4 and embedded PowerPC systems compile code using calling
15196 conventions that adheres to the March 1995 draft of the System V
15197 Application Binary Interface, PowerPC processor supplement. This is the
15198 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
15200 @item -mcall-sysv-eabi
15202 @opindex mcall-sysv-eabi
15203 @opindex mcall-eabi
15204 Specify both @option{-mcall-sysv} and @option{-meabi} options.
15206 @item -mcall-sysv-noeabi
15207 @opindex mcall-sysv-noeabi
15208 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
15210 @item -mcall-aixdesc
15212 On System V.4 and embedded PowerPC systems compile code for the AIX
15216 @opindex mcall-linux
15217 On System V.4 and embedded PowerPC systems compile code for the
15218 Linux-based GNU system.
15222 On System V.4 and embedded PowerPC systems compile code for the
15223 Hurd-based GNU system.
15225 @item -mcall-freebsd
15226 @opindex mcall-freebsd
15227 On System V.4 and embedded PowerPC systems compile code for the
15228 FreeBSD operating system.
15230 @item -mcall-netbsd
15231 @opindex mcall-netbsd
15232 On System V.4 and embedded PowerPC systems compile code for the
15233 NetBSD operating system.
15235 @item -mcall-openbsd
15236 @opindex mcall-netbsd
15237 On System V.4 and embedded PowerPC systems compile code for the
15238 OpenBSD operating system.
15240 @item -maix-struct-return
15241 @opindex maix-struct-return
15242 Return all structures in memory (as specified by the AIX ABI)@.
15244 @item -msvr4-struct-return
15245 @opindex msvr4-struct-return
15246 Return structures smaller than 8 bytes in registers (as specified by the
15249 @item -mabi=@var{abi-type}
15251 Extend the current ABI with a particular extension, or remove such extension.
15252 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
15253 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
15257 Extend the current ABI with SPE ABI extensions. This does not change
15258 the default ABI, instead it adds the SPE ABI extensions to the current
15262 @opindex mabi=no-spe
15263 Disable Booke SPE ABI extensions for the current ABI@.
15265 @item -mabi=ibmlongdouble
15266 @opindex mabi=ibmlongdouble
15267 Change the current ABI to use IBM extended precision long double.
15268 This is a PowerPC 32-bit SYSV ABI option.
15270 @item -mabi=ieeelongdouble
15271 @opindex mabi=ieeelongdouble
15272 Change the current ABI to use IEEE extended precision long double.
15273 This is a PowerPC 32-bit Linux ABI option.
15276 @itemx -mno-prototype
15277 @opindex mprototype
15278 @opindex mno-prototype
15279 On System V.4 and embedded PowerPC systems assume that all calls to
15280 variable argument functions are properly prototyped. Otherwise, the
15281 compiler must insert an instruction before every non prototyped call to
15282 set or clear bit 6 of the condition code register (@var{CR}) to
15283 indicate whether floating point values were passed in the floating point
15284 registers in case the function takes a variable arguments. With
15285 @option{-mprototype}, only calls to prototyped variable argument functions
15286 will set or clear the bit.
15290 On embedded PowerPC systems, assume that the startup module is called
15291 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
15292 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
15297 On embedded PowerPC systems, assume that the startup module is called
15298 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
15303 On embedded PowerPC systems, assume that the startup module is called
15304 @file{crt0.o} and the standard C libraries are @file{libads.a} and
15307 @item -myellowknife
15308 @opindex myellowknife
15309 On embedded PowerPC systems, assume that the startup module is called
15310 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
15315 On System V.4 and embedded PowerPC systems, specify that you are
15316 compiling for a VxWorks system.
15320 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
15321 header to indicate that @samp{eabi} extended relocations are used.
15327 On System V.4 and embedded PowerPC systems do (do not) adhere to the
15328 Embedded Applications Binary Interface (eabi) which is a set of
15329 modifications to the System V.4 specifications. Selecting @option{-meabi}
15330 means that the stack is aligned to an 8 byte boundary, a function
15331 @code{__eabi} is called to from @code{main} to set up the eabi
15332 environment, and the @option{-msdata} option can use both @code{r2} and
15333 @code{r13} to point to two separate small data areas. Selecting
15334 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
15335 do not call an initialization function from @code{main}, and the
15336 @option{-msdata} option will only use @code{r13} to point to a single
15337 small data area. The @option{-meabi} option is on by default if you
15338 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
15341 @opindex msdata=eabi
15342 On System V.4 and embedded PowerPC systems, put small initialized
15343 @code{const} global and static data in the @samp{.sdata2} section, which
15344 is pointed to by register @code{r2}. Put small initialized
15345 non-@code{const} global and static data in the @samp{.sdata} section,
15346 which is pointed to by register @code{r13}. Put small uninitialized
15347 global and static data in the @samp{.sbss} section, which is adjacent to
15348 the @samp{.sdata} section. The @option{-msdata=eabi} option is
15349 incompatible with the @option{-mrelocatable} option. The
15350 @option{-msdata=eabi} option also sets the @option{-memb} option.
15353 @opindex msdata=sysv
15354 On System V.4 and embedded PowerPC systems, put small global and static
15355 data in the @samp{.sdata} section, which is pointed to by register
15356 @code{r13}. Put small uninitialized global and static data in the
15357 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
15358 The @option{-msdata=sysv} option is incompatible with the
15359 @option{-mrelocatable} option.
15361 @item -msdata=default
15363 @opindex msdata=default
15365 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
15366 compile code the same as @option{-msdata=eabi}, otherwise compile code the
15367 same as @option{-msdata=sysv}.
15370 @opindex msdata=data
15371 On System V.4 and embedded PowerPC systems, put small global
15372 data in the @samp{.sdata} section. Put small uninitialized global
15373 data in the @samp{.sbss} section. Do not use register @code{r13}
15374 to address small data however. This is the default behavior unless
15375 other @option{-msdata} options are used.
15379 @opindex msdata=none
15381 On embedded PowerPC systems, put all initialized global and static data
15382 in the @samp{.data} section, and all uninitialized data in the
15383 @samp{.bss} section.
15387 @cindex smaller data references (PowerPC)
15388 @cindex .sdata/.sdata2 references (PowerPC)
15389 On embedded PowerPC systems, put global and static items less than or
15390 equal to @var{num} bytes into the small data or bss sections instead of
15391 the normal data or bss section. By default, @var{num} is 8. The
15392 @option{-G @var{num}} switch is also passed to the linker.
15393 All modules should be compiled with the same @option{-G @var{num}} value.
15396 @itemx -mno-regnames
15398 @opindex mno-regnames
15399 On System V.4 and embedded PowerPC systems do (do not) emit register
15400 names in the assembly language output using symbolic forms.
15403 @itemx -mno-longcall
15405 @opindex mno-longcall
15406 By default assume that all calls are far away so that a longer more
15407 expensive calling sequence is required. This is required for calls
15408 further than 32 megabytes (33,554,432 bytes) from the current location.
15409 A short call will be generated if the compiler knows
15410 the call cannot be that far away. This setting can be overridden by
15411 the @code{shortcall} function attribute, or by @code{#pragma
15414 Some linkers are capable of detecting out-of-range calls and generating
15415 glue code on the fly. On these systems, long calls are unnecessary and
15416 generate slower code. As of this writing, the AIX linker can do this,
15417 as can the GNU linker for PowerPC/64. It is planned to add this feature
15418 to the GNU linker for 32-bit PowerPC systems as well.
15420 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
15421 callee, L42'', plus a ``branch island'' (glue code). The two target
15422 addresses represent the callee and the ``branch island''. The
15423 Darwin/PPC linker will prefer the first address and generate a ``bl
15424 callee'' if the PPC ``bl'' instruction will reach the callee directly;
15425 otherwise, the linker will generate ``bl L42'' to call the ``branch
15426 island''. The ``branch island'' is appended to the body of the
15427 calling function; it computes the full 32-bit address of the callee
15430 On Mach-O (Darwin) systems, this option directs the compiler emit to
15431 the glue for every direct call, and the Darwin linker decides whether
15432 to use or discard it.
15434 In the future, we may cause GCC to ignore all longcall specifications
15435 when the linker is known to generate glue.
15437 @item -mtls-markers
15438 @itemx -mno-tls-markers
15439 @opindex mtls-markers
15440 @opindex mno-tls-markers
15441 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
15442 specifying the function argument. The relocation allows ld to
15443 reliably associate function call with argument setup instructions for
15444 TLS optimization, which in turn allows gcc to better schedule the
15449 Adds support for multithreading with the @dfn{pthreads} library.
15450 This option sets flags for both the preprocessor and linker.
15455 @subsection RX Options
15458 These command line options are defined for RX targets:
15461 @item -m64bit-doubles
15462 @itemx -m32bit-doubles
15463 @opindex m64bit-doubles
15464 @opindex m32bit-doubles
15465 Make the @code{double} data type be 64-bits (@option{-m64bit-doubles})
15466 or 32-bits (@option{-m32bit-doubles}) in size. The default is
15467 @option{-m32bit-doubles}. @emph{Note} RX floating point hardware only
15468 works on 32-bit values, which is why the default is
15469 @option{-m32bit-doubles}.
15475 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
15476 floating point hardware. The default is enabled for the @var{RX600}
15477 series and disabled for the @var{RX200} series.
15479 Floating point instructions will only be generated for 32-bit floating
15480 point values however, so if the @option{-m64bit-doubles} option is in
15481 use then the FPU hardware will not be used for doubles.
15483 @emph{Note} If the @option{-fpu} option is enabled then
15484 @option{-funsafe-math-optimizations} is also enabled automatically.
15485 This is because the RX FPU instructions are themselves unsafe.
15487 @item -mcpu=@var{name}
15488 @itemx -patch=@var{name}
15491 Selects the type of RX CPU to be targeted. Currently three types are
15492 supported, the generic @var{RX600} and @var{RX200} series hardware and
15493 the specific @var{RX610} cpu. The default is @var{RX600}.
15495 The only difference between @var{RX600} and @var{RX610} is that the
15496 @var{RX610} does not support the @code{MVTIPL} instruction.
15498 The @var{RX200} series does not have a hardware floating point unit
15499 and so @option{-nofpu} is enabled by default when this type is
15502 @item -mbig-endian-data
15503 @itemx -mlittle-endian-data
15504 @opindex mbig-endian-data
15505 @opindex mlittle-endian-data
15506 Store data (but not code) in the big-endian format. The default is
15507 @option{-mlittle-endian-data}, ie to store data in the little endian
15510 @item -msmall-data-limit=@var{N}
15511 @opindex msmall-data-limit
15512 Specifies the maximum size in bytes of global and static variables
15513 which can be placed into the small data area. Using the small data
15514 area can lead to smaller and faster code, but the size of area is
15515 limited and it is up to the programmer to ensure that the area does
15516 not overflow. Also when the small data area is used one of the RX's
15517 registers (@code{r13}) is reserved for use pointing to this area, so
15518 it is no longer available for use by the compiler. This could result
15519 in slower and/or larger code if variables which once could have been
15520 held in @code{r13} are now pushed onto the stack.
