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
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
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 -Wctor-dtor-privacy @gol
198 -Wnon-virtual-dtor -Wreorder @gol
199 -Weffc++ -Wstrict-null-sentinel @gol
200 -Wno-non-template-friend -Wold-style-cast @gol
201 -Woverloaded-virtual -Wno-pmf-conversions @gol
204 @item Objective-C and Objective-C++ Language Options
205 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
206 Objective-C and Objective-C++ Dialects}.
207 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
208 -fgnu-runtime -fnext-runtime @gol
209 -fno-nil-receivers @gol
210 -fobjc-call-cxx-cdtors @gol
211 -fobjc-direct-dispatch @gol
212 -fobjc-exceptions @gol
214 -freplace-objc-classes @gol
217 -Wassign-intercept @gol
218 -Wno-protocol -Wselector @gol
219 -Wstrict-selector-match @gol
220 -Wundeclared-selector}
222 @item Language Independent Options
223 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
224 @gccoptlist{-fmessage-length=@var{n} @gol
225 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
226 -fdiagnostics-show-option}
228 @item Warning Options
229 @xref{Warning Options,,Options to Request or Suppress Warnings}.
230 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
231 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
232 -Wno-attributes -Wno-builtin-macro-redefined @gol
233 -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual @gol
234 -Wchar-subscripts -Wclobbered -Wcomment @gol
235 -Wconversion -Wcoverage-mismatch -Wno-deprecated @gol
236 -Wno-deprecated-declarations -Wdisabled-optimization @gol
237 -Wno-div-by-zero -Wempty-body -Wenum-compare -Wno-endif-labels @gol
238 -Werror -Werror=* @gol
239 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
240 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
241 -Wformat-security -Wformat-y2k @gol
242 -Wframe-larger-than=@var{len} -Wjump-misses-init -Wignored-qualifiers @gol
243 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
244 -Winit-self -Winline @gol
245 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
246 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
247 -Wlogical-op -Wlong-long @gol
248 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
249 -Wmissing-format-attribute -Wmissing-include-dirs @gol
250 -Wmissing-noreturn -Wno-mudflap @gol
251 -Wno-multichar -Wnonnull -Wno-overflow @gol
252 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
253 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
254 -Wpointer-arith -Wno-pointer-to-int-cast @gol
255 -Wredundant-decls @gol
256 -Wreturn-type -Wsequence-point -Wshadow @gol
257 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
258 -Wstrict-aliasing -Wstrict-aliasing=n @gol
259 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
260 -Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand @gol
261 -Wsystem-headers -Wtrigraphs -Wtype-limits -Wundef -Wuninitialized @gol
262 -Wunknown-pragmas -Wno-pragmas -Wunreachable-code @gol
263 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
264 -Wunused-label -Wunused-parameter -Wno-unused-result -Wunused-value -Wunused-variable @gol
265 -Wvariadic-macros -Wvla @gol
266 -Wvolatile-register-var -Wwrite-strings}
268 @item C and Objective-C-only Warning Options
269 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
270 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
271 -Wold-style-declaration -Wold-style-definition @gol
272 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
273 -Wdeclaration-after-statement -Wpointer-sign}
275 @item Debugging Options
276 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
277 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
278 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
279 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
280 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
281 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
282 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
283 -fdump-statistics @gol
285 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
286 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
287 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
289 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
290 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
291 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
292 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
293 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
294 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
295 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
296 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
297 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
298 -fdump-tree-nrv -fdump-tree-vect @gol
299 -fdump-tree-sink @gol
300 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
301 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
302 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
303 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
304 -ftree-vectorizer-verbose=@var{n} @gol
305 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
306 -fdump-final-insns=@var{file} @gol
307 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
308 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
309 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
310 -fenable-icf-debug @gol
311 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
312 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
313 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
314 -ftest-coverage -ftime-report -fvar-tracking @gol
315 -fvar-tracking-assigments -fvar-tracking-assignments-toggle @gol
316 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
317 -ggdb -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
318 -gvms -gxcoff -gxcoff+ @gol
319 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
320 -fdebug-prefix-map=@var{old}=@var{new} @gol
321 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
322 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
323 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
324 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
325 -print-prog-name=@var{program} -print-search-dirs -Q @gol
326 -print-sysroot -print-sysroot-headers-suffix @gol
327 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
329 @item Optimization Options
330 @xref{Optimize Options,,Options that Control Optimization}.
332 -falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
333 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
334 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
335 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
336 -fcheck-data-deps -fconserve-stack -fcprop-registers -fcrossjumping @gol
337 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules -fcx-limited-range @gol
338 -fdata-sections -fdce -fdce @gol
339 -fdelayed-branch -fdelete-null-pointer-checks -fdse -fdse @gol
340 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffast-math @gol
341 -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
342 -fforward-propagate -ffunction-sections @gol
343 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm @gol
344 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
345 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
346 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg -fipa-pta @gol
347 -fipa-pure-const -fipa-reference -fipa-struct-reorg @gol
348 -fipa-type-escape -fira-algorithm=@var{algorithm} @gol
349 -fira-region=@var{region} -fira-coalesce @gol
350 -fira-loop-pressure -fno-ira-share-save-slots @gol
351 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
352 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
353 -floop-block -floop-interchange -floop-strip-mine -fgraphite-identity @gol
354 -floop-parallelize-all -flto -flto-compression-level -flto-report -fltrans @gol
355 -fltrans-output-list -fmerge-all-constants -fmerge-constants -fmodulo-sched @gol
356 -fmodulo-sched-allow-regmoves -fmove-loop-invariants -fmudflap @gol
357 -fmudflapir -fmudflapth -fno-branch-count-reg -fno-default-inline @gol
358 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
359 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
360 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
361 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
362 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
363 -fpeel-loops -fpredictive-commoning -fprefetch-loop-arrays @gol
364 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
365 -fprofile-generate=@var{path} @gol
366 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
367 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
368 -freorder-blocks-and-partition -freorder-functions @gol
369 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
370 -frounding-math -fsched2-use-superblocks @gol
371 -fsched2-use-traces -fsched-pressure @gol
372 -fsched-spec-load -fsched-spec-load-dangerous @gol
373 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
374 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
375 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
376 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
377 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
378 -fselective-scheduling -fselective-scheduling2 @gol
379 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
380 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
381 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
382 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer @gol
383 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
384 -ftree-copyrename -ftree-dce @gol
385 -ftree-dominator-opts -ftree-dse -ftree-forwprop -ftree-fre -ftree-loop-im @gol
386 -ftree-phiprop -ftree-loop-distribution @gol
387 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
388 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-pta -ftree-reassoc @gol
389 -ftree-sink -ftree-sra -ftree-switch-conversion @gol
390 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
391 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
392 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
393 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
394 -fwhole-program -fwhopr -fwpa -use-linker-plugin @gol
395 --param @var{name}=@var{value}
396 -O -O0 -O1 -O2 -O3 -Os}
398 @item Preprocessor Options
399 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
400 @gccoptlist{-A@var{question}=@var{answer} @gol
401 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
402 -C -dD -dI -dM -dN @gol
403 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
404 -idirafter @var{dir} @gol
405 -include @var{file} -imacros @var{file} @gol
406 -iprefix @var{file} -iwithprefix @var{dir} @gol
407 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
408 -imultilib @var{dir} -isysroot @var{dir} @gol
409 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
410 -P -fworking-directory -remap @gol
411 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
412 -Xpreprocessor @var{option}}
414 @item Assembler Option
415 @xref{Assembler Options,,Passing Options to the Assembler}.
416 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
419 @xref{Link Options,,Options for Linking}.
420 @gccoptlist{@var{object-file-name} -l@var{library} @gol
421 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
422 -s -static -static-libgcc -static-libstdc++ -shared @gol
423 -shared-libgcc -symbolic @gol
424 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
427 @item Directory Options
428 @xref{Directory Options,,Options for Directory Search}.
429 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir}
430 -specs=@var{file} -I- --sysroot=@var{dir}}
433 @c I wrote this xref this way to avoid overfull hbox. -- rms
434 @xref{Target Options}.
435 @gccoptlist{-V @var{version} -b @var{machine}}
437 @item Machine Dependent Options
438 @xref{Submodel Options,,Hardware Models and Configurations}.
439 @c This list is ordered alphanumerically by subsection name.
440 @c Try and put the significant identifier (CPU or system) first,
441 @c so users have a clue at guessing where the ones they want will be.
444 @gccoptlist{-EB -EL @gol
445 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
446 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
449 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
450 -mabi=@var{name} @gol
451 -mapcs-stack-check -mno-apcs-stack-check @gol
452 -mapcs-float -mno-apcs-float @gol
453 -mapcs-reentrant -mno-apcs-reentrant @gol
454 -msched-prolog -mno-sched-prolog @gol
455 -mlittle-endian -mbig-endian -mwords-little-endian @gol
456 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
457 -mfp16-format=@var{name}
458 -mthumb-interwork -mno-thumb-interwork @gol
459 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
460 -mstructure-size-boundary=@var{n} @gol
461 -mabort-on-noreturn @gol
462 -mlong-calls -mno-long-calls @gol
463 -msingle-pic-base -mno-single-pic-base @gol
464 -mpic-register=@var{reg} @gol
465 -mnop-fun-dllimport @gol
466 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
467 -mpoke-function-name @gol
469 -mtpcs-frame -mtpcs-leaf-frame @gol
470 -mcaller-super-interworking -mcallee-super-interworking @gol
472 -mword-relocations @gol
473 -mfix-cortex-m3-ldrd}
476 @gccoptlist{-mmcu=@var{mcu} -mno-interrupts @gol
477 -mcall-prologues -mtiny-stack -mint8}
479 @emph{Blackfin Options}
480 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
481 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
482 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
483 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
484 -mno-id-shared-library -mshared-library-id=@var{n} @gol
485 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
486 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
487 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
491 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
492 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
493 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
494 -mstack-align -mdata-align -mconst-align @gol
495 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
496 -melf -maout -melinux -mlinux -sim -sim2 @gol
497 -mmul-bug-workaround -mno-mul-bug-workaround}
500 @gccoptlist{-mmac -mpush-args}
502 @emph{Darwin Options}
503 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
504 -arch_only -bind_at_load -bundle -bundle_loader @gol
505 -client_name -compatibility_version -current_version @gol
507 -dependency-file -dylib_file -dylinker_install_name @gol
508 -dynamic -dynamiclib -exported_symbols_list @gol
509 -filelist -flat_namespace -force_cpusubtype_ALL @gol
510 -force_flat_namespace -headerpad_max_install_names @gol
512 -image_base -init -install_name -keep_private_externs @gol
513 -multi_module -multiply_defined -multiply_defined_unused @gol
514 -noall_load -no_dead_strip_inits_and_terms @gol
515 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
516 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
517 -private_bundle -read_only_relocs -sectalign @gol
518 -sectobjectsymbols -whyload -seg1addr @gol
519 -sectcreate -sectobjectsymbols -sectorder @gol
520 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
521 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
522 -segprot -segs_read_only_addr -segs_read_write_addr @gol
523 -single_module -static -sub_library -sub_umbrella @gol
524 -twolevel_namespace -umbrella -undefined @gol
525 -unexported_symbols_list -weak_reference_mismatches @gol
526 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
527 -mkernel -mone-byte-bool}
529 @emph{DEC Alpha Options}
530 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
531 -mieee -mieee-with-inexact -mieee-conformant @gol
532 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
533 -mtrap-precision=@var{mode} -mbuild-constants @gol
534 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
535 -mbwx -mmax -mfix -mcix @gol
536 -mfloat-vax -mfloat-ieee @gol
537 -mexplicit-relocs -msmall-data -mlarge-data @gol
538 -msmall-text -mlarge-text @gol
539 -mmemory-latency=@var{time}}
541 @emph{DEC Alpha/VMS Options}
542 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
545 @gccoptlist{-msmall-model -mno-lsim}
548 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
549 -mhard-float -msoft-float @gol
550 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
551 -mdouble -mno-double @gol
552 -mmedia -mno-media -mmuladd -mno-muladd @gol
553 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
