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 -Wcpp -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-assignments -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 @gol
920 -fleading-underscore -ftls-model=@var{model} @gol
921 -ftrapv -fwrapv -fbounds-check @gol
926 * Overall Options:: Controlling the kind of output:
927 an executable, object files, assembler files,
928 or preprocessed source.
929 * C Dialect Options:: Controlling the variant of C language compiled.
930 * C++ Dialect Options:: Variations on C++.
931 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
933 * Language Independent Options:: Controlling how diagnostics should be
935 * Warning Options:: How picky should the compiler be?
936 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
937 * Optimize Options:: How much optimization?
938 * Preprocessor Options:: Controlling header files and macro definitions.
939 Also, getting dependency information for Make.
940 * Assembler Options:: Passing options to the assembler.
941 * Link Options:: Specifying libraries and so on.
942 * Directory Options:: Where to find header files and libraries.
943 Where to find the compiler executable files.
944 * Spec Files:: How to pass switches to sub-processes.
945 * Target Options:: Running a cross-compiler, or an old version of GCC.
948 @node Overall Options
949 @section Options Controlling the Kind of Output
951 Compilation can involve up to four stages: preprocessing, compilation
952 proper, assembly and linking, always in that order. GCC is capable of
953 preprocessing and compiling several files either into several
954 assembler input files, or into one assembler input file; then each
955 assembler input file produces an object file, and linking combines all
956 the object files (those newly compiled, and those specified as input)
957 into an executable file.
959 @cindex file name suffix
960 For any given input file, the file name suffix determines what kind of
965 C source code which must be preprocessed.
968 C source code which should not be preprocessed.
971 C++ source code which should not be preprocessed.
974 Objective-C source code. Note that you must link with the @file{libobjc}
975 library to make an Objective-C program work.
978 Objective-C source code which should not be preprocessed.
982 Objective-C++ source code. Note that you must link with the @file{libobjc}
983 library to make an Objective-C++ program work. Note that @samp{.M} refers
984 to a literal capital M@.
987 Objective-C++ source code which should not be preprocessed.
990 C, C++, Objective-C or Objective-C++ header file to be turned into a
995 @itemx @var{file}.cxx
996 @itemx @var{file}.cpp
997 @itemx @var{file}.CPP
998 @itemx @var{file}.c++
1000 C++ source code which must be preprocessed. Note that in @samp{.cxx},
1001 the last two letters must both be literally @samp{x}. Likewise,
1002 @samp{.C} refers to a literal capital C@.
1006 Objective-C++ source code which must be preprocessed.
1008 @item @var{file}.mii
1009 Objective-C++ source code which should not be preprocessed.
1013 @itemx @var{file}.hp
1014 @itemx @var{file}.hxx
1015 @itemx @var{file}.hpp
1016 @itemx @var{file}.HPP
1017 @itemx @var{file}.h++
1018 @itemx @var{file}.tcc
1019 C++ header file to be turned into a precompiled header.
1022 @itemx @var{file}.for
1023 @itemx @var{file}.ftn
1024 Fixed form Fortran source code which should not be preprocessed.
1027 @itemx @var{file}.FOR
1028 @itemx @var{file}.fpp
1029 @itemx @var{file}.FPP
1030 @itemx @var{file}.FTN
1031 Fixed form Fortran source code which must be preprocessed (with the traditional
1034 @item @var{file}.f90
1035 @itemx @var{file}.f95
1036 @itemx @var{file}.f03
1037 @itemx @var{file}.f08
1038 Free form Fortran source code which should not be preprocessed.
1040 @item @var{file}.F90
1041 @itemx @var{file}.F95
1042 @itemx @var{file}.F03
1043 @itemx @var{file}.F08
1044 Free form Fortran source code which must be preprocessed (with the
1045 traditional preprocessor).
1047 @c FIXME: Descriptions of Java file types.
1053 @item @var{file}.ads
1054 Ada source code file which contains a library unit declaration (a
1055 declaration of a package, subprogram, or generic, or a generic
1056 instantiation), or a library unit renaming declaration (a package,
1057 generic, or subprogram renaming declaration). Such files are also
1060 @item @var{file}.adb
1061 Ada source code file containing a library unit body (a subprogram or
1062 package body). Such files are also called @dfn{bodies}.
1064 @c GCC also knows about some suffixes for languages not yet included:
1075 @itemx @var{file}.sx
1076 Assembler code which must be preprocessed.
1079 An object file to be fed straight into linking.
1080 Any file name with no recognized suffix is treated this way.
1084 You can specify the input language explicitly with the @option{-x} option:
1087 @item -x @var{language}
1088 Specify explicitly the @var{language} for the following input files
1089 (rather than letting the compiler choose a default based on the file
1090 name suffix). This option applies to all following input files until
1091 the next @option{-x} option. Possible values for @var{language} are:
1093 c c-header c-cpp-output
1094 c++ c++-header c++-cpp-output
1095 objective-c objective-c-header objective-c-cpp-output
1096 objective-c++ objective-c++-header objective-c++-cpp-output
1097 assembler assembler-with-cpp
1099 f77 f77-cpp-input f95 f95-cpp-input
1104 Turn off any specification of a language, so that subsequent files are
1105 handled according to their file name suffixes (as they are if @option{-x}
1106 has not been used at all).
1108 @item -pass-exit-codes
1109 @opindex pass-exit-codes
1110 Normally the @command{gcc} program will exit with the code of 1 if any
1111 phase of the compiler returns a non-success return code. If you specify
1112 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1113 numerically highest error produced by any phase that returned an error
1114 indication. The C, C++, and Fortran frontends return 4, if an internal
1115 compiler error is encountered.
1118 If you only want some of the stages of compilation, you can use
1119 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1120 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1121 @command{gcc} is to stop. Note that some combinations (for example,
1122 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1127 Compile or assemble the source files, but do not link. The linking
1128 stage simply is not done. The ultimate output is in the form of an
1129 object file for each source file.
1131 By default, the object file name for a source file is made by replacing
1132 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1134 Unrecognized input files, not requiring compilation or assembly, are
1139 Stop after the stage of compilation proper; do not assemble. The output
1140 is in the form of an assembler code file for each non-assembler input
1143 By default, the assembler file name for a source file is made by
1144 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1146 Input files that don't require compilation are ignored.
1150 Stop after the preprocessing stage; do not run the compiler proper. The
1151 output is in the form of preprocessed source code, which is sent to the
1154 Input files which don't require preprocessing are ignored.
1156 @cindex output file option
1159 Place output in file @var{file}. This applies regardless to whatever
1160 sort of output is being produced, whether it be an executable file,
1161 an object file, an assembler file or preprocessed C code.
1163 If @option{-o} is not specified, the default is to put an executable
1164 file in @file{a.out}, the object file for
1165 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1166 assembler file in @file{@var{source}.s}, a precompiled header file in
1167 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1172 Print (on standard error output) the commands executed to run the stages
1173 of compilation. Also print the version number of the compiler driver
1174 program and of the preprocessor and the compiler proper.
1178 Like @option{-v} except the commands are not executed and all command
1179 arguments are quoted. This is useful for shell scripts to capture the
1180 driver-generated command lines.
1184 Use pipes rather than temporary files for communication between the
1185 various stages of compilation. This fails to work on some systems where
1186 the assembler is unable to read from a pipe; but the GNU assembler has
1191 If you are compiling multiple source files, this option tells the driver
1192 to pass all the source files to the compiler at once (for those
1193 languages for which the compiler can handle this). This will allow
1194 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1195 language for which this is supported is C@. If you pass source files for
1196 multiple languages to the driver, using this option, the driver will invoke
1197 the compiler(s) that support IMA once each, passing each compiler all the
1198 source files appropriate for it. For those languages that do not support
1199 IMA this option will be ignored, and the compiler will be invoked once for
1200 each source file in that language. If you use this option in conjunction
1201 with @option{-save-temps}, the compiler will generate multiple
1203 (one for each source file), but only one (combined) @file{.o} or
1208 Print (on the standard output) a description of the command line options
1209 understood by @command{gcc}. If the @option{-v} option is also specified
1210 then @option{--help} will also be passed on to the various processes
1211 invoked by @command{gcc}, so that they can display the command line options
1212 they accept. If the @option{-Wextra} option has also been specified
1213 (prior to the @option{--help} option), then command line options which
1214 have no documentation associated with them will also be displayed.
1217 @opindex target-help
1218 Print (on the standard output) a description of target-specific command
1219 line options for each tool. For some targets extra target-specific
1220 information may also be printed.
1222 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1223 Print (on the standard output) a description of the command line
1224 options understood by the compiler that fit into all specified classes
1225 and qualifiers. These are the supported classes:
1228 @item @samp{optimizers}
1229 This will display all of the optimization options supported by the
1232 @item @samp{warnings}
1233 This will display all of the options controlling warning messages
1234 produced by the compiler.
1237 This will display target-specific options. Unlike the
1238 @option{--target-help} option however, target-specific options of the
1239 linker and assembler will not be displayed. This is because those
1240 tools do not currently support the extended @option{--help=} syntax.
1243 This will display the values recognized by the @option{--param}
1246 @item @var{language}
1247 This will display the options supported for @var{language}, where
1248 @var{language} is the name of one of the languages supported in this
1252 This will display the options that are common to all languages.
1255 These are the supported qualifiers:
1258 @item @samp{undocumented}
1259 Display only those options which are undocumented.
1262 Display options which take an argument that appears after an equal
1263 sign in the same continuous piece of text, such as:
1264 @samp{--help=target}.
1266 @item @samp{separate}
1267 Display options which take an argument that appears as a separate word
1268 following the original option, such as: @samp{-o output-file}.
1271 Thus for example to display all the undocumented target-specific
1272 switches supported by the compiler the following can be used:
1275 --help=target,undocumented
1278 The sense of a qualifier can be inverted by prefixing it with the
1279 @samp{^} character, so for example to display all binary warning
1280 options (i.e., ones that are either on or off and that do not take an
1281 argument), which have a description the following can be used:
1284 --help=warnings,^joined,^undocumented
1287 The argument to @option{--help=} should not consist solely of inverted
1290 Combining several classes is possible, although this usually
1291 restricts the output by so much that there is nothing to display. One
1292 case where it does work however is when one of the classes is
1293 @var{target}. So for example to display all the target-specific
1294 optimization options the following can be used:
1297 --help=target,optimizers
1300 The @option{--help=} option can be repeated on the command line. Each
1301 successive use will display its requested class of options, skipping
1302 those that have already been displayed.
1304 If the @option{-Q} option appears on the command line before the
1305 @option{--help=} option, then the descriptive text displayed by
1306 @option{--help=} is changed. Instead of describing the displayed
1307 options, an indication is given as to whether the option is enabled,
1308 disabled or set to a specific value (assuming that the compiler
1309 knows this at the point where the @option{--help=} option is used).
1311 Here is a truncated example from the ARM port of @command{gcc}:
1314 % gcc -Q -mabi=2 --help=target -c
1315 The following options are target specific:
1317 -mabort-on-noreturn [disabled]
1321 The output is sensitive to the effects of previous command line
1322 options, so for example it is possible to find out which optimizations
1323 are enabled at @option{-O2} by using:
1326 -Q -O2 --help=optimizers
1329 Alternatively you can discover which binary optimizations are enabled
1330 by @option{-O3} by using:
1333 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1334 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1335 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1338 @item -no-canonical-prefixes
1339 @opindex no-canonical-prefixes
1340 Do not expand any symbolic links, resolve references to @samp{/../}
1341 or @samp{/./}, or make the path absolute when generating a relative
1346 Display the version number and copyrights of the invoked GCC@.
1350 Invoke all subcommands under a wrapper program. It takes a single
1351 comma separated list as an argument, which will be used to invoke
1355 gcc -c t.c -wrapper gdb,--args
1358 This will invoke all subprograms of gcc under "gdb --args",
1359 thus cc1 invocation will be "gdb --args cc1 ...".
1361 @item -fplugin=@var{name}.so
1362 Load the plugin code in file @var{name}.so, assumed to be a
1363 shared object to be dlopen'd by the compiler. The base name of
1364 the shared object file is used to identify the plugin for the
1365 purposes of argument parsing (See
1366 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1367 Each plugin should define the callback functions specified in the
1370 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1371 Define an argument called @var{key} with a value of @var{value}
1372 for the plugin called @var{name}.
1374 @include @value{srcdir}/../libiberty/at-file.texi
1378 @section Compiling C++ Programs
1380 @cindex suffixes for C++ source
1381 @cindex C++ source file suffixes
1382 C++ source files conventionally use one of the suffixes @samp{.C},
1383 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1384 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1385 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1386 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1387 files with these names and compiles them as C++ programs even if you
1388 call the compiler the same way as for compiling C programs (usually
1389 with the name @command{gcc}).
1393 However, the use of @command{gcc} does not add the C++ library.
1394 @command{g++} is a program that calls GCC and treats @samp{.c},
1395 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1396 files unless @option{-x} is used, and automatically specifies linking
1397 against the C++ library. This program is also useful when
1398 precompiling a C header file with a @samp{.h} extension for use in C++
1399 compilations. On many systems, @command{g++} is also installed with
1400 the name @command{c++}.
1402 @cindex invoking @command{g++}
1403 When you compile C++ programs, you may specify many of the same
1404 command-line options that you use for compiling programs in any
1405 language; or command-line options meaningful for C and related
1406 languages; or options that are meaningful only for C++ programs.
1407 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1408 explanations of options for languages related to C@.
1409 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1410 explanations of options that are meaningful only for C++ programs.
1412 @node C Dialect Options
1413 @section Options Controlling C Dialect
1414 @cindex dialect options
1415 @cindex language dialect options
1416 @cindex options, dialect
1418 The following options control the dialect of C (or languages derived
1419 from C, such as C++, Objective-C and Objective-C++) that the compiler
1423 @cindex ANSI support
1427 In C mode, this is equivalent to @samp{-std=c90}. In C++ mode, it is
1428 equivalent to @samp{-std=c++98}.
1430 This turns off certain features of GCC that are incompatible with ISO
1431 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1432 such as the @code{asm} and @code{typeof} keywords, and
1433 predefined macros such as @code{unix} and @code{vax} that identify the
1434 type of system you are using. It also enables the undesirable and
1435 rarely used ISO trigraph feature. For the C compiler,
1436 it disables recognition of C++ style @samp{//} comments as well as
1437 the @code{inline} keyword.
1439 The alternate keywords @code{__asm__}, @code{__extension__},
1440 @code{__inline__} and @code{__typeof__} continue to work despite
1441 @option{-ansi}. You would not want to use them in an ISO C program, of
1442 course, but it is useful to put them in header files that might be included
1443 in compilations done with @option{-ansi}. Alternate predefined macros
1444 such as @code{__unix__} and @code{__vax__} are also available, with or
1445 without @option{-ansi}.
1447 The @option{-ansi} option does not cause non-ISO programs to be
1448 rejected gratuitously. For that, @option{-pedantic} is required in
1449 addition to @option{-ansi}. @xref{Warning Options}.
1451 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1452 option is used. Some header files may notice this macro and refrain
1453 from declaring certain functions or defining certain macros that the
1454 ISO standard doesn't call for; this is to avoid interfering with any
1455 programs that might use these names for other things.
1457 Functions that would normally be built in but do not have semantics
1458 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1459 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1460 built-in functions provided by GCC}, for details of the functions
1465 Determine the language standard. @xref{Standards,,Language Standards
1466 Supported by GCC}, for details of these standard versions. This option
1467 is currently only supported when compiling C or C++.
1469 The compiler can accept several base standards, such as @samp{c90} or
1470 @samp{c++98}, and GNU dialects of those standards, such as
1471 @samp{gnu90} or @samp{gnu++98}. By specifying a base standard, the
1472 compiler will accept all programs following that standard and those
1473 using GNU extensions that do not contradict it. For example,
1474 @samp{-std=c90} turns off certain features of GCC that are
1475 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1476 keywords, but not other GNU extensions that do not have a meaning in
1477 ISO C90, such as omitting the middle term of a @code{?:}
1478 expression. On the other hand, by specifying a GNU dialect of a
1479 standard, all features the compiler support are enabled, even when
1480 those features change the meaning of the base standard and some
1481 strict-conforming programs may be rejected. The particular standard
1482 is used by @option{-pedantic} to identify which features are GNU
1483 extensions given that version of the standard. For example
1484 @samp{-std=gnu90 -pedantic} would warn about C++ style @samp{//}
1485 comments, while @samp{-std=gnu99 -pedantic} would not.
1487 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.
1509 GNU dialect of ISO C90 (including some C99 features). This
1510 is the default for C code.
1514 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1515 this will become the default. The name @samp{gnu9x} is deprecated.
1518 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1522 GNU dialect of @option{-std=c++98}. This is the default for
1526 The working draft of the upcoming ISO C++0x standard. This option
1527 enables experimental features that are likely to be included in
1528 C++0x. The working draft is constantly changing, and any feature that is
1529 enabled by this flag may be removed from future versions of GCC if it is
1530 not part of the C++0x standard.
1533 GNU dialect of @option{-std=c++0x}. This option enables
1534 experimental features that may be removed in future versions of GCC.
1537 @item -fgnu89-inline
1538 @opindex fgnu89-inline
1539 The option @option{-fgnu89-inline} tells GCC to use the traditional
1540 GNU semantics for @code{inline} functions when in C99 mode.
1541 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1542 is accepted and ignored by GCC versions 4.1.3 up to but not including
1543 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1544 C99 mode. Using this option is roughly equivalent to adding the
1545 @code{gnu_inline} function attribute to all inline functions
1546 (@pxref{Function Attributes}).
1548 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1549 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1550 specifies the default behavior). This option was first supported in
1551 GCC 4.3. This option is not supported in @option{-std=c90} or
1552 @option{-std=gnu90} mode.
1554 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1555 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1556 in effect for @code{inline} functions. @xref{Common Predefined
1557 Macros,,,cpp,The C Preprocessor}.
1559 @item -aux-info @var{filename}
1561 Output to the given filename prototyped declarations for all functions
1562 declared and/or defined in a translation unit, including those in header
1563 files. This option is silently ignored in any language other than C@.
1565 Besides declarations, the file indicates, in comments, the origin of
1566 each declaration (source file and line), whether the declaration was
1567 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1568 @samp{O} for old, respectively, in the first character after the line
1569 number and the colon), and whether it came from a declaration or a
1570 definition (@samp{C} or @samp{F}, respectively, in the following
1571 character). In the case of function definitions, a K&R-style list of
1572 arguments followed by their declarations is also provided, inside
1573 comments, after the declaration.
1577 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1578 keyword, so that code can use these words as identifiers. You can use
1579 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1580 instead. @option{-ansi} implies @option{-fno-asm}.
1582 In C++, this switch only affects the @code{typeof} keyword, since
1583 @code{asm} and @code{inline} are standard keywords. You may want to
1584 use the @option{-fno-gnu-keywords} flag instead, which has the same
1585 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1586 switch only affects the @code{asm} and @code{typeof} keywords, since
1587 @code{inline} is a standard keyword in ISO C99.
1590 @itemx -fno-builtin-@var{function}
1591 @opindex fno-builtin
1592 @cindex built-in functions
1593 Don't recognize built-in functions that do not begin with
1594 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1595 functions provided by GCC}, for details of the functions affected,
1596 including those which are not built-in functions when @option{-ansi} or
1597 @option{-std} options for strict ISO C conformance are used because they
1598 do not have an ISO standard meaning.
1600 GCC normally generates special code to handle certain built-in functions
1601 more efficiently; for instance, calls to @code{alloca} may become single
1602 instructions that adjust the stack directly, and calls to @code{memcpy}
1603 may become inline copy loops. The resulting code is often both smaller
1604 and faster, but since the function calls no longer appear as such, you
1605 cannot set a breakpoint on those calls, nor can you change the behavior
1606 of the functions by linking with a different library. In addition,
1607 when a function is recognized as a built-in function, GCC may use
1608 information about that function to warn about problems with calls to
1609 that function, or to generate more efficient code, even if the
1610 resulting code still contains calls to that function. For example,
1611 warnings are given with @option{-Wformat} for bad calls to
1612 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1613 known not to modify global memory.
1615 With the @option{-fno-builtin-@var{function}} option
1616 only the built-in function @var{function} is
1617 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1618 function is named that is not built-in in this version of GCC, this
1619 option is ignored. There is no corresponding
1620 @option{-fbuiltin-@var{function}} option; if you wish to enable
1621 built-in functions selectively when using @option{-fno-builtin} or
1622 @option{-ffreestanding}, you may define macros such as:
1625 #define abs(n) __builtin_abs ((n))
1626 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1631 @cindex hosted environment
1633 Assert that compilation takes place in a hosted environment. This implies
1634 @option{-fbuiltin}. A hosted environment is one in which the
1635 entire standard library is available, and in which @code{main} has a return
1636 type of @code{int}. Examples are nearly everything except a kernel.
1637 This is equivalent to @option{-fno-freestanding}.
1639 @item -ffreestanding
1640 @opindex ffreestanding
1641 @cindex hosted environment
1643 Assert that compilation takes place in a freestanding environment. This
1644 implies @option{-fno-builtin}. A freestanding environment
1645 is one in which the standard library may not exist, and program startup may
1646 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1647 This is equivalent to @option{-fno-hosted}.
1649 @xref{Standards,,Language Standards Supported by GCC}, for details of
1650 freestanding and hosted environments.
1654 @cindex openmp parallel
1655 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1656 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1657 compiler generates parallel code according to the OpenMP Application
1658 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1659 implies @option{-pthread}, and thus is only supported on targets that
1660 have support for @option{-pthread}.
1662 @item -fms-extensions
1663 @opindex fms-extensions
1664 Accept some non-standard constructs used in Microsoft header files.
1666 Some cases of unnamed fields in structures and unions are only
1667 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1668 fields within structs/unions}, for details.
1672 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1673 options for strict ISO C conformance) implies @option{-trigraphs}.
1675 @item -no-integrated-cpp
1676 @opindex no-integrated-cpp
1677 Performs a compilation in two passes: preprocessing and compiling. This
1678 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1679 @option{-B} option. The user supplied compilation step can then add in
1680 an additional preprocessing step after normal preprocessing but before
1681 compiling. The default is to use the integrated cpp (internal cpp)
1683 The semantics of this option will change if "cc1", "cc1plus", and
1684 "cc1obj" are merged.
1686 @cindex traditional C language
1687 @cindex C language, traditional
1689 @itemx -traditional-cpp
1690 @opindex traditional-cpp
1691 @opindex traditional
1692 Formerly, these options caused GCC to attempt to emulate a pre-standard
1693 C compiler. They are now only supported with the @option{-E} switch.
1694 The preprocessor continues to support a pre-standard mode. See the GNU
1695 CPP manual for details.
1697 @item -fcond-mismatch
1698 @opindex fcond-mismatch
1699 Allow conditional expressions with mismatched types in the second and
1700 third arguments. The value of such an expression is void. This option
1701 is not supported for C++.
1703 @item -flax-vector-conversions
1704 @opindex flax-vector-conversions
1705 Allow implicit conversions between vectors with differing numbers of
1706 elements and/or incompatible element types. This option should not be
1709 @item -funsigned-char
1710 @opindex funsigned-char
1711 Let the type @code{char} be unsigned, like @code{unsigned char}.
1713 Each kind of machine has a default for what @code{char} should
1714 be. It is either like @code{unsigned char} by default or like
1715 @code{signed char} by default.
1717 Ideally, a portable program should always use @code{signed char} or
1718 @code{unsigned char} when it depends on the signedness of an object.
1719 But many programs have been written to use plain @code{char} and
1720 expect it to be signed, or expect it to be unsigned, depending on the
1721 machines they were written for. This option, and its inverse, let you
1722 make such a program work with the opposite default.
1724 The type @code{char} is always a distinct type from each of
1725 @code{signed char} or @code{unsigned char}, even though its behavior
1726 is always just like one of those two.
1729 @opindex fsigned-char
1730 Let the type @code{char} be signed, like @code{signed char}.
1732 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1733 the negative form of @option{-funsigned-char}. Likewise, the option
1734 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1736 @item -fsigned-bitfields
1737 @itemx -funsigned-bitfields
1738 @itemx -fno-signed-bitfields
1739 @itemx -fno-unsigned-bitfields
1740 @opindex fsigned-bitfields
1741 @opindex funsigned-bitfields
1742 @opindex fno-signed-bitfields
1743 @opindex fno-unsigned-bitfields
1744 These options control whether a bit-field is signed or unsigned, when the
1745 declaration does not use either @code{signed} or @code{unsigned}. By
1746 default, such a bit-field is signed, because this is consistent: the
1747 basic integer types such as @code{int} are signed types.
1750 @node C++ Dialect Options
1751 @section Options Controlling C++ Dialect
1753 @cindex compiler options, C++
1754 @cindex C++ options, command line
1755 @cindex options, C++
1756 This section describes the command-line options that are only meaningful
1757 for C++ programs; but you can also use most of the GNU compiler options
1758 regardless of what language your program is in. For example, you
1759 might compile a file @code{firstClass.C} like this:
1762 g++ -g -frepo -O -c firstClass.C
1766 In this example, only @option{-frepo} is an option meant
1767 only for C++ programs; you can use the other options with any
1768 language supported by GCC@.
1770 Here is a list of options that are @emph{only} for compiling C++ programs:
1774 @item -fabi-version=@var{n}
1775 @opindex fabi-version
1776 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1777 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1778 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1779 the version that conforms most closely to the C++ ABI specification.
1780 Therefore, the ABI obtained using version 0 will change as ABI bugs
1783 The default is version 2.
1785 Version 3 corrects an error in mangling a constant address as a
1788 Version 4 implements a standard mangling for vector types.
1790 See also @option{-Wabi}.
1792 @item -fno-access-control
1793 @opindex fno-access-control
1794 Turn off all access checking. This switch is mainly useful for working
1795 around bugs in the access control code.
1799 Check that the pointer returned by @code{operator new} is non-null
1800 before attempting to modify the storage allocated. This check is
1801 normally unnecessary because the C++ standard specifies that
1802 @code{operator new} will only return @code{0} if it is declared
1803 @samp{throw()}, in which case the compiler will always check the
1804 return value even without this option. In all other cases, when
1805 @code{operator new} has a non-empty exception specification, memory
1806 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1807 @samp{new (nothrow)}.
1809 @item -fconserve-space
1810 @opindex fconserve-space
1811 Put uninitialized or runtime-initialized global variables into the
1812 common segment, as C does. This saves space in the executable at the
1813 cost of not diagnosing duplicate definitions. If you compile with this
1814 flag and your program mysteriously crashes after @code{main()} has
1815 completed, you may have an object that is being destroyed twice because
1816 two definitions were merged.
1818 This option is no longer useful on most targets, now that support has
1819 been added for putting variables into BSS without making them common.
1821 @item -fno-deduce-init-list
1822 @opindex fno-deduce-init-list
1823 Disable deduction of a template type parameter as
1824 std::initializer_list from a brace-enclosed initializer list, i.e.
1827 template <class T> auto forward(T t) -> decltype (realfn (t))
1834 forward(@{1,2@}); // call forward<std::initializer_list<int>>
1838 This option is present because this deduction is an extension to the
1839 current specification in the C++0x working draft, and there was
1840 some concern about potential overload resolution problems.
1842 @item -ffriend-injection
1843 @opindex ffriend-injection
1844 Inject friend functions into the enclosing namespace, so that they are
1845 visible outside the scope of the class in which they are declared.
1846 Friend functions were documented to work this way in the old Annotated
1847 C++ Reference Manual, and versions of G++ before 4.1 always worked
1848 that way. However, in ISO C++ a friend function which is not declared
1849 in an enclosing scope can only be found using argument dependent
1850 lookup. This option causes friends to be injected as they were in
1853 This option is for compatibility, and may be removed in a future
1856 @item -fno-elide-constructors
1857 @opindex fno-elide-constructors
1858 The C++ standard allows an implementation to omit creating a temporary
1859 which is only used to initialize another object of the same type.
1860 Specifying this option disables that optimization, and forces G++ to
1861 call the copy constructor in all cases.
1863 @item -fno-enforce-eh-specs
1864 @opindex fno-enforce-eh-specs
1865 Don't generate code to check for violation of exception specifications
1866 at runtime. This option violates the C++ standard, but may be useful
1867 for reducing code size in production builds, much like defining
1868 @samp{NDEBUG}. This does not give user code permission to throw
1869 exceptions in violation of the exception specifications; the compiler
1870 will still optimize based on the specifications, so throwing an
1871 unexpected exception will result in undefined behavior.
1874 @itemx -fno-for-scope
1876 @opindex fno-for-scope
1877 If @option{-ffor-scope} is specified, the scope of variables declared in
1878 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1879 as specified by the C++ standard.
1880 If @option{-fno-for-scope} is specified, the scope of variables declared in
1881 a @i{for-init-statement} extends to the end of the enclosing scope,
1882 as was the case in old versions of G++, and other (traditional)
1883 implementations of C++.
1885 The default if neither flag is given to follow the standard,
1886 but to allow and give a warning for old-style code that would
1887 otherwise be invalid, or have different behavior.
1889 @item -fno-gnu-keywords
1890 @opindex fno-gnu-keywords
1891 Do not recognize @code{typeof} as a keyword, so that code can use this
1892 word as an identifier. You can use the keyword @code{__typeof__} instead.
1893 @option{-ansi} implies @option{-fno-gnu-keywords}.
1895 @item -fno-implicit-templates
1896 @opindex fno-implicit-templates
1897 Never emit code for non-inline templates which are instantiated
1898 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1899 @xref{Template Instantiation}, for more information.
1901 @item -fno-implicit-inline-templates
1902 @opindex fno-implicit-inline-templates
1903 Don't emit code for implicit instantiations of inline templates, either.
1904 The default is to handle inlines differently so that compiles with and
1905 without optimization will need the same set of explicit instantiations.
1907 @item -fno-implement-inlines
1908 @opindex fno-implement-inlines
1909 To save space, do not emit out-of-line copies of inline functions
1910 controlled by @samp{#pragma implementation}. This will cause linker
1911 errors if these functions are not inlined everywhere they are called.
1913 @item -fms-extensions
1914 @opindex fms-extensions
1915 Disable pedantic warnings about constructs used in MFC, such as implicit
1916 int and getting a pointer to member function via non-standard syntax.
1918 @item -fno-nonansi-builtins
1919 @opindex fno-nonansi-builtins
1920 Disable built-in declarations of functions that are not mandated by
1921 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1922 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1924 @item -fno-operator-names
1925 @opindex fno-operator-names
1926 Do not treat the operator name keywords @code{and}, @code{bitand},
1927 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1928 synonyms as keywords.
1930 @item -fno-optional-diags
1931 @opindex fno-optional-diags
1932 Disable diagnostics that the standard says a compiler does not need to
1933 issue. Currently, the only such diagnostic issued by G++ is the one for
1934 a name having multiple meanings within a class.
1937 @opindex fpermissive
1938 Downgrade some diagnostics about nonconformant code from errors to
1939 warnings. Thus, using @option{-fpermissive} will allow some
1940 nonconforming code to compile.
1942 @item -fno-pretty-templates
1943 @opindex fno-pretty-templates
1944 When an error message refers to a specialization of a function
1945 template, the compiler will normally print the signature of the
1946 template followed by the template arguments and any typedefs or
1947 typenames in the signature (e.g. @code{void f(T) [with T = int]}
1948 rather than @code{void f(int)}) so that it's clear which template is
1949 involved. When an error message refers to a specialization of a class
1950 template, the compiler will omit any template arguments which match
1951 the default template arguments for that template. If either of these
1952 behaviors make it harder to understand the error message rather than
1953 easier, using @option{-fno-pretty-templates} will disable them.
1957 Enable automatic template instantiation at link time. This option also
1958 implies @option{-fno-implicit-templates}. @xref{Template
1959 Instantiation}, for more information.
1963 Disable generation of information about every class with virtual
1964 functions for use by the C++ runtime type identification features
1965 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1966 of the language, you can save some space by using this flag. Note that
1967 exception handling uses the same information, but it will generate it as
1968 needed. The @samp{dynamic_cast} operator can still be used for casts that
1969 do not require runtime type information, i.e.@: casts to @code{void *} or to
1970 unambiguous base classes.
1974 Emit statistics about front-end processing at the end of the compilation.
1975 This information is generally only useful to the G++ development team.
1977 @item -ftemplate-depth=@var{n}
1978 @opindex ftemplate-depth
1979 Set the maximum instantiation depth for template classes to @var{n}.
1980 A limit on the template instantiation depth is needed to detect
1981 endless recursions during template class instantiation. ANSI/ISO C++
1982 conforming programs must not rely on a maximum depth greater than 17
1983 (changed to 1024 in C++0x).
1985 @item -fno-threadsafe-statics
1986 @opindex fno-threadsafe-statics
1987 Do not emit the extra code to use the routines specified in the C++
1988 ABI for thread-safe initialization of local statics. You can use this
1989 option to reduce code size slightly in code that doesn't need to be
1992 @item -fuse-cxa-atexit
1993 @opindex fuse-cxa-atexit
1994 Register destructors for objects with static storage duration with the
1995 @code{__cxa_atexit} function rather than the @code{atexit} function.
1996 This option is required for fully standards-compliant handling of static
1997 destructors, but will only work if your C library supports
1998 @code{__cxa_atexit}.
2000 @item -fno-use-cxa-get-exception-ptr
2001 @opindex fno-use-cxa-get-exception-ptr
2002 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2003 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
2004 if the runtime routine is not available.
2006 @item -fvisibility-inlines-hidden
2007 @opindex fvisibility-inlines-hidden
2008 This switch declares that the user does not attempt to compare
2009 pointers to inline methods where the addresses of the two functions
2010 were taken in different shared objects.
2012 The effect of this is that GCC may, effectively, mark inline methods with
2013 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2014 appear in the export table of a DSO and do not require a PLT indirection
2015 when used within the DSO@. Enabling this option can have a dramatic effect
2016 on load and link times of a DSO as it massively reduces the size of the
2017 dynamic export table when the library makes heavy use of templates.
2019 The behavior of this switch is not quite the same as marking the
2020 methods as hidden directly, because it does not affect static variables
2021 local to the function or cause the compiler to deduce that
2022 the function is defined in only one shared object.
2024 You may mark a method as having a visibility explicitly to negate the
2025 effect of the switch for that method. For example, if you do want to
2026 compare pointers to a particular inline method, you might mark it as
2027 having default visibility. Marking the enclosing class with explicit
2028 visibility will have no effect.
2030 Explicitly instantiated inline methods are unaffected by this option
2031 as their linkage might otherwise cross a shared library boundary.
2032 @xref{Template Instantiation}.
2034 @item -fvisibility-ms-compat
2035 @opindex fvisibility-ms-compat
2036 This flag attempts to use visibility settings to make GCC's C++
2037 linkage model compatible with that of Microsoft Visual Studio.
2039 The flag makes these changes to GCC's linkage model:
2043 It sets the default visibility to @code{hidden}, like
2044 @option{-fvisibility=hidden}.
2047 Types, but not their members, are not hidden by default.
2050 The One Definition Rule is relaxed for types without explicit
2051 visibility specifications which are defined in more than one different
2052 shared object: those declarations are permitted if they would have
2053 been permitted when this option was not used.
2056 In new code it is better to use @option{-fvisibility=hidden} and
2057 export those classes which are intended to be externally visible.
2058 Unfortunately it is possible for code to rely, perhaps accidentally,
2059 on the Visual Studio behavior.
2061 Among the consequences of these changes are that static data members
2062 of the same type with the same name but defined in different shared
2063 objects will be different, so changing one will not change the other;
2064 and that pointers to function members defined in different shared
2065 objects may not compare equal. When this flag is given, it is a
2066 violation of the ODR to define types with the same name differently.
2070 Do not use weak symbol support, even if it is provided by the linker.
2071 By default, G++ will use weak symbols if they are available. This
2072 option exists only for testing, and should not be used by end-users;
2073 it will result in inferior code and has no benefits. This option may
2074 be removed in a future release of G++.
2078 Do not search for header files in the standard directories specific to
2079 C++, but do still search the other standard directories. (This option
2080 is used when building the C++ library.)
2083 In addition, these optimization, warning, and code generation options
2084 have meanings only for C++ programs:
2087 @item -fno-default-inline
2088 @opindex fno-default-inline
2089 Do not assume @samp{inline} for functions defined inside a class scope.
2090 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2091 functions will have linkage like inline functions; they just won't be
2094 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2097 Warn when G++ generates code that is probably not compatible with the
2098 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2099 all such cases, there are probably some cases that are not warned about,
2100 even though G++ is generating incompatible code. There may also be
2101 cases where warnings are emitted even though the code that is generated
2104 You should rewrite your code to avoid these warnings if you are
2105 concerned about the fact that code generated by G++ may not be binary
2106 compatible with code generated by other compilers.
2108 The known incompatibilities in @option{-fabi-version=2} (the default) include:
2113 A template with a non-type template parameter of reference type is
2114 mangled incorrectly:
2117 template <int &> struct S @{@};
2121 This is fixed in @option{-fabi-version=3}.
2124 SIMD vector types declared using @code{__attribute ((vector_size))} are
2125 mangled in a non-standard way that does not allow for overloading of
2126 functions taking vectors of different sizes.
2128 The mangling is changed in @option{-fabi-version=4}.
2131 The known incompatibilities in @option{-fabi-version=1} include:
2136 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2137 pack data into the same byte as a base class. For example:
2140 struct A @{ virtual void f(); int f1 : 1; @};
2141 struct B : public A @{ int f2 : 1; @};
2145 In this case, G++ will place @code{B::f2} into the same byte
2146 as@code{A::f1}; other compilers will not. You can avoid this problem
2147 by explicitly padding @code{A} so that its size is a multiple of the
2148 byte size on your platform; that will cause G++ and other compilers to
2149 layout @code{B} identically.
2152 Incorrect handling of tail-padding for virtual bases. G++ does not use
2153 tail padding when laying out virtual bases. For example:
2156 struct A @{ virtual void f(); char c1; @};
2157 struct B @{ B(); char c2; @};
2158 struct C : public A, public virtual B @{@};
2162 In this case, G++ will not place @code{B} into the tail-padding for
2163 @code{A}; other compilers will. You can avoid this problem by
2164 explicitly padding @code{A} so that its size is a multiple of its
2165 alignment (ignoring virtual base classes); that will cause G++ and other
2166 compilers to layout @code{C} identically.
2169 Incorrect handling of bit-fields with declared widths greater than that
2170 of their underlying types, when the bit-fields appear in a union. For
2174 union U @{ int i : 4096; @};
2178 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2179 union too small by the number of bits in an @code{int}.
2182 Empty classes can be placed at incorrect offsets. For example:
2192 struct C : public B, public A @{@};
2196 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2197 it should be placed at offset zero. G++ mistakenly believes that the
2198 @code{A} data member of @code{B} is already at offset zero.
2201 Names of template functions whose types involve @code{typename} or
2202 template template parameters can be mangled incorrectly.
2205 template <typename Q>
2206 void f(typename Q::X) @{@}
2208 template <template <typename> class Q>
2209 void f(typename Q<int>::X) @{@}
2213 Instantiations of these templates may be mangled incorrectly.
2217 It also warns psABI related changes. The known psABI changes at this
2223 For SYSV/x86-64, when passing union with long double, it is changed to
2224 pass in memory as specified in psABI. For example:
2234 @code{union U} will always be passed in memory.
2238 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2239 @opindex Wctor-dtor-privacy
2240 @opindex Wno-ctor-dtor-privacy
2241 Warn when a class seems unusable because all the constructors or
2242 destructors in that class are private, and it has neither friends nor
2243 public static member functions.
2245 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2246 @opindex Wnon-virtual-dtor
2247 @opindex Wno-non-virtual-dtor
2248 Warn when a class has virtual functions and accessible non-virtual
2249 destructor, in which case it would be possible but unsafe to delete
2250 an instance of a derived class through a pointer to the base class.
2251 This warning is also enabled if -Weffc++ is specified.
2253 @item -Wreorder @r{(C++ and Objective-C++ only)}
2255 @opindex Wno-reorder
2256 @cindex reordering, warning
2257 @cindex warning for reordering of member initializers
2258 Warn when the order of member initializers given in the code does not
2259 match the order in which they must be executed. For instance:
2265 A(): j (0), i (1) @{ @}
2269 The compiler will rearrange the member initializers for @samp{i}
2270 and @samp{j} to match the declaration order of the members, emitting
2271 a warning to that effect. This warning is enabled by @option{-Wall}.
2274 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2277 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2280 Warn about violations of the following style guidelines from Scott Meyers'
2281 @cite{Effective C++} book:
2285 Item 11: Define a copy constructor and an assignment operator for classes
2286 with dynamically allocated memory.
2289 Item 12: Prefer initialization to assignment in constructors.
2292 Item 14: Make destructors virtual in base classes.
2295 Item 15: Have @code{operator=} return a reference to @code{*this}.
2298 Item 23: Don't try to return a reference when you must return an object.
2302 Also warn about violations of the following style guidelines from
2303 Scott Meyers' @cite{More Effective C++} book:
2307 Item 6: Distinguish between prefix and postfix forms of increment and
2308 decrement operators.
2311 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2315 When selecting this option, be aware that the standard library
2316 headers do not obey all of these guidelines; use @samp{grep -v}
2317 to filter out those warnings.
2319 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2320 @opindex Wstrict-null-sentinel
2321 @opindex Wno-strict-null-sentinel
2322 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2323 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2324 to @code{__null}. Although it is a null pointer constant not a null pointer,
2325 it is guaranteed to be of the same size as a pointer. But this use is
2326 not portable across different compilers.
2328 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2329 @opindex Wno-non-template-friend
2330 @opindex Wnon-template-friend
2331 Disable warnings when non-templatized friend functions are declared
2332 within a template. Since the advent of explicit template specification
2333 support in G++, if the name of the friend is an unqualified-id (i.e.,
2334 @samp{friend foo(int)}), the C++ language specification demands that the
2335 friend declare or define an ordinary, nontemplate function. (Section
2336 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2337 could be interpreted as a particular specialization of a templatized
2338 function. Because this non-conforming behavior is no longer the default
2339 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2340 check existing code for potential trouble spots and is on by default.
2341 This new compiler behavior can be turned off with
2342 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2343 but disables the helpful warning.
2345 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2346 @opindex Wold-style-cast
2347 @opindex Wno-old-style-cast
2348 Warn if an old-style (C-style) cast to a non-void type is used within
2349 a C++ program. The new-style casts (@samp{dynamic_cast},
2350 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2351 less vulnerable to unintended effects and much easier to search for.
2353 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2354 @opindex Woverloaded-virtual
2355 @opindex Wno-overloaded-virtual
2356 @cindex overloaded virtual fn, warning
2357 @cindex warning for overloaded virtual fn
2358 Warn when a function declaration hides virtual functions from a
2359 base class. For example, in:
2366 struct B: public A @{
2371 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2379 will fail to compile.
2381 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2382 @opindex Wno-pmf-conversions
2383 @opindex Wpmf-conversions
2384 Disable the diagnostic for converting a bound pointer to member function
2387 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2388 @opindex Wsign-promo
2389 @opindex Wno-sign-promo
2390 Warn when overload resolution chooses a promotion from unsigned or
2391 enumerated type to a signed type, over a conversion to an unsigned type of
2392 the same size. Previous versions of G++ would try to preserve
2393 unsignedness, but the standard mandates the current behavior.
2398 A& operator = (int);
2408 In this example, G++ will synthesize a default @samp{A& operator =
2409 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2412 @node Objective-C and Objective-C++ Dialect Options
2413 @section Options Controlling Objective-C and Objective-C++ Dialects
2415 @cindex compiler options, Objective-C and Objective-C++
2416 @cindex Objective-C and Objective-C++ options, command line
2417 @cindex options, Objective-C and Objective-C++
2418 (NOTE: This manual does not describe the Objective-C and Objective-C++
2419 languages themselves. See @xref{Standards,,Language Standards
2420 Supported by GCC}, for references.)
2422 This section describes the command-line options that are only meaningful
2423 for Objective-C and Objective-C++ programs, but you can also use most of
2424 the language-independent GNU compiler options.
2425 For example, you might compile a file @code{some_class.m} like this:
2428 gcc -g -fgnu-runtime -O -c some_class.m
2432 In this example, @option{-fgnu-runtime} is an option meant only for
2433 Objective-C and Objective-C++ programs; you can use the other options with
2434 any language supported by GCC@.
2436 Note that since Objective-C is an extension of the C language, Objective-C
2437 compilations may also use options specific to the C front-end (e.g.,
2438 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2439 C++-specific options (e.g., @option{-Wabi}).
2441 Here is a list of options that are @emph{only} for compiling Objective-C
2442 and Objective-C++ programs:
2445 @item -fconstant-string-class=@var{class-name}
2446 @opindex fconstant-string-class
2447 Use @var{class-name} as the name of the class to instantiate for each
2448 literal string specified with the syntax @code{@@"@dots{}"}. The default
2449 class name is @code{NXConstantString} if the GNU runtime is being used, and
2450 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2451 @option{-fconstant-cfstrings} option, if also present, will override the
2452 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2453 to be laid out as constant CoreFoundation strings.
2456 @opindex fgnu-runtime
2457 Generate object code compatible with the standard GNU Objective-C
2458 runtime. This is the default for most types of systems.
2460 @item -fnext-runtime
2461 @opindex fnext-runtime
2462 Generate output compatible with the NeXT runtime. This is the default
2463 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2464 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2467 @item -fno-nil-receivers
2468 @opindex fno-nil-receivers
2469 Assume that all Objective-C message dispatches (e.g.,
2470 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
2471 is not @code{nil}. This allows for more efficient entry points in the runtime
2472 to be used. Currently, this option is only available in conjunction with
2473 the NeXT runtime on Mac OS X 10.3 and later.
2475 @item -fobjc-call-cxx-cdtors
2476 @opindex fobjc-call-cxx-cdtors
2477 For each Objective-C class, check if any of its instance variables is a
2478 C++ object with a non-trivial default constructor. If so, synthesize a
2479 special @code{- (id) .cxx_construct} instance method that will run
2480 non-trivial default constructors on any such instance variables, in order,
2481 and then return @code{self}. Similarly, check if any instance variable
2482 is a C++ object with a non-trivial destructor, and if so, synthesize a
2483 special @code{- (void) .cxx_destruct} method that will run
2484 all such default destructors, in reverse order.
2486 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
2487 thusly generated will only operate on instance variables declared in the
2488 current Objective-C class, and not those inherited from superclasses. It
2489 is the responsibility of the Objective-C runtime to invoke all such methods
2490 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
2491 will be invoked by the runtime immediately after a new object
2492 instance is allocated; the @code{- (void) .cxx_destruct} methods will
2493 be invoked immediately before the runtime deallocates an object instance.
2495 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2496 support for invoking the @code{- (id) .cxx_construct} and
2497 @code{- (void) .cxx_destruct} methods.
2499 @item -fobjc-direct-dispatch
2500 @opindex fobjc-direct-dispatch
2501 Allow fast jumps to the message dispatcher. On Darwin this is
2502 accomplished via the comm page.
2504 @item -fobjc-exceptions
2505 @opindex fobjc-exceptions
2506 Enable syntactic support for structured exception handling in Objective-C,
2507 similar to what is offered by C++ and Java. This option is
2508 unavailable in conjunction with the NeXT runtime on Mac OS X 10.2 and
2517 @@catch (AnObjCClass *exc) @{
2524 @@catch (AnotherClass *exc) @{
2527 @@catch (id allOthers) @{
2537 The @code{@@throw} statement may appear anywhere in an Objective-C or
2538 Objective-C++ program; when used inside of a @code{@@catch} block, the
2539 @code{@@throw} may appear without an argument (as shown above), in which case
2540 the object caught by the @code{@@catch} will be rethrown.
2542 Note that only (pointers to) Objective-C objects may be thrown and
2543 caught using this scheme. When an object is thrown, it will be caught
2544 by the nearest @code{@@catch} clause capable of handling objects of that type,
2545 analogously to how @code{catch} blocks work in C++ and Java. A
2546 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2547 any and all Objective-C exceptions not caught by previous @code{@@catch}
2550 The @code{@@finally} clause, if present, will be executed upon exit from the
2551 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2552 regardless of whether any exceptions are thrown, caught or rethrown
2553 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2554 of the @code{finally} clause in Java.
2556 There are several caveats to using the new exception mechanism:
2560 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2561 idioms provided by the @code{NSException} class, the new
2562 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2563 systems, due to additional functionality needed in the (NeXT) Objective-C
2567 As mentioned above, the new exceptions do not support handling
2568 types other than Objective-C objects. Furthermore, when used from
2569 Objective-C++, the Objective-C exception model does not interoperate with C++
2570 exceptions at this time. This means you cannot @code{@@throw} an exception
2571 from Objective-C and @code{catch} it in C++, or vice versa
2572 (i.e., @code{throw @dots{} @@catch}).
2575 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2576 blocks for thread-safe execution:
2579 @@synchronized (ObjCClass *guard) @{
2584 Upon entering the @code{@@synchronized} block, a thread of execution shall
2585 first check whether a lock has been placed on the corresponding @code{guard}
2586 object by another thread. If it has, the current thread shall wait until
2587 the other thread relinquishes its lock. Once @code{guard} becomes available,
2588 the current thread will place its own lock on it, execute the code contained in
2589 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2590 making @code{guard} available to other threads).
2592 Unlike Java, Objective-C does not allow for entire methods to be marked
2593 @code{@@synchronized}. Note that throwing exceptions out of
2594 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2595 to be unlocked properly.
2599 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2601 @item -freplace-objc-classes
2602 @opindex freplace-objc-classes
2603 Emit a special marker instructing @command{ld(1)} not to statically link in
2604 the resulting object file, and allow @command{dyld(1)} to load it in at
2605 run time instead. This is used in conjunction with the Fix-and-Continue
2606 debugging mode, where the object file in question may be recompiled and
2607 dynamically reloaded in the course of program execution, without the need
2608 to restart the program itself. Currently, Fix-and-Continue functionality
2609 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2614 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2615 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2616 compile time) with static class references that get initialized at load time,
2617 which improves run-time performance. Specifying the @option{-fzero-link} flag
2618 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2619 to be retained. This is useful in Zero-Link debugging mode, since it allows
2620 for individual class implementations to be modified during program execution.
2624 Dump interface declarations for all classes seen in the source file to a
2625 file named @file{@var{sourcename}.decl}.
2627 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2628 @opindex Wassign-intercept
2629 @opindex Wno-assign-intercept
2630 Warn whenever an Objective-C assignment is being intercepted by the
2633 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2634 @opindex Wno-protocol
2636 If a class is declared to implement a protocol, a warning is issued for
2637 every method in the protocol that is not implemented by the class. The
2638 default behavior is to issue a warning for every method not explicitly
2639 implemented in the class, even if a method implementation is inherited
2640 from the superclass. If you use the @option{-Wno-protocol} option, then
2641 methods inherited from the superclass are considered to be implemented,
2642 and no warning is issued for them.
2644 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2646 @opindex Wno-selector
2647 Warn if multiple methods of different types for the same selector are
2648 found during compilation. The check is performed on the list of methods
2649 in the final stage of compilation. Additionally, a check is performed
2650 for each selector appearing in a @code{@@selector(@dots{})}
2651 expression, and a corresponding method for that selector has been found
2652 during compilation. Because these checks scan the method table only at
2653 the end of compilation, these warnings are not produced if the final
2654 stage of compilation is not reached, for example because an error is
2655 found during compilation, or because the @option{-fsyntax-only} option is
2658 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2659 @opindex Wstrict-selector-match
2660 @opindex Wno-strict-selector-match
2661 Warn if multiple methods with differing argument and/or return types are
2662 found for a given selector when attempting to send a message using this
2663 selector to a receiver of type @code{id} or @code{Class}. When this flag
2664 is off (which is the default behavior), the compiler will omit such warnings
2665 if any differences found are confined to types which share the same size
2668 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2669 @opindex Wundeclared-selector
2670 @opindex Wno-undeclared-selector
2671 Warn if a @code{@@selector(@dots{})} expression referring to an
2672 undeclared selector is found. A selector is considered undeclared if no
2673 method with that name has been declared before the
2674 @code{@@selector(@dots{})} expression, either explicitly in an
2675 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2676 an @code{@@implementation} section. This option always performs its
2677 checks as soon as a @code{@@selector(@dots{})} expression is found,
2678 while @option{-Wselector} only performs its checks in the final stage of
2679 compilation. This also enforces the coding style convention
2680 that methods and selectors must be declared before being used.
2682 @item -print-objc-runtime-info
2683 @opindex print-objc-runtime-info
2684 Generate C header describing the largest structure that is passed by
2689 @node Language Independent Options
2690 @section Options to Control Diagnostic Messages Formatting
2691 @cindex options to control diagnostics formatting
2692 @cindex diagnostic messages
2693 @cindex message formatting
2695 Traditionally, diagnostic messages have been formatted irrespective of
2696 the output device's aspect (e.g.@: its width, @dots{}). The options described
2697 below can be used to control the diagnostic messages formatting
2698 algorithm, e.g.@: how many characters per line, how often source location
2699 information should be reported. Right now, only the C++ front end can
2700 honor these options. However it is expected, in the near future, that
2701 the remaining front ends would be able to digest them correctly.
2704 @item -fmessage-length=@var{n}
2705 @opindex fmessage-length
2706 Try to format error messages so that they fit on lines of about @var{n}
2707 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2708 the front ends supported by GCC@. If @var{n} is zero, then no
2709 line-wrapping will be done; each error message will appear on a single
2712 @opindex fdiagnostics-show-location
2713 @item -fdiagnostics-show-location=once
2714 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2715 reporter to emit @emph{once} source location information; that is, in
2716 case the message is too long to fit on a single physical line and has to
2717 be wrapped, the source location won't be emitted (as prefix) again,
2718 over and over, in subsequent continuation lines. This is the default
2721 @item -fdiagnostics-show-location=every-line
2722 Only meaningful in line-wrapping mode. Instructs the diagnostic
2723 messages reporter to emit the same source location information (as
2724 prefix) for physical lines that result from the process of breaking
2725 a message which is too long to fit on a single line.
2727 @item -fdiagnostics-show-option
2728 @opindex fdiagnostics-show-option
2729 This option instructs the diagnostic machinery to add text to each
2730 diagnostic emitted, which indicates which command line option directly
2731 controls that diagnostic, when such an option is known to the
2732 diagnostic machinery.
2734 @item -Wcoverage-mismatch
2735 @opindex Wcoverage-mismatch
2736 Warn if feedback profiles do not match when using the
2737 @option{-fprofile-use} option.
2738 If a source file was changed between @option{-fprofile-gen} and
2739 @option{-fprofile-use}, the files with the profile feedback can fail
2740 to match the source file and GCC can not use the profile feedback
2741 information. By default, GCC emits an error message in this case.
2742 The option @option{-Wcoverage-mismatch} emits a warning instead of an
2743 error. GCC does not use appropriate feedback profiles, so using this
2744 option can result in poorly optimized code. This option is useful
2745 only in the case of very minor changes such as bug fixes to an
2750 @node Warning Options
2751 @section Options to Request or Suppress Warnings
2752 @cindex options to control warnings
2753 @cindex warning messages
2754 @cindex messages, warning
2755 @cindex suppressing warnings
2757 Warnings are diagnostic messages that report constructions which
2758 are not inherently erroneous but which are risky or suggest there
2759 may have been an error.
2761 The following language-independent options do not enable specific
2762 warnings but control the kinds of diagnostics produced by GCC.
2765 @cindex syntax checking
2767 @opindex fsyntax-only
2768 Check the code for syntax errors, but don't do anything beyond that.
2772 Inhibit all warning messages.
2777 Make all warnings into errors.
2782 Make the specified warning into an error. The specifier for a warning
2783 is appended, for example @option{-Werror=switch} turns the warnings
2784 controlled by @option{-Wswitch} into errors. This switch takes a
2785 negative form, to be used to negate @option{-Werror} for specific
2786 warnings, for example @option{-Wno-error=switch} makes
2787 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2788 is in effect. You can use the @option{-fdiagnostics-show-option}
2789 option to have each controllable warning amended with the option which
2790 controls it, to determine what to use with this option.
2792 Note that specifying @option{-Werror=}@var{foo} automatically implies
2793 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2796 @item -Wfatal-errors
2797 @opindex Wfatal-errors
2798 @opindex Wno-fatal-errors
2799 This option causes the compiler to abort compilation on the first error
2800 occurred rather than trying to keep going and printing further error
2805 You can request many specific warnings with options beginning
2806 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2807 implicit declarations. Each of these specific warning options also
2808 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2809 example, @option{-Wno-implicit}. This manual lists only one of the
2810 two forms, whichever is not the default. For further,
2811 language-specific options also refer to @ref{C++ Dialect Options} and
2812 @ref{Objective-C and Objective-C++ Dialect Options}.
2817 Issue all the warnings demanded by strict ISO C and ISO C++;
2818 reject all programs that use forbidden extensions, and some other
2819 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2820 version of the ISO C standard specified by any @option{-std} option used.
2822 Valid ISO C and ISO C++ programs should compile properly with or without
2823 this option (though a rare few will require @option{-ansi} or a
2824 @option{-std} option specifying the required version of ISO C)@. However,
2825 without this option, certain GNU extensions and traditional C and C++
2826 features are supported as well. With this option, they are rejected.
2828 @option{-pedantic} does not cause warning messages for use of the
2829 alternate keywords whose names begin and end with @samp{__}. Pedantic
2830 warnings are also disabled in the expression that follows
2831 @code{__extension__}. However, only system header files should use
2832 these escape routes; application programs should avoid them.
2833 @xref{Alternate Keywords}.
2835 Some users try to use @option{-pedantic} to check programs for strict ISO
2836 C conformance. They soon find that it does not do quite what they want:
2837 it finds some non-ISO practices, but not all---only those for which
2838 ISO C @emph{requires} a diagnostic, and some others for which
2839 diagnostics have been added.
2841 A feature to report any failure to conform to ISO C might be useful in
2842 some instances, but would require considerable additional work and would
2843 be quite different from @option{-pedantic}. We don't have plans to
2844 support such a feature in the near future.
2846 Where the standard specified with @option{-std} represents a GNU
2847 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
2848 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2849 extended dialect is based. Warnings from @option{-pedantic} are given
2850 where they are required by the base standard. (It would not make sense
2851 for such warnings to be given only for features not in the specified GNU
2852 C dialect, since by definition the GNU dialects of C include all
2853 features the compiler supports with the given option, and there would be
2854 nothing to warn about.)
2856 @item -pedantic-errors
2857 @opindex pedantic-errors
2858 Like @option{-pedantic}, except that errors are produced rather than
2864 This enables all the warnings about constructions that some users
2865 consider questionable, and that are easy to avoid (or modify to
2866 prevent the warning), even in conjunction with macros. This also
2867 enables some language-specific warnings described in @ref{C++ Dialect
2868 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2870 @option{-Wall} turns on the following warning flags:
2872 @gccoptlist{-Waddress @gol
2873 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2875 -Wchar-subscripts @gol
2876 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
2878 -Wimplicit-function-declaration @gol
2881 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2882 -Wmissing-braces @gol
2888 -Wsequence-point @gol
2889 -Wsign-compare @r{(only in C++)} @gol
2890 -Wstrict-aliasing @gol
2891 -Wstrict-overflow=1 @gol
2894 -Wuninitialized @gol
2895 -Wunknown-pragmas @gol
2896 -Wunused-function @gol
2899 -Wunused-variable @gol
2900 -Wvolatile-register-var @gol
2903 Note that some warning flags are not implied by @option{-Wall}. Some of
2904 them warn about constructions that users generally do not consider
2905 questionable, but which occasionally you might wish to check for;
2906 others warn about constructions that are necessary or hard to avoid in
2907 some cases, and there is no simple way to modify the code to suppress
2908 the warning. Some of them are enabled by @option{-Wextra} but many of
2909 them must be enabled individually.
2915 This enables some extra warning flags that are not enabled by
2916 @option{-Wall}. (This option used to be called @option{-W}. The older
2917 name is still supported, but the newer name is more descriptive.)
2919 @gccoptlist{-Wclobbered @gol
2921 -Wignored-qualifiers @gol
2922 -Wmissing-field-initializers @gol
2923 -Wmissing-parameter-type @r{(C only)} @gol
2924 -Wold-style-declaration @r{(C only)} @gol
2925 -Woverride-init @gol
2928 -Wuninitialized @gol
2929 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2932 The option @option{-Wextra} also prints warning messages for the
2938 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2939 @samp{>}, or @samp{>=}.
2942 (C++ only) An enumerator and a non-enumerator both appear in a
2943 conditional expression.
2946 (C++ only) Ambiguous virtual bases.
2949 (C++ only) Subscripting an array which has been declared @samp{register}.
2952 (C++ only) Taking the address of a variable which has been declared
2956 (C++ only) A base class is not initialized in a derived class' copy
2961 @item -Wchar-subscripts
2962 @opindex Wchar-subscripts
2963 @opindex Wno-char-subscripts
2964 Warn if an array subscript has type @code{char}. This is a common cause
2965 of error, as programmers often forget that this type is signed on some
2967 This warning is enabled by @option{-Wall}.
2971 @opindex Wno-comment
2972 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2973 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2974 This warning is enabled by @option{-Wall}.
2977 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
2979 Suppress warning messages emitted by @code{#warning} directives.
2984 @opindex ffreestanding
2985 @opindex fno-builtin
2986 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2987 the arguments supplied have types appropriate to the format string
2988 specified, and that the conversions specified in the format string make
2989 sense. This includes standard functions, and others specified by format
2990 attributes (@pxref{Function Attributes}), in the @code{printf},
2991 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2992 not in the C standard) families (or other target-specific families).
2993 Which functions are checked without format attributes having been
2994 specified depends on the standard version selected, and such checks of
2995 functions without the attribute specified are disabled by
2996 @option{-ffreestanding} or @option{-fno-builtin}.
2998 The formats are checked against the format features supported by GNU
2999 libc version 2.2. These include all ISO C90 and C99 features, as well
3000 as features from the Single Unix Specification and some BSD and GNU
3001 extensions. Other library implementations may not support all these
3002 features; GCC does not support warning about features that go beyond a
3003 particular library's limitations. However, if @option{-pedantic} is used
3004 with @option{-Wformat}, warnings will be given about format features not
3005 in the selected standard version (but not for @code{strfmon} formats,
3006 since those are not in any version of the C standard). @xref{C Dialect
3007 Options,,Options Controlling C Dialect}.
3009 Since @option{-Wformat} also checks for null format arguments for
3010 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
3012 @option{-Wformat} is included in @option{-Wall}. For more control over some
3013 aspects of format checking, the options @option{-Wformat-y2k},
3014 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
3015 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
3016 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
3019 @opindex Wformat-y2k
3020 @opindex Wno-format-y2k
3021 If @option{-Wformat} is specified, also warn about @code{strftime}
3022 formats which may yield only a two-digit year.
3024 @item -Wno-format-contains-nul
3025 @opindex Wno-format-contains-nul
3026 @opindex Wformat-contains-nul
3027 If @option{-Wformat} is specified, do not warn about format strings that
3030 @item -Wno-format-extra-args
3031 @opindex Wno-format-extra-args
3032 @opindex Wformat-extra-args
3033 If @option{-Wformat} is specified, do not warn about excess arguments to a
3034 @code{printf} or @code{scanf} format function. The C standard specifies
3035 that such arguments are ignored.
3037 Where the unused arguments lie between used arguments that are
3038 specified with @samp{$} operand number specifications, normally
3039 warnings are still given, since the implementation could not know what
3040 type to pass to @code{va_arg} to skip the unused arguments. However,
3041 in the case of @code{scanf} formats, this option will suppress the
3042 warning if the unused arguments are all pointers, since the Single
3043 Unix Specification says that such unused arguments are allowed.
3045 @item -Wno-format-zero-length @r{(C and Objective-C only)}
3046 @opindex Wno-format-zero-length
3047 @opindex Wformat-zero-length
3048 If @option{-Wformat} is specified, do not warn about zero-length formats.
3049 The C standard specifies that zero-length formats are allowed.
3051 @item -Wformat-nonliteral
3052 @opindex Wformat-nonliteral
3053 @opindex Wno-format-nonliteral
3054 If @option{-Wformat} is specified, also warn if the format string is not a
3055 string literal and so cannot be checked, unless the format function
3056 takes its format arguments as a @code{va_list}.
3058 @item -Wformat-security
3059 @opindex Wformat-security
3060 @opindex Wno-format-security
3061 If @option{-Wformat} is specified, also warn about uses of format
3062 functions that represent possible security problems. At present, this
3063 warns about calls to @code{printf} and @code{scanf} functions where the
3064 format string is not a string literal and there are no format arguments,
3065 as in @code{printf (foo);}. This may be a security hole if the format
3066 string came from untrusted input and contains @samp{%n}. (This is
3067 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3068 in future warnings may be added to @option{-Wformat-security} that are not
3069 included in @option{-Wformat-nonliteral}.)
3073 @opindex Wno-format=2
3074 Enable @option{-Wformat} plus format checks not included in
3075 @option{-Wformat}. Currently equivalent to @samp{-Wformat
3076 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
3078 @item -Wnonnull @r{(C and Objective-C only)}
3080 @opindex Wno-nonnull
3081 Warn about passing a null pointer for arguments marked as
3082 requiring a non-null value by the @code{nonnull} function attribute.
3084 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3085 can be disabled with the @option{-Wno-nonnull} option.
3087 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3089 @opindex Wno-init-self
3090 Warn about uninitialized variables which are initialized with themselves.
3091 Note this option can only be used with the @option{-Wuninitialized} option.
3093 For example, GCC will warn about @code{i} being uninitialized in the
3094 following snippet only when @option{-Winit-self} has been specified:
3105 @item -Wimplicit-int @r{(C and Objective-C only)}
3106 @opindex Wimplicit-int
3107 @opindex Wno-implicit-int
3108 Warn when a declaration does not specify a type.
3109 This warning is enabled by @option{-Wall}.
3111 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3112 @opindex Wimplicit-function-declaration
3113 @opindex Wno-implicit-function-declaration
3114 Give a warning whenever a function is used before being declared. In
3115 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3116 enabled by default and it is made into an error by
3117 @option{-pedantic-errors}. This warning is also enabled by
3122 @opindex Wno-implicit
3123 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3124 This warning is enabled by @option{-Wall}.
3126 @item -Wignored-qualifiers @r{(C and C++ only)}
3127 @opindex Wignored-qualifiers
3128 @opindex Wno-ignored-qualifiers
3129 Warn if the return type of a function has a type qualifier
3130 such as @code{const}. For ISO C such a type qualifier has no effect,
3131 since the value returned by a function is not an lvalue.
3132 For C++, the warning is only emitted for scalar types or @code{void}.
3133 ISO C prohibits qualified @code{void} return types on function
3134 definitions, so such return types always receive a warning
3135 even without this option.