15522 Note, common variables (variables which have not been initialised) and
15523 constants are not placed into the small data area as they are assigned
15524 to other sections in the output executable.
15526 The default value is zero, which disables this feature. Note, this
15527 feature is not enabled by default with higher optimization levels
15528 (@option{-O2} etc) because of the potentially detrimental effects of
15529 reserving register @code{r13}. It is up to the programmer to
15530 experiment and discover whether this feature is of benefit to their
15537 Use the simulator runtime. The default is to use the libgloss board
15540 @item -mas100-syntax
15541 @itemx -mno-as100-syntax
15542 @opindex mas100-syntax
15543 @opindex mno-as100-syntax
15544 When generating assembler output use a syntax that is compatible with
15545 Renesas's AS100 assembler. This syntax can also be handled by the GAS
15546 assembler but it has some restrictions so generating it is not the
15549 @item -mmax-constant-size=@var{N}
15550 @opindex mmax-constant-size
15551 Specifies the maxium size, in bytes, of a constant that can be used as
15552 an operand in a RX instruction. Although the RX instruction set does
15553 allow constants of up to 4 bytes in length to be used in instructions,
15554 a longer value equates to a longer instruction. Thus in some
15555 circumstances it can be beneficial to restrict the size of constants
15556 that are used in instructions. Constants that are too big are instead
15557 placed into a constant pool and referenced via register indirection.
15559 The value @var{N} can be between 0 and 4. A value of 0 (the default)
15560 or 4 means that constants of any size are allowed.
15564 Enable linker relaxation. Linker relaxation is a process whereby the
15565 linker will attempt to reduce the size of a program by finding shorter
15566 versions of various instructions. Disabled by default.
15568 @item -mint-register=@var{N}
15569 @opindex mint-register
15570 Specify the number of registers to reserve for fast interrupt handler
15571 functions. The value @var{N} can be between 0 and 4. A value of 1
15572 means that register @code{r13} will be reserved for ther exclusive use
15573 of fast interrupt handlers. A value of 2 reserves @code{r13} and
15574 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
15575 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
15576 A value of 0, the default, does not reserve any registers.
15578 @item -msave-acc-in-interrupts
15579 @opindex msave-acc-in-interrupts
15580 Specifies that interrupt handler functions should preserve the
15581 accumulator register. This is only necessary if normal code might use
15582 the accumulator register, for example because it performs 64-bit
15583 multiplications. The default is to ignore the accumulator as this
15584 makes the interrupt handlers faster.
15588 @emph{Note:} The generic GCC command line @option{-ffixed-@var{reg}}
15589 has special significance to the RX port when used with the
15590 @code{interrupt} function attribute. This attribute indicates a
15591 function intended to process fast interrupts. GCC will will ensure
15592 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
15593 and/or @code{r13} and only provided that the normal use of the
15594 corresponding registers have been restricted via the
15595 @option{-ffixed-@var{reg}} or @option{-mint-register} command line
15598 @node S/390 and zSeries Options
15599 @subsection S/390 and zSeries Options
15600 @cindex S/390 and zSeries Options
15602 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
15606 @itemx -msoft-float
15607 @opindex mhard-float
15608 @opindex msoft-float
15609 Use (do not use) the hardware floating-point instructions and registers
15610 for floating-point operations. When @option{-msoft-float} is specified,
15611 functions in @file{libgcc.a} will be used to perform floating-point
15612 operations. When @option{-mhard-float} is specified, the compiler
15613 generates IEEE floating-point instructions. This is the default.
15616 @itemx -mno-hard-dfp
15618 @opindex mno-hard-dfp
15619 Use (do not use) the hardware decimal-floating-point instructions for
15620 decimal-floating-point operations. When @option{-mno-hard-dfp} is
15621 specified, functions in @file{libgcc.a} will be used to perform
15622 decimal-floating-point operations. When @option{-mhard-dfp} is
15623 specified, the compiler generates decimal-floating-point hardware
15624 instructions. This is the default for @option{-march=z9-ec} or higher.
15626 @item -mlong-double-64
15627 @itemx -mlong-double-128
15628 @opindex mlong-double-64
15629 @opindex mlong-double-128
15630 These switches control the size of @code{long double} type. A size
15631 of 64bit makes the @code{long double} type equivalent to the @code{double}
15632 type. This is the default.
15635 @itemx -mno-backchain
15636 @opindex mbackchain
15637 @opindex mno-backchain
15638 Store (do not store) the address of the caller's frame as backchain pointer
15639 into the callee's stack frame.
15640 A backchain may be needed to allow debugging using tools that do not understand
15641 DWARF-2 call frame information.
15642 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
15643 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
15644 the backchain is placed into the topmost word of the 96/160 byte register
15647 In general, code compiled with @option{-mbackchain} is call-compatible with
15648 code compiled with @option{-mmo-backchain}; however, use of the backchain
15649 for debugging purposes usually requires that the whole binary is built with
15650 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
15651 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
15652 to build a linux kernel use @option{-msoft-float}.
15654 The default is to not maintain the backchain.
15656 @item -mpacked-stack
15657 @itemx -mno-packed-stack
15658 @opindex mpacked-stack
15659 @opindex mno-packed-stack
15660 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
15661 specified, the compiler uses the all fields of the 96/160 byte register save
15662 area only for their default purpose; unused fields still take up stack space.
15663 When @option{-mpacked-stack} is specified, register save slots are densely
15664 packed at the top of the register save area; unused space is reused for other
15665 purposes, allowing for more efficient use of the available stack space.
15666 However, when @option{-mbackchain} is also in effect, the topmost word of
15667 the save area is always used to store the backchain, and the return address
15668 register is always saved two words below the backchain.
15670 As long as the stack frame backchain is not used, code generated with
15671 @option{-mpacked-stack} is call-compatible with code generated with
15672 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
15673 S/390 or zSeries generated code that uses the stack frame backchain at run
15674 time, not just for debugging purposes. Such code is not call-compatible
15675 with code compiled with @option{-mpacked-stack}. Also, note that the
15676 combination of @option{-mbackchain},
15677 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
15678 to build a linux kernel use @option{-msoft-float}.
15680 The default is to not use the packed stack layout.
15683 @itemx -mno-small-exec
15684 @opindex msmall-exec
15685 @opindex mno-small-exec
15686 Generate (or do not generate) code using the @code{bras} instruction
15687 to do subroutine calls.
15688 This only works reliably if the total executable size does not
15689 exceed 64k. The default is to use the @code{basr} instruction instead,
15690 which does not have this limitation.
15696 When @option{-m31} is specified, generate code compliant to the
15697 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
15698 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
15699 particular to generate 64-bit instructions. For the @samp{s390}
15700 targets, the default is @option{-m31}, while the @samp{s390x}
15701 targets default to @option{-m64}.
15707 When @option{-mzarch} is specified, generate code using the
15708 instructions available on z/Architecture.
15709 When @option{-mesa} is specified, generate code using the
15710 instructions available on ESA/390. Note that @option{-mesa} is
15711 not possible with @option{-m64}.
15712 When generating code compliant to the GNU/Linux for S/390 ABI,
15713 the default is @option{-mesa}. When generating code compliant
15714 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
15720 Generate (or do not generate) code using the @code{mvcle} instruction
15721 to perform block moves. When @option{-mno-mvcle} is specified,
15722 use a @code{mvc} loop instead. This is the default unless optimizing for
15729 Print (or do not print) additional debug information when compiling.
15730 The default is to not print debug information.
15732 @item -march=@var{cpu-type}
15734 Generate code that will run on @var{cpu-type}, which is the name of a system
15735 representing a certain processor type. Possible values for
15736 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
15737 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
15738 When generating code using the instructions available on z/Architecture,
15739 the default is @option{-march=z900}. Otherwise, the default is
15740 @option{-march=g5}.
15742 @item -mtune=@var{cpu-type}
15744 Tune to @var{cpu-type} everything applicable about the generated code,
15745 except for the ABI and the set of available instructions.
15746 The list of @var{cpu-type} values is the same as for @option{-march}.
15747 The default is the value used for @option{-march}.
15750 @itemx -mno-tpf-trace
15751 @opindex mtpf-trace
15752 @opindex mno-tpf-trace
15753 Generate code that adds (does not add) in TPF OS specific branches to trace
15754 routines in the operating system. This option is off by default, even
15755 when compiling for the TPF OS@.
15758 @itemx -mno-fused-madd
15759 @opindex mfused-madd
15760 @opindex mno-fused-madd
15761 Generate code that uses (does not use) the floating point multiply and
15762 accumulate instructions. These instructions are generated by default if
15763 hardware floating point is used.
15765 @item -mwarn-framesize=@var{framesize}
15766 @opindex mwarn-framesize
15767 Emit a warning if the current function exceeds the given frame size. Because
15768 this is a compile time check it doesn't need to be a real problem when the program
15769 runs. It is intended to identify functions which most probably cause
15770 a stack overflow. It is useful to be used in an environment with limited stack
15771 size e.g.@: the linux kernel.
15773 @item -mwarn-dynamicstack
15774 @opindex mwarn-dynamicstack
15775 Emit a warning if the function calls alloca or uses dynamically
15776 sized arrays. This is generally a bad idea with a limited stack size.
15778 @item -mstack-guard=@var{stack-guard}
15779 @itemx -mstack-size=@var{stack-size}
15780 @opindex mstack-guard
15781 @opindex mstack-size
15782 If these options are provided the s390 back end emits additional instructions in
15783 the function prologue which trigger a trap if the stack size is @var{stack-guard}
15784 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
15785 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
15786 the frame size of the compiled function is chosen.
15787 These options are intended to be used to help debugging stack overflow problems.
15788 The additionally emitted code causes only little overhead and hence can also be
15789 used in production like systems without greater performance degradation. The given
15790 values have to be exact powers of 2 and @var{stack-size} has to be greater than
15791 @var{stack-guard} without exceeding 64k.
15792 In order to be efficient the extra code makes the assumption that the stack starts
15793 at an address aligned to the value given by @var{stack-size}.
15794 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
15797 @node Score Options
15798 @subsection Score Options
15799 @cindex Score Options
15801 These options are defined for Score implementations:
15806 Compile code for big endian mode. This is the default.
15810 Compile code for little endian mode.
15814 Disable generate bcnz instruction.
15818 Enable generate unaligned load and store instruction.
15822 Enable the use of multiply-accumulate instructions. Disabled by default.
15826 Specify the SCORE5 as the target architecture.
15830 Specify the SCORE5U of the target architecture.
15834 Specify the SCORE7 as the target architecture. This is the default.
15838 Specify the SCORE7D as the target architecture.
15842 @subsection SH Options
15844 These @samp{-m} options are defined for the SH implementations:
15849 Generate code for the SH1.
15853 Generate code for the SH2.
15856 Generate code for the SH2e.
15860 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
15861 that the floating-point unit is not used.