554 -mlinked-fp -mlong-calls -malign-labels @gol
555 -mlibrary-pic -macc-4 -macc-8 @gol
556 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
557 -moptimize-membar -mno-optimize-membar @gol
558 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
559 -mvliw-branch -mno-vliw-branch @gol
560 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
561 -mno-nested-cond-exec -mtomcat-stats @gol
565 @emph{GNU/Linux Options}
566 @gccoptlist{-muclibc}
568 @emph{H8/300 Options}
569 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
572 @gccoptlist{-march=@var{architecture-type} @gol
573 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
574 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
575 -mfixed-range=@var{register-range} @gol
576 -mjump-in-delay -mlinker-opt -mlong-calls @gol
577 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
578 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
579 -mno-jump-in-delay -mno-long-load-store @gol
580 -mno-portable-runtime -mno-soft-float @gol
581 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
582 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
583 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
584 -munix=@var{unix-std} -nolibdld -static -threads}
586 @emph{i386 and x86-64 Options}
587 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
588 -mfpmath=@var{unit} @gol
589 -masm=@var{dialect} -mno-fancy-math-387 @gol
590 -mno-fp-ret-in-387 -msoft-float @gol
591 -mno-wide-multiply -mrtd -malign-double @gol
592 -mpreferred-stack-boundary=@var{num}
593 -mincoming-stack-boundary=@var{num}
594 -mcld -mcx16 -msahf -mmovbe -mcrc32 -mrecip @gol
595 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
597 -msse4a -m3dnow -mpopcnt -mabm -mfma4 @gol
598 -mthreads -mno-align-stringops -minline-all-stringops @gol
599 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
600 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
601 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
602 -mveclibabi=@var{type} -mpc32 -mpc64 -mpc80 -mstackrealign @gol
603 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
604 -mcmodel=@var{code-model} -mabi=@var{name} @gol
605 -m32 -m64 -mlarge-data-threshold=@var{num} @gol
609 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
610 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
611 -mconstant-gp -mauto-pic -mfused-madd @gol
612 -minline-float-divide-min-latency @gol
613 -minline-float-divide-max-throughput @gol
614 -mno-inline-float-divide @gol
615 -minline-int-divide-min-latency @gol
616 -minline-int-divide-max-throughput @gol
617 -mno-inline-int-divide @gol
618 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
619 -mno-inline-sqrt @gol
620 -mdwarf2-asm -mearly-stop-bits @gol
621 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
622 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
623 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
624 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
625 -msched-spec-ldc -msched-spec-control-ldc @gol
626 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
627 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
628 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
629 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
631 @emph{IA-64/VMS Options}
632 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
634 @emph{M32R/D Options}
635 @gccoptlist{-m32r2 -m32rx -m32r @gol
637 -malign-loops -mno-align-loops @gol
638 -missue-rate=@var{number} @gol
639 -mbranch-cost=@var{number} @gol
640 -mmodel=@var{code-size-model-type} @gol
641 -msdata=@var{sdata-type} @gol
642 -mno-flush-func -mflush-func=@var{name} @gol
643 -mno-flush-trap -mflush-trap=@var{number} @gol
647 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
649 @emph{M680x0 Options}
650 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
651 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
652 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
653 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
654 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
655 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
656 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
657 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
660 @emph{M68hc1x Options}
661 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
662 -mauto-incdec -minmax -mlong-calls -mshort @gol
663 -msoft-reg-count=@var{count}}
666 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
667 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
668 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
669 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
670 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
673 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
674 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
675 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
676 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
680 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
681 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
682 -mips64 -mips64r2 @gol
683 -mips16 -mno-mips16 -mflip-mips16 @gol
684 -minterlink-mips16 -mno-interlink-mips16 @gol
685 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
686 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
687 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
688 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
689 -mfpu=@var{fpu-type} @gol
690 -msmartmips -mno-smartmips @gol
691 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
692 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
693 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
694 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
695 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
696 -membedded-data -mno-embedded-data @gol
697 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
698 -mcode-readable=@var{setting} @gol
699 -msplit-addresses -mno-split-addresses @gol
700 -mexplicit-relocs -mno-explicit-relocs @gol
701 -mcheck-zero-division -mno-check-zero-division @gol
702 -mdivide-traps -mdivide-breaks @gol
703 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
704 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
705 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
706 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
707 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
708 -mflush-func=@var{func} -mno-flush-func @gol
709 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
710 -mfp-exceptions -mno-fp-exceptions @gol
711 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
712 -mrelax-pic-calls -mno-relax-pic-calls}
715 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
716 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
717 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
718 -mno-base-addresses -msingle-exit -mno-single-exit}
720 @emph{MN10300 Options}
721 @gccoptlist{-mmult-bug -mno-mult-bug @gol
722 -mam33 -mno-am33 @gol
723 -mam33-2 -mno-am33-2 @gol
724 -mreturn-pointer-on-d0 @gol
727 @emph{PDP-11 Options}
728 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
729 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
730 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
731 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
732 -mbranch-expensive -mbranch-cheap @gol
733 -msplit -mno-split -munix-asm -mdec-asm}
735 @emph{picoChip Options}
736 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N}
737 -msymbol-as-address -mno-inefficient-warnings}
739 @emph{PowerPC Options}
740 See RS/6000 and PowerPC Options.
742 @emph{RS/6000 and PowerPC Options}
743 @gccoptlist{-mcpu=@var{cpu-type} @gol
744 -mtune=@var{cpu-type} @gol
745 -mpower -mno-power -mpower2 -mno-power2 @gol
746 -mpowerpc -mpowerpc64 -mno-powerpc @gol
747 -maltivec -mno-altivec @gol
748 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
749 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
750 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
751 -mfprnd -mno-fprnd @gol
752 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
753 -mnew-mnemonics -mold-mnemonics @gol
754 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
755 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
756 -malign-power -malign-natural @gol
757 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
758 -msingle-float -mdouble-float -msimple-fpu @gol
759 -mstring -mno-string -mupdate -mno-update @gol
760 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
761 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
762 -mstrict-align -mno-strict-align -mrelocatable @gol
763 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
764 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
765 -mdynamic-no-pic -maltivec -mswdiv @gol
766 -mprioritize-restricted-insns=@var{priority} @gol
767 -msched-costly-dep=@var{dependence_type} @gol
768 -minsert-sched-nops=@var{scheme} @gol
769 -mcall-sysv -mcall-netbsd @gol
770 -maix-struct-return -msvr4-struct-return @gol
771 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
772 -misel -mno-isel @gol
773 -misel=yes -misel=no @gol
775 -mspe=yes -mspe=no @gol
777 -mgen-cell-microcode -mwarn-cell-microcode @gol
778 -mvrsave -mno-vrsave @gol
779 -mmulhw -mno-mulhw @gol
780 -mdlmzb -mno-dlmzb @gol
781 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
782 -mprototype -mno-prototype @gol
783 -msim -mmvme -mads -myellowknife -memb -msdata @gol
784 -msdata=@var{opt} -mvxworks -G @var{num} -pthread}
787 @gccoptlist{-m64bit-doubles -m32bit-doubles -mieee -mno-ieee@gol
788 -mbig-endian-data -mlittle-endian-data @gol
791 -mas100-syntax -mno-as100-syntax@gol
793 -mmax-constant-size=@gol
796 @emph{S/390 and zSeries Options}
797 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
798 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
799 -mlong-double-64 -mlong-double-128 @gol
800 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
801 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
802 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
803 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
804 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
807 @gccoptlist{-meb -mel @gol
811 -mscore5 -mscore5u -mscore7 -mscore7d}
814 @gccoptlist{-m1 -m2 -m2e @gol
815 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
817 -m4-nofpu -m4-single-only -m4-single -m4 @gol
818 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
819 -m5-64media -m5-64media-nofpu @gol
820 -m5-32media -m5-32media-nofpu @gol
821 -m5-compact -m5-compact-nofpu @gol
822 -mb -ml -mdalign -mrelax @gol
823 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
824 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
825 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
826 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
827 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
831 @gccoptlist{-mcpu=@var{cpu-type} @gol
832 -mtune=@var{cpu-type} @gol
833 -mcmodel=@var{code-model} @gol
834 -m32 -m64 -mapp-regs -mno-app-regs @gol
835 -mfaster-structs -mno-faster-structs @gol
836 -mfpu -mno-fpu -mhard-float -msoft-float @gol
837 -mhard-quad-float -msoft-quad-float @gol
838 -mimpure-text -mno-impure-text -mlittle-endian @gol
839 -mstack-bias -mno-stack-bias @gol
840 -munaligned-doubles -mno-unaligned-doubles @gol
841 -mv8plus -mno-v8plus -mvis -mno-vis
842 -threads -pthreads -pthread}
845 @gccoptlist{-mwarn-reloc -merror-reloc @gol
846 -msafe-dma -munsafe-dma @gol
848 -msmall-mem -mlarge-mem -mstdmain @gol
849 -mfixed-range=@var{register-range} @gol
851 -maddress-space-conversion -mno-address-space-conversion @gol
852 -mcache-size=@var{cache-size} @gol
853 -matomic-updates -mno-atomic-updates}
855 @emph{System V Options}
856 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
859 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
860 -mprolog-function -mno-prolog-function -mspace @gol
861 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
862 -mapp-regs -mno-app-regs @gol
863 -mdisable-callt -mno-disable-callt @gol
869 @gccoptlist{-mg -mgnu -munix}
871 @emph{VxWorks Options}
872 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
873 -Xbind-lazy -Xbind-now}
875 @emph{x86-64 Options}
876 See i386 and x86-64 Options.
878 @emph{i386 and x86-64 Windows Options}
879 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll
880 -mnop-fun-dllimport -mthread -municode -mwin32 -mwindows}
882 @emph{Xstormy16 Options}
885 @emph{Xtensa Options}
886 @gccoptlist{-mconst16 -mno-const16 @gol
887 -mfused-madd -mno-fused-madd @gol
888 -mserialize-volatile -mno-serialize-volatile @gol
889 -mtext-section-literals -mno-text-section-literals @gol
890 -mtarget-align -mno-target-align @gol
891 -mlongcalls -mno-longcalls}
893 @emph{zSeries Options}
894 See S/390 and zSeries Options.
896 @item Code Generation Options
897 @xref{Code Gen Options,,Options for Code Generation Conventions}.
898 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
899 -ffixed-@var{reg} -fexceptions @gol
900 -fnon-call-exceptions -funwind-tables @gol
901 -fasynchronous-unwind-tables @gol
902 -finhibit-size-directive -finstrument-functions @gol
903 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
904 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
905 -fno-common -fno-ident @gol
906 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
907 -fno-jump-tables @gol
908 -frecord-gcc-switches @gol
909 -freg-struct-return -fshort-enums @gol
910 -fshort-double -fshort-wchar @gol
911 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
912 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
913 -fno-stack-limit -fargument-alias -fargument-noalias @gol
914 -fargument-noalias-global -fargument-noalias-anything @gol
915 -fleading-underscore -ftls-model=@var{model} @gol
916 -ftrapv -fwrapv -fbounds-check @gol
921 * Overall Options:: Controlling the kind of output:
922 an executable, object files, assembler files,
923 or preprocessed source.
924 * C Dialect Options:: Controlling the variant of C language compiled.
925 * C++ Dialect Options:: Variations on C++.
926 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
928 * Language Independent Options:: Controlling how diagnostics should be
930 * Warning Options:: How picky should the compiler be?
931 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
932 * Optimize Options:: How much optimization?
933 * Preprocessor Options:: Controlling header files and macro definitions.
934 Also, getting dependency information for Make.
935 * Assembler Options:: Passing options to the assembler.
936 * Link Options:: Specifying libraries and so on.
937 * Directory Options:: Where to find header files and libraries.
938 Where to find the compiler executable files.
939 * Spec Files:: How to pass switches to sub-processes.
940 * Target Options:: Running a cross-compiler, or an old version of GCC.
943 @node Overall Options
944 @section Options Controlling the Kind of Output
946 Compilation can involve up to four stages: preprocessing, compilation
947 proper, assembly and linking, always in that order. GCC is capable of
948 preprocessing and compiling several files either into several
949 assembler input files, or into one assembler input file; then each
950 assembler input file produces an object file, and linking combines all
951 the object files (those newly compiled, and those specified as input)
952 into an executable file.
954 @cindex file name suffix
955 For any given input file, the file name suffix determines what kind of
960 C source code which must be preprocessed.
963 C source code which should not be preprocessed.
966 C++ source code which should not be preprocessed.
969 Objective-C source code. Note that you must link with the @file{libobjc}
970 library to make an Objective-C program work.
973 Objective-C source code which should not be preprocessed.
977 Objective-C++ source code. Note that you must link with the @file{libobjc}
978 library to make an Objective-C++ program work. Note that @samp{.M} refers
979 to a literal capital M@.
982 Objective-C++ source code which should not be preprocessed.
985 C, C++, Objective-C or Objective-C++ header file to be turned into a
990 @itemx @var{file}.cxx
991 @itemx @var{file}.cpp
992 @itemx @var{file}.CPP
993 @itemx @var{file}.c++
995 C++ source code which must be preprocessed. Note that in @samp{.cxx},
996 the last two letters must both be literally @samp{x}. Likewise,
997 @samp{.C} refers to a literal capital C@.