3137 This warning is also enabled by @option{-Wextra}.
3142 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3143 a function with external linkage, returning int, taking either zero
3144 arguments, two, or three arguments of appropriate types. This warning
3145 is enabled by default in C++ and is enabled by either @option{-Wall}
3146 or @option{-pedantic}.
3148 @item -Wmissing-braces
3149 @opindex Wmissing-braces
3150 @opindex Wno-missing-braces
3151 Warn if an aggregate or union initializer is not fully bracketed. In
3152 the following example, the initializer for @samp{a} is not fully
3153 bracketed, but that for @samp{b} is fully bracketed.
3156 int a[2][2] = @{ 0, 1, 2, 3 @};
3157 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3160 This warning is enabled by @option{-Wall}.
3162 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3163 @opindex Wmissing-include-dirs
3164 @opindex Wno-missing-include-dirs
3165 Warn if a user-supplied include directory does not exist.
3168 @opindex Wparentheses
3169 @opindex Wno-parentheses
3170 Warn if parentheses are omitted in certain contexts, such
3171 as when there is an assignment in a context where a truth value
3172 is expected, or when operators are nested whose precedence people
3173 often get confused about.
3175 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3176 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3177 interpretation from that of ordinary mathematical notation.
3179 Also warn about constructions where there may be confusion to which
3180 @code{if} statement an @code{else} branch belongs. Here is an example of
3195 In C/C++, every @code{else} branch belongs to the innermost possible
3196 @code{if} statement, which in this example is @code{if (b)}. This is
3197 often not what the programmer expected, as illustrated in the above
3198 example by indentation the programmer chose. When there is the
3199 potential for this confusion, GCC will issue a warning when this flag
3200 is specified. To eliminate the warning, add explicit braces around
3201 the innermost @code{if} statement so there is no way the @code{else}
3202 could belong to the enclosing @code{if}. The resulting code would
3219 This warning is enabled by @option{-Wall}.
3221 @item -Wsequence-point
3222 @opindex Wsequence-point
3223 @opindex Wno-sequence-point
3224 Warn about code that may have undefined semantics because of violations
3225 of sequence point rules in the C and C++ standards.
3227 The C and C++ standards defines the order in which expressions in a C/C++
3228 program are evaluated in terms of @dfn{sequence points}, which represent
3229 a partial ordering between the execution of parts of the program: those
3230 executed before the sequence point, and those executed after it. These
3231 occur after the evaluation of a full expression (one which is not part
3232 of a larger expression), after the evaluation of the first operand of a
3233 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3234 function is called (but after the evaluation of its arguments and the
3235 expression denoting the called function), and in certain other places.
3236 Other than as expressed by the sequence point rules, the order of
3237 evaluation of subexpressions of an expression is not specified. All
3238 these rules describe only a partial order rather than a total order,
3239 since, for example, if two functions are called within one expression
3240 with no sequence point between them, the order in which the functions
3241 are called is not specified. However, the standards committee have
3242 ruled that function calls do not overlap.
3244 It is not specified when between sequence points modifications to the
3245 values of objects take effect. Programs whose behavior depends on this
3246 have undefined behavior; the C and C++ standards specify that ``Between
3247 the previous and next sequence point an object shall have its stored
3248 value modified at most once by the evaluation of an expression.
3249 Furthermore, the prior value shall be read only to determine the value
3250 to be stored.''. If a program breaks these rules, the results on any
3251 particular implementation are entirely unpredictable.
3253 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3254 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3255 diagnosed by this option, and it may give an occasional false positive
3256 result, but in general it has been found fairly effective at detecting
3257 this sort of problem in programs.
3259 The standard is worded confusingly, therefore there is some debate
3260 over the precise meaning of the sequence point rules in subtle cases.
3261 Links to discussions of the problem, including proposed formal
3262 definitions, may be found on the GCC readings page, at
3263 @w{@uref{http://gcc.gnu.org/readings.html}}.
3265 This warning is enabled by @option{-Wall} for C and C++.
3268 @opindex Wreturn-type
3269 @opindex Wno-return-type
3270 Warn whenever a function is defined with a return-type that defaults
3271 to @code{int}. Also warn about any @code{return} statement with no
3272 return-value in a function whose return-type is not @code{void}
3273 (falling off the end of the function body is considered returning
3274 without a value), and about a @code{return} statement with an
3275 expression in a function whose return-type is @code{void}.
3277 For C++, a function without return type always produces a diagnostic
3278 message, even when @option{-Wno-return-type} is specified. The only
3279 exceptions are @samp{main} and functions defined in system headers.
3281 This warning is enabled by @option{-Wall}.
3286 Warn whenever a @code{switch} statement has an index of enumerated type
3287 and lacks a @code{case} for one or more of the named codes of that
3288 enumeration. (The presence of a @code{default} label prevents this
3289 warning.) @code{case} labels outside the enumeration range also
3290 provoke warnings when this option is used (even if there is a
3291 @code{default} label).
3292 This warning is enabled by @option{-Wall}.
3294 @item -Wswitch-default
3295 @opindex Wswitch-default
3296 @opindex Wno-switch-default
3297 Warn whenever a @code{switch} statement does not have a @code{default}
3301 @opindex Wswitch-enum
3302 @opindex Wno-switch-enum
3303 Warn whenever a @code{switch} statement has an index of enumerated type
3304 and lacks a @code{case} for one or more of the named codes of that
3305 enumeration. @code{case} labels outside the enumeration range also
3306 provoke warnings when this option is used. The only difference
3307 between @option{-Wswitch} and this option is that this option gives a
3308 warning about an omitted enumeration code even if there is a
3309 @code{default} label.
3311 @item -Wsync-nand @r{(C and C++ only)}
3313 @opindex Wno-sync-nand
3314 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3315 built-in functions are used. These functions changed semantics in GCC 4.4.
3319 @opindex Wno-trigraphs
3320 Warn if any trigraphs are encountered that might change the meaning of
3321 the program (trigraphs within comments are not warned about).
3322 This warning is enabled by @option{-Wall}.
3324 @item -Wunused-function
3325 @opindex Wunused-function
3326 @opindex Wno-unused-function
3327 Warn whenever a static function is declared but not defined or a
3328 non-inline static function is unused.
3329 This warning is enabled by @option{-Wall}.
3331 @item -Wunused-label
3332 @opindex Wunused-label
3333 @opindex Wno-unused-label
3334 Warn whenever a label is declared but not used.
3335 This warning is enabled by @option{-Wall}.
3337 To suppress this warning use the @samp{unused} attribute
3338 (@pxref{Variable Attributes}).
3340 @item -Wunused-parameter
3341 @opindex Wunused-parameter
3342 @opindex Wno-unused-parameter
3343 Warn whenever a function parameter is unused aside from its declaration.
3345 To suppress this warning use the @samp{unused} attribute
3346 (@pxref{Variable Attributes}).
3348 @item -Wno-unused-result
3349 @opindex Wunused-result
3350 @opindex Wno-unused-result
3351 Do not warn if a caller of a function marked with attribute
3352 @code{warn_unused_result} (@pxref{Variable Attributes}) does not use
3353 its return value. The default is @option{-Wunused-result}.
3355 @item -Wunused-variable
3356 @opindex Wunused-variable
3357 @opindex Wno-unused-variable
3358 Warn whenever a local variable or non-constant static variable is unused
3359 aside from its declaration.
3360 This warning is enabled by @option{-Wall}.
3362 To suppress this warning use the @samp{unused} attribute
3363 (@pxref{Variable Attributes}).
3365 @item -Wunused-value
3366 @opindex Wunused-value
3367 @opindex Wno-unused-value
3368 Warn whenever a statement computes a result that is explicitly not
3369 used. To suppress this warning cast the unused expression to
3370 @samp{void}. This includes an expression-statement or the left-hand
3371 side of a comma expression that contains no side effects. For example,
3372 an expression such as @samp{x[i,j]} will cause a warning, while
3373 @samp{x[(void)i,j]} will not.
3375 This warning is enabled by @option{-Wall}.
3380 All the above @option{-Wunused} options combined.
3382 In order to get a warning about an unused function parameter, you must
3383 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3384 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3386 @item -Wuninitialized
3387 @opindex Wuninitialized
3388 @opindex Wno-uninitialized
3389 Warn if an automatic variable is used without first being initialized
3390 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3391 warn if a non-static reference or non-static @samp{const} member
3392 appears in a class without constructors.
3394 If you want to warn about code which uses the uninitialized value of the
3395 variable in its own initializer, use the @option{-Winit-self} option.
3397 These warnings occur for individual uninitialized or clobbered
3398 elements of structure, union or array variables as well as for
3399 variables which are uninitialized or clobbered as a whole. They do
3400 not occur for variables or elements declared @code{volatile}. Because
3401 these warnings depend on optimization, the exact variables or elements
3402 for which there are warnings will depend on the precise optimization
3403 options and version of GCC used.
3405 Note that there may be no warning about a variable that is used only
3406 to compute a value that itself is never used, because such
3407 computations may be deleted by data flow analysis before the warnings
3410 These warnings are made optional because GCC is not smart
3411 enough to see all the reasons why the code might be correct
3412 despite appearing to have an error. Here is one example of how
3433 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3434 always initialized, but GCC doesn't know this. Here is
3435 another common case:
3440 if (change_y) save_y = y, y = new_y;
3442 if (change_y) y = save_y;
3447 This has no bug because @code{save_y} is used only if it is set.
3449 @cindex @code{longjmp} warnings
3450 This option also warns when a non-volatile automatic variable might be
3451 changed by a call to @code{longjmp}. These warnings as well are possible
3452 only in optimizing compilation.
3454 The compiler sees only the calls to @code{setjmp}. It cannot know
3455 where @code{longjmp} will be called; in fact, a signal handler could
3456 call it at any point in the code. As a result, you may get a warning
3457 even when there is in fact no problem because @code{longjmp} cannot
3458 in fact be called at the place which would cause a problem.
3460 Some spurious warnings can be avoided if you declare all the functions
3461 you use that never return as @code{noreturn}. @xref{Function
3464 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3466 @item -Wunknown-pragmas
3467 @opindex Wunknown-pragmas
3468 @opindex Wno-unknown-pragmas
3469 @cindex warning for unknown pragmas
3470 @cindex unknown pragmas, warning
3471 @cindex pragmas, warning of unknown
3472 Warn when a #pragma directive is encountered which is not understood by
3473 GCC@. If this command line option is used, warnings will even be issued
3474 for unknown pragmas in system header files. This is not the case if
3475 the warnings were only enabled by the @option{-Wall} command line option.
3478 @opindex Wno-pragmas
3480 Do not warn about misuses of pragmas, such as incorrect parameters,
3481 invalid syntax, or conflicts between pragmas. See also
3482 @samp{-Wunknown-pragmas}.
3484 @item -Wstrict-aliasing
3485 @opindex Wstrict-aliasing
3486 @opindex Wno-strict-aliasing
3487 This option is only active when @option{-fstrict-aliasing} is active.
3488 It warns about code which might break the strict aliasing rules that the
3489 compiler is using for optimization. The warning does not catch all
3490 cases, but does attempt to catch the more common pitfalls. It is
3491 included in @option{-Wall}.
3492 It is equivalent to @option{-Wstrict-aliasing=3}
3494 @item -Wstrict-aliasing=n
3495 @opindex Wstrict-aliasing=n
3496 @opindex Wno-strict-aliasing=n
3497 This option is only active when @option{-fstrict-aliasing} is active.
3498 It warns about code which might break the strict aliasing rules that the
3499 compiler is using for optimization.
3500 Higher levels correspond to higher accuracy (fewer false positives).
3501 Higher levels also correspond to more effort, similar to the way -O works.
3502 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3505 Level 1: Most aggressive, quick, least accurate.
3506 Possibly useful when higher levels
3507 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3508 false negatives. However, it has many false positives.
3509 Warns for all pointer conversions between possibly incompatible types,
3510 even if never dereferenced. Runs in the frontend only.
3512 Level 2: Aggressive, quick, not too precise.
3513 May still have many false positives (not as many as level 1 though),
3514 and few false negatives (but possibly more than level 1).
3515 Unlike level 1, it only warns when an address is taken. Warns about
3516 incomplete types. Runs in the frontend only.
3518 Level 3 (default for @option{-Wstrict-aliasing}):
3519 Should have very few false positives and few false
3520 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3521 Takes care of the common pun+dereference pattern in the frontend:
3522 @code{*(int*)&some_float}.
3523 If optimization is enabled, it also runs in the backend, where it deals
3524 with multiple statement cases using flow-sensitive points-to information.
3525 Only warns when the converted pointer is dereferenced.
3526 Does not warn about incomplete types.
3528 @item -Wstrict-overflow
3529 @itemx -Wstrict-overflow=@var{n}
3530 @opindex Wstrict-overflow
3531 @opindex Wno-strict-overflow
3532 This option is only active when @option{-fstrict-overflow} is active.
3533 It warns about cases where the compiler optimizes based on the
3534 assumption that signed overflow does not occur. Note that it does not
3535 warn about all cases where the code might overflow: it only warns
3536 about cases where the compiler implements some optimization. Thus
3537 this warning depends on the optimization level.
3539 An optimization which assumes that signed overflow does not occur is
3540 perfectly safe if the values of the variables involved are such that
3541 overflow never does, in fact, occur. Therefore this warning can
3542 easily give a false positive: a warning about code which is not
3543 actually a problem. To help focus on important issues, several
3544 warning levels are defined. No warnings are issued for the use of
3545 undefined signed overflow when estimating how many iterations a loop
3546 will require, in particular when determining whether a loop will be
3550 @item -Wstrict-overflow=1
3551 Warn about cases which are both questionable and easy to avoid. For
3552 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3553 compiler will simplify this to @code{1}. This level of
3554 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3555 are not, and must be explicitly requested.
3557 @item -Wstrict-overflow=2
3558 Also warn about other cases where a comparison is simplified to a
3559 constant. For example: @code{abs (x) >= 0}. This can only be
3560 simplified when @option{-fstrict-overflow} is in effect, because
3561 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3562 zero. @option{-Wstrict-overflow} (with no level) is the same as
3563 @option{-Wstrict-overflow=2}.
3565 @item -Wstrict-overflow=3
3566 Also warn about other cases where a comparison is simplified. For
3567 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3569 @item -Wstrict-overflow=4
3570 Also warn about other simplifications not covered by the above cases.
3571 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3573 @item -Wstrict-overflow=5
3574 Also warn about cases where the compiler reduces the magnitude of a
3575 constant involved in a comparison. For example: @code{x + 2 > y} will
3576 be simplified to @code{x + 1 >= y}. This is reported only at the
3577 highest warning level because this simplification applies to many
3578 comparisons, so this warning level will give a very large number of
3582 @item -Warray-bounds
3583 @opindex Wno-array-bounds
3584 @opindex Warray-bounds
3585 This option is only active when @option{-ftree-vrp} is active
3586 (default for -O2 and above). It warns about subscripts to arrays
3587 that are always out of bounds. This warning is enabled by @option{-Wall}.
3589 @item -Wno-div-by-zero
3590 @opindex Wno-div-by-zero
3591 @opindex Wdiv-by-zero
3592 Do not warn about compile-time integer division by zero. Floating point
3593 division by zero is not warned about, as it can be a legitimate way of
3594 obtaining infinities and NaNs.
3596 @item -Wsystem-headers
3597 @opindex Wsystem-headers
3598 @opindex Wno-system-headers
3599 @cindex warnings from system headers
3600 @cindex system headers, warnings from
3601 Print warning messages for constructs found in system header files.
3602 Warnings from system headers are normally suppressed, on the assumption
3603 that they usually do not indicate real problems and would only make the
3604 compiler output harder to read. Using this command line option tells
3605 GCC to emit warnings from system headers as if they occurred in user
3606 code. However, note that using @option{-Wall} in conjunction with this
3607 option will @emph{not} warn about unknown pragmas in system
3608 headers---for that, @option{-Wunknown-pragmas} must also be used.
3611 @opindex Wfloat-equal
3612 @opindex Wno-float-equal
3613 Warn if floating point values are used in equality comparisons.
3615 The idea behind this is that sometimes it is convenient (for the
3616 programmer) to consider floating-point values as approximations to
3617 infinitely precise real numbers. If you are doing this, then you need
3618 to compute (by analyzing the code, or in some other way) the maximum or
3619 likely maximum error that the computation introduces, and allow for it
3620 when performing comparisons (and when producing output, but that's a
3621 different problem). In particular, instead of testing for equality, you
3622 would check to see whether the two values have ranges that overlap; and
3623 this is done with the relational operators, so equality comparisons are
3626 @item -Wtraditional @r{(C and Objective-C only)}
3627 @opindex Wtraditional
3628 @opindex Wno-traditional
3629 Warn about certain constructs that behave differently in traditional and
3630 ISO C@. Also warn about ISO C constructs that have no traditional C
3631 equivalent, and/or problematic constructs which should be avoided.
3635 Macro parameters that appear within string literals in the macro body.
3636 In traditional C macro replacement takes place within string literals,
3637 but does not in ISO C@.
3640 In traditional C, some preprocessor directives did not exist.
3641 Traditional preprocessors would only consider a line to be a directive
3642 if the @samp{#} appeared in column 1 on the line. Therefore
3643 @option{-Wtraditional} warns about directives that traditional C
3644 understands but would ignore because the @samp{#} does not appear as the
3645 first character on the line. It also suggests you hide directives like
3646 @samp{#pragma} not understood by traditional C by indenting them. Some
3647 traditional implementations would not recognize @samp{#elif}, so it
3648 suggests avoiding it altogether.
3651 A function-like macro that appears without arguments.
3654 The unary plus operator.
3657 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3658 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3659 constants.) Note, these suffixes appear in macros defined in the system
3660 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3661 Use of these macros in user code might normally lead to spurious
3662 warnings, however GCC's integrated preprocessor has enough context to
3663 avoid warning in these cases.
3666 A function declared external in one block and then used after the end of
3670 A @code{switch} statement has an operand of type @code{long}.
3673 A non-@code{static} function declaration follows a @code{static} one.
3674 This construct is not accepted by some traditional C compilers.
3677 The ISO type of an integer constant has a different width or
3678 signedness from its traditional type. This warning is only issued if
3679 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3680 typically represent bit patterns, are not warned about.
3683 Usage of ISO string concatenation is detected.
3686 Initialization of automatic aggregates.
3689 Identifier conflicts with labels. Traditional C lacks a separate
3690 namespace for labels.
3693 Initialization of unions. If the initializer is zero, the warning is
3694 omitted. This is done under the assumption that the zero initializer in
3695 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3696 initializer warnings and relies on default initialization to zero in the
3700 Conversions by prototypes between fixed/floating point values and vice
3701 versa. The absence of these prototypes when compiling with traditional
3702 C would cause serious problems. This is a subset of the possible
3703 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3706 Use of ISO C style function definitions. This warning intentionally is
3707 @emph{not} issued for prototype declarations or variadic functions
3708 because these ISO C features will appear in your code when using
3709 libiberty's traditional C compatibility macros, @code{PARAMS} and
3710 @code{VPARAMS}. This warning is also bypassed for nested functions
3711 because that feature is already a GCC extension and thus not relevant to
3712 traditional C compatibility.
3715 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3716 @opindex Wtraditional-conversion
3717 @opindex Wno-traditional-conversion
3718 Warn if a prototype causes a type conversion that is different from what
3719 would happen to the same argument in the absence of a prototype. This
3720 includes conversions of fixed point to floating and vice versa, and
3721 conversions changing the width or signedness of a fixed point argument
3722 except when the same as the default promotion.
3724 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3725 @opindex Wdeclaration-after-statement
3726 @opindex Wno-declaration-after-statement
3727 Warn when a declaration is found after a statement in a block. This
3728 construct, known from C++, was introduced with ISO C99 and is by default
3729 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3730 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3735 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3737 @item -Wno-endif-labels
3738 @opindex Wno-endif-labels
3739 @opindex Wendif-labels
3740 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3745 Warn whenever a local variable shadows another local variable, parameter or
3746 global variable or whenever a built-in function is shadowed.
3748 @item -Wlarger-than=@var{len}
3749 @opindex Wlarger-than=@var{len}
3750 @opindex Wlarger-than-@var{len}
3751 Warn whenever an object of larger than @var{len} bytes is defined.
3753 @item -Wframe-larger-than=@var{len}
3754 @opindex Wframe-larger-than
3755 Warn if the size of a function frame is larger than @var{len} bytes.
3756 The computation done to determine the stack frame size is approximate
3757 and not conservative.
3758 The actual requirements may be somewhat greater than @var{len}
3759 even if you do not get a warning. In addition, any space allocated
3760 via @code{alloca}, variable-length arrays, or related constructs
3761 is not included by the compiler when determining
3762 whether or not to issue a warning.
3764 @item -Wunsafe-loop-optimizations
3765 @opindex Wunsafe-loop-optimizations
3766 @opindex Wno-unsafe-loop-optimizations
3767 Warn if the loop cannot be optimized because the compiler could not
3768 assume anything on the bounds of the loop indices. With
3769 @option{-funsafe-loop-optimizations} warn if the compiler made
3772 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
3773 @opindex Wno-pedantic-ms-format
3774 @opindex Wpedantic-ms-format
3775 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
3776 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
3777 depending on the MS runtime, when you are using the options @option{-Wformat}
3778 and @option{-pedantic} without gnu-extensions.
3780 @item -Wpointer-arith
3781 @opindex Wpointer-arith
3782 @opindex Wno-pointer-arith
3783 Warn about anything that depends on the ``size of'' a function type or
3784 of @code{void}. GNU C assigns these types a size of 1, for
3785 convenience in calculations with @code{void *} pointers and pointers
3786 to functions. In C++, warn also when an arithmetic operation involves
3787 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3790 @opindex Wtype-limits
3791 @opindex Wno-type-limits
3792 Warn if a comparison is always true or always false due to the limited
3793 range of the data type, but do not warn for constant expressions. For
3794 example, warn if an unsigned variable is compared against zero with
3795 @samp{<} or @samp{>=}. This warning is also enabled by
3798 @item -Wbad-function-cast @r{(C and Objective-C only)}
3799 @opindex Wbad-function-cast
3800 @opindex Wno-bad-function-cast
3801 Warn whenever a function call is cast to a non-matching type.
3802 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3804 @item -Wc++-compat @r{(C and Objective-C only)}
3805 Warn about ISO C constructs that are outside of the common subset of
3806 ISO C and ISO C++, e.g.@: request for implicit conversion from
3807 @code{void *} to a pointer to non-@code{void} type.
3809 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3810 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3811 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3812 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3816 @opindex Wno-cast-qual
3817 Warn whenever a pointer is cast so as to remove a type qualifier from
3818 the target type. For example, warn if a @code{const char *} is cast
3819 to an ordinary @code{char *}.
3821 Also warn when making a cast which introduces a type qualifier in an
3822 unsafe way. For example, casting @code{char **} to @code{const char **}
3823 is unsafe, as in this example:
3826 /* p is char ** value. */
3827 const char **q = (const char **) p;
3828 /* Assignment of readonly string to const char * is OK. */
3830 /* Now char** pointer points to read-only memory. */
3835 @opindex Wcast-align
3836 @opindex Wno-cast-align
3837 Warn whenever a pointer is cast such that the required alignment of the
3838 target is increased. For example, warn if a @code{char *} is cast to
3839 an @code{int *} on machines where integers can only be accessed at
3840 two- or four-byte boundaries.
3842 @item -Wwrite-strings
3843 @opindex Wwrite-strings
3844 @opindex Wno-write-strings
3845 When compiling C, give string constants the type @code{const
3846 char[@var{length}]} so that copying the address of one into a
3847 non-@code{const} @code{char *} pointer will get a warning. These
3848 warnings will help you find at compile time code that can try to write
3849 into a string constant, but only if you have been very careful about
3850 using @code{const} in declarations and prototypes. Otherwise, it will
3851 just be a nuisance. This is why we did not make @option{-Wall} request
3854 When compiling C++, warn about the deprecated conversion from string
3855 literals to @code{char *}. This warning is enabled by default for C++
3860 @opindex Wno-clobbered
3861 Warn for variables that might be changed by @samp{longjmp} or
3862 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
3865 @opindex Wconversion
3866 @opindex Wno-conversion
3867 Warn for implicit conversions that may alter a value. This includes
3868 conversions between real and integer, like @code{abs (x)} when
3869 @code{x} is @code{double}; conversions between signed and unsigned,
3870 like @code{unsigned ui = -1}; and conversions to smaller types, like
3871 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
3872 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
3873 changed by the conversion like in @code{abs (2.0)}. Warnings about
3874 conversions between signed and unsigned integers can be disabled by
3875 using @option{-Wno-sign-conversion}.
3877 For C++, also warn for confusing overload resolution for user-defined
3878 conversions; and conversions that will never use a type conversion
3879 operator: conversions to @code{void}, the same type, a base class or a
3880 reference to them. Warnings about conversions between signed and
3881 unsigned integers are disabled by default in C++ unless
3882 @option{-Wsign-conversion} is explicitly enabled.
3884 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
3885 @opindex Wconversion-null
3886 @opindex Wno-conversion-null
3887 Do not warn for conversions between @code{NULL} and non-pointer
3888 types. @option{-Wconversion-null} is enabled by default.
3891 @opindex Wempty-body
3892 @opindex Wno-empty-body
3893 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
3894 while} statement. This warning is also enabled by @option{-Wextra}.
3896 @item -Wenum-compare
3897 @opindex Wenum-compare
3898 @opindex Wno-enum-compare
3899 Warn about a comparison between values of different enum types. In C++
3900 this warning is enabled by default. In C this warning is enabled by
3903 @item -Wjump-misses-init @r{(C, Objective-C only)}
3904 @opindex Wjump-misses-init
3905 @opindex Wno-jump-misses-init
3906 Warn if a @code{goto} statement or a @code{switch} statement jumps
3907 forward across the initialization of a variable, or jumps backward to a
3908 label after the variable has been initialized. This only warns about
3909 variables which are initialized when they are declared. This warning is
3910 only supported for C and Objective C; in C++ this sort of branch is an
3913 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
3914 can be disabled with the @option{-Wno-jump-misses-init} option.
3916 @item -Wsign-compare
3917 @opindex Wsign-compare
3918 @opindex Wno-sign-compare
3919 @cindex warning for comparison of signed and unsigned values
3920 @cindex comparison of signed and unsigned values, warning
3921 @cindex signed and unsigned values, comparison warning
3922 Warn when a comparison between signed and unsigned values could produce
3923 an incorrect result when the signed value is converted to unsigned.
3924 This warning is also enabled by @option{-Wextra}; to get the other warnings
3925 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3927 @item -Wsign-conversion
3928 @opindex Wsign-conversion
3929 @opindex Wno-sign-conversion
3930 Warn for implicit conversions that may change the sign of an integer
3931 value, like assigning a signed integer expression to an unsigned
3932 integer variable. An explicit cast silences the warning. In C, this
3933 option is enabled also by @option{-Wconversion}.
3937 @opindex Wno-address
3938 Warn about suspicious uses of memory addresses. These include using
3939 the address of a function in a conditional expression, such as
3940 @code{void func(void); if (func)}, and comparisons against the memory
3941 address of a string literal, such as @code{if (x == "abc")}. Such
3942 uses typically indicate a programmer error: the address of a function
3943 always evaluates to true, so their use in a conditional usually
3944 indicate that the programmer forgot the parentheses in a function
3945 call; and comparisons against string literals result in unspecified
3946 behavior and are not portable in C, so they usually indicate that the
3947 programmer intended to use @code{strcmp}. This warning is enabled by
3951 @opindex Wlogical-op
3952 @opindex Wno-logical-op
3953 Warn about suspicious uses of logical operators in expressions.
3954 This includes using logical operators in contexts where a
3955 bit-wise operator is likely to be expected.
3957 @item -Waggregate-return
3958 @opindex Waggregate-return
3959 @opindex Wno-aggregate-return
3960 Warn if any functions that return structures or unions are defined or
3961 called. (In languages where you can return an array, this also elicits
3964 @item -Wno-attributes
3965 @opindex Wno-attributes
3966 @opindex Wattributes
3967 Do not warn if an unexpected @code{__attribute__} is used, such as
3968 unrecognized attributes, function attributes applied to variables,
3969 etc. This will not stop errors for incorrect use of supported
3972 @item -Wno-builtin-macro-redefined
3973 @opindex Wno-builtin-macro-redefined
3974 @opindex Wbuiltin-macro-redefined
3975 Do not warn if certain built-in macros are redefined. This suppresses
3976 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
3977 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
3979 @item -Wstrict-prototypes @r{(C and Objective-C only)}
3980 @opindex Wstrict-prototypes
3981 @opindex Wno-strict-prototypes
3982 Warn if a function is declared or defined without specifying the
3983 argument types. (An old-style function definition is permitted without
3984 a warning if preceded by a declaration which specifies the argument
3987 @item -Wold-style-declaration @r{(C and Objective-C only)}
3988 @opindex Wold-style-declaration
3989 @opindex Wno-old-style-declaration
3990 Warn for obsolescent usages, according to the C Standard, in a
3991 declaration. For example, warn if storage-class specifiers like
3992 @code{static} are not the first things in a declaration. This warning
3993 is also enabled by @option{-Wextra}.
3995 @item -Wold-style-definition @r{(C and Objective-C only)}
3996 @opindex Wold-style-definition
3997 @opindex Wno-old-style-definition
3998 Warn if an old-style function definition is used. A warning is given
3999 even if there is a previous prototype.
4001 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
4002 @opindex Wmissing-parameter-type
4003 @opindex Wno-missing-parameter-type
4004 A function parameter is declared without a type specifier in K&R-style
4011 This warning is also enabled by @option{-Wextra}.
4013 @item -Wmissing-prototypes @r{(C and Objective-C only)}
4014 @opindex Wmissing-prototypes
4015 @opindex Wno-missing-prototypes
4016 Warn if a global function is defined without a previous prototype
4017 declaration. This warning is issued even if the definition itself
4018 provides a prototype. The aim is to detect global functions that fail
4019 to be declared in header files.
4021 @item -Wmissing-declarations
4022 @opindex Wmissing-declarations
4023 @opindex Wno-missing-declarations
4024 Warn if a global function is defined without a previous declaration.
4025 Do so even if the definition itself provides a prototype.
4026 Use this option to detect global functions that are not declared in
4027 header files. In C++, no warnings are issued for function templates,
4028 or for inline functions, or for functions in anonymous namespaces.
4030 @item -Wmissing-field-initializers
4031 @opindex Wmissing-field-initializers
4032 @opindex Wno-missing-field-initializers
4036 Warn if a structure's initializer has some fields missing. For
4037 example, the following code would cause such a warning, because
4038 @code{x.h} is implicitly zero:
4041 struct s @{ int f, g, h; @};
4042 struct s x = @{ 3, 4 @};
4045 This option does not warn about designated initializers, so the following
4046 modification would not trigger a warning:
4049 struct s @{ int f, g, h; @};
4050 struct s x = @{ .f = 3, .g = 4 @};
4053 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
4054 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
4056 @item -Wmissing-noreturn
4057 @opindex Wmissing-noreturn
4058 @opindex Wno-missing-noreturn
4059 Warn about functions which might be candidates for attribute @code{noreturn}.
4060 Note these are only possible candidates, not absolute ones. Care should
4061 be taken to manually verify functions actually do not ever return before
4062 adding the @code{noreturn} attribute, otherwise subtle code generation
4063 bugs could be introduced. You will not get a warning for @code{main} in
4064 hosted C environments.
4066 @item -Wmissing-format-attribute
4067 @opindex Wmissing-format-attribute
4068 @opindex Wno-missing-format-attribute
4071 Warn about function pointers which might be candidates for @code{format}
4072 attributes. Note these are only possible candidates, not absolute ones.
4073 GCC will guess that function pointers with @code{format} attributes that
4074 are used in assignment, initialization, parameter passing or return
4075 statements should have a corresponding @code{format} attribute in the
4076 resulting type. I.e.@: the left-hand side of the assignment or
4077 initialization, the type of the parameter variable, or the return type
4078 of the containing function respectively should also have a @code{format}
4079 attribute to avoid the warning.
4081 GCC will also warn about function definitions which might be
4082 candidates for @code{format} attributes. Again, these are only
4083 possible candidates. GCC will guess that @code{format} attributes
4084 might be appropriate for any function that calls a function like
4085 @code{vprintf} or @code{vscanf}, but this might not always be the
4086 case, and some functions for which @code{format} attributes are
4087 appropriate may not be detected.
4089 @item -Wno-multichar
4090 @opindex Wno-multichar
4092 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4093 Usually they indicate a typo in the user's code, as they have
4094 implementation-defined values, and should not be used in portable code.
4096 @item -Wnormalized=<none|id|nfc|nfkc>
4097 @opindex Wnormalized=
4100 @cindex character set, input normalization
4101 In ISO C and ISO C++, two identifiers are different if they are
4102 different sequences of characters. However, sometimes when characters
4103 outside the basic ASCII character set are used, you can have two
4104 different character sequences that look the same. To avoid confusion,
4105 the ISO 10646 standard sets out some @dfn{normalization rules} which
4106 when applied ensure that two sequences that look the same are turned into
4107 the same sequence. GCC can warn you if you are using identifiers which
4108 have not been normalized; this option controls that warning.
4110 There are four levels of warning that GCC supports. The default is
4111 @option{-Wnormalized=nfc}, which warns about any identifier which is
4112 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4113 recommended form for most uses.
4115 Unfortunately, there are some characters which ISO C and ISO C++ allow
4116 in identifiers that when turned into NFC aren't allowable as
4117 identifiers. That is, there's no way to use these symbols in portable
4118 ISO C or C++ and have all your identifiers in NFC@.
4119 @option{-Wnormalized=id} suppresses the warning for these characters.
4120 It is hoped that future versions of the standards involved will correct
4121 this, which is why this option is not the default.
4123 You can switch the warning off for all characters by writing
4124 @option{-Wnormalized=none}. You would only want to do this if you
4125 were using some other normalization scheme (like ``D''), because
4126 otherwise you can easily create bugs that are literally impossible to see.
4128 Some characters in ISO 10646 have distinct meanings but look identical
4129 in some fonts or display methodologies, especially once formatting has
4130 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4131 LETTER N'', will display just like a regular @code{n} which has been
4132 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4133 normalization scheme to convert all these into a standard form as
4134 well, and GCC will warn if your code is not in NFKC if you use
4135 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4136 about every identifier that contains the letter O because it might be
4137 confused with the digit 0, and so is not the default, but may be
4138 useful as a local coding convention if the programming environment is
4139 unable to be fixed to display these characters distinctly.
4141 @item -Wno-deprecated
4142 @opindex Wno-deprecated
4143 @opindex Wdeprecated
4144 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4146 @item -Wno-deprecated-declarations
4147 @opindex Wno-deprecated-declarations
4148 @opindex Wdeprecated-declarations
4149 Do not warn about uses of functions (@pxref{Function Attributes}),
4150 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4151 Attributes}) marked as deprecated by using the @code{deprecated}
4155 @opindex Wno-overflow
4157 Do not warn about compile-time overflow in constant expressions.
4159 @item -Woverride-init @r{(C and Objective-C only)}
4160 @opindex Woverride-init
4161 @opindex Wno-override-init
4165 Warn if an initialized field without side effects is overridden when
4166 using designated initializers (@pxref{Designated Inits, , Designated
4169 This warning is included in @option{-Wextra}. To get other
4170 @option{-Wextra} warnings without this one, use @samp{-Wextra
4171 -Wno-override-init}.
4176 Warn if a structure is given the packed attribute, but the packed
4177 attribute has no effect on the layout or size of the structure.
4178 Such structures may be mis-aligned for little benefit. For
4179 instance, in this code, the variable @code{f.x} in @code{struct bar}
4180 will be misaligned even though @code{struct bar} does not itself
4181 have the packed attribute:
4188 @} __attribute__((packed));
4196 @item -Wpacked-bitfield-compat
4197 @opindex Wpacked-bitfield-compat
4198 @opindex Wno-packed-bitfield-compat
4199 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4200 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4201 the change can lead to differences in the structure layout. GCC
4202 informs you when the offset of such a field has changed in GCC 4.4.
4203 For example there is no longer a 4-bit padding between field @code{a}
4204 and @code{b} in this structure:
4211 @} __attribute__ ((packed));
4214 This warning is enabled by default. Use
4215 @option{-Wno-packed-bitfield-compat} to disable this warning.
4220 Warn if padding is included in a structure, either to align an element
4221 of the structure or to align the whole structure. Sometimes when this
4222 happens it is possible to rearrange the fields of the structure to
4223 reduce the padding and so make the structure smaller.
4225 @item -Wredundant-decls
4226 @opindex Wredundant-decls
4227 @opindex Wno-redundant-decls
4228 Warn if anything is declared more than once in the same scope, even in
4229 cases where multiple declaration is valid and changes nothing.
4231 @item -Wnested-externs @r{(C and Objective-C only)}
4232 @opindex Wnested-externs
4233 @opindex Wno-nested-externs
4234 Warn if an @code{extern} declaration is encountered within a function.
4239 Warn if a function can not be inlined and it was declared as inline.
4240 Even with this option, the compiler will not warn about failures to
4241 inline functions declared in system headers.
4243 The compiler uses a variety of heuristics to determine whether or not
4244 to inline a function. For example, the compiler takes into account
4245 the size of the function being inlined and the amount of inlining
4246 that has already been done in the current function. Therefore,
4247 seemingly insignificant changes in the source program can cause the
4248 warnings produced by @option{-Winline} to appear or disappear.
4250 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4251 @opindex Wno-invalid-offsetof
4252 @opindex Winvalid-offsetof
4253 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4254 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4255 to a non-POD type is undefined. In existing C++ implementations,
4256 however, @samp{offsetof} typically gives meaningful results even when
4257 applied to certain kinds of non-POD types. (Such as a simple
4258 @samp{struct} that fails to be a POD type only by virtue of having a
4259 constructor.) This flag is for users who are aware that they are
4260 writing nonportable code and who have deliberately chosen to ignore the
4263 The restrictions on @samp{offsetof} may be relaxed in a future version
4264 of the C++ standard.
4266 @item -Wno-int-to-pointer-cast @r{(C and Objective-C only)}
4267 @opindex Wno-int-to-pointer-cast
4268 @opindex Wint-to-pointer-cast
4269 Suppress warnings from casts to pointer type of an integer of a
4272 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4273 @opindex Wno-pointer-to-int-cast
4274 @opindex Wpointer-to-int-cast
4275 Suppress warnings from casts from a pointer to an integer type of a
4279 @opindex Winvalid-pch
4280 @opindex Wno-invalid-pch
4281 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4282 the search path but can't be used.
4286 @opindex Wno-long-long
4287 Warn if @samp{long long} type is used. This is enabled by either
4288 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4289 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4291 @item -Wvariadic-macros
4292 @opindex Wvariadic-macros
4293 @opindex Wno-variadic-macros
4294 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4295 alternate syntax when in pedantic ISO C99 mode. This is default.
4296 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4301 Warn if variable length array is used in the code.
4302 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4303 the variable length array.
4305 @item -Wvolatile-register-var
4306 @opindex Wvolatile-register-var
4307 @opindex Wno-volatile-register-var
4308 Warn if a register variable is declared volatile. The volatile
4309 modifier does not inhibit all optimizations that may eliminate reads
4310 and/or writes to register variables. This warning is enabled by
4313 @item -Wdisabled-optimization
4314 @opindex Wdisabled-optimization
4315 @opindex Wno-disabled-optimization
4316 Warn if a requested optimization pass is disabled. This warning does
4317 not generally indicate that there is anything wrong with your code; it
4318 merely indicates that GCC's optimizers were unable to handle the code
4319 effectively. Often, the problem is that your code is too big or too
4320 complex; GCC will refuse to optimize programs when the optimization
4321 itself is likely to take inordinate amounts of time.
4323 @item -Wpointer-sign @r{(C and Objective-C only)}
4324 @opindex Wpointer-sign
4325 @opindex Wno-pointer-sign
4326 Warn for pointer argument passing or assignment with different signedness.
4327 This option is only supported for C and Objective-C@. It is implied by
4328 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4329 @option{-Wno-pointer-sign}.
4331 @item -Wstack-protector
4332 @opindex Wstack-protector
4333 @opindex Wno-stack-protector
4334 This option is only active when @option{-fstack-protector} is active. It
4335 warns about functions that will not be protected against stack smashing.
4338 @opindex Wno-mudflap
4339 Suppress warnings about constructs that cannot be instrumented by
4342 @item -Woverlength-strings
4343 @opindex Woverlength-strings
4344 @opindex Wno-overlength-strings
4345 Warn about string constants which are longer than the ``minimum
4346 maximum'' length specified in the C standard. Modern compilers
4347 generally allow string constants which are much longer than the
4348 standard's minimum limit, but very portable programs should avoid
4349 using longer strings.
4351 The limit applies @emph{after} string constant concatenation, and does
4352 not count the trailing NUL@. In C90, the limit was 509 characters; in
4353 C99, it was raised to 4095. C++98 does not specify a normative
4354 minimum maximum, so we do not diagnose overlength strings in C++@.
4356 This option is implied by @option{-pedantic}, and can be disabled with
4357 @option{-Wno-overlength-strings}.
4359 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
4360 @opindex Wunsuffixed-float-constants
4362 GCC will issue a warning for any floating constant that does not have
4363 a suffix. When used together with @option{-Wsystem-headers} it will
4364 warn about such constants in system header files. This can be useful
4365 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4366 from the decimal floating-point extension to C99.
4369 @node Debugging Options
4370 @section Options for Debugging Your Program or GCC
4371 @cindex options, debugging
4372 @cindex debugging information options
4374 GCC has various special options that are used for debugging
4375 either your program or GCC:
4380 Produce debugging information in the operating system's native format
4381 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4384 On most systems that use stabs format, @option{-g} enables use of extra
4385 debugging information that only GDB can use; this extra information
4386 makes debugging work better in GDB but will probably make other debuggers
4388 refuse to read the program. If you want to control for certain whether
4389 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4390 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4392 GCC allows you to use @option{-g} with
4393 @option{-O}. The shortcuts taken by optimized code may occasionally
4394 produce surprising results: some variables you declared may not exist
4395 at all; flow of control may briefly move where you did not expect it;
4396 some statements may not be executed because they compute constant
4397 results or their values were already at hand; some statements may
4398 execute in different places because they were moved out of loops.
4400 Nevertheless it proves possible to debug optimized output. This makes
4401 it reasonable to use the optimizer for programs that might have bugs.
4403 The following options are useful when GCC is generated with the
4404 capability for more than one debugging format.
4408 Produce debugging information for use by GDB@. This means to use the
4409 most expressive format available (DWARF 2, stabs, or the native format
4410 if neither of those are supported), including GDB extensions if at all
4415 Produce debugging information in stabs format (if that is supported),
4416 without GDB extensions. This is the format used by DBX on most BSD
4417 systems. On MIPS, Alpha and System V Release 4 systems this option
4418 produces stabs debugging output which is not understood by DBX or SDB@.
4419 On System V Release 4 systems this option requires the GNU assembler.
4421 @item -feliminate-unused-debug-symbols
4422 @opindex feliminate-unused-debug-symbols
4423 Produce debugging information in stabs format (if that is supported),
4424 for only symbols that are actually used.
4426 @item -femit-class-debug-always
4427 Instead of emitting debugging information for a C++ class in only one
4428 object file, emit it in all object files using the class. This option
4429 should be used only with debuggers that are unable to handle the way GCC
4430 normally emits debugging information for classes because using this
4431 option will increase the size of debugging information by as much as a
4436 Produce debugging information in stabs format (if that is supported),
4437 using GNU extensions understood only by the GNU debugger (GDB)@. The
4438 use of these extensions is likely to make other debuggers crash or
4439 refuse to read the program.
4443 Produce debugging information in COFF format (if that is supported).
4444 This is the format used by SDB on most System V systems prior to
4449 Produce debugging information in XCOFF format (if that is supported).
4450 This is the format used by the DBX debugger on IBM RS/6000 systems.
4454 Produce debugging information in XCOFF format (if that is supported),
4455 using GNU extensions understood only by the GNU debugger (GDB)@. The
4456 use of these extensions is likely to make other debuggers crash or
4457 refuse to read the program, and may cause assemblers other than the GNU
4458 assembler (GAS) to fail with an error.
4460 @item -gdwarf-@var{version}
4461 @opindex gdwarf-@var{version}
4462 Produce debugging information in DWARF format (if that is
4463 supported). This is the format used by DBX on IRIX 6. The value
4464 of @var{version} may be either 2, 3 or 4; the default version is 2.
4466 Note that with DWARF version 2 some ports require, and will always
4467 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4469 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
4470 for maximum benefit.
4472 @item -gstrict-dwarf
4473 @opindex gstrict-dwarf
4474 Disallow using extensions of later DWARF standard version than selected
4475 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
4476 DWARF extensions from later standard versions is allowed.
4478 @item -gno-strict-dwarf
4479 @opindex gno-strict-dwarf
4480 Allow using extensions of later DWARF standard version than selected with
4481 @option{-gdwarf-@var{version}}.
4485 Produce debugging information in VMS debug format (if that is
4486 supported). This is the format used by DEBUG on VMS systems.
4489 @itemx -ggdb@var{level}
4490 @itemx -gstabs@var{level}
4491 @itemx -gcoff@var{level}
4492 @itemx -gxcoff@var{level}
4493 @itemx -gvms@var{level}
4494 Request debugging information and also use @var{level} to specify how
4495 much information. The default level is 2.
4497 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4500 Level 1 produces minimal information, enough for making backtraces in
4501 parts of the program that you don't plan to debug. This includes
4502 descriptions of functions and external variables, but no information
4503 about local variables and no line numbers.
4505 Level 3 includes extra information, such as all the macro definitions
4506 present in the program. Some debuggers support macro expansion when
4507 you use @option{-g3}.
4509 @option{-gdwarf-2} does not accept a concatenated debug level, because
4510 GCC used to support an option @option{-gdwarf} that meant to generate
4511 debug information in version 1 of the DWARF format (which is very
4512 different from version 2), and it would have been too confusing. That
4513 debug format is long obsolete, but the option cannot be changed now.
4514 Instead use an additional @option{-g@var{level}} option to change the
4515 debug level for DWARF.
4519 Turn off generation of debug info, if leaving out this option would have
4520 generated it, or turn it on at level 2 otherwise. The position of this
4521 argument in the command line does not matter, it takes effect after all
4522 other options are processed, and it does so only once, no matter how
4523 many times it is given. This is mainly intended to be used with
4524 @option{-fcompare-debug}.
4526 @item -fdump-final-insns@r{[}=@var{file}@r{]}
4527 @opindex fdump-final-insns
4528 Dump the final internal representation (RTL) to @var{file}. If the
4529 optional argument is omitted (or if @var{file} is @code{.}), the name
4530 of the dump file will be determined by appending @code{.gkd} to the
4531 compilation output file name.
4533 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4534 @opindex fcompare-debug
4535 @opindex fno-compare-debug
4536 If no error occurs during compilation, run the compiler a second time,
4537 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4538 passed to the second compilation. Dump the final internal
4539 representation in both compilations, and print an error if they differ.
4541 If the equal sign is omitted, the default @option{-gtoggle} is used.
4543 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4544 and nonzero, implicitly enables @option{-fcompare-debug}. If
4545 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4546 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4549 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4550 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4551 of the final representation and the second compilation, preventing even
4552 @env{GCC_COMPARE_DEBUG} from taking effect.
4554 To verify full coverage during @option{-fcompare-debug} testing, set
4555 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4556 which GCC will reject as an invalid option in any actual compilation
4557 (rather than preprocessing, assembly or linking). To get just a
4558 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4559 not overridden} will do.
4561 @item -fcompare-debug-second
4562 @opindex fcompare-debug-second
4563 This option is implicitly passed to the compiler for the second
4564 compilation requested by @option{-fcompare-debug}, along with options to
4565 silence warnings, and omitting other options that would cause
4566 side-effect compiler outputs to files or to the standard output. Dump
4567 files and preserved temporary files are renamed so as to contain the
4568 @code{.gk} additional extension during the second compilation, to avoid
4569 overwriting those generated by the first.
4571 When this option is passed to the compiler driver, it causes the
4572 @emph{first} compilation to be skipped, which makes it useful for little
4573 other than debugging the compiler proper.
4575 @item -feliminate-dwarf2-dups
4576 @opindex feliminate-dwarf2-dups
4577 Compress DWARF2 debugging information by eliminating duplicated
4578 information about each symbol. This option only makes sense when
4579 generating DWARF2 debugging information with @option{-gdwarf-2}.
4581 @item -femit-struct-debug-baseonly
4582 Emit debug information for struct-like types
4583 only when the base name of the compilation source file
4584 matches the base name of file in which the struct was defined.
4586 This option substantially reduces the size of debugging information,
4587 but at significant potential loss in type information to the debugger.
4588 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4589 See @option{-femit-struct-debug-detailed} for more detailed control.
4591 This option works only with DWARF 2.
4593 @item -femit-struct-debug-reduced
4594 Emit debug information for struct-like types
4595 only when the base name of the compilation source file
4596 matches the base name of file in which the type was defined,
4597 unless the struct is a template or defined in a system header.
4599 This option significantly reduces the size of debugging information,
4600 with some potential loss in type information to the debugger.
4601 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4602 See @option{-femit-struct-debug-detailed} for more detailed control.
4604 This option works only with DWARF 2.
4606 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4607 Specify the struct-like types
4608 for which the compiler will generate debug information.
4609 The intent is to reduce duplicate struct debug information
4610 between different object files within the same program.
4612 This option is a detailed version of
4613 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4614 which will serve for most needs.
4616 A specification has the syntax
4617 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4619 The optional first word limits the specification to
4620 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4621 A struct type is used directly when it is the type of a variable, member.
4622 Indirect uses arise through pointers to structs.
4623 That is, when use of an incomplete struct would be legal, the use is indirect.
4625 @samp{struct one direct; struct two * indirect;}.
4627 The optional second word limits the specification to
4628 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4629 Generic structs are a bit complicated to explain.
4630 For C++, these are non-explicit specializations of template classes,
4631 or non-template classes within the above.
4632 Other programming languages have generics,
4633 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4635 The third word specifies the source files for those
4636 structs for which the compiler will emit debug information.
4637 The values @samp{none} and @samp{any} have the normal meaning.
4638 The value @samp{base} means that
4639 the base of name of the file in which the type declaration appears
4640 must match the base of the name of the main compilation file.
4641 In practice, this means that
4642 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4643 but types declared in other header will not.
4644 The value @samp{sys} means those types satisfying @samp{base}
4645 or declared in system or compiler headers.
4647 You may need to experiment to determine the best settings for your application.
4649 The default is @samp{-femit-struct-debug-detailed=all}.
4651 This option works only with DWARF 2.
4653 @item -fenable-icf-debug
4654 @opindex fenable-icf-debug
4655 Generate additional debug information to support identical code folding (ICF).
4656 This option only works with DWARF version 2 or higher.
4658 @item -fno-merge-debug-strings
4659 @opindex fmerge-debug-strings
4660 @opindex fno-merge-debug-strings
4661 Direct the linker to not merge together strings in the debugging
4662 information which are identical in different object files. Merging is
4663 not supported by all assemblers or linkers. Merging decreases the size
4664 of the debug information in the output file at the cost of increasing
4665 link processing time. Merging is enabled by default.
4667 @item -fdebug-prefix-map=@var{old}=@var{new}
4668 @opindex fdebug-prefix-map
4669 When compiling files in directory @file{@var{old}}, record debugging
4670 information describing them as in @file{@var{new}} instead.
4672 @item -fno-dwarf2-cfi-asm
4673 @opindex fdwarf2-cfi-asm
4674 @opindex fno-dwarf2-cfi-asm
4675 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4676 instead of using GAS @code{.cfi_*} directives.
4678 @cindex @command{prof}
4681 Generate extra code to write profile information suitable for the
4682 analysis program @command{prof}. You must use this option when compiling
4683 the source files you want data about, and you must also use it when
4686 @cindex @command{gprof}
4689 Generate extra code to write profile information suitable for the
4690 analysis program @command{gprof}. You must use this option when compiling
4691 the source files you want data about, and you must also use it when
4696 Makes the compiler print out each function name as it is compiled, and
4697 print some statistics about each pass when it finishes.
4700 @opindex ftime-report
4701 Makes the compiler print some statistics about the time consumed by each
4702 pass when it finishes.
4705 @opindex fmem-report
4706 Makes the compiler print some statistics about permanent memory
4707 allocation when it finishes.
4709 @item -fpre-ipa-mem-report
4710 @opindex fpre-ipa-mem-report
4711 @item -fpost-ipa-mem-report
4712 @opindex fpost-ipa-mem-report
4713 Makes the compiler print some statistics about permanent memory
4714 allocation before or after interprocedural optimization.
4716 @item -fprofile-arcs
4717 @opindex fprofile-arcs
4718 Add code so that program flow @dfn{arcs} are instrumented. During
4719 execution the program records how many times each branch and call is
4720 executed and how many times it is taken or returns. When the compiled
4721 program exits it saves this data to a file called
4722 @file{@var{auxname}.gcda} for each source file. The data may be used for
4723 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4724 test coverage analysis (@option{-ftest-coverage}). Each object file's
4725 @var{auxname} is generated from the name of the output file, if
4726 explicitly specified and it is not the final executable, otherwise it is
4727 the basename of the source file. In both cases any suffix is removed
4728 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4729 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4730 @xref{Cross-profiling}.
4732 @cindex @command{gcov}
4736 This option is used to compile and link code instrumented for coverage
4737 analysis. The option is a synonym for @option{-fprofile-arcs}
4738 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4739 linking). See the documentation for those options for more details.
4744 Compile the source files with @option{-fprofile-arcs} plus optimization
4745 and code generation options. For test coverage analysis, use the
4746 additional @option{-ftest-coverage} option. You do not need to profile
4747 every source file in a program.
4750 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4751 (the latter implies the former).
4754 Run the program on a representative workload to generate the arc profile
4755 information. This may be repeated any number of times. You can run
4756 concurrent instances of your program, and provided that the file system
4757 supports locking, the data files will be correctly updated. Also
4758 @code{fork} calls are detected and correctly handled (double counting
4762 For profile-directed optimizations, compile the source files again with
4763 the same optimization and code generation options plus
4764 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4765 Control Optimization}).
4768 For test coverage analysis, use @command{gcov} to produce human readable
4769 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4770 @command{gcov} documentation for further information.
4774 With @option{-fprofile-arcs}, for each function of your program GCC
4775 creates a program flow graph, then finds a spanning tree for the graph.
4776 Only arcs that are not on the spanning tree have to be instrumented: the
4777 compiler adds code to count the number of times that these arcs are
4778 executed. When an arc is the only exit or only entrance to a block, the
4779 instrumentation code can be added to the block; otherwise, a new basic
4780 block must be created to hold the instrumentation code.
4783 @item -ftest-coverage
4784 @opindex ftest-coverage
4785 Produce a notes file that the @command{gcov} code-coverage utility
4786 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4787 show program coverage. Each source file's note file is called
4788 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4789 above for a description of @var{auxname} and instructions on how to
4790 generate test coverage data. Coverage data will match the source files
4791 more closely, if you do not optimize.
4793 @item -fdbg-cnt-list
4794 @opindex fdbg-cnt-list
4795 Print the name and the counter upperbound for all debug counters.
4797 @item -fdbg-cnt=@var{counter-value-list}
4799 Set the internal debug counter upperbound. @var{counter-value-list}
4800 is a comma-separated list of @var{name}:@var{value} pairs
4801 which sets the upperbound of each debug counter @var{name} to @var{value}.
4802 All debug counters have the initial upperbound of @var{UINT_MAX},
4803 thus dbg_cnt() returns true always unless the upperbound is set by this option.
4804 e.g. With -fdbg-cnt=dce:10,tail_call:0
4805 dbg_cnt(dce) will return true only for first 10 invocations
4806 and dbg_cnt(tail_call) will return false always.
4808 @item -d@var{letters}
4809 @itemx -fdump-rtl-@var{pass}
4811 Says to make debugging dumps during compilation at times specified by
4812 @var{letters}. This is used for debugging the RTL-based passes of the
4813 compiler. The file names for most of the dumps are made by appending
4814 a pass number and a word to the @var{dumpname}, and the files are
4815 created in the directory of the output file. @var{dumpname} is
4816 generated from the name of the output file, if explicitly specified
4817 and it is not an executable, otherwise it is the basename of the
4818 source file. These switches may have different effects when
4819 @option{-E} is used for preprocessing.
4821 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
4822 @option{-d} option @var{letters}. Here are the possible
4823 letters for use in @var{pass} and @var{letters}, and their meanings:
4827 @item -fdump-rtl-alignments
4828 @opindex fdump-rtl-alignments
4829 Dump after branch alignments have been computed.
4831 @item -fdump-rtl-asmcons
4832 @opindex fdump-rtl-asmcons
4833 Dump after fixing rtl statements that have unsatisfied in/out constraints.
4835 @item -fdump-rtl-auto_inc_dec
4836 @opindex fdump-rtl-auto_inc_dec
4837 Dump after auto-inc-dec discovery. This pass is only run on
4838 architectures that have auto inc or auto dec instructions.
4840 @item -fdump-rtl-barriers
4841 @opindex fdump-rtl-barriers
4842 Dump after cleaning up the barrier instructions.
4844 @item -fdump-rtl-bbpart
4845 @opindex fdump-rtl-bbpart
4846 Dump after partitioning hot and cold basic blocks.
4848 @item -fdump-rtl-bbro
4849 @opindex fdump-rtl-bbro
4850 Dump after block reordering.
4852 @item -fdump-rtl-btl1
4853 @itemx -fdump-rtl-btl2
4854 @opindex fdump-rtl-btl2
4855 @opindex fdump-rtl-btl2
4856 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
4857 after the two branch
4858 target load optimization passes.
4860 @item -fdump-rtl-bypass
4861 @opindex fdump-rtl-bypass
4862 Dump after jump bypassing and control flow optimizations.
4864 @item -fdump-rtl-combine
4865 @opindex fdump-rtl-combine
4866 Dump after the RTL instruction combination pass.
4868 @item -fdump-rtl-compgotos
4869 @opindex fdump-rtl-compgotos
4870 Dump after duplicating the computed gotos.
4872 @item -fdump-rtl-ce1
4873 @itemx -fdump-rtl-ce2
4874 @itemx -fdump-rtl-ce3
4875 @opindex fdump-rtl-ce1
4876 @opindex fdump-rtl-ce2
4877 @opindex fdump-rtl-ce3
4878 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
4879 @option{-fdump-rtl-ce3} enable dumping after the three
4880 if conversion passes.
4882 @itemx -fdump-rtl-cprop_hardreg
4883 @opindex fdump-rtl-cprop_hardreg
4884 Dump after hard register copy propagation.
4886 @itemx -fdump-rtl-csa
4887 @opindex fdump-rtl-csa
4888 Dump after combining stack adjustments.
4890 @item -fdump-rtl-cse1
4891 @itemx -fdump-rtl-cse2
4892 @opindex fdump-rtl-cse1
4893 @opindex fdump-rtl-cse2
4894 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
4895 the two common sub-expression elimination passes.
4897 @itemx -fdump-rtl-dce
4898 @opindex fdump-rtl-dce
4899 Dump after the standalone dead code elimination passes.
4901 @itemx -fdump-rtl-dbr
4902 @opindex fdump-rtl-dbr
4903 Dump after delayed branch scheduling.
4905 @item -fdump-rtl-dce1
4906 @itemx -fdump-rtl-dce2
4907 @opindex fdump-rtl-dce1
4908 @opindex fdump-rtl-dce2
4909 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
4910 the two dead store elimination passes.
4913 @opindex fdump-rtl-eh
4914 Dump after finalization of EH handling code.
4916 @item -fdump-rtl-eh_ranges
4917 @opindex fdump-rtl-eh_ranges
4918 Dump after conversion of EH handling range regions.
4920 @item -fdump-rtl-expand
4921 @opindex fdump-rtl-expand
4922 Dump after RTL generation.
4924 @item -fdump-rtl-fwprop1
4925 @itemx -fdump-rtl-fwprop2
4926 @opindex fdump-rtl-fwprop1
4927 @opindex fdump-rtl-fwprop2
4928 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
4929 dumping after the two forward propagation passes.
4931 @item -fdump-rtl-gcse1
4932 @itemx -fdump-rtl-gcse2
4933 @opindex fdump-rtl-gcse1
4934 @opindex fdump-rtl-gcse2
4935 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
4936 after global common subexpression elimination.
4938 @item -fdump-rtl-init-regs
4939 @opindex fdump-rtl-init-regs
4940 Dump after the initialization of the registers.
4942 @item -fdump-rtl-initvals
4943 @opindex fdump-rtl-initvals
4944 Dump after the computation of the initial value sets.
4946 @itemx -fdump-rtl-into_cfglayout
4947 @opindex fdump-rtl-into_cfglayout
4948 Dump after converting to cfglayout mode.
4950 @item -fdump-rtl-ira
4951 @opindex fdump-rtl-ira
4952 Dump after iterated register allocation.
4954 @item -fdump-rtl-jump
4955 @opindex fdump-rtl-jump
4956 Dump after the second jump optimization.
4958 @item -fdump-rtl-loop2
4959 @opindex fdump-rtl-loop2
4960 @option{-fdump-rtl-loop2} enables dumping after the rtl
4961 loop optimization passes.
4963 @item -fdump-rtl-mach
4964 @opindex fdump-rtl-mach
4965 Dump after performing the machine dependent reorganization pass, if that
4968 @item -fdump-rtl-mode_sw
4969 @opindex fdump-rtl-mode_sw
4970 Dump after removing redundant mode switches.
4972 @item -fdump-rtl-rnreg
4973 @opindex fdump-rtl-rnreg
4974 Dump after register renumbering.
4976 @itemx -fdump-rtl-outof_cfglayout
4977 @opindex fdump-rtl-outof_cfglayout
4978 Dump after converting from cfglayout mode.
4980 @item -fdump-rtl-peephole2
4981 @opindex fdump-rtl-peephole2
4982 Dump after the peephole pass.
4984 @item -fdump-rtl-postreload
4985 @opindex fdump-rtl-postreload
4986 Dump after post-reload optimizations.
4988 @itemx -fdump-rtl-pro_and_epilogue
4989 @opindex fdump-rtl-pro_and_epilogue
4990 Dump after generating the function pro and epilogues.
4992 @item -fdump-rtl-regmove
4993 @opindex fdump-rtl-regmove
4994 Dump after the register move pass.
4996 @item -fdump-rtl-sched1
4997 @itemx -fdump-rtl-sched2
4998 @opindex fdump-rtl-sched1
4999 @opindex fdump-rtl-sched2
5000 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
5001 after the basic block scheduling passes.
5003 @item -fdump-rtl-see
5004 @opindex fdump-rtl-see
5005 Dump after sign extension elimination.
5007 @item -fdump-rtl-seqabstr
5008 @opindex fdump-rtl-seqabstr
5009 Dump after common sequence discovery.
5011 @item -fdump-rtl-shorten
5012 @opindex fdump-rtl-shorten
5013 Dump after shortening branches.
5015 @item -fdump-rtl-sibling
5016 @opindex fdump-rtl-sibling
5017 Dump after sibling call optimizations.
5019 @item -fdump-rtl-split1
5020 @itemx -fdump-rtl-split2
5021 @itemx -fdump-rtl-split3
5022 @itemx -fdump-rtl-split4
5023 @itemx -fdump-rtl-split5
5024 @opindex fdump-rtl-split1
5025 @opindex fdump-rtl-split2
5026 @opindex fdump-rtl-split3
5027 @opindex fdump-rtl-split4
5028 @opindex fdump-rtl-split5
5029 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
5030 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
5031 @option{-fdump-rtl-split5} enable dumping after five rounds of
5032 instruction splitting.
5034 @item -fdump-rtl-sms
5035 @opindex fdump-rtl-sms
5036 Dump after modulo scheduling. This pass is only run on some
5039 @item -fdump-rtl-stack
5040 @opindex fdump-rtl-stack
5041 Dump after conversion from GCC's "flat register file" registers to the
5042 x87's stack-like registers. This pass is only run on x86 variants.
5044 @item -fdump-rtl-subreg1
5045 @itemx -fdump-rtl-subreg2
5046 @opindex fdump-rtl-subreg1
5047 @opindex fdump-rtl-subreg2
5048 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
5049 the two subreg expansion passes.
5051 @item -fdump-rtl-unshare
5052 @opindex fdump-rtl-unshare
5053 Dump after all rtl has been unshared.
5055 @item -fdump-rtl-vartrack
5056 @opindex fdump-rtl-vartrack
5057 Dump after variable tracking.
5059 @item -fdump-rtl-vregs
5060 @opindex fdump-rtl-vregs
5061 Dump after converting virtual registers to hard registers.
5063 @item -fdump-rtl-web
5064 @opindex fdump-rtl-web
5065 Dump after live range splitting.
5067 @item -fdump-rtl-regclass
5068 @itemx -fdump-rtl-subregs_of_mode_init
5069 @itemx -fdump-rtl-subregs_of_mode_finish
5070 @itemx -fdump-rtl-dfinit
5071 @itemx -fdump-rtl-dfinish
5072 @opindex fdump-rtl-regclass
5073 @opindex fdump-rtl-subregs_of_mode_init
5074 @opindex fdump-rtl-subregs_of_mode_finish
5075 @opindex fdump-rtl-dfinit
5076 @opindex fdump-rtl-dfinish
5077 These dumps are defined but always produce empty files.
5079 @item -fdump-rtl-all
5080 @opindex fdump-rtl-all
5081 Produce all the dumps listed above.
5085 Annotate the assembler output with miscellaneous debugging information.
5089 Dump all macro definitions, at the end of preprocessing, in addition to
5094 Produce a core dump whenever an error occurs.
5098 Print statistics on memory usage, at the end of the run, to
5103 Annotate the assembler output with a comment indicating which
5104 pattern and alternative was used. The length of each instruction is
5109 Dump the RTL in the assembler output as a comment before each instruction.
5110 Also turns on @option{-dp} annotation.
5114 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
5115 dump a representation of the control flow graph suitable for viewing with VCG
5116 to @file{@var{file}.@var{pass}.vcg}.
5120 Just generate RTL for a function instead of compiling it. Usually used
5121 with @option{-fdump-rtl-expand}.
5125 Dump debugging information during parsing, to standard error.
5129 @opindex fdump-noaddr
5130 When doing debugging dumps, suppress address output. This makes it more
5131 feasible to use diff on debugging dumps for compiler invocations with
5132 different compiler binaries and/or different
5133 text / bss / data / heap / stack / dso start locations.