15863 @item -m2a-single-only
15864 @opindex m2a-single-only
15865 Generate code for the SH2a-FPU, in such a way that no double-precision
15866 floating point operations are used.
15869 @opindex m2a-single
15870 Generate code for the SH2a-FPU assuming the floating-point unit is in
15871 single-precision mode by default.
15875 Generate code for the SH2a-FPU assuming the floating-point unit is in
15876 double-precision mode by default.
15880 Generate code for the SH3.
15884 Generate code for the SH3e.
15888 Generate code for the SH4 without a floating-point unit.
15890 @item -m4-single-only
15891 @opindex m4-single-only
15892 Generate code for the SH4 with a floating-point unit that only
15893 supports single-precision arithmetic.
15897 Generate code for the SH4 assuming the floating-point unit is in
15898 single-precision mode by default.
15902 Generate code for the SH4.
15906 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
15907 floating-point unit is not used.
15909 @item -m4a-single-only
15910 @opindex m4a-single-only
15911 Generate code for the SH4a, in such a way that no double-precision
15912 floating point operations are used.
15915 @opindex m4a-single
15916 Generate code for the SH4a assuming the floating-point unit is in
15917 single-precision mode by default.
15921 Generate code for the SH4a.
15925 Same as @option{-m4a-nofpu}, except that it implicitly passes
15926 @option{-dsp} to the assembler. GCC doesn't generate any DSP
15927 instructions at the moment.
15931 Compile code for the processor in big endian mode.
15935 Compile code for the processor in little endian mode.
15939 Align doubles at 64-bit boundaries. Note that this changes the calling
15940 conventions, and thus some functions from the standard C library will
15941 not work unless you recompile it first with @option{-mdalign}.
15945 Shorten some address references at link time, when possible; uses the
15946 linker option @option{-relax}.
15950 Use 32-bit offsets in @code{switch} tables. The default is to use
15955 Enable the use of bit manipulation instructions on SH2A.
15959 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
15960 alignment constraints.
15964 Comply with the calling conventions defined by Renesas.
15968 Comply with the calling conventions defined by Renesas.
15972 Comply with the calling conventions defined for GCC before the Renesas
15973 conventions were available. This option is the default for all
15974 targets of the SH toolchain except for @samp{sh-symbianelf}.
15977 @opindex mnomacsave
15978 Mark the @code{MAC} register as call-clobbered, even if
15979 @option{-mhitachi} is given.
15983 Increase IEEE-compliance of floating-point code.
15984 At the moment, this is equivalent to @option{-fno-finite-math-only}.
15985 When generating 16 bit SH opcodes, getting IEEE-conforming results for
15986 comparisons of NANs / infinities incurs extra overhead in every
15987 floating point comparison, therefore the default is set to
15988 @option{-ffinite-math-only}.
15990 @item -minline-ic_invalidate
15991 @opindex minline-ic_invalidate
15992 Inline code to invalidate instruction cache entries after setting up
15993 nested function trampolines.
15994 This option has no effect if -musermode is in effect and the selected
15995 code generation option (e.g. -m4) does not allow the use of the icbi
15997 If the selected code generation option does not allow the use of the icbi
15998 instruction, and -musermode is not in effect, the inlined code will
15999 manipulate the instruction cache address array directly with an associative
16000 write. This not only requires privileged mode, but it will also
16001 fail if the cache line had been mapped via the TLB and has become unmapped.
16005 Dump instruction size and location in the assembly code.
16008 @opindex mpadstruct
16009 This option is deprecated. It pads structures to multiple of 4 bytes,
16010 which is incompatible with the SH ABI@.
16014 Optimize for space instead of speed. Implied by @option{-Os}.
16017 @opindex mprefergot
16018 When generating position-independent code, emit function calls using
16019 the Global Offset Table instead of the Procedure Linkage Table.
16023 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
16024 if the inlined code would not work in user mode.
16025 This is the default when the target is @code{sh-*-linux*}.
16027 @item -multcost=@var{number}
16028 @opindex multcost=@var{number}
16029 Set the cost to assume for a multiply insn.
16031 @item -mdiv=@var{strategy}
16032 @opindex mdiv=@var{strategy}
16033 Set the division strategy to use for SHmedia code. @var{strategy} must be
16034 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
16035 inv:call2, inv:fp .
16036 "fp" performs the operation in floating point. This has a very high latency,
16037 but needs only a few instructions, so it might be a good choice if
16038 your code has enough easily exploitable ILP to allow the compiler to
16039 schedule the floating point instructions together with other instructions.
16040 Division by zero causes a floating point exception.
16041 "inv" uses integer operations to calculate the inverse of the divisor,
16042 and then multiplies the dividend with the inverse. This strategy allows
16043 cse and hoisting of the inverse calculation. Division by zero calculates
16044 an unspecified result, but does not trap.
16045 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
16046 have been found, or if the entire operation has been hoisted to the same
16047 place, the last stages of the inverse calculation are intertwined with the
16048 final multiply to reduce the overall latency, at the expense of using a few
16049 more instructions, and thus offering fewer scheduling opportunities with
16051 "call" calls a library function that usually implements the inv:minlat
16053 This gives high code density for m5-*media-nofpu compilations.
16054 "call2" uses a different entry point of the same library function, where it
16055 assumes that a pointer to a lookup table has already been set up, which
16056 exposes the pointer load to cse / code hoisting optimizations.
16057 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
16058 code generation, but if the code stays unoptimized, revert to the "call",
16059 "call2", or "fp" strategies, respectively. Note that the
16060 potentially-trapping side effect of division by zero is carried by a
16061 separate instruction, so it is possible that all the integer instructions
16062 are hoisted out, but the marker for the side effect stays where it is.
16063 A recombination to fp operations or a call is not possible in that case.
16064 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
16065 that the inverse calculation was nor separated from the multiply, they speed
16066 up division where the dividend fits into 20 bits (plus sign where applicable),
16067 by inserting a test to skip a number of operations in this case; this test
16068 slows down the case of larger dividends. inv20u assumes the case of a such
16069 a small dividend to be unlikely, and inv20l assumes it to be likely.
16071 @item -mdivsi3_libfunc=@var{name}
16072 @opindex mdivsi3_libfunc=@var{name}
16073 Set the name of the library function used for 32 bit signed division to
16074 @var{name}. This only affect the name used in the call and inv:call
16075 division strategies, and the compiler will still expect the same
16076 sets of input/output/clobbered registers as if this option was not present.
16078 @item -mfixed-range=@var{register-range}
16079 @opindex mfixed-range
16080 Generate code treating the given register range as fixed registers.
16081 A fixed register is one that the register allocator can not use. This is
16082 useful when compiling kernel code. A register range is specified as
16083 two registers separated by a dash. Multiple register ranges can be
16084 specified separated by a comma.
16086 @item -madjust-unroll
16087 @opindex madjust-unroll
16088 Throttle unrolling to avoid thrashing target registers.
16089 This option only has an effect if the gcc code base supports the
16090 TARGET_ADJUST_UNROLL_MAX target hook.
16092 @item -mindexed-addressing
16093 @opindex mindexed-addressing
16094 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
16095 This is only safe if the hardware and/or OS implement 32 bit wrap-around
16096 semantics for the indexed addressing mode. The architecture allows the
16097 implementation of processors with 64 bit MMU, which the OS could use to
16098 get 32 bit addressing, but since no current hardware implementation supports
16099 this or any other way to make the indexed addressing mode safe to use in
16100 the 32 bit ABI, the default is -mno-indexed-addressing.
16102 @item -mgettrcost=@var{number}
16103 @opindex mgettrcost=@var{number}
16104 Set the cost assumed for the gettr instruction to @var{number}.
16105 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
16109 Assume pt* instructions won't trap. This will generally generate better
16110 scheduled code, but is unsafe on current hardware. The current architecture
16111 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
16112 This has the unintentional effect of making it unsafe to schedule ptabs /
16113 ptrel before a branch, or hoist it out of a loop. For example,
16114 __do_global_ctors, a part of libgcc that runs constructors at program
16115 startup, calls functions in a list which is delimited by @minus{}1. With the
16116 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
16117 That means that all the constructors will be run a bit quicker, but when
16118 the loop comes to the end of the list, the program crashes because ptabs
16119 loads @minus{}1 into a target register. Since this option is unsafe for any
16120 hardware implementing the current architecture specification, the default
16121 is -mno-pt-fixed. Unless the user specifies a specific cost with
16122 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
16123 this deters register allocation using target registers for storing
16126 @item -minvalid-symbols
16127 @opindex minvalid-symbols
16128 Assume symbols might be invalid. Ordinary function symbols generated by
16129 the compiler will always be valid to load with movi/shori/ptabs or
16130 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
16131 to generate symbols that will cause ptabs / ptrel to trap.
16132 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
16133 It will then prevent cross-basic-block cse, hoisting and most scheduling
16134 of symbol loads. The default is @option{-mno-invalid-symbols}.
16137 @node SPARC Options
16138 @subsection SPARC Options
16139 @cindex SPARC options
16141 These @samp{-m} options are supported on the SPARC:
16144 @item -mno-app-regs
16146 @opindex mno-app-regs
16148 Specify @option{-mapp-regs} to generate output using the global registers
16149 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
16152 To be fully SVR4 ABI compliant at the cost of some performance loss,
16153 specify @option{-mno-app-regs}. You should compile libraries and system
16154 software with this option.
16157 @itemx -mhard-float
16159 @opindex mhard-float
16160 Generate output containing floating point instructions. This is the
16164 @itemx -msoft-float
16166 @opindex msoft-float
16167 Generate output containing library calls for floating point.
16168 @strong{Warning:} the requisite libraries are not available for all SPARC
16169 targets. Normally the facilities of the machine's usual C compiler are
16170 used, but this cannot be done directly in cross-compilation. You must make
16171 your own arrangements to provide suitable library functions for
16172 cross-compilation. The embedded targets @samp{sparc-*-aout} and
16173 @samp{sparclite-*-*} do provide software floating point support.
16175 @option{-msoft-float} changes the calling convention in the output file;
16176 therefore, it is only useful if you compile @emph{all} of a program with
16177 this option. In particular, you need to compile @file{libgcc.a}, the
16178 library that comes with GCC, with @option{-msoft-float} in order for
16181 @item -mhard-quad-float
16182 @opindex mhard-quad-float
16183 Generate output containing quad-word (long double) floating point
16186 @item -msoft-quad-float
16187 @opindex msoft-quad-float
16188 Generate output containing library calls for quad-word (long double)
16189 floating point instructions. The functions called are those specified
16190 in the SPARC ABI@. This is the default.