1001 Objective-C++ source code which must be preprocessed.
1003 @item @var{file}.mii
1004 Objective-C++ source code which should not be preprocessed.
1008 @itemx @var{file}.hp
1009 @itemx @var{file}.hxx
1010 @itemx @var{file}.hpp
1011 @itemx @var{file}.HPP
1012 @itemx @var{file}.h++
1013 @itemx @var{file}.tcc
1014 C++ header file to be turned into a precompiled header.
1017 @itemx @var{file}.for
1018 @itemx @var{file}.ftn
1019 Fixed form Fortran source code which should not be preprocessed.
1022 @itemx @var{file}.FOR
1023 @itemx @var{file}.fpp
1024 @itemx @var{file}.FPP
1025 @itemx @var{file}.FTN
1026 Fixed form Fortran source code which must be preprocessed (with the traditional
1029 @item @var{file}.f90
1030 @itemx @var{file}.f95
1031 @itemx @var{file}.f03
1032 @itemx @var{file}.f08
1033 Free form Fortran source code which should not be preprocessed.
1035 @item @var{file}.F90
1036 @itemx @var{file}.F95
1037 @itemx @var{file}.F03
1038 @itemx @var{file}.F08
1039 Free form Fortran source code which must be preprocessed (with the
1040 traditional preprocessor).
1042 @c FIXME: Descriptions of Java file types.
1048 @item @var{file}.ads
1049 Ada source code file which contains a library unit declaration (a
1050 declaration of a package, subprogram, or generic, or a generic
1051 instantiation), or a library unit renaming declaration (a package,
1052 generic, or subprogram renaming declaration). Such files are also
1055 @item @var{file}.adb
1056 Ada source code file containing a library unit body (a subprogram or
1057 package body). Such files are also called @dfn{bodies}.
1059 @c GCC also knows about some suffixes for languages not yet included:
1070 @itemx @var{file}.sx
1071 Assembler code which must be preprocessed.
1074 An object file to be fed straight into linking.
1075 Any file name with no recognized suffix is treated this way.
1079 You can specify the input language explicitly with the @option{-x} option:
1082 @item -x @var{language}
1083 Specify explicitly the @var{language} for the following input files
1084 (rather than letting the compiler choose a default based on the file
1085 name suffix). This option applies to all following input files until
1086 the next @option{-x} option. Possible values for @var{language} are:
1088 c c-header c-cpp-output
1089 c++ c++-header c++-cpp-output
1090 objective-c objective-c-header objective-c-cpp-output
1091 objective-c++ objective-c++-header objective-c++-cpp-output
1092 assembler assembler-with-cpp
1094 f77 f77-cpp-input f95 f95-cpp-input
1099 Turn off any specification of a language, so that subsequent files are
1100 handled according to their file name suffixes (as they are if @option{-x}
1101 has not been used at all).
1103 @item -pass-exit-codes
1104 @opindex pass-exit-codes
1105 Normally the @command{gcc} program will exit with the code of 1 if any
1106 phase of the compiler returns a non-success return code. If you specify
1107 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1108 numerically highest error produced by any phase that returned an error
1109 indication. The C, C++, and Fortran frontends return 4, if an internal
1110 compiler error is encountered.
1113 If you only want some of the stages of compilation, you can use
1114 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1115 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1116 @command{gcc} is to stop. Note that some combinations (for example,
1117 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1122 Compile or assemble the source files, but do not link. The linking
1123 stage simply is not done. The ultimate output is in the form of an
1124 object file for each source file.
1126 By default, the object file name for a source file is made by replacing
1127 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1129 Unrecognized input files, not requiring compilation or assembly, are
1134 Stop after the stage of compilation proper; do not assemble. The output
1135 is in the form of an assembler code file for each non-assembler input
1138 By default, the assembler file name for a source file is made by
1139 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1141 Input files that don't require compilation are ignored.
1145 Stop after the preprocessing stage; do not run the compiler proper. The
1146 output is in the form of preprocessed source code, which is sent to the
1149 Input files which don't require preprocessing are ignored.
1151 @cindex output file option
1154 Place output in file @var{file}. This applies regardless to whatever
1155 sort of output is being produced, whether it be an executable file,
1156 an object file, an assembler file or preprocessed C code.
1158 If @option{-o} is not specified, the default is to put an executable
1159 file in @file{a.out}, the object file for
1160 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1161 assembler file in @file{@var{source}.s}, a precompiled header file in
1162 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1167 Print (on standard error output) the commands executed to run the stages
1168 of compilation. Also print the version number of the compiler driver
1169 program and of the preprocessor and the compiler proper.
1173 Like @option{-v} except the commands are not executed and all command
1174 arguments are quoted. This is useful for shell scripts to capture the
1175 driver-generated command lines.
1179 Use pipes rather than temporary files for communication between the
1180 various stages of compilation. This fails to work on some systems where
1181 the assembler is unable to read from a pipe; but the GNU assembler has
1186 If you are compiling multiple source files, this option tells the driver
1187 to pass all the source files to the compiler at once (for those
1188 languages for which the compiler can handle this). This will allow
1189 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1190 language for which this is supported is C@. If you pass source files for
1191 multiple languages to the driver, using this option, the driver will invoke
1192 the compiler(s) that support IMA once each, passing each compiler all the
1193 source files appropriate for it. For those languages that do not support
1194 IMA this option will be ignored, and the compiler will be invoked once for
1195 each source file in that language. If you use this option in conjunction
1196 with @option{-save-temps}, the compiler will generate multiple
1198 (one for each source file), but only one (combined) @file{.o} or
1203 Print (on the standard output) a description of the command line options
1204 understood by @command{gcc}. If the @option{-v} option is also specified
1205 then @option{--help} will also be passed on to the various processes
1206 invoked by @command{gcc}, so that they can display the command line options
1207 they accept. If the @option{-Wextra} option has also been specified
1208 (prior to the @option{--help} option), then command line options which
1209 have no documentation associated with them will also be displayed.
1212 @opindex target-help
1213 Print (on the standard output) a description of target-specific command
1214 line options for each tool. For some targets extra target-specific
1215 information may also be printed.
1217 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1218 Print (on the standard output) a description of the command line
1219 options understood by the compiler that fit into all specified classes
1220 and qualifiers. These are the supported classes:
1223 @item @samp{optimizers}
1224 This will display all of the optimization options supported by the
1227 @item @samp{warnings}
1228 This will display all of the options controlling warning messages
1229 produced by the compiler.
1232 This will display target-specific options. Unlike the
1233 @option{--target-help} option however, target-specific options of the
1234 linker and assembler will not be displayed. This is because those
1235 tools do not currently support the extended @option{--help=} syntax.
1238 This will display the values recognized by the @option{--param}
1241 @item @var{language}
1242 This will display the options supported for @var{language}, where
1243 @var{language} is the name of one of the languages supported in this
1247 This will display the options that are common to all languages.
1250 These are the supported qualifiers:
1253 @item @samp{undocumented}
1254 Display only those options which are undocumented.
1257 Display options which take an argument that appears after an equal
1258 sign in the same continuous piece of text, such as:
1259 @samp{--help=target}.
1261 @item @samp{separate}
1262 Display options which take an argument that appears as a separate word
1263 following the original option, such as: @samp{-o output-file}.
1266 Thus for example to display all the undocumented target-specific
1267 switches supported by the compiler the following can be used:
1270 --help=target,undocumented
1273 The sense of a qualifier can be inverted by prefixing it with the
1274 @samp{^} character, so for example to display all binary warning
1275 options (i.e., ones that are either on or off and that do not take an
1276 argument), which have a description the following can be used:
1279 --help=warnings,^joined,^undocumented
1282 The argument to @option{--help=} should not consist solely of inverted
1285 Combining several classes is possible, although this usually
1286 restricts the output by so much that there is nothing to display. One
1287 case where it does work however is when one of the classes is
1288 @var{target}. So for example to display all the target-specific
1289 optimization options the following can be used:
1292 --help=target,optimizers
1295 The @option{--help=} option can be repeated on the command line. Each
1296 successive use will display its requested class of options, skipping
1297 those that have already been displayed.
1299 If the @option{-Q} option appears on the command line before the
1300 @option{--help=} option, then the descriptive text displayed by
1301 @option{--help=} is changed. Instead of describing the displayed
1302 options, an indication is given as to whether the option is enabled,
1303 disabled or set to a specific value (assuming that the compiler
1304 knows this at the point where the @option{--help=} option is used).
1306 Here is a truncated example from the ARM port of @command{gcc}:
1309 % gcc -Q -mabi=2 --help=target -c
1310 The following options are target specific:
1312 -mabort-on-noreturn [disabled]
1316 The output is sensitive to the effects of previous command line
1317 options, so for example it is possible to find out which optimizations
1318 are enabled at @option{-O2} by using:
1321 -Q -O2 --help=optimizers
1324 Alternatively you can discover which binary optimizations are enabled
1325 by @option{-O3} by using:
1328 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1329 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1330 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1333 @item -no-canonical-prefixes
1334 @opindex no-canonical-prefixes
1335 Do not expand any symbolic links, resolve references to @samp{/../}
1336 or @samp{/./}, or make the path absolute when generating a relative
1341 Display the version number and copyrights of the invoked GCC@.
1345 Invoke all subcommands under a wrapper program. It takes a single
1346 comma separated list as an argument, which will be used to invoke
1350 gcc -c t.c -wrapper gdb,--args
1353 This will invoke all subprograms of gcc under "gdb --args",
1354 thus cc1 invocation will be "gdb --args cc1 ...".
1356 @item -fplugin=@var{name}.so
1357 Load the plugin code in file @var{name}.so, assumed to be a
1358 shared object to be dlopen'd by the compiler. The base name of
1359 the shared object file is used to identify the plugin for the
1360 purposes of argument parsing (See
1361 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1362 Each plugin should define the callback functions specified in the
1365 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1366 Define an argument called @var{key} with a value of @var{value}
1367 for the plugin called @var{name}.
1369 @include @value{srcdir}/../libiberty/at-file.texi
1373 @section Compiling C++ Programs
1375 @cindex suffixes for C++ source
1376 @cindex C++ source file suffixes
1377 C++ source files conventionally use one of the suffixes @samp{.C},
1378 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1379 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1380 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1381 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1382 files with these names and compiles them as C++ programs even if you
1383 call the compiler the same way as for compiling C programs (usually
1384 with the name @command{gcc}).
1388 However, the use of @command{gcc} does not add the C++ library.
1389 @command{g++} is a program that calls GCC and treats @samp{.c},
1390 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1391 files unless @option{-x} is used, and automatically specifies linking
1392 against the C++ library. This program is also useful when
1393 precompiling a C header file with a @samp{.h} extension for use in C++
1394 compilations. On many systems, @command{g++} is also installed with
1395 the name @command{c++}.
1397 @cindex invoking @command{g++}
1398 When you compile C++ programs, you may specify many of the same
1399 command-line options that you use for compiling programs in any
1400 language; or command-line options meaningful for C and related
1401 languages; or options that are meaningful only for C++ programs.
1402 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1403 explanations of options for languages related to C@.
1404 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1405 explanations of options that are meaningful only for C++ programs.
1407 @node C Dialect Options
1408 @section Options Controlling C Dialect
1409 @cindex dialect options
1410 @cindex language dialect options
1411 @cindex options, dialect
1413 The following options control the dialect of C (or languages derived
1414 from C, such as C++, Objective-C and Objective-C++) that the compiler
1418 @cindex ANSI support
1422 In C mode, this is equivalent to @samp{-std=c89}. In C++ mode, it is
1423 equivalent to @samp{-std=c++98}.
1425 This turns off certain features of GCC that are incompatible with ISO
1426 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1427 such as the @code{asm} and @code{typeof} keywords, and
1428 predefined macros such as @code{unix} and @code{vax} that identify the
1429 type of system you are using. It also enables the undesirable and
1430 rarely used ISO trigraph feature. For the C compiler,
1431 it disables recognition of C++ style @samp{//} comments as well as
1432 the @code{inline} keyword.
1434 The alternate keywords @code{__asm__}, @code{__extension__},
1435 @code{__inline__} and @code{__typeof__} continue to work despite
1436 @option{-ansi}. You would not want to use them in an ISO C program, of
1437 course, but it is useful to put them in header files that might be included
1438 in compilations done with @option{-ansi}. Alternate predefined macros
1439 such as @code{__unix__} and @code{__vax__} are also available, with or
1440 without @option{-ansi}.
1442 The @option{-ansi} option does not cause non-ISO programs to be
1443 rejected gratuitously. For that, @option{-pedantic} is required in
1444 addition to @option{-ansi}. @xref{Warning Options}.
1446 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1447 option is used. Some header files may notice this macro and refrain
1448 from declaring certain functions or defining certain macros that the
1449 ISO standard doesn't call for; this is to avoid interfering with any
1450 programs that might use these names for other things.
1452 Functions that would normally be built in but do not have semantics
1453 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1454 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1455 built-in functions provided by GCC}, for details of the functions
1460 Determine the language standard. @xref{Standards,,Language Standards
1461 Supported by GCC}, for details of these standard versions. This option
1462 is currently only supported when compiling C or C++.