5135 @item -fdump-unnumbered
5136 @opindex fdump-unnumbered
5137 When doing debugging dumps, suppress instruction numbers and address output.
5138 This makes it more feasible to use diff on debugging dumps for compiler
5139 invocations with different options, in particular with and without
5142 @item -fdump-unnumbered-links
5143 @opindex fdump-unnumbered-links
5144 When doing debugging dumps (see @option{-d} option above), suppress
5145 instruction numbers for the links to the previous and next instructions
5148 @item -fdump-translation-unit @r{(C++ only)}
5149 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5150 @opindex fdump-translation-unit
5151 Dump a representation of the tree structure for the entire translation
5152 unit to a file. The file name is made by appending @file{.tu} to the
5153 source file name, and the file is created in the same directory as the
5154 output file. If the @samp{-@var{options}} form is used, @var{options}
5155 controls the details of the dump as described for the
5156 @option{-fdump-tree} options.
5158 @item -fdump-class-hierarchy @r{(C++ only)}
5159 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5160 @opindex fdump-class-hierarchy
5161 Dump a representation of each class's hierarchy and virtual function
5162 table layout to a file. The file name is made by appending
5163 @file{.class} to the source file name, and the file is created in the
5164 same directory as the output file. If the @samp{-@var{options}} form
5165 is used, @var{options} controls the details of the dump as described
5166 for the @option{-fdump-tree} options.
5168 @item -fdump-ipa-@var{switch}
5170 Control the dumping at various stages of inter-procedural analysis
5171 language tree to a file. The file name is generated by appending a
5172 switch specific suffix to the source file name, and the file is created
5173 in the same directory as the output file. The following dumps are
5178 Enables all inter-procedural analysis dumps.
5181 Dumps information about call-graph optimization, unused function removal,
5182 and inlining decisions.
5185 Dump after function inlining.
5189 @item -fdump-statistics-@var{option}
5190 @opindex fdump-statistics
5191 Enable and control dumping of pass statistics in a separate file. The
5192 file name is generated by appending a suffix ending in
5193 @samp{.statistics} to the source file name, and the file is created in
5194 the same directory as the output file. If the @samp{-@var{option}}
5195 form is used, @samp{-stats} will cause counters to be summed over the
5196 whole compilation unit while @samp{-details} will dump every event as
5197 the passes generate them. The default with no option is to sum
5198 counters for each function compiled.
5200 @item -fdump-tree-@var{switch}
5201 @itemx -fdump-tree-@var{switch}-@var{options}
5203 Control the dumping at various stages of processing the intermediate
5204 language tree to a file. The file name is generated by appending a
5205 switch specific suffix to the source file name, and the file is
5206 created in the same directory as the output file. If the
5207 @samp{-@var{options}} form is used, @var{options} is a list of
5208 @samp{-} separated options that control the details of the dump. Not
5209 all options are applicable to all dumps, those which are not
5210 meaningful will be ignored. The following options are available
5214 Print the address of each node. Usually this is not meaningful as it
5215 changes according to the environment and source file. Its primary use
5216 is for tying up a dump file with a debug environment.
5218 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
5219 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
5220 use working backward from mangled names in the assembly file.
5222 Inhibit dumping of members of a scope or body of a function merely
5223 because that scope has been reached. Only dump such items when they
5224 are directly reachable by some other path. When dumping pretty-printed
5225 trees, this option inhibits dumping the bodies of control structures.
5227 Print a raw representation of the tree. By default, trees are
5228 pretty-printed into a C-like representation.
5230 Enable more detailed dumps (not honored by every dump option).
5232 Enable dumping various statistics about the pass (not honored by every dump
5235 Enable showing basic block boundaries (disabled in raw dumps).
5237 Enable showing virtual operands for every statement.
5239 Enable showing line numbers for statements.
5241 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5243 Enable showing the tree dump for each statement.
5245 Enable showing the EH region number holding each statement.
5247 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5248 and @option{lineno}.
5251 The following tree dumps are possible:
5255 @opindex fdump-tree-original
5256 Dump before any tree based optimization, to @file{@var{file}.original}.
5259 @opindex fdump-tree-optimized
5260 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5263 @opindex fdump-tree-gimple
5264 Dump each function before and after the gimplification pass to a file. The
5265 file name is made by appending @file{.gimple} to the source file name.
5268 @opindex fdump-tree-cfg
5269 Dump the control flow graph of each function to a file. The file name is
5270 made by appending @file{.cfg} to the source file name.
5273 @opindex fdump-tree-vcg
5274 Dump the control flow graph of each function to a file in VCG format. The
5275 file name is made by appending @file{.vcg} to the source file name. Note
5276 that if the file contains more than one function, the generated file cannot
5277 be used directly by VCG@. You will need to cut and paste each function's
5278 graph into its own separate file first.
5281 @opindex fdump-tree-ch
5282 Dump each function after copying loop headers. The file name is made by
5283 appending @file{.ch} to the source file name.
5286 @opindex fdump-tree-ssa
5287 Dump SSA related information to a file. The file name is made by appending
5288 @file{.ssa} to the source file name.
5291 @opindex fdump-tree-alias
5292 Dump aliasing information for each function. The file name is made by
5293 appending @file{.alias} to the source file name.
5296 @opindex fdump-tree-ccp
5297 Dump each function after CCP@. The file name is made by appending
5298 @file{.ccp} to the source file name.
5301 @opindex fdump-tree-storeccp
5302 Dump each function after STORE-CCP@. The file name is made by appending
5303 @file{.storeccp} to the source file name.
5306 @opindex fdump-tree-pre
5307 Dump trees after partial redundancy elimination. The file name is made
5308 by appending @file{.pre} to the source file name.
5311 @opindex fdump-tree-fre
5312 Dump trees after full redundancy elimination. The file name is made
5313 by appending @file{.fre} to the source file name.
5316 @opindex fdump-tree-copyprop
5317 Dump trees after copy propagation. The file name is made
5318 by appending @file{.copyprop} to the source file name.
5320 @item store_copyprop
5321 @opindex fdump-tree-store_copyprop
5322 Dump trees after store copy-propagation. The file name is made
5323 by appending @file{.store_copyprop} to the source file name.
5326 @opindex fdump-tree-dce
5327 Dump each function after dead code elimination. The file name is made by
5328 appending @file{.dce} to the source file name.
5331 @opindex fdump-tree-mudflap
5332 Dump each function after adding mudflap instrumentation. The file name is
5333 made by appending @file{.mudflap} to the source file name.
5336 @opindex fdump-tree-sra
5337 Dump each function after performing scalar replacement of aggregates. The
5338 file name is made by appending @file{.sra} to the source file name.
5341 @opindex fdump-tree-sink
5342 Dump each function after performing code sinking. The file name is made
5343 by appending @file{.sink} to the source file name.
5346 @opindex fdump-tree-dom
5347 Dump each function after applying dominator tree optimizations. The file
5348 name is made by appending @file{.dom} to the source file name.
5351 @opindex fdump-tree-dse
5352 Dump each function after applying dead store elimination. The file
5353 name is made by appending @file{.dse} to the source file name.
5356 @opindex fdump-tree-phiopt
5357 Dump each function after optimizing PHI nodes into straightline code. The file
5358 name is made by appending @file{.phiopt} to the source file name.
5361 @opindex fdump-tree-forwprop
5362 Dump each function after forward propagating single use variables. The file
5363 name is made by appending @file{.forwprop} to the source file name.
5366 @opindex fdump-tree-copyrename
5367 Dump each function after applying the copy rename optimization. The file
5368 name is made by appending @file{.copyrename} to the source file name.
5371 @opindex fdump-tree-nrv
5372 Dump each function after applying the named return value optimization on
5373 generic trees. The file name is made by appending @file{.nrv} to the source
5377 @opindex fdump-tree-vect
5378 Dump each function after applying vectorization of loops. The file name is
5379 made by appending @file{.vect} to the source file name.
5382 @opindex fdump-tree-slp
5383 Dump each function after applying vectorization of basic blocks. The file name
5384 is made by appending @file{.slp} to the source file name.
5387 @opindex fdump-tree-vrp
5388 Dump each function after Value Range Propagation (VRP). The file name
5389 is made by appending @file{.vrp} to the source file name.
5392 @opindex fdump-tree-all
5393 Enable all the available tree dumps with the flags provided in this option.
5396 @item -ftree-vectorizer-verbose=@var{n}
5397 @opindex ftree-vectorizer-verbose
5398 This option controls the amount of debugging output the vectorizer prints.
5399 This information is written to standard error, unless
5400 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5401 in which case it is output to the usual dump listing file, @file{.vect}.
5402 For @var{n}=0 no diagnostic information is reported.
5403 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5404 and the total number of loops that got vectorized.
5405 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5406 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5407 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5408 level that @option{-fdump-tree-vect-stats} uses.
5409 Higher verbosity levels mean either more information dumped for each
5410 reported loop, or same amount of information reported for more loops:
5411 if @var{n}=3, vectorizer cost model information is reported.
5412 If @var{n}=4, alignment related information is added to the reports.
5413 If @var{n}=5, data-references related information (e.g.@: memory dependences,
5414 memory access-patterns) is added to the reports.
5415 If @var{n}=6, the vectorizer reports also non-vectorized inner-most loops
5416 that did not pass the first analysis phase (i.e., may not be countable, or
5417 may have complicated control-flow).
5418 If @var{n}=7, the vectorizer reports also non-vectorized nested loops.
5419 If @var{n}=8, SLP related information is added to the reports.
5420 For @var{n}=9, all the information the vectorizer generates during its
5421 analysis and transformation is reported. This is the same verbosity level
5422 that @option{-fdump-tree-vect-details} uses.
5424 @item -frandom-seed=@var{string}
5425 @opindex frandom-seed
5426 This option provides a seed that GCC uses when it would otherwise use
5427 random numbers. It is used to generate certain symbol names
5428 that have to be different in every compiled file. It is also used to
5429 place unique stamps in coverage data files and the object files that
5430 produce them. You can use the @option{-frandom-seed} option to produce
5431 reproducibly identical object files.
5433 The @var{string} should be different for every file you compile.
5435 @item -fsched-verbose=@var{n}
5436 @opindex fsched-verbose
5437 On targets that use instruction scheduling, this option controls the
5438 amount of debugging output the scheduler prints. This information is
5439 written to standard error, unless @option{-fdump-rtl-sched1} or
5440 @option{-fdump-rtl-sched2} is specified, in which case it is output
5441 to the usual dump listing file, @file{.sched1} or @file{.sched2}
5442 respectively. However for @var{n} greater than nine, the output is
5443 always printed to standard error.
5445 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5446 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5447 For @var{n} greater than one, it also output basic block probabilities,
5448 detailed ready list information and unit/insn info. For @var{n} greater
5449 than two, it includes RTL at abort point, control-flow and regions info.
5450 And for @var{n} over four, @option{-fsched-verbose} also includes
5454 @itemx -save-temps=cwd
5456 Store the usual ``temporary'' intermediate files permanently; place them
5457 in the current directory and name them based on the source file. Thus,
5458 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5459 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5460 preprocessed @file{foo.i} output file even though the compiler now
5461 normally uses an integrated preprocessor.
5463 When used in combination with the @option{-x} command line option,
5464 @option{-save-temps} is sensible enough to avoid over writing an
5465 input source file with the same extension as an intermediate file.
5466 The corresponding intermediate file may be obtained by renaming the
5467 source file before using @option{-save-temps}.
5469 If you invoke GCC in parallel, compiling several different source
5470 files that share a common base name in different subdirectories or the
5471 same source file compiled for multiple output destinations, it is
5472 likely that the different parallel compilers will interfere with each
5473 other, and overwrite the temporary files. For instance:
5476 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5477 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5480 may result in @file{foo.i} and @file{foo.o} being written to
5481 simultaneously by both compilers.
5483 @item -save-temps=obj
5484 @opindex save-temps=obj
5485 Store the usual ``temporary'' intermediate files permanently. If the
5486 @option{-o} option is used, the temporary files are based on the
5487 object file. If the @option{-o} option is not used, the
5488 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5493 gcc -save-temps=obj -c foo.c
5494 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5495 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5498 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5499 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5500 @file{dir2/yfoobar.o}.
5502 @item -time@r{[}=@var{file}@r{]}
5504 Report the CPU time taken by each subprocess in the compilation
5505 sequence. For C source files, this is the compiler proper and assembler
5506 (plus the linker if linking is done).
5508 Without the specification of an output file, the output looks like this:
5515 The first number on each line is the ``user time'', that is time spent
5516 executing the program itself. The second number is ``system time'',
5517 time spent executing operating system routines on behalf of the program.
5518 Both numbers are in seconds.
5520 With the specification of an output file, the output is appended to the
5521 named file, and it looks like this:
5524 0.12 0.01 cc1 @var{options}
5525 0.00 0.01 as @var{options}
5528 The ``user time'' and the ``system time'' are moved before the program
5529 name, and the options passed to the program are displayed, so that one
5530 can later tell what file was being compiled, and with which options.
5532 @item -fvar-tracking
5533 @opindex fvar-tracking
5534 Run variable tracking pass. It computes where variables are stored at each
5535 position in code. Better debugging information is then generated
5536 (if the debugging information format supports this information).
5538 It is enabled by default when compiling with optimization (@option{-Os},
5539 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5540 the debug info format supports it.
5542 @item -fvar-tracking-assignments
5543 @opindex fvar-tracking-assignments
5544 @opindex fno-var-tracking-assignments
5545 Annotate assignments to user variables early in the compilation and
5546 attempt to carry the annotations over throughout the compilation all the
5547 way to the end, in an attempt to improve debug information while
5548 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
5550 It can be enabled even if var-tracking is disabled, in which case
5551 annotations will be created and maintained, but discarded at the end.
5553 @item -fvar-tracking-assignments-toggle
5554 @opindex fvar-tracking-assignments-toggle
5555 @opindex fno-var-tracking-assignments-toggle
5556 Toggle @option{-fvar-tracking-assignments}, in the same way that
5557 @option{-gtoggle} toggles @option{-g}.
5559 @item -print-file-name=@var{library}
5560 @opindex print-file-name
5561 Print the full absolute name of the library file @var{library} that
5562 would be used when linking---and don't do anything else. With this
5563 option, GCC does not compile or link anything; it just prints the
5566 @item -print-multi-directory
5567 @opindex print-multi-directory
5568 Print the directory name corresponding to the multilib selected by any
5569 other switches present in the command line. This directory is supposed
5570 to exist in @env{GCC_EXEC_PREFIX}.
5572 @item -print-multi-lib
5573 @opindex print-multi-lib
5574 Print the mapping from multilib directory names to compiler switches
5575 that enable them. The directory name is separated from the switches by
5576 @samp{;}, and each switch starts with an @samp{@@} instead of the
5577 @samp{-}, without spaces between multiple switches. This is supposed to
5578 ease shell-processing.
5580 @item -print-multi-os-directory
5581 @opindex print-multi-os-directory
5582 Print the path to OS libraries for the selected
5583 multilib, relative to some @file{lib} subdirectory. If OS libraries are
5584 present in the @file{lib} subdirectory and no multilibs are used, this is
5585 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
5586 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
5587 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
5588 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
5590 @item -print-prog-name=@var{program}
5591 @opindex print-prog-name
5592 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5594 @item -print-libgcc-file-name
5595 @opindex print-libgcc-file-name
5596 Same as @option{-print-file-name=libgcc.a}.
5598 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5599 but you do want to link with @file{libgcc.a}. You can do
5602 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5605 @item -print-search-dirs
5606 @opindex print-search-dirs
5607 Print the name of the configured installation directory and a list of
5608 program and library directories @command{gcc} will search---and don't do anything else.
5610 This is useful when @command{gcc} prints the error message
5611 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5612 To resolve this you either need to put @file{cpp0} and the other compiler
5613 components where @command{gcc} expects to find them, or you can set the environment
5614 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5615 Don't forget the trailing @samp{/}.
5616 @xref{Environment Variables}.
5618 @item -print-sysroot
5619 @opindex print-sysroot
5620 Print the target sysroot directory that will be used during
5621 compilation. This is the target sysroot specified either at configure
5622 time or using the @option{--sysroot} option, possibly with an extra
5623 suffix that depends on compilation options. If no target sysroot is
5624 specified, the option prints nothing.
5626 @item -print-sysroot-headers-suffix
5627 @opindex print-sysroot-headers-suffix
5628 Print the suffix added to the target sysroot when searching for
5629 headers, or give an error if the compiler is not configured with such
5630 a suffix---and don't do anything else.
5633 @opindex dumpmachine
5634 Print the compiler's target machine (for example,
5635 @samp{i686-pc-linux-gnu})---and don't do anything else.
5638 @opindex dumpversion
5639 Print the compiler version (for example, @samp{3.0})---and don't do
5644 Print the compiler's built-in specs---and don't do anything else. (This
5645 is used when GCC itself is being built.) @xref{Spec Files}.
5647 @item -feliminate-unused-debug-types
5648 @opindex feliminate-unused-debug-types
5649 Normally, when producing DWARF2 output, GCC will emit debugging
5650 information for all types declared in a compilation
5651 unit, regardless of whether or not they are actually used
5652 in that compilation unit. Sometimes this is useful, such as
5653 if, in the debugger, you want to cast a value to a type that is
5654 not actually used in your program (but is declared). More often,
5655 however, this results in a significant amount of wasted space.
5656 With this option, GCC will avoid producing debug symbol output
5657 for types that are nowhere used in the source file being compiled.
5660 @node Optimize Options
5661 @section Options That Control Optimization
5662 @cindex optimize options
5663 @cindex options, optimization
5665 These options control various sorts of optimizations.
5667 Without any optimization option, the compiler's goal is to reduce the
5668 cost of compilation and to make debugging produce the expected
5669 results. Statements are independent: if you stop the program with a
5670 breakpoint between statements, you can then assign a new value to any
5671 variable or change the program counter to any other statement in the
5672 function and get exactly the results you would expect from the source
5675 Turning on optimization flags makes the compiler attempt to improve
5676 the performance and/or code size at the expense of compilation time
5677 and possibly the ability to debug the program.
5679 The compiler performs optimization based on the knowledge it has of the
5680 program. Compiling multiple files at once to a single output file mode allows
5681 the compiler to use information gained from all of the files when compiling
5684 Not all optimizations are controlled directly by a flag. Only
5685 optimizations that have a flag are listed in this section.
5687 Most optimizations are only enabled if an @option{-O} level is set on
5688 the command line. Otherwise they are disabled, even if individual
5689 optimization flags are specified.
5691 Depending on the target and how GCC was configured, a slightly different
5692 set of optimizations may be enabled at each @option{-O} level than
5693 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
5694 to find out the exact set of optimizations that are enabled at each level.
5695 @xref{Overall Options}, for examples.
5702 Optimize. Optimizing compilation takes somewhat more time, and a lot
5703 more memory for a large function.
5705 With @option{-O}, the compiler tries to reduce code size and execution
5706 time, without performing any optimizations that take a great deal of
5709 @option{-O} turns on the following optimization flags:
5712 -fcprop-registers @gol
5715 -fdelayed-branch @gol
5717 -fguess-branch-probability @gol
5718 -fif-conversion2 @gol
5719 -fif-conversion @gol
5720 -fipa-pure-const @gol
5721 -fipa-reference @gol
5723 -fsplit-wide-types @gol
5724 -ftree-builtin-call-dce @gol
5727 -ftree-copyrename @gol
5729 -ftree-dominator-opts @gol
5731 -ftree-forwprop @gol
5739 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5740 where doing so does not interfere with debugging.
5744 Optimize even more. GCC performs nearly all supported optimizations
5745 that do not involve a space-speed tradeoff.
5746 As compared to @option{-O}, this option increases both compilation time
5747 and the performance of the generated code.
5749 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5750 also turns on the following optimization flags:
5751 @gccoptlist{-fthread-jumps @gol
5752 -falign-functions -falign-jumps @gol
5753 -falign-loops -falign-labels @gol
5756 -fcse-follow-jumps -fcse-skip-blocks @gol
5757 -fdelete-null-pointer-checks @gol
5758 -fexpensive-optimizations @gol
5759 -fgcse -fgcse-lm @gol
5760 -finline-small-functions @gol
5761 -findirect-inlining @gol
5763 -foptimize-sibling-calls @gol
5766 -freorder-blocks -freorder-functions @gol
5767 -frerun-cse-after-loop @gol
5768 -fsched-interblock -fsched-spec @gol
5769 -fschedule-insns -fschedule-insns2 @gol
5770 -fstrict-aliasing -fstrict-overflow @gol
5771 -ftree-switch-conversion @gol
5775 Please note the warning under @option{-fgcse} about
5776 invoking @option{-O2} on programs that use computed gotos.
5780 Optimize yet more. @option{-O3} turns on all optimizations specified
5781 by @option{-O2} and also turns on the @option{-finline-functions},
5782 @option{-funswitch-loops}, @option{-fpredictive-commoning},
5783 @option{-fgcse-after-reload} and @option{-ftree-vectorize} options.
5787 Reduce compilation time and make debugging produce the expected
5788 results. This is the default.
5792 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5793 do not typically increase code size. It also performs further
5794 optimizations designed to reduce code size.
5796 @option{-Os} disables the following optimization flags:
5797 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5798 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5799 -fprefetch-loop-arrays -ftree-vect-loop-version}
5801 If you use multiple @option{-O} options, with or without level numbers,
5802 the last such option is the one that is effective.
5805 Options of the form @option{-f@var{flag}} specify machine-independent
5806 flags. Most flags have both positive and negative forms; the negative
5807 form of @option{-ffoo} would be @option{-fno-foo}. In the table
5808 below, only one of the forms is listed---the one you typically will
5809 use. You can figure out the other form by either removing @samp{no-}
5812 The following options control specific optimizations. They are either
5813 activated by @option{-O} options or are related to ones that are. You
5814 can use the following flags in the rare cases when ``fine-tuning'' of
5815 optimizations to be performed is desired.
5818 @item -fno-default-inline
5819 @opindex fno-default-inline
5820 Do not make member functions inline by default merely because they are
5821 defined inside the class scope (C++ only). Otherwise, when you specify
5822 @w{@option{-O}}, member functions defined inside class scope are compiled
5823 inline by default; i.e., you don't need to add @samp{inline} in front of
5824 the member function name.
5826 @item -fno-defer-pop
5827 @opindex fno-defer-pop
5828 Always pop the arguments to each function call as soon as that function
5829 returns. For machines which must pop arguments after a function call,
5830 the compiler normally lets arguments accumulate on the stack for several
5831 function calls and pops them all at once.
5833 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5835 @item -fforward-propagate
5836 @opindex fforward-propagate
5837 Perform a forward propagation pass on RTL@. The pass tries to combine two
5838 instructions and checks if the result can be simplified. If loop unrolling
5839 is active, two passes are performed and the second is scheduled after
5842 This option is enabled by default at optimization levels @option{-O},
5843 @option{-O2}, @option{-O3}, @option{-Os}.
5845 @item -fomit-frame-pointer
5846 @opindex fomit-frame-pointer
5847 Don't keep the frame pointer in a register for functions that
5848 don't need one. This avoids the instructions to save, set up and
5849 restore frame pointers; it also makes an extra register available
5850 in many functions. @strong{It also makes debugging impossible on
5853 On some machines, such as the VAX, this flag has no effect, because
5854 the standard calling sequence automatically handles the frame pointer
5855 and nothing is saved by pretending it doesn't exist. The
5856 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
5857 whether a target machine supports this flag. @xref{Registers,,Register
5858 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
5860 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5862 @item -foptimize-sibling-calls
5863 @opindex foptimize-sibling-calls
5864 Optimize sibling and tail recursive calls.
5866 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5870 Don't pay attention to the @code{inline} keyword. Normally this option
5871 is used to keep the compiler from expanding any functions inline.
5872 Note that if you are not optimizing, no functions can be expanded inline.
5874 @item -finline-small-functions
5875 @opindex finline-small-functions
5876 Integrate functions into their callers when their body is smaller than expected
5877 function call code (so overall size of program gets smaller). The compiler
5878 heuristically decides which functions are simple enough to be worth integrating
5881 Enabled at level @option{-O2}.
5883 @item -findirect-inlining
5884 @opindex findirect-inlining
5885 Inline also indirect calls that are discovered to be known at compile
5886 time thanks to previous inlining. This option has any effect only
5887 when inlining itself is turned on by the @option{-finline-functions}
5888 or @option{-finline-small-functions} options.
5890 Enabled at level @option{-O2}.
5892 @item -finline-functions
5893 @opindex finline-functions
5894 Integrate all simple functions into their callers. The compiler
5895 heuristically decides which functions are simple enough to be worth
5896 integrating in this way.
5898 If all calls to a given function are integrated, and the function is
5899 declared @code{static}, then the function is normally not output as
5900 assembler code in its own right.
5902 Enabled at level @option{-O3}.
5904 @item -finline-functions-called-once
5905 @opindex finline-functions-called-once
5906 Consider all @code{static} functions called once for inlining into their
5907 caller even if they are not marked @code{inline}. If a call to a given
5908 function is integrated, then the function is not output as assembler code
5911 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
5913 @item -fearly-inlining
5914 @opindex fearly-inlining
5915 Inline functions marked by @code{always_inline} and functions whose body seems
5916 smaller than the function call overhead early before doing
5917 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
5918 makes profiling significantly cheaper and usually inlining faster on programs
5919 having large chains of nested wrapper functions.
5925 Perform interprocedural scalar replacement of aggregates, removal of
5926 unused parameters and replacement of parameters passed by reference
5927 by parameters passed by value.
5929 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
5931 @item -finline-limit=@var{n}
5932 @opindex finline-limit
5933 By default, GCC limits the size of functions that can be inlined. This flag
5934 allows coarse control of this limit. @var{n} is the size of functions that
5935 can be inlined in number of pseudo instructions.
5937 Inlining is actually controlled by a number of parameters, which may be
5938 specified individually by using @option{--param @var{name}=@var{value}}.
5939 The @option{-finline-limit=@var{n}} option sets some of these parameters
5943 @item max-inline-insns-single
5944 is set to @var{n}/2.
5945 @item max-inline-insns-auto
5946 is set to @var{n}/2.
5949 See below for a documentation of the individual
5950 parameters controlling inlining and for the defaults of these parameters.
5952 @emph{Note:} there may be no value to @option{-finline-limit} that results
5953 in default behavior.
5955 @emph{Note:} pseudo instruction represents, in this particular context, an
5956 abstract measurement of function's size. In no way does it represent a count
5957 of assembly instructions and as such its exact meaning might change from one
5958 release to an another.
5960 @item -fkeep-inline-functions
5961 @opindex fkeep-inline-functions
5962 In C, emit @code{static} functions that are declared @code{inline}
5963 into the object file, even if the function has been inlined into all
5964 of its callers. This switch does not affect functions using the
5965 @code{extern inline} extension in GNU C90@. In C++, emit any and all
5966 inline functions into the object file.
5968 @item -fkeep-static-consts
5969 @opindex fkeep-static-consts
5970 Emit variables declared @code{static const} when optimization isn't turned
5971 on, even if the variables aren't referenced.
5973 GCC enables this option by default. If you want to force the compiler to
5974 check if the variable was referenced, regardless of whether or not
5975 optimization is turned on, use the @option{-fno-keep-static-consts} option.
5977 @item -fmerge-constants
5978 @opindex fmerge-constants
5979 Attempt to merge identical constants (string constants and floating point
5980 constants) across compilation units.
5982 This option is the default for optimized compilation if the assembler and
5983 linker support it. Use @option{-fno-merge-constants} to inhibit this
5986 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5988 @item -fmerge-all-constants
5989 @opindex fmerge-all-constants
5990 Attempt to merge identical constants and identical variables.
5992 This option implies @option{-fmerge-constants}. In addition to
5993 @option{-fmerge-constants} this considers e.g.@: even constant initialized
5994 arrays or initialized constant variables with integral or floating point
5995 types. Languages like C or C++ require each variable, including multiple
5996 instances of the same variable in recursive calls, to have distinct locations,
5997 so using this option will result in non-conforming
6000 @item -fmodulo-sched
6001 @opindex fmodulo-sched
6002 Perform swing modulo scheduling immediately before the first scheduling
6003 pass. This pass looks at innermost loops and reorders their
6004 instructions by overlapping different iterations.
6006 @item -fmodulo-sched-allow-regmoves
6007 @opindex fmodulo-sched-allow-regmoves
6008 Perform more aggressive SMS based modulo scheduling with register moves
6009 allowed. By setting this flag certain anti-dependences edges will be
6010 deleted which will trigger the generation of reg-moves based on the
6011 life-range analysis. This option is effective only with
6012 @option{-fmodulo-sched} enabled.
6014 @item -fno-branch-count-reg
6015 @opindex fno-branch-count-reg
6016 Do not use ``decrement and branch'' instructions on a count register,
6017 but instead generate a sequence of instructions that decrement a
6018 register, compare it against zero, then branch based upon the result.
6019 This option is only meaningful on architectures that support such
6020 instructions, which include x86, PowerPC, IA-64 and S/390.
6022 The default is @option{-fbranch-count-reg}.
6024 @item -fno-function-cse
6025 @opindex fno-function-cse
6026 Do not put function addresses in registers; make each instruction that
6027 calls a constant function contain the function's address explicitly.
6029 This option results in less efficient code, but some strange hacks
6030 that alter the assembler output may be confused by the optimizations
6031 performed when this option is not used.
6033 The default is @option{-ffunction-cse}
6035 @item -fno-zero-initialized-in-bss
6036 @opindex fno-zero-initialized-in-bss
6037 If the target supports a BSS section, GCC by default puts variables that
6038 are initialized to zero into BSS@. This can save space in the resulting
6041 This option turns off this behavior because some programs explicitly
6042 rely on variables going to the data section. E.g., so that the
6043 resulting executable can find the beginning of that section and/or make
6044 assumptions based on that.
6046 The default is @option{-fzero-initialized-in-bss}.
6048 @item -fmudflap -fmudflapth -fmudflapir
6052 @cindex bounds checking
6054 For front-ends that support it (C and C++), instrument all risky
6055 pointer/array dereferencing operations, some standard library
6056 string/heap functions, and some other associated constructs with
6057 range/validity tests. Modules so instrumented should be immune to
6058 buffer overflows, invalid heap use, and some other classes of C/C++
6059 programming errors. The instrumentation relies on a separate runtime
6060 library (@file{libmudflap}), which will be linked into a program if
6061 @option{-fmudflap} is given at link time. Run-time behavior of the
6062 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
6063 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
6066 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
6067 link if your program is multi-threaded. Use @option{-fmudflapir}, in
6068 addition to @option{-fmudflap} or @option{-fmudflapth}, if
6069 instrumentation should ignore pointer reads. This produces less
6070 instrumentation (and therefore faster execution) and still provides
6071 some protection against outright memory corrupting writes, but allows
6072 erroneously read data to propagate within a program.
6074 @item -fthread-jumps
6075 @opindex fthread-jumps
6076 Perform optimizations where we check to see if a jump branches to a
6077 location where another comparison subsumed by the first is found. If
6078 so, the first branch is redirected to either the destination of the
6079 second branch or a point immediately following it, depending on whether
6080 the condition is known to be true or false.
6082 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6084 @item -fsplit-wide-types
6085 @opindex fsplit-wide-types
6086 When using a type that occupies multiple registers, such as @code{long
6087 long} on a 32-bit system, split the registers apart and allocate them
6088 independently. This normally generates better code for those types,
6089 but may make debugging more difficult.
6091 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
6094 @item -fcse-follow-jumps
6095 @opindex fcse-follow-jumps
6096 In common subexpression elimination (CSE), scan through jump instructions
6097 when the target of the jump is not reached by any other path. For
6098 example, when CSE encounters an @code{if} statement with an
6099 @code{else} clause, CSE will follow the jump when the condition
6102 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6104 @item -fcse-skip-blocks
6105 @opindex fcse-skip-blocks
6106 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
6107 follow jumps which conditionally skip over blocks. When CSE
6108 encounters a simple @code{if} statement with no else clause,
6109 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6110 body of the @code{if}.
6112 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6114 @item -frerun-cse-after-loop
6115 @opindex frerun-cse-after-loop
6116 Re-run common subexpression elimination after loop optimizations has been
6119 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6123 Perform a global common subexpression elimination pass.
6124 This pass also performs global constant and copy propagation.
6126 @emph{Note:} When compiling a program using computed gotos, a GCC
6127 extension, you may get better runtime performance if you disable
6128 the global common subexpression elimination pass by adding
6129 @option{-fno-gcse} to the command line.
6131 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6135 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
6136 attempt to move loads which are only killed by stores into themselves. This
6137 allows a loop containing a load/store sequence to be changed to a load outside
6138 the loop, and a copy/store within the loop.
6140 Enabled by default when gcse is enabled.
6144 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6145 global common subexpression elimination. This pass will attempt to move
6146 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6147 loops containing a load/store sequence can be changed to a load before
6148 the loop and a store after the loop.
6150 Not enabled at any optimization level.
6154 When @option{-fgcse-las} is enabled, the global common subexpression
6155 elimination pass eliminates redundant loads that come after stores to the
6156 same memory location (both partial and full redundancies).
6158 Not enabled at any optimization level.
6160 @item -fgcse-after-reload
6161 @opindex fgcse-after-reload
6162 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6163 pass is performed after reload. The purpose of this pass is to cleanup
6166 @item -funsafe-loop-optimizations
6167 @opindex funsafe-loop-optimizations
6168 If given, the loop optimizer will assume that loop indices do not
6169 overflow, and that the loops with nontrivial exit condition are not
6170 infinite. This enables a wider range of loop optimizations even if
6171 the loop optimizer itself cannot prove that these assumptions are valid.
6172 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
6173 if it finds this kind of loop.
6175 @item -fcrossjumping
6176 @opindex fcrossjumping
6177 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6178 resulting code may or may not perform better than without cross-jumping.
6180 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6182 @item -fauto-inc-dec
6183 @opindex fauto-inc-dec
6184 Combine increments or decrements of addresses with memory accesses.
6185 This pass is always skipped on architectures that do not have
6186 instructions to support this. Enabled by default at @option{-O} and
6187 higher on architectures that support this.
6191 Perform dead code elimination (DCE) on RTL@.
6192 Enabled by default at @option{-O} and higher.
6196 Perform dead store elimination (DSE) on RTL@.
6197 Enabled by default at @option{-O} and higher.
6199 @item -fif-conversion
6200 @opindex fif-conversion
6201 Attempt to transform conditional jumps into branch-less equivalents. This
6202 include use of conditional moves, min, max, set flags and abs instructions, and
6203 some tricks doable by standard arithmetics. The use of conditional execution
6204 on chips where it is available is controlled by @code{if-conversion2}.
6206 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6208 @item -fif-conversion2
6209 @opindex fif-conversion2
6210 Use conditional execution (where available) to transform conditional jumps into
6211 branch-less equivalents.
6213 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6215 @item -fdelete-null-pointer-checks
6216 @opindex fdelete-null-pointer-checks
6217 Assume that programs cannot safely dereference null pointers, and that
6218 no code or data element resides there. This enables simple constant
6219 folding optimizations at all optimization levels. In addition, other
6220 optimization passes in GCC use this flag to control global dataflow
6221 analyses that eliminate useless checks for null pointers; these assume
6222 that if a pointer is checked after it has already been dereferenced,
6225 Note however that in some environments this assumption is not true.
6226 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6227 for programs which depend on that behavior.
6229 Some targets, especially embedded ones, disable this option at all levels.
6230 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6231 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6232 are enabled independently at different optimization levels.
6234 @item -fexpensive-optimizations
6235 @opindex fexpensive-optimizations
6236 Perform a number of minor optimizations that are relatively expensive.
6238 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6240 @item -foptimize-register-move
6242 @opindex foptimize-register-move
6244 Attempt to reassign register numbers in move instructions and as
6245 operands of other simple instructions in order to maximize the amount of
6246 register tying. This is especially helpful on machines with two-operand
6249 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6252 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6254 @item -fira-algorithm=@var{algorithm}
6255 Use specified coloring algorithm for the integrated register
6256 allocator. The @var{algorithm} argument should be @code{priority} or
6257 @code{CB}. The first algorithm specifies Chow's priority coloring,
6258 the second one specifies Chaitin-Briggs coloring. The second
6259 algorithm can be unimplemented for some architectures. If it is
6260 implemented, it is the default because Chaitin-Briggs coloring as a
6261 rule generates a better code.
6263 @item -fira-region=@var{region}
6264 Use specified regions for the integrated register allocator. The
6265 @var{region} argument should be one of @code{all}, @code{mixed}, or
6266 @code{one}. The first value means using all loops as register
6267 allocation regions, the second value which is the default means using
6268 all loops except for loops with small register pressure as the
6269 regions, and third one means using all function as a single region.
6270 The first value can give best result for machines with small size and
6271 irregular register set, the third one results in faster and generates
6272 decent code and the smallest size code, and the default value usually
6273 give the best results in most cases and for most architectures.
6275 @item -fira-coalesce
6276 @opindex fira-coalesce
6277 Do optimistic register coalescing. This option might be profitable for
6278 architectures with big regular register files.
6280 @item -fira-loop-pressure
6281 @opindex fira-loop-pressure
6282 Use IRA to evaluate register pressure in loops for decision to move
6283 loop invariants. Usage of this option usually results in generation
6284 of faster and smaller code on machines with big register files (>= 32
6285 registers) but it can slow compiler down.
6287 This option is enabled at level @option{-O3} for some targets.
6289 @item -fno-ira-share-save-slots
6290 @opindex fno-ira-share-save-slots
6291 Switch off sharing stack slots used for saving call used hard
6292 registers living through a call. Each hard register will get a
6293 separate stack slot and as a result function stack frame will be
6296 @item -fno-ira-share-spill-slots
6297 @opindex fno-ira-share-spill-slots
6298 Switch off sharing stack slots allocated for pseudo-registers. Each
6299 pseudo-register which did not get a hard register will get a separate
6300 stack slot and as a result function stack frame will be bigger.
6302 @item -fira-verbose=@var{n}
6303 @opindex fira-verbose
6304 Set up how verbose dump file for the integrated register allocator
6305 will be. Default value is 5. If the value is greater or equal to 10,
6306 the dump file will be stderr as if the value were @var{n} minus 10.
6308 @item -fdelayed-branch
6309 @opindex fdelayed-branch
6310 If supported for the target machine, attempt to reorder instructions
6311 to exploit instruction slots available after delayed branch
6314 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6316 @item -fschedule-insns
6317 @opindex fschedule-insns
6318 If supported for the target machine, attempt to reorder instructions to
6319 eliminate execution stalls due to required data being unavailable. This
6320 helps machines that have slow floating point or memory load instructions
6321 by allowing other instructions to be issued until the result of the load
6322 or floating point instruction is required.
6324 Enabled at levels @option{-O2}, @option{-O3}.
6326 @item -fschedule-insns2
6327 @opindex fschedule-insns2
6328 Similar to @option{-fschedule-insns}, but requests an additional pass of
6329 instruction scheduling after register allocation has been done. This is
6330 especially useful on machines with a relatively small number of
6331 registers and where memory load instructions take more than one cycle.
6333 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6335 @item -fno-sched-interblock
6336 @opindex fno-sched-interblock
6337 Don't schedule instructions across basic blocks. This is normally
6338 enabled by default when scheduling before register allocation, i.e.@:
6339 with @option{-fschedule-insns} or at @option{-O2} or higher.
6341 @item -fno-sched-spec
6342 @opindex fno-sched-spec
6343 Don't allow speculative motion of non-load instructions. This is normally
6344 enabled by default when scheduling before register allocation, i.e.@:
6345 with @option{-fschedule-insns} or at @option{-O2} or higher.
6347 @item -fsched-pressure
6348 @opindex fsched-pressure
6349 Enable register pressure sensitive insn scheduling before the register
6350 allocation. This only makes sense when scheduling before register
6351 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
6352 @option{-O2} or higher. Usage of this option can improve the
6353 generated code and decrease its size by preventing register pressure
6354 increase above the number of available hard registers and as a
6355 consequence register spills in the register allocation.
6357 @item -fsched-spec-load
6358 @opindex fsched-spec-load
6359 Allow speculative motion of some load instructions. This only makes
6360 sense when scheduling before register allocation, i.e.@: with
6361 @option{-fschedule-insns} or at @option{-O2} or higher.
6363 @item -fsched-spec-load-dangerous
6364 @opindex fsched-spec-load-dangerous
6365 Allow speculative motion of more load instructions. This only makes
6366 sense when scheduling before register allocation, i.e.@: with
6367 @option{-fschedule-insns} or at @option{-O2} or higher.
6369 @item -fsched-stalled-insns
6370 @itemx -fsched-stalled-insns=@var{n}
6371 @opindex fsched-stalled-insns
6372 Define how many insns (if any) can be moved prematurely from the queue
6373 of stalled insns into the ready list, during the second scheduling pass.
6374 @option{-fno-sched-stalled-insns} means that no insns will be moved
6375 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6376 on how many queued insns can be moved prematurely.
6377 @option{-fsched-stalled-insns} without a value is equivalent to
6378 @option{-fsched-stalled-insns=1}.
6380 @item -fsched-stalled-insns-dep
6381 @itemx -fsched-stalled-insns-dep=@var{n}
6382 @opindex fsched-stalled-insns-dep
6383 Define how many insn groups (cycles) will be examined for a dependency
6384 on a stalled insn that is candidate for premature removal from the queue
6385 of stalled insns. This has an effect only during the second scheduling pass,
6386 and only if @option{-fsched-stalled-insns} is used.
6387 @option{-fno-sched-stalled-insns-dep} is equivalent to
6388 @option{-fsched-stalled-insns-dep=0}.
6389 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6390 @option{-fsched-stalled-insns-dep=1}.
6392 @item -fsched2-use-superblocks
6393 @opindex fsched2-use-superblocks
6394 When scheduling after register allocation, do use superblock scheduling
6395 algorithm. Superblock scheduling allows motion across basic block boundaries
6396 resulting on faster schedules. This option is experimental, as not all machine
6397 descriptions used by GCC model the CPU closely enough to avoid unreliable
6398 results from the algorithm.
6400 This only makes sense when scheduling after register allocation, i.e.@: with
6401 @option{-fschedule-insns2} or at @option{-O2} or higher.
6403 @item -fsched-group-heuristic
6404 @opindex fsched-group-heuristic
6405 Enable the group heuristic in the scheduler. This heuristic favors
6406 the instruction that belongs to a schedule group. This is enabled
6407 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6408 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6410 @item -fsched-critical-path-heuristic
6411 @opindex fsched-critical-path-heuristic
6412 Enable the critical-path heuristic in the scheduler. This heuristic favors
6413 instructions on the critical path. This is enabled by default when
6414 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6415 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6417 @item -fsched-spec-insn-heuristic
6418 @opindex fsched-spec-insn-heuristic
6419 Enable the speculative instruction heuristic in the scheduler. This
6420 heuristic favors speculative instructions with greater dependency weakness.
6421 This is enabled by default when scheduling is enabled, i.e.@:
6422 with @option{-fschedule-insns} or @option{-fschedule-insns2}
6423 or at @option{-O2} or higher.
6425 @item -fsched-rank-heuristic
6426 @opindex fsched-rank-heuristic
6427 Enable the rank heuristic in the scheduler. This heuristic favors
6428 the instruction belonging to a basic block with greater size or frequency.
6429 This is enabled by default when scheduling is enabled, i.e.@:
6430 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6431 at @option{-O2} or higher.
6433 @item -fsched-last-insn-heuristic
6434 @opindex fsched-last-insn-heuristic
6435 Enable the last-instruction heuristic in the scheduler. This heuristic
6436 favors the instruction that is less dependent on the last instruction
6437 scheduled. This is enabled by default when scheduling is enabled,
6438 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6439 at @option{-O2} or higher.
6441 @item -fsched-dep-count-heuristic
6442 @opindex fsched-dep-count-heuristic
6443 Enable the dependent-count heuristic in the scheduler. This heuristic
6444 favors the instruction that has more instructions depending on it.
6445 This is enabled by default when scheduling is enabled, i.e.@:
6446 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6447 at @option{-O2} or higher.
6449 @item -freschedule-modulo-scheduled-loops
6450 @opindex freschedule-modulo-scheduled-loops
6451 The modulo scheduling comes before the traditional scheduling, if a loop
6452 was modulo scheduled we may want to prevent the later scheduling passes
6453 from changing its schedule, we use this option to control that.
6455 @item -fselective-scheduling
6456 @opindex fselective-scheduling
6457 Schedule instructions using selective scheduling algorithm. Selective
6458 scheduling runs instead of the first scheduler pass.
6460 @item -fselective-scheduling2
6461 @opindex fselective-scheduling2
6462 Schedule instructions using selective scheduling algorithm. Selective
6463 scheduling runs instead of the second scheduler pass.
6465 @item -fsel-sched-pipelining
6466 @opindex fsel-sched-pipelining
6467 Enable software pipelining of innermost loops during selective scheduling.
6468 This option has no effect until one of @option{-fselective-scheduling} or
6469 @option{-fselective-scheduling2} is turned on.
6471 @item -fsel-sched-pipelining-outer-loops
6472 @opindex fsel-sched-pipelining-outer-loops
6473 When pipelining loops during selective scheduling, also pipeline outer loops.
6474 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6476 @item -fcaller-saves
6477 @opindex fcaller-saves
6478 Enable values to be allocated in registers that will be clobbered by
6479 function calls, by emitting extra instructions to save and restore the
6480 registers around such calls. Such allocation is done only when it
6481 seems to result in better code than would otherwise be produced.
6483 This option is always enabled by default on certain machines, usually
6484 those which have no call-preserved registers to use instead.
6486 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6488 @item -fconserve-stack
6489 @opindex fconserve-stack
6490 Attempt to minimize stack usage. The compiler will attempt to use less
6491 stack space, even if that makes the program slower. This option
6492 implies setting the @option{large-stack-frame} parameter to 100
6493 and the @option{large-stack-frame-growth} parameter to 400.
6495 @item -ftree-reassoc
6496 @opindex ftree-reassoc
6497 Perform reassociation on trees. This flag is enabled by default
6498 at @option{-O} and higher.
6502 Perform partial redundancy elimination (PRE) on trees. This flag is
6503 enabled by default at @option{-O2} and @option{-O3}.
6505 @item -ftree-forwprop
6506 @opindex ftree-forwprop
6507 Perform forward propagation on trees. This flag is enabled by default
6508 at @option{-O} and higher.
6512 Perform full redundancy elimination (FRE) on trees. The difference
6513 between FRE and PRE is that FRE only considers expressions
6514 that are computed on all paths leading to the redundant computation.
6515 This analysis is faster than PRE, though it exposes fewer redundancies.
6516 This flag is enabled by default at @option{-O} and higher.
6518 @item -ftree-phiprop
6519 @opindex ftree-phiprop
6520 Perform hoisting of loads from conditional pointers on trees. This
6521 pass is enabled by default at @option{-O} and higher.
6523 @item -ftree-copy-prop
6524 @opindex ftree-copy-prop
6525 Perform copy propagation on trees. This pass eliminates unnecessary
6526 copy operations. This flag is enabled by default at @option{-O} and
6529 @item -fipa-pure-const
6530 @opindex fipa-pure-const
6531 Discover which functions are pure or constant.
6532 Enabled by default at @option{-O} and higher.
6534 @item -fipa-reference
6535 @opindex fipa-reference
6536 Discover which static variables do not escape cannot escape the
6538 Enabled by default at @option{-O} and higher.
6540 @item -fipa-struct-reorg
6541 @opindex fipa-struct-reorg
6542 Perform structure reorganization optimization, that change C-like structures
6543 layout in order to better utilize spatial locality. This transformation is
6544 affective for programs containing arrays of structures. Available in two
6545 compilation modes: profile-based (enabled with @option{-fprofile-generate})
6546 or static (which uses built-in heuristics). Require @option{-fipa-type-escape}
6547 to provide the safety of this transformation. It works only in whole program
6548 mode, so it requires @option{-fwhole-program} and @option{-combine} to be
6549 enabled. Structures considered @samp{cold} by this transformation are not
6550 affected (see @option{--param struct-reorg-cold-struct-ratio=@var{value}}).
6552 With this flag, the program debug info reflects a new structure layout.
6556 Perform interprocedural pointer analysis. This option is experimental
6557 and does not affect generated code.
6561 Perform interprocedural constant propagation.
6562 This optimization analyzes the program to determine when values passed
6563 to functions are constants and then optimizes accordingly.
6564 This optimization can substantially increase performance
6565 if the application has constants passed to functions.
6566 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
6568 @item -fipa-cp-clone
6569 @opindex fipa-cp-clone
6570 Perform function cloning to make interprocedural constant propagation stronger.
6571 When enabled, interprocedural constant propagation will perform function cloning
6572 when externally visible function can be called with constant arguments.
6573 Because this optimization can create multiple copies of functions,
6574 it may significantly increase code size
6575 (see @option{--param ipcp-unit-growth=@var{value}}).
6576 This flag is enabled by default at @option{-O3}.
6578 @item -fipa-matrix-reorg
6579 @opindex fipa-matrix-reorg
6580 Perform matrix flattening and transposing.
6581 Matrix flattening tries to replace an @math{m}-dimensional matrix
6582 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
6583 This reduces the level of indirection needed for accessing the elements
6584 of the matrix. The second optimization is matrix transposing that
6585 attempts to change the order of the matrix's dimensions in order to
6586 improve cache locality.
6587 Both optimizations need the @option{-fwhole-program} flag.
6588 Transposing is enabled only if profiling information is available.
6592 Perform forward store motion on trees. This flag is
6593 enabled by default at @option{-O} and higher.
6597 Perform sparse conditional constant propagation (CCP) on trees. This
6598 pass only operates on local scalar variables and is enabled by default
6599 at @option{-O} and higher.
6601 @item -ftree-switch-conversion
6602 Perform conversion of simple initializations in a switch to
6603 initializations from a scalar array. This flag is enabled by default
6604 at @option{-O2} and higher.
6608 Perform dead code elimination (DCE) on trees. This flag is enabled by
6609 default at @option{-O} and higher.
6611 @item -ftree-builtin-call-dce
6612 @opindex ftree-builtin-call-dce
6613 Perform conditional dead code elimination (DCE) for calls to builtin functions
6614 that may set @code{errno} but are otherwise side-effect free. This flag is
6615 enabled by default at @option{-O2} and higher if @option{-Os} is not also
6618 @item -ftree-dominator-opts
6619 @opindex ftree-dominator-opts
6620 Perform a variety of simple scalar cleanups (constant/copy
6621 propagation, redundancy elimination, range propagation and expression
6622 simplification) based on a dominator tree traversal. This also
6623 performs jump threading (to reduce jumps to jumps). This flag is
6624 enabled by default at @option{-O} and higher.
6628 Perform dead store elimination (DSE) on trees. A dead store is a store into
6629 a memory location which will later be overwritten by another store without
6630 any intervening loads. In this case the earlier store can be deleted. This
6631 flag is enabled by default at @option{-O} and higher.
6635 Perform loop header copying on trees. This is beneficial since it increases
6636 effectiveness of code motion optimizations. It also saves one jump. This flag
6637 is enabled by default at @option{-O} and higher. It is not enabled
6638 for @option{-Os}, since it usually increases code size.
6640 @item -ftree-loop-optimize
6641 @opindex ftree-loop-optimize
6642 Perform loop optimizations on trees. This flag is enabled by default
6643 at @option{-O} and higher.
6645 @item -ftree-loop-linear
6646 @opindex ftree-loop-linear
6647 Perform linear loop transformations on tree. This flag can improve cache
6648 performance and allow further loop optimizations to take place.
6650 @item -floop-interchange
6651 Perform loop interchange transformations on loops. Interchanging two
6652 nested loops switches the inner and outer loops. For example, given a
6657 A(J, I) = A(J, I) * C
6661 loop interchange will transform the loop as if the user had written:
6665 A(J, I) = A(J, I) * C
6669 which can be beneficial when @code{N} is larger than the caches,
6670 because in Fortran, the elements of an array are stored in memory
6671 contiguously by column, and the original loop iterates over rows,
6672 potentially creating at each access a cache miss. This optimization
6673 applies to all the languages supported by GCC and is not limited to
6674 Fortran. To use this code transformation, GCC has to be configured
6675 with @option{--with-ppl} and @option{--with-cloog} to enable the
6676 Graphite loop transformation infrastructure.
6678 @item -floop-strip-mine
6679 Perform loop strip mining transformations on loops. Strip mining
6680 splits a loop into two nested loops. The outer loop has strides
6681 equal to the strip size and the inner loop has strides of the
6682 original loop within a strip. The strip length can be changed
6683 using the @option{loop-block-tile-size} parameter. For example,
6690 loop strip mining will transform the loop as if the user had written:
6693 DO I = II, min (II + 50, N)
6698 This optimization applies to all the languages supported by GCC and is
6699 not limited to Fortran. To use this code transformation, GCC has to
6700 be configured with @option{--with-ppl} and @option{--with-cloog} to
6701 enable the Graphite loop transformation infrastructure.
6704 Perform loop blocking transformations on loops. Blocking strip mines
6705 each loop in the loop nest such that the memory accesses of the
6706 element loops fit inside caches. The strip length can be changed
6707 using the @option{loop-block-tile-size} parameter. For example, given
6712 A(J, I) = B(I) + C(J)
6716 loop blocking will transform the loop as if the user had written:
6720 DO I = II, min (II + 50, N)
6721 DO J = JJ, min (JJ + 50, M)
6722 A(J, I) = B(I) + C(J)
6728 which can be beneficial when @code{M} is larger than the caches,
6729 because the innermost loop will iterate over a smaller amount of data
6730 that can be kept in the caches. This optimization applies to all the
6731 languages supported by GCC and is not limited to Fortran. To use this
6732 code transformation, GCC has to be configured with @option{--with-ppl}
6733 and @option{--with-cloog} to enable the Graphite loop transformation
6736 @item -fgraphite-identity
6737 @opindex fgraphite-identity
6738 Enable the identity transformation for graphite. For every SCoP we generate
6739 the polyhedral representation and transform it back to gimple. Using
6740 @option{-fgraphite-identity} we can check the costs or benefits of the
6741 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
6742 are also performed by the code generator CLooG, like index splitting and
6743 dead code elimination in loops.
6745 @item -floop-parallelize-all
6746 Use the Graphite data dependence analysis to identify loops that can
6747 be parallelized. Parallelize all the loops that can be analyzed to
6748 not contain loop carried dependences without checking that it is
6749 profitable to parallelize the loops.
6751 @item -fcheck-data-deps
6752 @opindex fcheck-data-deps
6753 Compare the results of several data dependence analyzers. This option
6754 is used for debugging the data dependence analyzers.
6756 @item -ftree-loop-distribution
6757 Perform loop distribution. This flag can improve cache performance on
6758 big loop bodies and allow further loop optimizations, like
6759 parallelization or vectorization, to take place. For example, the loop
6776 @item -ftree-loop-im
6777 @opindex ftree-loop-im
6778 Perform loop invariant motion on trees. This pass moves only invariants that
6779 would be hard to handle at RTL level (function calls, operations that expand to
6780 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
6781 operands of conditions that are invariant out of the loop, so that we can use
6782 just trivial invariantness analysis in loop unswitching. The pass also includes
6785 @item -ftree-loop-ivcanon
6786 @opindex ftree-loop-ivcanon
6787 Create a canonical counter for number of iterations in the loop for that
6788 determining number of iterations requires complicated analysis. Later
6789 optimizations then may determine the number easily. Useful especially
6790 in connection with unrolling.
6794 Perform induction variable optimizations (strength reduction, induction
6795 variable merging and induction variable elimination) on trees.
6797 @item -ftree-parallelize-loops=n
6798 @opindex ftree-parallelize-loops
6799 Parallelize loops, i.e., split their iteration space to run in n threads.
6800 This is only possible for loops whose iterations are independent
6801 and can be arbitrarily reordered. The optimization is only
6802 profitable on multiprocessor machines, for loops that are CPU-intensive,
6803 rather than constrained e.g.@: by memory bandwidth. This option
6804 implies @option{-pthread}, and thus is only supported on targets
6805 that have support for @option{-pthread}.
6809 Perform function-local points-to analysis on trees. This flag is
6810 enabled by default at @option{-O} and higher.
6814 Perform scalar replacement of aggregates. This pass replaces structure
6815 references with scalars to prevent committing structures to memory too
6816 early. This flag is enabled by default at @option{-O} and higher.
6818 @item -ftree-copyrename
6819 @opindex ftree-copyrename
6820 Perform copy renaming on trees. This pass attempts to rename compiler
6821 temporaries to other variables at copy locations, usually resulting in
6822 variable names which more closely resemble the original variables. This flag
6823 is enabled by default at @option{-O} and higher.
6827 Perform temporary expression replacement during the SSA->normal phase. Single
6828 use/single def temporaries are replaced at their use location with their
6829 defining expression. This results in non-GIMPLE code, but gives the expanders
6830 much more complex trees to work on resulting in better RTL generation. This is
6831 enabled by default at @option{-O} and higher.
6833 @item -ftree-vectorize
6834 @opindex ftree-vectorize
6835 Perform loop vectorization on trees. This flag is enabled by default at
6838 @item -ftree-slp-vectorize
6839 @opindex ftree-slp-vectorize
6840 Perform basic block vectorization on trees. This flag is enabled by default at
6841 @option{-O3} and when @option{-ftree-vectorize} is enabled.
6843 @item -ftree-vect-loop-version
6844 @opindex ftree-vect-loop-version
6845 Perform loop versioning when doing loop vectorization on trees. When a loop
6846 appears to be vectorizable except that data alignment or data dependence cannot
6847 be determined at compile time then vectorized and non-vectorized versions of
6848 the loop are generated along with runtime checks for alignment or dependence
6849 to control which version is executed. This option is enabled by default
6850 except at level @option{-Os} where it is disabled.
6852 @item -fvect-cost-model
6853 @opindex fvect-cost-model
6854 Enable cost model for vectorization.
6858 Perform Value Range Propagation on trees. This is similar to the
6859 constant propagation pass, but instead of values, ranges of values are
6860 propagated. This allows the optimizers to remove unnecessary range
6861 checks like array bound checks and null pointer checks. This is
6862 enabled by default at @option{-O2} and higher. Null pointer check
6863 elimination is only done if @option{-fdelete-null-pointer-checks} is
6868 Perform tail duplication to enlarge superblock size. This transformation
6869 simplifies the control flow of the function allowing other optimizations to do
6872 @item -funroll-loops
6873 @opindex funroll-loops
6874 Unroll loops whose number of iterations can be determined at compile
6875 time or upon entry to the loop. @option{-funroll-loops} implies
6876 @option{-frerun-cse-after-loop}. This option makes code larger,
6877 and may or may not make it run faster.
6879 @item -funroll-all-loops
6880 @opindex funroll-all-loops
6881 Unroll all loops, even if their number of iterations is uncertain when
6882 the loop is entered. This usually makes programs run more slowly.
6883 @option{-funroll-all-loops} implies the same options as
6884 @option{-funroll-loops},
6886 @item -fsplit-ivs-in-unroller
6887 @opindex fsplit-ivs-in-unroller
6888 Enables expressing of values of induction variables in later iterations
6889 of the unrolled loop using the value in the first iteration. This breaks
6890 long dependency chains, thus improving efficiency of the scheduling passes.
6892 Combination of @option{-fweb} and CSE is often sufficient to obtain the
6893 same effect. However in cases the loop body is more complicated than
6894 a single basic block, this is not reliable. It also does not work at all
6895 on some of the architectures due to restrictions in the CSE pass.
6897 This optimization is enabled by default.
6899 @item -fvariable-expansion-in-unroller
6900 @opindex fvariable-expansion-in-unroller
6901 With this option, the compiler will create multiple copies of some
6902 local variables when unrolling a loop which can result in superior code.
6904 @item -fpredictive-commoning
6905 @opindex fpredictive-commoning
6906 Perform predictive commoning optimization, i.e., reusing computations
6907 (especially memory loads and stores) performed in previous
6908 iterations of loops.
6910 This option is enabled at level @option{-O3}.
6912 @item -fprefetch-loop-arrays
6913 @opindex fprefetch-loop-arrays
6914 If supported by the target machine, generate instructions to prefetch
6915 memory to improve the performance of loops that access large arrays.
6917 This option may generate better or worse code; results are highly
6918 dependent on the structure of loops within the source code.
6920 Disabled at level @option{-Os}.
6923 @itemx -fno-peephole2
6924 @opindex fno-peephole
6925 @opindex fno-peephole2
6926 Disable any machine-specific peephole optimizations. The difference
6927 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
6928 are implemented in the compiler; some targets use one, some use the
6929 other, a few use both.
6931 @option{-fpeephole} is enabled by default.
6932 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6934 @item -fno-guess-branch-probability
6935 @opindex fno-guess-branch-probability
6936 Do not guess branch probabilities using heuristics.
6938 GCC will use heuristics to guess branch probabilities if they are
6939 not provided by profiling feedback (@option{-fprofile-arcs}). These
6940 heuristics are based on the control flow graph. If some branch probabilities
6941 are specified by @samp{__builtin_expect}, then the heuristics will be
6942 used to guess branch probabilities for the rest of the control flow graph,
6943 taking the @samp{__builtin_expect} info into account. The interactions
6944 between the heuristics and @samp{__builtin_expect} can be complex, and in
6945 some cases, it may be useful to disable the heuristics so that the effects
6946 of @samp{__builtin_expect} are easier to understand.
6948 The default is @option{-fguess-branch-probability} at levels
6949 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6951 @item -freorder-blocks
6952 @opindex freorder-blocks
6953 Reorder basic blocks in the compiled function in order to reduce number of
6954 taken branches and improve code locality.
6956 Enabled at levels @option{-O2}, @option{-O3}.
6958 @item -freorder-blocks-and-partition
6959 @opindex freorder-blocks-and-partition
6960 In addition to reordering basic blocks in the compiled function, in order
6961 to reduce number of taken branches, partitions hot and cold basic blocks
6962 into separate sections of the assembly and .o files, to improve
6963 paging and cache locality performance.
6965 This optimization is automatically turned off in the presence of
6966 exception handling, for linkonce sections, for functions with a user-defined
6967 section attribute and on any architecture that does not support named
6970 @item -freorder-functions
6971 @opindex freorder-functions
6972 Reorder functions in the object file in order to
6973 improve code locality. This is implemented by using special
6974 subsections @code{.text.hot} for most frequently executed functions and
6975 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
6976 the linker so object file format must support named sections and linker must
6977 place them in a reasonable way.
6979 Also profile feedback must be available in to make this option effective. See
6980 @option{-fprofile-arcs} for details.
6982 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6984 @item -fstrict-aliasing
6985 @opindex fstrict-aliasing
6986 Allow the compiler to assume the strictest aliasing rules applicable to
6987 the language being compiled. For C (and C++), this activates
6988 optimizations based on the type of expressions. In particular, an
6989 object of one type is assumed never to reside at the same address as an
6990 object of a different type, unless the types are almost the same. For
6991 example, an @code{unsigned int} can alias an @code{int}, but not a
6992 @code{void*} or a @code{double}. A character type may alias any other
6995 @anchor{Type-punning}Pay special attention to code like this:
7008 The practice of reading from a different union member than the one most
7009 recently written to (called ``type-punning'') is common. Even with
7010 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
7011 is accessed through the union type. So, the code above will work as
7012 expected. @xref{Structures unions enumerations and bit-fields
7013 implementation}. However, this code might not:
7024 Similarly, access by taking the address, casting the resulting pointer
7025 and dereferencing the result has undefined behavior, even if the cast
7026 uses a union type, e.g.:
7030 return ((union a_union *) &d)->i;
7034 The @option{-fstrict-aliasing} option is enabled at levels
7035 @option{-O2}, @option{-O3}, @option{-Os}.
7037 @item -fstrict-overflow
7038 @opindex fstrict-overflow
7039 Allow the compiler to assume strict signed overflow rules, depending
7040 on the language being compiled. For C (and C++) this means that
7041 overflow when doing arithmetic with signed numbers is undefined, which
7042 means that the compiler may assume that it will not happen. This
7043 permits various optimizations. For example, the compiler will assume
7044 that an expression like @code{i + 10 > i} will always be true for
7045 signed @code{i}. This assumption is only valid if signed overflow is
7046 undefined, as the expression is false if @code{i + 10} overflows when
7047 using twos complement arithmetic. When this option is in effect any
7048 attempt to determine whether an operation on signed numbers will
7049 overflow must be written carefully to not actually involve overflow.
7051 This option also allows the compiler to assume strict pointer
7052 semantics: given a pointer to an object, if adding an offset to that
7053 pointer does not produce a pointer to the same object, the addition is
7054 undefined. This permits the compiler to conclude that @code{p + u >
7055 p} is always true for a pointer @code{p} and unsigned integer
7056 @code{u}. This assumption is only valid because pointer wraparound is
7057 undefined, as the expression is false if @code{p + u} overflows using
7058 twos complement arithmetic.
7060 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
7061 that integer signed overflow is fully defined: it wraps. When
7062 @option{-fwrapv} is used, there is no difference between
7063 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
7064 integers. With @option{-fwrapv} certain types of overflow are
7065 permitted. For example, if the compiler gets an overflow when doing
7066 arithmetic on constants, the overflowed value can still be used with
7067 @option{-fwrapv}, but not otherwise.
7069 The @option{-fstrict-overflow} option is enabled at levels
7070 @option{-O2}, @option{-O3}, @option{-Os}.
7072 @item -falign-functions
7073 @itemx -falign-functions=@var{n}
7074 @opindex falign-functions
7075 Align the start of functions to the next power-of-two greater than
7076 @var{n}, skipping up to @var{n} bytes. For instance,
7077 @option{-falign-functions=32} aligns functions to the next 32-byte
7078 boundary, but @option{-falign-functions=24} would align to the next
7079 32-byte boundary only if this can be done by skipping 23 bytes or less.
7081 @option{-fno-align-functions} and @option{-falign-functions=1} are
7082 equivalent and mean that functions will not be aligned.
7084 Some assemblers only support this flag when @var{n} is a power of two;
7085 in that case, it is rounded up.
7087 If @var{n} is not specified or is zero, use a machine-dependent default.
7089 Enabled at levels @option{-O2}, @option{-O3}.
7091 @item -falign-labels
7092 @itemx -falign-labels=@var{n}
7093 @opindex falign-labels
7094 Align all branch targets to a power-of-two boundary, skipping up to
7095 @var{n} bytes like @option{-falign-functions}. This option can easily
7096 make code slower, because it must insert dummy operations for when the
7097 branch target is reached in the usual flow of the code.
7099 @option{-fno-align-labels} and @option{-falign-labels=1} are
7100 equivalent and mean that labels will not be aligned.
7102 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
7103 are greater than this value, then their values are used instead.
7105 If @var{n} is not specified or is zero, use a machine-dependent default
7106 which is very likely to be @samp{1}, meaning no alignment.
7108 Enabled at levels @option{-O2}, @option{-O3}.
7111 @itemx -falign-loops=@var{n}
7112 @opindex falign-loops
7113 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7114 like @option{-falign-functions}. The hope is that the loop will be
7115 executed many times, which will make up for any execution of the dummy
7118 @option{-fno-align-loops} and @option{-falign-loops=1} are
7119 equivalent and mean that loops will not be aligned.