16192 As of this writing, there are no SPARC implementations that have hardware
16193 support for the quad-word floating point instructions. They all invoke
16194 a trap handler for one of these instructions, and then the trap handler
16195 emulates the effect of the instruction. Because of the trap handler overhead,
16196 this is much slower than calling the ABI library routines. Thus the
16197 @option{-msoft-quad-float} option is the default.
16199 @item -mno-unaligned-doubles
16200 @itemx -munaligned-doubles
16201 @opindex mno-unaligned-doubles
16202 @opindex munaligned-doubles
16203 Assume that doubles have 8 byte alignment. This is the default.
16205 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
16206 alignment only if they are contained in another type, or if they have an
16207 absolute address. Otherwise, it assumes they have 4 byte alignment.
16208 Specifying this option avoids some rare compatibility problems with code
16209 generated by other compilers. It is not the default because it results
16210 in a performance loss, especially for floating point code.
16212 @item -mno-faster-structs
16213 @itemx -mfaster-structs
16214 @opindex mno-faster-structs
16215 @opindex mfaster-structs
16216 With @option{-mfaster-structs}, the compiler assumes that structures
16217 should have 8 byte alignment. This enables the use of pairs of
16218 @code{ldd} and @code{std} instructions for copies in structure
16219 assignment, in place of twice as many @code{ld} and @code{st} pairs.
16220 However, the use of this changed alignment directly violates the SPARC
16221 ABI@. Thus, it's intended only for use on targets where the developer
16222 acknowledges that their resulting code will not be directly in line with
16223 the rules of the ABI@.
16225 @item -mimpure-text
16226 @opindex mimpure-text
16227 @option{-mimpure-text}, used in addition to @option{-shared}, tells
16228 the compiler to not pass @option{-z text} to the linker when linking a
16229 shared object. Using this option, you can link position-dependent
16230 code into a shared object.
16232 @option{-mimpure-text} suppresses the ``relocations remain against
16233 allocatable but non-writable sections'' linker error message.
16234 However, the necessary relocations will trigger copy-on-write, and the
16235 shared object is not actually shared across processes. Instead of
16236 using @option{-mimpure-text}, you should compile all source code with
16237 @option{-fpic} or @option{-fPIC}.
16239 This option is only available on SunOS and Solaris.
16241 @item -mcpu=@var{cpu_type}
16243 Set the instruction set, register set, and instruction scheduling parameters
16244 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
16245 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
16246 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
16247 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
16248 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
16250 Default instruction scheduling parameters are used for values that select
16251 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
16252 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
16254 Here is a list of each supported architecture and their supported
16259 v8: supersparc, hypersparc
16260 sparclite: f930, f934, sparclite86x
16262 v9: ultrasparc, ultrasparc3, niagara, niagara2
16265 By default (unless configured otherwise), GCC generates code for the V7
16266 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
16267 additionally optimizes it for the Cypress CY7C602 chip, as used in the
16268 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
16269 SPARCStation 1, 2, IPX etc.
16271 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
16272 architecture. The only difference from V7 code is that the compiler emits
16273 the integer multiply and integer divide instructions which exist in SPARC-V8
16274 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
16275 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
16278 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
16279 the SPARC architecture. This adds the integer multiply, integer divide step
16280 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
16281 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
16282 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
16283 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
16284 MB86934 chip, which is the more recent SPARClite with FPU@.
16286 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
16287 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
16288 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
16289 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
16290 optimizes it for the TEMIC SPARClet chip.
16292 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
16293 architecture. This adds 64-bit integer and floating-point move instructions,
16294 3 additional floating-point condition code registers and conditional move
16295 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
16296 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
16297 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
16298 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
16299 @option{-mcpu=niagara}, the compiler additionally optimizes it for
16300 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
16301 additionally optimizes it for Sun UltraSPARC T2 chips.
16303 @item -mtune=@var{cpu_type}
16305 Set the instruction scheduling parameters for machine type
16306 @var{cpu_type}, but do not set the instruction set or register set that the
16307 option @option{-mcpu=@var{cpu_type}} would.
16309 The same values for @option{-mcpu=@var{cpu_type}} can be used for
16310 @option{-mtune=@var{cpu_type}}, but the only useful values are those
16311 that select a particular cpu implementation. Those are @samp{cypress},
16312 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
16313 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
16314 @samp{ultrasparc3}, @samp{niagara}, and @samp{niagara2}.
16319 @opindex mno-v8plus
16320 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
16321 difference from the V8 ABI is that the global and out registers are
16322 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
16323 mode for all SPARC-V9 processors.
16329 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
16330 Visual Instruction Set extensions. The default is @option{-mno-vis}.
16333 These @samp{-m} options are supported in addition to the above
16334 on SPARC-V9 processors in 64-bit environments:
16337 @item -mlittle-endian
16338 @opindex mlittle-endian
16339 Generate code for a processor running in little-endian mode. It is only
16340 available for a few configurations and most notably not on Solaris and Linux.
16346 Generate code for a 32-bit or 64-bit environment.
16347 The 32-bit environment sets int, long and pointer to 32 bits.
16348 The 64-bit environment sets int to 32 bits and long and pointer
16351 @item -mcmodel=medlow
16352 @opindex mcmodel=medlow
16353 Generate code for the Medium/Low code model: 64-bit addresses, programs
16354 must be linked in the low 32 bits of memory. Programs can be statically
16355 or dynamically linked.
16357 @item -mcmodel=medmid
16358 @opindex mcmodel=medmid
16359 Generate code for the Medium/Middle code model: 64-bit addresses, programs
16360 must be linked in the low 44 bits of memory, the text and data segments must
16361 be less than 2GB in size and the data segment must be located within 2GB of
16364 @item -mcmodel=medany
16365 @opindex mcmodel=medany
16366 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
16367 may be linked anywhere in memory, the text and data segments must be less
16368 than 2GB in size and the data segment must be located within 2GB of the
16371 @item -mcmodel=embmedany
16372 @opindex mcmodel=embmedany
16373 Generate code for the Medium/Anywhere code model for embedded systems:
16374 64-bit addresses, the text and data segments must be less than 2GB in
16375 size, both starting anywhere in memory (determined at link time). The
16376 global register %g4 points to the base of the data segment. Programs
16377 are statically linked and PIC is not supported.
16380 @itemx -mno-stack-bias
16381 @opindex mstack-bias
16382 @opindex mno-stack-bias
16383 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
16384 frame pointer if present, are offset by @minus{}2047 which must be added back
16385 when making stack frame references. This is the default in 64-bit mode.
16386 Otherwise, assume no such offset is present.
16389 These switches are supported in addition to the above on Solaris:
16394 Add support for multithreading using the Solaris threads library. This
16395 option sets flags for both the preprocessor and linker. This option does
16396 not affect the thread safety of object code produced by the compiler or
16397 that of libraries supplied with it.
16401 Add support for multithreading using the POSIX threads library. This
16402 option sets flags for both the preprocessor and linker. This option does
16403 not affect the thread safety of object code produced by the compiler or
16404 that of libraries supplied with it.
16408 This is a synonym for @option{-pthreads}.
16412 @subsection SPU Options
16413 @cindex SPU options
16415 These @samp{-m} options are supported on the SPU:
16419 @itemx -merror-reloc
16420 @opindex mwarn-reloc
16421 @opindex merror-reloc
16423 The loader for SPU does not handle dynamic relocations. By default, GCC
16424 will give an error when it generates code that requires a dynamic
16425 relocation. @option{-mno-error-reloc} disables the error,
16426 @option{-mwarn-reloc} will generate a warning instead.
16429 @itemx -munsafe-dma
16431 @opindex munsafe-dma
16433 Instructions which initiate or test completion of DMA must not be
16434 reordered with respect to loads and stores of the memory which is being
16435 accessed. Users typically address this problem using the volatile
16436 keyword, but that can lead to inefficient code in places where the
16437 memory is known to not change. Rather than mark the memory as volatile
16438 we treat the DMA instructions as potentially effecting all memory. With
16439 @option{-munsafe-dma} users must use the volatile keyword to protect
16442 @item -mbranch-hints
16443 @opindex mbranch-hints
16445 By default, GCC will generate a branch hint instruction to avoid
16446 pipeline stalls for always taken or probably taken branches. A hint
16447 will not be generated closer than 8 instructions away from its branch.
16448 There is little reason to disable them, except for debugging purposes,
16449 or to make an object a little bit smaller.
16453 @opindex msmall-mem
16454 @opindex mlarge-mem
16456 By default, GCC generates code assuming that addresses are never larger
16457 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
16458 a full 32 bit address.
16463 By default, GCC links against startup code that assumes the SPU-style
16464 main function interface (which has an unconventional parameter list).
16465 With @option{-mstdmain}, GCC will link your program against startup
16466 code that assumes a C99-style interface to @code{main}, including a
16467 local copy of @code{argv} strings.
16469 @item -mfixed-range=@var{register-range}
16470 @opindex mfixed-range
16471 Generate code treating the given register range as fixed registers.
16472 A fixed register is one that the register allocator can not use. This is
16473 useful when compiling kernel code. A register range is specified as
16474 two registers separated by a dash. Multiple register ranges can be
16475 specified separated by a comma.
16481 Compile code assuming that pointers to the PPU address space accessed
16482 via the @code{__ea} named address space qualifier are either 32 or 64
16483 bits wide. The default is 32 bits. As this is an ABI changing option,
16484 all object code in an executable must be compiled with the same setting.
16486 @item -maddress-space-conversion
16487 @itemx -mno-address-space-conversion
16488 @opindex maddress-space-conversion
16489 @opindex mno-address-space-conversion
16490 Allow/disallow treating the @code{__ea} address space as superset
16491 of the generic address space. This enables explicit type casts
16492 between @code{__ea} and generic pointer as well as implicit
16493 conversions of generic pointers to @code{__ea} pointers. The
16494 default is to allow address space pointer conversions.
16496 @item -mcache-size=@var{cache-size}
16497 @opindex mcache-size
16498 This option controls the version of libgcc that the compiler links to an
16499 executable and selects a software-managed cache for accessing variables
16500 in the @code{__ea} address space with a particular cache size. Possible
16501 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
16502 and @samp{128}. The default cache size is 64KB.
16504 @item -matomic-updates
16505 @itemx -mno-atomic-updates
16506 @opindex matomic-updates
16507 @opindex mno-atomic-updates
16508 This option controls the version of libgcc that the compiler links to an
16509 executable and selects whether atomic updates to the software-managed
16510 cache of PPU-side variables are used. If you use atomic updates, changes
16511 to a PPU variable from SPU code using the @code{__ea} named address space
16512 qualifier will not interfere with changes to other PPU variables residing
16513 in the same cache line from PPU code. If you do not use atomic updates,
16514 such interference may occur; however, writing back cache lines will be
16515 more efficient. The default behavior is to use atomic updates.