1464 The compiler can accept several base standards, such as @samp{c89} or
1465 @samp{c++98}, and GNU dialects of those standards, such as
1466 @samp{gnu89} or @samp{gnu++98}. By specifying a base standard, the
1467 compiler will accept all programs following that standard and those
1468 using GNU extensions that do not contradict it. For example,
1469 @samp{-std=c89} turns off certain features of GCC that are
1470 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1471 keywords, but not other GNU extensions that do not have a meaning in
1472 ISO C90, such as omitting the middle term of a @code{?:}
1473 expression. On the other hand, by specifying a GNU dialect of a
1474 standard, all features the compiler support are enabled, even when
1475 those features change the meaning of the base standard and some
1476 strict-conforming programs may be rejected. The particular standard
1477 is used by @option{-pedantic} to identify which features are GNU
1478 extensions given that version of the standard. For example
1479 @samp{-std=gnu89 -pedantic} would warn about C++ style @samp{//}
1480 comments, while @samp{-std=gnu99 -pedantic} would not.
1482 A value for this option must be provided; possible values are
1487 Support all ISO C90 programs (certain GNU extensions that conflict
1488 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1490 @item iso9899:199409
1491 ISO C90 as modified in amendment 1.
1497 ISO C99. Note that this standard is not yet fully supported; see
1498 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1499 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1502 GNU dialect of ISO C90 (including some C99 features). This
1503 is the default for C code.
1507 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1508 this will become the default. The name @samp{gnu9x} is deprecated.
1511 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1515 GNU dialect of @option{-std=c++98}. This is the default for
1519 The working draft of the upcoming ISO C++0x standard. This option
1520 enables experimental features that are likely to be included in
1521 C++0x. The working draft is constantly changing, and any feature that is
1522 enabled by this flag may be removed from future versions of GCC if it is
1523 not part of the C++0x standard.
1526 GNU dialect of @option{-std=c++0x}. This option enables
1527 experimental features that may be removed in future versions of GCC.
1530 @item -fgnu89-inline
1531 @opindex fgnu89-inline
1532 The option @option{-fgnu89-inline} tells GCC to use the traditional
1533 GNU semantics for @code{inline} functions when in C99 mode.
1534 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1535 is accepted and ignored by GCC versions 4.1.3 up to but not including
1536 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1537 C99 mode. Using this option is roughly equivalent to adding the
1538 @code{gnu_inline} function attribute to all inline functions
1539 (@pxref{Function Attributes}).
1541 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1542 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1543 specifies the default behavior). This option was first supported in
1544 GCC 4.3. This option is not supported in C89 or gnu89 mode.
1546 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1547 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1548 in effect for @code{inline} functions. @xref{Common Predefined
1549 Macros,,,cpp,The C Preprocessor}.
1551 @item -aux-info @var{filename}
1553 Output to the given filename prototyped declarations for all functions
1554 declared and/or defined in a translation unit, including those in header
1555 files. This option is silently ignored in any language other than C@.
1557 Besides declarations, the file indicates, in comments, the origin of
1558 each declaration (source file and line), whether the declaration was
1559 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1560 @samp{O} for old, respectively, in the first character after the line
1561 number and the colon), and whether it came from a declaration or a
1562 definition (@samp{C} or @samp{F}, respectively, in the following
1563 character). In the case of function definitions, a K&R-style list of
1564 arguments followed by their declarations is also provided, inside
1565 comments, after the declaration.
1569 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1570 keyword, so that code can use these words as identifiers. You can use
1571 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1572 instead. @option{-ansi} implies @option{-fno-asm}.
1574 In C++, this switch only affects the @code{typeof} keyword, since
1575 @code{asm} and @code{inline} are standard keywords. You may want to
1576 use the @option{-fno-gnu-keywords} flag instead, which has the same
1577 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1578 switch only affects the @code{asm} and @code{typeof} keywords, since
1579 @code{inline} is a standard keyword in ISO C99.
1582 @itemx -fno-builtin-@var{function}
1583 @opindex fno-builtin
1584 @cindex built-in functions
1585 Don't recognize built-in functions that do not begin with
1586 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1587 functions provided by GCC}, for details of the functions affected,
1588 including those which are not built-in functions when @option{-ansi} or
1589 @option{-std} options for strict ISO C conformance are used because they
1590 do not have an ISO standard meaning.
1592 GCC normally generates special code to handle certain built-in functions
1593 more efficiently; for instance, calls to @code{alloca} may become single
1594 instructions that adjust the stack directly, and calls to @code{memcpy}
1595 may become inline copy loops. The resulting code is often both smaller
1596 and faster, but since the function calls no longer appear as such, you
1597 cannot set a breakpoint on those calls, nor can you change the behavior
1598 of the functions by linking with a different library. In addition,
1599 when a function is recognized as a built-in function, GCC may use
1600 information about that function to warn about problems with calls to
1601 that function, or to generate more efficient code, even if the
1602 resulting code still contains calls to that function. For example,
1603 warnings are given with @option{-Wformat} for bad calls to
1604 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1605 known not to modify global memory.
1607 With the @option{-fno-builtin-@var{function}} option
1608 only the built-in function @var{function} is
1609 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1610 function is named that is not built-in in this version of GCC, this
1611 option is ignored. There is no corresponding
1612 @option{-fbuiltin-@var{function}} option; if you wish to enable
1613 built-in functions selectively when using @option{-fno-builtin} or
1614 @option{-ffreestanding}, you may define macros such as:
1617 #define abs(n) __builtin_abs ((n))
1618 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1623 @cindex hosted environment
1625 Assert that compilation takes place in a hosted environment. This implies
1626 @option{-fbuiltin}. A hosted environment is one in which the
1627 entire standard library is available, and in which @code{main} has a return
1628 type of @code{int}. Examples are nearly everything except a kernel.
1629 This is equivalent to @option{-fno-freestanding}.
1631 @item -ffreestanding
1632 @opindex ffreestanding
1633 @cindex hosted environment
1635 Assert that compilation takes place in a freestanding environment. This
1636 implies @option{-fno-builtin}. A freestanding environment
1637 is one in which the standard library may not exist, and program startup may
1638 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1639 This is equivalent to @option{-fno-hosted}.
1641 @xref{Standards,,Language Standards Supported by GCC}, for details of
1642 freestanding and hosted environments.
1646 @cindex openmp parallel
1647 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1648 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1649 compiler generates parallel code according to the OpenMP Application
1650 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1651 implies @option{-pthread}, and thus is only supported on targets that
1652 have support for @option{-pthread}.
1654 @item -fms-extensions
1655 @opindex fms-extensions
1656 Accept some non-standard constructs used in Microsoft header files.
1658 Some cases of unnamed fields in structures and unions are only
1659 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1660 fields within structs/unions}, for details.
1664 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1665 options for strict ISO C conformance) implies @option{-trigraphs}.
1667 @item -no-integrated-cpp
1668 @opindex no-integrated-cpp
1669 Performs a compilation in two passes: preprocessing and compiling. This
1670 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1671 @option{-B} option. The user supplied compilation step can then add in
1672 an additional preprocessing step after normal preprocessing but before
1673 compiling. The default is to use the integrated cpp (internal cpp)
1675 The semantics of this option will change if "cc1", "cc1plus", and
1676 "cc1obj" are merged.
1678 @cindex traditional C language
1679 @cindex C language, traditional
1681 @itemx -traditional-cpp
1682 @opindex traditional-cpp
1683 @opindex traditional
1684 Formerly, these options caused GCC to attempt to emulate a pre-standard
1685 C compiler. They are now only supported with the @option{-E} switch.
1686 The preprocessor continues to support a pre-standard mode. See the GNU
1687 CPP manual for details.
1689 @item -fcond-mismatch
1690 @opindex fcond-mismatch
1691 Allow conditional expressions with mismatched types in the second and
1692 third arguments. The value of such an expression is void. This option
1693 is not supported for C++.
1695 @item -flax-vector-conversions
1696 @opindex flax-vector-conversions
1697 Allow implicit conversions between vectors with differing numbers of
1698 elements and/or incompatible element types. This option should not be
1701 @item -funsigned-char
1702 @opindex funsigned-char
1703 Let the type @code{char} be unsigned, like @code{unsigned char}.
1705 Each kind of machine has a default for what @code{char} should
1706 be. It is either like @code{unsigned char} by default or like
1707 @code{signed char} by default.
1709 Ideally, a portable program should always use @code{signed char} or
1710 @code{unsigned char} when it depends on the signedness of an object.
1711 But many programs have been written to use plain @code{char} and
1712 expect it to be signed, or expect it to be unsigned, depending on the
1713 machines they were written for. This option, and its inverse, let you
1714 make such a program work with the opposite default.
1716 The type @code{char} is always a distinct type from each of
1717 @code{signed char} or @code{unsigned char}, even though its behavior
1718 is always just like one of those two.
1721 @opindex fsigned-char
1722 Let the type @code{char} be signed, like @code{signed char}.
1724 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1725 the negative form of @option{-funsigned-char}. Likewise, the option
1726 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1728 @item -fsigned-bitfields
1729 @itemx -funsigned-bitfields
1730 @itemx -fno-signed-bitfields
1731 @itemx -fno-unsigned-bitfields
1732 @opindex fsigned-bitfields
1733 @opindex funsigned-bitfields
1734 @opindex fno-signed-bitfields
1735 @opindex fno-unsigned-bitfields
1736 These options control whether a bit-field is signed or unsigned, when the
1737 declaration does not use either @code{signed} or @code{unsigned}. By
1738 default, such a bit-field is signed, because this is consistent: the
1739 basic integer types such as @code{int} are signed types.
1742 @node C++ Dialect Options
1743 @section Options Controlling C++ Dialect
1745 @cindex compiler options, C++
1746 @cindex C++ options, command line
1747 @cindex options, C++
1748 This section describes the command-line options that are only meaningful
1749 for C++ programs; but you can also use most of the GNU compiler options
1750 regardless of what language your program is in. For example, you
1751 might compile a file @code{firstClass.C} like this:
1754 g++ -g -frepo -O -c firstClass.C
1758 In this example, only @option{-frepo} is an option meant
1759 only for C++ programs; you can use the other options with any
1760 language supported by GCC@.
1762 Here is a list of options that are @emph{only} for compiling C++ programs:
1766 @item -fabi-version=@var{n}
1767 @opindex fabi-version
1768 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1769 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1770 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1771 the version that conforms most closely to the C++ ABI specification.
1772 Therefore, the ABI obtained using version 0 will change as ABI bugs
1775 The default is version 2.
1777 @item -fno-access-control
1778 @opindex fno-access-control
1779 Turn off all access checking. This switch is mainly useful for working
1780 around bugs in the access control code.
1784 Check that the pointer returned by @code{operator new} is non-null
1785 before attempting to modify the storage allocated. This check is
1786 normally unnecessary because the C++ standard specifies that
1787 @code{operator new} will only return @code{0} if it is declared
1788 @samp{throw()}, in which case the compiler will always check the
1789 return value even without this option. In all other cases, when
1790 @code{operator new} has a non-empty exception specification, memory
1791 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1792 @samp{new (nothrow)}.
1794 @item -fconserve-space
1795 @opindex fconserve-space
1796 Put uninitialized or runtime-initialized global variables into the
1797 common segment, as C does. This saves space in the executable at the
1798 cost of not diagnosing duplicate definitions. If you compile with this
1799 flag and your program mysteriously crashes after @code{main()} has
1800 completed, you may have an object that is being destroyed twice because
1801 two definitions were merged.
1803 This option is no longer useful on most targets, now that support has
1804 been added for putting variables into BSS without making them common.
1806 @item -fno-deduce-init-list
1807 @opindex fno-deduce-init-list
1808 Disable deduction of a template type parameter as
1809 std::initializer_list from a brace-enclosed initializer list, i.e.
1812 template <class T> auto forward(T t) -> decltype (realfn (t))
1819 forward(@{1,2@}); // call forward<std::initializer_list<int>>
1823 This option is present because this deduction is an extension to the
1824 current specification in the C++0x working draft, and there was
1825 some concern about potential overload resolution problems.
1827 @item -ffriend-injection
1828 @opindex ffriend-injection
1829 Inject friend functions into the enclosing namespace, so that they are
1830 visible outside the scope of the class in which they are declared.
1831 Friend functions were documented to work this way in the old Annotated
1832 C++ Reference Manual, and versions of G++ before 4.1 always worked
1833 that way. However, in ISO C++ a friend function which is not declared
1834 in an enclosing scope can only be found using argument dependent
1835 lookup. This option causes friends to be injected as they were in
1838 This option is for compatibility, and may be removed in a future
1841 @item -fno-elide-constructors
1842 @opindex fno-elide-constructors
1843 The C++ standard allows an implementation to omit creating a temporary
1844 which is only used to initialize another object of the same type.
1845 Specifying this option disables that optimization, and forces G++ to
1846 call the copy constructor in all cases.
1848 @item -fno-enforce-eh-specs
1849 @opindex fno-enforce-eh-specs
1850 Don't generate code to check for violation of exception specifications
1851 at runtime. This option violates the C++ standard, but may be useful
1852 for reducing code size in production builds, much like defining
1853 @samp{NDEBUG}. This does not give user code permission to throw
1854 exceptions in violation of the exception specifications; the compiler
1855 will still optimize based on the specifications, so throwing an
1856 unexpected exception will result in undefined behavior.