7121 If @var{n} is not specified or is zero, use a machine-dependent default.
7123 Enabled at levels @option{-O2}, @option{-O3}.
7126 @itemx -falign-jumps=@var{n}
7127 @opindex falign-jumps
7128 Align branch targets to a power-of-two boundary, for branch targets
7129 where the targets can only be reached by jumping, skipping up to @var{n}
7130 bytes like @option{-falign-functions}. In this case, no dummy operations
7133 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7134 equivalent and mean that loops will not be aligned.
7136 If @var{n} is not specified or is zero, use a machine-dependent default.
7138 Enabled at levels @option{-O2}, @option{-O3}.
7140 @item -funit-at-a-time
7141 @opindex funit-at-a-time
7142 This option is left for compatibility reasons. @option{-funit-at-a-time}
7143 has no effect, while @option{-fno-unit-at-a-time} implies
7144 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7148 @item -fno-toplevel-reorder
7149 @opindex fno-toplevel-reorder
7150 Do not reorder top-level functions, variables, and @code{asm}
7151 statements. Output them in the same order that they appear in the
7152 input file. When this option is used, unreferenced static variables
7153 will not be removed. This option is intended to support existing code
7154 which relies on a particular ordering. For new code, it is better to
7157 Enabled at level @option{-O0}. When disabled explicitly, it also imply
7158 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
7163 Constructs webs as commonly used for register allocation purposes and assign
7164 each web individual pseudo register. This allows the register allocation pass
7165 to operate on pseudos directly, but also strengthens several other optimization
7166 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7167 however, make debugging impossible, since variables will no longer stay in a
7170 Enabled by default with @option{-funroll-loops}.
7172 @item -fwhole-program
7173 @opindex fwhole-program
7174 Assume that the current compilation unit represents the whole program being
7175 compiled. All public functions and variables with the exception of @code{main}
7176 and those merged by attribute @code{externally_visible} become static functions
7177 and in effect are optimized more aggressively by interprocedural optimizers.
7178 While this option is equivalent to proper use of the @code{static} keyword for
7179 programs consisting of a single file, in combination with option
7180 @option{-combine}, @option{-flto} or @option{-fwhopr} this flag can be used to
7181 compile many smaller scale programs since the functions and variables become
7182 local for the whole combined compilation unit, not for the single source file
7185 This option implies @option{-fwhole-file} for Fortran programs.
7189 This option runs the standard link-time optimizer. When invoked
7190 with source code, it generates GIMPLE (one of GCC's internal
7191 representations) and writes it to special ELF sections in the object
7192 file. When the object files are linked together, all the function
7193 bodies are read from these ELF sections and instantiated as if they
7194 had been part of the same translation unit.
7196 To use the link-timer optimizer, @option{-flto} needs to be specified at
7197 compile time and during the final link. For example,
7200 gcc -c -O2 -flto foo.c
7201 gcc -c -O2 -flto bar.c
7202 gcc -o myprog -flto -O2 foo.o bar.o
7205 The first two invocations to GCC will save a bytecode representation
7206 of GIMPLE into special ELF sections inside @file{foo.o} and
7207 @file{bar.o}. The final invocation will read the GIMPLE bytecode from
7208 @file{foo.o} and @file{bar.o}, merge the two files into a single
7209 internal image, and compile the result as usual. Since both
7210 @file{foo.o} and @file{bar.o} are merged into a single image, this
7211 causes all the inter-procedural analyses and optimizations in GCC to
7212 work across the two files as if they were a single one. This means,
7213 for example, that the inliner will be able to inline functions in
7214 @file{bar.o} into functions in @file{foo.o} and vice-versa.
7216 Another (simpler) way to enable link-time optimization is,
7219 gcc -o myprog -flto -O2 foo.c bar.c
7222 The above will generate bytecode for @file{foo.c} and @file{bar.c},
7223 merge them together into a single GIMPLE representation and optimize
7224 them as usual to produce @file{myprog}.
7226 The only important thing to keep in mind is that to enable link-time
7227 optimizations the @option{-flto} flag needs to be passed to both the
7228 compile and the link commands.
7230 Note that when a file is compiled with @option{-flto}, the generated
7231 object file will be larger than a regular object file because it will
7232 contain GIMPLE bytecodes and the usual final code. This means that
7233 object files with LTO information can be linked as a normal object
7234 file. So, in the previous example, if the final link is done with
7237 gcc -o myprog foo.o bar.o
7240 The only difference will be that no inter-procedural optimizations
7241 will be applied to produce @file{myprog}. The two object files
7242 @file{foo.o} and @file{bar.o} will be simply sent to the regular
7245 Additionally, the optimization flags used to compile individual files
7246 are not necessarily related to those used at link-time. For instance,
7249 gcc -c -O0 -flto foo.c
7250 gcc -c -O0 -flto bar.c
7251 gcc -o myprog -flto -O3 foo.o bar.o
7254 This will produce individual object files with unoptimized assembler
7255 code, but the resulting binary @file{myprog} will be optimized at
7256 @option{-O3}. Now, if the final binary is generated without
7257 @option{-flto}, then @file{myprog} will not be optimized.
7259 When producing the final binary with @option{-flto}, GCC will only
7260 apply link-time optimizations to those files that contain bytecode.
7261 Therefore, you can mix and match object files and libraries with
7262 GIMPLE bytecodes and final object code. GCC will automatically select
7263 which files to optimize in LTO mode and which files to link without
7266 There are some code generation flags that GCC will preserve when
7267 generating bytecodes, as they need to be used during the final link
7268 stage. Currently, the following options are saved into the GIMPLE
7269 bytecode files: @option{-fPIC}, @option{-fcommon} and all the
7270 @option{-m} target flags.
7272 At link time, these options are read-in and reapplied. Note that the
7273 current implementation makes no attempt at recognizing conflicting
7274 values for these options. If two or more files have a conflicting
7275 value (e.g., one file is compiled with @option{-fPIC} and another
7276 isn't), the compiler will simply use the last value read from the
7277 bytecode files. It is recommended, then, that all the files
7278 participating in the same link be compiled with the same options.
7280 Another feature of LTO is that it is possible to apply interprocedural
7281 optimizations on files written in different languages. This requires
7282 some support in the language front end. Currently, the C, C++ and
7283 Fortran front ends are capable of emitting GIMPLE bytecodes, so
7284 something like this should work
7289 gfortran -c -flto baz.f90
7290 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
7293 Notice that the final link is done with @command{g++} to get the C++
7294 runtime libraries and @option{-lgfortran} is added to get the Fortran
7295 runtime libraries. In general, when mixing languages in LTO mode, you
7296 should use the same link command used when mixing languages in a
7297 regular (non-LTO) compilation. This means that if your build process
7298 was mixing languages before, all you need to add is @option{-flto} to
7299 all the compile and link commands.
7301 If LTO encounters objects with C linkage declared with incompatible
7302 types in separate translation units to be linked together (undefined
7303 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
7304 issued. The behavior is still undefined at runtime.
7306 If object files containing GIMPLE bytecode are stored in a library
7307 archive, say @file{libfoo.a}, it is possible to extract and use them
7308 in an LTO link if you are using @command{gold} as the linker (which,
7309 in turn requires GCC to be configured with @option{--enable-gold}).
7310 To enable this feature, use the flag @option{-fuse-linker-plugin} at
7314 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
7317 With the linker plugin enabled, @command{gold} will extract the needed
7318 GIMPLE files from @file{libfoo.a} and pass them on to the running GCC
7319 to make them part of the aggregated GIMPLE image to be optimized.
7321 If you are not using @command{gold} and/or do not specify
7322 @option{-fuse-linker-plugin} then the objects inside @file{libfoo.a}
7323 will be extracted and linked as usual, but they will not participate
7324 in the LTO optimization process.
7326 Link time optimizations do not require the presence of the whole
7327 program to operate. If the program does not require any symbols to
7328 be exported, it is possible to combine @option{-flto} and
7329 @option{-fwhopr} with @option{-fwhole-program} to allow the
7330 interprocedural optimizers to use more aggressive assumptions which
7331 may lead to improved optimization opportunities.
7333 Regarding portability: the current implementation of LTO makes no
7334 attempt at generating bytecode that can be ported between different
7335 types of hosts. The bytecode files are versioned and there is a
7336 strict version check, so bytecode files generated in one version of
7337 GCC will not work with an older/newer version of GCC.
7339 Link time optimization does not play well with generating debugging
7340 information. Combining @option{-flto} or @option{-fwhopr} with
7341 @option{-g} is experimental.
7343 This option is disabled by default.
7347 This option is identical in functionality to @option{-flto} but it
7348 differs in how the final link stage is executed. Instead of loading
7349 all the function bodies in memory, the callgraph is analyzed and
7350 optimization decisions are made (whole program analysis or WPA). Once
7351 optimization decisions are made, the callgraph is partitioned and the
7352 different sections are compiled separately (local transformations or
7353 LTRANS)@. This process allows optimizations on very large programs
7354 that otherwise would not fit in memory. This option enables
7355 @option{-fwpa} and @option{-fltrans} automatically.
7357 Disabled by default.
7359 This option is experimental.
7363 This is an internal option used by GCC when compiling with
7364 @option{-fwhopr}. You should never need to use it.
7366 This option runs the link-time optimizer in the whole-program-analysis
7367 (WPA) mode, which reads in summary information from all inputs and
7368 performs a whole-program analysis based on summary information only.
7369 It generates object files for subsequent runs of the link-time
7370 optimizer where individual object files are optimized using both
7371 summary information from the WPA mode and the actual function bodies.
7372 It then drives the LTRANS phase.
7374 Disabled by default.
7378 This is an internal option used by GCC when compiling with
7379 @option{-fwhopr}. You should never need to use it.
7381 This option runs the link-time optimizer in the local-transformation (LTRANS)
7382 mode, which reads in output from a previous run of the LTO in WPA mode.
7383 In the LTRANS mode, LTO optimizes an object and produces the final assembly.
7385 Disabled by default.
7387 @item -fltrans-output-list=@var{file}
7388 @opindex fltrans-output-list
7389 This is an internal option used by GCC when compiling with
7390 @option{-fwhopr}. You should never need to use it.
7392 This option specifies a file to which the names of LTRANS output files are
7393 written. This option is only meaningful in conjunction with @option{-fwpa}.
7395 Disabled by default.
7397 @item -flto-compression-level=@var{n}
7398 This option specifies the level of compression used for intermediate
7399 language written to LTO object files, and is only meaningful in
7400 conjunction with LTO mode (@option{-fwhopr}, @option{-flto}). Valid
7401 values are 0 (no compression) to 9 (maximum compression). Values
7402 outside this range are clamped to either 0 or 9. If the option is not
7403 given, a default balanced compression setting is used.
7406 Prints a report with internal details on the workings of the link-time
7407 optimizer. The contents of this report vary from version to version,
7408 it is meant to be useful to GCC developers when processing object
7409 files in LTO mode (via @option{-fwhopr} or @option{-flto}).
7411 Disabled by default.
7413 @item -fuse-linker-plugin
7414 Enables the extraction of objects with GIMPLE bytecode information
7415 from library archives. This option relies on features available only
7416 in @command{gold}, so to use this you must configure GCC with
7417 @option{--enable-gold}. See @option{-flto} for a description on the
7418 effect of this flag and how to use it.
7420 Disabled by default.
7422 @item -fcprop-registers
7423 @opindex fcprop-registers
7424 After register allocation and post-register allocation instruction splitting,
7425 we perform a copy-propagation pass to try to reduce scheduling dependencies
7426 and occasionally eliminate the copy.
7428 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7430 @item -fprofile-correction
7431 @opindex fprofile-correction
7432 Profiles collected using an instrumented binary for multi-threaded programs may
7433 be inconsistent due to missed counter updates. When this option is specified,
7434 GCC will use heuristics to correct or smooth out such inconsistencies. By
7435 default, GCC will emit an error message when an inconsistent profile is detected.
7437 @item -fprofile-dir=@var{path}
7438 @opindex fprofile-dir
7440 Set the directory to search the profile data files in to @var{path}.
7441 This option affects only the profile data generated by
7442 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
7443 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
7444 and its related options.
7445 By default, GCC will use the current directory as @var{path}
7446 thus the profile data file will appear in the same directory as the object file.
7448 @item -fprofile-generate
7449 @itemx -fprofile-generate=@var{path}
7450 @opindex fprofile-generate
7452 Enable options usually used for instrumenting application to produce
7453 profile useful for later recompilation with profile feedback based
7454 optimization. You must use @option{-fprofile-generate} both when
7455 compiling and when linking your program.
7457 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
7459 If @var{path} is specified, GCC will look at the @var{path} to find
7460 the profile feedback data files. See @option{-fprofile-dir}.
7463 @itemx -fprofile-use=@var{path}
7464 @opindex fprofile-use
7465 Enable profile feedback directed optimizations, and optimizations
7466 generally profitable only with profile feedback available.
7468 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
7469 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
7471 By default, GCC emits an error message if the feedback profiles do not
7472 match the source code. This error can be turned into a warning by using
7473 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
7476 If @var{path} is specified, GCC will look at the @var{path} to find
7477 the profile feedback data files. See @option{-fprofile-dir}.
7480 The following options control compiler behavior regarding floating
7481 point arithmetic. These options trade off between speed and
7482 correctness. All must be specifically enabled.
7486 @opindex ffloat-store
7487 Do not store floating point variables in registers, and inhibit other
7488 options that might change whether a floating point value is taken from a
7491 @cindex floating point precision
7492 This option prevents undesirable excess precision on machines such as
7493 the 68000 where the floating registers (of the 68881) keep more
7494 precision than a @code{double} is supposed to have. Similarly for the
7495 x86 architecture. For most programs, the excess precision does only
7496 good, but a few programs rely on the precise definition of IEEE floating
7497 point. Use @option{-ffloat-store} for such programs, after modifying
7498 them to store all pertinent intermediate computations into variables.
7500 @item -fexcess-precision=@var{style}
7501 @opindex fexcess-precision
7502 This option allows further control over excess precision on machines
7503 where floating-point registers have more precision than the IEEE
7504 @code{float} and @code{double} types and the processor does not
7505 support operations rounding to those types. By default,
7506 @option{-fexcess-precision=fast} is in effect; this means that
7507 operations are carried out in the precision of the registers and that
7508 it is unpredictable when rounding to the types specified in the source
7509 code takes place. When compiling C, if
7510 @option{-fexcess-precision=standard} is specified then excess
7511 precision will follow the rules specified in ISO C99; in particular,
7512 both casts and assignments cause values to be rounded to their
7513 semantic types (whereas @option{-ffloat-store} only affects
7514 assignments). This option is enabled by default for C if a strict
7515 conformance option such as @option{-std=c99} is used.
7518 @option{-fexcess-precision=standard} is not implemented for languages
7519 other than C, and has no effect if
7520 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
7521 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
7522 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
7523 semantics apply without excess precision, and in the latter, rounding
7528 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
7529 @option{-ffinite-math-only}, @option{-fno-rounding-math},
7530 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
7532 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
7534 This option is not turned on by any @option{-O} option since
7535 it can result in incorrect output for programs which depend on
7536 an exact implementation of IEEE or ISO rules/specifications for
7537 math functions. It may, however, yield faster code for programs
7538 that do not require the guarantees of these specifications.
7540 @item -fno-math-errno
7541 @opindex fno-math-errno
7542 Do not set ERRNO after calling math functions that are executed
7543 with a single instruction, e.g., sqrt. A program that relies on
7544 IEEE exceptions for math error handling may want to use this flag
7545 for speed while maintaining IEEE arithmetic compatibility.
7547 This option is not turned on by any @option{-O} option since
7548 it can result in incorrect output for programs which depend on
7549 an exact implementation of IEEE or ISO rules/specifications for
7550 math functions. It may, however, yield faster code for programs
7551 that do not require the guarantees of these specifications.
7553 The default is @option{-fmath-errno}.
7555 On Darwin systems, the math library never sets @code{errno}. There is
7556 therefore no reason for the compiler to consider the possibility that
7557 it might, and @option{-fno-math-errno} is the default.
7559 @item -funsafe-math-optimizations
7560 @opindex funsafe-math-optimizations
7562 Allow optimizations for floating-point arithmetic that (a) assume
7563 that arguments and results are valid and (b) may violate IEEE or
7564 ANSI standards. When used at link-time, it may include libraries
7565 or startup files that change the default FPU control word or other
7566 similar optimizations.
7568 This option is not turned on by any @option{-O} option since
7569 it can result in incorrect output for programs which depend on
7570 an exact implementation of IEEE or ISO rules/specifications for
7571 math functions. It may, however, yield faster code for programs
7572 that do not require the guarantees of these specifications.
7573 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
7574 @option{-fassociative-math} and @option{-freciprocal-math}.
7576 The default is @option{-fno-unsafe-math-optimizations}.
7578 @item -fassociative-math
7579 @opindex fassociative-math
7581 Allow re-association of operands in series of floating-point operations.
7582 This violates the ISO C and C++ language standard by possibly changing
7583 computation result. NOTE: re-ordering may change the sign of zero as
7584 well as ignore NaNs and inhibit or create underflow or overflow (and
7585 thus cannot be used on a code which relies on rounding behavior like
7586 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
7587 and thus may not be used when ordered comparisons are required.
7588 This option requires that both @option{-fno-signed-zeros} and
7589 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
7590 much sense with @option{-frounding-math}. For Fortran the option
7591 is automatically enabled when both @option{-fno-signed-zeros} and
7592 @option{-fno-trapping-math} are in effect.
7594 The default is @option{-fno-associative-math}.
7596 @item -freciprocal-math
7597 @opindex freciprocal-math
7599 Allow the reciprocal of a value to be used instead of dividing by
7600 the value if this enables optimizations. For example @code{x / y}
7601 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
7602 is subject to common subexpression elimination. Note that this loses
7603 precision and increases the number of flops operating on the value.
7605 The default is @option{-fno-reciprocal-math}.
7607 @item -ffinite-math-only
7608 @opindex ffinite-math-only
7609 Allow optimizations for floating-point arithmetic that assume
7610 that arguments and results are not NaNs or +-Infs.
7612 This option is not turned on by any @option{-O} option since
7613 it can result in incorrect output for programs which depend on
7614 an exact implementation of IEEE or ISO rules/specifications for
7615 math functions. It may, however, yield faster code for programs
7616 that do not require the guarantees of these specifications.
7618 The default is @option{-fno-finite-math-only}.
7620 @item -fno-signed-zeros
7621 @opindex fno-signed-zeros
7622 Allow optimizations for floating point arithmetic that ignore the
7623 signedness of zero. IEEE arithmetic specifies the behavior of
7624 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
7625 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
7626 This option implies that the sign of a zero result isn't significant.
7628 The default is @option{-fsigned-zeros}.
7630 @item -fno-trapping-math
7631 @opindex fno-trapping-math
7632 Compile code assuming that floating-point operations cannot generate
7633 user-visible traps. These traps include division by zero, overflow,
7634 underflow, inexact result and invalid operation. This option requires
7635 that @option{-fno-signaling-nans} be in effect. Setting this option may
7636 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
7638 This option should never be turned on by any @option{-O} option since
7639 it can result in incorrect output for programs which depend on
7640 an exact implementation of IEEE or ISO rules/specifications for
7643 The default is @option{-ftrapping-math}.
7645 @item -frounding-math
7646 @opindex frounding-math
7647 Disable transformations and optimizations that assume default floating
7648 point rounding behavior. This is round-to-zero for all floating point
7649 to integer conversions, and round-to-nearest for all other arithmetic
7650 truncations. This option should be specified for programs that change
7651 the FP rounding mode dynamically, or that may be executed with a
7652 non-default rounding mode. This option disables constant folding of
7653 floating point expressions at compile-time (which may be affected by
7654 rounding mode) and arithmetic transformations that are unsafe in the
7655 presence of sign-dependent rounding modes.
7657 The default is @option{-fno-rounding-math}.
7659 This option is experimental and does not currently guarantee to
7660 disable all GCC optimizations that are affected by rounding mode.
7661 Future versions of GCC may provide finer control of this setting
7662 using C99's @code{FENV_ACCESS} pragma. This command line option
7663 will be used to specify the default state for @code{FENV_ACCESS}.
7665 @item -fsignaling-nans
7666 @opindex fsignaling-nans
7667 Compile code assuming that IEEE signaling NaNs may generate user-visible
7668 traps during floating-point operations. Setting this option disables
7669 optimizations that may change the number of exceptions visible with
7670 signaling NaNs. This option implies @option{-ftrapping-math}.
7672 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
7675 The default is @option{-fno-signaling-nans}.
7677 This option is experimental and does not currently guarantee to
7678 disable all GCC optimizations that affect signaling NaN behavior.
7680 @item -fsingle-precision-constant
7681 @opindex fsingle-precision-constant
7682 Treat floating point constant as single precision constant instead of
7683 implicitly converting it to double precision constant.
7685 @item -fcx-limited-range
7686 @opindex fcx-limited-range
7687 When enabled, this option states that a range reduction step is not
7688 needed when performing complex division. Also, there is no checking
7689 whether the result of a complex multiplication or division is @code{NaN
7690 + I*NaN}, with an attempt to rescue the situation in that case. The
7691 default is @option{-fno-cx-limited-range}, but is enabled by
7692 @option{-ffast-math}.
7694 This option controls the default setting of the ISO C99
7695 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
7698 @item -fcx-fortran-rules
7699 @opindex fcx-fortran-rules
7700 Complex multiplication and division follow Fortran rules. Range
7701 reduction is done as part of complex division, but there is no checking
7702 whether the result of a complex multiplication or division is @code{NaN
7703 + I*NaN}, with an attempt to rescue the situation in that case.
7705 The default is @option{-fno-cx-fortran-rules}.
7709 The following options control optimizations that may improve
7710 performance, but are not enabled by any @option{-O} options. This
7711 section includes experimental options that may produce broken code.
7714 @item -fbranch-probabilities
7715 @opindex fbranch-probabilities
7716 After running a program compiled with @option{-fprofile-arcs}
7717 (@pxref{Debugging Options,, Options for Debugging Your Program or
7718 @command{gcc}}), you can compile it a second time using
7719 @option{-fbranch-probabilities}, to improve optimizations based on
7720 the number of times each branch was taken. When the program
7721 compiled with @option{-fprofile-arcs} exits it saves arc execution
7722 counts to a file called @file{@var{sourcename}.gcda} for each source
7723 file. The information in this data file is very dependent on the
7724 structure of the generated code, so you must use the same source code
7725 and the same optimization options for both compilations.
7727 With @option{-fbranch-probabilities}, GCC puts a
7728 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
7729 These can be used to improve optimization. Currently, they are only
7730 used in one place: in @file{reorg.c}, instead of guessing which path a
7731 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
7732 exactly determine which path is taken more often.
7734 @item -fprofile-values
7735 @opindex fprofile-values
7736 If combined with @option{-fprofile-arcs}, it adds code so that some
7737 data about values of expressions in the program is gathered.
7739 With @option{-fbranch-probabilities}, it reads back the data gathered
7740 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
7741 notes to instructions for their later usage in optimizations.
7743 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
7747 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
7748 a code to gather information about values of expressions.
7750 With @option{-fbranch-probabilities}, it reads back the data gathered
7751 and actually performs the optimizations based on them.
7752 Currently the optimizations include specialization of division operation
7753 using the knowledge about the value of the denominator.
7755 @item -frename-registers
7756 @opindex frename-registers
7757 Attempt to avoid false dependencies in scheduled code by making use
7758 of registers left over after register allocation. This optimization
7759 will most benefit processors with lots of registers. Depending on the
7760 debug information format adopted by the target, however, it can
7761 make debugging impossible, since variables will no longer stay in
7762 a ``home register''.
7764 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
7768 Perform tail duplication to enlarge superblock size. This transformation
7769 simplifies the control flow of the function allowing other optimizations to do
7772 Enabled with @option{-fprofile-use}.
7774 @item -funroll-loops
7775 @opindex funroll-loops
7776 Unroll loops whose number of iterations can be determined at compile time or
7777 upon entry to the loop. @option{-funroll-loops} implies
7778 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
7779 It also turns on complete loop peeling (i.e.@: complete removal of loops with
7780 small constant number of iterations). This option makes code larger, and may
7781 or may not make it run faster.
7783 Enabled with @option{-fprofile-use}.
7785 @item -funroll-all-loops
7786 @opindex funroll-all-loops
7787 Unroll all loops, even if their number of iterations is uncertain when
7788 the loop is entered. This usually makes programs run more slowly.
7789 @option{-funroll-all-loops} implies the same options as
7790 @option{-funroll-loops}.
7793 @opindex fpeel-loops
7794 Peels the loops for that there is enough information that they do not
7795 roll much (from profile feedback). It also turns on complete loop peeling
7796 (i.e.@: complete removal of loops with small constant number of iterations).
7798 Enabled with @option{-fprofile-use}.
7800 @item -fmove-loop-invariants
7801 @opindex fmove-loop-invariants
7802 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
7803 at level @option{-O1}
7805 @item -funswitch-loops
7806 @opindex funswitch-loops
7807 Move branches with loop invariant conditions out of the loop, with duplicates
7808 of the loop on both branches (modified according to result of the condition).
7810 @item -ffunction-sections
7811 @itemx -fdata-sections
7812 @opindex ffunction-sections
7813 @opindex fdata-sections
7814 Place each function or data item into its own section in the output
7815 file if the target supports arbitrary sections. The name of the
7816 function or the name of the data item determines the section's name
7819 Use these options on systems where the linker can perform optimizations
7820 to improve locality of reference in the instruction space. Most systems
7821 using the ELF object format and SPARC processors running Solaris 2 have
7822 linkers with such optimizations. AIX may have these optimizations in
7825 Only use these options when there are significant benefits from doing
7826 so. When you specify these options, the assembler and linker will
7827 create larger object and executable files and will also be slower.
7828 You will not be able to use @code{gprof} on all systems if you
7829 specify this option and you may have problems with debugging if
7830 you specify both this option and @option{-g}.
7832 @item -fbranch-target-load-optimize
7833 @opindex fbranch-target-load-optimize
7834 Perform branch target register load optimization before prologue / epilogue
7836 The use of target registers can typically be exposed only during reload,
7837 thus hoisting loads out of loops and doing inter-block scheduling needs
7838 a separate optimization pass.
7840 @item -fbranch-target-load-optimize2
7841 @opindex fbranch-target-load-optimize2
7842 Perform branch target register load optimization after prologue / epilogue
7845 @item -fbtr-bb-exclusive
7846 @opindex fbtr-bb-exclusive
7847 When performing branch target register load optimization, don't reuse
7848 branch target registers in within any basic block.
7850 @item -fstack-protector
7851 @opindex fstack-protector
7852 Emit extra code to check for buffer overflows, such as stack smashing
7853 attacks. This is done by adding a guard variable to functions with
7854 vulnerable objects. This includes functions that call alloca, and
7855 functions with buffers larger than 8 bytes. The guards are initialized
7856 when a function is entered and then checked when the function exits.
7857 If a guard check fails, an error message is printed and the program exits.
7859 @item -fstack-protector-all
7860 @opindex fstack-protector-all
7861 Like @option{-fstack-protector} except that all functions are protected.
7863 @item -fsection-anchors
7864 @opindex fsection-anchors
7865 Try to reduce the number of symbolic address calculations by using
7866 shared ``anchor'' symbols to address nearby objects. This transformation
7867 can help to reduce the number of GOT entries and GOT accesses on some
7870 For example, the implementation of the following function @code{foo}:
7874 int foo (void) @{ return a + b + c; @}
7877 would usually calculate the addresses of all three variables, but if you
7878 compile it with @option{-fsection-anchors}, it will access the variables
7879 from a common anchor point instead. The effect is similar to the
7880 following pseudocode (which isn't valid C):
7885 register int *xr = &x;
7886 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
7890 Not all targets support this option.
7892 @item --param @var{name}=@var{value}
7894 In some places, GCC uses various constants to control the amount of
7895 optimization that is done. For example, GCC will not inline functions
7896 that contain more that a certain number of instructions. You can
7897 control some of these constants on the command-line using the
7898 @option{--param} option.
7900 The names of specific parameters, and the meaning of the values, are
7901 tied to the internals of the compiler, and are subject to change
7902 without notice in future releases.
7904 In each case, the @var{value} is an integer. The allowable choices for
7905 @var{name} are given in the following table:
7908 @item struct-reorg-cold-struct-ratio
7909 The threshold ratio (as a percentage) between a structure frequency
7910 and the frequency of the hottest structure in the program. This parameter
7911 is used by struct-reorg optimization enabled by @option{-fipa-struct-reorg}.
7912 We say that if the ratio of a structure frequency, calculated by profiling,
7913 to the hottest structure frequency in the program is less than this
7914 parameter, then structure reorganization is not applied to this structure.
7917 @item predictable-branch-outcome
7918 When branch is predicted to be taken with probability lower than this threshold
7919 (in percent), then it is considered well predictable. The default is 10.
7921 @item max-crossjump-edges
7922 The maximum number of incoming edges to consider for crossjumping.
7923 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
7924 the number of edges incoming to each block. Increasing values mean
7925 more aggressive optimization, making the compile time increase with
7926 probably small improvement in executable size.
7928 @item min-crossjump-insns
7929 The minimum number of instructions which must be matched at the end
7930 of two blocks before crossjumping will be performed on them. This
7931 value is ignored in the case where all instructions in the block being
7932 crossjumped from are matched. The default value is 5.
7934 @item max-grow-copy-bb-insns
7935 The maximum code size expansion factor when copying basic blocks
7936 instead of jumping. The expansion is relative to a jump instruction.
7937 The default value is 8.
7939 @item max-goto-duplication-insns
7940 The maximum number of instructions to duplicate to a block that jumps
7941 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
7942 passes, GCC factors computed gotos early in the compilation process,
7943 and unfactors them as late as possible. Only computed jumps at the
7944 end of a basic blocks with no more than max-goto-duplication-insns are
7945 unfactored. The default value is 8.
7947 @item max-delay-slot-insn-search
7948 The maximum number of instructions to consider when looking for an
7949 instruction to fill a delay slot. If more than this arbitrary number of
7950 instructions is searched, the time savings from filling the delay slot
7951 will be minimal so stop searching. Increasing values mean more
7952 aggressive optimization, making the compile time increase with probably
7953 small improvement in executable run time.
7955 @item max-delay-slot-live-search
7956 When trying to fill delay slots, the maximum number of instructions to
7957 consider when searching for a block with valid live register
7958 information. Increasing this arbitrarily chosen value means more
7959 aggressive optimization, increasing the compile time. This parameter
7960 should be removed when the delay slot code is rewritten to maintain the
7963 @item max-gcse-memory
7964 The approximate maximum amount of memory that will be allocated in
7965 order to perform the global common subexpression elimination
7966 optimization. If more memory than specified is required, the
7967 optimization will not be done.
7969 @item max-pending-list-length
7970 The maximum number of pending dependencies scheduling will allow
7971 before flushing the current state and starting over. Large functions
7972 with few branches or calls can create excessively large lists which
7973 needlessly consume memory and resources.
7975 @item max-inline-insns-single
7976 Several parameters control the tree inliner used in gcc.
7977 This number sets the maximum number of instructions (counted in GCC's
7978 internal representation) in a single function that the tree inliner
7979 will consider for inlining. This only affects functions declared
7980 inline and methods implemented in a class declaration (C++).
7981 The default value is 300.
7983 @item max-inline-insns-auto
7984 When you use @option{-finline-functions} (included in @option{-O3}),
7985 a lot of functions that would otherwise not be considered for inlining
7986 by the compiler will be investigated. To those functions, a different
7987 (more restrictive) limit compared to functions declared inline can
7989 The default value is 50.
7991 @item large-function-insns
7992 The limit specifying really large functions. For functions larger than this
7993 limit after inlining, inlining is constrained by
7994 @option{--param large-function-growth}. This parameter is useful primarily
7995 to avoid extreme compilation time caused by non-linear algorithms used by the
7997 The default value is 2700.
7999 @item large-function-growth
8000 Specifies maximal growth of large function caused by inlining in percents.
8001 The default value is 100 which limits large function growth to 2.0 times
8004 @item large-unit-insns
8005 The limit specifying large translation unit. Growth caused by inlining of
8006 units larger than this limit is limited by @option{--param inline-unit-growth}.
8007 For small units this might be too tight (consider unit consisting of function A
8008 that is inline and B that just calls A three time. If B is small relative to
8009 A, the growth of unit is 300\% and yet such inlining is very sane. For very
8010 large units consisting of small inlineable functions however the overall unit
8011 growth limit is needed to avoid exponential explosion of code size. Thus for
8012 smaller units, the size is increased to @option{--param large-unit-insns}
8013 before applying @option{--param inline-unit-growth}. The default is 10000
8015 @item inline-unit-growth
8016 Specifies maximal overall growth of the compilation unit caused by inlining.
8017 The default value is 30 which limits unit growth to 1.3 times the original
8020 @item ipcp-unit-growth
8021 Specifies maximal overall growth of the compilation unit caused by
8022 interprocedural constant propagation. The default value is 10 which limits
8023 unit growth to 1.1 times the original size.
8025 @item large-stack-frame
8026 The limit specifying large stack frames. While inlining the algorithm is trying
8027 to not grow past this limit too much. Default value is 256 bytes.
8029 @item large-stack-frame-growth
8030 Specifies maximal growth of large stack frames caused by inlining in percents.
8031 The default value is 1000 which limits large stack frame growth to 11 times
8034 @item max-inline-insns-recursive
8035 @itemx max-inline-insns-recursive-auto
8036 Specifies maximum number of instructions out-of-line copy of self recursive inline
8037 function can grow into by performing recursive inlining.
8039 For functions declared inline @option{--param max-inline-insns-recursive} is
8040 taken into account. For function not declared inline, recursive inlining
8041 happens only when @option{-finline-functions} (included in @option{-O3}) is
8042 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
8043 default value is 450.
8045 @item max-inline-recursive-depth
8046 @itemx max-inline-recursive-depth-auto
8047 Specifies maximum recursion depth used by the recursive inlining.
8049 For functions declared inline @option{--param max-inline-recursive-depth} is
8050 taken into account. For function not declared inline, recursive inlining
8051 happens only when @option{-finline-functions} (included in @option{-O3}) is
8052 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
8055 @item min-inline-recursive-probability
8056 Recursive inlining is profitable only for function having deep recursion
8057 in average and can hurt for function having little recursion depth by
8058 increasing the prologue size or complexity of function body to other
8061 When profile feedback is available (see @option{-fprofile-generate}) the actual
8062 recursion depth can be guessed from probability that function will recurse via
8063 given call expression. This parameter limits inlining only to call expression
8064 whose probability exceeds given threshold (in percents). The default value is
8067 @item early-inlining-insns
8068 Specify growth that early inliner can make. In effect it increases amount of
8069 inlining for code having large abstraction penalty. The default value is 8.
8071 @item max-early-inliner-iterations
8072 @itemx max-early-inliner-iterations
8073 Limit of iterations of early inliner. This basically bounds number of nested
8074 indirect calls early inliner can resolve. Deeper chains are still handled by
8077 @item min-vect-loop-bound
8078 The minimum number of iterations under which a loop will not get vectorized
8079 when @option{-ftree-vectorize} is used. The number of iterations after
8080 vectorization needs to be greater than the value specified by this option
8081 to allow vectorization. The default value is 0.
8083 @item max-unrolled-insns
8084 The maximum number of instructions that a loop should have if that loop
8085 is unrolled, and if the loop is unrolled, it determines how many times
8086 the loop code is unrolled.
8088 @item max-average-unrolled-insns
8089 The maximum number of instructions biased by probabilities of their execution
8090 that a loop should have if that loop is unrolled, and if the loop is unrolled,
8091 it determines how many times the loop code is unrolled.
8093 @item max-unroll-times
8094 The maximum number of unrollings of a single loop.
8096 @item max-peeled-insns
8097 The maximum number of instructions that a loop should have if that loop
8098 is peeled, and if the loop is peeled, it determines how many times
8099 the loop code is peeled.
8101 @item max-peel-times
8102 The maximum number of peelings of a single loop.
8104 @item max-completely-peeled-insns
8105 The maximum number of insns of a completely peeled loop.
8107 @item max-completely-peel-times
8108 The maximum number of iterations of a loop to be suitable for complete peeling.
8110 @item max-completely-peel-loop-nest-depth
8111 The maximum depth of a loop nest suitable for complete peeling.
8113 @item max-unswitch-insns
8114 The maximum number of insns of an unswitched loop.
8116 @item max-unswitch-level
8117 The maximum number of branches unswitched in a single loop.
8120 The minimum cost of an expensive expression in the loop invariant motion.
8122 @item iv-consider-all-candidates-bound
8123 Bound on number of candidates for induction variables below that
8124 all candidates are considered for each use in induction variable
8125 optimizations. Only the most relevant candidates are considered
8126 if there are more candidates, to avoid quadratic time complexity.
8128 @item iv-max-considered-uses
8129 The induction variable optimizations give up on loops that contain more
8130 induction variable uses.
8132 @item iv-always-prune-cand-set-bound
8133 If number of candidates in the set is smaller than this value,
8134 we always try to remove unnecessary ivs from the set during its
8135 optimization when a new iv is added to the set.
8137 @item scev-max-expr-size
8138 Bound on size of expressions used in the scalar evolutions analyzer.
8139 Large expressions slow the analyzer.
8141 @item omega-max-vars
8142 The maximum number of variables in an Omega constraint system.
8143 The default value is 128.
8145 @item omega-max-geqs
8146 The maximum number of inequalities in an Omega constraint system.
8147 The default value is 256.
8150 The maximum number of equalities in an Omega constraint system.
8151 The default value is 128.
8153 @item omega-max-wild-cards
8154 The maximum number of wildcard variables that the Omega solver will
8155 be able to insert. The default value is 18.
8157 @item omega-hash-table-size
8158 The size of the hash table in the Omega solver. The default value is
8161 @item omega-max-keys
8162 The maximal number of keys used by the Omega solver. The default
8165 @item omega-eliminate-redundant-constraints
8166 When set to 1, use expensive methods to eliminate all redundant
8167 constraints. The default value is 0.
8169 @item vect-max-version-for-alignment-checks
8170 The maximum number of runtime checks that can be performed when
8171 doing loop versioning for alignment in the vectorizer. See option
8172 ftree-vect-loop-version for more information.
8174 @item vect-max-version-for-alias-checks
8175 The maximum number of runtime checks that can be performed when
8176 doing loop versioning for alias in the vectorizer. See option
8177 ftree-vect-loop-version for more information.
8179 @item max-iterations-to-track
8181 The maximum number of iterations of a loop the brute force algorithm
8182 for analysis of # of iterations of the loop tries to evaluate.
8184 @item hot-bb-count-fraction
8185 Select fraction of the maximal count of repetitions of basic block in program
8186 given basic block needs to have to be considered hot.
8188 @item hot-bb-frequency-fraction
8189 Select fraction of the maximal frequency of executions of basic block in
8190 function given basic block needs to have to be considered hot
8192 @item max-predicted-iterations
8193 The maximum number of loop iterations we predict statically. This is useful
8194 in cases where function contain single loop with known bound and other loop
8195 with unknown. We predict the known number of iterations correctly, while
8196 the unknown number of iterations average to roughly 10. This means that the
8197 loop without bounds would appear artificially cold relative to the other one.
8199 @item align-threshold
8201 Select fraction of the maximal frequency of executions of basic block in
8202 function given basic block will get aligned.
8204 @item align-loop-iterations
8206 A loop expected to iterate at lest the selected number of iterations will get
8209 @item tracer-dynamic-coverage
8210 @itemx tracer-dynamic-coverage-feedback
8212 This value is used to limit superblock formation once the given percentage of
8213 executed instructions is covered. This limits unnecessary code size
8216 The @option{tracer-dynamic-coverage-feedback} is used only when profile
8217 feedback is available. The real profiles (as opposed to statically estimated
8218 ones) are much less balanced allowing the threshold to be larger value.
8220 @item tracer-max-code-growth
8221 Stop tail duplication once code growth has reached given percentage. This is
8222 rather hokey argument, as most of the duplicates will be eliminated later in
8223 cross jumping, so it may be set to much higher values than is the desired code
8226 @item tracer-min-branch-ratio
8228 Stop reverse growth when the reverse probability of best edge is less than this
8229 threshold (in percent).
8231 @item tracer-min-branch-ratio
8232 @itemx tracer-min-branch-ratio-feedback
8234 Stop forward growth if the best edge do have probability lower than this
8237 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
8238 compilation for profile feedback and one for compilation without. The value
8239 for compilation with profile feedback needs to be more conservative (higher) in
8240 order to make tracer effective.
8242 @item max-cse-path-length
8244 Maximum number of basic blocks on path that cse considers. The default is 10.
8247 The maximum instructions CSE process before flushing. The default is 1000.
8249 @item ggc-min-expand
8251 GCC uses a garbage collector to manage its own memory allocation. This
8252 parameter specifies the minimum percentage by which the garbage
8253 collector's heap should be allowed to expand between collections.
8254 Tuning this may improve compilation speed; it has no effect on code
8257 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
8258 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
8259 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
8260 GCC is not able to calculate RAM on a particular platform, the lower
8261 bound of 30% is used. Setting this parameter and
8262 @option{ggc-min-heapsize} to zero causes a full collection to occur at
8263 every opportunity. This is extremely slow, but can be useful for
8266 @item ggc-min-heapsize
8268 Minimum size of the garbage collector's heap before it begins bothering
8269 to collect garbage. The first collection occurs after the heap expands
8270 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
8271 tuning this may improve compilation speed, and has no effect on code
8274 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
8275 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
8276 with a lower bound of 4096 (four megabytes) and an upper bound of
8277 131072 (128 megabytes). If GCC is not able to calculate RAM on a
8278 particular platform, the lower bound is used. Setting this parameter
8279 very large effectively disables garbage collection. Setting this
8280 parameter and @option{ggc-min-expand} to zero causes a full collection
8281 to occur at every opportunity.
8283 @item max-reload-search-insns
8284 The maximum number of instruction reload should look backward for equivalent
8285 register. Increasing values mean more aggressive optimization, making the
8286 compile time increase with probably slightly better performance. The default
8289 @item max-cselib-memory-locations
8290 The maximum number of memory locations cselib should take into account.
8291 Increasing values mean more aggressive optimization, making the compile time
8292 increase with probably slightly better performance. The default value is 500.
8294 @item reorder-blocks-duplicate
8295 @itemx reorder-blocks-duplicate-feedback
8297 Used by basic block reordering pass to decide whether to use unconditional
8298 branch or duplicate the code on its destination. Code is duplicated when its
8299 estimated size is smaller than this value multiplied by the estimated size of
8300 unconditional jump in the hot spots of the program.
8302 The @option{reorder-block-duplicate-feedback} is used only when profile
8303 feedback is available and may be set to higher values than
8304 @option{reorder-block-duplicate} since information about the hot spots is more
8307 @item max-sched-ready-insns
8308 The maximum number of instructions ready to be issued the scheduler should
8309 consider at any given time during the first scheduling pass. Increasing
8310 values mean more thorough searches, making the compilation time increase
8311 with probably little benefit. The default value is 100.
8313 @item max-sched-region-blocks
8314 The maximum number of blocks in a region to be considered for
8315 interblock scheduling. The default value is 10.
8317 @item max-pipeline-region-blocks
8318 The maximum number of blocks in a region to be considered for
8319 pipelining in the selective scheduler. The default value is 15.
8321 @item max-sched-region-insns
8322 The maximum number of insns in a region to be considered for
8323 interblock scheduling. The default value is 100.
8325 @item max-pipeline-region-insns
8326 The maximum number of insns in a region to be considered for
8327 pipelining in the selective scheduler. The default value is 200.
8330 The minimum probability (in percents) of reaching a source block
8331 for interblock speculative scheduling. The default value is 40.
8333 @item max-sched-extend-regions-iters
8334 The maximum number of iterations through CFG to extend regions.
8335 0 - disable region extension,
8336 N - do at most N iterations.
8337 The default value is 0.
8339 @item max-sched-insn-conflict-delay
8340 The maximum conflict delay for an insn to be considered for speculative motion.
8341 The default value is 3.
8343 @item sched-spec-prob-cutoff
8344 The minimal probability of speculation success (in percents), so that
8345 speculative insn will be scheduled.
8346 The default value is 40.
8348 @item sched-mem-true-dep-cost
8349 Minimal distance (in CPU cycles) between store and load targeting same
8350 memory locations. The default value is 1.
8352 @item selsched-max-lookahead
8353 The maximum size of the lookahead window of selective scheduling. It is a
8354 depth of search for available instructions.
8355 The default value is 50.
8357 @item selsched-max-sched-times
8358 The maximum number of times that an instruction will be scheduled during
8359 selective scheduling. This is the limit on the number of iterations
8360 through which the instruction may be pipelined. The default value is 2.
8362 @item selsched-max-insns-to-rename
8363 The maximum number of best instructions in the ready list that are considered
8364 for renaming in the selective scheduler. The default value is 2.
8366 @item max-last-value-rtl
8367 The maximum size measured as number of RTLs that can be recorded in an expression
8368 in combiner for a pseudo register as last known value of that register. The default
8371 @item integer-share-limit
8372 Small integer constants can use a shared data structure, reducing the
8373 compiler's memory usage and increasing its speed. This sets the maximum
8374 value of a shared integer constant. The default value is 256.
8376 @item min-virtual-mappings
8377 Specifies the minimum number of virtual mappings in the incremental
8378 SSA updater that should be registered to trigger the virtual mappings
8379 heuristic defined by virtual-mappings-ratio. The default value is
8382 @item virtual-mappings-ratio
8383 If the number of virtual mappings is virtual-mappings-ratio bigger
8384 than the number of virtual symbols to be updated, then the incremental
8385 SSA updater switches to a full update for those symbols. The default
8388 @item ssp-buffer-size
8389 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
8390 protection when @option{-fstack-protection} is used.
8392 @item max-jump-thread-duplication-stmts
8393 Maximum number of statements allowed in a block that needs to be
8394 duplicated when threading jumps.
8396 @item max-fields-for-field-sensitive
8397 Maximum number of fields in a structure we will treat in
8398 a field sensitive manner during pointer analysis. The default is zero
8399 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
8401 @item prefetch-latency
8402 Estimate on average number of instructions that are executed before
8403 prefetch finishes. The distance we prefetch ahead is proportional
8404 to this constant. Increasing this number may also lead to less
8405 streams being prefetched (see @option{simultaneous-prefetches}).
8407 @item simultaneous-prefetches
8408 Maximum number of prefetches that can run at the same time.
8410 @item l1-cache-line-size
8411 The size of cache line in L1 cache, in bytes.
8414 The size of L1 cache, in kilobytes.
8417 The size of L2 cache, in kilobytes.
8419 @item min-insn-to-prefetch-ratio
8420 The minimum ratio between the number of instructions and the
8421 number of prefetches to enable prefetching in a loop with an
8424 @item prefetch-min-insn-to-mem-ratio
8425 The minimum ratio between the number of instructions and the
8426 number of memory references to enable prefetching in a loop.
8428 @item use-canonical-types
8429 Whether the compiler should use the ``canonical'' type system. By
8430 default, this should always be 1, which uses a more efficient internal
8431 mechanism for comparing types in C++ and Objective-C++. However, if
8432 bugs in the canonical type system are causing compilation failures,
8433 set this value to 0 to disable canonical types.
8435 @item switch-conversion-max-branch-ratio
8436 Switch initialization conversion will refuse to create arrays that are
8437 bigger than @option{switch-conversion-max-branch-ratio} times the number of
8438 branches in the switch.
8440 @item max-partial-antic-length
8441 Maximum length of the partial antic set computed during the tree
8442 partial redundancy elimination optimization (@option{-ftree-pre}) when
8443 optimizing at @option{-O3} and above. For some sorts of source code
8444 the enhanced partial redundancy elimination optimization can run away,
8445 consuming all of the memory available on the host machine. This
8446 parameter sets a limit on the length of the sets that are computed,
8447 which prevents the runaway behavior. Setting a value of 0 for
8448 this parameter will allow an unlimited set length.
8450 @item sccvn-max-scc-size
8451 Maximum size of a strongly connected component (SCC) during SCCVN
8452 processing. If this limit is hit, SCCVN processing for the whole
8453 function will not be done and optimizations depending on it will
8454 be disabled. The default maximum SCC size is 10000.
8456 @item ira-max-loops-num
8457 IRA uses a regional register allocation by default. If a function
8458 contains loops more than number given by the parameter, only at most
8459 given number of the most frequently executed loops will form regions
8460 for the regional register allocation. The default value of the
8463 @item ira-max-conflict-table-size
8464 Although IRA uses a sophisticated algorithm of compression conflict
8465 table, the table can be still big for huge functions. If the conflict
8466 table for a function could be more than size in MB given by the
8467 parameter, the conflict table is not built and faster, simpler, and
8468 lower quality register allocation algorithm will be used. The
8469 algorithm do not use pseudo-register conflicts. The default value of
8470 the parameter is 2000.
8472 @item ira-loop-reserved-regs
8473 IRA can be used to evaluate more accurate register pressure in loops
8474 for decision to move loop invariants (see @option{-O3}). The number
8475 of available registers reserved for some other purposes is described
8476 by this parameter. The default value of the parameter is 2 which is
8477 minimal number of registers needed for execution of typical
8478 instruction. This value is the best found from numerous experiments.
8480 @item loop-invariant-max-bbs-in-loop
8481 Loop invariant motion can be very expensive, both in compile time and
8482 in amount of needed compile time memory, with very large loops. Loops
8483 with more basic blocks than this parameter won't have loop invariant
8484 motion optimization performed on them. The default value of the
8485 parameter is 1000 for -O1 and 10000 for -O2 and above.
8487 @item max-vartrack-size
8488 Sets a maximum number of hash table slots to use during variable
8489 tracking dataflow analysis of any function. If this limit is exceeded
8490 with variable tracking at assignments enabled, analysis for that
8491 function is retried without it, after removing all debug insns from
8492 the function. If the limit is exceeded even without debug insns, var
8493 tracking analysis is completely disabled for the function. Setting
8494 the parameter to zero makes it unlimited.
8496 @item min-nondebug-insn-uid
8497 Use uids starting at this parameter for nondebug insns. The range below
8498 the parameter is reserved exclusively for debug insns created by
8499 @option{-fvar-tracking-assignments}, but debug insns may get
8500 (non-overlapping) uids above it if the reserved range is exhausted.
8502 @item ipa-sra-ptr-growth-factor
8503 IPA-SRA will replace a pointer to an aggregate with one or more new
8504 parameters only when their cumulative size is less or equal to
8505 @option{ipa-sra-ptr-growth-factor} times the size of the original
8508 @item graphite-max-nb-scop-params
8509 To avoid exponential effects in the Graphite loop transforms, the
8510 number of parameters in a Static Control Part (SCoP) is bounded. The
8511 default value is 10 parameters. A variable whose value is unknown at
8512 compile time and defined outside a SCoP is a parameter of the SCoP.
8514 @item graphite-max-bbs-per-function
8515 To avoid exponential effects in the detection of SCoPs, the size of
8516 the functions analyzed by Graphite is bounded. The default value is
8519 @item loop-block-tile-size
8520 Loop blocking or strip mining transforms, enabled with
8521 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
8522 loop in the loop nest by a given number of iterations. The strip
8523 length can be changed using the @option{loop-block-tile-size}
8524 parameter. The default value is 51 iterations.
8529 @node Preprocessor Options
8530 @section Options Controlling the Preprocessor
8531 @cindex preprocessor options
8532 @cindex options, preprocessor
8534 These options control the C preprocessor, which is run on each C source
8535 file before actual compilation.
8537 If you use the @option{-E} option, nothing is done except preprocessing.
8538 Some of these options make sense only together with @option{-E} because
8539 they cause the preprocessor output to be unsuitable for actual
8543 @item -Wp,@var{option}
8545 You can use @option{-Wp,@var{option}} to bypass the compiler driver
8546 and pass @var{option} directly through to the preprocessor. If
8547 @var{option} contains commas, it is split into multiple options at the
8548 commas. However, many options are modified, translated or interpreted
8549 by the compiler driver before being passed to the preprocessor, and
8550 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
8551 interface is undocumented and subject to change, so whenever possible
8552 you should avoid using @option{-Wp} and let the driver handle the
8555 @item -Xpreprocessor @var{option}
8556 @opindex Xpreprocessor
8557 Pass @var{option} as an option to the preprocessor. You can use this to
8558 supply system-specific preprocessor options which GCC does not know how to
8561 If you want to pass an option that takes an argument, you must use
8562 @option{-Xpreprocessor} twice, once for the option and once for the argument.
8565 @include cppopts.texi
8567 @node Assembler Options
8568 @section Passing Options to the Assembler
8570 @c prevent bad page break with this line
8571 You can pass options to the assembler.
8574 @item -Wa,@var{option}
8576 Pass @var{option} as an option to the assembler. If @var{option}
8577 contains commas, it is split into multiple options at the commas.
8579 @item -Xassembler @var{option}
8581 Pass @var{option} as an option to the assembler. You can use this to
8582 supply system-specific assembler options which GCC does not know how to
8585 If you want to pass an option that takes an argument, you must use
8586 @option{-Xassembler} twice, once for the option and once for the argument.
8591 @section Options for Linking
8592 @cindex link options
8593 @cindex options, linking
8595 These options come into play when the compiler links object files into
8596 an executable output file. They are meaningless if the compiler is
8597 not doing a link step.
8601 @item @var{object-file-name}
8602 A file name that does not end in a special recognized suffix is
8603 considered to name an object file or library. (Object files are
8604 distinguished from libraries by the linker according to the file
8605 contents.) If linking is done, these object files are used as input
8614 If any of these options is used, then the linker is not run, and
8615 object file names should not be used as arguments. @xref{Overall
8619 @item -l@var{library}
8620 @itemx -l @var{library}
8622 Search the library named @var{library} when linking. (The second
8623 alternative with the library as a separate argument is only for
8624 POSIX compliance and is not recommended.)
8626 It makes a difference where in the command you write this option; the
8627 linker searches and processes libraries and object files in the order they
8628 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
8629 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
8630 to functions in @samp{z}, those functions may not be loaded.
8632 The linker searches a standard list of directories for the library,
8633 which is actually a file named @file{lib@var{library}.a}. The linker
8634 then uses this file as if it had been specified precisely by name.
8636 The directories searched include several standard system directories
8637 plus any that you specify with @option{-L}.
8639 Normally the files found this way are library files---archive files
8640 whose members are object files. The linker handles an archive file by
8641 scanning through it for members which define symbols that have so far
8642 been referenced but not defined. But if the file that is found is an
8643 ordinary object file, it is linked in the usual fashion. The only
8644 difference between using an @option{-l} option and specifying a file name
8645 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
8646 and searches several directories.
8650 You need this special case of the @option{-l} option in order to
8651 link an Objective-C or Objective-C++ program.
8654 @opindex nostartfiles
8655 Do not use the standard system startup files when linking.
8656 The standard system libraries are used normally, unless @option{-nostdlib}
8657 or @option{-nodefaultlibs} is used.
8659 @item -nodefaultlibs
8660 @opindex nodefaultlibs
8661 Do not use the standard system libraries when linking.
8662 Only the libraries you specify will be passed to the linker, options
8663 specifying linkage of the system libraries, such as @code{-static-libgcc}
8664 or @code{-shared-libgcc}, will be ignored.
8665 The standard startup files are used normally, unless @option{-nostartfiles}
8666 is used. The compiler may generate calls to @code{memcmp},
8667 @code{memset}, @code{memcpy} and @code{memmove}.
8668 These entries are usually resolved by entries in
8669 libc. These entry points should be supplied through some other
8670 mechanism when this option is specified.
8674 Do not use the standard system startup files or libraries when linking.
8675 No startup files and only the libraries you specify will be passed to
8676 the linker, options specifying linkage of the system libraries, such as
8677 @code{-static-libgcc} or @code{-shared-libgcc}, will be ignored.
8678 The compiler may generate calls to @code{memcmp}, @code{memset},
8679 @code{memcpy} and @code{memmove}.
8680 These entries are usually resolved by entries in
8681 libc. These entry points should be supplied through some other
8682 mechanism when this option is specified.
8684 @cindex @option{-lgcc}, use with @option{-nostdlib}
8685 @cindex @option{-nostdlib} and unresolved references
8686 @cindex unresolved references and @option{-nostdlib}
8687 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
8688 @cindex @option{-nodefaultlibs} and unresolved references
8689 @cindex unresolved references and @option{-nodefaultlibs}
8690 One of the standard libraries bypassed by @option{-nostdlib} and
8691 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
8692 that GCC uses to overcome shortcomings of particular machines, or special
8693 needs for some languages.
8694 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
8695 Collection (GCC) Internals},
8696 for more discussion of @file{libgcc.a}.)
8697 In most cases, you need @file{libgcc.a} even when you want to avoid
8698 other standard libraries. In other words, when you specify @option{-nostdlib}
8699 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
8700 This ensures that you have no unresolved references to internal GCC
8701 library subroutines. (For example, @samp{__main}, used to ensure C++
8702 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
8703 GNU Compiler Collection (GCC) Internals}.)
8707 Produce a position independent executable on targets which support it.
8708 For predictable results, you must also specify the same set of options
8709 that were used to generate code (@option{-fpie}, @option{-fPIE},
8710 or model suboptions) when you specify this option.
8714 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
8715 that support it. This instructs the linker to add all symbols, not
8716 only used ones, to the dynamic symbol table. This option is needed
8717 for some uses of @code{dlopen} or to allow obtaining backtraces
8718 from within a program.
8722 Remove all symbol table and relocation information from the executable.
8726 On systems that support dynamic linking, this prevents linking with the shared
8727 libraries. On other systems, this option has no effect.
8731 Produce a shared object which can then be linked with other objects to
8732 form an executable. Not all systems support this option. For predictable
8733 results, you must also specify the same set of options that were used to
8734 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
8735 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
8736 needs to build supplementary stub code for constructors to work. On
8737 multi-libbed systems, @samp{gcc -shared} must select the correct support
8738 libraries to link against. Failing to supply the correct flags may lead
8739 to subtle defects. Supplying them in cases where they are not necessary
8742 @item -shared-libgcc
8743 @itemx -static-libgcc
8744 @opindex shared-libgcc
8745 @opindex static-libgcc
8746 On systems that provide @file{libgcc} as a shared library, these options
8747 force the use of either the shared or static version respectively.
8748 If no shared version of @file{libgcc} was built when the compiler was
8749 configured, these options have no effect.
8751 There are several situations in which an application should use the
8752 shared @file{libgcc} instead of the static version. The most common
8753 of these is when the application wishes to throw and catch exceptions
8754 across different shared libraries. In that case, each of the libraries
8755 as well as the application itself should use the shared @file{libgcc}.
8757 Therefore, the G++ and GCJ drivers automatically add
8758 @option{-shared-libgcc} whenever you build a shared library or a main
8759 executable, because C++ and Java programs typically use exceptions, so
8760 this is the right thing to do.
8762 If, instead, you use the GCC driver to create shared libraries, you may
8763 find that they will not always be linked with the shared @file{libgcc}.
8764 If GCC finds, at its configuration time, that you have a non-GNU linker
8765 or a GNU linker that does not support option @option{--eh-frame-hdr},
8766 it will link the shared version of @file{libgcc} into shared libraries
8767 by default. Otherwise, it will take advantage of the linker and optimize
8768 away the linking with the shared version of @file{libgcc}, linking with
8769 the static version of libgcc by default. This allows exceptions to
8770 propagate through such shared libraries, without incurring relocation
8771 costs at library load time.
8773 However, if a library or main executable is supposed to throw or catch
8774 exceptions, you must link it using the G++ or GCJ driver, as appropriate
8775 for the languages used in the program, or using the option
8776 @option{-shared-libgcc}, such that it is linked with the shared
8779 @item -static-libstdc++
8780 When the @command{g++} program is used to link a C++ program, it will
8781 normally automatically link against @option{libstdc++}. If
8782 @file{libstdc++} is available as a shared library, and the
8783 @option{-static} option is not used, then this will link against the
8784 shared version of @file{libstdc++}. That is normally fine. However, it
8785 is sometimes useful to freeze the version of @file{libstdc++} used by
8786 the program without going all the way to a fully static link. The
8787 @option{-static-libstdc++} option directs the @command{g++} driver to
8788 link @file{libstdc++} statically, without necessarily linking other
8789 libraries statically.
8793 Bind references to global symbols when building a shared object. Warn
8794 about any unresolved references (unless overridden by the link editor
8795 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
8798 @item -T @var{script}
8800 @cindex linker script
8801 Use @var{script} as the linker script. This option is supported by most
8802 systems using the GNU linker. On some targets, such as bare-board
8803 targets without an operating system, the @option{-T} option may be required
8804 when linking to avoid references to undefined symbols.
8806 @item -Xlinker @var{option}
8808 Pass @var{option} as an option to the linker. You can use this to
8809 supply system-specific linker options which GCC does not know how to
8812 If you want to pass an option that takes a separate argument, you must use
8813 @option{-Xlinker} twice, once for the option and once for the argument.
8814 For example, to pass @option{-assert definitions}, you must write
8815 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
8816 @option{-Xlinker "-assert definitions"}, because this passes the entire
8817 string as a single argument, which is not what the linker expects.
8819 When using the GNU linker, it is usually more convenient to pass
8820 arguments to linker options using the @option{@var{option}=@var{value}}
8821 syntax than as separate arguments. For example, you can specify
8822 @samp{-Xlinker -Map=output.map} rather than
8823 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
8824 this syntax for command-line options.
8826 @item -Wl,@var{option}
8828 Pass @var{option} as an option to the linker. If @var{option} contains
8829 commas, it is split into multiple options at the commas. You can use this
8830 syntax to pass an argument to the option.
8831 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
8832 linker. When using the GNU linker, you can also get the same effect with
8833 @samp{-Wl,-Map=output.map}.
8835 @item -u @var{symbol}
8837 Pretend the symbol @var{symbol} is undefined, to force linking of
8838 library modules to define it. You can use @option{-u} multiple times with
8839 different symbols to force loading of additional library modules.
8842 @node Directory Options
8843 @section Options for Directory Search
8844 @cindex directory options
8845 @cindex options, directory search
8848 These options specify directories to search for header files, for
8849 libraries and for parts of the compiler:
8854 Add the directory @var{dir} to the head of the list of directories to be
8855 searched for header files. This can be used to override a system header
8856 file, substituting your own version, since these directories are
8857 searched before the system header file directories. However, you should
8858 not use this option to add directories that contain vendor-supplied
8859 system header files (use @option{-isystem} for that). If you use more than
8860 one @option{-I} option, the directories are scanned in left-to-right
8861 order; the standard system directories come after.
8863 If a standard system include directory, or a directory specified with
8864 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
8865 option will be ignored. The directory will still be searched but as a
8866 system directory at its normal position in the system include chain.
8867 This is to ensure that GCC's procedure to fix buggy system headers and
8868 the ordering for the include_next directive are not inadvertently changed.
8869 If you really need to change the search order for system directories,
8870 use the @option{-nostdinc} and/or @option{-isystem} options.
8872 @item -iquote@var{dir}
8874 Add the directory @var{dir} to the head of the list of directories to
8875 be searched for header files only for the case of @samp{#include
8876 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
8877 otherwise just like @option{-I}.
8881 Add directory @var{dir} to the list of directories to be searched
8884 @item -B@var{prefix}
8886 This option specifies where to find the executables, libraries,
8887 include files, and data files of the compiler itself.
8889 The compiler driver program runs one or more of the subprograms
8890 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
8891 @var{prefix} as a prefix for each program it tries to run, both with and
8892 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
8894 For each subprogram to be run, the compiler driver first tries the
8895 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
8896 was not specified, the driver tries two standard prefixes, which are
8897 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
8898 those results in a file name that is found, the unmodified program
8899 name is searched for using the directories specified in your
8900 @env{PATH} environment variable.
8902 The compiler will check to see if the path provided by the @option{-B}
8903 refers to a directory, and if necessary it will add a directory
8904 separator character at the end of the path.
8906 @option{-B} prefixes that effectively specify directory names also apply
8907 to libraries in the linker, because the compiler translates these
8908 options into @option{-L} options for the linker. They also apply to
8909 includes files in the preprocessor, because the compiler translates these
8910 options into @option{-isystem} options for the preprocessor. In this case,
8911 the compiler appends @samp{include} to the prefix.
8913 The run-time support file @file{libgcc.a} can also be searched for using
8914 the @option{-B} prefix, if needed. If it is not found there, the two
8915 standard prefixes above are tried, and that is all. The file is left
8916 out of the link if it is not found by those means.
8918 Another way to specify a prefix much like the @option{-B} prefix is to use
8919 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
8922 As a special kludge, if the path provided by @option{-B} is
8923 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
8924 9, then it will be replaced by @file{[dir/]include}. This is to help
8925 with boot-strapping the compiler.
8927 @item -specs=@var{file}
8929 Process @var{file} after the compiler reads in the standard @file{specs}
8930 file, in order to override the defaults that the @file{gcc} driver
8931 program uses when determining what switches to pass to @file{cc1},
8932 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
8933 @option{-specs=@var{file}} can be specified on the command line, and they
8934 are processed in order, from left to right.
8936 @item --sysroot=@var{dir}
8938 Use @var{dir} as the logical root directory for headers and libraries.
8939 For example, if the compiler would normally search for headers in
8940 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
8941 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
8943 If you use both this option and the @option{-isysroot} option, then
8944 the @option{--sysroot} option will apply to libraries, but the
8945 @option{-isysroot} option will apply to header files.
8947 The GNU linker (beginning with version 2.16) has the necessary support
8948 for this option. If your linker does not support this option, the
8949 header file aspect of @option{--sysroot} will still work, but the
8950 library aspect will not.
8954 This option has been deprecated. Please use @option{-iquote} instead for
8955 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
8956 Any directories you specify with @option{-I} options before the @option{-I-}
8957 option are searched only for the case of @samp{#include "@var{file}"};
8958 they are not searched for @samp{#include <@var{file}>}.
8960 If additional directories are specified with @option{-I} options after
8961 the @option{-I-}, these directories are searched for all @samp{#include}
8962 directives. (Ordinarily @emph{all} @option{-I} directories are used
8965 In addition, the @option{-I-} option inhibits the use of the current
8966 directory (where the current input file came from) as the first search
8967 directory for @samp{#include "@var{file}"}. There is no way to
8968 override this effect of @option{-I-}. With @option{-I.} you can specify
8969 searching the directory which was current when the compiler was
8970 invoked. That is not exactly the same as what the preprocessor does
8971 by default, but it is often satisfactory.