16518 @itemx -mdual-nops=@var{n}
16519 @opindex mdual-nops
16520 By default, GCC will insert nops to increase dual issue when it expects
16521 it to increase performance. @var{n} can be a value from 0 to 10. A
16522 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
16523 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
16525 @item -mhint-max-nops=@var{n}
16526 @opindex mhint-max-nops
16527 Maximum number of nops to insert for a branch hint. A branch hint must
16528 be at least 8 instructions away from the branch it is effecting. GCC
16529 will insert up to @var{n} nops to enforce this, otherwise it will not
16530 generate the branch hint.
16532 @item -mhint-max-distance=@var{n}
16533 @opindex mhint-max-distance
16534 The encoding of the branch hint instruction limits the hint to be within
16535 256 instructions of the branch it is effecting. By default, GCC makes
16536 sure it is within 125.
16539 @opindex msafe-hints
16540 Work around a hardware bug which causes the SPU to stall indefinitely.
16541 By default, GCC will insert the @code{hbrp} instruction to make sure
16542 this stall won't happen.
16546 @node System V Options
16547 @subsection Options for System V
16549 These additional options are available on System V Release 4 for
16550 compatibility with other compilers on those systems:
16555 Create a shared object.
16556 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
16560 Identify the versions of each tool used by the compiler, in a
16561 @code{.ident} assembler directive in the output.
16565 Refrain from adding @code{.ident} directives to the output file (this is
16568 @item -YP,@var{dirs}
16570 Search the directories @var{dirs}, and no others, for libraries
16571 specified with @option{-l}.
16573 @item -Ym,@var{dir}
16575 Look in the directory @var{dir} to find the M4 preprocessor.
16576 The assembler uses this option.
16577 @c This is supposed to go with a -Yd for predefined M4 macro files, but
16578 @c the generic assembler that comes with Solaris takes just -Ym.
16582 @subsection V850 Options
16583 @cindex V850 Options
16585 These @samp{-m} options are defined for V850 implementations:
16589 @itemx -mno-long-calls
16590 @opindex mlong-calls
16591 @opindex mno-long-calls
16592 Treat all calls as being far away (near). If calls are assumed to be
16593 far away, the compiler will always load the functions address up into a
16594 register, and call indirect through the pointer.
16600 Do not optimize (do optimize) basic blocks that use the same index
16601 pointer 4 or more times to copy pointer into the @code{ep} register, and
16602 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
16603 option is on by default if you optimize.
16605 @item -mno-prolog-function
16606 @itemx -mprolog-function
16607 @opindex mno-prolog-function
16608 @opindex mprolog-function
16609 Do not use (do use) external functions to save and restore registers
16610 at the prologue and epilogue of a function. The external functions
16611 are slower, but use less code space if more than one function saves
16612 the same number of registers. The @option{-mprolog-function} option
16613 is on by default if you optimize.
16617 Try to make the code as small as possible. At present, this just turns
16618 on the @option{-mep} and @option{-mprolog-function} options.
16620 @item -mtda=@var{n}
16622 Put static or global variables whose size is @var{n} bytes or less into
16623 the tiny data area that register @code{ep} points to. The tiny data
16624 area can hold up to 256 bytes in total (128 bytes for byte references).
16626 @item -msda=@var{n}
16628 Put static or global variables whose size is @var{n} bytes or less into
16629 the small data area that register @code{gp} points to. The small data
16630 area can hold up to 64 kilobytes.
16632 @item -mzda=@var{n}
16634 Put static or global variables whose size is @var{n} bytes or less into
16635 the first 32 kilobytes of memory.
16639 Specify that the target processor is the V850.
16642 @opindex mbig-switch
16643 Generate code suitable for big switch tables. Use this option only if
16644 the assembler/linker complain about out of range branches within a switch
16649 This option will cause r2 and r5 to be used in the code generated by
16650 the compiler. This setting is the default.
16652 @item -mno-app-regs
16653 @opindex mno-app-regs
16654 This option will cause r2 and r5 to be treated as fixed registers.
16658 Specify that the target processor is the V850E1. The preprocessor
16659 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
16660 this option is used.
16664 Specify that the target processor is the V850E@. The preprocessor
16665 constant @samp{__v850e__} will be defined if this option is used.
16667 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
16668 are defined then a default target processor will be chosen and the
16669 relevant @samp{__v850*__} preprocessor constant will be defined.
16671 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
16672 defined, regardless of which processor variant is the target.
16674 @item -mdisable-callt
16675 @opindex mdisable-callt
16676 This option will suppress generation of the CALLT instruction for the
16677 v850e and v850e1 flavors of the v850 architecture. The default is
16678 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
16683 @subsection VAX Options
16684 @cindex VAX options
16686 These @samp{-m} options are defined for the VAX:
16691 Do not output certain jump instructions (@code{aobleq} and so on)
16692 that the Unix assembler for the VAX cannot handle across long
16697 Do output those jump instructions, on the assumption that you
16698 will assemble with the GNU assembler.
16702 Output code for g-format floating point numbers instead of d-format.
16705 @node VxWorks Options
16706 @subsection VxWorks Options
16707 @cindex VxWorks Options
16709 The options in this section are defined for all VxWorks targets.
16710 Options specific to the target hardware are listed with the other
16711 options for that target.
16716 GCC can generate code for both VxWorks kernels and real time processes
16717 (RTPs). This option switches from the former to the latter. It also
16718 defines the preprocessor macro @code{__RTP__}.
16721 @opindex non-static
16722 Link an RTP executable against shared libraries rather than static
16723 libraries. The options @option{-static} and @option{-shared} can
16724 also be used for RTPs (@pxref{Link Options}); @option{-static}
16731 These options are passed down to the linker. They are defined for
16732 compatibility with Diab.
16735 @opindex Xbind-lazy
16736 Enable lazy binding of function calls. This option is equivalent to
16737 @option{-Wl,-z,now} and is defined for compatibility with Diab.
16741 Disable lazy binding of function calls. This option is the default and
16742 is defined for compatibility with Diab.
16745 @node x86-64 Options
16746 @subsection x86-64 Options
16747 @cindex x86-64 options
16749 These are listed under @xref{i386 and x86-64 Options}.
16751 @node i386 and x86-64 Windows Options
16752 @subsection i386 and x86-64 Windows Options
16753 @cindex i386 and x86-64 Windows Options
16755 These additional options are available for Windows targets:
16760 This option is available for Cygwin and MinGW targets. It
16761 specifies that a console application is to be generated, by
16762 instructing the linker to set the PE header subsystem type
16763 required for console applications.
16764 This is the default behavior for Cygwin and MinGW targets.
16768 This option is available for Cygwin targets. It specifies that
16769 the Cygwin internal interface is to be used for predefined
16770 preprocessor macros, C runtime libraries and related linker
16771 paths and options. For Cygwin targets this is the default behavior.
16772 This option is deprecated and will be removed in a future release.
16775 @opindex mno-cygwin
16776 This option is available for Cygwin targets. It specifies that
16777 the MinGW internal interface is to be used instead of Cygwin's, by
16778 setting MinGW-related predefined macros and linker paths and default
16780 This option is deprecated and will be removed in a future release.
16784 This option is available for Cygwin and MinGW targets. It
16785 specifies that a DLL - a dynamic link library - is to be
16786 generated, enabling the selection of the required runtime
16787 startup object and entry point.
16789 @item -mnop-fun-dllimport
16790 @opindex mnop-fun-dllimport
16791 This option is available for Cygwin and MinGW targets. It
16792 specifies that the dllimport attribute should be ignored.
16796 This option is available for MinGW targets. It specifies
16797 that MinGW-specific thread support is to be used.
16801 This option is available for mingw-w64 targets. It specifies
16802 that the UNICODE macro is getting pre-defined and that the
16803 unicode capable runtime startup code is chosen.
16807 This option is available for Cygwin and MinGW targets. It
16808 specifies that the typical Windows pre-defined macros are to
16809 be set in the pre-processor, but does not influence the choice
16810 of runtime library/startup code.
16814 This option is available for Cygwin and MinGW targets. It
16815 specifies that a GUI application is to be generated by
16816 instructing the linker to set the PE header subsystem type
16819 @item -fno-set-stack-executable
16820 @opindex fno-set-stack-executable
16821 This option is available for MinGW targets. It specifies that
16822 the executable flag for stack used by nested functions isn't
16823 set. This is necessary for binaries running in kernel mode of
16824 Windows, as there the user32 API, which is used to set executable
16825 privileges, isn't available.
16827 @item -mpe-aligned-commons
16828 @opindex mpe-aligned-commons
16829 This option is available for Cygwin and MinGW targets. It
16830 specifies that the GNU extension to the PE file format that
16831 permits the correct alignment of COMMON variables should be
16832 used when generating code. It will be enabled by default if
16833 GCC detects that the target assembler found during configuration
16834 supports the feature.
16837 See also under @ref{i386 and x86-64 Options} for standard options.
16839 @node Xstormy16 Options
16840 @subsection Xstormy16 Options
16841 @cindex Xstormy16 Options
16843 These options are defined for Xstormy16:
16848 Choose startup files and linker script suitable for the simulator.
16851 @node Xtensa Options
16852 @subsection Xtensa Options
16853 @cindex Xtensa Options
16855 These options are supported for Xtensa targets:
16859 @itemx -mno-const16
16861 @opindex mno-const16
16862 Enable or disable use of @code{CONST16} instructions for loading
16863 constant values. The @code{CONST16} instruction is currently not a
16864 standard option from Tensilica. When enabled, @code{CONST16}
16865 instructions are always used in place of the standard @code{L32R}
16866 instructions. The use of @code{CONST16} is enabled by default only if
16867 the @code{L32R} instruction is not available.
16870 @itemx -mno-fused-madd
16871 @opindex mfused-madd
16872 @opindex mno-fused-madd
16873 Enable or disable use of fused multiply/add and multiply/subtract
16874 instructions in the floating-point option. This has no effect if the
16875 floating-point option is not also enabled. Disabling fused multiply/add
16876 and multiply/subtract instructions forces the compiler to use separate
16877 instructions for the multiply and add/subtract operations. This may be
16878 desirable in some cases where strict IEEE 754-compliant results are
16879 required: the fused multiply add/subtract instructions do not round the
16880 intermediate result, thereby producing results with @emph{more} bits of
16881 precision than specified by the IEEE standard. Disabling fused multiply
16882 add/subtract instructions also ensures that the program output is not
16883 sensitive to the compiler's ability to combine multiply and add/subtract
16886 @item -mserialize-volatile
16887 @itemx -mno-serialize-volatile
16888 @opindex mserialize-volatile
16889 @opindex mno-serialize-volatile
16890 When this option is enabled, GCC inserts @code{MEMW} instructions before
16891 @code{volatile} memory references to guarantee sequential consistency.