1859 @itemx -fno-for-scope
1861 @opindex fno-for-scope
1862 If @option{-ffor-scope} is specified, the scope of variables declared in
1863 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1864 as specified by the C++ standard.
1865 If @option{-fno-for-scope} is specified, the scope of variables declared in
1866 a @i{for-init-statement} extends to the end of the enclosing scope,
1867 as was the case in old versions of G++, and other (traditional)
1868 implementations of C++.
1870 The default if neither flag is given to follow the standard,
1871 but to allow and give a warning for old-style code that would
1872 otherwise be invalid, or have different behavior.
1874 @item -fno-gnu-keywords
1875 @opindex fno-gnu-keywords
1876 Do not recognize @code{typeof} as a keyword, so that code can use this
1877 word as an identifier. You can use the keyword @code{__typeof__} instead.
1878 @option{-ansi} implies @option{-fno-gnu-keywords}.
1880 @item -fno-implicit-templates
1881 @opindex fno-implicit-templates
1882 Never emit code for non-inline templates which are instantiated
1883 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1884 @xref{Template Instantiation}, for more information.
1886 @item -fno-implicit-inline-templates
1887 @opindex fno-implicit-inline-templates
1888 Don't emit code for implicit instantiations of inline templates, either.
1889 The default is to handle inlines differently so that compiles with and
1890 without optimization will need the same set of explicit instantiations.
1892 @item -fno-implement-inlines
1893 @opindex fno-implement-inlines
1894 To save space, do not emit out-of-line copies of inline functions
1895 controlled by @samp{#pragma implementation}. This will cause linker
1896 errors if these functions are not inlined everywhere they are called.
1898 @item -fms-extensions
1899 @opindex fms-extensions
1900 Disable pedantic warnings about constructs used in MFC, such as implicit
1901 int and getting a pointer to member function via non-standard syntax.
1903 @item -fno-nonansi-builtins
1904 @opindex fno-nonansi-builtins
1905 Disable built-in declarations of functions that are not mandated by
1906 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1907 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1909 @item -fno-operator-names
1910 @opindex fno-operator-names
1911 Do not treat the operator name keywords @code{and}, @code{bitand},
1912 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1913 synonyms as keywords.
1915 @item -fno-optional-diags
1916 @opindex fno-optional-diags
1917 Disable diagnostics that the standard says a compiler does not need to
1918 issue. Currently, the only such diagnostic issued by G++ is the one for
1919 a name having multiple meanings within a class.
1922 @opindex fpermissive
1923 Downgrade some diagnostics about nonconformant code from errors to
1924 warnings. Thus, using @option{-fpermissive} will allow some
1925 nonconforming code to compile.
1927 @item -fno-pretty-templates
1928 @opindex fno-pretty-templates
1929 When an error message refers to a specialization of a function
1930 template, the compiler will normally print the signature of the
1931 template followed by the template arguments and any typedefs or
1932 typenames in the signature (e.g. @code{void f(T) [with T = int]}
1933 rather than @code{void f(int)}) so that it's clear which template is
1934 involved. When an error message refers to a specialization of a class
1935 template, the compiler will omit any template arguments which match
1936 the default template arguments for that template. If either of these
1937 behaviors make it harder to understand the error message rather than
1938 easier, using @option{-fno-pretty-templates} will disable them.
1942 Enable automatic template instantiation at link time. This option also
1943 implies @option{-fno-implicit-templates}. @xref{Template
1944 Instantiation}, for more information.
1948 Disable generation of information about every class with virtual
1949 functions for use by the C++ runtime type identification features
1950 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1951 of the language, you can save some space by using this flag. Note that
1952 exception handling uses the same information, but it will generate it as
1953 needed. The @samp{dynamic_cast} operator can still be used for casts that
1954 do not require runtime type information, i.e.@: casts to @code{void *} or to
1955 unambiguous base classes.
1959 Emit statistics about front-end processing at the end of the compilation.
1960 This information is generally only useful to the G++ development team.
1962 @item -ftemplate-depth-@var{n}
1963 @opindex ftemplate-depth
1964 Set the maximum instantiation depth for template classes to @var{n}.
1965 A limit on the template instantiation depth is needed to detect
1966 endless recursions during template class instantiation. ANSI/ISO C++
1967 conforming programs must not rely on a maximum depth greater than 17
1968 (changed to 1024 in C++0x).
1970 @item -fno-threadsafe-statics
1971 @opindex fno-threadsafe-statics
1972 Do not emit the extra code to use the routines specified in the C++
1973 ABI for thread-safe initialization of local statics. You can use this
1974 option to reduce code size slightly in code that doesn't need to be
1977 @item -fuse-cxa-atexit
1978 @opindex fuse-cxa-atexit
1979 Register destructors for objects with static storage duration with the
1980 @code{__cxa_atexit} function rather than the @code{atexit} function.
1981 This option is required for fully standards-compliant handling of static
1982 destructors, but will only work if your C library supports
1983 @code{__cxa_atexit}.
1985 @item -fno-use-cxa-get-exception-ptr
1986 @opindex fno-use-cxa-get-exception-ptr
1987 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
1988 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
1989 if the runtime routine is not available.
1991 @item -fvisibility-inlines-hidden
1992 @opindex fvisibility-inlines-hidden
1993 This switch declares that the user does not attempt to compare
1994 pointers to inline methods where the addresses of the two functions
1995 were taken in different shared objects.
1997 The effect of this is that GCC may, effectively, mark inline methods with
1998 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1999 appear in the export table of a DSO and do not require a PLT indirection
2000 when used within the DSO@. Enabling this option can have a dramatic effect
2001 on load and link times of a DSO as it massively reduces the size of the
2002 dynamic export table when the library makes heavy use of templates.
2004 The behavior of this switch is not quite the same as marking the
2005 methods as hidden directly, because it does not affect static variables
2006 local to the function or cause the compiler to deduce that
2007 the function is defined in only one shared object.
2009 You may mark a method as having a visibility explicitly to negate the
2010 effect of the switch for that method. For example, if you do want to
2011 compare pointers to a particular inline method, you might mark it as
2012 having default visibility. Marking the enclosing class with explicit
2013 visibility will have no effect.
2015 Explicitly instantiated inline methods are unaffected by this option
2016 as their linkage might otherwise cross a shared library boundary.
2017 @xref{Template Instantiation}.
2019 @item -fvisibility-ms-compat
2020 @opindex fvisibility-ms-compat
2021 This flag attempts to use visibility settings to make GCC's C++
2022 linkage model compatible with that of Microsoft Visual Studio.
2024 The flag makes these changes to GCC's linkage model:
2028 It sets the default visibility to @code{hidden}, like
2029 @option{-fvisibility=hidden}.
2032 Types, but not their members, are not hidden by default.
2035 The One Definition Rule is relaxed for types without explicit
2036 visibility specifications which are defined in more than one different
2037 shared object: those declarations are permitted if they would have
2038 been permitted when this option was not used.
2041 In new code it is better to use @option{-fvisibility=hidden} and
2042 export those classes which are intended to be externally visible.
2043 Unfortunately it is possible for code to rely, perhaps accidentally,
2044 on the Visual Studio behavior.
2046 Among the consequences of these changes are that static data members
2047 of the same type with the same name but defined in different shared
2048 objects will be different, so changing one will not change the other;
2049 and that pointers to function members defined in different shared
2050 objects may not compare equal. When this flag is given, it is a
2051 violation of the ODR to define types with the same name differently.
2055 Do not use weak symbol support, even if it is provided by the linker.
2056 By default, G++ will use weak symbols if they are available. This
2057 option exists only for testing, and should not be used by end-users;
2058 it will result in inferior code and has no benefits. This option may
2059 be removed in a future release of G++.
2063 Do not search for header files in the standard directories specific to
2064 C++, but do still search the other standard directories. (This option
2065 is used when building the C++ library.)
2068 In addition, these optimization, warning, and code generation options
2069 have meanings only for C++ programs:
2072 @item -fno-default-inline
2073 @opindex fno-default-inline
2074 Do not assume @samp{inline} for functions defined inside a class scope.
2075 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2076 functions will have linkage like inline functions; they just won't be
2079 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2082 Warn when G++ generates code that is probably not compatible with the
2083 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2084 all such cases, there are probably some cases that are not warned about,
2085 even though G++ is generating incompatible code. There may also be
2086 cases where warnings are emitted even though the code that is generated
2089 You should rewrite your code to avoid these warnings if you are
2090 concerned about the fact that code generated by G++ may not be binary
2091 compatible with code generated by other compilers.
2093 The known incompatibilities at this point include:
2098 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2099 pack data into the same byte as a base class. For example:
2102 struct A @{ virtual void f(); int f1 : 1; @};
2103 struct B : public A @{ int f2 : 1; @};
2107 In this case, G++ will place @code{B::f2} into the same byte
2108 as@code{A::f1}; other compilers will not. You can avoid this problem
2109 by explicitly padding @code{A} so that its size is a multiple of the
2110 byte size on your platform; that will cause G++ and other compilers to
2111 layout @code{B} identically.
2114 Incorrect handling of tail-padding for virtual bases. G++ does not use
2115 tail padding when laying out virtual bases. For example:
2118 struct A @{ virtual void f(); char c1; @};
2119 struct B @{ B(); char c2; @};
2120 struct C : public A, public virtual B @{@};
2124 In this case, G++ will not place @code{B} into the tail-padding for
2125 @code{A}; other compilers will. You can avoid this problem by
2126 explicitly padding @code{A} so that its size is a multiple of its
2127 alignment (ignoring virtual base classes); that will cause G++ and other
2128 compilers to layout @code{C} identically.
2131 Incorrect handling of bit-fields with declared widths greater than that
2132 of their underlying types, when the bit-fields appear in a union. For
2136 union U @{ int i : 4096; @};
2140 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2141 union too small by the number of bits in an @code{int}.
2144 Empty classes can be placed at incorrect offsets. For example:
2154 struct C : public B, public A @{@};
2158 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2159 it should be placed at offset zero. G++ mistakenly believes that the
2160 @code{A} data member of @code{B} is already at offset zero.
2163 Names of template functions whose types involve @code{typename} or
2164 template template parameters can be mangled incorrectly.
2167 template <typename Q>
2168 void f(typename Q::X) @{@}
2170 template <template <typename> class Q>
2171 void f(typename Q<int>::X) @{@}
2175 Instantiations of these templates may be mangled incorrectly.
2179 It also warns psABI related changes. The known psABI changes at this
2185 For SYSV/x86-64, when passing union with long double, it is changed to
2186 pass in memory as specified in psABI. For example:
2196 @code{union U} will always be passed in memory.
2200 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2201 @opindex Wctor-dtor-privacy
2202 @opindex Wno-ctor-dtor-privacy
2203 Warn when a class seems unusable because all the constructors or
2204 destructors in that class are private, and it has neither friends nor
2205 public static member functions.
2207 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2208 @opindex Wnon-virtual-dtor
2209 @opindex Wno-non-virtual-dtor
2210 Warn when a class has virtual functions and accessible non-virtual
2211 destructor, in which case it would be possible but unsafe to delete
2212 an instance of a derived class through a pointer to the base class.
2213 This warning is also enabled if -Weffc++ is specified.
2215 @item -Wreorder @r{(C++ and Objective-C++ only)}
2217 @opindex Wno-reorder
2218 @cindex reordering, warning
2219 @cindex warning for reordering of member initializers
2220 Warn when the order of member initializers given in the code does not
2221 match the order in which they must be executed. For instance:
2227 A(): j (0), i (1) @{ @}
2231 The compiler will rearrange the member initializers for @samp{i}
2232 and @samp{j} to match the declaration order of the members, emitting
2233 a warning to that effect. This warning is enabled by @option{-Wall}.
2236 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2239 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2242 Warn about violations of the following style guidelines from Scott Meyers'
2243 @cite{Effective C++} book:
2247 Item 11: Define a copy constructor and an assignment operator for classes
2248 with dynamically allocated memory.
2251 Item 12: Prefer initialization to assignment in constructors.
2254 Item 14: Make destructors virtual in base classes.
2257 Item 15: Have @code{operator=} return a reference to @code{*this}.
2260 Item 23: Don't try to return a reference when you must return an object.
2264 Also warn about violations of the following style guidelines from
2265 Scott Meyers' @cite{More Effective C++} book:
2269 Item 6: Distinguish between prefix and postfix forms of increment and
2270 decrement operators.
2273 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2277 When selecting this option, be aware that the standard library
2278 headers do not obey all of these guidelines; use @samp{grep -v}
2279 to filter out those warnings.
2281 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2282 @opindex Wstrict-null-sentinel
2283 @opindex Wno-strict-null-sentinel
2284 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2285 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2286 to @code{__null}. Although it is a null pointer constant not a null pointer,
2287 it is guaranteed to be of the same size as a pointer. But this use is
2288 not portable across different compilers.