8973 @option{-I-} does not inhibit the use of the standard system directories
8974 for header files. Thus, @option{-I-} and @option{-nostdinc} are
8981 @section Specifying subprocesses and the switches to pass to them
8984 @command{gcc} is a driver program. It performs its job by invoking a
8985 sequence of other programs to do the work of compiling, assembling and
8986 linking. GCC interprets its command-line parameters and uses these to
8987 deduce which programs it should invoke, and which command-line options
8988 it ought to place on their command lines. This behavior is controlled
8989 by @dfn{spec strings}. In most cases there is one spec string for each
8990 program that GCC can invoke, but a few programs have multiple spec
8991 strings to control their behavior. The spec strings built into GCC can
8992 be overridden by using the @option{-specs=} command-line switch to specify
8995 @dfn{Spec files} are plaintext files that are used to construct spec
8996 strings. They consist of a sequence of directives separated by blank
8997 lines. The type of directive is determined by the first non-whitespace
8998 character on the line and it can be one of the following:
9001 @item %@var{command}
9002 Issues a @var{command} to the spec file processor. The commands that can
9006 @item %include <@var{file}>
9008 Search for @var{file} and insert its text at the current point in the
9011 @item %include_noerr <@var{file}>
9012 @cindex %include_noerr
9013 Just like @samp{%include}, but do not generate an error message if the include
9014 file cannot be found.
9016 @item %rename @var{old_name} @var{new_name}
9018 Rename the spec string @var{old_name} to @var{new_name}.
9022 @item *[@var{spec_name}]:
9023 This tells the compiler to create, override or delete the named spec
9024 string. All lines after this directive up to the next directive or
9025 blank line are considered to be the text for the spec string. If this
9026 results in an empty string then the spec will be deleted. (Or, if the
9027 spec did not exist, then nothing will happened.) Otherwise, if the spec
9028 does not currently exist a new spec will be created. If the spec does
9029 exist then its contents will be overridden by the text of this
9030 directive, unless the first character of that text is the @samp{+}
9031 character, in which case the text will be appended to the spec.
9033 @item [@var{suffix}]:
9034 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
9035 and up to the next directive or blank line are considered to make up the
9036 spec string for the indicated suffix. When the compiler encounters an
9037 input file with the named suffix, it will processes the spec string in
9038 order to work out how to compile that file. For example:
9045 This says that any input file whose name ends in @samp{.ZZ} should be
9046 passed to the program @samp{z-compile}, which should be invoked with the
9047 command-line switch @option{-input} and with the result of performing the
9048 @samp{%i} substitution. (See below.)
9050 As an alternative to providing a spec string, the text that follows a
9051 suffix directive can be one of the following:
9054 @item @@@var{language}
9055 This says that the suffix is an alias for a known @var{language}. This is
9056 similar to using the @option{-x} command-line switch to GCC to specify a
9057 language explicitly. For example:
9064 Says that .ZZ files are, in fact, C++ source files.
9067 This causes an error messages saying:
9070 @var{name} compiler not installed on this system.
9074 GCC already has an extensive list of suffixes built into it.
9075 This directive will add an entry to the end of the list of suffixes, but
9076 since the list is searched from the end backwards, it is effectively
9077 possible to override earlier entries using this technique.
9081 GCC has the following spec strings built into it. Spec files can
9082 override these strings or create their own. Note that individual
9083 targets can also add their own spec strings to this list.
9086 asm Options to pass to the assembler
9087 asm_final Options to pass to the assembler post-processor
9088 cpp Options to pass to the C preprocessor
9089 cc1 Options to pass to the C compiler
9090 cc1plus Options to pass to the C++ compiler
9091 endfile Object files to include at the end of the link
9092 link Options to pass to the linker
9093 lib Libraries to include on the command line to the linker
9094 libgcc Decides which GCC support library to pass to the linker
9095 linker Sets the name of the linker
9096 predefines Defines to be passed to the C preprocessor
9097 signed_char Defines to pass to CPP to say whether @code{char} is signed
9099 startfile Object files to include at the start of the link
9102 Here is a small example of a spec file:
9108 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
9111 This example renames the spec called @samp{lib} to @samp{old_lib} and
9112 then overrides the previous definition of @samp{lib} with a new one.
9113 The new definition adds in some extra command-line options before
9114 including the text of the old definition.
9116 @dfn{Spec strings} are a list of command-line options to be passed to their
9117 corresponding program. In addition, the spec strings can contain
9118 @samp{%}-prefixed sequences to substitute variable text or to
9119 conditionally insert text into the command line. Using these constructs
9120 it is possible to generate quite complex command lines.
9122 Here is a table of all defined @samp{%}-sequences for spec
9123 strings. Note that spaces are not generated automatically around the
9124 results of expanding these sequences. Therefore you can concatenate them
9125 together or combine them with constant text in a single argument.
9129 Substitute one @samp{%} into the program name or argument.
9132 Substitute the name of the input file being processed.
9135 Substitute the basename of the input file being processed.
9136 This is the substring up to (and not including) the last period
9137 and not including the directory.
9140 This is the same as @samp{%b}, but include the file suffix (text after
9144 Marks the argument containing or following the @samp{%d} as a
9145 temporary file name, so that that file will be deleted if GCC exits
9146 successfully. Unlike @samp{%g}, this contributes no text to the
9149 @item %g@var{suffix}
9150 Substitute a file name that has suffix @var{suffix} and is chosen
9151 once per compilation, and mark the argument in the same way as
9152 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
9153 name is now chosen in a way that is hard to predict even when previously
9154 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
9155 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
9156 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
9157 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
9158 was simply substituted with a file name chosen once per compilation,
9159 without regard to any appended suffix (which was therefore treated
9160 just like ordinary text), making such attacks more likely to succeed.
9162 @item %u@var{suffix}
9163 Like @samp{%g}, but generates a new temporary file name even if
9164 @samp{%u@var{suffix}} was already seen.
9166 @item %U@var{suffix}
9167 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
9168 new one if there is no such last file name. In the absence of any
9169 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
9170 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
9171 would involve the generation of two distinct file names, one
9172 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
9173 simply substituted with a file name chosen for the previous @samp{%u},
9174 without regard to any appended suffix.
9176 @item %j@var{suffix}
9177 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
9178 writable, and if save-temps is off; otherwise, substitute the name
9179 of a temporary file, just like @samp{%u}. This temporary file is not
9180 meant for communication between processes, but rather as a junk
9183 @item %|@var{suffix}
9184 @itemx %m@var{suffix}
9185 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
9186 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
9187 all. These are the two most common ways to instruct a program that it
9188 should read from standard input or write to standard output. If you
9189 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
9190 construct: see for example @file{f/lang-specs.h}.
9192 @item %.@var{SUFFIX}
9193 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
9194 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
9195 terminated by the next space or %.
9198 Marks the argument containing or following the @samp{%w} as the
9199 designated output file of this compilation. This puts the argument
9200 into the sequence of arguments that @samp{%o} will substitute later.
9203 Substitutes the names of all the output files, with spaces
9204 automatically placed around them. You should write spaces
9205 around the @samp{%o} as well or the results are undefined.
9206 @samp{%o} is for use in the specs for running the linker.
9207 Input files whose names have no recognized suffix are not compiled
9208 at all, but they are included among the output files, so they will
9212 Substitutes the suffix for object files. Note that this is
9213 handled specially when it immediately follows @samp{%g, %u, or %U},
9214 because of the need for those to form complete file names. The
9215 handling is such that @samp{%O} is treated exactly as if it had already
9216 been substituted, except that @samp{%g, %u, and %U} do not currently
9217 support additional @var{suffix} characters following @samp{%O} as they would
9218 following, for example, @samp{.o}.
9221 Substitutes the standard macro predefinitions for the
9222 current target machine. Use this when running @code{cpp}.
9225 Like @samp{%p}, but puts @samp{__} before and after the name of each
9226 predefined macro, except for macros that start with @samp{__} or with
9227 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
9231 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
9232 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
9233 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
9234 and @option{-imultilib} as necessary.
9237 Current argument is the name of a library or startup file of some sort.
9238 Search for that file in a standard list of directories and substitute
9239 the full name found. The current working directory is included in the
9240 list of directories scanned.
9243 Current argument is the name of a linker script. Search for that file
9244 in the current list of directories to scan for libraries. If the file
9245 is located insert a @option{--script} option into the command line
9246 followed by the full path name found. If the file is not found then
9247 generate an error message. Note: the current working directory is not
9251 Print @var{str} as an error message. @var{str} is terminated by a newline.
9252 Use this when inconsistent options are detected.
9255 Substitute the contents of spec string @var{name} at this point.
9258 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
9260 @item %x@{@var{option}@}
9261 Accumulate an option for @samp{%X}.
9264 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
9268 Output the accumulated assembler options specified by @option{-Wa}.
9271 Output the accumulated preprocessor options specified by @option{-Wp}.
9274 Process the @code{asm} spec. This is used to compute the
9275 switches to be passed to the assembler.
9278 Process the @code{asm_final} spec. This is a spec string for
9279 passing switches to an assembler post-processor, if such a program is
9283 Process the @code{link} spec. This is the spec for computing the
9284 command line passed to the linker. Typically it will make use of the
9285 @samp{%L %G %S %D and %E} sequences.
9288 Dump out a @option{-L} option for each directory that GCC believes might
9289 contain startup files. If the target supports multilibs then the
9290 current multilib directory will be prepended to each of these paths.
9293 Process the @code{lib} spec. This is a spec string for deciding which
9294 libraries should be included on the command line to the linker.
9297 Process the @code{libgcc} spec. This is a spec string for deciding
9298 which GCC support library should be included on the command line to the linker.
9301 Process the @code{startfile} spec. This is a spec for deciding which
9302 object files should be the first ones passed to the linker. Typically
9303 this might be a file named @file{crt0.o}.
9306 Process the @code{endfile} spec. This is a spec string that specifies
9307 the last object files that will be passed to the linker.
9310 Process the @code{cpp} spec. This is used to construct the arguments
9311 to be passed to the C preprocessor.
9314 Process the @code{cc1} spec. This is used to construct the options to be
9315 passed to the actual C compiler (@samp{cc1}).
9318 Process the @code{cc1plus} spec. This is used to construct the options to be
9319 passed to the actual C++ compiler (@samp{cc1plus}).
9322 Substitute the variable part of a matched option. See below.
9323 Note that each comma in the substituted string is replaced by
9327 Remove all occurrences of @code{-S} from the command line. Note---this
9328 command is position dependent. @samp{%} commands in the spec string
9329 before this one will see @code{-S}, @samp{%} commands in the spec string
9330 after this one will not.
9332 @item %:@var{function}(@var{args})
9333 Call the named function @var{function}, passing it @var{args}.
9334 @var{args} is first processed as a nested spec string, then split
9335 into an argument vector in the usual fashion. The function returns
9336 a string which is processed as if it had appeared literally as part
9337 of the current spec.
9339 The following built-in spec functions are provided:
9343 The @code{getenv} spec function takes two arguments: an environment
9344 variable name and a string. If the environment variable is not
9345 defined, a fatal error is issued. Otherwise, the return value is the
9346 value of the environment variable concatenated with the string. For
9347 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
9350 %:getenv(TOPDIR /include)
9353 expands to @file{/path/to/top/include}.
9355 @item @code{if-exists}
9356 The @code{if-exists} spec function takes one argument, an absolute
9357 pathname to a file. If the file exists, @code{if-exists} returns the
9358 pathname. Here is a small example of its usage:
9362 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
9365 @item @code{if-exists-else}
9366 The @code{if-exists-else} spec function is similar to the @code{if-exists}
9367 spec function, except that it takes two arguments. The first argument is
9368 an absolute pathname to a file. If the file exists, @code{if-exists-else}
9369 returns the pathname. If it does not exist, it returns the second argument.
9370 This way, @code{if-exists-else} can be used to select one file or another,
9371 based on the existence of the first. Here is a small example of its usage:
9375 crt0%O%s %:if-exists(crti%O%s) \
9376 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
9379 @item @code{replace-outfile}
9380 The @code{replace-outfile} spec function takes two arguments. It looks for the
9381 first argument in the outfiles array and replaces it with the second argument. Here
9382 is a small example of its usage:
9385 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
9388 @item @code{print-asm-header}
9389 The @code{print-asm-header} function takes no arguments and simply
9390 prints a banner like:
9396 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
9399 It is used to separate compiler options from assembler options
9400 in the @option{--target-help} output.
9404 Substitutes the @code{-S} switch, if that switch was given to GCC@.
9405 If that switch was not specified, this substitutes nothing. Note that
9406 the leading dash is omitted when specifying this option, and it is
9407 automatically inserted if the substitution is performed. Thus the spec
9408 string @samp{%@{foo@}} would match the command-line option @option{-foo}
9409 and would output the command line option @option{-foo}.
9411 @item %W@{@code{S}@}
9412 Like %@{@code{S}@} but mark last argument supplied within as a file to be
9415 @item %@{@code{S}*@}
9416 Substitutes all the switches specified to GCC whose names start
9417 with @code{-S}, but which also take an argument. This is used for
9418 switches like @option{-o}, @option{-D}, @option{-I}, etc.
9419 GCC considers @option{-o foo} as being
9420 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
9421 text, including the space. Thus two arguments would be generated.
9423 @item %@{@code{S}*&@code{T}*@}
9424 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
9425 (the order of @code{S} and @code{T} in the spec is not significant).
9426 There can be any number of ampersand-separated variables; for each the
9427 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
9429 @item %@{@code{S}:@code{X}@}
9430 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
9432 @item %@{!@code{S}:@code{X}@}
9433 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
9435 @item %@{@code{S}*:@code{X}@}
9436 Substitutes @code{X} if one or more switches whose names start with
9437 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
9438 once, no matter how many such switches appeared. However, if @code{%*}
9439 appears somewhere in @code{X}, then @code{X} will be substituted once
9440 for each matching switch, with the @code{%*} replaced by the part of
9441 that switch that matched the @code{*}.
9443 @item %@{.@code{S}:@code{X}@}
9444 Substitutes @code{X}, if processing a file with suffix @code{S}.
9446 @item %@{!.@code{S}:@code{X}@}
9447 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
9449 @item %@{,@code{S}:@code{X}@}
9450 Substitutes @code{X}, if processing a file for language @code{S}.
9452 @item %@{!,@code{S}:@code{X}@}
9453 Substitutes @code{X}, if not processing a file for language @code{S}.
9455 @item %@{@code{S}|@code{P}:@code{X}@}
9456 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
9457 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
9458 @code{*} sequences as well, although they have a stronger binding than
9459 the @samp{|}. If @code{%*} appears in @code{X}, all of the
9460 alternatives must be starred, and only the first matching alternative
9463 For example, a spec string like this:
9466 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
9469 will output the following command-line options from the following input
9470 command-line options:
9475 -d fred.c -foo -baz -boggle
9476 -d jim.d -bar -baz -boggle
9479 @item %@{S:X; T:Y; :D@}
9481 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
9482 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
9483 be as many clauses as you need. This may be combined with @code{.},
9484 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
9489 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
9490 construct may contain other nested @samp{%} constructs or spaces, or
9491 even newlines. They are processed as usual, as described above.
9492 Trailing white space in @code{X} is ignored. White space may also
9493 appear anywhere on the left side of the colon in these constructs,
9494 except between @code{.} or @code{*} and the corresponding word.
9496 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
9497 handled specifically in these constructs. If another value of
9498 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
9499 @option{-W} switch is found later in the command line, the earlier
9500 switch value is ignored, except with @{@code{S}*@} where @code{S} is
9501 just one letter, which passes all matching options.
9503 The character @samp{|} at the beginning of the predicate text is used to
9504 indicate that a command should be piped to the following command, but
9505 only if @option{-pipe} is specified.
9507 It is built into GCC which switches take arguments and which do not.
9508 (You might think it would be useful to generalize this to allow each
9509 compiler's spec to say which switches take arguments. But this cannot
9510 be done in a consistent fashion. GCC cannot even decide which input
9511 files have been specified without knowing which switches take arguments,
9512 and it must know which input files to compile in order to tell which
9515 GCC also knows implicitly that arguments starting in @option{-l} are to be
9516 treated as compiler output files, and passed to the linker in their
9517 proper position among the other output files.
9519 @c man begin OPTIONS
9521 @node Target Options
9522 @section Specifying Target Machine and Compiler Version
9523 @cindex target options
9524 @cindex cross compiling
9525 @cindex specifying machine version
9526 @cindex specifying compiler version and target machine
9527 @cindex compiler version, specifying
9528 @cindex target machine, specifying
9530 The usual way to run GCC is to run the executable called @file{gcc}, or
9531 @file{<machine>-gcc} when cross-compiling, or
9532 @file{<machine>-gcc-<version>} to run a version other than the one that
9533 was installed last. Sometimes this is inconvenient, so GCC provides
9534 options that will switch to another cross-compiler or version.
9537 @item -b @var{machine}
9539 The argument @var{machine} specifies the target machine for compilation.
9541 The value to use for @var{machine} is the same as was specified as the
9542 machine type when configuring GCC as a cross-compiler. For
9543 example, if a cross-compiler was configured with @samp{configure
9544 arm-elf}, meaning to compile for an arm processor with elf binaries,
9545 then you would specify @option{-b arm-elf} to run that cross compiler.
9546 Because there are other options beginning with @option{-b}, the
9547 configuration must contain a hyphen, or @option{-b} alone should be one
9548 argument followed by the configuration in the next argument.
9550 @item -V @var{version}
9552 The argument @var{version} specifies which version of GCC to run.
9553 This is useful when multiple versions are installed. For example,
9554 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
9557 The @option{-V} and @option{-b} options work by running the
9558 @file{<machine>-gcc-<version>} executable, so there's no real reason to
9559 use them if you can just run that directly.
9561 @node Submodel Options
9562 @section Hardware Models and Configurations
9563 @cindex submodel options
9564 @cindex specifying hardware config
9565 @cindex hardware models and configurations, specifying
9566 @cindex machine dependent options
9568 Earlier we discussed the standard option @option{-b} which chooses among
9569 different installed compilers for completely different target
9570 machines, such as VAX vs.@: 68000 vs.@: 80386.
9572 In addition, each of these target machine types can have its own
9573 special options, starting with @samp{-m}, to choose among various
9574 hardware models or configurations---for example, 68010 vs 68020,
9575 floating coprocessor or none. A single installed version of the
9576 compiler can compile for any model or configuration, according to the
9579 Some configurations of the compiler also support additional special
9580 options, usually for compatibility with other compilers on the same
9583 @c This list is ordered alphanumerically by subsection name.
9584 @c It should be the same order and spelling as these options are listed
9585 @c in Machine Dependent Options
9591 * Blackfin Options::
9595 * DEC Alpha Options::
9596 * DEC Alpha/VMS Options::
9599 * GNU/Linux Options::
9602 * i386 and x86-64 Options::
9603 * i386 and x86-64 Windows Options::
9605 * IA-64/VMS Options::
9617 * picoChip Options::
9619 * RS/6000 and PowerPC Options::
9621 * S/390 and zSeries Options::
9626 * System V Options::
9631 * Xstormy16 Options::
9637 @subsection ARC Options
9640 These options are defined for ARC implementations:
9645 Compile code for little endian mode. This is the default.
9649 Compile code for big endian mode.
9652 @opindex mmangle-cpu
9653 Prepend the name of the cpu to all public symbol names.
9654 In multiple-processor systems, there are many ARC variants with different
9655 instruction and register set characteristics. This flag prevents code
9656 compiled for one cpu to be linked with code compiled for another.
9657 No facility exists for handling variants that are ``almost identical''.
9658 This is an all or nothing option.
9660 @item -mcpu=@var{cpu}
9662 Compile code for ARC variant @var{cpu}.
9663 Which variants are supported depend on the configuration.
9664 All variants support @option{-mcpu=base}, this is the default.
9666 @item -mtext=@var{text-section}
9667 @itemx -mdata=@var{data-section}
9668 @itemx -mrodata=@var{readonly-data-section}
9672 Put functions, data, and readonly data in @var{text-section},
9673 @var{data-section}, and @var{readonly-data-section} respectively
9674 by default. This can be overridden with the @code{section} attribute.
9675 @xref{Variable Attributes}.
9677 @item -mfix-cortex-m3-ldrd
9678 @opindex mfix-cortex-m3-ldrd
9679 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
9680 with overlapping destination and base registers are used. This option avoids
9681 generating these instructions. This option is enabled by default when
9682 @option{-mcpu=cortex-m3} is specified.
9687 @subsection ARM Options
9690 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
9694 @item -mabi=@var{name}
9696 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
9697 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
9700 @opindex mapcs-frame
9701 Generate a stack frame that is compliant with the ARM Procedure Call
9702 Standard for all functions, even if this is not strictly necessary for
9703 correct execution of the code. Specifying @option{-fomit-frame-pointer}
9704 with this option will cause the stack frames not to be generated for
9705 leaf functions. The default is @option{-mno-apcs-frame}.
9709 This is a synonym for @option{-mapcs-frame}.
9712 @c not currently implemented
9713 @item -mapcs-stack-check
9714 @opindex mapcs-stack-check
9715 Generate code to check the amount of stack space available upon entry to
9716 every function (that actually uses some stack space). If there is
9717 insufficient space available then either the function
9718 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
9719 called, depending upon the amount of stack space required. The run time
9720 system is required to provide these functions. The default is
9721 @option{-mno-apcs-stack-check}, since this produces smaller code.
9723 @c not currently implemented
9725 @opindex mapcs-float
9726 Pass floating point arguments using the float point registers. This is
9727 one of the variants of the APCS@. This option is recommended if the
9728 target hardware has a floating point unit or if a lot of floating point
9729 arithmetic is going to be performed by the code. The default is
9730 @option{-mno-apcs-float}, since integer only code is slightly increased in
9731 size if @option{-mapcs-float} is used.
9733 @c not currently implemented
9734 @item -mapcs-reentrant
9735 @opindex mapcs-reentrant
9736 Generate reentrant, position independent code. The default is
9737 @option{-mno-apcs-reentrant}.
9740 @item -mthumb-interwork
9741 @opindex mthumb-interwork
9742 Generate code which supports calling between the ARM and Thumb
9743 instruction sets. Without this option the two instruction sets cannot
9744 be reliably used inside one program. The default is
9745 @option{-mno-thumb-interwork}, since slightly larger code is generated
9746 when @option{-mthumb-interwork} is specified.
9748 @item -mno-sched-prolog
9749 @opindex mno-sched-prolog
9750 Prevent the reordering of instructions in the function prolog, or the
9751 merging of those instruction with the instructions in the function's
9752 body. This means that all functions will start with a recognizable set
9753 of instructions (or in fact one of a choice from a small set of
9754 different function prologues), and this information can be used to
9755 locate the start if functions inside an executable piece of code. The
9756 default is @option{-msched-prolog}.
9758 @item -mfloat-abi=@var{name}
9760 Specifies which floating-point ABI to use. Permissible values
9761 are: @samp{soft}, @samp{softfp} and @samp{hard}.
9763 Specifying @samp{soft} causes GCC to generate output containing
9764 library calls for floating-point operations.
9765 @samp{softfp} allows the generation of code using hardware floating-point
9766 instructions, but still uses the soft-float calling conventions.
9767 @samp{hard} allows generation of floating-point instructions
9768 and uses FPU-specific calling conventions.
9770 The default depends on the specific target configuration. Note that
9771 the hard-float and soft-float ABIs are not link-compatible; you must
9772 compile your entire program with the same ABI, and link with a
9773 compatible set of libraries.
9776 @opindex mhard-float
9777 Equivalent to @option{-mfloat-abi=hard}.
9780 @opindex msoft-float
9781 Equivalent to @option{-mfloat-abi=soft}.
9783 @item -mlittle-endian
9784 @opindex mlittle-endian
9785 Generate code for a processor running in little-endian mode. This is
9786 the default for all standard configurations.
9789 @opindex mbig-endian
9790 Generate code for a processor running in big-endian mode; the default is
9791 to compile code for a little-endian processor.
9793 @item -mwords-little-endian
9794 @opindex mwords-little-endian
9795 This option only applies when generating code for big-endian processors.
9796 Generate code for a little-endian word order but a big-endian byte
9797 order. That is, a byte order of the form @samp{32107654}. Note: this
9798 option should only be used if you require compatibility with code for
9799 big-endian ARM processors generated by versions of the compiler prior to
9802 @item -mcpu=@var{name}
9804 This specifies the name of the target ARM processor. GCC uses this name
9805 to determine what kind of instructions it can emit when generating
9806 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
9807 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
9808 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
9809 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
9810 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
9812 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
9813 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
9814 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
9815 @samp{strongarm1110},
9816 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
9817 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
9818 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
9819 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
9820 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
9821 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
9822 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
9823 @samp{cortex-a5}, @samp{cortex-a8}, @samp{cortex-a9},
9824 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-m3},
9827 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9829 @item -mtune=@var{name}
9831 This option is very similar to the @option{-mcpu=} option, except that
9832 instead of specifying the actual target processor type, and hence
9833 restricting which instructions can be used, it specifies that GCC should
9834 tune the performance of the code as if the target were of the type
9835 specified in this option, but still choosing the instructions that it
9836 will generate based on the cpu specified by a @option{-mcpu=} option.
9837 For some ARM implementations better performance can be obtained by using
9840 @item -march=@var{name}
9842 This specifies the name of the target ARM architecture. GCC uses this
9843 name to determine what kind of instructions it can emit when generating
9844 assembly code. This option can be used in conjunction with or instead
9845 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
9846 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
9847 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
9848 @samp{armv6}, @samp{armv6j},
9849 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
9850 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
9851 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9853 @item -mfpu=@var{name}
9854 @itemx -mfpe=@var{number}
9855 @itemx -mfp=@var{number}
9859 This specifies what floating point hardware (or hardware emulation) is
9860 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
9861 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-fp16},
9862 @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd}, @samp{vfpv3xd-fp16},
9863 @samp{neon}, @samp{neon-fp16}, @samp{vfpv4}, @samp{vfpv4-d16},
9864 @samp{fpv4-sp-d16} and @samp{neon-vfpv4}.
9865 @option{-mfp} and @option{-mfpe} are synonyms for
9866 @option{-mfpu}=@samp{fpe}@var{number}, for compatibility with older versions
9869 If @option{-msoft-float} is specified this specifies the format of
9870 floating point values.
9872 @item -mfp16-format=@var{name}
9873 @opindex mfp16-format
9874 Specify the format of the @code{__fp16} half-precision floating-point type.
9875 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
9876 the default is @samp{none}, in which case the @code{__fp16} type is not
9877 defined. @xref{Half-Precision}, for more information.
9879 @item -mstructure-size-boundary=@var{n}
9880 @opindex mstructure-size-boundary
9881 The size of all structures and unions will be rounded up to a multiple
9882 of the number of bits set by this option. Permissible values are 8, 32
9883 and 64. The default value varies for different toolchains. For the COFF
9884 targeted toolchain the default value is 8. A value of 64 is only allowed
9885 if the underlying ABI supports it.
9887 Specifying the larger number can produce faster, more efficient code, but
9888 can also increase the size of the program. Different values are potentially
9889 incompatible. Code compiled with one value cannot necessarily expect to
9890 work with code or libraries compiled with another value, if they exchange
9891 information using structures or unions.
9893 @item -mabort-on-noreturn
9894 @opindex mabort-on-noreturn
9895 Generate a call to the function @code{abort} at the end of a
9896 @code{noreturn} function. It will be executed if the function tries to
9900 @itemx -mno-long-calls
9901 @opindex mlong-calls
9902 @opindex mno-long-calls
9903 Tells the compiler to perform function calls by first loading the
9904 address of the function into a register and then performing a subroutine
9905 call on this register. This switch is needed if the target function
9906 will lie outside of the 64 megabyte addressing range of the offset based
9907 version of subroutine call instruction.
9909 Even if this switch is enabled, not all function calls will be turned
9910 into long calls. The heuristic is that static functions, functions
9911 which have the @samp{short-call} attribute, functions that are inside
9912 the scope of a @samp{#pragma no_long_calls} directive and functions whose
9913 definitions have already been compiled within the current compilation
9914 unit, will not be turned into long calls. The exception to this rule is
9915 that weak function definitions, functions with the @samp{long-call}
9916 attribute or the @samp{section} attribute, and functions that are within
9917 the scope of a @samp{#pragma long_calls} directive, will always be
9918 turned into long calls.
9920 This feature is not enabled by default. Specifying
9921 @option{-mno-long-calls} will restore the default behavior, as will
9922 placing the function calls within the scope of a @samp{#pragma
9923 long_calls_off} directive. Note these switches have no effect on how
9924 the compiler generates code to handle function calls via function
9927 @item -msingle-pic-base
9928 @opindex msingle-pic-base
9929 Treat the register used for PIC addressing as read-only, rather than
9930 loading it in the prologue for each function. The run-time system is
9931 responsible for initializing this register with an appropriate value
9932 before execution begins.
9934 @item -mpic-register=@var{reg}
9935 @opindex mpic-register
9936 Specify the register to be used for PIC addressing. The default is R10
9937 unless stack-checking is enabled, when R9 is used.
9939 @item -mcirrus-fix-invalid-insns
9940 @opindex mcirrus-fix-invalid-insns
9941 @opindex mno-cirrus-fix-invalid-insns
9942 Insert NOPs into the instruction stream to in order to work around
9943 problems with invalid Maverick instruction combinations. This option
9944 is only valid if the @option{-mcpu=ep9312} option has been used to
9945 enable generation of instructions for the Cirrus Maverick floating
9946 point co-processor. This option is not enabled by default, since the
9947 problem is only present in older Maverick implementations. The default
9948 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
9951 @item -mpoke-function-name
9952 @opindex mpoke-function-name
9953 Write the name of each function into the text section, directly
9954 preceding the function prologue. The generated code is similar to this:
9958 .ascii "arm_poke_function_name", 0
9961 .word 0xff000000 + (t1 - t0)
9962 arm_poke_function_name
9964 stmfd sp!, @{fp, ip, lr, pc@}
9968 When performing a stack backtrace, code can inspect the value of
9969 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
9970 location @code{pc - 12} and the top 8 bits are set, then we know that
9971 there is a function name embedded immediately preceding this location
9972 and has length @code{((pc[-3]) & 0xff000000)}.
9976 Generate code for the Thumb instruction set. The default is to
9977 use the 32-bit ARM instruction set.
9978 This option automatically enables either 16-bit Thumb-1 or
9979 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
9980 and @option{-march=@var{name}} options. This option is not passed to the
9981 assembler. If you want to force assembler files to be interpreted as Thumb code,
9982 either add a @samp{.thumb} directive to the source or pass the @option{-mthumb}
9983 option directly to the assembler by prefixing it with @option{-Wa}.
9986 @opindex mtpcs-frame
9987 Generate a stack frame that is compliant with the Thumb Procedure Call
9988 Standard for all non-leaf functions. (A leaf function is one that does
9989 not call any other functions.) The default is @option{-mno-tpcs-frame}.
9991 @item -mtpcs-leaf-frame
9992 @opindex mtpcs-leaf-frame
9993 Generate a stack frame that is compliant with the Thumb Procedure Call
9994 Standard for all leaf functions. (A leaf function is one that does
9995 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
9997 @item -mcallee-super-interworking
9998 @opindex mcallee-super-interworking
9999 Gives all externally visible functions in the file being compiled an ARM
10000 instruction set header which switches to Thumb mode before executing the
10001 rest of the function. This allows these functions to be called from
10002 non-interworking code. This option is not valid in AAPCS configurations
10003 because interworking is enabled by default.
10005 @item -mcaller-super-interworking
10006 @opindex mcaller-super-interworking
10007 Allows calls via function pointers (including virtual functions) to
10008 execute correctly regardless of whether the target code has been
10009 compiled for interworking or not. There is a small overhead in the cost
10010 of executing a function pointer if this option is enabled. This option
10011 is not valid in AAPCS configurations because interworking is enabled
10014 @item -mtp=@var{name}
10016 Specify the access model for the thread local storage pointer. The valid
10017 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
10018 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
10019 (supported in the arm6k architecture), and @option{auto}, which uses the
10020 best available method for the selected processor. The default setting is
10023 @item -mword-relocations
10024 @opindex mword-relocations
10025 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
10026 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
10027 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
10033 @subsection AVR Options
10034 @cindex AVR Options
10036 These options are defined for AVR implementations:
10039 @item -mmcu=@var{mcu}
10041 Specify ATMEL AVR instruction set or MCU type.
10043 Instruction set avr1 is for the minimal AVR core, not supported by the C
10044 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
10045 attiny11, attiny12, attiny15, attiny28).
10047 Instruction set avr2 (default) is for the classic AVR core with up to
10048 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
10049 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
10050 at90c8534, at90s8535).
10052 Instruction set avr3 is for the classic AVR core with up to 128K program
10053 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
10055 Instruction set avr4 is for the enhanced AVR core with up to 8K program
10056 memory space (MCU types: atmega8, atmega83, atmega85).
10058 Instruction set avr5 is for the enhanced AVR core with up to 128K program
10059 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
10060 atmega64, atmega128, at43usb355, at94k).
10062 @item -mno-interrupts
10063 @opindex mno-interrupts
10064 Generated code is not compatible with hardware interrupts.
10065 Code size will be smaller.
10067 @item -mcall-prologues
10068 @opindex mcall-prologues
10069 Functions prologues/epilogues expanded as call to appropriate
10070 subroutines. Code size will be smaller.
10073 @opindex mtiny-stack
10074 Change only the low 8 bits of the stack pointer.
10078 Assume int to be 8 bit integer. This affects the sizes of all types: A
10079 char will be 1 byte, an int will be 1 byte, a long will be 2 bytes
10080 and long long will be 4 bytes. Please note that this option does not
10081 comply to the C standards, but it will provide you with smaller code
10085 @node Blackfin Options
10086 @subsection Blackfin Options
10087 @cindex Blackfin Options
10090 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
10092 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
10093 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
10094 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
10095 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
10096 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
10097 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
10098 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
10100 The optional @var{sirevision} specifies the silicon revision of the target
10101 Blackfin processor. Any workarounds available for the targeted silicon revision
10102 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
10103 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
10104 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
10105 hexadecimal digits representing the major and minor numbers in the silicon
10106 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
10107 is not defined. If @var{sirevision} is @samp{any}, the
10108 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
10109 If this optional @var{sirevision} is not used, GCC assumes the latest known
10110 silicon revision of the targeted Blackfin processor.
10112 Support for @samp{bf561} is incomplete. For @samp{bf561},
10113 Only the processor macro is defined.
10114 Without this option, @samp{bf532} is used as the processor by default.
10115 The corresponding predefined processor macros for @var{cpu} is to
10116 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
10117 provided by libgloss to be linked in if @option{-msim} is not given.
10121 Specifies that the program will be run on the simulator. This causes
10122 the simulator BSP provided by libgloss to be linked in. This option
10123 has effect only for @samp{bfin-elf} toolchain.
10124 Certain other options, such as @option{-mid-shared-library} and
10125 @option{-mfdpic}, imply @option{-msim}.
10127 @item -momit-leaf-frame-pointer
10128 @opindex momit-leaf-frame-pointer
10129 Don't keep the frame pointer in a register for leaf functions. This
10130 avoids the instructions to save, set up and restore frame pointers and
10131 makes an extra register available in leaf functions. The option
10132 @option{-fomit-frame-pointer} removes the frame pointer for all functions
10133 which might make debugging harder.
10135 @item -mspecld-anomaly
10136 @opindex mspecld-anomaly
10137 When enabled, the compiler will ensure that the generated code does not
10138 contain speculative loads after jump instructions. If this option is used,
10139 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
10141 @item -mno-specld-anomaly
10142 @opindex mno-specld-anomaly
10143 Don't generate extra code to prevent speculative loads from occurring.
10145 @item -mcsync-anomaly
10146 @opindex mcsync-anomaly
10147 When enabled, the compiler will ensure that the generated code does not
10148 contain CSYNC or SSYNC instructions too soon after conditional branches.
10149 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
10151 @item -mno-csync-anomaly
10152 @opindex mno-csync-anomaly
10153 Don't generate extra code to prevent CSYNC or SSYNC instructions from
10154 occurring too soon after a conditional branch.
10158 When enabled, the compiler is free to take advantage of the knowledge that
10159 the entire program fits into the low 64k of memory.
10162 @opindex mno-low-64k
10163 Assume that the program is arbitrarily large. This is the default.
10165 @item -mstack-check-l1
10166 @opindex mstack-check-l1
10167 Do stack checking using information placed into L1 scratchpad memory by the
10170 @item -mid-shared-library
10171 @opindex mid-shared-library
10172 Generate code that supports shared libraries via the library ID method.
10173 This allows for execute in place and shared libraries in an environment
10174 without virtual memory management. This option implies @option{-fPIC}.
10175 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10177 @item -mno-id-shared-library
10178 @opindex mno-id-shared-library
10179 Generate code that doesn't assume ID based shared libraries are being used.
10180 This is the default.
10182 @item -mleaf-id-shared-library
10183 @opindex mleaf-id-shared-library
10184 Generate code that supports shared libraries via the library ID method,
10185 but assumes that this library or executable won't link against any other
10186 ID shared libraries. That allows the compiler to use faster code for jumps
10189 @item -mno-leaf-id-shared-library
10190 @opindex mno-leaf-id-shared-library
10191 Do not assume that the code being compiled won't link against any ID shared
10192 libraries. Slower code will be generated for jump and call insns.
10194 @item -mshared-library-id=n
10195 @opindex mshared-library-id
10196 Specified the identification number of the ID based shared library being
10197 compiled. Specifying a value of 0 will generate more compact code, specifying
10198 other values will force the allocation of that number to the current
10199 library but is no more space or time efficient than omitting this option.
10203 Generate code that allows the data segment to be located in a different
10204 area of memory from the text segment. This allows for execute in place in
10205 an environment without virtual memory management by eliminating relocations
10206 against the text section.
10208 @item -mno-sep-data
10209 @opindex mno-sep-data
10210 Generate code that assumes that the data segment follows the text segment.
10211 This is the default.
10214 @itemx -mno-long-calls
10215 @opindex mlong-calls
10216 @opindex mno-long-calls
10217 Tells the compiler to perform function calls by first loading the
10218 address of the function into a register and then performing a subroutine
10219 call on this register. This switch is needed if the target function
10220 will lie outside of the 24 bit addressing range of the offset based
10221 version of subroutine call instruction.
10223 This feature is not enabled by default. Specifying
10224 @option{-mno-long-calls} will restore the default behavior. Note these
10225 switches have no effect on how the compiler generates code to handle
10226 function calls via function pointers.
10230 Link with the fast floating-point library. This library relaxes some of
10231 the IEEE floating-point standard's rules for checking inputs against
10232 Not-a-Number (NAN), in the interest of performance.
10235 @opindex minline-plt
10236 Enable inlining of PLT entries in function calls to functions that are
10237 not known to bind locally. It has no effect without @option{-mfdpic}.
10240 @opindex mmulticore
10241 Build standalone application for multicore Blackfin processor. Proper
10242 start files and link scripts will be used to support multicore.
10243 This option defines @code{__BFIN_MULTICORE}. It can only be used with
10244 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
10245 @option{-mcorea} or @option{-mcoreb}. If it's used without
10246 @option{-mcorea} or @option{-mcoreb}, single application/dual core
10247 programming model is used. In this model, the main function of Core B
10248 should be named as coreb_main. If it's used with @option{-mcorea} or
10249 @option{-mcoreb}, one application per core programming model is used.
10250 If this option is not used, single core application programming
10255 Build standalone application for Core A of BF561 when using
10256 one application per core programming model. Proper start files
10257 and link scripts will be used to support Core A. This option
10258 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
10262 Build standalone application for Core B of BF561 when using
10263 one application per core programming model. Proper start files
10264 and link scripts will be used to support Core B. This option
10265 defines @code{__BFIN_COREB}. When this option is used, coreb_main
10266 should be used instead of main. It must be used with
10267 @option{-mmulticore}.
10271 Build standalone application for SDRAM. Proper start files and
10272 link scripts will be used to put the application into SDRAM.
10273 Loader should initialize SDRAM before loading the application
10274 into SDRAM. This option defines @code{__BFIN_SDRAM}.
10278 Assume that ICPLBs are enabled at runtime. This has an effect on certain
10279 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
10280 are enabled; for standalone applications the default is off.
10284 @subsection CRIS Options
10285 @cindex CRIS Options
10287 These options are defined specifically for the CRIS ports.
10290 @item -march=@var{architecture-type}
10291 @itemx -mcpu=@var{architecture-type}
10294 Generate code for the specified architecture. The choices for
10295 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
10296 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
10297 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
10300 @item -mtune=@var{architecture-type}
10302 Tune to @var{architecture-type} everything applicable about the generated
10303 code, except for the ABI and the set of available instructions. The
10304 choices for @var{architecture-type} are the same as for
10305 @option{-march=@var{architecture-type}}.
10307 @item -mmax-stack-frame=@var{n}
10308 @opindex mmax-stack-frame
10309 Warn when the stack frame of a function exceeds @var{n} bytes.
10315 The options @option{-metrax4} and @option{-metrax100} are synonyms for
10316 @option{-march=v3} and @option{-march=v8} respectively.
10318 @item -mmul-bug-workaround
10319 @itemx -mno-mul-bug-workaround
10320 @opindex mmul-bug-workaround
10321 @opindex mno-mul-bug-workaround
10322 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
10323 models where it applies. This option is active by default.
10327 Enable CRIS-specific verbose debug-related information in the assembly
10328 code. This option also has the effect to turn off the @samp{#NO_APP}
10329 formatted-code indicator to the assembler at the beginning of the
10334 Do not use condition-code results from previous instruction; always emit
10335 compare and test instructions before use of condition codes.
10337 @item -mno-side-effects
10338 @opindex mno-side-effects
10339 Do not emit instructions with side-effects in addressing modes other than
10342 @item -mstack-align
10343 @itemx -mno-stack-align
10344 @itemx -mdata-align
10345 @itemx -mno-data-align
10346 @itemx -mconst-align
10347 @itemx -mno-const-align
10348 @opindex mstack-align
10349 @opindex mno-stack-align
10350 @opindex mdata-align
10351 @opindex mno-data-align
10352 @opindex mconst-align
10353 @opindex mno-const-align
10354 These options (no-options) arranges (eliminate arrangements) for the
10355 stack-frame, individual data and constants to be aligned for the maximum
10356 single data access size for the chosen CPU model. The default is to
10357 arrange for 32-bit alignment. ABI details such as structure layout are
10358 not affected by these options.
10366 Similar to the stack- data- and const-align options above, these options
10367 arrange for stack-frame, writable data and constants to all be 32-bit,
10368 16-bit or 8-bit aligned. The default is 32-bit alignment.
10370 @item -mno-prologue-epilogue
10371 @itemx -mprologue-epilogue
10372 @opindex mno-prologue-epilogue
10373 @opindex mprologue-epilogue
10374 With @option{-mno-prologue-epilogue}, the normal function prologue and
10375 epilogue that sets up the stack-frame are omitted and no return
10376 instructions or return sequences are generated in the code. Use this
10377 option only together with visual inspection of the compiled code: no
10378 warnings or errors are generated when call-saved registers must be saved,
10379 or storage for local variable needs to be allocated.
10383 @opindex mno-gotplt
10385 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
10386 instruction sequences that load addresses for functions from the PLT part
10387 of the GOT rather than (traditional on other architectures) calls to the
10388 PLT@. The default is @option{-mgotplt}.
10392 Legacy no-op option only recognized with the cris-axis-elf and
10393 cris-axis-linux-gnu targets.
10397 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
10401 This option, recognized for the cris-axis-elf arranges
10402 to link with input-output functions from a simulator library. Code,
10403 initialized data and zero-initialized data are allocated consecutively.
10407 Like @option{-sim}, but pass linker options to locate initialized data at
10408 0x40000000 and zero-initialized data at 0x80000000.
10412 @subsection CRX Options
10413 @cindex CRX Options
10415 These options are defined specifically for the CRX ports.
10421 Enable the use of multiply-accumulate instructions. Disabled by default.
10424 @opindex mpush-args
10425 Push instructions will be used to pass outgoing arguments when functions
10426 are called. Enabled by default.
10429 @node Darwin Options
10430 @subsection Darwin Options
10431 @cindex Darwin options
10433 These options are defined for all architectures running the Darwin operating
10436 FSF GCC on Darwin does not create ``fat'' object files; it will create
10437 an object file for the single architecture that it was built to
10438 target. Apple's GCC on Darwin does create ``fat'' files if multiple
10439 @option{-arch} options are used; it does so by running the compiler or
10440 linker multiple times and joining the results together with
10443 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
10444 @samp{i686}) is determined by the flags that specify the ISA
10445 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
10446 @option{-force_cpusubtype_ALL} option can be used to override this.
10448 The Darwin tools vary in their behavior when presented with an ISA
10449 mismatch. The assembler, @file{as}, will only permit instructions to
10450 be used that are valid for the subtype of the file it is generating,
10451 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
10452 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
10453 and print an error if asked to create a shared library with a less
10454 restrictive subtype than its input files (for instance, trying to put
10455 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
10456 for executables, @file{ld}, will quietly give the executable the most
10457 restrictive subtype of any of its input files.
10462 Add the framework directory @var{dir} to the head of the list of
10463 directories to be searched for header files. These directories are
10464 interleaved with those specified by @option{-I} options and are
10465 scanned in a left-to-right order.
10467 A framework directory is a directory with frameworks in it. A
10468 framework is a directory with a @samp{"Headers"} and/or
10469 @samp{"PrivateHeaders"} directory contained directly in it that ends
10470 in @samp{".framework"}. The name of a framework is the name of this
10471 directory excluding the @samp{".framework"}. Headers associated with
10472 the framework are found in one of those two directories, with
10473 @samp{"Headers"} being searched first. A subframework is a framework
10474 directory that is in a framework's @samp{"Frameworks"} directory.
10475 Includes of subframework headers can only appear in a header of a
10476 framework that contains the subframework, or in a sibling subframework
10477 header. Two subframeworks are siblings if they occur in the same
10478 framework. A subframework should not have the same name as a
10479 framework, a warning will be issued if this is violated. Currently a
10480 subframework cannot have subframeworks, in the future, the mechanism
10481 may be extended to support this. The standard frameworks can be found
10482 in @samp{"/System/Library/Frameworks"} and
10483 @samp{"/Library/Frameworks"}. An example include looks like
10484 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
10485 the name of the framework and header.h is found in the
10486 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
10488 @item -iframework@var{dir}
10489 @opindex iframework
10490 Like @option{-F} except the directory is a treated as a system
10491 directory. The main difference between this @option{-iframework} and
10492 @option{-F} is that with @option{-iframework} the compiler does not
10493 warn about constructs contained within header files found via
10494 @var{dir}. This option is valid only for the C family of languages.
10498 Emit debugging information for symbols that are used. For STABS
10499 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
10500 This is by default ON@.
10504 Emit debugging information for all symbols and types.
10506 @item -mmacosx-version-min=@var{version}
10507 The earliest version of MacOS X that this executable will run on
10508 is @var{version}. Typical values of @var{version} include @code{10.1},
10509 @code{10.2}, and @code{10.3.9}.
10511 If the compiler was built to use the system's headers by default,
10512 then the default for this option is the system version on which the
10513 compiler is running, otherwise the default is to make choices which
10514 are compatible with as many systems and code bases as possible.
10518 Enable kernel development mode. The @option{-mkernel} option sets
10519 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
10520 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
10521 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
10522 applicable. This mode also sets @option{-mno-altivec},
10523 @option{-msoft-float}, @option{-fno-builtin} and
10524 @option{-mlong-branch} for PowerPC targets.
10526 @item -mone-byte-bool
10527 @opindex mone-byte-bool
10528 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
10529 By default @samp{sizeof(bool)} is @samp{4} when compiling for
10530 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
10531 option has no effect on x86.
10533 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
10534 to generate code that is not binary compatible with code generated
10535 without that switch. Using this switch may require recompiling all
10536 other modules in a program, including system libraries. Use this
10537 switch to conform to a non-default data model.
10539 @item -mfix-and-continue
10540 @itemx -ffix-and-continue
10541 @itemx -findirect-data
10542 @opindex mfix-and-continue
10543 @opindex ffix-and-continue
10544 @opindex findirect-data
10545 Generate code suitable for fast turn around development. Needed to
10546 enable gdb to dynamically load @code{.o} files into already running
10547 programs. @option{-findirect-data} and @option{-ffix-and-continue}
10548 are provided for backwards compatibility.
10552 Loads all members of static archive libraries.
10553 See man ld(1) for more information.
10555 @item -arch_errors_fatal
10556 @opindex arch_errors_fatal
10557 Cause the errors having to do with files that have the wrong architecture
10560 @item -bind_at_load
10561 @opindex bind_at_load
10562 Causes the output file to be marked such that the dynamic linker will
10563 bind all undefined references when the file is loaded or launched.
10567 Produce a Mach-o bundle format file.
10568 See man ld(1) for more information.
10570 @item -bundle_loader @var{executable}
10571 @opindex bundle_loader
10572 This option specifies the @var{executable} that will be loading the build
10573 output file being linked. See man ld(1) for more information.
10576 @opindex dynamiclib
10577 When passed this option, GCC will produce a dynamic library instead of
10578 an executable when linking, using the Darwin @file{libtool} command.
10580 @item -force_cpusubtype_ALL
10581 @opindex force_cpusubtype_ALL
10582 This causes GCC's output file to have the @var{ALL} subtype, instead of
10583 one controlled by the @option{-mcpu} or @option{-march} option.
10585 @item -allowable_client @var{client_name}
10586 @itemx -client_name
10587 @itemx -compatibility_version
10588 @itemx -current_version
10590 @itemx -dependency-file
10592 @itemx -dylinker_install_name
10594 @itemx -exported_symbols_list
10596 @itemx -flat_namespace
10597 @itemx -force_flat_namespace
10598 @itemx -headerpad_max_install_names
10601 @itemx -install_name
10602 @itemx -keep_private_externs
10603 @itemx -multi_module
10604 @itemx -multiply_defined
10605 @itemx -multiply_defined_unused
10607 @itemx -no_dead_strip_inits_and_terms
10608 @itemx -nofixprebinding
10609 @itemx -nomultidefs
10611 @itemx -noseglinkedit
10612 @itemx -pagezero_size
10614 @itemx -prebind_all_twolevel_modules
10615 @itemx -private_bundle
10616 @itemx -read_only_relocs
10618 @itemx -sectobjectsymbols
10622 @itemx -sectobjectsymbols
10625 @itemx -segs_read_only_addr
10626 @itemx -segs_read_write_addr
10627 @itemx -seg_addr_table
10628 @itemx -seg_addr_table_filename
10629 @itemx -seglinkedit
10631 @itemx -segs_read_only_addr
10632 @itemx -segs_read_write_addr
10633 @itemx -single_module
10635 @itemx -sub_library
10636 @itemx -sub_umbrella
10637 @itemx -twolevel_namespace
10640 @itemx -unexported_symbols_list
10641 @itemx -weak_reference_mismatches
10642 @itemx -whatsloaded
10643 @opindex allowable_client
10644 @opindex client_name
10645 @opindex compatibility_version
10646 @opindex current_version
10647 @opindex dead_strip
10648 @opindex dependency-file
10649 @opindex dylib_file
10650 @opindex dylinker_install_name
10652 @opindex exported_symbols_list
10654 @opindex flat_namespace
10655 @opindex force_flat_namespace
10656 @opindex headerpad_max_install_names
10657 @opindex image_base
10659 @opindex install_name
10660 @opindex keep_private_externs
10661 @opindex multi_module
10662 @opindex multiply_defined
10663 @opindex multiply_defined_unused
10664 @opindex noall_load
10665 @opindex no_dead_strip_inits_and_terms
10666 @opindex nofixprebinding
10667 @opindex nomultidefs
10669 @opindex noseglinkedit
10670 @opindex pagezero_size
10672 @opindex prebind_all_twolevel_modules
10673 @opindex private_bundle
10674 @opindex read_only_relocs
10676 @opindex sectobjectsymbols
10679 @opindex sectcreate
10680 @opindex sectobjectsymbols
10683 @opindex segs_read_only_addr
10684 @opindex segs_read_write_addr
10685 @opindex seg_addr_table
10686 @opindex seg_addr_table_filename
10687 @opindex seglinkedit
10689 @opindex segs_read_only_addr
10690 @opindex segs_read_write_addr
10691 @opindex single_module
10693 @opindex sub_library
10694 @opindex sub_umbrella
10695 @opindex twolevel_namespace
10698 @opindex unexported_symbols_list
10699 @opindex weak_reference_mismatches
10700 @opindex whatsloaded
10701 These options are passed to the Darwin linker. The Darwin linker man page
10702 describes them in detail.
10705 @node DEC Alpha Options
10706 @subsection DEC Alpha Options
10708 These @samp{-m} options are defined for the DEC Alpha implementations:
10711 @item -mno-soft-float
10712 @itemx -msoft-float
10713 @opindex mno-soft-float
10714 @opindex msoft-float
10715 Use (do not use) the hardware floating-point instructions for
10716 floating-point operations. When @option{-msoft-float} is specified,
10717 functions in @file{libgcc.a} will be used to perform floating-point
10718 operations. Unless they are replaced by routines that emulate the
10719 floating-point operations, or compiled in such a way as to call such
10720 emulations routines, these routines will issue floating-point
10721 operations. If you are compiling for an Alpha without floating-point
10722 operations, you must ensure that the library is built so as not to call
10725 Note that Alpha implementations without floating-point operations are
10726 required to have floating-point registers.
10729 @itemx -mno-fp-regs
10731 @opindex mno-fp-regs
10732 Generate code that uses (does not use) the floating-point register set.
10733 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
10734 register set is not used, floating point operands are passed in integer
10735 registers as if they were integers and floating-point results are passed
10736 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
10737 so any function with a floating-point argument or return value called by code
10738 compiled with @option{-mno-fp-regs} must also be compiled with that
10741 A typical use of this option is building a kernel that does not use,
10742 and hence need not save and restore, any floating-point registers.
10746 The Alpha architecture implements floating-point hardware optimized for
10747 maximum performance. It is mostly compliant with the IEEE floating
10748 point standard. However, for full compliance, software assistance is
10749 required. This option generates code fully IEEE compliant code
10750 @emph{except} that the @var{inexact-flag} is not maintained (see below).
10751 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
10752 defined during compilation. The resulting code is less efficient but is
10753 able to correctly support denormalized numbers and exceptional IEEE
10754 values such as not-a-number and plus/minus infinity. Other Alpha
10755 compilers call this option @option{-ieee_with_no_inexact}.
10757 @item -mieee-with-inexact
10758 @opindex mieee-with-inexact
10759 This is like @option{-mieee} except the generated code also maintains
10760 the IEEE @var{inexact-flag}. Turning on this option causes the
10761 generated code to implement fully-compliant IEEE math. In addition to
10762 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
10763 macro. On some Alpha implementations the resulting code may execute
10764 significantly slower than the code generated by default. Since there is
10765 very little code that depends on the @var{inexact-flag}, you should
10766 normally not specify this option. Other Alpha compilers call this
10767 option @option{-ieee_with_inexact}.
10769 @item -mfp-trap-mode=@var{trap-mode}
10770 @opindex mfp-trap-mode
10771 This option controls what floating-point related traps are enabled.
10772 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
10773 The trap mode can be set to one of four values:
10777 This is the default (normal) setting. The only traps that are enabled
10778 are the ones that cannot be disabled in software (e.g., division by zero
10782 In addition to the traps enabled by @samp{n}, underflow traps are enabled
10786 Like @samp{u}, but the instructions are marked to be safe for software
10787 completion (see Alpha architecture manual for details).
10790 Like @samp{su}, but inexact traps are enabled as well.
10793 @item -mfp-rounding-mode=@var{rounding-mode}
10794 @opindex mfp-rounding-mode
10795 Selects the IEEE rounding mode. Other Alpha compilers call this option
10796 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
10801 Normal IEEE rounding mode. Floating point numbers are rounded towards
10802 the nearest machine number or towards the even machine number in case
10806 Round towards minus infinity.
10809 Chopped rounding mode. Floating point numbers are rounded towards zero.
10812 Dynamic rounding mode. A field in the floating point control register
10813 (@var{fpcr}, see Alpha architecture reference manual) controls the
10814 rounding mode in effect. The C library initializes this register for
10815 rounding towards plus infinity. Thus, unless your program modifies the
10816 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
10819 @item -mtrap-precision=@var{trap-precision}
10820 @opindex mtrap-precision
10821 In the Alpha architecture, floating point traps are imprecise. This
10822 means without software assistance it is impossible to recover from a
10823 floating trap and program execution normally needs to be terminated.
10824 GCC can generate code that can assist operating system trap handlers
10825 in determining the exact location that caused a floating point trap.
10826 Depending on the requirements of an application, different levels of
10827 precisions can be selected:
10831 Program precision. This option is the default and means a trap handler
10832 can only identify which program caused a floating point exception.
10835 Function precision. The trap handler can determine the function that
10836 caused a floating point exception.
10839 Instruction precision. The trap handler can determine the exact
10840 instruction that caused a floating point exception.
10843 Other Alpha compilers provide the equivalent options called
10844 @option{-scope_safe} and @option{-resumption_safe}.
10846 @item -mieee-conformant
10847 @opindex mieee-conformant
10848 This option marks the generated code as IEEE conformant. You must not
10849 use this option unless you also specify @option{-mtrap-precision=i} and either
10850 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
10851 is to emit the line @samp{.eflag 48} in the function prologue of the
10852 generated assembly file. Under DEC Unix, this has the effect that
10853 IEEE-conformant math library routines will be linked in.
10855 @item -mbuild-constants
10856 @opindex mbuild-constants
10857 Normally GCC examines a 32- or 64-bit integer constant to
10858 see if it can construct it from smaller constants in two or three
10859 instructions. If it cannot, it will output the constant as a literal and
10860 generate code to load it from the data segment at runtime.
10862 Use this option to require GCC to construct @emph{all} integer constants
10863 using code, even if it takes more instructions (the maximum is six).
10865 You would typically use this option to build a shared library dynamic
10866 loader. Itself a shared library, it must relocate itself in memory
10867 before it can find the variables and constants in its own data segment.
10873 Select whether to generate code to be assembled by the vendor-supplied
10874 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
10892 Indicate whether GCC should generate code to use the optional BWX,
10893 CIX, FIX and MAX instruction sets. The default is to use the instruction
10894 sets supported by the CPU type specified via @option{-mcpu=} option or that
10895 of the CPU on which GCC was built if none was specified.
10898 @itemx -mfloat-ieee
10899 @opindex mfloat-vax
10900 @opindex mfloat-ieee
10901 Generate code that uses (does not use) VAX F and G floating point
10902 arithmetic instead of IEEE single and double precision.
10904 @item -mexplicit-relocs
10905 @itemx -mno-explicit-relocs
10906 @opindex mexplicit-relocs
10907 @opindex mno-explicit-relocs
10908 Older Alpha assemblers provided no way to generate symbol relocations
10909 except via assembler macros. Use of these macros does not allow
10910 optimal instruction scheduling. GNU binutils as of version 2.12
10911 supports a new syntax that allows the compiler to explicitly mark
10912 which relocations should apply to which instructions. This option
10913 is mostly useful for debugging, as GCC detects the capabilities of
10914 the assembler when it is built and sets the default accordingly.
10917 @itemx -mlarge-data
10918 @opindex msmall-data
10919 @opindex mlarge-data
10920 When @option{-mexplicit-relocs} is in effect, static data is
10921 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
10922 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
10923 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
10924 16-bit relocations off of the @code{$gp} register. This limits the
10925 size of the small data area to 64KB, but allows the variables to be
10926 directly accessed via a single instruction.
10928 The default is @option{-mlarge-data}. With this option the data area
10929 is limited to just below 2GB@. Programs that require more than 2GB of
10930 data must use @code{malloc} or @code{mmap} to allocate the data in the
10931 heap instead of in the program's data segment.
10933 When generating code for shared libraries, @option{-fpic} implies
10934 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
10937 @itemx -mlarge-text
10938 @opindex msmall-text
10939 @opindex mlarge-text
10940 When @option{-msmall-text} is used, the compiler assumes that the
10941 code of the entire program (or shared library) fits in 4MB, and is
10942 thus reachable with a branch instruction. When @option{-msmall-data}
10943 is used, the compiler can assume that all local symbols share the
10944 same @code{$gp} value, and thus reduce the number of instructions
10945 required for a function call from 4 to 1.
10947 The default is @option{-mlarge-text}.
10949 @item -mcpu=@var{cpu_type}
10951 Set the instruction set and instruction scheduling parameters for
10952 machine type @var{cpu_type}. You can specify either the @samp{EV}
10953 style name or the corresponding chip number. GCC supports scheduling
10954 parameters for the EV4, EV5 and EV6 family of processors and will
10955 choose the default values for the instruction set from the processor
10956 you specify. If you do not specify a processor type, GCC will default
10957 to the processor on which the compiler was built.
10959 Supported values for @var{cpu_type} are
10965 Schedules as an EV4 and has no instruction set extensions.
10969 Schedules as an EV5 and has no instruction set extensions.
10973 Schedules as an EV5 and supports the BWX extension.
10978 Schedules as an EV5 and supports the BWX and MAX extensions.
10982 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
10986 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
10989 Native Linux/GNU toolchains also support the value @samp{native},
10990 which selects the best architecture option for the host processor.
10991 @option{-mcpu=native} has no effect if GCC does not recognize
10994 @item -mtune=@var{cpu_type}
10996 Set only the instruction scheduling parameters for machine type
10997 @var{cpu_type}. The instruction set is not changed.
10999 Native Linux/GNU toolchains also support the value @samp{native},
11000 which selects the best architecture option for the host processor.
11001 @option{-mtune=native} has no effect if GCC does not recognize
11004 @item -mmemory-latency=@var{time}
11005 @opindex mmemory-latency
11006 Sets the latency the scheduler should assume for typical memory
11007 references as seen by the application. This number is highly
11008 dependent on the memory access patterns used by the application
11009 and the size of the external cache on the machine.
11011 Valid options for @var{time} are
11015 A decimal number representing clock cycles.
11021 The compiler contains estimates of the number of clock cycles for
11022 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
11023 (also called Dcache, Scache, and Bcache), as well as to main memory.
11024 Note that L3 is only valid for EV5.
11029 @node DEC Alpha/VMS Options
11030 @subsection DEC Alpha/VMS Options
11032 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
11035 @item -mvms-return-codes
11036 @opindex mvms-return-codes
11037 Return VMS condition codes from main. The default is to return POSIX
11038 style condition (e.g.@: error) codes.
11040 @item -mdebug-main=@var{prefix}
11041 @opindex mdebug-main=@var{prefix}
11042 Flag the first routine whose name starts with @var{prefix} as the main
11043 routine for the debugger.
11047 Default to 64bit memory allocation routines.
11051 @subsection FR30 Options
11052 @cindex FR30 Options
11054 These options are defined specifically for the FR30 port.
11058 @item -msmall-model
11059 @opindex msmall-model
11060 Use the small address space model. This can produce smaller code, but
11061 it does assume that all symbolic values and addresses will fit into a
11066 Assume that run-time support has been provided and so there is no need
11067 to include the simulator library (@file{libsim.a}) on the linker
11073 @subsection FRV Options
11074 @cindex FRV Options
11080 Only use the first 32 general purpose registers.
11085 Use all 64 general purpose registers.
11090 Use only the first 32 floating point registers.
11095 Use all 64 floating point registers
11098 @opindex mhard-float
11100 Use hardware instructions for floating point operations.
11103 @opindex msoft-float
11105 Use library routines for floating point operations.
11110 Dynamically allocate condition code registers.
11115 Do not try to dynamically allocate condition code registers, only
11116 use @code{icc0} and @code{fcc0}.
11121 Change ABI to use double word insns.
11126 Do not use double word instructions.
11131 Use floating point double instructions.
11134 @opindex mno-double
11136 Do not use floating point double instructions.
11141 Use media instructions.
11146 Do not use media instructions.
11151 Use multiply and add/subtract instructions.
11154 @opindex mno-muladd
11156 Do not use multiply and add/subtract instructions.
11161 Select the FDPIC ABI, that uses function descriptors to represent
11162 pointers to functions. Without any PIC/PIE-related options, it
11163 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
11164 assumes GOT entries and small data are within a 12-bit range from the
11165 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
11166 are computed with 32 bits.
11167 With a @samp{bfin-elf} target, this option implies @option{-msim}.
11170 @opindex minline-plt
11172 Enable inlining of PLT entries in function calls to functions that are
11173 not known to bind locally. It has no effect without @option{-mfdpic}.
11174 It's enabled by default if optimizing for speed and compiling for
11175 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
11176 optimization option such as @option{-O3} or above is present in the
11182 Assume a large TLS segment when generating thread-local code.
11187 Do not assume a large TLS segment when generating thread-local code.
11192 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
11193 that is known to be in read-only sections. It's enabled by default,
11194 except for @option{-fpic} or @option{-fpie}: even though it may help
11195 make the global offset table smaller, it trades 1 instruction for 4.
11196 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
11197 one of which may be shared by multiple symbols, and it avoids the need
11198 for a GOT entry for the referenced symbol, so it's more likely to be a
11199 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
11201 @item -multilib-library-pic
11202 @opindex multilib-library-pic
11204 Link with the (library, not FD) pic libraries. It's implied by
11205 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
11206 @option{-fpic} without @option{-mfdpic}. You should never have to use
11210 @opindex mlinked-fp
11212 Follow the EABI requirement of always creating a frame pointer whenever
11213 a stack frame is allocated. This option is enabled by default and can
11214 be disabled with @option{-mno-linked-fp}.
11217 @opindex mlong-calls
11219 Use indirect addressing to call functions outside the current
11220 compilation unit. This allows the functions to be placed anywhere
11221 within the 32-bit address space.
11223 @item -malign-labels
11224 @opindex malign-labels
11226 Try to align labels to an 8-byte boundary by inserting nops into the
11227 previous packet. This option only has an effect when VLIW packing
11228 is enabled. It doesn't create new packets; it merely adds nops to
11231 @item -mlibrary-pic
11232 @opindex mlibrary-pic
11234 Generate position-independent EABI code.
11239 Use only the first four media accumulator registers.
11244 Use all eight media accumulator registers.
11249 Pack VLIW instructions.
11254 Do not pack VLIW instructions.
11257 @opindex mno-eflags
11259 Do not mark ABI switches in e_flags.
11262 @opindex mcond-move
11264 Enable the use of conditional-move instructions (default).
11266 This switch is mainly for debugging the compiler and will likely be removed
11267 in a future version.
11269 @item -mno-cond-move
11270 @opindex mno-cond-move
11272 Disable the use of conditional-move instructions.
11274 This switch is mainly for debugging the compiler and will likely be removed
11275 in a future version.
11280 Enable the use of conditional set instructions (default).
11282 This switch is mainly for debugging the compiler and will likely be removed
11283 in a future version.
11288 Disable the use of conditional set instructions.
11290 This switch is mainly for debugging the compiler and will likely be removed
11291 in a future version.
11294 @opindex mcond-exec
11296 Enable the use of conditional execution (default).
11298 This switch is mainly for debugging the compiler and will likely be removed
11299 in a future version.
11301 @item -mno-cond-exec
11302 @opindex mno-cond-exec
11304 Disable the use of conditional execution.
11306 This switch is mainly for debugging the compiler and will likely be removed
11307 in a future version.