16892 The default is @option{-mserialize-volatile}. Use
16893 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
16895 @item -mtext-section-literals
16896 @itemx -mno-text-section-literals
16897 @opindex mtext-section-literals
16898 @opindex mno-text-section-literals
16899 Control the treatment of literal pools. The default is
16900 @option{-mno-text-section-literals}, which places literals in a separate
16901 section in the output file. This allows the literal pool to be placed
16902 in a data RAM/ROM, and it also allows the linker to combine literal
16903 pools from separate object files to remove redundant literals and
16904 improve code size. With @option{-mtext-section-literals}, the literals
16905 are interspersed in the text section in order to keep them as close as
16906 possible to their references. This may be necessary for large assembly
16909 @item -mtarget-align
16910 @itemx -mno-target-align
16911 @opindex mtarget-align
16912 @opindex mno-target-align
16913 When this option is enabled, GCC instructs the assembler to
16914 automatically align instructions to reduce branch penalties at the
16915 expense of some code density. The assembler attempts to widen density
16916 instructions to align branch targets and the instructions following call
16917 instructions. If there are not enough preceding safe density
16918 instructions to align a target, no widening will be performed. The
16919 default is @option{-mtarget-align}. These options do not affect the
16920 treatment of auto-aligned instructions like @code{LOOP}, which the
16921 assembler will always align, either by widening density instructions or
16922 by inserting no-op instructions.
16925 @itemx -mno-longcalls
16926 @opindex mlongcalls
16927 @opindex mno-longcalls
16928 When this option is enabled, GCC instructs the assembler to translate
16929 direct calls to indirect calls unless it can determine that the target
16930 of a direct call is in the range allowed by the call instruction. This
16931 translation typically occurs for calls to functions in other source
16932 files. Specifically, the assembler translates a direct @code{CALL}
16933 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
16934 The default is @option{-mno-longcalls}. This option should be used in
16935 programs where the call target can potentially be out of range. This
16936 option is implemented in the assembler, not the compiler, so the
16937 assembly code generated by GCC will still show direct call
16938 instructions---look at the disassembled object code to see the actual
16939 instructions. Note that the assembler will use an indirect call for
16940 every cross-file call, not just those that really will be out of range.
16943 @node zSeries Options
16944 @subsection zSeries Options
16945 @cindex zSeries options
16947 These are listed under @xref{S/390 and zSeries Options}.
16949 @node Code Gen Options
16950 @section Options for Code Generation Conventions
16951 @cindex code generation conventions
16952 @cindex options, code generation
16953 @cindex run-time options
16955 These machine-independent options control the interface conventions
16956 used in code generation.
16958 Most of them have both positive and negative forms; the negative form
16959 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
16960 one of the forms is listed---the one which is not the default. You
16961 can figure out the other form by either removing @samp{no-} or adding
16965 @item -fbounds-check
16966 @opindex fbounds-check
16967 For front-ends that support it, generate additional code to check that
16968 indices used to access arrays are within the declared range. This is
16969 currently only supported by the Java and Fortran front-ends, where
16970 this option defaults to true and false respectively.
16974 This option generates traps for signed overflow on addition, subtraction,
16975 multiplication operations.
16979 This option instructs the compiler to assume that signed arithmetic
16980 overflow of addition, subtraction and multiplication wraps around
16981 using twos-complement representation. This flag enables some optimizations
16982 and disables others. This option is enabled by default for the Java
16983 front-end, as required by the Java language specification.
16986 @opindex fexceptions
16987 Enable exception handling. Generates extra code needed to propagate
16988 exceptions. For some targets, this implies GCC will generate frame
16989 unwind information for all functions, which can produce significant data
16990 size overhead, although it does not affect execution. If you do not
16991 specify this option, GCC will enable it by default for languages like
16992 C++ which normally require exception handling, and disable it for
16993 languages like C that do not normally require it. However, you may need
16994 to enable this option when compiling C code that needs to interoperate
16995 properly with exception handlers written in C++. You may also wish to
16996 disable this option if you are compiling older C++ programs that don't
16997 use exception handling.
16999 @item -fnon-call-exceptions
17000 @opindex fnon-call-exceptions
17001 Generate code that allows trapping instructions to throw exceptions.
17002 Note that this requires platform-specific runtime support that does
17003 not exist everywhere. Moreover, it only allows @emph{trapping}
17004 instructions to throw exceptions, i.e.@: memory references or floating
17005 point instructions. It does not allow exceptions to be thrown from
17006 arbitrary signal handlers such as @code{SIGALRM}.
17008 @item -funwind-tables
17009 @opindex funwind-tables
17010 Similar to @option{-fexceptions}, except that it will just generate any needed
17011 static data, but will not affect the generated code in any other way.
17012 You will normally not enable this option; instead, a language processor
17013 that needs this handling would enable it on your behalf.
17015 @item -fasynchronous-unwind-tables
17016 @opindex fasynchronous-unwind-tables
17017 Generate unwind table in dwarf2 format, if supported by target machine. The
17018 table is exact at each instruction boundary, so it can be used for stack
17019 unwinding from asynchronous events (such as debugger or garbage collector).
17021 @item -fpcc-struct-return
17022 @opindex fpcc-struct-return
17023 Return ``short'' @code{struct} and @code{union} values in memory like
17024 longer ones, rather than in registers. This convention is less
17025 efficient, but it has the advantage of allowing intercallability between
17026 GCC-compiled files and files compiled with other compilers, particularly
17027 the Portable C Compiler (pcc).
17029 The precise convention for returning structures in memory depends
17030 on the target configuration macros.
17032 Short structures and unions are those whose size and alignment match
17033 that of some integer type.
17035 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
17036 switch is not binary compatible with code compiled with the
17037 @option{-freg-struct-return} switch.
17038 Use it to conform to a non-default application binary interface.
17040 @item -freg-struct-return
17041 @opindex freg-struct-return
17042 Return @code{struct} and @code{union} values in registers when possible.
17043 This is more efficient for small structures than
17044 @option{-fpcc-struct-return}.
17046 If you specify neither @option{-fpcc-struct-return} nor
17047 @option{-freg-struct-return}, GCC defaults to whichever convention is
17048 standard for the target. If there is no standard convention, GCC
17049 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
17050 the principal compiler. In those cases, we can choose the standard, and
17051 we chose the more efficient register return alternative.
17053 @strong{Warning:} code compiled with the @option{-freg-struct-return}
17054 switch is not binary compatible with code compiled with the
17055 @option{-fpcc-struct-return} switch.
17056 Use it to conform to a non-default application binary interface.
17058 @item -fshort-enums
17059 @opindex fshort-enums
17060 Allocate to an @code{enum} type only as many bytes as it needs for the
17061 declared range of possible values. Specifically, the @code{enum} type
17062 will be equivalent to the smallest integer type which has enough room.
17064 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
17065 code that is not binary compatible with code generated without that switch.
17066 Use it to conform to a non-default application binary interface.
17068 @item -fshort-double
17069 @opindex fshort-double
17070 Use the same size for @code{double} as for @code{float}.
17072 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
17073 code that is not binary compatible with code generated without that switch.
17074 Use it to conform to a non-default application binary interface.
17076 @item -fshort-wchar
17077 @opindex fshort-wchar
17078 Override the underlying type for @samp{wchar_t} to be @samp{short
17079 unsigned int} instead of the default for the target. This option is
17080 useful for building programs to run under WINE@.
17082 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
17083 code that is not binary compatible with code generated without that switch.
17084 Use it to conform to a non-default application binary interface.
17087 @opindex fno-common
17088 In C code, controls the placement of uninitialized global variables.
17089 Unix C compilers have traditionally permitted multiple definitions of
17090 such variables in different compilation units by placing the variables
17092 This is the behavior specified by @option{-fcommon}, and is the default
17093 for GCC on most targets.
17094 On the other hand, this behavior is not required by ISO C, and on some
17095 targets may carry a speed or code size penalty on variable references.
17096 The @option{-fno-common} option specifies that the compiler should place
17097 uninitialized global variables in the data section of the object file,
17098 rather than generating them as common blocks.
17099 This has the effect that if the same variable is declared
17100 (without @code{extern}) in two different compilations,
17101 you will get a multiple-definition error when you link them.
17102 In this case, you must compile with @option{-fcommon} instead.
17103 Compiling with @option{-fno-common} is useful on targets for which
17104 it provides better performance, or if you wish to verify that the
17105 program will work on other systems which always treat uninitialized
17106 variable declarations this way.
17110 Ignore the @samp{#ident} directive.
17112 @item -finhibit-size-directive
17113 @opindex finhibit-size-directive
17114 Don't output a @code{.size} assembler directive, or anything else that
17115 would cause trouble if the function is split in the middle, and the
17116 two halves are placed at locations far apart in memory. This option is
17117 used when compiling @file{crtstuff.c}; you should not need to use it
17120 @item -fverbose-asm
17121 @opindex fverbose-asm
17122 Put extra commentary information in the generated assembly code to
17123 make it more readable. This option is generally only of use to those
17124 who actually need to read the generated assembly code (perhaps while
17125 debugging the compiler itself).
17127 @option{-fno-verbose-asm}, the default, causes the
17128 extra information to be omitted and is useful when comparing two assembler
17131 @item -frecord-gcc-switches
17132 @opindex frecord-gcc-switches
17133 This switch causes the command line that was used to invoke the
17134 compiler to be recorded into the object file that is being created.
17135 This switch is only implemented on some targets and the exact format
17136 of the recording is target and binary file format dependent, but it
17137 usually takes the form of a section containing ASCII text. This
17138 switch is related to the @option{-fverbose-asm} switch, but that
17139 switch only records information in the assembler output file as
17140 comments, so it never reaches the object file.
17144 @cindex global offset table
17146 Generate position-independent code (PIC) suitable for use in a shared
17147 library, if supported for the target machine. Such code accesses all
17148 constant addresses through a global offset table (GOT)@. The dynamic
17149 loader resolves the GOT entries when the program starts (the dynamic
17150 loader is not part of GCC; it is part of the operating system). If
17151 the GOT size for the linked executable exceeds a machine-specific
17152 maximum size, you get an error message from the linker indicating that
17153 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
17154 instead. (These maximums are 8k on the SPARC and 32k
17155 on the m68k and RS/6000. The 386 has no such limit.)
17157 Position-independent code requires special support, and therefore works
17158 only on certain machines. For the 386, GCC supports PIC for System V
17159 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
17160 position-independent.
17162 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
17167 If supported for the target machine, emit position-independent code,
17168 suitable for dynamic linking and avoiding any limit on the size of the
17169 global offset table. This option makes a difference on the m68k,
17170 PowerPC and SPARC@.
17172 Position-independent code requires special support, and therefore works
17173 only on certain machines.
17175 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
17182 These options are similar to @option{-fpic} and @option{-fPIC}, but
17183 generated position independent code can be only linked into executables.
17184 Usually these options are used when @option{-pie} GCC option will be
17185 used during linking.
17187 @option{-fpie} and @option{-fPIE} both define the macros
17188 @code{__pie__} and @code{__PIE__}. The macros have the value 1
17189 for @option{-fpie} and 2 for @option{-fPIE}.