2290 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2291 @opindex Wno-non-template-friend
2292 @opindex Wnon-template-friend
2293 Disable warnings when non-templatized friend functions are declared
2294 within a template. Since the advent of explicit template specification
2295 support in G++, if the name of the friend is an unqualified-id (i.e.,
2296 @samp{friend foo(int)}), the C++ language specification demands that the
2297 friend declare or define an ordinary, nontemplate function. (Section
2298 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2299 could be interpreted as a particular specialization of a templatized
2300 function. Because this non-conforming behavior is no longer the default
2301 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2302 check existing code for potential trouble spots and is on by default.
2303 This new compiler behavior can be turned off with
2304 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2305 but disables the helpful warning.
2307 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2308 @opindex Wold-style-cast
2309 @opindex Wno-old-style-cast
2310 Warn if an old-style (C-style) cast to a non-void type is used within
2311 a C++ program. The new-style casts (@samp{dynamic_cast},
2312 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2313 less vulnerable to unintended effects and much easier to search for.
2315 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2316 @opindex Woverloaded-virtual
2317 @opindex Wno-overloaded-virtual
2318 @cindex overloaded virtual fn, warning
2319 @cindex warning for overloaded virtual fn
2320 Warn when a function declaration hides virtual functions from a
2321 base class. For example, in:
2328 struct B: public A @{
2333 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2341 will fail to compile.
2343 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2344 @opindex Wno-pmf-conversions
2345 @opindex Wpmf-conversions
2346 Disable the diagnostic for converting a bound pointer to member function
2349 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2350 @opindex Wsign-promo
2351 @opindex Wno-sign-promo
2352 Warn when overload resolution chooses a promotion from unsigned or
2353 enumerated type to a signed type, over a conversion to an unsigned type of
2354 the same size. Previous versions of G++ would try to preserve
2355 unsignedness, but the standard mandates the current behavior.
2360 A& operator = (int);
2370 In this example, G++ will synthesize a default @samp{A& operator =
2371 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2374 @node Objective-C and Objective-C++ Dialect Options
2375 @section Options Controlling Objective-C and Objective-C++ Dialects
2377 @cindex compiler options, Objective-C and Objective-C++
2378 @cindex Objective-C and Objective-C++ options, command line
2379 @cindex options, Objective-C and Objective-C++
2380 (NOTE: This manual does not describe the Objective-C and Objective-C++
2381 languages themselves. See @xref{Standards,,Language Standards
2382 Supported by GCC}, for references.)
2384 This section describes the command-line options that are only meaningful
2385 for Objective-C and Objective-C++ programs, but you can also use most of
2386 the language-independent GNU compiler options.
2387 For example, you might compile a file @code{some_class.m} like this:
2390 gcc -g -fgnu-runtime -O -c some_class.m
2394 In this example, @option{-fgnu-runtime} is an option meant only for
2395 Objective-C and Objective-C++ programs; you can use the other options with
2396 any language supported by GCC@.
2398 Note that since Objective-C is an extension of the C language, Objective-C
2399 compilations may also use options specific to the C front-end (e.g.,
2400 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2401 C++-specific options (e.g., @option{-Wabi}).
2403 Here is a list of options that are @emph{only} for compiling Objective-C
2404 and Objective-C++ programs:
2407 @item -fconstant-string-class=@var{class-name}
2408 @opindex fconstant-string-class
2409 Use @var{class-name} as the name of the class to instantiate for each
2410 literal string specified with the syntax @code{@@"@dots{}"}. The default
2411 class name is @code{NXConstantString} if the GNU runtime is being used, and
2412 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2413 @option{-fconstant-cfstrings} option, if also present, will override the
2414 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2415 to be laid out as constant CoreFoundation strings.
2418 @opindex fgnu-runtime
2419 Generate object code compatible with the standard GNU Objective-C
2420 runtime. This is the default for most types of systems.
2422 @item -fnext-runtime
2423 @opindex fnext-runtime
2424 Generate output compatible with the NeXT runtime. This is the default
2425 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2426 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2429 @item -fno-nil-receivers
2430 @opindex fno-nil-receivers
2431 Assume that all Objective-C message dispatches (e.g.,
2432 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
2433 is not @code{nil}. This allows for more efficient entry points in the runtime
2434 to be used. Currently, this option is only available in conjunction with
2435 the NeXT runtime on Mac OS X 10.3 and later.
2437 @item -fobjc-call-cxx-cdtors
2438 @opindex fobjc-call-cxx-cdtors
2439 For each Objective-C class, check if any of its instance variables is a
2440 C++ object with a non-trivial default constructor. If so, synthesize a
2441 special @code{- (id) .cxx_construct} instance method that will run
2442 non-trivial default constructors on any such instance variables, in order,
2443 and then return @code{self}. Similarly, check if any instance variable
2444 is a C++ object with a non-trivial destructor, and if so, synthesize a
2445 special @code{- (void) .cxx_destruct} method that will run
2446 all such default destructors, in reverse order.
2448 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
2449 thusly generated will only operate on instance variables declared in the
2450 current Objective-C class, and not those inherited from superclasses. It
2451 is the responsibility of the Objective-C runtime to invoke all such methods
2452 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
2453 will be invoked by the runtime immediately after a new object
2454 instance is allocated; the @code{- (void) .cxx_destruct} methods will
2455 be invoked immediately before the runtime deallocates an object instance.
2457 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2458 support for invoking the @code{- (id) .cxx_construct} and
2459 @code{- (void) .cxx_destruct} methods.
2461 @item -fobjc-direct-dispatch
2462 @opindex fobjc-direct-dispatch
2463 Allow fast jumps to the message dispatcher. On Darwin this is
2464 accomplished via the comm page.
2466 @item -fobjc-exceptions
2467 @opindex fobjc-exceptions
2468 Enable syntactic support for structured exception handling in Objective-C,
2469 similar to what is offered by C++ and Java. This option is
2470 unavailable in conjunction with the NeXT runtime on Mac OS X 10.2 and
2479 @@catch (AnObjCClass *exc) @{
2486 @@catch (AnotherClass *exc) @{
2489 @@catch (id allOthers) @{
2499 The @code{@@throw} statement may appear anywhere in an Objective-C or
2500 Objective-C++ program; when used inside of a @code{@@catch} block, the
2501 @code{@@throw} may appear without an argument (as shown above), in which case
2502 the object caught by the @code{@@catch} will be rethrown.
2504 Note that only (pointers to) Objective-C objects may be thrown and
2505 caught using this scheme. When an object is thrown, it will be caught
2506 by the nearest @code{@@catch} clause capable of handling objects of that type,
2507 analogously to how @code{catch} blocks work in C++ and Java. A
2508 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2509 any and all Objective-C exceptions not caught by previous @code{@@catch}
2512 The @code{@@finally} clause, if present, will be executed upon exit from the
2513 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2514 regardless of whether any exceptions are thrown, caught or rethrown
2515 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2516 of the @code{finally} clause in Java.
2518 There are several caveats to using the new exception mechanism:
2522 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2523 idioms provided by the @code{NSException} class, the new
2524 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2525 systems, due to additional functionality needed in the (NeXT) Objective-C
2529 As mentioned above, the new exceptions do not support handling
2530 types other than Objective-C objects. Furthermore, when used from
2531 Objective-C++, the Objective-C exception model does not interoperate with C++
2532 exceptions at this time. This means you cannot @code{@@throw} an exception
2533 from Objective-C and @code{catch} it in C++, or vice versa
2534 (i.e., @code{throw @dots{} @@catch}).
2537 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2538 blocks for thread-safe execution:
2541 @@synchronized (ObjCClass *guard) @{
2546 Upon entering the @code{@@synchronized} block, a thread of execution shall
2547 first check whether a lock has been placed on the corresponding @code{guard}
2548 object by another thread. If it has, the current thread shall wait until
2549 the other thread relinquishes its lock. Once @code{guard} becomes available,
2550 the current thread will place its own lock on it, execute the code contained in
2551 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2552 making @code{guard} available to other threads).
2554 Unlike Java, Objective-C does not allow for entire methods to be marked
2555 @code{@@synchronized}. Note that throwing exceptions out of
2556 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2557 to be unlocked properly.
2561 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2563 @item -freplace-objc-classes
2564 @opindex freplace-objc-classes
2565 Emit a special marker instructing @command{ld(1)} not to statically link in
2566 the resulting object file, and allow @command{dyld(1)} to load it in at
2567 run time instead. This is used in conjunction with the Fix-and-Continue
2568 debugging mode, where the object file in question may be recompiled and
2569 dynamically reloaded in the course of program execution, without the need
2570 to restart the program itself. Currently, Fix-and-Continue functionality
2571 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2576 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2577 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2578 compile time) with static class references that get initialized at load time,
2579 which improves run-time performance. Specifying the @option{-fzero-link} flag
2580 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2581 to be retained. This is useful in Zero-Link debugging mode, since it allows
2582 for individual class implementations to be modified during program execution.
2586 Dump interface declarations for all classes seen in the source file to a
2587 file named @file{@var{sourcename}.decl}.
2589 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2590 @opindex Wassign-intercept
2591 @opindex Wno-assign-intercept
2592 Warn whenever an Objective-C assignment is being intercepted by the
2595 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2596 @opindex Wno-protocol
2598 If a class is declared to implement a protocol, a warning is issued for
2599 every method in the protocol that is not implemented by the class. The
2600 default behavior is to issue a warning for every method not explicitly
2601 implemented in the class, even if a method implementation is inherited
2602 from the superclass. If you use the @option{-Wno-protocol} option, then
2603 methods inherited from the superclass are considered to be implemented,
2604 and no warning is issued for them.
2606 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2608 @opindex Wno-selector
2609 Warn if multiple methods of different types for the same selector are
2610 found during compilation. The check is performed on the list of methods
2611 in the final stage of compilation. Additionally, a check is performed
2612 for each selector appearing in a @code{@@selector(@dots{})}
2613 expression, and a corresponding method for that selector has been found
2614 during compilation. Because these checks scan the method table only at
2615 the end of compilation, these warnings are not produced if the final
2616 stage of compilation is not reached, for example because an error is
2617 found during compilation, or because the @option{-fsyntax-only} option is
2620 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2621 @opindex Wstrict-selector-match
2622 @opindex Wno-strict-selector-match
2623 Warn if multiple methods with differing argument and/or return types are
2624 found for a given selector when attempting to send a message using this
2625 selector to a receiver of type @code{id} or @code{Class}. When this flag
2626 is off (which is the default behavior), the compiler will omit such warnings
2627 if any differences found are confined to types which share the same size
2630 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2631 @opindex Wundeclared-selector
2632 @opindex Wno-undeclared-selector
2633 Warn if a @code{@@selector(@dots{})} expression referring to an
2634 undeclared selector is found. A selector is considered undeclared if no
2635 method with that name has been declared before the
2636 @code{@@selector(@dots{})} expression, either explicitly in an
2637 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2638 an @code{@@implementation} section. This option always performs its
2639 checks as soon as a @code{@@selector(@dots{})} expression is found,
2640 while @option{-Wselector} only performs its checks in the final stage of
2641 compilation. This also enforces the coding style convention
2642 that methods and selectors must be declared before being used.
2644 @item -print-objc-runtime-info
2645 @opindex print-objc-runtime-info
2646 Generate C header describing the largest structure that is passed by
2651 @node Language Independent Options
2652 @section Options to Control Diagnostic Messages Formatting
2653 @cindex options to control diagnostics formatting
2654 @cindex diagnostic messages
2655 @cindex message formatting
2657 Traditionally, diagnostic messages have been formatted irrespective of
2658 the output device's aspect (e.g.@: its width, @dots{}). The options described
2659 below can be used to control the diagnostic messages formatting
2660 algorithm, e.g.@: how many characters per line, how often source location
2661 information should be reported. Right now, only the C++ front end can
2662 honor these options. However it is expected, in the near future, that
2663 the remaining front ends would be able to digest them correctly.
2666 @item -fmessage-length=@var{n}
2667 @opindex fmessage-length
2668 Try to format error messages so that they fit on lines of about @var{n}
2669 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2670 the front ends supported by GCC@. If @var{n} is zero, then no
2671 line-wrapping will be done; each error message will appear on a single
2674 @opindex fdiagnostics-show-location
2675 @item -fdiagnostics-show-location=once
2676 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2677 reporter to emit @emph{once} source location information; that is, in
2678 case the message is too long to fit on a single physical line and has to
2679 be wrapped, the source location won't be emitted (as prefix) again,
2680 over and over, in subsequent continuation lines. This is the default
2683 @item -fdiagnostics-show-location=every-line
2684 Only meaningful in line-wrapping mode. Instructs the diagnostic
2685 messages reporter to emit the same source location information (as
2686 prefix) for physical lines that result from the process of breaking
2687 a message which is too long to fit on a single line.