11309 @item -mvliw-branch
11310 @opindex mvliw-branch
11312 Run a pass to pack branches into VLIW instructions (default).
11314 This switch is mainly for debugging the compiler and will likely be removed
11315 in a future version.
11317 @item -mno-vliw-branch
11318 @opindex mno-vliw-branch
11320 Do not run a pass to pack branches into VLIW instructions.
11322 This switch is mainly for debugging the compiler and will likely be removed
11323 in a future version.
11325 @item -mmulti-cond-exec
11326 @opindex mmulti-cond-exec
11328 Enable optimization of @code{&&} and @code{||} in conditional execution
11331 This switch is mainly for debugging the compiler and will likely be removed
11332 in a future version.
11334 @item -mno-multi-cond-exec
11335 @opindex mno-multi-cond-exec
11337 Disable optimization of @code{&&} and @code{||} in conditional execution.
11339 This switch is mainly for debugging the compiler and will likely be removed
11340 in a future version.
11342 @item -mnested-cond-exec
11343 @opindex mnested-cond-exec
11345 Enable nested conditional execution optimizations (default).
11347 This switch is mainly for debugging the compiler and will likely be removed
11348 in a future version.
11350 @item -mno-nested-cond-exec
11351 @opindex mno-nested-cond-exec
11353 Disable nested conditional execution optimizations.
11355 This switch is mainly for debugging the compiler and will likely be removed
11356 in a future version.
11358 @item -moptimize-membar
11359 @opindex moptimize-membar
11361 This switch removes redundant @code{membar} instructions from the
11362 compiler generated code. It is enabled by default.
11364 @item -mno-optimize-membar
11365 @opindex mno-optimize-membar
11367 This switch disables the automatic removal of redundant @code{membar}
11368 instructions from the generated code.
11370 @item -mtomcat-stats
11371 @opindex mtomcat-stats
11373 Cause gas to print out tomcat statistics.
11375 @item -mcpu=@var{cpu}
11378 Select the processor type for which to generate code. Possible values are
11379 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
11380 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
11384 @node GNU/Linux Options
11385 @subsection GNU/Linux Options
11387 These @samp{-m} options are defined for GNU/Linux targets:
11392 Use the GNU C library instead of uClibc. This is the default except
11393 on @samp{*-*-linux-*uclibc*} targets.
11397 Use uClibc instead of the GNU C library. This is the default on
11398 @samp{*-*-linux-*uclibc*} targets.
11401 @node H8/300 Options
11402 @subsection H8/300 Options
11404 These @samp{-m} options are defined for the H8/300 implementations:
11409 Shorten some address references at link time, when possible; uses the
11410 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
11411 ld, Using ld}, for a fuller description.
11415 Generate code for the H8/300H@.
11419 Generate code for the H8S@.
11423 Generate code for the H8S and H8/300H in the normal mode. This switch
11424 must be used either with @option{-mh} or @option{-ms}.
11428 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
11432 Make @code{int} data 32 bits by default.
11435 @opindex malign-300
11436 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
11437 The default for the H8/300H and H8S is to align longs and floats on 4
11439 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
11440 This option has no effect on the H8/300.
11444 @subsection HPPA Options
11445 @cindex HPPA Options
11447 These @samp{-m} options are defined for the HPPA family of computers:
11450 @item -march=@var{architecture-type}
11452 Generate code for the specified architecture. The choices for
11453 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
11454 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
11455 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
11456 architecture option for your machine. Code compiled for lower numbered
11457 architectures will run on higher numbered architectures, but not the
11460 @item -mpa-risc-1-0
11461 @itemx -mpa-risc-1-1
11462 @itemx -mpa-risc-2-0
11463 @opindex mpa-risc-1-0
11464 @opindex mpa-risc-1-1
11465 @opindex mpa-risc-2-0
11466 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
11469 @opindex mbig-switch
11470 Generate code suitable for big switch tables. Use this option only if
11471 the assembler/linker complain about out of range branches within a switch
11474 @item -mjump-in-delay
11475 @opindex mjump-in-delay
11476 Fill delay slots of function calls with unconditional jump instructions
11477 by modifying the return pointer for the function call to be the target
11478 of the conditional jump.
11480 @item -mdisable-fpregs
11481 @opindex mdisable-fpregs
11482 Prevent floating point registers from being used in any manner. This is
11483 necessary for compiling kernels which perform lazy context switching of
11484 floating point registers. If you use this option and attempt to perform
11485 floating point operations, the compiler will abort.
11487 @item -mdisable-indexing
11488 @opindex mdisable-indexing
11489 Prevent the compiler from using indexing address modes. This avoids some
11490 rather obscure problems when compiling MIG generated code under MACH@.
11492 @item -mno-space-regs
11493 @opindex mno-space-regs
11494 Generate code that assumes the target has no space registers. This allows
11495 GCC to generate faster indirect calls and use unscaled index address modes.
11497 Such code is suitable for level 0 PA systems and kernels.
11499 @item -mfast-indirect-calls
11500 @opindex mfast-indirect-calls
11501 Generate code that assumes calls never cross space boundaries. This
11502 allows GCC to emit code which performs faster indirect calls.
11504 This option will not work in the presence of shared libraries or nested
11507 @item -mfixed-range=@var{register-range}
11508 @opindex mfixed-range
11509 Generate code treating the given register range as fixed registers.
11510 A fixed register is one that the register allocator can not use. This is
11511 useful when compiling kernel code. A register range is specified as
11512 two registers separated by a dash. Multiple register ranges can be
11513 specified separated by a comma.
11515 @item -mlong-load-store
11516 @opindex mlong-load-store
11517 Generate 3-instruction load and store sequences as sometimes required by
11518 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
11521 @item -mportable-runtime
11522 @opindex mportable-runtime
11523 Use the portable calling conventions proposed by HP for ELF systems.
11527 Enable the use of assembler directives only GAS understands.
11529 @item -mschedule=@var{cpu-type}
11531 Schedule code according to the constraints for the machine type
11532 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
11533 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
11534 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
11535 proper scheduling option for your machine. The default scheduling is
11539 @opindex mlinker-opt
11540 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
11541 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
11542 linkers in which they give bogus error messages when linking some programs.
11545 @opindex msoft-float
11546 Generate output containing library calls for floating point.
11547 @strong{Warning:} the requisite libraries are not available for all HPPA
11548 targets. Normally the facilities of the machine's usual C compiler are
11549 used, but this cannot be done directly in cross-compilation. You must make
11550 your own arrangements to provide suitable library functions for
11553 @option{-msoft-float} changes the calling convention in the output file;
11554 therefore, it is only useful if you compile @emph{all} of a program with
11555 this option. In particular, you need to compile @file{libgcc.a}, the
11556 library that comes with GCC, with @option{-msoft-float} in order for
11561 Generate the predefine, @code{_SIO}, for server IO@. The default is
11562 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
11563 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
11564 options are available under HP-UX and HI-UX@.
11568 Use GNU ld specific options. This passes @option{-shared} to ld when
11569 building a shared library. It is the default when GCC is configured,
11570 explicitly or implicitly, with the GNU linker. This option does not
11571 have any affect on which ld is called, it only changes what parameters
11572 are passed to that ld. The ld that is called is determined by the
11573 @option{--with-ld} configure option, GCC's program search path, and
11574 finally by the user's @env{PATH}. The linker used by GCC can be printed
11575 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
11576 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11580 Use HP ld specific options. This passes @option{-b} to ld when building
11581 a shared library and passes @option{+Accept TypeMismatch} to ld on all
11582 links. It is the default when GCC is configured, explicitly or
11583 implicitly, with the HP linker. This option does not have any affect on
11584 which ld is called, it only changes what parameters are passed to that
11585 ld. The ld that is called is determined by the @option{--with-ld}
11586 configure option, GCC's program search path, and finally by the user's
11587 @env{PATH}. The linker used by GCC can be printed using @samp{which
11588 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
11589 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11592 @opindex mno-long-calls
11593 Generate code that uses long call sequences. This ensures that a call
11594 is always able to reach linker generated stubs. The default is to generate
11595 long calls only when the distance from the call site to the beginning
11596 of the function or translation unit, as the case may be, exceeds a
11597 predefined limit set by the branch type being used. The limits for
11598 normal calls are 7,600,000 and 240,000 bytes, respectively for the
11599 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
11602 Distances are measured from the beginning of functions when using the
11603 @option{-ffunction-sections} option, or when using the @option{-mgas}
11604 and @option{-mno-portable-runtime} options together under HP-UX with
11607 It is normally not desirable to use this option as it will degrade
11608 performance. However, it may be useful in large applications,
11609 particularly when partial linking is used to build the application.
11611 The types of long calls used depends on the capabilities of the
11612 assembler and linker, and the type of code being generated. The
11613 impact on systems that support long absolute calls, and long pic
11614 symbol-difference or pc-relative calls should be relatively small.
11615 However, an indirect call is used on 32-bit ELF systems in pic code
11616 and it is quite long.
11618 @item -munix=@var{unix-std}
11620 Generate compiler predefines and select a startfile for the specified
11621 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
11622 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
11623 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
11624 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
11625 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
11628 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
11629 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
11630 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
11631 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
11632 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
11633 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
11635 It is @emph{important} to note that this option changes the interfaces
11636 for various library routines. It also affects the operational behavior
11637 of the C library. Thus, @emph{extreme} care is needed in using this
11640 Library code that is intended to operate with more than one UNIX
11641 standard must test, set and restore the variable @var{__xpg4_extended_mask}
11642 as appropriate. Most GNU software doesn't provide this capability.
11646 Suppress the generation of link options to search libdld.sl when the
11647 @option{-static} option is specified on HP-UX 10 and later.
11651 The HP-UX implementation of setlocale in libc has a dependency on
11652 libdld.sl. There isn't an archive version of libdld.sl. Thus,
11653 when the @option{-static} option is specified, special link options
11654 are needed to resolve this dependency.
11656 On HP-UX 10 and later, the GCC driver adds the necessary options to
11657 link with libdld.sl when the @option{-static} option is specified.
11658 This causes the resulting binary to be dynamic. On the 64-bit port,
11659 the linkers generate dynamic binaries by default in any case. The
11660 @option{-nolibdld} option can be used to prevent the GCC driver from
11661 adding these link options.
11665 Add support for multithreading with the @dfn{dce thread} library
11666 under HP-UX@. This option sets flags for both the preprocessor and
11670 @node i386 and x86-64 Options
11671 @subsection Intel 386 and AMD x86-64 Options
11672 @cindex i386 Options
11673 @cindex x86-64 Options
11674 @cindex Intel 386 Options
11675 @cindex AMD x86-64 Options
11677 These @samp{-m} options are defined for the i386 and x86-64 family of
11681 @item -mtune=@var{cpu-type}
11683 Tune to @var{cpu-type} everything applicable about the generated code, except
11684 for the ABI and the set of available instructions. The choices for
11685 @var{cpu-type} are:
11688 Produce code optimized for the most common IA32/AMD64/EM64T processors.
11689 If you know the CPU on which your code will run, then you should use
11690 the corresponding @option{-mtune} option instead of
11691 @option{-mtune=generic}. But, if you do not know exactly what CPU users
11692 of your application will have, then you should use this option.
11694 As new processors are deployed in the marketplace, the behavior of this
11695 option will change. Therefore, if you upgrade to a newer version of
11696 GCC, the code generated option will change to reflect the processors
11697 that were most common when that version of GCC was released.
11699 There is no @option{-march=generic} option because @option{-march}
11700 indicates the instruction set the compiler can use, and there is no
11701 generic instruction set applicable to all processors. In contrast,
11702 @option{-mtune} indicates the processor (or, in this case, collection of
11703 processors) for which the code is optimized.
11705 This selects the CPU to tune for at compilation time by determining
11706 the processor type of the compiling machine. Using @option{-mtune=native}
11707 will produce code optimized for the local machine under the constraints
11708 of the selected instruction set. Using @option{-march=native} will
11709 enable all instruction subsets supported by the local machine (hence
11710 the result might not run on different machines).
11712 Original Intel's i386 CPU@.
11714 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
11715 @item i586, pentium
11716 Intel Pentium CPU with no MMX support.
11718 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
11720 Intel PentiumPro CPU@.
11722 Same as @code{generic}, but when used as @code{march} option, PentiumPro
11723 instruction set will be used, so the code will run on all i686 family chips.
11725 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
11726 @item pentium3, pentium3m
11727 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
11730 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
11731 support. Used by Centrino notebooks.
11732 @item pentium4, pentium4m
11733 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
11735 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
11738 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
11739 SSE2 and SSE3 instruction set support.
11741 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
11742 instruction set support.
11744 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
11745 instruction set support.
11747 AMD K6 CPU with MMX instruction set support.
11749 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
11750 @item athlon, athlon-tbird
11751 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
11753 @item athlon-4, athlon-xp, athlon-mp
11754 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
11755 instruction set support.
11756 @item k8, opteron, athlon64, athlon-fx
11757 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
11758 MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit instruction set extensions.)
11759 @item k8-sse3, opteron-sse3, athlon64-sse3
11760 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
11761 @item amdfam10, barcelona
11762 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
11763 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
11764 instruction set extensions.)
11766 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
11769 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
11770 instruction set support.
11772 Via C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
11773 implemented for this chip.)
11775 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
11776 implemented for this chip.)
11778 Embedded AMD CPU with MMX and 3DNow!@: instruction set support.
11781 While picking a specific @var{cpu-type} will schedule things appropriately
11782 for that particular chip, the compiler will not generate any code that
11783 does not run on the i386 without the @option{-march=@var{cpu-type}} option
11786 @item -march=@var{cpu-type}
11788 Generate instructions for the machine type @var{cpu-type}. The choices
11789 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
11790 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
11792 @item -mcpu=@var{cpu-type}
11794 A deprecated synonym for @option{-mtune}.
11796 @item -mfpmath=@var{unit}
11798 Generate floating point arithmetics for selected unit @var{unit}. The choices
11799 for @var{unit} are:
11803 Use the standard 387 floating point coprocessor present majority of chips and
11804 emulated otherwise. Code compiled with this option will run almost everywhere.
11805 The temporary results are computed in 80bit precision instead of precision
11806 specified by the type resulting in slightly different results compared to most
11807 of other chips. See @option{-ffloat-store} for more detailed description.
11809 This is the default choice for i386 compiler.
11812 Use scalar floating point instructions present in the SSE instruction set.
11813 This instruction set is supported by Pentium3 and newer chips, in the AMD line
11814 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
11815 instruction set supports only single precision arithmetics, thus the double and
11816 extended precision arithmetics is still done using 387. Later version, present
11817 only in Pentium4 and the future AMD x86-64 chips supports double precision
11820 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
11821 or @option{-msse2} switches to enable SSE extensions and make this option
11822 effective. For the x86-64 compiler, these extensions are enabled by default.
11824 The resulting code should be considerably faster in the majority of cases and avoid
11825 the numerical instability problems of 387 code, but may break some existing
11826 code that expects temporaries to be 80bit.
11828 This is the default choice for the x86-64 compiler.
11833 Attempt to utilize both instruction sets at once. This effectively double the
11834 amount of available registers and on chips with separate execution units for
11835 387 and SSE the execution resources too. Use this option with care, as it is
11836 still experimental, because the GCC register allocator does not model separate
11837 functional units well resulting in instable performance.
11840 @item -masm=@var{dialect}
11841 @opindex masm=@var{dialect}
11842 Output asm instructions using selected @var{dialect}. Supported
11843 choices are @samp{intel} or @samp{att} (the default one). Darwin does
11844 not support @samp{intel}.
11847 @itemx -mno-ieee-fp
11849 @opindex mno-ieee-fp
11850 Control whether or not the compiler uses IEEE floating point
11851 comparisons. These handle correctly the case where the result of a
11852 comparison is unordered.
11855 @opindex msoft-float
11856 Generate output containing library calls for floating point.
11857 @strong{Warning:} the requisite libraries are not part of GCC@.
11858 Normally the facilities of the machine's usual C compiler are used, but
11859 this can't be done directly in cross-compilation. You must make your
11860 own arrangements to provide suitable library functions for
11863 On machines where a function returns floating point results in the 80387
11864 register stack, some floating point opcodes may be emitted even if
11865 @option{-msoft-float} is used.
11867 @item -mno-fp-ret-in-387
11868 @opindex mno-fp-ret-in-387
11869 Do not use the FPU registers for return values of functions.
11871 The usual calling convention has functions return values of types
11872 @code{float} and @code{double} in an FPU register, even if there
11873 is no FPU@. The idea is that the operating system should emulate
11876 The option @option{-mno-fp-ret-in-387} causes such values to be returned
11877 in ordinary CPU registers instead.
11879 @item -mno-fancy-math-387
11880 @opindex mno-fancy-math-387
11881 Some 387 emulators do not support the @code{sin}, @code{cos} and
11882 @code{sqrt} instructions for the 387. Specify this option to avoid
11883 generating those instructions. This option is the default on FreeBSD,
11884 OpenBSD and NetBSD@. This option is overridden when @option{-march}
11885 indicates that the target cpu will always have an FPU and so the
11886 instruction will not need emulation. As of revision 2.6.1, these
11887 instructions are not generated unless you also use the
11888 @option{-funsafe-math-optimizations} switch.
11890 @item -malign-double
11891 @itemx -mno-align-double
11892 @opindex malign-double
11893 @opindex mno-align-double
11894 Control whether GCC aligns @code{double}, @code{long double}, and
11895 @code{long long} variables on a two word boundary or a one word
11896 boundary. Aligning @code{double} variables on a two word boundary will
11897 produce code that runs somewhat faster on a @samp{Pentium} at the
11898 expense of more memory.
11900 On x86-64, @option{-malign-double} is enabled by default.
11902 @strong{Warning:} if you use the @option{-malign-double} switch,
11903 structures containing the above types will be aligned differently than
11904 the published application binary interface specifications for the 386
11905 and will not be binary compatible with structures in code compiled
11906 without that switch.
11908 @item -m96bit-long-double
11909 @itemx -m128bit-long-double
11910 @opindex m96bit-long-double
11911 @opindex m128bit-long-double
11912 These switches control the size of @code{long double} type. The i386
11913 application binary interface specifies the size to be 96 bits,
11914 so @option{-m96bit-long-double} is the default in 32 bit mode.
11916 Modern architectures (Pentium and newer) would prefer @code{long double}
11917 to be aligned to an 8 or 16 byte boundary. In arrays or structures
11918 conforming to the ABI, this would not be possible. So specifying a
11919 @option{-m128bit-long-double} will align @code{long double}
11920 to a 16 byte boundary by padding the @code{long double} with an additional
11923 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
11924 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
11926 Notice that neither of these options enable any extra precision over the x87
11927 standard of 80 bits for a @code{long double}.
11929 @strong{Warning:} if you override the default value for your target ABI, the
11930 structures and arrays containing @code{long double} variables will change
11931 their size as well as function calling convention for function taking
11932 @code{long double} will be modified. Hence they will not be binary
11933 compatible with arrays or structures in code compiled without that switch.
11935 @item -mlarge-data-threshold=@var{number}
11936 @opindex mlarge-data-threshold=@var{number}
11937 When @option{-mcmodel=medium} is specified, the data greater than
11938 @var{threshold} are placed in large data section. This value must be the
11939 same across all object linked into the binary and defaults to 65535.
11943 Use a different function-calling convention, in which functions that
11944 take a fixed number of arguments return with the @code{ret} @var{num}
11945 instruction, which pops their arguments while returning. This saves one
11946 instruction in the caller since there is no need to pop the arguments
11949 You can specify that an individual function is called with this calling
11950 sequence with the function attribute @samp{stdcall}. You can also
11951 override the @option{-mrtd} option by using the function attribute
11952 @samp{cdecl}. @xref{Function Attributes}.
11954 @strong{Warning:} this calling convention is incompatible with the one
11955 normally used on Unix, so you cannot use it if you need to call
11956 libraries compiled with the Unix compiler.
11958 Also, you must provide function prototypes for all functions that
11959 take variable numbers of arguments (including @code{printf});
11960 otherwise incorrect code will be generated for calls to those
11963 In addition, seriously incorrect code will result if you call a
11964 function with too many arguments. (Normally, extra arguments are
11965 harmlessly ignored.)
11967 @item -mregparm=@var{num}
11969 Control how many registers are used to pass integer arguments. By
11970 default, no registers are used to pass arguments, and at most 3
11971 registers can be used. You can control this behavior for a specific
11972 function by using the function attribute @samp{regparm}.
11973 @xref{Function Attributes}.
11975 @strong{Warning:} if you use this switch, and
11976 @var{num} is nonzero, then you must build all modules with the same
11977 value, including any libraries. This includes the system libraries and
11981 @opindex msseregparm
11982 Use SSE register passing conventions for float and double arguments
11983 and return values. You can control this behavior for a specific
11984 function by using the function attribute @samp{sseregparm}.
11985 @xref{Function Attributes}.
11987 @strong{Warning:} if you use this switch then you must build all
11988 modules with the same value, including any libraries. This includes
11989 the system libraries and startup modules.
11998 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
11999 is specified, the significands of results of floating-point operations are
12000 rounded to 24 bits (single precision); @option{-mpc64} rounds the
12001 significands of results of floating-point operations to 53 bits (double
12002 precision) and @option{-mpc80} rounds the significands of results of
12003 floating-point operations to 64 bits (extended double precision), which is
12004 the default. When this option is used, floating-point operations in higher
12005 precisions are not available to the programmer without setting the FPU
12006 control word explicitly.
12008 Setting the rounding of floating-point operations to less than the default
12009 80 bits can speed some programs by 2% or more. Note that some mathematical
12010 libraries assume that extended precision (80 bit) floating-point operations
12011 are enabled by default; routines in such libraries could suffer significant
12012 loss of accuracy, typically through so-called "catastrophic cancellation",
12013 when this option is used to set the precision to less than extended precision.
12015 @item -mstackrealign
12016 @opindex mstackrealign
12017 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
12018 option will generate an alternate prologue and epilogue that realigns the
12019 runtime stack if necessary. This supports mixing legacy codes that keep
12020 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
12021 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
12022 applicable to individual functions.
12024 @item -mpreferred-stack-boundary=@var{num}
12025 @opindex mpreferred-stack-boundary
12026 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
12027 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
12028 the default is 4 (16 bytes or 128 bits).
12030 @item -mincoming-stack-boundary=@var{num}
12031 @opindex mincoming-stack-boundary
12032 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
12033 boundary. If @option{-mincoming-stack-boundary} is not specified,
12034 the one specified by @option{-mpreferred-stack-boundary} will be used.
12036 On Pentium and PentiumPro, @code{double} and @code{long double} values
12037 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
12038 suffer significant run time performance penalties. On Pentium III, the
12039 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
12040 properly if it is not 16 byte aligned.
12042 To ensure proper alignment of this values on the stack, the stack boundary
12043 must be as aligned as that required by any value stored on the stack.
12044 Further, every function must be generated such that it keeps the stack
12045 aligned. Thus calling a function compiled with a higher preferred
12046 stack boundary from a function compiled with a lower preferred stack
12047 boundary will most likely misalign the stack. It is recommended that
12048 libraries that use callbacks always use the default setting.
12050 This extra alignment does consume extra stack space, and generally
12051 increases code size. Code that is sensitive to stack space usage, such
12052 as embedded systems and operating system kernels, may want to reduce the
12053 preferred alignment to @option{-mpreferred-stack-boundary=2}.
12097 These switches enable or disable the use of instructions in the MMX,
12098 SSE, SSE2, SSE3, SSSE3, SSE4.1, AVX, AES, PCLMUL, SSE4A, FMA4, XOP,
12099 LWP, ABM or 3DNow!@: extended instruction sets.
12100 These extensions are also available as built-in functions: see
12101 @ref{X86 Built-in Functions}, for details of the functions enabled and
12102 disabled by these switches.
12104 To have SSE/SSE2 instructions generated automatically from floating-point
12105 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
12107 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
12108 generates new AVX instructions or AVX equivalence for all SSEx instructions
12111 These options will enable GCC to use these extended instructions in
12112 generated code, even without @option{-mfpmath=sse}. Applications which
12113 perform runtime CPU detection must compile separate files for each
12114 supported architecture, using the appropriate flags. In particular,
12115 the file containing the CPU detection code should be compiled without
12119 @itemx -mno-fused-madd
12120 @opindex mfused-madd
12121 @opindex mno-fused-madd
12122 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
12123 instructions. The default is to use these instructions.
12127 This option instructs GCC to emit a @code{cld} instruction in the prologue
12128 of functions that use string instructions. String instructions depend on
12129 the DF flag to select between autoincrement or autodecrement mode. While the
12130 ABI specifies the DF flag to be cleared on function entry, some operating
12131 systems violate this specification by not clearing the DF flag in their
12132 exception dispatchers. The exception handler can be invoked with the DF flag
12133 set which leads to wrong direction mode, when string instructions are used.
12134 This option can be enabled by default on 32-bit x86 targets by configuring
12135 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
12136 instructions can be suppressed with the @option{-mno-cld} compiler option
12141 This option will enable GCC to use CMPXCHG16B instruction in generated code.
12142 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
12143 data types. This is useful for high resolution counters that could be updated
12144 by multiple processors (or cores). This instruction is generated as part of
12145 atomic built-in functions: see @ref{Atomic Builtins} for details.
12149 This option will enable GCC to use SAHF instruction in generated 64-bit code.
12150 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
12151 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
12152 SAHF are load and store instructions, respectively, for certain status flags.
12153 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
12154 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
12158 This option will enable GCC to use movbe instruction to implement
12159 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
12163 This option will enable built-in functions, @code{__builtin_ia32_crc32qi},
12164 @code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and
12165 @code{__builtin_ia32_crc32di} to generate the crc32 machine instruction.
12169 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
12170 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
12171 to increase precision instead of DIVSS and SQRTSS (and their vectorized
12172 variants) for single precision floating point arguments. These instructions
12173 are generated only when @option{-funsafe-math-optimizations} is enabled
12174 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
12175 Note that while the throughput of the sequence is higher than the throughput
12176 of the non-reciprocal instruction, the precision of the sequence can be
12177 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
12179 Note that GCC implements 1.0f/sqrtf(x) in terms of RSQRTSS (or RSQRTPS)
12180 already with @option{-ffast-math} (or the above option combination), and
12181 doesn't need @option{-mrecip}.
12183 @item -mveclibabi=@var{type}
12184 @opindex mveclibabi
12185 Specifies the ABI type to use for vectorizing intrinsics using an
12186 external library. Supported types are @code{svml} for the Intel short
12187 vector math library and @code{acml} for the AMD math core library style
12188 of interfacing. GCC will currently emit calls to @code{vmldExp2},
12189 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
12190 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
12191 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
12192 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
12193 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
12194 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
12195 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
12196 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
12197 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
12198 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
12199 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
12200 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
12201 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
12202 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
12203 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
12204 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
12205 compatible library will have to be specified at link time.
12207 @item -mabi=@var{name}
12209 Generate code for the specified calling convention. Permissible values
12210 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
12211 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
12212 ABI when targeting Windows. On all other systems, the default is the
12213 SYSV ABI. You can control this behavior for a specific function by
12214 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
12215 @xref{Function Attributes}.
12218 @itemx -mno-push-args
12219 @opindex mpush-args
12220 @opindex mno-push-args
12221 Use PUSH operations to store outgoing parameters. This method is shorter
12222 and usually equally fast as method using SUB/MOV operations and is enabled
12223 by default. In some cases disabling it may improve performance because of
12224 improved scheduling and reduced dependencies.
12226 @item -maccumulate-outgoing-args
12227 @opindex maccumulate-outgoing-args
12228 If enabled, the maximum amount of space required for outgoing arguments will be
12229 computed in the function prologue. This is faster on most modern CPUs
12230 because of reduced dependencies, improved scheduling and reduced stack usage
12231 when preferred stack boundary is not equal to 2. The drawback is a notable
12232 increase in code size. This switch implies @option{-mno-push-args}.
12236 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
12237 on thread-safe exception handling must compile and link all code with the
12238 @option{-mthreads} option. When compiling, @option{-mthreads} defines
12239 @option{-D_MT}; when linking, it links in a special thread helper library
12240 @option{-lmingwthrd} which cleans up per thread exception handling data.
12242 @item -mno-align-stringops
12243 @opindex mno-align-stringops
12244 Do not align destination of inlined string operations. This switch reduces
12245 code size and improves performance in case the destination is already aligned,
12246 but GCC doesn't know about it.
12248 @item -minline-all-stringops
12249 @opindex minline-all-stringops
12250 By default GCC inlines string operations only when destination is known to be
12251 aligned at least to 4 byte boundary. This enables more inlining, increase code
12252 size, but may improve performance of code that depends on fast memcpy, strlen
12253 and memset for short lengths.
12255 @item -minline-stringops-dynamically
12256 @opindex minline-stringops-dynamically
12257 For string operation of unknown size, inline runtime checks so for small
12258 blocks inline code is used, while for large blocks library call is used.
12260 @item -mstringop-strategy=@var{alg}
12261 @opindex mstringop-strategy=@var{alg}
12262 Overwrite internal decision heuristic about particular algorithm to inline
12263 string operation with. The allowed values are @code{rep_byte},
12264 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
12265 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
12266 expanding inline loop, @code{libcall} for always expanding library call.
12268 @item -momit-leaf-frame-pointer
12269 @opindex momit-leaf-frame-pointer
12270 Don't keep the frame pointer in a register for leaf functions. This
12271 avoids the instructions to save, set up and restore frame pointers and
12272 makes an extra register available in leaf functions. The option
12273 @option{-fomit-frame-pointer} removes the frame pointer for all functions
12274 which might make debugging harder.
12276 @item -mtls-direct-seg-refs
12277 @itemx -mno-tls-direct-seg-refs
12278 @opindex mtls-direct-seg-refs
12279 Controls whether TLS variables may be accessed with offsets from the
12280 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
12281 or whether the thread base pointer must be added. Whether or not this
12282 is legal depends on the operating system, and whether it maps the
12283 segment to cover the entire TLS area.
12285 For systems that use GNU libc, the default is on.
12288 @itemx -mno-sse2avx
12290 Specify that the assembler should encode SSE instructions with VEX
12291 prefix. The option @option{-mavx} turns this on by default.
12294 These @samp{-m} switches are supported in addition to the above
12295 on AMD x86-64 processors in 64-bit environments.
12302 Generate code for a 32-bit or 64-bit environment.
12303 The 32-bit environment sets int, long and pointer to 32 bits and
12304 generates code that runs on any i386 system.
12305 The 64-bit environment sets int to 32 bits and long and pointer
12306 to 64 bits and generates code for AMD's x86-64 architecture. For
12307 darwin only the -m64 option turns off the @option{-fno-pic} and
12308 @option{-mdynamic-no-pic} options.
12310 @item -mno-red-zone
12311 @opindex mno-red-zone
12312 Do not use a so called red zone for x86-64 code. The red zone is mandated
12313 by the x86-64 ABI, it is a 128-byte area beyond the location of the
12314 stack pointer that will not be modified by signal or interrupt handlers
12315 and therefore can be used for temporary data without adjusting the stack
12316 pointer. The flag @option{-mno-red-zone} disables this red zone.
12318 @item -mcmodel=small
12319 @opindex mcmodel=small
12320 Generate code for the small code model: the program and its symbols must
12321 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
12322 Programs can be statically or dynamically linked. This is the default
12325 @item -mcmodel=kernel
12326 @opindex mcmodel=kernel
12327 Generate code for the kernel code model. The kernel runs in the
12328 negative 2 GB of the address space.
12329 This model has to be used for Linux kernel code.
12331 @item -mcmodel=medium
12332 @opindex mcmodel=medium
12333 Generate code for the medium model: The program is linked in the lower 2
12334 GB of the address space. Small symbols are also placed there. Symbols
12335 with sizes larger than @option{-mlarge-data-threshold} are put into
12336 large data or bss sections and can be located above 2GB. Programs can
12337 be statically or dynamically linked.
12339 @item -mcmodel=large
12340 @opindex mcmodel=large
12341 Generate code for the large model: This model makes no assumptions
12342 about addresses and sizes of sections.
12345 @node IA-64 Options
12346 @subsection IA-64 Options
12347 @cindex IA-64 Options
12349 These are the @samp{-m} options defined for the Intel IA-64 architecture.
12353 @opindex mbig-endian
12354 Generate code for a big endian target. This is the default for HP-UX@.
12356 @item -mlittle-endian
12357 @opindex mlittle-endian
12358 Generate code for a little endian target. This is the default for AIX5
12364 @opindex mno-gnu-as
12365 Generate (or don't) code for the GNU assembler. This is the default.
12366 @c Also, this is the default if the configure option @option{--with-gnu-as}
12372 @opindex mno-gnu-ld
12373 Generate (or don't) code for the GNU linker. This is the default.
12374 @c Also, this is the default if the configure option @option{--with-gnu-ld}
12379 Generate code that does not use a global pointer register. The result
12380 is not position independent code, and violates the IA-64 ABI@.
12382 @item -mvolatile-asm-stop
12383 @itemx -mno-volatile-asm-stop
12384 @opindex mvolatile-asm-stop
12385 @opindex mno-volatile-asm-stop
12386 Generate (or don't) a stop bit immediately before and after volatile asm
12389 @item -mregister-names
12390 @itemx -mno-register-names
12391 @opindex mregister-names
12392 @opindex mno-register-names
12393 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
12394 the stacked registers. This may make assembler output more readable.
12400 Disable (or enable) optimizations that use the small data section. This may
12401 be useful for working around optimizer bugs.
12403 @item -mconstant-gp
12404 @opindex mconstant-gp
12405 Generate code that uses a single constant global pointer value. This is
12406 useful when compiling kernel code.
12410 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
12411 This is useful when compiling firmware code.
12413 @item -minline-float-divide-min-latency
12414 @opindex minline-float-divide-min-latency
12415 Generate code for inline divides of floating point values
12416 using the minimum latency algorithm.
12418 @item -minline-float-divide-max-throughput
12419 @opindex minline-float-divide-max-throughput
12420 Generate code for inline divides of floating point values
12421 using the maximum throughput algorithm.
12423 @item -mno-inline-float-divide
12424 @opindex mno-inline-float-divide
12425 Do not generate inline code for divides of floating point values.
12427 @item -minline-int-divide-min-latency
12428 @opindex minline-int-divide-min-latency
12429 Generate code for inline divides of integer values
12430 using the minimum latency algorithm.
12432 @item -minline-int-divide-max-throughput
12433 @opindex minline-int-divide-max-throughput
12434 Generate code for inline divides of integer values
12435 using the maximum throughput algorithm.
12437 @item -mno-inline-int-divide
12438 @opindex mno-inline-int-divide
12439 Do not generate inline code for divides of integer values.
12441 @item -minline-sqrt-min-latency
12442 @opindex minline-sqrt-min-latency
12443 Generate code for inline square roots
12444 using the minimum latency algorithm.
12446 @item -minline-sqrt-max-throughput
12447 @opindex minline-sqrt-max-throughput
12448 Generate code for inline square roots
12449 using the maximum throughput algorithm.
12451 @item -mno-inline-sqrt
12452 @opindex mno-inline-sqrt
12453 Do not generate inline code for sqrt.
12456 @itemx -mno-fused-madd
12457 @opindex mfused-madd
12458 @opindex mno-fused-madd
12459 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
12460 instructions. The default is to use these instructions.
12462 @item -mno-dwarf2-asm
12463 @itemx -mdwarf2-asm
12464 @opindex mno-dwarf2-asm
12465 @opindex mdwarf2-asm
12466 Don't (or do) generate assembler code for the DWARF2 line number debugging
12467 info. This may be useful when not using the GNU assembler.
12469 @item -mearly-stop-bits
12470 @itemx -mno-early-stop-bits
12471 @opindex mearly-stop-bits
12472 @opindex mno-early-stop-bits
12473 Allow stop bits to be placed earlier than immediately preceding the
12474 instruction that triggered the stop bit. This can improve instruction
12475 scheduling, but does not always do so.
12477 @item -mfixed-range=@var{register-range}
12478 @opindex mfixed-range
12479 Generate code treating the given register range as fixed registers.
12480 A fixed register is one that the register allocator can not use. This is
12481 useful when compiling kernel code. A register range is specified as
12482 two registers separated by a dash. Multiple register ranges can be
12483 specified separated by a comma.
12485 @item -mtls-size=@var{tls-size}
12487 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
12490 @item -mtune=@var{cpu-type}
12492 Tune the instruction scheduling for a particular CPU, Valid values are
12493 itanium, itanium1, merced, itanium2, and mckinley.
12499 Generate code for a 32-bit or 64-bit environment.
12500 The 32-bit environment sets int, long and pointer to 32 bits.
12501 The 64-bit environment sets int to 32 bits and long and pointer
12502 to 64 bits. These are HP-UX specific flags.
12504 @item -mno-sched-br-data-spec
12505 @itemx -msched-br-data-spec
12506 @opindex mno-sched-br-data-spec
12507 @opindex msched-br-data-spec
12508 (Dis/En)able data speculative scheduling before reload.
12509 This will result in generation of the ld.a instructions and
12510 the corresponding check instructions (ld.c / chk.a).
12511 The default is 'disable'.
12513 @item -msched-ar-data-spec
12514 @itemx -mno-sched-ar-data-spec
12515 @opindex msched-ar-data-spec
12516 @opindex mno-sched-ar-data-spec
12517 (En/Dis)able data speculative scheduling after reload.
12518 This will result in generation of the ld.a instructions and
12519 the corresponding check instructions (ld.c / chk.a).
12520 The default is 'enable'.
12522 @item -mno-sched-control-spec
12523 @itemx -msched-control-spec
12524 @opindex mno-sched-control-spec
12525 @opindex msched-control-spec
12526 (Dis/En)able control speculative scheduling. This feature is
12527 available only during region scheduling (i.e.@: before reload).
12528 This will result in generation of the ld.s instructions and
12529 the corresponding check instructions chk.s .
12530 The default is 'disable'.
12532 @item -msched-br-in-data-spec
12533 @itemx -mno-sched-br-in-data-spec
12534 @opindex msched-br-in-data-spec
12535 @opindex mno-sched-br-in-data-spec
12536 (En/Dis)able speculative scheduling of the instructions that
12537 are dependent on the data speculative loads before reload.
12538 This is effective only with @option{-msched-br-data-spec} enabled.
12539 The default is 'enable'.
12541 @item -msched-ar-in-data-spec
12542 @itemx -mno-sched-ar-in-data-spec
12543 @opindex msched-ar-in-data-spec
12544 @opindex mno-sched-ar-in-data-spec
12545 (En/Dis)able speculative scheduling of the instructions that
12546 are dependent on the data speculative loads after reload.
12547 This is effective only with @option{-msched-ar-data-spec} enabled.
12548 The default is 'enable'.
12550 @item -msched-in-control-spec
12551 @itemx -mno-sched-in-control-spec
12552 @opindex msched-in-control-spec
12553 @opindex mno-sched-in-control-spec
12554 (En/Dis)able speculative scheduling of the instructions that
12555 are dependent on the control speculative loads.
12556 This is effective only with @option{-msched-control-spec} enabled.
12557 The default is 'enable'.
12559 @item -mno-sched-prefer-non-data-spec-insns
12560 @itemx -msched-prefer-non-data-spec-insns
12561 @opindex mno-sched-prefer-non-data-spec-insns
12562 @opindex msched-prefer-non-data-spec-insns
12563 If enabled, data speculative instructions will be chosen for schedule
12564 only if there are no other choices at the moment. This will make
12565 the use of the data speculation much more conservative.
12566 The default is 'disable'.
12568 @item -mno-sched-prefer-non-control-spec-insns
12569 @itemx -msched-prefer-non-control-spec-insns
12570 @opindex mno-sched-prefer-non-control-spec-insns
12571 @opindex msched-prefer-non-control-spec-insns
12572 If enabled, control speculative instructions will be chosen for schedule
12573 only if there are no other choices at the moment. This will make
12574 the use of the control speculation much more conservative.
12575 The default is 'disable'.
12577 @item -mno-sched-count-spec-in-critical-path
12578 @itemx -msched-count-spec-in-critical-path
12579 @opindex mno-sched-count-spec-in-critical-path
12580 @opindex msched-count-spec-in-critical-path
12581 If enabled, speculative dependencies will be considered during
12582 computation of the instructions priorities. This will make the use of the
12583 speculation a bit more conservative.
12584 The default is 'disable'.
12586 @item -msched-spec-ldc
12587 @opindex msched-spec-ldc
12588 Use a simple data speculation check. This option is on by default.
12590 @item -msched-control-spec-ldc
12591 @opindex msched-spec-ldc
12592 Use a simple check for control speculation. This option is on by default.
12594 @item -msched-stop-bits-after-every-cycle
12595 @opindex msched-stop-bits-after-every-cycle
12596 Place a stop bit after every cycle when scheduling. This option is on
12599 @item -msched-fp-mem-deps-zero-cost
12600 @opindex msched-fp-mem-deps-zero-cost
12601 Assume that floating-point stores and loads are not likely to cause a conflict
12602 when placed into the same instruction group. This option is disabled by
12605 @item -msel-sched-dont-check-control-spec
12606 @opindex msel-sched-dont-check-control-spec
12607 Generate checks for control speculation in selective scheduling.
12608 This flag is disabled by default.
12610 @item -msched-max-memory-insns=@var{max-insns}
12611 @opindex msched-max-memory-insns
12612 Limit on the number of memory insns per instruction group, giving lower
12613 priority to subsequent memory insns attempting to schedule in the same
12614 instruction group. Frequently useful to prevent cache bank conflicts.
12615 The default value is 1.
12617 @item -msched-max-memory-insns-hard-limit
12618 @opindex msched-max-memory-insns-hard-limit
12619 Disallow more than `msched-max-memory-insns' in instruction group.
12620 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
12621 when limit is reached but may still schedule memory operations.
12625 @node IA-64/VMS Options
12626 @subsection IA-64/VMS Options
12628 These @samp{-m} options are defined for the IA-64/VMS implementations:
12631 @item -mvms-return-codes
12632 @opindex mvms-return-codes
12633 Return VMS condition codes from main. The default is to return POSIX
12634 style condition (e.g.@ error) codes.
12636 @item -mdebug-main=@var{prefix}
12637 @opindex mdebug-main=@var{prefix}
12638 Flag the first routine whose name starts with @var{prefix} as the main
12639 routine for the debugger.
12643 Default to 64bit memory allocation routines.
12647 @subsection LM32 Options
12648 @cindex LM32 options
12650 These @option{-m} options are defined for the Lattice Mico32 architecture:
12653 @item -mbarrel-shift-enabled
12654 @opindex mbarrel-shift-enabled
12655 Enable barrel-shift instructions.
12657 @item -mdivide-enabled
12658 @opindex mdivide-enabled
12659 Enable divide and modulus instructions.
12661 @item -mmultiply-enabled
12662 @opindex multiply-enabled
12663 Enable multiply instructions.
12665 @item -msign-extend-enabled
12666 @opindex msign-extend-enabled
12667 Enable sign extend instructions.
12669 @item -muser-enabled
12670 @opindex muser-enabled
12671 Enable user-defined instructions.
12676 @subsection M32C Options
12677 @cindex M32C options
12680 @item -mcpu=@var{name}
12682 Select the CPU for which code is generated. @var{name} may be one of
12683 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
12684 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
12685 the M32C/80 series.
12689 Specifies that the program will be run on the simulator. This causes
12690 an alternate runtime library to be linked in which supports, for
12691 example, file I/O@. You must not use this option when generating
12692 programs that will run on real hardware; you must provide your own
12693 runtime library for whatever I/O functions are needed.
12695 @item -memregs=@var{number}
12697 Specifies the number of memory-based pseudo-registers GCC will use
12698 during code generation. These pseudo-registers will be used like real
12699 registers, so there is a tradeoff between GCC's ability to fit the
12700 code into available registers, and the performance penalty of using
12701 memory instead of registers. Note that all modules in a program must
12702 be compiled with the same value for this option. Because of that, you
12703 must not use this option with the default runtime libraries gcc
12708 @node M32R/D Options
12709 @subsection M32R/D Options
12710 @cindex M32R/D options
12712 These @option{-m} options are defined for Renesas M32R/D architectures:
12717 Generate code for the M32R/2@.
12721 Generate code for the M32R/X@.
12725 Generate code for the M32R@. This is the default.
12727 @item -mmodel=small
12728 @opindex mmodel=small
12729 Assume all objects live in the lower 16MB of memory (so that their addresses
12730 can be loaded with the @code{ld24} instruction), and assume all subroutines
12731 are reachable with the @code{bl} instruction.
12732 This is the default.
12734 The addressability of a particular object can be set with the
12735 @code{model} attribute.
12737 @item -mmodel=medium
12738 @opindex mmodel=medium
12739 Assume objects may be anywhere in the 32-bit address space (the compiler
12740 will generate @code{seth/add3} instructions to load their addresses), and
12741 assume all subroutines are reachable with the @code{bl} instruction.
12743 @item -mmodel=large
12744 @opindex mmodel=large
12745 Assume objects may be anywhere in the 32-bit address space (the compiler
12746 will generate @code{seth/add3} instructions to load their addresses), and
12747 assume subroutines may not be reachable with the @code{bl} instruction
12748 (the compiler will generate the much slower @code{seth/add3/jl}
12749 instruction sequence).
12752 @opindex msdata=none
12753 Disable use of the small data area. Variables will be put into
12754 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
12755 @code{section} attribute has been specified).
12756 This is the default.
12758 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
12759 Objects may be explicitly put in the small data area with the
12760 @code{section} attribute using one of these sections.
12762 @item -msdata=sdata
12763 @opindex msdata=sdata
12764 Put small global and static data in the small data area, but do not
12765 generate special code to reference them.
12768 @opindex msdata=use
12769 Put small global and static data in the small data area, and generate
12770 special instructions to reference them.
12774 @cindex smaller data references
12775 Put global and static objects less than or equal to @var{num} bytes
12776 into the small data or bss sections instead of the normal data or bss
12777 sections. The default value of @var{num} is 8.
12778 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
12779 for this option to have any effect.
12781 All modules should be compiled with the same @option{-G @var{num}} value.
12782 Compiling with different values of @var{num} may or may not work; if it
12783 doesn't the linker will give an error message---incorrect code will not be
12788 Makes the M32R specific code in the compiler display some statistics
12789 that might help in debugging programs.
12791 @item -malign-loops
12792 @opindex malign-loops
12793 Align all loops to a 32-byte boundary.
12795 @item -mno-align-loops
12796 @opindex mno-align-loops
12797 Do not enforce a 32-byte alignment for loops. This is the default.
12799 @item -missue-rate=@var{number}
12800 @opindex missue-rate=@var{number}
12801 Issue @var{number} instructions per cycle. @var{number} can only be 1
12804 @item -mbranch-cost=@var{number}
12805 @opindex mbranch-cost=@var{number}
12806 @var{number} can only be 1 or 2. If it is 1 then branches will be
12807 preferred over conditional code, if it is 2, then the opposite will
12810 @item -mflush-trap=@var{number}
12811 @opindex mflush-trap=@var{number}
12812 Specifies the trap number to use to flush the cache. The default is
12813 12. Valid numbers are between 0 and 15 inclusive.
12815 @item -mno-flush-trap
12816 @opindex mno-flush-trap
12817 Specifies that the cache cannot be flushed by using a trap.
12819 @item -mflush-func=@var{name}
12820 @opindex mflush-func=@var{name}
12821 Specifies the name of the operating system function to call to flush
12822 the cache. The default is @emph{_flush_cache}, but a function call
12823 will only be used if a trap is not available.
12825 @item -mno-flush-func
12826 @opindex mno-flush-func
12827 Indicates that there is no OS function for flushing the cache.
12831 @node M680x0 Options
12832 @subsection M680x0 Options
12833 @cindex M680x0 options
12835 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
12836 The default settings depend on which architecture was selected when
12837 the compiler was configured; the defaults for the most common choices
12841 @item -march=@var{arch}
12843 Generate code for a specific M680x0 or ColdFire instruction set
12844 architecture. Permissible values of @var{arch} for M680x0
12845 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
12846 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
12847 architectures are selected according to Freescale's ISA classification
12848 and the permissible values are: @samp{isaa}, @samp{isaaplus},
12849 @samp{isab} and @samp{isac}.
12851 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
12852 code for a ColdFire target. The @var{arch} in this macro is one of the
12853 @option{-march} arguments given above.
12855 When used together, @option{-march} and @option{-mtune} select code
12856 that runs on a family of similar processors but that is optimized
12857 for a particular microarchitecture.
12859 @item -mcpu=@var{cpu}
12861 Generate code for a specific M680x0 or ColdFire processor.
12862 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
12863 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
12864 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
12865 below, which also classifies the CPUs into families:
12867 @multitable @columnfractions 0.20 0.80
12868 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
12869 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
12870 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
12871 @item @samp{5206e} @tab @samp{5206e}
12872 @item @samp{5208} @tab @samp{5207} @samp{5208}
12873 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
12874 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
12875 @item @samp{5216} @tab @samp{5214} @samp{5216}
12876 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
12877 @item @samp{5225} @tab @samp{5224} @samp{5225}
12878 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
12879 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
12880 @item @samp{5249} @tab @samp{5249}
12881 @item @samp{5250} @tab @samp{5250}
12882 @item @samp{5271} @tab @samp{5270} @samp{5271}
12883 @item @samp{5272} @tab @samp{5272}
12884 @item @samp{5275} @tab @samp{5274} @samp{5275}
12885 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
12886 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
12887 @item @samp{5307} @tab @samp{5307}
12888 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
12889 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
12890 @item @samp{5407} @tab @samp{5407}
12891 @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}
12894 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
12895 @var{arch} is compatible with @var{cpu}. Other combinations of
12896 @option{-mcpu} and @option{-march} are rejected.
12898 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
12899 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
12900 where the value of @var{family} is given by the table above.
12902 @item -mtune=@var{tune}
12904 Tune the code for a particular microarchitecture, within the
12905 constraints set by @option{-march} and @option{-mcpu}.
12906 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
12907 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
12908 and @samp{cpu32}. The ColdFire microarchitectures
12909 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
12911 You can also use @option{-mtune=68020-40} for code that needs
12912 to run relatively well on 68020, 68030 and 68040 targets.
12913 @option{-mtune=68020-60} is similar but includes 68060 targets
12914 as well. These two options select the same tuning decisions as
12915 @option{-m68020-40} and @option{-m68020-60} respectively.
12917 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
12918 when tuning for 680x0 architecture @var{arch}. It also defines
12919 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
12920 option is used. If gcc is tuning for a range of architectures,
12921 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
12922 it defines the macros for every architecture in the range.
12924 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
12925 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
12926 of the arguments given above.
12932 Generate output for a 68000. This is the default
12933 when the compiler is configured for 68000-based systems.
12934 It is equivalent to @option{-march=68000}.
12936 Use this option for microcontrollers with a 68000 or EC000 core,
12937 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
12941 Generate output for a 68010. This is the default
12942 when the compiler is configured for 68010-based systems.
12943 It is equivalent to @option{-march=68010}.
12949 Generate output for a 68020. This is the default
12950 when the compiler is configured for 68020-based systems.
12951 It is equivalent to @option{-march=68020}.
12955 Generate output for a 68030. This is the default when the compiler is
12956 configured for 68030-based systems. It is equivalent to
12957 @option{-march=68030}.
12961 Generate output for a 68040. This is the default when the compiler is
12962 configured for 68040-based systems. It is equivalent to
12963 @option{-march=68040}.
12965 This option inhibits the use of 68881/68882 instructions that have to be
12966 emulated by software on the 68040. Use this option if your 68040 does not
12967 have code to emulate those instructions.
12971 Generate output for a 68060. This is the default when the compiler is
12972 configured for 68060-based systems. It is equivalent to
12973 @option{-march=68060}.
12975 This option inhibits the use of 68020 and 68881/68882 instructions that
12976 have to be emulated by software on the 68060. Use this option if your 68060
12977 does not have code to emulate those instructions.
12981 Generate output for a CPU32. This is the default
12982 when the compiler is configured for CPU32-based systems.
12983 It is equivalent to @option{-march=cpu32}.
12985 Use this option for microcontrollers with a
12986 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
12987 68336, 68340, 68341, 68349 and 68360.
12991 Generate output for a 520X ColdFire CPU@. This is the default
12992 when the compiler is configured for 520X-based systems.
12993 It is equivalent to @option{-mcpu=5206}, and is now deprecated
12994 in favor of that option.
12996 Use this option for microcontroller with a 5200 core, including
12997 the MCF5202, MCF5203, MCF5204 and MCF5206.
13001 Generate output for a 5206e ColdFire CPU@. The option is now
13002 deprecated in favor of the equivalent @option{-mcpu=5206e}.
13006 Generate output for a member of the ColdFire 528X family.
13007 The option is now deprecated in favor of the equivalent
13008 @option{-mcpu=528x}.
13012 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
13013 in favor of the equivalent @option{-mcpu=5307}.
13017 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
13018 in favor of the equivalent @option{-mcpu=5407}.
13022 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
13023 This includes use of hardware floating point instructions.
13024 The option is equivalent to @option{-mcpu=547x}, and is now
13025 deprecated in favor of that option.
13029 Generate output for a 68040, without using any of the new instructions.
13030 This results in code which can run relatively efficiently on either a
13031 68020/68881 or a 68030 or a 68040. The generated code does use the
13032 68881 instructions that are emulated on the 68040.
13034 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
13038 Generate output for a 68060, without using any of the new instructions.
13039 This results in code which can run relatively efficiently on either a
13040 68020/68881 or a 68030 or a 68040. The generated code does use the
13041 68881 instructions that are emulated on the 68060.
13043 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
13047 @opindex mhard-float
13049 Generate floating-point instructions. This is the default for 68020
13050 and above, and for ColdFire devices that have an FPU@. It defines the
13051 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
13052 on ColdFire targets.
13055 @opindex msoft-float
13056 Do not generate floating-point instructions; use library calls instead.
13057 This is the default for 68000, 68010, and 68832 targets. It is also
13058 the default for ColdFire devices that have no FPU.
13064 Generate (do not generate) ColdFire hardware divide and remainder
13065 instructions. If @option{-march} is used without @option{-mcpu},
13066 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
13067 architectures. Otherwise, the default is taken from the target CPU
13068 (either the default CPU, or the one specified by @option{-mcpu}). For
13069 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
13070 @option{-mcpu=5206e}.
13072 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
13076 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13077 Additionally, parameters passed on the stack are also aligned to a
13078 16-bit boundary even on targets whose API mandates promotion to 32-bit.
13082 Do not consider type @code{int} to be 16 bits wide. This is the default.
13085 @itemx -mno-bitfield
13086 @opindex mnobitfield
13087 @opindex mno-bitfield
13088 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
13089 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
13093 Do use the bit-field instructions. The @option{-m68020} option implies
13094 @option{-mbitfield}. This is the default if you use a configuration
13095 designed for a 68020.
13099 Use a different function-calling convention, in which functions
13100 that take a fixed number of arguments return with the @code{rtd}
13101 instruction, which pops their arguments while returning. This
13102 saves one instruction in the caller since there is no need to pop
13103 the arguments there.
13105 This calling convention is incompatible with the one normally
13106 used on Unix, so you cannot use it if you need to call libraries
13107 compiled with the Unix compiler.
13109 Also, you must provide function prototypes for all functions that
13110 take variable numbers of arguments (including @code{printf});
13111 otherwise incorrect code will be generated for calls to those
13114 In addition, seriously incorrect code will result if you call a
13115 function with too many arguments. (Normally, extra arguments are
13116 harmlessly ignored.)
13118 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
13119 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
13123 Do not use the calling conventions selected by @option{-mrtd}.
13124 This is the default.
13127 @itemx -mno-align-int
13128 @opindex malign-int
13129 @opindex mno-align-int
13130 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
13131 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
13132 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
13133 Aligning variables on 32-bit boundaries produces code that runs somewhat
13134 faster on processors with 32-bit busses at the expense of more memory.
13136 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
13137 align structures containing the above types differently than
13138 most published application binary interface specifications for the m68k.
13142 Use the pc-relative addressing mode of the 68000 directly, instead of
13143 using a global offset table. At present, this option implies @option{-fpic},
13144 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
13145 not presently supported with @option{-mpcrel}, though this could be supported for
13146 68020 and higher processors.
13148 @item -mno-strict-align
13149 @itemx -mstrict-align
13150 @opindex mno-strict-align
13151 @opindex mstrict-align
13152 Do not (do) assume that unaligned memory references will be handled by
13156 Generate code that allows the data segment to be located in a different
13157 area of memory from the text segment. This allows for execute in place in
13158 an environment without virtual memory management. This option implies
13161 @item -mno-sep-data
13162 Generate code that assumes that the data segment follows the text segment.
13163 This is the default.
13165 @item -mid-shared-library
13166 Generate code that supports shared libraries via the library ID method.
13167 This allows for execute in place and shared libraries in an environment
13168 without virtual memory management. This option implies @option{-fPIC}.
13170 @item -mno-id-shared-library
13171 Generate code that doesn't assume ID based shared libraries are being used.
13172 This is the default.
13174 @item -mshared-library-id=n
13175 Specified the identification number of the ID based shared library being
13176 compiled. Specifying a value of 0 will generate more compact code, specifying
13177 other values will force the allocation of that number to the current
13178 library but is no more space or time efficient than omitting this option.
13184 When generating position-independent code for ColdFire, generate code
13185 that works if the GOT has more than 8192 entries. This code is
13186 larger and slower than code generated without this option. On M680x0
13187 processors, this option is not needed; @option{-fPIC} suffices.
13189 GCC normally uses a single instruction to load values from the GOT@.
13190 While this is relatively efficient, it only works if the GOT
13191 is smaller than about 64k. Anything larger causes the linker
13192 to report an error such as:
13194 @cindex relocation truncated to fit (ColdFire)
13196 relocation truncated to fit: R_68K_GOT16O foobar
13199 If this happens, you should recompile your code with @option{-mxgot}.
13200 It should then work with very large GOTs. However, code generated with
13201 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
13202 the value of a global symbol.
13204 Note that some linkers, including newer versions of the GNU linker,
13205 can create multiple GOTs and sort GOT entries. If you have such a linker,
13206 you should only need to use @option{-mxgot} when compiling a single
13207 object file that accesses more than 8192 GOT entries. Very few do.
13209 These options have no effect unless GCC is generating
13210 position-independent code.
13214 @node M68hc1x Options
13215 @subsection M68hc1x Options
13216 @cindex M68hc1x options
13218 These are the @samp{-m} options defined for the 68hc11 and 68hc12
13219 microcontrollers. The default values for these options depends on
13220 which style of microcontroller was selected when the compiler was configured;
13221 the defaults for the most common choices are given below.
13228 Generate output for a 68HC11. This is the default
13229 when the compiler is configured for 68HC11-based systems.
13235 Generate output for a 68HC12. This is the default
13236 when the compiler is configured for 68HC12-based systems.
13242 Generate output for a 68HCS12.
13244 @item -mauto-incdec
13245 @opindex mauto-incdec
13246 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
13253 Enable the use of 68HC12 min and max instructions.
13256 @itemx -mno-long-calls
13257 @opindex mlong-calls
13258 @opindex mno-long-calls
13259 Treat all calls as being far away (near). If calls are assumed to be
13260 far away, the compiler will use the @code{call} instruction to
13261 call a function and the @code{rtc} instruction for returning.
13265 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13267 @item -msoft-reg-count=@var{count}
13268 @opindex msoft-reg-count
13269 Specify the number of pseudo-soft registers which are used for the
13270 code generation. The maximum number is 32. Using more pseudo-soft
13271 register may or may not result in better code depending on the program.
13272 The default is 4 for 68HC11 and 2 for 68HC12.
13276 @node MCore Options
13277 @subsection MCore Options
13278 @cindex MCore options
13280 These are the @samp{-m} options defined for the Motorola M*Core
13286 @itemx -mno-hardlit
13288 @opindex mno-hardlit
13289 Inline constants into the code stream if it can be done in two
13290 instructions or less.
13296 Use the divide instruction. (Enabled by default).
13298 @item -mrelax-immediate
13299 @itemx -mno-relax-immediate
13300 @opindex mrelax-immediate
13301 @opindex mno-relax-immediate
13302 Allow arbitrary sized immediates in bit operations.
13304 @item -mwide-bitfields
13305 @itemx -mno-wide-bitfields
13306 @opindex mwide-bitfields
13307 @opindex mno-wide-bitfields
13308 Always treat bit-fields as int-sized.
13310 @item -m4byte-functions
13311 @itemx -mno-4byte-functions
13312 @opindex m4byte-functions
13313 @opindex mno-4byte-functions
13314 Force all functions to be aligned to a four byte boundary.
13316 @item -mcallgraph-data
13317 @itemx -mno-callgraph-data
13318 @opindex mcallgraph-data
13319 @opindex mno-callgraph-data
13320 Emit callgraph information.
13323 @itemx -mno-slow-bytes
13324 @opindex mslow-bytes
13325 @opindex mno-slow-bytes
13326 Prefer word access when reading byte quantities.
13328 @item -mlittle-endian
13329 @itemx -mbig-endian
13330 @opindex mlittle-endian
13331 @opindex mbig-endian
13332 Generate code for a little endian target.
13338 Generate code for the 210 processor.
13342 Assume that run-time support has been provided and so omit the
13343 simulator library (@file{libsim.a)} from the linker command line.
13345 @item -mstack-increment=@var{size}
13346 @opindex mstack-increment
13347 Set the maximum amount for a single stack increment operation. Large
13348 values can increase the speed of programs which contain functions
13349 that need a large amount of stack space, but they can also trigger a
13350 segmentation fault if the stack is extended too much. The default
13356 @subsection MeP Options
13357 @cindex MeP options
13363 Enables the @code{abs} instruction, which is the absolute difference
13364 between two registers.
13368 Enables all the optional instructions - average, multiply, divide, bit
13369 operations, leading zero, absolute difference, min/max, clip, and
13375 Enables the @code{ave} instruction, which computes the average of two
13378 @item -mbased=@var{n}
13380 Variables of size @var{n} bytes or smaller will be placed in the
13381 @code{.based} section by default. Based variables use the @code{$tp}
13382 register as a base register, and there is a 128 byte limit to the
13383 @code{.based} section.
13387 Enables the bit operation instructions - bit test (@code{btstm}), set
13388 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
13389 test-and-set (@code{tas}).
13391 @item -mc=@var{name}
13393 Selects which section constant data will be placed in. @var{name} may
13394 be @code{tiny}, @code{near}, or @code{far}.
13398 Enables the @code{clip} instruction. Note that @code{-mclip} is not
13399 useful unless you also provide @code{-mminmax}.
13401 @item -mconfig=@var{name}
13403 Selects one of the build-in core configurations. Each MeP chip has
13404 one or more modules in it; each module has a core CPU and a variety of
13405 coprocessors, optional instructions, and peripherals. The
13406 @code{MeP-Integrator} tool, not part of GCC, provides these
13407 configurations through this option; using this option is the same as
13408 using all the corresponding command line options. The default
13409 configuration is @code{default}.
13413 Enables the coprocessor instructions. By default, this is a 32-bit
13414 coprocessor. Note that the coprocessor is normally enabled via the
13415 @code{-mconfig=} option.
13419 Enables the 32-bit coprocessor's instructions.
13423 Enables the 64-bit coprocessor's instructions.
13427 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
13431 Causes constant variables to be placed in the @code{.near} section.
13435 Enables the @code{div} and @code{divu} instructions.
13439 Generate big-endian code.
13443 Generate little-endian code.
13445 @item -mio-volatile
13446 @opindex mio-volatile
13447 Tells the compiler that any variable marked with the @code{io}
13448 attribute is to be considered volatile.
13452 Causes variables to be assigned to the @code{.far} section by default.
13456 Enables the @code{leadz} (leading zero) instruction.
13460 Causes variables to be assigned to the @code{.near} section by default.
13464 Enables the @code{min} and @code{max} instructions.
13468 Enables the multiplication and multiply-accumulate instructions.
13472 Disables all the optional instructions enabled by @code{-mall-opts}.
13476 Enables the @code{repeat} and @code{erepeat} instructions, used for
13477 low-overhead looping.
13481 Causes all variables to default to the @code{.tiny} section. Note
13482 that there is a 65536 byte limit to this section. Accesses to these
13483 variables use the @code{%gp} base register.
13487 Enables the saturation instructions. Note that the compiler does not
13488 currently generate these itself, but this option is included for
13489 compatibility with other tools, like @code{as}.
13493 Link the SDRAM-based runtime instead of the default ROM-based runtime.
13497 Link the simulator runtime libraries.
13501 Link the simulator runtime libraries, excluding built-in support
13502 for reset and exception vectors and tables.
13506 Causes all functions to default to the @code{.far} section. Without
13507 this option, functions default to the @code{.near} section.
13509 @item -mtiny=@var{n}
13511 Variables that are @var{n} bytes or smaller will be allocated to the
13512 @code{.tiny} section. These variables use the @code{$gp} base
13513 register. The default for this option is 4, but note that there's a
13514 65536 byte limit to the @code{.tiny} section.
13519 @subsection MIPS Options
13520 @cindex MIPS options
13526 Generate big-endian code.
13530 Generate little-endian code. This is the default for @samp{mips*el-*-*}
13533 @item -march=@var{arch}
13535 Generate code that will run on @var{arch}, which can be the name of a
13536 generic MIPS ISA, or the name of a particular processor.
13538 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
13539 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
13540 The processor names are:
13541 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
13542 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
13543 @samp{5kc}, @samp{5kf},
13545 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
13546 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
13547 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
13548 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
13549 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
13550 @samp{loongson2e}, @samp{loongson2f},
13554 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
13555 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
13556 @samp{rm7000}, @samp{rm9000},
13557 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
13560 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
13561 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
13563 The special value @samp{from-abi} selects the
13564 most compatible architecture for the selected ABI (that is,
13565 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
13567 Native Linux/GNU toolchains also support the value @samp{native},
13568 which selects the best architecture option for the host processor.
13569 @option{-march=native} has no effect if GCC does not recognize
13572 In processor names, a final @samp{000} can be abbreviated as @samp{k}
13573 (for example, @samp{-march=r2k}). Prefixes are optional, and
13574 @samp{vr} may be written @samp{r}.
13576 Names of the form @samp{@var{n}f2_1} refer to processors with
13577 FPUs clocked at half the rate of the core, names of the form
13578 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
13579 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
13580 processors with FPUs clocked a ratio of 3:2 with respect to the core.
13581 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
13582 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
13583 accepted as synonyms for @samp{@var{n}f1_1}.
13585 GCC defines two macros based on the value of this option. The first
13586 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
13587 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
13588 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
13589 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
13590 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
13592 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
13593 above. In other words, it will have the full prefix and will not
13594 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
13595 the macro names the resolved architecture (either @samp{"mips1"} or
13596 @samp{"mips3"}). It names the default architecture when no
13597 @option{-march} option is given.