17191 @item -fno-jump-tables
17192 @opindex fno-jump-tables
17193 Do not use jump tables for switch statements even where it would be
17194 more efficient than other code generation strategies. This option is
17195 of use in conjunction with @option{-fpic} or @option{-fPIC} for
17196 building code which forms part of a dynamic linker and cannot
17197 reference the address of a jump table. On some targets, jump tables
17198 do not require a GOT and this option is not needed.
17200 @item -ffixed-@var{reg}
17202 Treat the register named @var{reg} as a fixed register; generated code
17203 should never refer to it (except perhaps as a stack pointer, frame
17204 pointer or in some other fixed role).
17206 @var{reg} must be the name of a register. The register names accepted
17207 are machine-specific and are defined in the @code{REGISTER_NAMES}
17208 macro in the machine description macro file.
17210 This flag does not have a negative form, because it specifies a
17213 @item -fcall-used-@var{reg}
17214 @opindex fcall-used
17215 Treat the register named @var{reg} as an allocable register that is
17216 clobbered by function calls. It may be allocated for temporaries or
17217 variables that do not live across a call. Functions compiled this way
17218 will not save and restore the register @var{reg}.
17220 It is an error to used this flag with the frame pointer or stack pointer.
17221 Use of this flag for other registers that have fixed pervasive roles in
17222 the machine's execution model will produce disastrous results.
17224 This flag does not have a negative form, because it specifies a
17227 @item -fcall-saved-@var{reg}
17228 @opindex fcall-saved
17229 Treat the register named @var{reg} as an allocable register saved by
17230 functions. It may be allocated even for temporaries or variables that
17231 live across a call. Functions compiled this way will save and restore
17232 the register @var{reg} if they use it.
17234 It is an error to used this flag with the frame pointer or stack pointer.
17235 Use of this flag for other registers that have fixed pervasive roles in
17236 the machine's execution model will produce disastrous results.
17238 A different sort of disaster will result from the use of this flag for
17239 a register in which function values may be returned.
17241 This flag does not have a negative form, because it specifies a
17244 @item -fpack-struct[=@var{n}]
17245 @opindex fpack-struct
17246 Without a value specified, pack all structure members together without
17247 holes. When a value is specified (which must be a small power of two), pack
17248 structure members according to this value, representing the maximum
17249 alignment (that is, objects with default alignment requirements larger than
17250 this will be output potentially unaligned at the next fitting location.
17252 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
17253 code that is not binary compatible with code generated without that switch.
17254 Additionally, it makes the code suboptimal.
17255 Use it to conform to a non-default application binary interface.
17257 @item -finstrument-functions
17258 @opindex finstrument-functions
17259 Generate instrumentation calls for entry and exit to functions. Just
17260 after function entry and just before function exit, the following
17261 profiling functions will be called with the address of the current
17262 function and its call site. (On some platforms,
17263 @code{__builtin_return_address} does not work beyond the current
17264 function, so the call site information may not be available to the
17265 profiling functions otherwise.)
17268 void __cyg_profile_func_enter (void *this_fn,
17270 void __cyg_profile_func_exit (void *this_fn,
17274 The first argument is the address of the start of the current function,
17275 which may be looked up exactly in the symbol table.
17277 This instrumentation is also done for functions expanded inline in other
17278 functions. The profiling calls will indicate where, conceptually, the
17279 inline function is entered and exited. This means that addressable
17280 versions of such functions must be available. If all your uses of a
17281 function are expanded inline, this may mean an additional expansion of
17282 code size. If you use @samp{extern inline} in your C code, an
17283 addressable version of such functions must be provided. (This is
17284 normally the case anyways, but if you get lucky and the optimizer always
17285 expands the functions inline, you might have gotten away without
17286 providing static copies.)
17288 A function may be given the attribute @code{no_instrument_function}, in
17289 which case this instrumentation will not be done. This can be used, for
17290 example, for the profiling functions listed above, high-priority
17291 interrupt routines, and any functions from which the profiling functions
17292 cannot safely be called (perhaps signal handlers, if the profiling
17293 routines generate output or allocate memory).
17295 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
17296 @opindex finstrument-functions-exclude-file-list
17298 Set the list of functions that are excluded from instrumentation (see
17299 the description of @code{-finstrument-functions}). If the file that
17300 contains a function definition matches with one of @var{file}, then
17301 that function is not instrumented. The match is done on substrings:
17302 if the @var{file} parameter is a substring of the file name, it is
17303 considered to be a match.
17306 @code{-finstrument-functions-exclude-file-list=/bits/stl,include/sys}
17307 will exclude any inline function defined in files whose pathnames
17308 contain @code{/bits/stl} or @code{include/sys}.
17310 If, for some reason, you want to include letter @code{','} in one of
17311 @var{sym}, write @code{'\,'}. For example,
17312 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
17313 (note the single quote surrounding the option).
17315 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
17316 @opindex finstrument-functions-exclude-function-list
17318 This is similar to @code{-finstrument-functions-exclude-file-list},
17319 but this option sets the list of function names to be excluded from
17320 instrumentation. The function name to be matched is its user-visible
17321 name, such as @code{vector<int> blah(const vector<int> &)}, not the
17322 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
17323 match is done on substrings: if the @var{sym} parameter is a substring
17324 of the function name, it is considered to be a match. For C99 and C++
17325 extended identifiers, the function name must be given in UTF-8, not
17326 using universal character names.
17328 @item -fstack-check
17329 @opindex fstack-check
17330 Generate code to verify that you do not go beyond the boundary of the
17331 stack. You should specify this flag if you are running in an
17332 environment with multiple threads, but only rarely need to specify it in
17333 a single-threaded environment since stack overflow is automatically
17334 detected on nearly all systems if there is only one stack.
17336 Note that this switch does not actually cause checking to be done; the
17337 operating system or the language runtime must do that. The switch causes
17338 generation of code to ensure that they see the stack being extended.
17340 You can additionally specify a string parameter: @code{no} means no
17341 checking, @code{generic} means force the use of old-style checking,
17342 @code{specific} means use the best checking method and is equivalent
17343 to bare @option{-fstack-check}.
17345 Old-style checking is a generic mechanism that requires no specific
17346 target support in the compiler but comes with the following drawbacks:
17350 Modified allocation strategy for large objects: they will always be
17351 allocated dynamically if their size exceeds a fixed threshold.
17354 Fixed limit on the size of the static frame of functions: when it is
17355 topped by a particular function, stack checking is not reliable and
17356 a warning is issued by the compiler.
17359 Inefficiency: because of both the modified allocation strategy and the
17360 generic implementation, the performances of the code are hampered.
17363 Note that old-style stack checking is also the fallback method for
17364 @code{specific} if no target support has been added in the compiler.
17366 @item -fstack-limit-register=@var{reg}
17367 @itemx -fstack-limit-symbol=@var{sym}
17368 @itemx -fno-stack-limit
17369 @opindex fstack-limit-register
17370 @opindex fstack-limit-symbol
17371 @opindex fno-stack-limit
17372 Generate code to ensure that the stack does not grow beyond a certain value,
17373 either the value of a register or the address of a symbol. If the stack
17374 would grow beyond the value, a signal is raised. For most targets,
17375 the signal is raised before the stack overruns the boundary, so
17376 it is possible to catch the signal without taking special precautions.
17378 For instance, if the stack starts at absolute address @samp{0x80000000}
17379 and grows downwards, you can use the flags
17380 @option{-fstack-limit-symbol=__stack_limit} and
17381 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
17382 of 128KB@. Note that this may only work with the GNU linker.
17384 @cindex aliasing of parameters
17385 @cindex parameters, aliased
17386 @item -fargument-alias
17387 @itemx -fargument-noalias
17388 @itemx -fargument-noalias-global
17389 @itemx -fargument-noalias-anything
17390 @opindex fargument-alias
17391 @opindex fargument-noalias
17392 @opindex fargument-noalias-global
17393 @opindex fargument-noalias-anything
17394 Specify the possible relationships among parameters and between
17395 parameters and global data.
17397 @option{-fargument-alias} specifies that arguments (parameters) may
17398 alias each other and may alias global storage.@*
17399 @option{-fargument-noalias} specifies that arguments do not alias
17400 each other, but may alias global storage.@*
17401 @option{-fargument-noalias-global} specifies that arguments do not
17402 alias each other and do not alias global storage.
17403 @option{-fargument-noalias-anything} specifies that arguments do not
17404 alias any other storage.
17406 Each language will automatically use whatever option is required by
17407 the language standard. You should not need to use these options yourself.
17409 @item -fleading-underscore
17410 @opindex fleading-underscore
17411 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
17412 change the way C symbols are represented in the object file. One use
17413 is to help link with legacy assembly code.
17415 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
17416 generate code that is not binary compatible with code generated without that
17417 switch. Use it to conform to a non-default application binary interface.
17418 Not all targets provide complete support for this switch.
17420 @item -ftls-model=@var{model}
17421 @opindex ftls-model
17422 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
17423 The @var{model} argument should be one of @code{global-dynamic},
17424 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
17426 The default without @option{-fpic} is @code{initial-exec}; with
17427 @option{-fpic} the default is @code{global-dynamic}.
17429 @item -fvisibility=@var{default|internal|hidden|protected}
17430 @opindex fvisibility
17431 Set the default ELF image symbol visibility to the specified option---all
17432 symbols will be marked with this unless overridden within the code.
17433 Using this feature can very substantially improve linking and
17434 load times of shared object libraries, produce more optimized
17435 code, provide near-perfect API export and prevent symbol clashes.
17436 It is @strong{strongly} recommended that you use this in any shared objects
17439 Despite the nomenclature, @code{default} always means public ie;
17440 available to be linked against from outside the shared object.
17441 @code{protected} and @code{internal} are pretty useless in real-world
17442 usage so the only other commonly used option will be @code{hidden}.
17443 The default if @option{-fvisibility} isn't specified is
17444 @code{default}, i.e., make every
17445 symbol public---this causes the same behavior as previous versions of
17448 A good explanation of the benefits offered by ensuring ELF
17449 symbols have the correct visibility is given by ``How To Write
17450 Shared Libraries'' by Ulrich Drepper (which can be found at
17451 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
17452 solution made possible by this option to marking things hidden when
17453 the default is public is to make the default hidden and mark things
17454 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
17455 and @code{__attribute__ ((visibility("default")))} instead of
17456 @code{__declspec(dllexport)} you get almost identical semantics with
17457 identical syntax. This is a great boon to those working with
17458 cross-platform projects.
17460 For those adding visibility support to existing code, you may find
17461 @samp{#pragma GCC visibility} of use. This works by you enclosing
17462 the declarations you wish to set visibility for with (for example)
17463 @samp{#pragma GCC visibility push(hidden)} and
17464 @samp{#pragma GCC visibility pop}.