2689 @item -fdiagnostics-show-option
2690 @opindex fdiagnostics-show-option
2691 This option instructs the diagnostic machinery to add text to each
2692 diagnostic emitted, which indicates which command line option directly
2693 controls that diagnostic, when such an option is known to the
2694 diagnostic machinery.
2696 @item -Wcoverage-mismatch
2697 @opindex Wcoverage-mismatch
2698 Warn if feedback profiles do not match when using the
2699 @option{-fprofile-use} option.
2700 If a source file was changed between @option{-fprofile-gen} and
2701 @option{-fprofile-use}, the files with the profile feedback can fail
2702 to match the source file and GCC can not use the profile feedback
2703 information. By default, GCC emits an error message in this case.
2704 The option @option{-Wcoverage-mismatch} emits a warning instead of an
2705 error. GCC does not use appropriate feedback profiles, so using this
2706 option can result in poorly optimized code. This option is useful
2707 only in the case of very minor changes such as bug fixes to an
2712 @node Warning Options
2713 @section Options to Request or Suppress Warnings
2714 @cindex options to control warnings
2715 @cindex warning messages
2716 @cindex messages, warning
2717 @cindex suppressing warnings
2719 Warnings are diagnostic messages that report constructions which
2720 are not inherently erroneous but which are risky or suggest there
2721 may have been an error.
2723 The following language-independent options do not enable specific
2724 warnings but control the kinds of diagnostics produced by GCC.
2727 @cindex syntax checking
2729 @opindex fsyntax-only
2730 Check the code for syntax errors, but don't do anything beyond that.
2734 Inhibit all warning messages.
2739 Make all warnings into errors.
2744 Make the specified warning into an error. The specifier for a warning
2745 is appended, for example @option{-Werror=switch} turns the warnings
2746 controlled by @option{-Wswitch} into errors. This switch takes a
2747 negative form, to be used to negate @option{-Werror} for specific
2748 warnings, for example @option{-Wno-error=switch} makes
2749 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2750 is in effect. You can use the @option{-fdiagnostics-show-option}
2751 option to have each controllable warning amended with the option which
2752 controls it, to determine what to use with this option.
2754 Note that specifying @option{-Werror=}@var{foo} automatically implies
2755 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2758 @item -Wfatal-errors
2759 @opindex Wfatal-errors
2760 @opindex Wno-fatal-errors
2761 This option causes the compiler to abort compilation on the first error
2762 occurred rather than trying to keep going and printing further error
2767 You can request many specific warnings with options beginning
2768 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2769 implicit declarations. Each of these specific warning options also
2770 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2771 example, @option{-Wno-implicit}. This manual lists only one of the
2772 two forms, whichever is not the default. For further,
2773 language-specific options also refer to @ref{C++ Dialect Options} and
2774 @ref{Objective-C and Objective-C++ Dialect Options}.
2779 Issue all the warnings demanded by strict ISO C and ISO C++;
2780 reject all programs that use forbidden extensions, and some other
2781 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2782 version of the ISO C standard specified by any @option{-std} option used.
2784 Valid ISO C and ISO C++ programs should compile properly with or without
2785 this option (though a rare few will require @option{-ansi} or a
2786 @option{-std} option specifying the required version of ISO C)@. However,
2787 without this option, certain GNU extensions and traditional C and C++
2788 features are supported as well. With this option, they are rejected.
2790 @option{-pedantic} does not cause warning messages for use of the
2791 alternate keywords whose names begin and end with @samp{__}. Pedantic
2792 warnings are also disabled in the expression that follows
2793 @code{__extension__}. However, only system header files should use
2794 these escape routes; application programs should avoid them.
2795 @xref{Alternate Keywords}.
2797 Some users try to use @option{-pedantic} to check programs for strict ISO
2798 C conformance. They soon find that it does not do quite what they want:
2799 it finds some non-ISO practices, but not all---only those for which
2800 ISO C @emph{requires} a diagnostic, and some others for which
2801 diagnostics have been added.
2803 A feature to report any failure to conform to ISO C might be useful in
2804 some instances, but would require considerable additional work and would
2805 be quite different from @option{-pedantic}. We don't have plans to
2806 support such a feature in the near future.
2808 Where the standard specified with @option{-std} represents a GNU
2809 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2810 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2811 extended dialect is based. Warnings from @option{-pedantic} are given
2812 where they are required by the base standard. (It would not make sense
2813 for such warnings to be given only for features not in the specified GNU
2814 C dialect, since by definition the GNU dialects of C include all
2815 features the compiler supports with the given option, and there would be
2816 nothing to warn about.)
2818 @item -pedantic-errors
2819 @opindex pedantic-errors
2820 Like @option{-pedantic}, except that errors are produced rather than
2826 This enables all the warnings about constructions that some users
2827 consider questionable, and that are easy to avoid (or modify to
2828 prevent the warning), even in conjunction with macros. This also
2829 enables some language-specific warnings described in @ref{C++ Dialect
2830 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2832 @option{-Wall} turns on the following warning flags:
2834 @gccoptlist{-Waddress @gol
2835 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2837 -Wchar-subscripts @gol
2838 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
2840 -Wimplicit-function-declaration @gol
2843 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2844 -Wmissing-braces @gol
2850 -Wsequence-point @gol
2851 -Wsign-compare @r{(only in C++)} @gol
2852 -Wstrict-aliasing @gol
2853 -Wstrict-overflow=1 @gol
2856 -Wuninitialized @gol
2857 -Wunknown-pragmas @gol
2858 -Wunused-function @gol
2861 -Wunused-variable @gol
2862 -Wvolatile-register-var @gol
2865 Note that some warning flags are not implied by @option{-Wall}. Some of
2866 them warn about constructions that users generally do not consider
2867 questionable, but which occasionally you might wish to check for;
2868 others warn about constructions that are necessary or hard to avoid in
2869 some cases, and there is no simple way to modify the code to suppress
2870 the warning. Some of them are enabled by @option{-Wextra} but many of
2871 them must be enabled individually.
2877 This enables some extra warning flags that are not enabled by
2878 @option{-Wall}. (This option used to be called @option{-W}. The older
2879 name is still supported, but the newer name is more descriptive.)
2881 @gccoptlist{-Wclobbered @gol
2883 -Wignored-qualifiers @gol
2884 -Wmissing-field-initializers @gol
2885 -Wmissing-parameter-type @r{(C only)} @gol
2886 -Wold-style-declaration @r{(C only)} @gol
2887 -Woverride-init @gol
2890 -Wuninitialized @gol
2891 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2894 The option @option{-Wextra} also prints warning messages for the
2900 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2901 @samp{>}, or @samp{>=}.
2904 (C++ only) An enumerator and a non-enumerator both appear in a
2905 conditional expression.
2908 (C++ only) Ambiguous virtual bases.
2911 (C++ only) Subscripting an array which has been declared @samp{register}.
2914 (C++ only) Taking the address of a variable which has been declared
2918 (C++ only) A base class is not initialized in a derived class' copy
2923 @item -Wchar-subscripts
2924 @opindex Wchar-subscripts
2925 @opindex Wno-char-subscripts
2926 Warn if an array subscript has type @code{char}. This is a common cause
2927 of error, as programmers often forget that this type is signed on some
2929 This warning is enabled by @option{-Wall}.
2933 @opindex Wno-comment
2934 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2935 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2936 This warning is enabled by @option{-Wall}.
2941 @opindex ffreestanding
2942 @opindex fno-builtin
2943 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2944 the arguments supplied have types appropriate to the format string
2945 specified, and that the conversions specified in the format string make
2946 sense. This includes standard functions, and others specified by format
2947 attributes (@pxref{Function Attributes}), in the @code{printf},
2948 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2949 not in the C standard) families (or other target-specific families).
2950 Which functions are checked without format attributes having been
2951 specified depends on the standard version selected, and such checks of
2952 functions without the attribute specified are disabled by
2953 @option{-ffreestanding} or @option{-fno-builtin}.
2955 The formats are checked against the format features supported by GNU
2956 libc version 2.2. These include all ISO C90 and C99 features, as well
2957 as features from the Single Unix Specification and some BSD and GNU
2958 extensions. Other library implementations may not support all these
2959 features; GCC does not support warning about features that go beyond a
2960 particular library's limitations. However, if @option{-pedantic} is used
2961 with @option{-Wformat}, warnings will be given about format features not
2962 in the selected standard version (but not for @code{strfmon} formats,
2963 since those are not in any version of the C standard). @xref{C Dialect
2964 Options,,Options Controlling C Dialect}.
2966 Since @option{-Wformat} also checks for null format arguments for
2967 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2969 @option{-Wformat} is included in @option{-Wall}. For more control over some
2970 aspects of format checking, the options @option{-Wformat-y2k},
2971 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2972 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2973 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2976 @opindex Wformat-y2k
2977 @opindex Wno-format-y2k
2978 If @option{-Wformat} is specified, also warn about @code{strftime}
2979 formats which may yield only a two-digit year.
2981 @item -Wno-format-contains-nul
2982 @opindex Wno-format-contains-nul
2983 @opindex Wformat-contains-nul
2984 If @option{-Wformat} is specified, do not warn about format strings that
2987 @item -Wno-format-extra-args
2988 @opindex Wno-format-extra-args
2989 @opindex Wformat-extra-args
2990 If @option{-Wformat} is specified, do not warn about excess arguments to a
2991 @code{printf} or @code{scanf} format function. The C standard specifies
2992 that such arguments are ignored.
2994 Where the unused arguments lie between used arguments that are
2995 specified with @samp{$} operand number specifications, normally
2996 warnings are still given, since the implementation could not know what
2997 type to pass to @code{va_arg} to skip the unused arguments. However,
2998 in the case of @code{scanf} formats, this option will suppress the
2999 warning if the unused arguments are all pointers, since the Single
3000 Unix Specification says that such unused arguments are allowed.
3002 @item -Wno-format-zero-length @r{(C and Objective-C only)}
3003 @opindex Wno-format-zero-length
3004 @opindex Wformat-zero-length
3005 If @option{-Wformat} is specified, do not warn about zero-length formats.
3006 The C standard specifies that zero-length formats are allowed.
3008 @item -Wformat-nonliteral
3009 @opindex Wformat-nonliteral
3010 @opindex Wno-format-nonliteral
3011 If @option{-Wformat} is specified, also warn if the format string is not a
3012 string literal and so cannot be checked, unless the format function
3013 takes its format arguments as a @code{va_list}.
3015 @item -Wformat-security
3016 @opindex Wformat-security
3017 @opindex Wno-format-security
3018 If @option{-Wformat} is specified, also warn about uses of format
3019 functions that represent possible security problems. At present, this
3020 warns about calls to @code{printf} and @code{scanf} functions where the
3021 format string is not a string literal and there are no format arguments,
3022 as in @code{printf (foo);}. This may be a security hole if the format
3023 string came from untrusted input and contains @samp{%n}. (This is
3024 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3025 in future warnings may be added to @option{-Wformat-security} that are not
3026 included in @option{-Wformat-nonliteral}.)
3030 @opindex Wno-format=2
3031 Enable @option{-Wformat} plus format checks not included in
3032 @option{-Wformat}. Currently equivalent to @samp{-Wformat
3033 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
3035 @item -Wnonnull @r{(C and Objective-C only)}
3037 @opindex Wno-nonnull
3038 Warn about passing a null pointer for arguments marked as
3039 requiring a non-null value by the @code{nonnull} function attribute.
3041 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3042 can be disabled with the @option{-Wno-nonnull} option.
3044 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3046 @opindex Wno-init-self
3047 Warn about uninitialized variables which are initialized with themselves.
3048 Note this option can only be used with the @option{-Wuninitialized} option.
3050 For example, GCC will warn about @code{i} being uninitialized in the
3051 following snippet only when @option{-Winit-self} has been specified:
3062 @item -Wimplicit-int @r{(C and Objective-C only)}
3063 @opindex Wimplicit-int
3064 @opindex Wno-implicit-int
3065 Warn when a declaration does not specify a type.
3066 This warning is enabled by @option{-Wall}.
3068 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3069 @opindex Wimplicit-function-declaration
3070 @opindex Wno-implicit-function-declaration
3071 Give a warning whenever a function is used before being declared. In
3072 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3073 enabled by default and it is made into an error by
3074 @option{-pedantic-errors}. This warning is also enabled by
3079 @opindex Wno-implicit
3080 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3081 This warning is enabled by @option{-Wall}.
3083 @item -Wignored-qualifiers @r{(C and C++ only)}
3084 @opindex Wignored-qualifiers
3085 @opindex Wno-ignored-qualifiers
3086 Warn if the return type of a function has a type qualifier
3087 such as @code{const}. For ISO C such a type qualifier has no effect,
3088 since the value returned by a function is not an lvalue.
3089 For C++, the warning is only emitted for scalar types or @code{void}.
3090 ISO C prohibits qualified @code{void} return types on function
3091 definitions, so such return types always receive a warning
3092 even without this option.