13599 @item -mtune=@var{arch}
13601 Optimize for @var{arch}. Among other things, this option controls
13602 the way instructions are scheduled, and the perceived cost of arithmetic
13603 operations. The list of @var{arch} values is the same as for
13606 When this option is not used, GCC will optimize for the processor
13607 specified by @option{-march}. By using @option{-march} and
13608 @option{-mtune} together, it is possible to generate code that will
13609 run on a family of processors, but optimize the code for one
13610 particular member of that family.
13612 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
13613 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
13614 @samp{-march} ones described above.
13618 Equivalent to @samp{-march=mips1}.
13622 Equivalent to @samp{-march=mips2}.
13626 Equivalent to @samp{-march=mips3}.
13630 Equivalent to @samp{-march=mips4}.
13634 Equivalent to @samp{-march=mips32}.
13638 Equivalent to @samp{-march=mips32r2}.
13642 Equivalent to @samp{-march=mips64}.
13646 Equivalent to @samp{-march=mips64r2}.
13651 @opindex mno-mips16
13652 Generate (do not generate) MIPS16 code. If GCC is targetting a
13653 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
13655 MIPS16 code generation can also be controlled on a per-function basis
13656 by means of @code{mips16} and @code{nomips16} attributes.
13657 @xref{Function Attributes}, for more information.
13659 @item -mflip-mips16
13660 @opindex mflip-mips16
13661 Generate MIPS16 code on alternating functions. This option is provided
13662 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
13663 not intended for ordinary use in compiling user code.
13665 @item -minterlink-mips16
13666 @itemx -mno-interlink-mips16
13667 @opindex minterlink-mips16
13668 @opindex mno-interlink-mips16
13669 Require (do not require) that non-MIPS16 code be link-compatible with
13672 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
13673 it must either use a call or an indirect jump. @option{-minterlink-mips16}
13674 therefore disables direct jumps unless GCC knows that the target of the
13675 jump is not MIPS16.
13687 Generate code for the given ABI@.
13689 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
13690 generates 64-bit code when you select a 64-bit architecture, but you
13691 can use @option{-mgp32} to get 32-bit code instead.
13693 For information about the O64 ABI, see
13694 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
13696 GCC supports a variant of the o32 ABI in which floating-point registers
13697 are 64 rather than 32 bits wide. You can select this combination with
13698 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
13699 and @samp{mfhc1} instructions and is therefore only supported for
13700 MIPS32R2 processors.
13702 The register assignments for arguments and return values remain the
13703 same, but each scalar value is passed in a single 64-bit register
13704 rather than a pair of 32-bit registers. For example, scalar
13705 floating-point values are returned in @samp{$f0} only, not a
13706 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
13707 remains the same, but all 64 bits are saved.
13710 @itemx -mno-abicalls
13712 @opindex mno-abicalls
13713 Generate (do not generate) code that is suitable for SVR4-style
13714 dynamic objects. @option{-mabicalls} is the default for SVR4-based
13719 Generate (do not generate) code that is fully position-independent,
13720 and that can therefore be linked into shared libraries. This option
13721 only affects @option{-mabicalls}.
13723 All @option{-mabicalls} code has traditionally been position-independent,
13724 regardless of options like @option{-fPIC} and @option{-fpic}. However,
13725 as an extension, the GNU toolchain allows executables to use absolute
13726 accesses for locally-binding symbols. It can also use shorter GP
13727 initialization sequences and generate direct calls to locally-defined
13728 functions. This mode is selected by @option{-mno-shared}.
13730 @option{-mno-shared} depends on binutils 2.16 or higher and generates
13731 objects that can only be linked by the GNU linker. However, the option
13732 does not affect the ABI of the final executable; it only affects the ABI
13733 of relocatable objects. Using @option{-mno-shared} will generally make
13734 executables both smaller and quicker.
13736 @option{-mshared} is the default.
13742 Assume (do not assume) that the static and dynamic linkers
13743 support PLTs and copy relocations. This option only affects
13744 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
13745 has no effect without @samp{-msym32}.
13747 You can make @option{-mplt} the default by configuring
13748 GCC with @option{--with-mips-plt}. The default is
13749 @option{-mno-plt} otherwise.
13755 Lift (do not lift) the usual restrictions on the size of the global
13758 GCC normally uses a single instruction to load values from the GOT@.
13759 While this is relatively efficient, it will only work if the GOT
13760 is smaller than about 64k. Anything larger will cause the linker
13761 to report an error such as:
13763 @cindex relocation truncated to fit (MIPS)
13765 relocation truncated to fit: R_MIPS_GOT16 foobar
13768 If this happens, you should recompile your code with @option{-mxgot}.
13769 It should then work with very large GOTs, although it will also be
13770 less efficient, since it will take three instructions to fetch the
13771 value of a global symbol.
13773 Note that some linkers can create multiple GOTs. If you have such a
13774 linker, you should only need to use @option{-mxgot} when a single object
13775 file accesses more than 64k's worth of GOT entries. Very few do.
13777 These options have no effect unless GCC is generating position
13782 Assume that general-purpose registers are 32 bits wide.
13786 Assume that general-purpose registers are 64 bits wide.
13790 Assume that floating-point registers are 32 bits wide.
13794 Assume that floating-point registers are 64 bits wide.
13797 @opindex mhard-float
13798 Use floating-point coprocessor instructions.
13801 @opindex msoft-float
13802 Do not use floating-point coprocessor instructions. Implement
13803 floating-point calculations using library calls instead.
13805 @item -msingle-float
13806 @opindex msingle-float
13807 Assume that the floating-point coprocessor only supports single-precision
13810 @item -mdouble-float
13811 @opindex mdouble-float
13812 Assume that the floating-point coprocessor supports double-precision
13813 operations. This is the default.
13819 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
13820 implement atomic memory built-in functions. When neither option is
13821 specified, GCC will use the instructions if the target architecture
13824 @option{-mllsc} is useful if the runtime environment can emulate the
13825 instructions and @option{-mno-llsc} can be useful when compiling for
13826 nonstandard ISAs. You can make either option the default by
13827 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
13828 respectively. @option{--with-llsc} is the default for some
13829 configurations; see the installation documentation for details.
13835 Use (do not use) revision 1 of the MIPS DSP ASE@.
13836 @xref{MIPS DSP Built-in Functions}. This option defines the
13837 preprocessor macro @samp{__mips_dsp}. It also defines
13838 @samp{__mips_dsp_rev} to 1.
13844 Use (do not use) revision 2 of the MIPS DSP ASE@.
13845 @xref{MIPS DSP Built-in Functions}. This option defines the
13846 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
13847 It also defines @samp{__mips_dsp_rev} to 2.
13850 @itemx -mno-smartmips
13851 @opindex msmartmips
13852 @opindex mno-smartmips
13853 Use (do not use) the MIPS SmartMIPS ASE.
13855 @item -mpaired-single
13856 @itemx -mno-paired-single
13857 @opindex mpaired-single
13858 @opindex mno-paired-single
13859 Use (do not use) paired-single floating-point instructions.
13860 @xref{MIPS Paired-Single Support}. This option requires
13861 hardware floating-point support to be enabled.
13867 Use (do not use) MIPS Digital Media Extension instructions.
13868 This option can only be used when generating 64-bit code and requires
13869 hardware floating-point support to be enabled.
13874 @opindex mno-mips3d
13875 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
13876 The option @option{-mips3d} implies @option{-mpaired-single}.
13882 Use (do not use) MT Multithreading instructions.
13886 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
13887 an explanation of the default and the way that the pointer size is
13892 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
13894 The default size of @code{int}s, @code{long}s and pointers depends on
13895 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
13896 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
13897 32-bit @code{long}s. Pointers are the same size as @code{long}s,
13898 or the same size as integer registers, whichever is smaller.
13904 Assume (do not assume) that all symbols have 32-bit values, regardless
13905 of the selected ABI@. This option is useful in combination with
13906 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
13907 to generate shorter and faster references to symbolic addresses.
13911 Put definitions of externally-visible data in a small data section
13912 if that data is no bigger than @var{num} bytes. GCC can then access
13913 the data more efficiently; see @option{-mgpopt} for details.
13915 The default @option{-G} option depends on the configuration.
13917 @item -mlocal-sdata
13918 @itemx -mno-local-sdata
13919 @opindex mlocal-sdata
13920 @opindex mno-local-sdata
13921 Extend (do not extend) the @option{-G} behavior to local data too,
13922 such as to static variables in C@. @option{-mlocal-sdata} is the
13923 default for all configurations.
13925 If the linker complains that an application is using too much small data,
13926 you might want to try rebuilding the less performance-critical parts with
13927 @option{-mno-local-sdata}. You might also want to build large
13928 libraries with @option{-mno-local-sdata}, so that the libraries leave
13929 more room for the main program.
13931 @item -mextern-sdata
13932 @itemx -mno-extern-sdata
13933 @opindex mextern-sdata
13934 @opindex mno-extern-sdata
13935 Assume (do not assume) that externally-defined data will be in
13936 a small data section if that data is within the @option{-G} limit.
13937 @option{-mextern-sdata} is the default for all configurations.
13939 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
13940 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
13941 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
13942 is placed in a small data section. If @var{Var} is defined by another
13943 module, you must either compile that module with a high-enough
13944 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
13945 definition. If @var{Var} is common, you must link the application
13946 with a high-enough @option{-G} setting.
13948 The easiest way of satisfying these restrictions is to compile
13949 and link every module with the same @option{-G} option. However,
13950 you may wish to build a library that supports several different
13951 small data limits. You can do this by compiling the library with
13952 the highest supported @option{-G} setting and additionally using
13953 @option{-mno-extern-sdata} to stop the library from making assumptions
13954 about externally-defined data.
13960 Use (do not use) GP-relative accesses for symbols that are known to be
13961 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
13962 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
13965 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
13966 might not hold the value of @code{_gp}. For example, if the code is
13967 part of a library that might be used in a boot monitor, programs that
13968 call boot monitor routines will pass an unknown value in @code{$gp}.
13969 (In such situations, the boot monitor itself would usually be compiled
13970 with @option{-G0}.)
13972 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
13973 @option{-mno-extern-sdata}.
13975 @item -membedded-data
13976 @itemx -mno-embedded-data
13977 @opindex membedded-data
13978 @opindex mno-embedded-data
13979 Allocate variables to the read-only data section first if possible, then
13980 next in the small data section if possible, otherwise in data. This gives
13981 slightly slower code than the default, but reduces the amount of RAM required
13982 when executing, and thus may be preferred for some embedded systems.
13984 @item -muninit-const-in-rodata
13985 @itemx -mno-uninit-const-in-rodata
13986 @opindex muninit-const-in-rodata
13987 @opindex mno-uninit-const-in-rodata
13988 Put uninitialized @code{const} variables in the read-only data section.
13989 This option is only meaningful in conjunction with @option{-membedded-data}.
13991 @item -mcode-readable=@var{setting}
13992 @opindex mcode-readable
13993 Specify whether GCC may generate code that reads from executable sections.
13994 There are three possible settings:
13997 @item -mcode-readable=yes
13998 Instructions may freely access executable sections. This is the
14001 @item -mcode-readable=pcrel
14002 MIPS16 PC-relative load instructions can access executable sections,
14003 but other instructions must not do so. This option is useful on 4KSc
14004 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
14005 It is also useful on processors that can be configured to have a dual
14006 instruction/data SRAM interface and that, like the M4K, automatically
14007 redirect PC-relative loads to the instruction RAM.
14009 @item -mcode-readable=no
14010 Instructions must not access executable sections. This option can be
14011 useful on targets that are configured to have a dual instruction/data
14012 SRAM interface but that (unlike the M4K) do not automatically redirect
14013 PC-relative loads to the instruction RAM.
14016 @item -msplit-addresses
14017 @itemx -mno-split-addresses
14018 @opindex msplit-addresses
14019 @opindex mno-split-addresses
14020 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
14021 relocation operators. This option has been superseded by
14022 @option{-mexplicit-relocs} but is retained for backwards compatibility.
14024 @item -mexplicit-relocs
14025 @itemx -mno-explicit-relocs
14026 @opindex mexplicit-relocs
14027 @opindex mno-explicit-relocs
14028 Use (do not use) assembler relocation operators when dealing with symbolic
14029 addresses. The alternative, selected by @option{-mno-explicit-relocs},
14030 is to use assembler macros instead.
14032 @option{-mexplicit-relocs} is the default if GCC was configured
14033 to use an assembler that supports relocation operators.
14035 @item -mcheck-zero-division
14036 @itemx -mno-check-zero-division
14037 @opindex mcheck-zero-division
14038 @opindex mno-check-zero-division
14039 Trap (do not trap) on integer division by zero.
14041 The default is @option{-mcheck-zero-division}.
14043 @item -mdivide-traps
14044 @itemx -mdivide-breaks
14045 @opindex mdivide-traps
14046 @opindex mdivide-breaks
14047 MIPS systems check for division by zero by generating either a
14048 conditional trap or a break instruction. Using traps results in
14049 smaller code, but is only supported on MIPS II and later. Also, some
14050 versions of the Linux kernel have a bug that prevents trap from
14051 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
14052 allow conditional traps on architectures that support them and
14053 @option{-mdivide-breaks} to force the use of breaks.
14055 The default is usually @option{-mdivide-traps}, but this can be
14056 overridden at configure time using @option{--with-divide=breaks}.
14057 Divide-by-zero checks can be completely disabled using
14058 @option{-mno-check-zero-division}.
14063 @opindex mno-memcpy
14064 Force (do not force) the use of @code{memcpy()} for non-trivial block
14065 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
14066 most constant-sized copies.
14069 @itemx -mno-long-calls
14070 @opindex mlong-calls
14071 @opindex mno-long-calls
14072 Disable (do not disable) use of the @code{jal} instruction. Calling
14073 functions using @code{jal} is more efficient but requires the caller
14074 and callee to be in the same 256 megabyte segment.
14076 This option has no effect on abicalls code. The default is
14077 @option{-mno-long-calls}.
14083 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
14084 instructions, as provided by the R4650 ISA@.
14087 @itemx -mno-fused-madd
14088 @opindex mfused-madd
14089 @opindex mno-fused-madd
14090 Enable (disable) use of the floating point multiply-accumulate
14091 instructions, when they are available. The default is
14092 @option{-mfused-madd}.
14094 When multiply-accumulate instructions are used, the intermediate
14095 product is calculated to infinite precision and is not subject to
14096 the FCSR Flush to Zero bit. This may be undesirable in some
14101 Tell the MIPS assembler to not run its preprocessor over user
14102 assembler files (with a @samp{.s} suffix) when assembling them.
14105 @itemx -mno-fix-r4000
14106 @opindex mfix-r4000
14107 @opindex mno-fix-r4000
14108 Work around certain R4000 CPU errata:
14111 A double-word or a variable shift may give an incorrect result if executed
14112 immediately after starting an integer division.
14114 A double-word or a variable shift may give an incorrect result if executed
14115 while an integer multiplication is in progress.
14117 An integer division may give an incorrect result if started in a delay slot
14118 of a taken branch or a jump.
14122 @itemx -mno-fix-r4400
14123 @opindex mfix-r4400
14124 @opindex mno-fix-r4400
14125 Work around certain R4400 CPU errata:
14128 A double-word or a variable shift may give an incorrect result if executed
14129 immediately after starting an integer division.
14133 @itemx -mno-fix-r10000
14134 @opindex mfix-r10000
14135 @opindex mno-fix-r10000
14136 Work around certain R10000 errata:
14139 @code{ll}/@code{sc} sequences may not behave atomically on revisions
14140 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
14143 This option can only be used if the target architecture supports
14144 branch-likely instructions. @option{-mfix-r10000} is the default when
14145 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
14149 @itemx -mno-fix-vr4120
14150 @opindex mfix-vr4120
14151 Work around certain VR4120 errata:
14154 @code{dmultu} does not always produce the correct result.
14156 @code{div} and @code{ddiv} do not always produce the correct result if one
14157 of the operands is negative.
14159 The workarounds for the division errata rely on special functions in
14160 @file{libgcc.a}. At present, these functions are only provided by
14161 the @code{mips64vr*-elf} configurations.
14163 Other VR4120 errata require a nop to be inserted between certain pairs of
14164 instructions. These errata are handled by the assembler, not by GCC itself.
14167 @opindex mfix-vr4130
14168 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
14169 workarounds are implemented by the assembler rather than by GCC,
14170 although GCC will avoid using @code{mflo} and @code{mfhi} if the
14171 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
14172 instructions are available instead.
14175 @itemx -mno-fix-sb1
14177 Work around certain SB-1 CPU core errata.
14178 (This flag currently works around the SB-1 revision 2
14179 ``F1'' and ``F2'' floating point errata.)
14181 @item -mr10k-cache-barrier=@var{setting}
14182 @opindex mr10k-cache-barrier
14183 Specify whether GCC should insert cache barriers to avoid the
14184 side-effects of speculation on R10K processors.
14186 In common with many processors, the R10K tries to predict the outcome
14187 of a conditional branch and speculatively executes instructions from
14188 the ``taken'' branch. It later aborts these instructions if the
14189 predicted outcome was wrong. However, on the R10K, even aborted
14190 instructions can have side effects.
14192 This problem only affects kernel stores and, depending on the system,
14193 kernel loads. As an example, a speculatively-executed store may load
14194 the target memory into cache and mark the cache line as dirty, even if
14195 the store itself is later aborted. If a DMA operation writes to the
14196 same area of memory before the ``dirty'' line is flushed, the cached
14197 data will overwrite the DMA-ed data. See the R10K processor manual
14198 for a full description, including other potential problems.
14200 One workaround is to insert cache barrier instructions before every memory
14201 access that might be speculatively executed and that might have side
14202 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
14203 controls GCC's implementation of this workaround. It assumes that
14204 aborted accesses to any byte in the following regions will not have
14209 the memory occupied by the current function's stack frame;
14212 the memory occupied by an incoming stack argument;
14215 the memory occupied by an object with a link-time-constant address.
14218 It is the kernel's responsibility to ensure that speculative
14219 accesses to these regions are indeed safe.
14221 If the input program contains a function declaration such as:
14227 then the implementation of @code{foo} must allow @code{j foo} and
14228 @code{jal foo} to be executed speculatively. GCC honors this
14229 restriction for functions it compiles itself. It expects non-GCC
14230 functions (such as hand-written assembly code) to do the same.
14232 The option has three forms:
14235 @item -mr10k-cache-barrier=load-store
14236 Insert a cache barrier before a load or store that might be
14237 speculatively executed and that might have side effects even
14240 @item -mr10k-cache-barrier=store
14241 Insert a cache barrier before a store that might be speculatively
14242 executed and that might have side effects even if aborted.
14244 @item -mr10k-cache-barrier=none
14245 Disable the insertion of cache barriers. This is the default setting.
14248 @item -mflush-func=@var{func}
14249 @itemx -mno-flush-func
14250 @opindex mflush-func
14251 Specifies the function to call to flush the I and D caches, or to not
14252 call any such function. If called, the function must take the same
14253 arguments as the common @code{_flush_func()}, that is, the address of the
14254 memory range for which the cache is being flushed, the size of the
14255 memory range, and the number 3 (to flush both caches). The default
14256 depends on the target GCC was configured for, but commonly is either
14257 @samp{_flush_func} or @samp{__cpu_flush}.
14259 @item mbranch-cost=@var{num}
14260 @opindex mbranch-cost
14261 Set the cost of branches to roughly @var{num} ``simple'' instructions.
14262 This cost is only a heuristic and is not guaranteed to produce
14263 consistent results across releases. A zero cost redundantly selects
14264 the default, which is based on the @option{-mtune} setting.
14266 @item -mbranch-likely
14267 @itemx -mno-branch-likely
14268 @opindex mbranch-likely
14269 @opindex mno-branch-likely
14270 Enable or disable use of Branch Likely instructions, regardless of the
14271 default for the selected architecture. By default, Branch Likely
14272 instructions may be generated if they are supported by the selected
14273 architecture. An exception is for the MIPS32 and MIPS64 architectures
14274 and processors which implement those architectures; for those, Branch
14275 Likely instructions will not be generated by default because the MIPS32
14276 and MIPS64 architectures specifically deprecate their use.
14278 @item -mfp-exceptions
14279 @itemx -mno-fp-exceptions
14280 @opindex mfp-exceptions
14281 Specifies whether FP exceptions are enabled. This affects how we schedule
14282 FP instructions for some processors. The default is that FP exceptions are
14285 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
14286 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
14289 @item -mvr4130-align
14290 @itemx -mno-vr4130-align
14291 @opindex mvr4130-align
14292 The VR4130 pipeline is two-way superscalar, but can only issue two
14293 instructions together if the first one is 8-byte aligned. When this
14294 option is enabled, GCC will align pairs of instructions that it
14295 thinks should execute in parallel.
14297 This option only has an effect when optimizing for the VR4130.
14298 It normally makes code faster, but at the expense of making it bigger.
14299 It is enabled by default at optimization level @option{-O3}.
14304 Enable (disable) generation of @code{synci} instructions on
14305 architectures that support it. The @code{synci} instructions (if
14306 enabled) will be generated when @code{__builtin___clear_cache()} is
14309 This option defaults to @code{-mno-synci}, but the default can be
14310 overridden by configuring with @code{--with-synci}.
14312 When compiling code for single processor systems, it is generally safe
14313 to use @code{synci}. However, on many multi-core (SMP) systems, it
14314 will not invalidate the instruction caches on all cores and may lead
14315 to undefined behavior.
14317 @item -mrelax-pic-calls
14318 @itemx -mno-relax-pic-calls
14319 @opindex mrelax-pic-calls
14320 Try to turn PIC calls that are normally dispatched via register
14321 @code{$25} into direct calls. This is only possible if the linker can
14322 resolve the destination at link-time and if the destination is within
14323 range for a direct call.
14325 @option{-mrelax-pic-calls} is the default if GCC was configured to use
14326 an assembler and a linker that supports the @code{.reloc} assembly
14327 directive and @code{-mexplicit-relocs} is in effect. With
14328 @code{-mno-explicit-relocs}, this optimization can be performed by the
14329 assembler and the linker alone without help from the compiler.
14331 @item -mmcount-ra-address
14332 @itemx -mno-mcount-ra-address
14333 @opindex mmcount-ra-address
14334 @opindex mno-mcount-ra-address
14335 Emit (do not emit) code that allows @code{_mcount} to modify the
14336 calling function's return address. When enabled, this option extends
14337 the usual @code{_mcount} interface with a new @var{ra-address}
14338 parameter, which has type @code{intptr_t *} and is passed in register
14339 @code{$12}. @code{_mcount} can then modify the return address by
14340 doing both of the following:
14343 Returning the new address in register @code{$31}.
14345 Storing the new address in @code{*@var{ra-address}},
14346 if @var{ra-address} is nonnull.
14349 The default is @option{-mno-mcount-ra-address}.
14354 @subsection MMIX Options
14355 @cindex MMIX Options
14357 These options are defined for the MMIX:
14361 @itemx -mno-libfuncs
14363 @opindex mno-libfuncs
14364 Specify that intrinsic library functions are being compiled, passing all
14365 values in registers, no matter the size.
14368 @itemx -mno-epsilon
14370 @opindex mno-epsilon
14371 Generate floating-point comparison instructions that compare with respect
14372 to the @code{rE} epsilon register.
14374 @item -mabi=mmixware
14376 @opindex mabi=mmixware
14378 Generate code that passes function parameters and return values that (in
14379 the called function) are seen as registers @code{$0} and up, as opposed to
14380 the GNU ABI which uses global registers @code{$231} and up.
14382 @item -mzero-extend
14383 @itemx -mno-zero-extend
14384 @opindex mzero-extend
14385 @opindex mno-zero-extend
14386 When reading data from memory in sizes shorter than 64 bits, use (do not
14387 use) zero-extending load instructions by default, rather than
14388 sign-extending ones.
14391 @itemx -mno-knuthdiv
14393 @opindex mno-knuthdiv
14394 Make the result of a division yielding a remainder have the same sign as
14395 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
14396 remainder follows the sign of the dividend. Both methods are
14397 arithmetically valid, the latter being almost exclusively used.
14399 @item -mtoplevel-symbols
14400 @itemx -mno-toplevel-symbols
14401 @opindex mtoplevel-symbols
14402 @opindex mno-toplevel-symbols
14403 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
14404 code can be used with the @code{PREFIX} assembly directive.
14408 Generate an executable in the ELF format, rather than the default
14409 @samp{mmo} format used by the @command{mmix} simulator.
14411 @item -mbranch-predict
14412 @itemx -mno-branch-predict
14413 @opindex mbranch-predict
14414 @opindex mno-branch-predict
14415 Use (do not use) the probable-branch instructions, when static branch
14416 prediction indicates a probable branch.
14418 @item -mbase-addresses
14419 @itemx -mno-base-addresses
14420 @opindex mbase-addresses
14421 @opindex mno-base-addresses
14422 Generate (do not generate) code that uses @emph{base addresses}. Using a
14423 base address automatically generates a request (handled by the assembler
14424 and the linker) for a constant to be set up in a global register. The
14425 register is used for one or more base address requests within the range 0
14426 to 255 from the value held in the register. The generally leads to short
14427 and fast code, but the number of different data items that can be
14428 addressed is limited. This means that a program that uses lots of static
14429 data may require @option{-mno-base-addresses}.
14431 @item -msingle-exit
14432 @itemx -mno-single-exit
14433 @opindex msingle-exit
14434 @opindex mno-single-exit
14435 Force (do not force) generated code to have a single exit point in each
14439 @node MN10300 Options
14440 @subsection MN10300 Options
14441 @cindex MN10300 options
14443 These @option{-m} options are defined for Matsushita MN10300 architectures:
14448 Generate code to avoid bugs in the multiply instructions for the MN10300
14449 processors. This is the default.
14451 @item -mno-mult-bug
14452 @opindex mno-mult-bug
14453 Do not generate code to avoid bugs in the multiply instructions for the
14454 MN10300 processors.
14458 Generate code which uses features specific to the AM33 processor.
14462 Do not generate code which uses features specific to the AM33 processor. This
14465 @item -mreturn-pointer-on-d0
14466 @opindex mreturn-pointer-on-d0
14467 When generating a function which returns a pointer, return the pointer
14468 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
14469 only in a0, and attempts to call such functions without a prototype
14470 would result in errors. Note that this option is on by default; use
14471 @option{-mno-return-pointer-on-d0} to disable it.
14475 Do not link in the C run-time initialization object file.
14479 Indicate to the linker that it should perform a relaxation optimization pass
14480 to shorten branches, calls and absolute memory addresses. This option only
14481 has an effect when used on the command line for the final link step.
14483 This option makes symbolic debugging impossible.
14486 @node PDP-11 Options
14487 @subsection PDP-11 Options
14488 @cindex PDP-11 Options
14490 These options are defined for the PDP-11:
14495 Use hardware FPP floating point. This is the default. (FIS floating
14496 point on the PDP-11/40 is not supported.)
14499 @opindex msoft-float
14500 Do not use hardware floating point.
14504 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
14508 Return floating-point results in memory. This is the default.
14512 Generate code for a PDP-11/40.
14516 Generate code for a PDP-11/45. This is the default.
14520 Generate code for a PDP-11/10.
14522 @item -mbcopy-builtin
14523 @opindex mbcopy-builtin
14524 Use inline @code{movmemhi} patterns for copying memory. This is the
14529 Do not use inline @code{movmemhi} patterns for copying memory.
14535 Use 16-bit @code{int}. This is the default.
14541 Use 32-bit @code{int}.
14544 @itemx -mno-float32
14546 @opindex mno-float32
14547 Use 64-bit @code{float}. This is the default.
14550 @itemx -mno-float64
14552 @opindex mno-float64
14553 Use 32-bit @code{float}.
14557 Use @code{abshi2} pattern. This is the default.
14561 Do not use @code{abshi2} pattern.
14563 @item -mbranch-expensive
14564 @opindex mbranch-expensive
14565 Pretend that branches are expensive. This is for experimenting with
14566 code generation only.
14568 @item -mbranch-cheap
14569 @opindex mbranch-cheap
14570 Do not pretend that branches are expensive. This is the default.
14574 Generate code for a system with split I&D@.
14578 Generate code for a system without split I&D@. This is the default.
14582 Use Unix assembler syntax. This is the default when configured for
14583 @samp{pdp11-*-bsd}.
14587 Use DEC assembler syntax. This is the default when configured for any
14588 PDP-11 target other than @samp{pdp11-*-bsd}.
14591 @node picoChip Options
14592 @subsection picoChip Options
14593 @cindex picoChip options
14595 These @samp{-m} options are defined for picoChip implementations:
14599 @item -mae=@var{ae_type}
14601 Set the instruction set, register set, and instruction scheduling
14602 parameters for array element type @var{ae_type}. Supported values
14603 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
14605 @option{-mae=ANY} selects a completely generic AE type. Code
14606 generated with this option will run on any of the other AE types. The
14607 code will not be as efficient as it would be if compiled for a specific
14608 AE type, and some types of operation (e.g., multiplication) will not
14609 work properly on all types of AE.
14611 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
14612 for compiled code, and is the default.
14614 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
14615 option may suffer from poor performance of byte (char) manipulation,
14616 since the DSP AE does not provide hardware support for byte load/stores.
14618 @item -msymbol-as-address
14619 Enable the compiler to directly use a symbol name as an address in a
14620 load/store instruction, without first loading it into a
14621 register. Typically, the use of this option will generate larger
14622 programs, which run faster than when the option isn't used. However, the
14623 results vary from program to program, so it is left as a user option,
14624 rather than being permanently enabled.
14626 @item -mno-inefficient-warnings
14627 Disables warnings about the generation of inefficient code. These
14628 warnings can be generated, for example, when compiling code which
14629 performs byte-level memory operations on the MAC AE type. The MAC AE has
14630 no hardware support for byte-level memory operations, so all byte
14631 load/stores must be synthesized from word load/store operations. This is
14632 inefficient and a warning will be generated indicating to the programmer
14633 that they should rewrite the code to avoid byte operations, or to target
14634 an AE type which has the necessary hardware support. This option enables
14635 the warning to be turned off.
14639 @node PowerPC Options
14640 @subsection PowerPC Options
14641 @cindex PowerPC options
14643 These are listed under @xref{RS/6000 and PowerPC Options}.
14645 @node RS/6000 and PowerPC Options
14646 @subsection IBM RS/6000 and PowerPC Options
14647 @cindex RS/6000 and PowerPC Options
14648 @cindex IBM RS/6000 and PowerPC Options
14650 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
14657 @itemx -mno-powerpc
14658 @itemx -mpowerpc-gpopt
14659 @itemx -mno-powerpc-gpopt
14660 @itemx -mpowerpc-gfxopt
14661 @itemx -mno-powerpc-gfxopt
14663 @itemx -mno-powerpc64
14667 @itemx -mno-popcntb
14669 @itemx -mno-popcntd
14677 @itemx -mno-hard-dfp
14681 @opindex mno-power2
14683 @opindex mno-powerpc
14684 @opindex mpowerpc-gpopt
14685 @opindex mno-powerpc-gpopt
14686 @opindex mpowerpc-gfxopt
14687 @opindex mno-powerpc-gfxopt
14688 @opindex mpowerpc64
14689 @opindex mno-powerpc64
14693 @opindex mno-popcntb
14695 @opindex mno-popcntd
14701 @opindex mno-mfpgpr
14703 @opindex mno-hard-dfp
14704 GCC supports two related instruction set architectures for the
14705 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
14706 instructions supported by the @samp{rios} chip set used in the original
14707 RS/6000 systems and the @dfn{PowerPC} instruction set is the
14708 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
14709 the IBM 4xx, 6xx, and follow-on microprocessors.
14711 Neither architecture is a subset of the other. However there is a
14712 large common subset of instructions supported by both. An MQ
14713 register is included in processors supporting the POWER architecture.
14715 You use these options to specify which instructions are available on the
14716 processor you are using. The default value of these options is
14717 determined when configuring GCC@. Specifying the
14718 @option{-mcpu=@var{cpu_type}} overrides the specification of these
14719 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
14720 rather than the options listed above.
14722 The @option{-mpower} option allows GCC to generate instructions that
14723 are found only in the POWER architecture and to use the MQ register.
14724 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
14725 to generate instructions that are present in the POWER2 architecture but
14726 not the original POWER architecture.
14728 The @option{-mpowerpc} option allows GCC to generate instructions that
14729 are found only in the 32-bit subset of the PowerPC architecture.
14730 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
14731 GCC to use the optional PowerPC architecture instructions in the
14732 General Purpose group, including floating-point square root. Specifying
14733 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
14734 use the optional PowerPC architecture instructions in the Graphics
14735 group, including floating-point select.
14737 The @option{-mmfcrf} option allows GCC to generate the move from
14738 condition register field instruction implemented on the POWER4
14739 processor and other processors that support the PowerPC V2.01
14741 The @option{-mpopcntb} option allows GCC to generate the popcount and
14742 double precision FP reciprocal estimate instruction implemented on the
14743 POWER5 processor and other processors that support the PowerPC V2.02
14745 The @option{-mpopcntd} option allows GCC to generate the popcount
14746 instruction implemented on the POWER7 processor and other processors
14747 that support the PowerPC V2.06 architecture.
14748 The @option{-mfprnd} option allows GCC to generate the FP round to
14749 integer instructions implemented on the POWER5+ processor and other
14750 processors that support the PowerPC V2.03 architecture.
14751 The @option{-mcmpb} option allows GCC to generate the compare bytes
14752 instruction implemented on the POWER6 processor and other processors
14753 that support the PowerPC V2.05 architecture.
14754 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
14755 general purpose register instructions implemented on the POWER6X
14756 processor and other processors that support the extended PowerPC V2.05
14758 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
14759 point instructions implemented on some POWER processors.
14761 The @option{-mpowerpc64} option allows GCC to generate the additional
14762 64-bit instructions that are found in the full PowerPC64 architecture
14763 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
14764 @option{-mno-powerpc64}.
14766 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
14767 will use only the instructions in the common subset of both
14768 architectures plus some special AIX common-mode calls, and will not use
14769 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
14770 permits GCC to use any instruction from either architecture and to
14771 allow use of the MQ register; specify this for the Motorola MPC601.
14773 @item -mnew-mnemonics
14774 @itemx -mold-mnemonics
14775 @opindex mnew-mnemonics
14776 @opindex mold-mnemonics
14777 Select which mnemonics to use in the generated assembler code. With
14778 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
14779 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
14780 assembler mnemonics defined for the POWER architecture. Instructions
14781 defined in only one architecture have only one mnemonic; GCC uses that
14782 mnemonic irrespective of which of these options is specified.
14784 GCC defaults to the mnemonics appropriate for the architecture in
14785 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
14786 value of these option. Unless you are building a cross-compiler, you
14787 should normally not specify either @option{-mnew-mnemonics} or
14788 @option{-mold-mnemonics}, but should instead accept the default.
14790 @item -mcpu=@var{cpu_type}
14792 Set architecture type, register usage, choice of mnemonics, and
14793 instruction scheduling parameters for machine type @var{cpu_type}.
14794 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
14795 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
14796 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
14797 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
14798 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
14799 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
14800 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{ec603e}, @samp{G3},
14801 @samp{G4}, @samp{G5}, @samp{power}, @samp{power2}, @samp{power3},
14802 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x},
14803 @samp{power7}, @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
14804 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
14806 @option{-mcpu=common} selects a completely generic processor. Code
14807 generated under this option will run on any POWER or PowerPC processor.
14808 GCC will use only the instructions in the common subset of both
14809 architectures, and will not use the MQ register. GCC assumes a generic
14810 processor model for scheduling purposes.
14812 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
14813 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
14814 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
14815 types, with an appropriate, generic processor model assumed for
14816 scheduling purposes.
14818 The other options specify a specific processor. Code generated under
14819 those options will run best on that processor, and may not run at all on
14822 The @option{-mcpu} options automatically enable or disable the
14825 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
14826 -mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol
14827 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
14828 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx}
14830 The particular options set for any particular CPU will vary between
14831 compiler versions, depending on what setting seems to produce optimal
14832 code for that CPU; it doesn't necessarily reflect the actual hardware's
14833 capabilities. If you wish to set an individual option to a particular
14834 value, you may specify it after the @option{-mcpu} option, like
14835 @samp{-mcpu=970 -mno-altivec}.
14837 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
14838 not enabled or disabled by the @option{-mcpu} option at present because
14839 AIX does not have full support for these options. You may still
14840 enable or disable them individually if you're sure it'll work in your
14843 @item -mtune=@var{cpu_type}
14845 Set the instruction scheduling parameters for machine type
14846 @var{cpu_type}, but do not set the architecture type, register usage, or
14847 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
14848 values for @var{cpu_type} are used for @option{-mtune} as for
14849 @option{-mcpu}. If both are specified, the code generated will use the
14850 architecture, registers, and mnemonics set by @option{-mcpu}, but the
14851 scheduling parameters set by @option{-mtune}.
14857 Generate code to compute division as reciprocal estimate and iterative
14858 refinement, creating opportunities for increased throughput. This
14859 feature requires: optional PowerPC Graphics instruction set for single
14860 precision and FRE instruction for double precision, assuming divides
14861 cannot generate user-visible traps, and the domain values not include
14862 Infinities, denormals or zero denominator.
14865 @itemx -mno-altivec
14867 @opindex mno-altivec
14868 Generate code that uses (does not use) AltiVec instructions, and also
14869 enable the use of built-in functions that allow more direct access to
14870 the AltiVec instruction set. You may also need to set
14871 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
14877 @opindex mno-vrsave
14878 Generate VRSAVE instructions when generating AltiVec code.
14880 @item -mgen-cell-microcode
14881 @opindex mgen-cell-microcode
14882 Generate Cell microcode instructions
14884 @item -mwarn-cell-microcode
14885 @opindex mwarn-cell-microcode
14886 Warning when a Cell microcode instruction is going to emitted. An example
14887 of a Cell microcode instruction is a variable shift.
14890 @opindex msecure-plt
14891 Generate code that allows ld and ld.so to build executables and shared
14892 libraries with non-exec .plt and .got sections. This is a PowerPC
14893 32-bit SYSV ABI option.
14897 Generate code that uses a BSS .plt section that ld.so fills in, and
14898 requires .plt and .got sections that are both writable and executable.
14899 This is a PowerPC 32-bit SYSV ABI option.
14905 This switch enables or disables the generation of ISEL instructions.
14907 @item -misel=@var{yes/no}
14908 This switch has been deprecated. Use @option{-misel} and
14909 @option{-mno-isel} instead.
14915 This switch enables or disables the generation of SPE simd
14921 @opindex mno-paired
14922 This switch enables or disables the generation of PAIRED simd
14925 @item -mspe=@var{yes/no}
14926 This option has been deprecated. Use @option{-mspe} and
14927 @option{-mno-spe} instead.
14933 Generate code that uses (does not use) vector/scalar (VSX)
14934 instructions, and also enable the use of built-in functions that allow
14935 more direct access to the VSX instruction set.
14937 @item -mfloat-gprs=@var{yes/single/double/no}
14938 @itemx -mfloat-gprs
14939 @opindex mfloat-gprs
14940 This switch enables or disables the generation of floating point
14941 operations on the general purpose registers for architectures that
14944 The argument @var{yes} or @var{single} enables the use of
14945 single-precision floating point operations.
14947 The argument @var{double} enables the use of single and
14948 double-precision floating point operations.
14950 The argument @var{no} disables floating point operations on the
14951 general purpose registers.
14953 This option is currently only available on the MPC854x.
14959 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
14960 targets (including GNU/Linux). The 32-bit environment sets int, long
14961 and pointer to 32 bits and generates code that runs on any PowerPC
14962 variant. The 64-bit environment sets int to 32 bits and long and
14963 pointer to 64 bits, and generates code for PowerPC64, as for
14964 @option{-mpowerpc64}.
14967 @itemx -mno-fp-in-toc
14968 @itemx -mno-sum-in-toc
14969 @itemx -mminimal-toc
14971 @opindex mno-fp-in-toc
14972 @opindex mno-sum-in-toc
14973 @opindex mminimal-toc
14974 Modify generation of the TOC (Table Of Contents), which is created for
14975 every executable file. The @option{-mfull-toc} option is selected by
14976 default. In that case, GCC will allocate at least one TOC entry for
14977 each unique non-automatic variable reference in your program. GCC
14978 will also place floating-point constants in the TOC@. However, only
14979 16,384 entries are available in the TOC@.
14981 If you receive a linker error message that saying you have overflowed
14982 the available TOC space, you can reduce the amount of TOC space used
14983 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
14984 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
14985 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
14986 generate code to calculate the sum of an address and a constant at
14987 run-time instead of putting that sum into the TOC@. You may specify one
14988 or both of these options. Each causes GCC to produce very slightly
14989 slower and larger code at the expense of conserving TOC space.
14991 If you still run out of space in the TOC even when you specify both of
14992 these options, specify @option{-mminimal-toc} instead. This option causes
14993 GCC to make only one TOC entry for every file. When you specify this
14994 option, GCC will produce code that is slower and larger but which
14995 uses extremely little TOC space. You may wish to use this option
14996 only on files that contain less frequently executed code.
15002 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
15003 @code{long} type, and the infrastructure needed to support them.
15004 Specifying @option{-maix64} implies @option{-mpowerpc64} and
15005 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
15006 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
15009 @itemx -mno-xl-compat
15010 @opindex mxl-compat
15011 @opindex mno-xl-compat
15012 Produce code that conforms more closely to IBM XL compiler semantics
15013 when using AIX-compatible ABI@. Pass floating-point arguments to
15014 prototyped functions beyond the register save area (RSA) on the stack
15015 in addition to argument FPRs. Do not assume that most significant
15016 double in 128-bit long double value is properly rounded when comparing
15017 values and converting to double. Use XL symbol names for long double
15020 The AIX calling convention was extended but not initially documented to
15021 handle an obscure K&R C case of calling a function that takes the
15022 address of its arguments with fewer arguments than declared. IBM XL
15023 compilers access floating point arguments which do not fit in the
15024 RSA from the stack when a subroutine is compiled without
15025 optimization. Because always storing floating-point arguments on the
15026 stack is inefficient and rarely needed, this option is not enabled by
15027 default and only is necessary when calling subroutines compiled by IBM
15028 XL compilers without optimization.
15032 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
15033 application written to use message passing with special startup code to
15034 enable the application to run. The system must have PE installed in the
15035 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
15036 must be overridden with the @option{-specs=} option to specify the
15037 appropriate directory location. The Parallel Environment does not
15038 support threads, so the @option{-mpe} option and the @option{-pthread}
15039 option are incompatible.
15041 @item -malign-natural
15042 @itemx -malign-power
15043 @opindex malign-natural
15044 @opindex malign-power
15045 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
15046 @option{-malign-natural} overrides the ABI-defined alignment of larger
15047 types, such as floating-point doubles, on their natural size-based boundary.
15048 The option @option{-malign-power} instructs GCC to follow the ABI-specified
15049 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
15051 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
15055 @itemx -mhard-float
15056 @opindex msoft-float
15057 @opindex mhard-float
15058 Generate code that does not use (uses) the floating-point register set.
15059 Software floating point emulation is provided if you use the
15060 @option{-msoft-float} option, and pass the option to GCC when linking.
15062 @item -msingle-float
15063 @itemx -mdouble-float
15064 @opindex msingle-float
15065 @opindex mdouble-float
15066 Generate code for single or double-precision floating point operations.
15067 @option{-mdouble-float} implies @option{-msingle-float}.
15070 @opindex msimple-fpu
15071 Do not generate sqrt and div instructions for hardware floating point unit.
15075 Specify type of floating point unit. Valid values are @var{sp_lite}
15076 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
15077 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
15078 and @var{dp_full} (equivalent to -mdouble-float).
15081 @opindex mxilinx-fpu
15082 Perform optimizations for floating point unit on Xilinx PPC 405/440.
15085 @itemx -mno-multiple
15087 @opindex mno-multiple
15088 Generate code that uses (does not use) the load multiple word
15089 instructions and the store multiple word instructions. These
15090 instructions are generated by default on POWER systems, and not
15091 generated on PowerPC systems. Do not use @option{-mmultiple} on little
15092 endian PowerPC systems, since those instructions do not work when the
15093 processor is in little endian mode. The exceptions are PPC740 and
15094 PPC750 which permit the instructions usage in little endian mode.
15099 @opindex mno-string
15100 Generate code that uses (does not use) the load string instructions
15101 and the store string word instructions to save multiple registers and
15102 do small block moves. These instructions are generated by default on
15103 POWER systems, and not generated on PowerPC systems. Do not use
15104 @option{-mstring} on little endian PowerPC systems, since those
15105 instructions do not work when the processor is in little endian mode.
15106 The exceptions are PPC740 and PPC750 which permit the instructions
15107 usage in little endian mode.
15112 @opindex mno-update
15113 Generate code that uses (does not use) the load or store instructions
15114 that update the base register to the address of the calculated memory
15115 location. These instructions are generated by default. If you use
15116 @option{-mno-update}, there is a small window between the time that the
15117 stack pointer is updated and the address of the previous frame is
15118 stored, which means code that walks the stack frame across interrupts or
15119 signals may get corrupted data.
15121 @item -mavoid-indexed-addresses
15122 @itemx -mno-avoid-indexed-addresses
15123 @opindex mavoid-indexed-addresses
15124 @opindex mno-avoid-indexed-addresses
15125 Generate code that tries to avoid (not avoid) the use of indexed load
15126 or store instructions. These instructions can incur a performance
15127 penalty on Power6 processors in certain situations, such as when
15128 stepping through large arrays that cross a 16M boundary. This option
15129 is enabled by default when targetting Power6 and disabled otherwise.
15132 @itemx -mno-fused-madd
15133 @opindex mfused-madd
15134 @opindex mno-fused-madd
15135 Generate code that uses (does not use) the floating point multiply and
15136 accumulate instructions. These instructions are generated by default if
15137 hardware floating is used.
15143 Generate code that uses (does not use) the half-word multiply and
15144 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
15145 These instructions are generated by default when targetting those
15152 Generate code that uses (does not use) the string-search @samp{dlmzb}
15153 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
15154 generated by default when targetting those processors.
15156 @item -mno-bit-align
15158 @opindex mno-bit-align
15159 @opindex mbit-align
15160 On System V.4 and embedded PowerPC systems do not (do) force structures
15161 and unions that contain bit-fields to be aligned to the base type of the
15164 For example, by default a structure containing nothing but 8
15165 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
15166 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
15167 the structure would be aligned to a 1 byte boundary and be one byte in
15170 @item -mno-strict-align
15171 @itemx -mstrict-align
15172 @opindex mno-strict-align
15173 @opindex mstrict-align
15174 On System V.4 and embedded PowerPC systems do not (do) assume that
15175 unaligned memory references will be handled by the system.
15177 @item -mrelocatable
15178 @itemx -mno-relocatable
15179 @opindex mrelocatable
15180 @opindex mno-relocatable
15181 On embedded PowerPC systems generate code that allows (does not allow)
15182 the program to be relocated to a different address at runtime. If you
15183 use @option{-mrelocatable} on any module, all objects linked together must
15184 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
15186 @item -mrelocatable-lib
15187 @itemx -mno-relocatable-lib
15188 @opindex mrelocatable-lib
15189 @opindex mno-relocatable-lib
15190 On embedded PowerPC systems generate code that allows (does not allow)
15191 the program to be relocated to a different address at runtime. Modules
15192 compiled with @option{-mrelocatable-lib} can be linked with either modules
15193 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
15194 with modules compiled with the @option{-mrelocatable} options.
15200 On System V.4 and embedded PowerPC systems do not (do) assume that
15201 register 2 contains a pointer to a global area pointing to the addresses
15202 used in the program.
15205 @itemx -mlittle-endian
15207 @opindex mlittle-endian
15208 On System V.4 and embedded PowerPC systems compile code for the
15209 processor in little endian mode. The @option{-mlittle-endian} option is
15210 the same as @option{-mlittle}.
15213 @itemx -mbig-endian
15215 @opindex mbig-endian
15216 On System V.4 and embedded PowerPC systems compile code for the
15217 processor in big endian mode. The @option{-mbig-endian} option is
15218 the same as @option{-mbig}.
15220 @item -mdynamic-no-pic
15221 @opindex mdynamic-no-pic
15222 On Darwin and Mac OS X systems, compile code so that it is not
15223 relocatable, but that its external references are relocatable. The
15224 resulting code is suitable for applications, but not shared
15227 @item -mprioritize-restricted-insns=@var{priority}
15228 @opindex mprioritize-restricted-insns
15229 This option controls the priority that is assigned to
15230 dispatch-slot restricted instructions during the second scheduling
15231 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
15232 @var{no/highest/second-highest} priority to dispatch slot restricted
15235 @item -msched-costly-dep=@var{dependence_type}
15236 @opindex msched-costly-dep
15237 This option controls which dependences are considered costly
15238 by the target during instruction scheduling. The argument
15239 @var{dependence_type} takes one of the following values:
15240 @var{no}: no dependence is costly,
15241 @var{all}: all dependences are costly,
15242 @var{true_store_to_load}: a true dependence from store to load is costly,
15243 @var{store_to_load}: any dependence from store to load is costly,
15244 @var{number}: any dependence which latency >= @var{number} is costly.
15246 @item -minsert-sched-nops=@var{scheme}
15247 @opindex minsert-sched-nops
15248 This option controls which nop insertion scheme will be used during
15249 the second scheduling pass. The argument @var{scheme} takes one of the
15251 @var{no}: Don't insert nops.
15252 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
15253 according to the scheduler's grouping.
15254 @var{regroup_exact}: Insert nops to force costly dependent insns into
15255 separate groups. Insert exactly as many nops as needed to force an insn
15256 to a new group, according to the estimated processor grouping.
15257 @var{number}: Insert nops to force costly dependent insns into
15258 separate groups. Insert @var{number} nops to force an insn to a new group.
15261 @opindex mcall-sysv
15262 On System V.4 and embedded PowerPC systems compile code using calling
15263 conventions that adheres to the March 1995 draft of the System V
15264 Application Binary Interface, PowerPC processor supplement. This is the
15265 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
15267 @item -mcall-sysv-eabi
15269 @opindex mcall-sysv-eabi
15270 @opindex mcall-eabi
15271 Specify both @option{-mcall-sysv} and @option{-meabi} options.
15273 @item -mcall-sysv-noeabi
15274 @opindex mcall-sysv-noeabi
15275 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
15277 @item -mcall-aixdesc
15279 On System V.4 and embedded PowerPC systems compile code for the AIX
15283 @opindex mcall-linux
15284 On System V.4 and embedded PowerPC systems compile code for the
15285 Linux-based GNU system.
15289 On System V.4 and embedded PowerPC systems compile code for the
15290 Hurd-based GNU system.
15292 @item -mcall-freebsd
15293 @opindex mcall-freebsd
15294 On System V.4 and embedded PowerPC systems compile code for the
15295 FreeBSD operating system.
15297 @item -mcall-netbsd
15298 @opindex mcall-netbsd
15299 On System V.4 and embedded PowerPC systems compile code for the
15300 NetBSD operating system.
15302 @item -mcall-openbsd
15303 @opindex mcall-netbsd
15304 On System V.4 and embedded PowerPC systems compile code for the
15305 OpenBSD operating system.
15307 @item -maix-struct-return
15308 @opindex maix-struct-return
15309 Return all structures in memory (as specified by the AIX ABI)@.
15311 @item -msvr4-struct-return
15312 @opindex msvr4-struct-return
15313 Return structures smaller than 8 bytes in registers (as specified by the
15316 @item -mabi=@var{abi-type}
15318 Extend the current ABI with a particular extension, or remove such extension.
15319 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
15320 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
15324 Extend the current ABI with SPE ABI extensions. This does not change
15325 the default ABI, instead it adds the SPE ABI extensions to the current
15329 @opindex mabi=no-spe
15330 Disable Booke SPE ABI extensions for the current ABI@.
15332 @item -mabi=ibmlongdouble
15333 @opindex mabi=ibmlongdouble
15334 Change the current ABI to use IBM extended precision long double.
15335 This is a PowerPC 32-bit SYSV ABI option.
15337 @item -mabi=ieeelongdouble
15338 @opindex mabi=ieeelongdouble
15339 Change the current ABI to use IEEE extended precision long double.
15340 This is a PowerPC 32-bit Linux ABI option.
15343 @itemx -mno-prototype
15344 @opindex mprototype
15345 @opindex mno-prototype
15346 On System V.4 and embedded PowerPC systems assume that all calls to
15347 variable argument functions are properly prototyped. Otherwise, the
15348 compiler must insert an instruction before every non prototyped call to
15349 set or clear bit 6 of the condition code register (@var{CR}) to
15350 indicate whether floating point values were passed in the floating point
15351 registers in case the function takes a variable arguments. With
15352 @option{-mprototype}, only calls to prototyped variable argument functions
15353 will set or clear the bit.
15357 On embedded PowerPC systems, assume that the startup module is called
15358 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
15359 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
15364 On embedded PowerPC systems, assume that the startup module is called
15365 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
15370 On embedded PowerPC systems, assume that the startup module is called
15371 @file{crt0.o} and the standard C libraries are @file{libads.a} and
15374 @item -myellowknife
15375 @opindex myellowknife
15376 On embedded PowerPC systems, assume that the startup module is called
15377 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
15382 On System V.4 and embedded PowerPC systems, specify that you are
15383 compiling for a VxWorks system.
15387 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
15388 header to indicate that @samp{eabi} extended relocations are used.
15394 On System V.4 and embedded PowerPC systems do (do not) adhere to the
15395 Embedded Applications Binary Interface (eabi) which is a set of
15396 modifications to the System V.4 specifications. Selecting @option{-meabi}
15397 means that the stack is aligned to an 8 byte boundary, a function
15398 @code{__eabi} is called to from @code{main} to set up the eabi
15399 environment, and the @option{-msdata} option can use both @code{r2} and
15400 @code{r13} to point to two separate small data areas. Selecting
15401 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
15402 do not call an initialization function from @code{main}, and the
15403 @option{-msdata} option will only use @code{r13} to point to a single
15404 small data area. The @option{-meabi} option is on by default if you
15405 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
15408 @opindex msdata=eabi
15409 On System V.4 and embedded PowerPC systems, put small initialized
15410 @code{const} global and static data in the @samp{.sdata2} section, which
15411 is pointed to by register @code{r2}. Put small initialized
15412 non-@code{const} global and static data in the @samp{.sdata} section,
15413 which is pointed to by register @code{r13}. Put small uninitialized
15414 global and static data in the @samp{.sbss} section, which is adjacent to
15415 the @samp{.sdata} section. The @option{-msdata=eabi} option is
15416 incompatible with the @option{-mrelocatable} option. The
15417 @option{-msdata=eabi} option also sets the @option{-memb} option.
15420 @opindex msdata=sysv
15421 On System V.4 and embedded PowerPC systems, put small global and static
15422 data in the @samp{.sdata} section, which is pointed to by register
15423 @code{r13}. Put small uninitialized global and static data in the
15424 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
15425 The @option{-msdata=sysv} option is incompatible with the
15426 @option{-mrelocatable} option.
15428 @item -msdata=default
15430 @opindex msdata=default
15432 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
15433 compile code the same as @option{-msdata=eabi}, otherwise compile code the
15434 same as @option{-msdata=sysv}.
15437 @opindex msdata=data
15438 On System V.4 and embedded PowerPC systems, put small global
15439 data in the @samp{.sdata} section. Put small uninitialized global
15440 data in the @samp{.sbss} section. Do not use register @code{r13}
15441 to address small data however. This is the default behavior unless
15442 other @option{-msdata} options are used.
15446 @opindex msdata=none
15448 On embedded PowerPC systems, put all initialized global and static data
15449 in the @samp{.data} section, and all uninitialized data in the
15450 @samp{.bss} section.
15454 @cindex smaller data references (PowerPC)
15455 @cindex .sdata/.sdata2 references (PowerPC)
15456 On embedded PowerPC systems, put global and static items less than or
15457 equal to @var{num} bytes into the small data or bss sections instead of
15458 the normal data or bss section. By default, @var{num} is 8. The
15459 @option{-G @var{num}} switch is also passed to the linker.
15460 All modules should be compiled with the same @option{-G @var{num}} value.
15463 @itemx -mno-regnames
15465 @opindex mno-regnames
15466 On System V.4 and embedded PowerPC systems do (do not) emit register
15467 names in the assembly language output using symbolic forms.
15470 @itemx -mno-longcall
15472 @opindex mno-longcall
15473 By default assume that all calls are far away so that a longer more
15474 expensive calling sequence is required. This is required for calls
15475 further than 32 megabytes (33,554,432 bytes) from the current location.
15476 A short call will be generated if the compiler knows
15477 the call cannot be that far away. This setting can be overridden by
15478 the @code{shortcall} function attribute, or by @code{#pragma
15481 Some linkers are capable of detecting out-of-range calls and generating
15482 glue code on the fly. On these systems, long calls are unnecessary and
15483 generate slower code. As of this writing, the AIX linker can do this,
15484 as can the GNU linker for PowerPC/64. It is planned to add this feature
15485 to the GNU linker for 32-bit PowerPC systems as well.
15487 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
15488 callee, L42'', plus a ``branch island'' (glue code). The two target
15489 addresses represent the callee and the ``branch island''. The
15490 Darwin/PPC linker will prefer the first address and generate a ``bl
15491 callee'' if the PPC ``bl'' instruction will reach the callee directly;
15492 otherwise, the linker will generate ``bl L42'' to call the ``branch
15493 island''. The ``branch island'' is appended to the body of the
15494 calling function; it computes the full 32-bit address of the callee
15497 On Mach-O (Darwin) systems, this option directs the compiler emit to
15498 the glue for every direct call, and the Darwin linker decides whether
15499 to use or discard it.
15501 In the future, we may cause GCC to ignore all longcall specifications
15502 when the linker is known to generate glue.
15504 @item -mtls-markers
15505 @itemx -mno-tls-markers
15506 @opindex mtls-markers
15507 @opindex mno-tls-markers
15508 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
15509 specifying the function argument. The relocation allows ld to
15510 reliably associate function call with argument setup instructions for
15511 TLS optimization, which in turn allows gcc to better schedule the
15516 Adds support for multithreading with the @dfn{pthreads} library.
15517 This option sets flags for both the preprocessor and linker.
15522 @subsection RX Options
15525 These command line options are defined for RX targets:
15528 @item -m64bit-doubles
15529 @itemx -m32bit-doubles
15530 @opindex m64bit-doubles
15531 @opindex m32bit-doubles
15532 Make the @code{double} data type be 64-bits (@option{-m64bit-doubles})
15533 or 32-bits (@option{-m32bit-doubles}) in size. The default is
15534 @option{-m32bit-doubles}. @emph{Note} RX floating point hardware only
15535 works on 32-bit values, which is why the default is
15536 @option{-m32bit-doubles}.
15542 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
15543 floating point hardware. The default is enabled for the @var{RX600}
15544 series and disabled for the @var{RX200} series.
15546 Floating point instructions will only be generated for 32-bit floating
15547 point values however, so if the @option{-m64bit-doubles} option is in
15548 use then the FPU hardware will not be used for doubles.
15550 @emph{Note} If the @option{-fpu} option is enabled then
15551 @option{-funsafe-math-optimizations} is also enabled automatically.
15552 This is because the RX FPU instructions are themselves unsafe.
15554 @item -mcpu=@var{name}
15555 @itemx -patch=@var{name}
15558 Selects the type of RX CPU to be targeted. Currently three types are
15559 supported, the generic @var{RX600} and @var{RX200} series hardware and
15560 the specific @var{RX610} cpu. The default is @var{RX600}.
15562 The only difference between @var{RX600} and @var{RX610} is that the
15563 @var{RX610} does not support the @code{MVTIPL} instruction.
15565 The @var{RX200} series does not have a hardware floating point unit
15566 and so @option{-nofpu} is enabled by default when this type is
15569 @item -mbig-endian-data
15570 @itemx -mlittle-endian-data
15571 @opindex mbig-endian-data
15572 @opindex mlittle-endian-data
15573 Store data (but not code) in the big-endian format. The default is
15574 @option{-mlittle-endian-data}, ie to store data in the little endian
15577 @item -msmall-data-limit=@var{N}
15578 @opindex msmall-data-limit
15579 Specifies the maximum size in bytes of global and static variables
15580 which can be placed into the small data area. Using the small data
15581 area can lead to smaller and faster code, but the size of area is
15582 limited and it is up to the programmer to ensure that the area does
15583 not overflow. Also when the small data area is used one of the RX's
15584 registers (@code{r13}) is reserved for use pointing to this area, so
15585 it is no longer available for use by the compiler. This could result
15586 in slower and/or larger code if variables which once could have been
15587 held in @code{r13} are now pushed onto the stack.
15589 Note, common variables (variables which have not been initialised) and
15590 constants are not placed into the small data area as they are assigned
15591 to other sections in the output executable.
15593 The default value is zero, which disables this feature. Note, this
15594 feature is not enabled by default with higher optimization levels
15595 (@option{-O2} etc) because of the potentially detrimental effects of
15596 reserving register @code{r13}. It is up to the programmer to
15597 experiment and discover whether this feature is of benefit to their
15604 Use the simulator runtime. The default is to use the libgloss board
15607 @item -mas100-syntax
15608 @itemx -mno-as100-syntax
15609 @opindex mas100-syntax
15610 @opindex mno-as100-syntax
15611 When generating assembler output use a syntax that is compatible with
15612 Renesas's AS100 assembler. This syntax can also be handled by the GAS
15613 assembler but it has some restrictions so generating it is not the
15616 @item -mmax-constant-size=@var{N}
15617 @opindex mmax-constant-size
15618 Specifies the maximum size, in bytes, of a constant that can be used as
15619 an operand in a RX instruction. Although the RX instruction set does
15620 allow constants of up to 4 bytes in length to be used in instructions,
15621 a longer value equates to a longer instruction. Thus in some
15622 circumstances it can be beneficial to restrict the size of constants
15623 that are used in instructions. Constants that are too big are instead
15624 placed into a constant pool and referenced via register indirection.
15626 The value @var{N} can be between 0 and 4. A value of 0 (the default)
15627 or 4 means that constants of any size are allowed.
15631 Enable linker relaxation. Linker relaxation is a process whereby the
15632 linker will attempt to reduce the size of a program by finding shorter
15633 versions of various instructions. Disabled by default.
15635 @item -mint-register=@var{N}
15636 @opindex mint-register
15637 Specify the number of registers to reserve for fast interrupt handler
15638 functions. The value @var{N} can be between 0 and 4. A value of 1
15639 means that register @code{r13} will be reserved for the exclusive use
15640 of fast interrupt handlers. A value of 2 reserves @code{r13} and
15641 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
15642 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
15643 A value of 0, the default, does not reserve any registers.
15645 @item -msave-acc-in-interrupts
15646 @opindex msave-acc-in-interrupts
15647 Specifies that interrupt handler functions should preserve the
15648 accumulator register. This is only necessary if normal code might use
15649 the accumulator register, for example because it performs 64-bit
15650 multiplications. The default is to ignore the accumulator as this
15651 makes the interrupt handlers faster.
15655 @emph{Note:} The generic GCC command line @option{-ffixed-@var{reg}}
15656 has special significance to the RX port when used with the
15657 @code{interrupt} function attribute. This attribute indicates a
15658 function intended to process fast interrupts. GCC will will ensure
15659 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
15660 and/or @code{r13} and only provided that the normal use of the
15661 corresponding registers have been restricted via the
15662 @option{-ffixed-@var{reg}} or @option{-mint-register} command line
15665 @node S/390 and zSeries Options
15666 @subsection S/390 and zSeries Options
15667 @cindex S/390 and zSeries Options
15669 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
15673 @itemx -msoft-float
15674 @opindex mhard-float
15675 @opindex msoft-float
15676 Use (do not use) the hardware floating-point instructions and registers
15677 for floating-point operations. When @option{-msoft-float} is specified,
15678 functions in @file{libgcc.a} will be used to perform floating-point
15679 operations. When @option{-mhard-float} is specified, the compiler
15680 generates IEEE floating-point instructions. This is the default.
15683 @itemx -mno-hard-dfp
15685 @opindex mno-hard-dfp
15686 Use (do not use) the hardware decimal-floating-point instructions for
15687 decimal-floating-point operations. When @option{-mno-hard-dfp} is
15688 specified, functions in @file{libgcc.a} will be used to perform
15689 decimal-floating-point operations. When @option{-mhard-dfp} is
15690 specified, the compiler generates decimal-floating-point hardware
15691 instructions. This is the default for @option{-march=z9-ec} or higher.
15693 @item -mlong-double-64
15694 @itemx -mlong-double-128
15695 @opindex mlong-double-64
15696 @opindex mlong-double-128
15697 These switches control the size of @code{long double} type. A size
15698 of 64bit makes the @code{long double} type equivalent to the @code{double}
15699 type. This is the default.
15702 @itemx -mno-backchain
15703 @opindex mbackchain
15704 @opindex mno-backchain
15705 Store (do not store) the address of the caller's frame as backchain pointer
15706 into the callee's stack frame.
15707 A backchain may be needed to allow debugging using tools that do not understand
15708 DWARF-2 call frame information.
15709 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
15710 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
15711 the backchain is placed into the topmost word of the 96/160 byte register
15714 In general, code compiled with @option{-mbackchain} is call-compatible with
15715 code compiled with @option{-mmo-backchain}; however, use of the backchain
15716 for debugging purposes usually requires that the whole binary is built with
15717 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
15718 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
15719 to build a linux kernel use @option{-msoft-float}.
15721 The default is to not maintain the backchain.
15723 @item -mpacked-stack
15724 @itemx -mno-packed-stack
15725 @opindex mpacked-stack
15726 @opindex mno-packed-stack
15727 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
15728 specified, the compiler uses the all fields of the 96/160 byte register save
15729 area only for their default purpose; unused fields still take up stack space.
15730 When @option{-mpacked-stack} is specified, register save slots are densely
15731 packed at the top of the register save area; unused space is reused for other
15732 purposes, allowing for more efficient use of the available stack space.
15733 However, when @option{-mbackchain} is also in effect, the topmost word of
15734 the save area is always used to store the backchain, and the return address
15735 register is always saved two words below the backchain.
15737 As long as the stack frame backchain is not used, code generated with
15738 @option{-mpacked-stack} is call-compatible with code generated with
15739 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
15740 S/390 or zSeries generated code that uses the stack frame backchain at run
15741 time, not just for debugging purposes. Such code is not call-compatible
15742 with code compiled with @option{-mpacked-stack}. Also, note that the
15743 combination of @option{-mbackchain},
15744 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
15745 to build a linux kernel use @option{-msoft-float}.
15747 The default is to not use the packed stack layout.
15750 @itemx -mno-small-exec
15751 @opindex msmall-exec
15752 @opindex mno-small-exec
15753 Generate (or do not generate) code using the @code{bras} instruction
15754 to do subroutine calls.
15755 This only works reliably if the total executable size does not
15756 exceed 64k. The default is to use the @code{basr} instruction instead,
15757 which does not have this limitation.