17465 Bear in mind that symbol visibility should be viewed @strong{as
17466 part of the API interface contract} and thus all new code should
17467 always specify visibility when it is not the default ie; declarations
17468 only for use within the local DSO should @strong{always} be marked explicitly
17469 as hidden as so to avoid PLT indirection overheads---making this
17470 abundantly clear also aids readability and self-documentation of the code.
17471 Note that due to ISO C++ specification requirements, operator new and
17472 operator delete must always be of default visibility.
17474 Be aware that headers from outside your project, in particular system
17475 headers and headers from any other library you use, may not be
17476 expecting to be compiled with visibility other than the default. You
17477 may need to explicitly say @samp{#pragma GCC visibility push(default)}
17478 before including any such headers.
17480 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
17481 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
17482 no modifications. However, this means that calls to @samp{extern}
17483 functions with no explicit visibility will use the PLT, so it is more
17484 effective to use @samp{__attribute ((visibility))} and/or
17485 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
17486 declarations should be treated as hidden.
17488 Note that @samp{-fvisibility} does affect C++ vague linkage
17489 entities. This means that, for instance, an exception class that will
17490 be thrown between DSOs must be explicitly marked with default
17491 visibility so that the @samp{type_info} nodes will be unified between
17494 An overview of these techniques, their benefits and how to use them
17495 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
17501 @node Environment Variables
17502 @section Environment Variables Affecting GCC
17503 @cindex environment variables
17505 @c man begin ENVIRONMENT
17506 This section describes several environment variables that affect how GCC
17507 operates. Some of them work by specifying directories or prefixes to use
17508 when searching for various kinds of files. Some are used to specify other
17509 aspects of the compilation environment.
17511 Note that you can also specify places to search using options such as
17512 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
17513 take precedence over places specified using environment variables, which
17514 in turn take precedence over those specified by the configuration of GCC@.
17515 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
17516 GNU Compiler Collection (GCC) Internals}.
17521 @c @itemx LC_COLLATE
17523 @c @itemx LC_MONETARY
17524 @c @itemx LC_NUMERIC
17529 @c @findex LC_COLLATE
17530 @findex LC_MESSAGES
17531 @c @findex LC_MONETARY
17532 @c @findex LC_NUMERIC
17536 These environment variables control the way that GCC uses
17537 localization information that allow GCC to work with different
17538 national conventions. GCC inspects the locale categories
17539 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
17540 so. These locale categories can be set to any value supported by your
17541 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
17542 Kingdom encoded in UTF-8.
17544 The @env{LC_CTYPE} environment variable specifies character
17545 classification. GCC uses it to determine the character boundaries in
17546 a string; this is needed for some multibyte encodings that contain quote
17547 and escape characters that would otherwise be interpreted as a string
17550 The @env{LC_MESSAGES} environment variable specifies the language to
17551 use in diagnostic messages.
17553 If the @env{LC_ALL} environment variable is set, it overrides the value
17554 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
17555 and @env{LC_MESSAGES} default to the value of the @env{LANG}
17556 environment variable. If none of these variables are set, GCC
17557 defaults to traditional C English behavior.
17561 If @env{TMPDIR} is set, it specifies the directory to use for temporary
17562 files. GCC uses temporary files to hold the output of one stage of
17563 compilation which is to be used as input to the next stage: for example,
17564 the output of the preprocessor, which is the input to the compiler
17567 @item GCC_EXEC_PREFIX
17568 @findex GCC_EXEC_PREFIX
17569 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
17570 names of the subprograms executed by the compiler. No slash is added
17571 when this prefix is combined with the name of a subprogram, but you can
17572 specify a prefix that ends with a slash if you wish.
17574 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
17575 an appropriate prefix to use based on the pathname it was invoked with.
17577 If GCC cannot find the subprogram using the specified prefix, it
17578 tries looking in the usual places for the subprogram.
17580 The default value of @env{GCC_EXEC_PREFIX} is
17581 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
17582 the installed compiler. In many cases @var{prefix} is the value
17583 of @code{prefix} when you ran the @file{configure} script.
17585 Other prefixes specified with @option{-B} take precedence over this prefix.
17587 This prefix is also used for finding files such as @file{crt0.o} that are
17590 In addition, the prefix is used in an unusual way in finding the
17591 directories to search for header files. For each of the standard
17592 directories whose name normally begins with @samp{/usr/local/lib/gcc}
17593 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
17594 replacing that beginning with the specified prefix to produce an
17595 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
17596 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
17597 These alternate directories are searched first; the standard directories
17598 come next. If a standard directory begins with the configured
17599 @var{prefix} then the value of @var{prefix} is replaced by
17600 @env{GCC_EXEC_PREFIX} when looking for header files.
17602 @item COMPILER_PATH
17603 @findex COMPILER_PATH
17604 The value of @env{COMPILER_PATH} is a colon-separated list of
17605 directories, much like @env{PATH}. GCC tries the directories thus
17606 specified when searching for subprograms, if it can't find the
17607 subprograms using @env{GCC_EXEC_PREFIX}.
17610 @findex LIBRARY_PATH
17611 The value of @env{LIBRARY_PATH} is a colon-separated list of
17612 directories, much like @env{PATH}. When configured as a native compiler,
17613 GCC tries the directories thus specified when searching for special
17614 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
17615 using GCC also uses these directories when searching for ordinary
17616 libraries for the @option{-l} option (but directories specified with
17617 @option{-L} come first).
17621 @cindex locale definition
17622 This variable is used to pass locale information to the compiler. One way in
17623 which this information is used is to determine the character set to be used
17624 when character literals, string literals and comments are parsed in C and C++.
17625 When the compiler is configured to allow multibyte characters,
17626 the following values for @env{LANG} are recognized:
17630 Recognize JIS characters.
17632 Recognize SJIS characters.
17634 Recognize EUCJP characters.
17637 If @env{LANG} is not defined, or if it has some other value, then the
17638 compiler will use mblen and mbtowc as defined by the default locale to
17639 recognize and translate multibyte characters.
17643 Some additional environments variables affect the behavior of the
17646 @include cppenv.texi
17650 @node Precompiled Headers
17651 @section Using Precompiled Headers
17652 @cindex precompiled headers
17653 @cindex speed of compilation
17655 Often large projects have many header files that are included in every
17656 source file. The time the compiler takes to process these header files
17657 over and over again can account for nearly all of the time required to
17658 build the project. To make builds faster, GCC allows users to
17659 `precompile' a header file; then, if builds can use the precompiled
17660 header file they will be much faster.
17662 To create a precompiled header file, simply compile it as you would any
17663 other file, if necessary using the @option{-x} option to make the driver
17664 treat it as a C or C++ header file. You will probably want to use a
17665 tool like @command{make} to keep the precompiled header up-to-date when
17666 the headers it contains change.
17668 A precompiled header file will be searched for when @code{#include} is
17669 seen in the compilation. As it searches for the included file
17670 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
17671 compiler looks for a precompiled header in each directory just before it
17672 looks for the include file in that directory. The name searched for is
17673 the name specified in the @code{#include} with @samp{.gch} appended. If
17674 the precompiled header file can't be used, it is ignored.
17676 For instance, if you have @code{#include "all.h"}, and you have
17677 @file{all.h.gch} in the same directory as @file{all.h}, then the
17678 precompiled header file will be used if possible, and the original
17679 header will be used otherwise.
17681 Alternatively, you might decide to put the precompiled header file in a
17682 directory and use @option{-I} to ensure that directory is searched
17683 before (or instead of) the directory containing the original header.
17684 Then, if you want to check that the precompiled header file is always
17685 used, you can put a file of the same name as the original header in this
17686 directory containing an @code{#error} command.
17688 This also works with @option{-include}. So yet another way to use
17689 precompiled headers, good for projects not designed with precompiled
17690 header files in mind, is to simply take most of the header files used by
17691 a project, include them from another header file, precompile that header
17692 file, and @option{-include} the precompiled header. If the header files
17693 have guards against multiple inclusion, they will be skipped because
17694 they've already been included (in the precompiled header).
17696 If you need to precompile the same header file for different
17697 languages, targets, or compiler options, you can instead make a
17698 @emph{directory} named like @file{all.h.gch}, and put each precompiled
17699 header in the directory, perhaps using @option{-o}. It doesn't matter
17700 what you call the files in the directory, every precompiled header in
17701 the directory will be considered. The first precompiled header
17702 encountered in the directory that is valid for this compilation will
17703 be used; they're searched in no particular order.
17705 There are many other possibilities, limited only by your imagination,
17706 good sense, and the constraints of your build system.
17708 A precompiled header file can be used only when these conditions apply:
17712 Only one precompiled header can be used in a particular compilation.
17715 A precompiled header can't be used once the first C token is seen. You
17716 can have preprocessor directives before a precompiled header; you can
17717 even include a precompiled header from inside another header, so long as
17718 there are no C tokens before the @code{#include}.
17721 The precompiled header file must be produced for the same language as
17722 the current compilation. You can't use a C precompiled header for a C++
17726 The precompiled header file must have been produced by the same compiler
17727 binary as the current compilation is using.
17730 Any macros defined before the precompiled header is included must
17731 either be defined in the same way as when the precompiled header was
17732 generated, or must not affect the precompiled header, which usually
17733 means that they don't appear in the precompiled header at all.
17735 The @option{-D} option is one way to define a macro before a
17736 precompiled header is included; using a @code{#define} can also do it.
17737 There are also some options that define macros implicitly, like
17738 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
17741 @item If debugging information is output when using the precompiled
17742 header, using @option{-g} or similar, the same kind of debugging information
17743 must have been output when building the precompiled header. However,
17744 a precompiled header built using @option{-g} can be used in a compilation
17745 when no debugging information is being output.
17747 @item The same @option{-m} options must generally be used when building
17748 and using the precompiled header. @xref{Submodel Options},
17749 for any cases where this rule is relaxed.
17751 @item Each of the following options must be the same when building and using
17752 the precompiled header:
17754 @gccoptlist{-fexceptions}
17757 Some other command-line options starting with @option{-f},
17758 @option{-p}, or @option{-O} must be defined in the same way as when
17759 the precompiled header was generated. At present, it's not clear
17760 which options are safe to change and which are not; the safest choice
17761 is to use exactly the same options when generating and using the
17762 precompiled header. The following are known to be safe:
17764 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
17765 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
17766 -fsched-verbose=<number> -fschedule-insns -fvisibility= @gol
17771 For all of these except the last, the compiler will automatically
17772 ignore the precompiled header if the conditions aren't met. If you
17773 find an option combination that doesn't work and doesn't cause the
17774 precompiled header to be ignored, please consider filing a bug report,
17777 If you do use differing options when generating and using the
17778 precompiled header, the actual behavior will be a mixture of the
17779 behavior for the options. For instance, if you use @option{-g} to
17780 generate the precompiled header but not when using it, you may or may
17781 not get debugging information for routines in the precompiled header.