15763 When @option{-m31} is specified, generate code compliant to the
15764 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
15765 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
15766 particular to generate 64-bit instructions. For the @samp{s390}
15767 targets, the default is @option{-m31}, while the @samp{s390x}
15768 targets default to @option{-m64}.
15774 When @option{-mzarch} is specified, generate code using the
15775 instructions available on z/Architecture.
15776 When @option{-mesa} is specified, generate code using the
15777 instructions available on ESA/390. Note that @option{-mesa} is
15778 not possible with @option{-m64}.
15779 When generating code compliant to the GNU/Linux for S/390 ABI,
15780 the default is @option{-mesa}. When generating code compliant
15781 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
15787 Generate (or do not generate) code using the @code{mvcle} instruction
15788 to perform block moves. When @option{-mno-mvcle} is specified,
15789 use a @code{mvc} loop instead. This is the default unless optimizing for
15796 Print (or do not print) additional debug information when compiling.
15797 The default is to not print debug information.
15799 @item -march=@var{cpu-type}
15801 Generate code that will run on @var{cpu-type}, which is the name of a system
15802 representing a certain processor type. Possible values for
15803 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
15804 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
15805 When generating code using the instructions available on z/Architecture,
15806 the default is @option{-march=z900}. Otherwise, the default is
15807 @option{-march=g5}.
15809 @item -mtune=@var{cpu-type}
15811 Tune to @var{cpu-type} everything applicable about the generated code,
15812 except for the ABI and the set of available instructions.
15813 The list of @var{cpu-type} values is the same as for @option{-march}.
15814 The default is the value used for @option{-march}.
15817 @itemx -mno-tpf-trace
15818 @opindex mtpf-trace
15819 @opindex mno-tpf-trace
15820 Generate code that adds (does not add) in TPF OS specific branches to trace
15821 routines in the operating system. This option is off by default, even
15822 when compiling for the TPF OS@.
15825 @itemx -mno-fused-madd
15826 @opindex mfused-madd
15827 @opindex mno-fused-madd
15828 Generate code that uses (does not use) the floating point multiply and
15829 accumulate instructions. These instructions are generated by default if
15830 hardware floating point is used.
15832 @item -mwarn-framesize=@var{framesize}
15833 @opindex mwarn-framesize
15834 Emit a warning if the current function exceeds the given frame size. Because
15835 this is a compile time check it doesn't need to be a real problem when the program
15836 runs. It is intended to identify functions which most probably cause
15837 a stack overflow. It is useful to be used in an environment with limited stack
15838 size e.g.@: the linux kernel.
15840 @item -mwarn-dynamicstack
15841 @opindex mwarn-dynamicstack
15842 Emit a warning if the function calls alloca or uses dynamically
15843 sized arrays. This is generally a bad idea with a limited stack size.
15845 @item -mstack-guard=@var{stack-guard}
15846 @itemx -mstack-size=@var{stack-size}
15847 @opindex mstack-guard
15848 @opindex mstack-size
15849 If these options are provided the s390 back end emits additional instructions in
15850 the function prologue which trigger a trap if the stack size is @var{stack-guard}
15851 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
15852 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
15853 the frame size of the compiled function is chosen.
15854 These options are intended to be used to help debugging stack overflow problems.
15855 The additionally emitted code causes only little overhead and hence can also be
15856 used in production like systems without greater performance degradation. The given
15857 values have to be exact powers of 2 and @var{stack-size} has to be greater than
15858 @var{stack-guard} without exceeding 64k.
15859 In order to be efficient the extra code makes the assumption that the stack starts
15860 at an address aligned to the value given by @var{stack-size}.
15861 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
15864 @node Score Options
15865 @subsection Score Options
15866 @cindex Score Options
15868 These options are defined for Score implementations:
15873 Compile code for big endian mode. This is the default.
15877 Compile code for little endian mode.
15881 Disable generate bcnz instruction.
15885 Enable generate unaligned load and store instruction.
15889 Enable the use of multiply-accumulate instructions. Disabled by default.
15893 Specify the SCORE5 as the target architecture.
15897 Specify the SCORE5U of the target architecture.
15901 Specify the SCORE7 as the target architecture. This is the default.
15905 Specify the SCORE7D as the target architecture.
15909 @subsection SH Options
15911 These @samp{-m} options are defined for the SH implementations:
15916 Generate code for the SH1.
15920 Generate code for the SH2.
15923 Generate code for the SH2e.
15927 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
15928 that the floating-point unit is not used.
15930 @item -m2a-single-only
15931 @opindex m2a-single-only
15932 Generate code for the SH2a-FPU, in such a way that no double-precision
15933 floating point operations are used.
15936 @opindex m2a-single
15937 Generate code for the SH2a-FPU assuming the floating-point unit is in
15938 single-precision mode by default.
15942 Generate code for the SH2a-FPU assuming the floating-point unit is in
15943 double-precision mode by default.
15947 Generate code for the SH3.
15951 Generate code for the SH3e.
15955 Generate code for the SH4 without a floating-point unit.
15957 @item -m4-single-only
15958 @opindex m4-single-only
15959 Generate code for the SH4 with a floating-point unit that only
15960 supports single-precision arithmetic.
15964 Generate code for the SH4 assuming the floating-point unit is in
15965 single-precision mode by default.
15969 Generate code for the SH4.
15973 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
15974 floating-point unit is not used.
15976 @item -m4a-single-only
15977 @opindex m4a-single-only
15978 Generate code for the SH4a, in such a way that no double-precision
15979 floating point operations are used.
15982 @opindex m4a-single
15983 Generate code for the SH4a assuming the floating-point unit is in
15984 single-precision mode by default.
15988 Generate code for the SH4a.
15992 Same as @option{-m4a-nofpu}, except that it implicitly passes
15993 @option{-dsp} to the assembler. GCC doesn't generate any DSP
15994 instructions at the moment.
15998 Compile code for the processor in big endian mode.
16002 Compile code for the processor in little endian mode.
16006 Align doubles at 64-bit boundaries. Note that this changes the calling
16007 conventions, and thus some functions from the standard C library will
16008 not work unless you recompile it first with @option{-mdalign}.
16012 Shorten some address references at link time, when possible; uses the
16013 linker option @option{-relax}.
16017 Use 32-bit offsets in @code{switch} tables. The default is to use
16022 Enable the use of bit manipulation instructions on SH2A.
16026 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
16027 alignment constraints.
16031 Comply with the calling conventions defined by Renesas.
16035 Comply with the calling conventions defined by Renesas.
16039 Comply with the calling conventions defined for GCC before the Renesas
16040 conventions were available. This option is the default for all
16041 targets of the SH toolchain except for @samp{sh-symbianelf}.
16044 @opindex mnomacsave
16045 Mark the @code{MAC} register as call-clobbered, even if
16046 @option{-mhitachi} is given.
16050 Increase IEEE-compliance of floating-point code.
16051 At the moment, this is equivalent to @option{-fno-finite-math-only}.
16052 When generating 16 bit SH opcodes, getting IEEE-conforming results for
16053 comparisons of NANs / infinities incurs extra overhead in every
16054 floating point comparison, therefore the default is set to
16055 @option{-ffinite-math-only}.
16057 @item -minline-ic_invalidate
16058 @opindex minline-ic_invalidate
16059 Inline code to invalidate instruction cache entries after setting up
16060 nested function trampolines.
16061 This option has no effect if -musermode is in effect and the selected
16062 code generation option (e.g. -m4) does not allow the use of the icbi
16064 If the selected code generation option does not allow the use of the icbi
16065 instruction, and -musermode is not in effect, the inlined code will
16066 manipulate the instruction cache address array directly with an associative
16067 write. This not only requires privileged mode, but it will also
16068 fail if the cache line had been mapped via the TLB and has become unmapped.
16072 Dump instruction size and location in the assembly code.
16075 @opindex mpadstruct
16076 This option is deprecated. It pads structures to multiple of 4 bytes,
16077 which is incompatible with the SH ABI@.
16081 Optimize for space instead of speed. Implied by @option{-Os}.
16084 @opindex mprefergot
16085 When generating position-independent code, emit function calls using
16086 the Global Offset Table instead of the Procedure Linkage Table.
16090 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
16091 if the inlined code would not work in user mode.
16092 This is the default when the target is @code{sh-*-linux*}.
16094 @item -multcost=@var{number}
16095 @opindex multcost=@var{number}
16096 Set the cost to assume for a multiply insn.
16098 @item -mdiv=@var{strategy}
16099 @opindex mdiv=@var{strategy}
16100 Set the division strategy to use for SHmedia code. @var{strategy} must be
16101 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
16102 inv:call2, inv:fp .
16103 "fp" performs the operation in floating point. This has a very high latency,
16104 but needs only a few instructions, so it might be a good choice if
16105 your code has enough easily exploitable ILP to allow the compiler to
16106 schedule the floating point instructions together with other instructions.
16107 Division by zero causes a floating point exception.
16108 "inv" uses integer operations to calculate the inverse of the divisor,
16109 and then multiplies the dividend with the inverse. This strategy allows
16110 cse and hoisting of the inverse calculation. Division by zero calculates
16111 an unspecified result, but does not trap.
16112 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
16113 have been found, or if the entire operation has been hoisted to the same
16114 place, the last stages of the inverse calculation are intertwined with the
16115 final multiply to reduce the overall latency, at the expense of using a few
16116 more instructions, and thus offering fewer scheduling opportunities with
16118 "call" calls a library function that usually implements the inv:minlat
16120 This gives high code density for m5-*media-nofpu compilations.
16121 "call2" uses a different entry point of the same library function, where it
16122 assumes that a pointer to a lookup table has already been set up, which
16123 exposes the pointer load to cse / code hoisting optimizations.
16124 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
16125 code generation, but if the code stays unoptimized, revert to the "call",
16126 "call2", or "fp" strategies, respectively. Note that the
16127 potentially-trapping side effect of division by zero is carried by a
16128 separate instruction, so it is possible that all the integer instructions
16129 are hoisted out, but the marker for the side effect stays where it is.
16130 A recombination to fp operations or a call is not possible in that case.
16131 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
16132 that the inverse calculation was nor separated from the multiply, they speed
16133 up division where the dividend fits into 20 bits (plus sign where applicable),
16134 by inserting a test to skip a number of operations in this case; this test
16135 slows down the case of larger dividends. inv20u assumes the case of a such
16136 a small dividend to be unlikely, and inv20l assumes it to be likely.
16138 @item -mdivsi3_libfunc=@var{name}
16139 @opindex mdivsi3_libfunc=@var{name}
16140 Set the name of the library function used for 32 bit signed division to
16141 @var{name}. This only affect the name used in the call and inv:call
16142 division strategies, and the compiler will still expect the same
16143 sets of input/output/clobbered registers as if this option was not present.
16145 @item -mfixed-range=@var{register-range}
16146 @opindex mfixed-range
16147 Generate code treating the given register range as fixed registers.
16148 A fixed register is one that the register allocator can not use. This is
16149 useful when compiling kernel code. A register range is specified as
16150 two registers separated by a dash. Multiple register ranges can be
16151 specified separated by a comma.
16153 @item -madjust-unroll
16154 @opindex madjust-unroll
16155 Throttle unrolling to avoid thrashing target registers.
16156 This option only has an effect if the gcc code base supports the
16157 TARGET_ADJUST_UNROLL_MAX target hook.
16159 @item -mindexed-addressing
16160 @opindex mindexed-addressing
16161 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
16162 This is only safe if the hardware and/or OS implement 32 bit wrap-around
16163 semantics for the indexed addressing mode. The architecture allows the
16164 implementation of processors with 64 bit MMU, which the OS could use to
16165 get 32 bit addressing, but since no current hardware implementation supports
16166 this or any other way to make the indexed addressing mode safe to use in
16167 the 32 bit ABI, the default is -mno-indexed-addressing.
16169 @item -mgettrcost=@var{number}
16170 @opindex mgettrcost=@var{number}
16171 Set the cost assumed for the gettr instruction to @var{number}.
16172 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
16176 Assume pt* instructions won't trap. This will generally generate better
16177 scheduled code, but is unsafe on current hardware. The current architecture
16178 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
16179 This has the unintentional effect of making it unsafe to schedule ptabs /
16180 ptrel before a branch, or hoist it out of a loop. For example,
16181 __do_global_ctors, a part of libgcc that runs constructors at program
16182 startup, calls functions in a list which is delimited by @minus{}1. With the
16183 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
16184 That means that all the constructors will be run a bit quicker, but when
16185 the loop comes to the end of the list, the program crashes because ptabs
16186 loads @minus{}1 into a target register. Since this option is unsafe for any
16187 hardware implementing the current architecture specification, the default
16188 is -mno-pt-fixed. Unless the user specifies a specific cost with
16189 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
16190 this deters register allocation using target registers for storing
16193 @item -minvalid-symbols
16194 @opindex minvalid-symbols
16195 Assume symbols might be invalid. Ordinary function symbols generated by
16196 the compiler will always be valid to load with movi/shori/ptabs or
16197 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
16198 to generate symbols that will cause ptabs / ptrel to trap.
16199 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
16200 It will then prevent cross-basic-block cse, hoisting and most scheduling
16201 of symbol loads. The default is @option{-mno-invalid-symbols}.
16204 @node SPARC Options
16205 @subsection SPARC Options
16206 @cindex SPARC options
16208 These @samp{-m} options are supported on the SPARC:
16211 @item -mno-app-regs
16213 @opindex mno-app-regs
16215 Specify @option{-mapp-regs} to generate output using the global registers
16216 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
16219 To be fully SVR4 ABI compliant at the cost of some performance loss,
16220 specify @option{-mno-app-regs}. You should compile libraries and system
16221 software with this option.
16224 @itemx -mhard-float
16226 @opindex mhard-float
16227 Generate output containing floating point instructions. This is the
16231 @itemx -msoft-float
16233 @opindex msoft-float
16234 Generate output containing library calls for floating point.
16235 @strong{Warning:} the requisite libraries are not available for all SPARC
16236 targets. Normally the facilities of the machine's usual C compiler are
16237 used, but this cannot be done directly in cross-compilation. You must make
16238 your own arrangements to provide suitable library functions for
16239 cross-compilation. The embedded targets @samp{sparc-*-aout} and
16240 @samp{sparclite-*-*} do provide software floating point support.
16242 @option{-msoft-float} changes the calling convention in the output file;
16243 therefore, it is only useful if you compile @emph{all} of a program with
16244 this option. In particular, you need to compile @file{libgcc.a}, the
16245 library that comes with GCC, with @option{-msoft-float} in order for
16248 @item -mhard-quad-float
16249 @opindex mhard-quad-float
16250 Generate output containing quad-word (long double) floating point
16253 @item -msoft-quad-float
16254 @opindex msoft-quad-float
16255 Generate output containing library calls for quad-word (long double)
16256 floating point instructions. The functions called are those specified
16257 in the SPARC ABI@. This is the default.
16259 As of this writing, there are no SPARC implementations that have hardware
16260 support for the quad-word floating point instructions. They all invoke
16261 a trap handler for one of these instructions, and then the trap handler
16262 emulates the effect of the instruction. Because of the trap handler overhead,
16263 this is much slower than calling the ABI library routines. Thus the
16264 @option{-msoft-quad-float} option is the default.
16266 @item -mno-unaligned-doubles
16267 @itemx -munaligned-doubles
16268 @opindex mno-unaligned-doubles
16269 @opindex munaligned-doubles
16270 Assume that doubles have 8 byte alignment. This is the default.
16272 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
16273 alignment only if they are contained in another type, or if they have an
16274 absolute address. Otherwise, it assumes they have 4 byte alignment.
16275 Specifying this option avoids some rare compatibility problems with code
16276 generated by other compilers. It is not the default because it results
16277 in a performance loss, especially for floating point code.
16279 @item -mno-faster-structs
16280 @itemx -mfaster-structs
16281 @opindex mno-faster-structs
16282 @opindex mfaster-structs
16283 With @option{-mfaster-structs}, the compiler assumes that structures
16284 should have 8 byte alignment. This enables the use of pairs of
16285 @code{ldd} and @code{std} instructions for copies in structure
16286 assignment, in place of twice as many @code{ld} and @code{st} pairs.
16287 However, the use of this changed alignment directly violates the SPARC
16288 ABI@. Thus, it's intended only for use on targets where the developer
16289 acknowledges that their resulting code will not be directly in line with
16290 the rules of the ABI@.
16292 @item -mimpure-text
16293 @opindex mimpure-text
16294 @option{-mimpure-text}, used in addition to @option{-shared}, tells
16295 the compiler to not pass @option{-z text} to the linker when linking a
16296 shared object. Using this option, you can link position-dependent
16297 code into a shared object.
16299 @option{-mimpure-text} suppresses the ``relocations remain against
16300 allocatable but non-writable sections'' linker error message.
16301 However, the necessary relocations will trigger copy-on-write, and the
16302 shared object is not actually shared across processes. Instead of
16303 using @option{-mimpure-text}, you should compile all source code with
16304 @option{-fpic} or @option{-fPIC}.
16306 This option is only available on SunOS and Solaris.
16308 @item -mcpu=@var{cpu_type}
16310 Set the instruction set, register set, and instruction scheduling parameters
16311 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
16312 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
16313 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
16314 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
16315 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
16317 Default instruction scheduling parameters are used for values that select
16318 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
16319 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
16321 Here is a list of each supported architecture and their supported
16326 v8: supersparc, hypersparc
16327 sparclite: f930, f934, sparclite86x
16329 v9: ultrasparc, ultrasparc3, niagara, niagara2
16332 By default (unless configured otherwise), GCC generates code for the V7
16333 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
16334 additionally optimizes it for the Cypress CY7C602 chip, as used in the
16335 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
16336 SPARCStation 1, 2, IPX etc.
16338 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
16339 architecture. The only difference from V7 code is that the compiler emits
16340 the integer multiply and integer divide instructions which exist in SPARC-V8
16341 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
16342 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
16345 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
16346 the SPARC architecture. This adds the integer multiply, integer divide step
16347 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
16348 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
16349 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
16350 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
16351 MB86934 chip, which is the more recent SPARClite with FPU@.
16353 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
16354 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
16355 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
16356 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
16357 optimizes it for the TEMIC SPARClet chip.
16359 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
16360 architecture. This adds 64-bit integer and floating-point move instructions,
16361 3 additional floating-point condition code registers and conditional move
16362 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
16363 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
16364 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
16365 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
16366 @option{-mcpu=niagara}, the compiler additionally optimizes it for
16367 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
16368 additionally optimizes it for Sun UltraSPARC T2 chips.
16370 @item -mtune=@var{cpu_type}
16372 Set the instruction scheduling parameters for machine type
16373 @var{cpu_type}, but do not set the instruction set or register set that the
16374 option @option{-mcpu=@var{cpu_type}} would.
16376 The same values for @option{-mcpu=@var{cpu_type}} can be used for
16377 @option{-mtune=@var{cpu_type}}, but the only useful values are those
16378 that select a particular cpu implementation. Those are @samp{cypress},
16379 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
16380 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
16381 @samp{ultrasparc3}, @samp{niagara}, and @samp{niagara2}.
16386 @opindex mno-v8plus
16387 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
16388 difference from the V8 ABI is that the global and out registers are
16389 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
16390 mode for all SPARC-V9 processors.
16396 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
16397 Visual Instruction Set extensions. The default is @option{-mno-vis}.
16400 These @samp{-m} options are supported in addition to the above
16401 on SPARC-V9 processors in 64-bit environments:
16404 @item -mlittle-endian
16405 @opindex mlittle-endian
16406 Generate code for a processor running in little-endian mode. It is only
16407 available for a few configurations and most notably not on Solaris and Linux.
16413 Generate code for a 32-bit or 64-bit environment.
16414 The 32-bit environment sets int, long and pointer to 32 bits.
16415 The 64-bit environment sets int to 32 bits and long and pointer
16418 @item -mcmodel=medlow
16419 @opindex mcmodel=medlow
16420 Generate code for the Medium/Low code model: 64-bit addresses, programs
16421 must be linked in the low 32 bits of memory. Programs can be statically
16422 or dynamically linked.
16424 @item -mcmodel=medmid
16425 @opindex mcmodel=medmid
16426 Generate code for the Medium/Middle code model: 64-bit addresses, programs
16427 must be linked in the low 44 bits of memory, the text and data segments must
16428 be less than 2GB in size and the data segment must be located within 2GB of
16431 @item -mcmodel=medany
16432 @opindex mcmodel=medany
16433 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
16434 may be linked anywhere in memory, the text and data segments must be less
16435 than 2GB in size and the data segment must be located within 2GB of the
16438 @item -mcmodel=embmedany
16439 @opindex mcmodel=embmedany
16440 Generate code for the Medium/Anywhere code model for embedded systems:
16441 64-bit addresses, the text and data segments must be less than 2GB in
16442 size, both starting anywhere in memory (determined at link time). The
16443 global register %g4 points to the base of the data segment. Programs
16444 are statically linked and PIC is not supported.
16447 @itemx -mno-stack-bias
16448 @opindex mstack-bias
16449 @opindex mno-stack-bias
16450 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
16451 frame pointer if present, are offset by @minus{}2047 which must be added back
16452 when making stack frame references. This is the default in 64-bit mode.
16453 Otherwise, assume no such offset is present.
16456 These switches are supported in addition to the above on Solaris:
16461 Add support for multithreading using the Solaris threads library. This
16462 option sets flags for both the preprocessor and linker. This option does
16463 not affect the thread safety of object code produced by the compiler or
16464 that of libraries supplied with it.
16468 Add support for multithreading using the POSIX threads library. This
16469 option sets flags for both the preprocessor and linker. This option does
16470 not affect the thread safety of object code produced by the compiler or
16471 that of libraries supplied with it.
16475 This is a synonym for @option{-pthreads}.
16479 @subsection SPU Options
16480 @cindex SPU options
16482 These @samp{-m} options are supported on the SPU:
16486 @itemx -merror-reloc
16487 @opindex mwarn-reloc
16488 @opindex merror-reloc
16490 The loader for SPU does not handle dynamic relocations. By default, GCC
16491 will give an error when it generates code that requires a dynamic
16492 relocation. @option{-mno-error-reloc} disables the error,
16493 @option{-mwarn-reloc} will generate a warning instead.
16496 @itemx -munsafe-dma
16498 @opindex munsafe-dma
16500 Instructions which initiate or test completion of DMA must not be
16501 reordered with respect to loads and stores of the memory which is being
16502 accessed. Users typically address this problem using the volatile
16503 keyword, but that can lead to inefficient code in places where the
16504 memory is known to not change. Rather than mark the memory as volatile
16505 we treat the DMA instructions as potentially effecting all memory. With
16506 @option{-munsafe-dma} users must use the volatile keyword to protect
16509 @item -mbranch-hints
16510 @opindex mbranch-hints
16512 By default, GCC will generate a branch hint instruction to avoid
16513 pipeline stalls for always taken or probably taken branches. A hint
16514 will not be generated closer than 8 instructions away from its branch.
16515 There is little reason to disable them, except for debugging purposes,
16516 or to make an object a little bit smaller.
16520 @opindex msmall-mem
16521 @opindex mlarge-mem
16523 By default, GCC generates code assuming that addresses are never larger
16524 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
16525 a full 32 bit address.
16530 By default, GCC links against startup code that assumes the SPU-style
16531 main function interface (which has an unconventional parameter list).
16532 With @option{-mstdmain}, GCC will link your program against startup
16533 code that assumes a C99-style interface to @code{main}, including a
16534 local copy of @code{argv} strings.
16536 @item -mfixed-range=@var{register-range}
16537 @opindex mfixed-range
16538 Generate code treating the given register range as fixed registers.
16539 A fixed register is one that the register allocator can not use. This is
16540 useful when compiling kernel code. A register range is specified as
16541 two registers separated by a dash. Multiple register ranges can be
16542 specified separated by a comma.
16548 Compile code assuming that pointers to the PPU address space accessed
16549 via the @code{__ea} named address space qualifier are either 32 or 64
16550 bits wide. The default is 32 bits. As this is an ABI changing option,
16551 all object code in an executable must be compiled with the same setting.
16553 @item -maddress-space-conversion
16554 @itemx -mno-address-space-conversion
16555 @opindex maddress-space-conversion
16556 @opindex mno-address-space-conversion
16557 Allow/disallow treating the @code{__ea} address space as superset
16558 of the generic address space. This enables explicit type casts
16559 between @code{__ea} and generic pointer as well as implicit
16560 conversions of generic pointers to @code{__ea} pointers. The
16561 default is to allow address space pointer conversions.
16563 @item -mcache-size=@var{cache-size}
16564 @opindex mcache-size
16565 This option controls the version of libgcc that the compiler links to an
16566 executable and selects a software-managed cache for accessing variables
16567 in the @code{__ea} address space with a particular cache size. Possible
16568 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
16569 and @samp{128}. The default cache size is 64KB.
16571 @item -matomic-updates
16572 @itemx -mno-atomic-updates
16573 @opindex matomic-updates
16574 @opindex mno-atomic-updates
16575 This option controls the version of libgcc that the compiler links to an
16576 executable and selects whether atomic updates to the software-managed
16577 cache of PPU-side variables are used. If you use atomic updates, changes
16578 to a PPU variable from SPU code using the @code{__ea} named address space
16579 qualifier will not interfere with changes to other PPU variables residing
16580 in the same cache line from PPU code. If you do not use atomic updates,
16581 such interference may occur; however, writing back cache lines will be
16582 more efficient. The default behavior is to use atomic updates.
16585 @itemx -mdual-nops=@var{n}
16586 @opindex mdual-nops
16587 By default, GCC will insert nops to increase dual issue when it expects
16588 it to increase performance. @var{n} can be a value from 0 to 10. A
16589 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
16590 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
16592 @item -mhint-max-nops=@var{n}
16593 @opindex mhint-max-nops
16594 Maximum number of nops to insert for a branch hint. A branch hint must
16595 be at least 8 instructions away from the branch it is effecting. GCC
16596 will insert up to @var{n} nops to enforce this, otherwise it will not
16597 generate the branch hint.
16599 @item -mhint-max-distance=@var{n}
16600 @opindex mhint-max-distance
16601 The encoding of the branch hint instruction limits the hint to be within
16602 256 instructions of the branch it is effecting. By default, GCC makes
16603 sure it is within 125.
16606 @opindex msafe-hints
16607 Work around a hardware bug which causes the SPU to stall indefinitely.
16608 By default, GCC will insert the @code{hbrp} instruction to make sure
16609 this stall won't happen.
16613 @node System V Options
16614 @subsection Options for System V
16616 These additional options are available on System V Release 4 for
16617 compatibility with other compilers on those systems:
16622 Create a shared object.
16623 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
16627 Identify the versions of each tool used by the compiler, in a
16628 @code{.ident} assembler directive in the output.
16632 Refrain from adding @code{.ident} directives to the output file (this is
16635 @item -YP,@var{dirs}
16637 Search the directories @var{dirs}, and no others, for libraries
16638 specified with @option{-l}.
16640 @item -Ym,@var{dir}
16642 Look in the directory @var{dir} to find the M4 preprocessor.
16643 The assembler uses this option.
16644 @c This is supposed to go with a -Yd for predefined M4 macro files, but
16645 @c the generic assembler that comes with Solaris takes just -Ym.
16649 @subsection V850 Options
16650 @cindex V850 Options
16652 These @samp{-m} options are defined for V850 implementations:
16656 @itemx -mno-long-calls
16657 @opindex mlong-calls
16658 @opindex mno-long-calls
16659 Treat all calls as being far away (near). If calls are assumed to be
16660 far away, the compiler will always load the functions address up into a
16661 register, and call indirect through the pointer.
16667 Do not optimize (do optimize) basic blocks that use the same index
16668 pointer 4 or more times to copy pointer into the @code{ep} register, and
16669 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
16670 option is on by default if you optimize.
16672 @item -mno-prolog-function
16673 @itemx -mprolog-function
16674 @opindex mno-prolog-function
16675 @opindex mprolog-function
16676 Do not use (do use) external functions to save and restore registers
16677 at the prologue and epilogue of a function. The external functions
16678 are slower, but use less code space if more than one function saves
16679 the same number of registers. The @option{-mprolog-function} option
16680 is on by default if you optimize.
16684 Try to make the code as small as possible. At present, this just turns
16685 on the @option{-mep} and @option{-mprolog-function} options.
16687 @item -mtda=@var{n}
16689 Put static or global variables whose size is @var{n} bytes or less into
16690 the tiny data area that register @code{ep} points to. The tiny data
16691 area can hold up to 256 bytes in total (128 bytes for byte references).
16693 @item -msda=@var{n}
16695 Put static or global variables whose size is @var{n} bytes or less into
16696 the small data area that register @code{gp} points to. The small data
16697 area can hold up to 64 kilobytes.
16699 @item -mzda=@var{n}
16701 Put static or global variables whose size is @var{n} bytes or less into
16702 the first 32 kilobytes of memory.
16706 Specify that the target processor is the V850.
16709 @opindex mbig-switch
16710 Generate code suitable for big switch tables. Use this option only if
16711 the assembler/linker complain about out of range branches within a switch
16716 This option will cause r2 and r5 to be used in the code generated by
16717 the compiler. This setting is the default.
16719 @item -mno-app-regs
16720 @opindex mno-app-regs
16721 This option will cause r2 and r5 to be treated as fixed registers.
16725 Specify that the target processor is the V850E1. The preprocessor
16726 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
16727 this option is used.
16731 Specify that the target processor is the V850E@. The preprocessor
16732 constant @samp{__v850e__} will be defined if this option is used.
16734 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
16735 are defined then a default target processor will be chosen and the
16736 relevant @samp{__v850*__} preprocessor constant will be defined.
16738 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
16739 defined, regardless of which processor variant is the target.
16741 @item -mdisable-callt
16742 @opindex mdisable-callt
16743 This option will suppress generation of the CALLT instruction for the
16744 v850e and v850e1 flavors of the v850 architecture. The default is
16745 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
16750 @subsection VAX Options
16751 @cindex VAX options
16753 These @samp{-m} options are defined for the VAX:
16758 Do not output certain jump instructions (@code{aobleq} and so on)
16759 that the Unix assembler for the VAX cannot handle across long
16764 Do output those jump instructions, on the assumption that you
16765 will assemble with the GNU assembler.
16769 Output code for g-format floating point numbers instead of d-format.
16772 @node VxWorks Options
16773 @subsection VxWorks Options
16774 @cindex VxWorks Options
16776 The options in this section are defined for all VxWorks targets.
16777 Options specific to the target hardware are listed with the other
16778 options for that target.
16783 GCC can generate code for both VxWorks kernels and real time processes
16784 (RTPs). This option switches from the former to the latter. It also
16785 defines the preprocessor macro @code{__RTP__}.
16788 @opindex non-static
16789 Link an RTP executable against shared libraries rather than static
16790 libraries. The options @option{-static} and @option{-shared} can
16791 also be used for RTPs (@pxref{Link Options}); @option{-static}
16798 These options are passed down to the linker. They are defined for
16799 compatibility with Diab.
16802 @opindex Xbind-lazy
16803 Enable lazy binding of function calls. This option is equivalent to
16804 @option{-Wl,-z,now} and is defined for compatibility with Diab.
16808 Disable lazy binding of function calls. This option is the default and
16809 is defined for compatibility with Diab.
16812 @node x86-64 Options
16813 @subsection x86-64 Options
16814 @cindex x86-64 options
16816 These are listed under @xref{i386 and x86-64 Options}.
16818 @node i386 and x86-64 Windows Options
16819 @subsection i386 and x86-64 Windows Options
16820 @cindex i386 and x86-64 Windows Options
16822 These additional options are available for Windows targets:
16827 This option is available for Cygwin and MinGW targets. It
16828 specifies that a console application is to be generated, by
16829 instructing the linker to set the PE header subsystem type
16830 required for console applications.
16831 This is the default behavior for Cygwin and MinGW targets.
16835 This option is available for Cygwin targets. It specifies that
16836 the Cygwin internal interface is to be used for predefined
16837 preprocessor macros, C runtime libraries and related linker
16838 paths and options. For Cygwin targets this is the default behavior.
16839 This option is deprecated and will be removed in a future release.
16842 @opindex mno-cygwin
16843 This option is available for Cygwin targets. It specifies that
16844 the MinGW internal interface is to be used instead of Cygwin's, by
16845 setting MinGW-related predefined macros and linker paths and default
16847 This option is deprecated and will be removed in a future release.
16851 This option is available for Cygwin and MinGW targets. It
16852 specifies that a DLL - a dynamic link library - is to be
16853 generated, enabling the selection of the required runtime
16854 startup object and entry point.
16856 @item -mnop-fun-dllimport
16857 @opindex mnop-fun-dllimport
16858 This option is available for Cygwin and MinGW targets. It
16859 specifies that the dllimport attribute should be ignored.
16863 This option is available for MinGW targets. It specifies
16864 that MinGW-specific thread support is to be used.
16868 This option is available for mingw-w64 targets. It specifies
16869 that the UNICODE macro is getting pre-defined and that the
16870 unicode capable runtime startup code is chosen.
16874 This option is available for Cygwin and MinGW targets. It
16875 specifies that the typical Windows pre-defined macros are to
16876 be set in the pre-processor, but does not influence the choice
16877 of runtime library/startup code.
16881 This option is available for Cygwin and MinGW targets. It
16882 specifies that a GUI application is to be generated by
16883 instructing the linker to set the PE header subsystem type
16886 @item -fno-set-stack-executable
16887 @opindex fno-set-stack-executable
16888 This option is available for MinGW targets. It specifies that
16889 the executable flag for stack used by nested functions isn't
16890 set. This is necessary for binaries running in kernel mode of
16891 Windows, as there the user32 API, which is used to set executable
16892 privileges, isn't available.
16894 @item -mpe-aligned-commons
16895 @opindex mpe-aligned-commons
16896 This option is available for Cygwin and MinGW targets. It
16897 specifies that the GNU extension to the PE file format that
16898 permits the correct alignment of COMMON variables should be
16899 used when generating code. It will be enabled by default if
16900 GCC detects that the target assembler found during configuration
16901 supports the feature.
16904 See also under @ref{i386 and x86-64 Options} for standard options.
16906 @node Xstormy16 Options
16907 @subsection Xstormy16 Options
16908 @cindex Xstormy16 Options
16910 These options are defined for Xstormy16:
16915 Choose startup files and linker script suitable for the simulator.
16918 @node Xtensa Options
16919 @subsection Xtensa Options
16920 @cindex Xtensa Options
16922 These options are supported for Xtensa targets:
16926 @itemx -mno-const16
16928 @opindex mno-const16
16929 Enable or disable use of @code{CONST16} instructions for loading
16930 constant values. The @code{CONST16} instruction is currently not a
16931 standard option from Tensilica. When enabled, @code{CONST16}
16932 instructions are always used in place of the standard @code{L32R}
16933 instructions. The use of @code{CONST16} is enabled by default only if
16934 the @code{L32R} instruction is not available.
16937 @itemx -mno-fused-madd
16938 @opindex mfused-madd
16939 @opindex mno-fused-madd
16940 Enable or disable use of fused multiply/add and multiply/subtract
16941 instructions in the floating-point option. This has no effect if the
16942 floating-point option is not also enabled. Disabling fused multiply/add
16943 and multiply/subtract instructions forces the compiler to use separate
16944 instructions for the multiply and add/subtract operations. This may be
16945 desirable in some cases where strict IEEE 754-compliant results are
16946 required: the fused multiply add/subtract instructions do not round the
16947 intermediate result, thereby producing results with @emph{more} bits of
16948 precision than specified by the IEEE standard. Disabling fused multiply
16949 add/subtract instructions also ensures that the program output is not
16950 sensitive to the compiler's ability to combine multiply and add/subtract
16953 @item -mserialize-volatile
16954 @itemx -mno-serialize-volatile
16955 @opindex mserialize-volatile
16956 @opindex mno-serialize-volatile
16957 When this option is enabled, GCC inserts @code{MEMW} instructions before
16958 @code{volatile} memory references to guarantee sequential consistency.
16959 The default is @option{-mserialize-volatile}. Use
16960 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
16962 @item -mtext-section-literals
16963 @itemx -mno-text-section-literals
16964 @opindex mtext-section-literals
16965 @opindex mno-text-section-literals
16966 Control the treatment of literal pools. The default is
16967 @option{-mno-text-section-literals}, which places literals in a separate
16968 section in the output file. This allows the literal pool to be placed
16969 in a data RAM/ROM, and it also allows the linker to combine literal
16970 pools from separate object files to remove redundant literals and
16971 improve code size. With @option{-mtext-section-literals}, the literals
16972 are interspersed in the text section in order to keep them as close as
16973 possible to their references. This may be necessary for large assembly
16976 @item -mtarget-align
16977 @itemx -mno-target-align
16978 @opindex mtarget-align
16979 @opindex mno-target-align
16980 When this option is enabled, GCC instructs the assembler to
16981 automatically align instructions to reduce branch penalties at the
16982 expense of some code density. The assembler attempts to widen density
16983 instructions to align branch targets and the instructions following call
16984 instructions. If there are not enough preceding safe density
16985 instructions to align a target, no widening will be performed. The
16986 default is @option{-mtarget-align}. These options do not affect the
16987 treatment of auto-aligned instructions like @code{LOOP}, which the
16988 assembler will always align, either by widening density instructions or
16989 by inserting no-op instructions.
16992 @itemx -mno-longcalls
16993 @opindex mlongcalls
16994 @opindex mno-longcalls
16995 When this option is enabled, GCC instructs the assembler to translate
16996 direct calls to indirect calls unless it can determine that the target
16997 of a direct call is in the range allowed by the call instruction. This
16998 translation typically occurs for calls to functions in other source
16999 files. Specifically, the assembler translates a direct @code{CALL}
17000 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
17001 The default is @option{-mno-longcalls}. This option should be used in
17002 programs where the call target can potentially be out of range. This
17003 option is implemented in the assembler, not the compiler, so the
17004 assembly code generated by GCC will still show direct call
17005 instructions---look at the disassembled object code to see the actual
17006 instructions. Note that the assembler will use an indirect call for
17007 every cross-file call, not just those that really will be out of range.
17010 @node zSeries Options
17011 @subsection zSeries Options
17012 @cindex zSeries options
17014 These are listed under @xref{S/390 and zSeries Options}.
17016 @node Code Gen Options
17017 @section Options for Code Generation Conventions
17018 @cindex code generation conventions
17019 @cindex options, code generation
17020 @cindex run-time options
17022 These machine-independent options control the interface conventions
17023 used in code generation.
17025 Most of them have both positive and negative forms; the negative form
17026 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
17027 one of the forms is listed---the one which is not the default. You
17028 can figure out the other form by either removing @samp{no-} or adding
17032 @item -fbounds-check
17033 @opindex fbounds-check
17034 For front-ends that support it, generate additional code to check that
17035 indices used to access arrays are within the declared range. This is
17036 currently only supported by the Java and Fortran front-ends, where
17037 this option defaults to true and false respectively.
17041 This option generates traps for signed overflow on addition, subtraction,
17042 multiplication operations.
17046 This option instructs the compiler to assume that signed arithmetic
17047 overflow of addition, subtraction and multiplication wraps around
17048 using twos-complement representation. This flag enables some optimizations
17049 and disables others. This option is enabled by default for the Java
17050 front-end, as required by the Java language specification.
17053 @opindex fexceptions
17054 Enable exception handling. Generates extra code needed to propagate
17055 exceptions. For some targets, this implies GCC will generate frame
17056 unwind information for all functions, which can produce significant data
17057 size overhead, although it does not affect execution. If you do not
17058 specify this option, GCC will enable it by default for languages like
17059 C++ which normally require exception handling, and disable it for
17060 languages like C that do not normally require it. However, you may need
17061 to enable this option when compiling C code that needs to interoperate
17062 properly with exception handlers written in C++. You may also wish to
17063 disable this option if you are compiling older C++ programs that don't
17064 use exception handling.
17066 @item -fnon-call-exceptions
17067 @opindex fnon-call-exceptions
17068 Generate code that allows trapping instructions to throw exceptions.
17069 Note that this requires platform-specific runtime support that does
17070 not exist everywhere. Moreover, it only allows @emph{trapping}
17071 instructions to throw exceptions, i.e.@: memory references or floating
17072 point instructions. It does not allow exceptions to be thrown from
17073 arbitrary signal handlers such as @code{SIGALRM}.
17075 @item -funwind-tables
17076 @opindex funwind-tables
17077 Similar to @option{-fexceptions}, except that it will just generate any needed
17078 static data, but will not affect the generated code in any other way.
17079 You will normally not enable this option; instead, a language processor
17080 that needs this handling would enable it on your behalf.
17082 @item -fasynchronous-unwind-tables
17083 @opindex fasynchronous-unwind-tables
17084 Generate unwind table in dwarf2 format, if supported by target machine. The
17085 table is exact at each instruction boundary, so it can be used for stack
17086 unwinding from asynchronous events (such as debugger or garbage collector).
17088 @item -fpcc-struct-return
17089 @opindex fpcc-struct-return
17090 Return ``short'' @code{struct} and @code{union} values in memory like
17091 longer ones, rather than in registers. This convention is less
17092 efficient, but it has the advantage of allowing intercallability between
17093 GCC-compiled files and files compiled with other compilers, particularly
17094 the Portable C Compiler (pcc).
17096 The precise convention for returning structures in memory depends
17097 on the target configuration macros.
17099 Short structures and unions are those whose size and alignment match
17100 that of some integer type.
17102 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
17103 switch is not binary compatible with code compiled with the
17104 @option{-freg-struct-return} switch.
17105 Use it to conform to a non-default application binary interface.
17107 @item -freg-struct-return
17108 @opindex freg-struct-return
17109 Return @code{struct} and @code{union} values in registers when possible.
17110 This is more efficient for small structures than
17111 @option{-fpcc-struct-return}.
17113 If you specify neither @option{-fpcc-struct-return} nor
17114 @option{-freg-struct-return}, GCC defaults to whichever convention is
17115 standard for the target. If there is no standard convention, GCC
17116 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
17117 the principal compiler. In those cases, we can choose the standard, and
17118 we chose the more efficient register return alternative.
17120 @strong{Warning:} code compiled with the @option{-freg-struct-return}
17121 switch is not binary compatible with code compiled with the
17122 @option{-fpcc-struct-return} switch.
17123 Use it to conform to a non-default application binary interface.
17125 @item -fshort-enums
17126 @opindex fshort-enums
17127 Allocate to an @code{enum} type only as many bytes as it needs for the
17128 declared range of possible values. Specifically, the @code{enum} type
17129 will be equivalent to the smallest integer type which has enough room.
17131 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
17132 code that is not binary compatible with code generated without that switch.
17133 Use it to conform to a non-default application binary interface.
17135 @item -fshort-double
17136 @opindex fshort-double
17137 Use the same size for @code{double} as for @code{float}.
17139 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
17140 code that is not binary compatible with code generated without that switch.
17141 Use it to conform to a non-default application binary interface.
17143 @item -fshort-wchar
17144 @opindex fshort-wchar
17145 Override the underlying type for @samp{wchar_t} to be @samp{short
17146 unsigned int} instead of the default for the target. This option is
17147 useful for building programs to run under WINE@.
17149 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
17150 code that is not binary compatible with code generated without that switch.
17151 Use it to conform to a non-default application binary interface.
17154 @opindex fno-common
17155 In C code, controls the placement of uninitialized global variables.
17156 Unix C compilers have traditionally permitted multiple definitions of
17157 such variables in different compilation units by placing the variables
17159 This is the behavior specified by @option{-fcommon}, and is the default
17160 for GCC on most targets.
17161 On the other hand, this behavior is not required by ISO C, and on some
17162 targets may carry a speed or code size penalty on variable references.
17163 The @option{-fno-common} option specifies that the compiler should place
17164 uninitialized global variables in the data section of the object file,
17165 rather than generating them as common blocks.
17166 This has the effect that if the same variable is declared
17167 (without @code{extern}) in two different compilations,
17168 you will get a multiple-definition error when you link them.
17169 In this case, you must compile with @option{-fcommon} instead.
17170 Compiling with @option{-fno-common} is useful on targets for which
17171 it provides better performance, or if you wish to verify that the
17172 program will work on other systems which always treat uninitialized
17173 variable declarations this way.
17177 Ignore the @samp{#ident} directive.
17179 @item -finhibit-size-directive
17180 @opindex finhibit-size-directive
17181 Don't output a @code{.size} assembler directive, or anything else that
17182 would cause trouble if the function is split in the middle, and the
17183 two halves are placed at locations far apart in memory. This option is
17184 used when compiling @file{crtstuff.c}; you should not need to use it
17187 @item -fverbose-asm
17188 @opindex fverbose-asm
17189 Put extra commentary information in the generated assembly code to
17190 make it more readable. This option is generally only of use to those
17191 who actually need to read the generated assembly code (perhaps while
17192 debugging the compiler itself).
17194 @option{-fno-verbose-asm}, the default, causes the
17195 extra information to be omitted and is useful when comparing two assembler
17198 @item -frecord-gcc-switches
17199 @opindex frecord-gcc-switches
17200 This switch causes the command line that was used to invoke the
17201 compiler to be recorded into the object file that is being created.
17202 This switch is only implemented on some targets and the exact format
17203 of the recording is target and binary file format dependent, but it
17204 usually takes the form of a section containing ASCII text. This
17205 switch is related to the @option{-fverbose-asm} switch, but that
17206 switch only records information in the assembler output file as
17207 comments, so it never reaches the object file.
17211 @cindex global offset table
17213 Generate position-independent code (PIC) suitable for use in a shared
17214 library, if supported for the target machine. Such code accesses all
17215 constant addresses through a global offset table (GOT)@. The dynamic
17216 loader resolves the GOT entries when the program starts (the dynamic
17217 loader is not part of GCC; it is part of the operating system). If
17218 the GOT size for the linked executable exceeds a machine-specific
17219 maximum size, you get an error message from the linker indicating that
17220 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
17221 instead. (These maximums are 8k on the SPARC and 32k
17222 on the m68k and RS/6000. The 386 has no such limit.)
17224 Position-independent code requires special support, and therefore works
17225 only on certain machines. For the 386, GCC supports PIC for System V
17226 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
17227 position-independent.
17229 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
17234 If supported for the target machine, emit position-independent code,
17235 suitable for dynamic linking and avoiding any limit on the size of the
17236 global offset table. This option makes a difference on the m68k,
17237 PowerPC and SPARC@.
17239 Position-independent code requires special support, and therefore works
17240 only on certain machines.
17242 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
17249 These options are similar to @option{-fpic} and @option{-fPIC}, but
17250 generated position independent code can be only linked into executables.
17251 Usually these options are used when @option{-pie} GCC option will be
17252 used during linking.
17254 @option{-fpie} and @option{-fPIE} both define the macros
17255 @code{__pie__} and @code{__PIE__}. The macros have the value 1
17256 for @option{-fpie} and 2 for @option{-fPIE}.
17258 @item -fno-jump-tables
17259 @opindex fno-jump-tables
17260 Do not use jump tables for switch statements even where it would be
17261 more efficient than other code generation strategies. This option is
17262 of use in conjunction with @option{-fpic} or @option{-fPIC} for
17263 building code which forms part of a dynamic linker and cannot
17264 reference the address of a jump table. On some targets, jump tables
17265 do not require a GOT and this option is not needed.
17267 @item -ffixed-@var{reg}
17269 Treat the register named @var{reg} as a fixed register; generated code
17270 should never refer to it (except perhaps as a stack pointer, frame
17271 pointer or in some other fixed role).
17273 @var{reg} must be the name of a register. The register names accepted
17274 are machine-specific and are defined in the @code{REGISTER_NAMES}
17275 macro in the machine description macro file.
17277 This flag does not have a negative form, because it specifies a
17280 @item -fcall-used-@var{reg}
17281 @opindex fcall-used
17282 Treat the register named @var{reg} as an allocable register that is
17283 clobbered by function calls. It may be allocated for temporaries or
17284 variables that do not live across a call. Functions compiled this way
17285 will not save and restore the register @var{reg}.
17287 It is an error to used this flag with the frame pointer or stack pointer.
17288 Use of this flag for other registers that have fixed pervasive roles in
17289 the machine's execution model will produce disastrous results.
17291 This flag does not have a negative form, because it specifies a
17294 @item -fcall-saved-@var{reg}
17295 @opindex fcall-saved
17296 Treat the register named @var{reg} as an allocable register saved by
17297 functions. It may be allocated even for temporaries or variables that
17298 live across a call. Functions compiled this way will save and restore
17299 the register @var{reg} if they use it.
17301 It is an error to used this flag with the frame pointer or stack pointer.
17302 Use of this flag for other registers that have fixed pervasive roles in
17303 the machine's execution model will produce disastrous results.
17305 A different sort of disaster will result from the use of this flag for
17306 a register in which function values may be returned.
17308 This flag does not have a negative form, because it specifies a
17311 @item -fpack-struct[=@var{n}]
17312 @opindex fpack-struct
17313 Without a value specified, pack all structure members together without
17314 holes. When a value is specified (which must be a small power of two), pack
17315 structure members according to this value, representing the maximum
17316 alignment (that is, objects with default alignment requirements larger than
17317 this will be output potentially unaligned at the next fitting location.
17319 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
17320 code that is not binary compatible with code generated without that switch.
17321 Additionally, it makes the code suboptimal.
17322 Use it to conform to a non-default application binary interface.
17324 @item -finstrument-functions
17325 @opindex finstrument-functions
17326 Generate instrumentation calls for entry and exit to functions. Just
17327 after function entry and just before function exit, the following
17328 profiling functions will be called with the address of the current
17329 function and its call site. (On some platforms,
17330 @code{__builtin_return_address} does not work beyond the current
17331 function, so the call site information may not be available to the
17332 profiling functions otherwise.)
17335 void __cyg_profile_func_enter (void *this_fn,
17337 void __cyg_profile_func_exit (void *this_fn,
17341 The first argument is the address of the start of the current function,
17342 which may be looked up exactly in the symbol table.
17344 This instrumentation is also done for functions expanded inline in other
17345 functions. The profiling calls will indicate where, conceptually, the
17346 inline function is entered and exited. This means that addressable
17347 versions of such functions must be available. If all your uses of a
17348 function are expanded inline, this may mean an additional expansion of
17349 code size. If you use @samp{extern inline} in your C code, an
17350 addressable version of such functions must be provided. (This is
17351 normally the case anyways, but if you get lucky and the optimizer always
17352 expands the functions inline, you might have gotten away without
17353 providing static copies.)
17355 A function may be given the attribute @code{no_instrument_function}, in
17356 which case this instrumentation will not be done. This can be used, for
17357 example, for the profiling functions listed above, high-priority
17358 interrupt routines, and any functions from which the profiling functions
17359 cannot safely be called (perhaps signal handlers, if the profiling
17360 routines generate output or allocate memory).
17362 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
17363 @opindex finstrument-functions-exclude-file-list
17365 Set the list of functions that are excluded from instrumentation (see
17366 the description of @code{-finstrument-functions}). If the file that
17367 contains a function definition matches with one of @var{file}, then
17368 that function is not instrumented. The match is done on substrings:
17369 if the @var{file} parameter is a substring of the file name, it is
17370 considered to be a match.
17373 @code{-finstrument-functions-exclude-file-list=/bits/stl,include/sys}
17374 will exclude any inline function defined in files whose pathnames
17375 contain @code{/bits/stl} or @code{include/sys}.
17377 If, for some reason, you want to include letter @code{','} in one of
17378 @var{sym}, write @code{'\,'}. For example,
17379 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
17380 (note the single quote surrounding the option).
17382 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
17383 @opindex finstrument-functions-exclude-function-list
17385 This is similar to @code{-finstrument-functions-exclude-file-list},
17386 but this option sets the list of function names to be excluded from
17387 instrumentation. The function name to be matched is its user-visible
17388 name, such as @code{vector<int> blah(const vector<int> &)}, not the
17389 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
17390 match is done on substrings: if the @var{sym} parameter is a substring
17391 of the function name, it is considered to be a match. For C99 and C++
17392 extended identifiers, the function name must be given in UTF-8, not
17393 using universal character names.
17395 @item -fstack-check
17396 @opindex fstack-check
17397 Generate code to verify that you do not go beyond the boundary of the
17398 stack. You should specify this flag if you are running in an
17399 environment with multiple threads, but only rarely need to specify it in
17400 a single-threaded environment since stack overflow is automatically
17401 detected on nearly all systems if there is only one stack.
17403 Note that this switch does not actually cause checking to be done; the
17404 operating system or the language runtime must do that. The switch causes
17405 generation of code to ensure that they see the stack being extended.
17407 You can additionally specify a string parameter: @code{no} means no
17408 checking, @code{generic} means force the use of old-style checking,
17409 @code{specific} means use the best checking method and is equivalent
17410 to bare @option{-fstack-check}.
17412 Old-style checking is a generic mechanism that requires no specific
17413 target support in the compiler but comes with the following drawbacks:
17417 Modified allocation strategy for large objects: they will always be
17418 allocated dynamically if their size exceeds a fixed threshold.
17421 Fixed limit on the size of the static frame of functions: when it is
17422 topped by a particular function, stack checking is not reliable and
17423 a warning is issued by the compiler.
17426 Inefficiency: because of both the modified allocation strategy and the
17427 generic implementation, the performances of the code are hampered.
17430 Note that old-style stack checking is also the fallback method for
17431 @code{specific} if no target support has been added in the compiler.
17433 @item -fstack-limit-register=@var{reg}
17434 @itemx -fstack-limit-symbol=@var{sym}
17435 @itemx -fno-stack-limit
17436 @opindex fstack-limit-register
17437 @opindex fstack-limit-symbol
17438 @opindex fno-stack-limit
17439 Generate code to ensure that the stack does not grow beyond a certain value,
17440 either the value of a register or the address of a symbol. If the stack
17441 would grow beyond the value, a signal is raised. For most targets,
17442 the signal is raised before the stack overruns the boundary, so
17443 it is possible to catch the signal without taking special precautions.
17445 For instance, if the stack starts at absolute address @samp{0x80000000}
17446 and grows downwards, you can use the flags
17447 @option{-fstack-limit-symbol=__stack_limit} and
17448 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
17449 of 128KB@. Note that this may only work with the GNU linker.
17451 @item -fleading-underscore
17452 @opindex fleading-underscore
17453 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
17454 change the way C symbols are represented in the object file. One use
17455 is to help link with legacy assembly code.
17457 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
17458 generate code that is not binary compatible with code generated without that
17459 switch. Use it to conform to a non-default application binary interface.
17460 Not all targets provide complete support for this switch.
17462 @item -ftls-model=@var{model}
17463 @opindex ftls-model
17464 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
17465 The @var{model} argument should be one of @code{global-dynamic},
17466 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
17468 The default without @option{-fpic} is @code{initial-exec}; with
17469 @option{-fpic} the default is @code{global-dynamic}.
17471 @item -fvisibility=@var{default|internal|hidden|protected}
17472 @opindex fvisibility
17473 Set the default ELF image symbol visibility to the specified option---all
17474 symbols will be marked with this unless overridden within the code.
17475 Using this feature can very substantially improve linking and
17476 load times of shared object libraries, produce more optimized
17477 code, provide near-perfect API export and prevent symbol clashes.
17478 It is @strong{strongly} recommended that you use this in any shared objects
17481 Despite the nomenclature, @code{default} always means public ie;
17482 available to be linked against from outside the shared object.
17483 @code{protected} and @code{internal} are pretty useless in real-world
17484 usage so the only other commonly used option will be @code{hidden}.
17485 The default if @option{-fvisibility} isn't specified is
17486 @code{default}, i.e., make every
17487 symbol public---this causes the same behavior as previous versions of
17490 A good explanation of the benefits offered by ensuring ELF
17491 symbols have the correct visibility is given by ``How To Write
17492 Shared Libraries'' by Ulrich Drepper (which can be found at
17493 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
17494 solution made possible by this option to marking things hidden when
17495 the default is public is to make the default hidden and mark things
17496 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
17497 and @code{__attribute__ ((visibility("default")))} instead of
17498 @code{__declspec(dllexport)} you get almost identical semantics with
17499 identical syntax. This is a great boon to those working with
17500 cross-platform projects.
17502 For those adding visibility support to existing code, you may find
17503 @samp{#pragma GCC visibility} of use. This works by you enclosing
17504 the declarations you wish to set visibility for with (for example)
17505 @samp{#pragma GCC visibility push(hidden)} and
17506 @samp{#pragma GCC visibility pop}.
17507 Bear in mind that symbol visibility should be viewed @strong{as
17508 part of the API interface contract} and thus all new code should
17509 always specify visibility when it is not the default ie; declarations
17510 only for use within the local DSO should @strong{always} be marked explicitly
17511 as hidden as so to avoid PLT indirection overheads---making this
17512 abundantly clear also aids readability and self-documentation of the code.
17513 Note that due to ISO C++ specification requirements, operator new and
17514 operator delete must always be of default visibility.
17516 Be aware that headers from outside your project, in particular system
17517 headers and headers from any other library you use, may not be
17518 expecting to be compiled with visibility other than the default. You
17519 may need to explicitly say @samp{#pragma GCC visibility push(default)}
17520 before including any such headers.
17522 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
17523 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
17524 no modifications. However, this means that calls to @samp{extern}
17525 functions with no explicit visibility will use the PLT, so it is more
17526 effective to use @samp{__attribute ((visibility))} and/or
17527 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
17528 declarations should be treated as hidden.
17530 Note that @samp{-fvisibility} does affect C++ vague linkage
17531 entities. This means that, for instance, an exception class that will
17532 be thrown between DSOs must be explicitly marked with default
17533 visibility so that the @samp{type_info} nodes will be unified between
17536 An overview of these techniques, their benefits and how to use them
17537 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
17543 @node Environment Variables
17544 @section Environment Variables Affecting GCC
17545 @cindex environment variables
17547 @c man begin ENVIRONMENT
17548 This section describes several environment variables that affect how GCC
17549 operates. Some of them work by specifying directories or prefixes to use
17550 when searching for various kinds of files. Some are used to specify other
17551 aspects of the compilation environment.
17553 Note that you can also specify places to search using options such as
17554 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
17555 take precedence over places specified using environment variables, which
17556 in turn take precedence over those specified by the configuration of GCC@.
17557 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
17558 GNU Compiler Collection (GCC) Internals}.
17563 @c @itemx LC_COLLATE
17565 @c @itemx LC_MONETARY
17566 @c @itemx LC_NUMERIC
17571 @c @findex LC_COLLATE
17572 @findex LC_MESSAGES
17573 @c @findex LC_MONETARY
17574 @c @findex LC_NUMERIC
17578 These environment variables control the way that GCC uses
17579 localization information that allow GCC to work with different
17580 national conventions. GCC inspects the locale categories
17581 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
17582 so. These locale categories can be set to any value supported by your
17583 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
17584 Kingdom encoded in UTF-8.
17586 The @env{LC_CTYPE} environment variable specifies character
17587 classification. GCC uses it to determine the character boundaries in
17588 a string; this is needed for some multibyte encodings that contain quote
17589 and escape characters that would otherwise be interpreted as a string
17592 The @env{LC_MESSAGES} environment variable specifies the language to
17593 use in diagnostic messages.
17595 If the @env{LC_ALL} environment variable is set, it overrides the value
17596 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
17597 and @env{LC_MESSAGES} default to the value of the @env{LANG}
17598 environment variable. If none of these variables are set, GCC
17599 defaults to traditional C English behavior.
17603 If @env{TMPDIR} is set, it specifies the directory to use for temporary
17604 files. GCC uses temporary files to hold the output of one stage of
17605 compilation which is to be used as input to the next stage: for example,
17606 the output of the preprocessor, which is the input to the compiler
17609 @item GCC_EXEC_PREFIX
17610 @findex GCC_EXEC_PREFIX
17611 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
17612 names of the subprograms executed by the compiler. No slash is added
17613 when this prefix is combined with the name of a subprogram, but you can
17614 specify a prefix that ends with a slash if you wish.
17616 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
17617 an appropriate prefix to use based on the pathname it was invoked with.
17619 If GCC cannot find the subprogram using the specified prefix, it
17620 tries looking in the usual places for the subprogram.
17622 The default value of @env{GCC_EXEC_PREFIX} is
17623 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
17624 the installed compiler. In many cases @var{prefix} is the value
17625 of @code{prefix} when you ran the @file{configure} script.
17627 Other prefixes specified with @option{-B} take precedence over this prefix.
17629 This prefix is also used for finding files such as @file{crt0.o} that are
17632 In addition, the prefix is used in an unusual way in finding the
17633 directories to search for header files. For each of the standard
17634 directories whose name normally begins with @samp{/usr/local/lib/gcc}
17635 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
17636 replacing that beginning with the specified prefix to produce an
17637 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
17638 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
17639 These alternate directories are searched first; the standard directories
17640 come next. If a standard directory begins with the configured
17641 @var{prefix} then the value of @var{prefix} is replaced by
17642 @env{GCC_EXEC_PREFIX} when looking for header files.
17644 @item COMPILER_PATH
17645 @findex COMPILER_PATH
17646 The value of @env{COMPILER_PATH} is a colon-separated list of
17647 directories, much like @env{PATH}. GCC tries the directories thus
17648 specified when searching for subprograms, if it can't find the
17649 subprograms using @env{GCC_EXEC_PREFIX}.
17652 @findex LIBRARY_PATH
17653 The value of @env{LIBRARY_PATH} is a colon-separated list of
17654 directories, much like @env{PATH}. When configured as a native compiler,
17655 GCC tries the directories thus specified when searching for special
17656 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
17657 using GCC also uses these directories when searching for ordinary
17658 libraries for the @option{-l} option (but directories specified with
17659 @option{-L} come first).
17663 @cindex locale definition
17664 This variable is used to pass locale information to the compiler. One way in
17665 which this information is used is to determine the character set to be used
17666 when character literals, string literals and comments are parsed in C and C++.
17667 When the compiler is configured to allow multibyte characters,
17668 the following values for @env{LANG} are recognized:
17672 Recognize JIS characters.
17674 Recognize SJIS characters.
17676 Recognize EUCJP characters.
17679 If @env{LANG} is not defined, or if it has some other value, then the
17680 compiler will use mblen and mbtowc as defined by the default locale to
17681 recognize and translate multibyte characters.
17685 Some additional environments variables affect the behavior of the
17688 @include cppenv.texi
17692 @node Precompiled Headers
17693 @section Using Precompiled Headers
17694 @cindex precompiled headers
17695 @cindex speed of compilation
17697 Often large projects have many header files that are included in every
17698 source file. The time the compiler takes to process these header files
17699 over and over again can account for nearly all of the time required to
17700 build the project. To make builds faster, GCC allows users to
17701 `precompile' a header file; then, if builds can use the precompiled
17702 header file they will be much faster.
17704 To create a precompiled header file, simply compile it as you would any
17705 other file, if necessary using the @option{-x} option to make the driver
17706 treat it as a C or C++ header file. You will probably want to use a
17707 tool like @command{make} to keep the precompiled header up-to-date when
17708 the headers it contains change.
17710 A precompiled header file will be searched for when @code{#include} is
17711 seen in the compilation. As it searches for the included file
17712 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
17713 compiler looks for a precompiled header in each directory just before it
17714 looks for the include file in that directory. The name searched for is
17715 the name specified in the @code{#include} with @samp{.gch} appended. If
17716 the precompiled header file can't be used, it is ignored.
17718 For instance, if you have @code{#include "all.h"}, and you have
17719 @file{all.h.gch} in the same directory as @file{all.h}, then the
17720 precompiled header file will be used if possible, and the original
17721 header will be used otherwise.
17723 Alternatively, you might decide to put the precompiled header file in a
17724 directory and use @option{-I} to ensure that directory is searched
17725 before (or instead of) the directory containing the original header.
17726 Then, if you want to check that the precompiled header file is always
17727 used, you can put a file of the same name as the original header in this
17728 directory containing an @code{#error} command.
17730 This also works with @option{-include}. So yet another way to use
17731 precompiled headers, good for projects not designed with precompiled
17732 header files in mind, is to simply take most of the header files used by
17733 a project, include them from another header file, precompile that header
17734 file, and @option{-include} the precompiled header. If the header files
17735 have guards against multiple inclusion, they will be skipped because
17736 they've already been included (in the precompiled header).
17738 If you need to precompile the same header file for different
17739 languages, targets, or compiler options, you can instead make a
17740 @emph{directory} named like @file{all.h.gch}, and put each precompiled
17741 header in the directory, perhaps using @option{-o}. It doesn't matter
17742 what you call the files in the directory, every precompiled header in
17743 the directory will be considered. The first precompiled header
17744 encountered in the directory that is valid for this compilation will
17745 be used; they're searched in no particular order.
17747 There are many other possibilities, limited only by your imagination,
17748 good sense, and the constraints of your build system.
17750 A precompiled header file can be used only when these conditions apply:
17754 Only one precompiled header can be used in a particular compilation.
17757 A precompiled header can't be used once the first C token is seen. You
17758 can have preprocessor directives before a precompiled header; you can
17759 even include a precompiled header from inside another header, so long as
17760 there are no C tokens before the @code{#include}.
17763 The precompiled header file must be produced for the same language as
17764 the current compilation. You can't use a C precompiled header for a C++
17768 The precompiled header file must have been produced by the same compiler
17769 binary as the current compilation is using.
17772 Any macros defined before the precompiled header is included must
17773 either be defined in the same way as when the precompiled header was
17774 generated, or must not affect the precompiled header, which usually
17775 means that they don't appear in the precompiled header at all.
17777 The @option{-D} option is one way to define a macro before a
17778 precompiled header is included; using a @code{#define} can also do it.
17779 There are also some options that define macros implicitly, like
17780 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
17783 @item If debugging information is output when using the precompiled
17784 header, using @option{-g} or similar, the same kind of debugging information
17785 must have been output when building the precompiled header. However,
17786 a precompiled header built using @option{-g} can be used in a compilation
17787 when no debugging information is being output.
17789 @item The same @option{-m} options must generally be used when building
17790 and using the precompiled header. @xref{Submodel Options},
17791 for any cases where this rule is relaxed.
17793 @item Each of the following options must be the same when building and using
17794 the precompiled header:
17796 @gccoptlist{-fexceptions}
17799 Some other command-line options starting with @option{-f},
17800 @option{-p}, or @option{-O} must be defined in the same way as when
17801 the precompiled header was generated. At present, it's not clear
17802 which options are safe to change and which are not; the safest choice
17803 is to use exactly the same options when generating and using the
17804 precompiled header. The following are known to be safe:
17806 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
17807 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
17808 -fsched-verbose=<number> -fschedule-insns -fvisibility= @gol
17813 For all of these except the last, the compiler will automatically
17814 ignore the precompiled header if the conditions aren't met. If you
17815 find an option combination that doesn't work and doesn't cause the
17816 precompiled header to be ignored, please consider filing a bug report,
17819 If you do use differing options when generating and using the
17820 precompiled header, the actual behavior will be a mixture of the
17821 behavior for the options. For instance, if you use @option{-g} to
17822 generate the precompiled header but not when using it, you may or may
17823 not get debugging information for routines in the precompiled header.