3094 This warning is also enabled by @option{-Wextra}.
3099 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3100 a function with external linkage, returning int, taking either zero
3101 arguments, two, or three arguments of appropriate types. This warning
3102 is enabled by default in C++ and is enabled by either @option{-Wall}
3103 or @option{-pedantic}.
3105 @item -Wmissing-braces
3106 @opindex Wmissing-braces
3107 @opindex Wno-missing-braces
3108 Warn if an aggregate or union initializer is not fully bracketed. In
3109 the following example, the initializer for @samp{a} is not fully
3110 bracketed, but that for @samp{b} is fully bracketed.
3113 int a[2][2] = @{ 0, 1, 2, 3 @};
3114 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3117 This warning is enabled by @option{-Wall}.
3119 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3120 @opindex Wmissing-include-dirs
3121 @opindex Wno-missing-include-dirs
3122 Warn if a user-supplied include directory does not exist.
3125 @opindex Wparentheses
3126 @opindex Wno-parentheses
3127 Warn if parentheses are omitted in certain contexts, such
3128 as when there is an assignment in a context where a truth value
3129 is expected, or when operators are nested whose precedence people
3130 often get confused about.
3132 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3133 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3134 interpretation from that of ordinary mathematical notation.
3136 Also warn about constructions where there may be confusion to which
3137 @code{if} statement an @code{else} branch belongs. Here is an example of
3152 In C/C++, every @code{else} branch belongs to the innermost possible
3153 @code{if} statement, which in this example is @code{if (b)}. This is
3154 often not what the programmer expected, as illustrated in the above
3155 example by indentation the programmer chose. When there is the
3156 potential for this confusion, GCC will issue a warning when this flag
3157 is specified. To eliminate the warning, add explicit braces around
3158 the innermost @code{if} statement so there is no way the @code{else}
3159 could belong to the enclosing @code{if}. The resulting code would
3176 This warning is enabled by @option{-Wall}.
3178 @item -Wsequence-point
3179 @opindex Wsequence-point
3180 @opindex Wno-sequence-point
3181 Warn about code that may have undefined semantics because of violations
3182 of sequence point rules in the C and C++ standards.
3184 The C and C++ standards defines the order in which expressions in a C/C++
3185 program are evaluated in terms of @dfn{sequence points}, which represent
3186 a partial ordering between the execution of parts of the program: those
3187 executed before the sequence point, and those executed after it. These
3188 occur after the evaluation of a full expression (one which is not part
3189 of a larger expression), after the evaluation of the first operand of a
3190 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3191 function is called (but after the evaluation of its arguments and the
3192 expression denoting the called function), and in certain other places.
3193 Other than as expressed by the sequence point rules, the order of
3194 evaluation of subexpressions of an expression is not specified. All
3195 these rules describe only a partial order rather than a total order,
3196 since, for example, if two functions are called within one expression
3197 with no sequence point between them, the order in which the functions
3198 are called is not specified. However, the standards committee have
3199 ruled that function calls do not overlap.
3201 It is not specified when between sequence points modifications to the
3202 values of objects take effect. Programs whose behavior depends on this
3203 have undefined behavior; the C and C++ standards specify that ``Between
3204 the previous and next sequence point an object shall have its stored
3205 value modified at most once by the evaluation of an expression.
3206 Furthermore, the prior value shall be read only to determine the value
3207 to be stored.''. If a program breaks these rules, the results on any
3208 particular implementation are entirely unpredictable.
3210 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3211 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3212 diagnosed by this option, and it may give an occasional false positive
3213 result, but in general it has been found fairly effective at detecting
3214 this sort of problem in programs.
3216 The standard is worded confusingly, therefore there is some debate
3217 over the precise meaning of the sequence point rules in subtle cases.
3218 Links to discussions of the problem, including proposed formal
3219 definitions, may be found on the GCC readings page, at
3220 @w{@uref{http://gcc.gnu.org/readings.html}}.
3222 This warning is enabled by @option{-Wall} for C and C++.
3225 @opindex Wreturn-type
3226 @opindex Wno-return-type
3227 Warn whenever a function is defined with a return-type that defaults
3228 to @code{int}. Also warn about any @code{return} statement with no
3229 return-value in a function whose return-type is not @code{void}
3230 (falling off the end of the function body is considered returning
3231 without a value), and about a @code{return} statement with an
3232 expression in a function whose return-type is @code{void}.
3234 For C++, a function without return type always produces a diagnostic
3235 message, even when @option{-Wno-return-type} is specified. The only
3236 exceptions are @samp{main} and functions defined in system headers.
3238 This warning is enabled by @option{-Wall}.
3243 Warn whenever a @code{switch} statement has an index of enumerated type
3244 and lacks a @code{case} for one or more of the named codes of that
3245 enumeration. (The presence of a @code{default} label prevents this
3246 warning.) @code{case} labels outside the enumeration range also
3247 provoke warnings when this option is used (even if there is a
3248 @code{default} label).
3249 This warning is enabled by @option{-Wall}.
3251 @item -Wswitch-default
3252 @opindex Wswitch-default
3253 @opindex Wno-switch-default
3254 Warn whenever a @code{switch} statement does not have a @code{default}
3258 @opindex Wswitch-enum
3259 @opindex Wno-switch-enum
3260 Warn whenever a @code{switch} statement has an index of enumerated type
3261 and lacks a @code{case} for one or more of the named codes of that
3262 enumeration. @code{case} labels outside the enumeration range also
3263 provoke warnings when this option is used. The only difference
3264 between @option{-Wswitch} and this option is that this option gives a
3265 warning about an omitted enumeration code even if there is a
3266 @code{default} label.
3268 @item -Wsync-nand @r{(C and C++ only)}
3270 @opindex Wno-sync-nand
3271 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3272 built-in functions are used. These functions changed semantics in GCC 4.4.
3276 @opindex Wno-trigraphs
3277 Warn if any trigraphs are encountered that might change the meaning of
3278 the program (trigraphs within comments are not warned about).
3279 This warning is enabled by @option{-Wall}.
3281 @item -Wunused-function
3282 @opindex Wunused-function
3283 @opindex Wno-unused-function
3284 Warn whenever a static function is declared but not defined or a
3285 non-inline static function is unused.
3286 This warning is enabled by @option{-Wall}.
3288 @item -Wunused-label
3289 @opindex Wunused-label
3290 @opindex Wno-unused-label
3291 Warn whenever a label is declared but not used.
3292 This warning is enabled by @option{-Wall}.
3294 To suppress this warning use the @samp{unused} attribute
3295 (@pxref{Variable Attributes}).
3297 @item -Wunused-parameter
3298 @opindex Wunused-parameter
3299 @opindex Wno-unused-parameter
3300 Warn whenever a function parameter is unused aside from its declaration.
3302 To suppress this warning use the @samp{unused} attribute
3303 (@pxref{Variable Attributes}).
3305 @item -Wno-unused-result
3306 @opindex Wunused-result
3307 @opindex Wno-unused-result
3308 Do not warn if a caller of a function marked with attribute
3309 @code{warn_unused_result} (@pxref{Variable Attributes}) does not use
3310 its return value. The default is @option{-Wunused-result}.
3312 @item -Wunused-variable
3313 @opindex Wunused-variable
3314 @opindex Wno-unused-variable
3315 Warn whenever a local variable or non-constant static variable is unused
3316 aside from its declaration.
3317 This warning is enabled by @option{-Wall}.
3319 To suppress this warning use the @samp{unused} attribute
3320 (@pxref{Variable Attributes}).
3322 @item -Wunused-value
3323 @opindex Wunused-value
3324 @opindex Wno-unused-value
3325 Warn whenever a statement computes a result that is explicitly not
3326 used. To suppress this warning cast the unused expression to
3327 @samp{void}. This includes an expression-statement or the left-hand
3328 side of a comma expression that contains no side effects. For example,
3329 an expression such as @samp{x[i,j]} will cause a warning, while
3330 @samp{x[(void)i,j]} will not.
3332 This warning is enabled by @option{-Wall}.
3337 All the above @option{-Wunused} options combined.
3339 In order to get a warning about an unused function parameter, you must
3340 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3341 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3343 @item -Wuninitialized
3344 @opindex Wuninitialized
3345 @opindex Wno-uninitialized
3346 Warn if an automatic variable is used without first being initialized
3347 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3348 warn if a non-static reference or non-static @samp{const} member
3349 appears in a class without constructors.
3351 If you want to warn about code which uses the uninitialized value of the
3352 variable in its own initializer, use the @option{-Winit-self} option.
3354 These warnings occur for individual uninitialized or clobbered
3355 elements of structure, union or array variables as well as for
3356 variables which are uninitialized or clobbered as a whole. They do
3357 not occur for variables or elements declared @code{volatile}. Because
3358 these warnings depend on optimization, the exact variables or elements
3359 for which there are warnings will depend on the precise optimization
3360 options and version of GCC used.
3362 Note that there may be no warning about a variable that is used only
3363 to compute a value that itself is never used, because such
3364 computations may be deleted by data flow analysis before the warnings
3367 These warnings are made optional because GCC is not smart
3368 enough to see all the reasons why the code might be correct
3369 despite appearing to have an error. Here is one example of how
3390 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3391 always initialized, but GCC doesn't know this. Here is
3392 another common case:
3397 if (change_y) save_y = y, y = new_y;
3399 if (change_y) y = save_y;
3404 This has no bug because @code{save_y} is used only if it is set.
3406 @cindex @code{longjmp} warnings
3407 This option also warns when a non-volatile automatic variable might be
3408 changed by a call to @code{longjmp}. These warnings as well are possible
3409 only in optimizing compilation.
3411 The compiler sees only the calls to @code{setjmp}. It cannot know
3412 where @code{longjmp} will be called; in fact, a signal handler could
3413 call it at any point in the code. As a result, you may get a warning
3414 even when there is in fact no problem because @code{longjmp} cannot
3415 in fact be called at the place which would cause a problem.
3417 Some spurious warnings can be avoided if you declare all the functions
3418 you use that never return as @code{noreturn}. @xref{Function
3421 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3423 @item -Wunknown-pragmas
3424 @opindex Wunknown-pragmas
3425 @opindex Wno-unknown-pragmas
3426 @cindex warning for unknown pragmas
3427 @cindex unknown pragmas, warning
3428 @cindex pragmas, warning of unknown
3429 Warn when a #pragma directive is encountered which is not understood by
3430 GCC@. If this command line option is used, warnings will even be issued
3431 for unknown pragmas in system header files. This is not the case if
3432 the warnings were only enabled by the @option{-Wall} command line option.
3435 @opindex Wno-pragmas
3437 Do not warn about misuses of pragmas, such as incorrect parameters,
3438 invalid syntax, or conflicts between pragmas. See also
3439 @samp{-Wunknown-pragmas}.
3441 @item -Wstrict-aliasing
3442 @opindex Wstrict-aliasing
3443 @opindex Wno-strict-aliasing
3444 This option is only active when @option{-fstrict-aliasing} is active.
3445 It warns about code which might break the strict aliasing rules that the
3446 compiler is using for optimization. The warning does not catch all
3447 cases, but does attempt to catch the more common pitfalls. It is
3448 included in @option{-Wall}.
3449 It is equivalent to @option{-Wstrict-aliasing=3}
3451 @item -Wstrict-aliasing=n
3452 @opindex Wstrict-aliasing=n
3453 @opindex Wno-strict-aliasing=n
3454 This option is only active when @option{-fstrict-aliasing} is active.
3455 It warns about code which might break the strict aliasing rules that the
3456 compiler is using for optimization.
3457 Higher levels correspond to higher accuracy (fewer false positives).
3458 Higher levels also correspond to more effort, similar to the way -O works.
3459 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3462 Level 1: Most aggressive, quick, least accurate.
3463 Possibly useful when higher levels
3464 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3465 false negatives. However, it has many false positives.
3466 Warns for all pointer conversions between possibly incompatible types,
3467 even if never dereferenced. Runs in the frontend only.
3469 Level 2: Aggressive, quick, not too precise.
3470 May still have many false positives (not as many as level 1 though),
3471 and few false negatives (but possibly more than level 1).
3472 Unlike level 1, it only warns when an address is taken. Warns about
3473 incomplete types. Runs in the frontend only.
3475 Level 3 (default for @option{-Wstrict-aliasing}):
3476 Should have very few false positives and few false
3477 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3478 Takes care of the common punn+dereference pattern in the frontend:
3479 @code{*(int*)&some_float}.
3480 If optimization is enabled, it also runs in the backend, where it deals
3481 with multiple statement cases using flow-sensitive points-to information.
3482 Only warns when the converted pointer is dereferenced.
3483 Does not warn about incomplete types.
3485 @item -Wstrict-overflow
3486 @itemx -Wstrict-overflow=@var{n}
3487 @opindex Wstrict-overflow
3488 @opindex Wno-strict-overflow
3489 This option is only active when @option{-fstrict-overflow} is active.
3490 It warns about cases where the compiler optimizes based on the
3491 assumption that signed overflow does not occur. Note that it does not
3492 warn about all cases where the code might overflow: it only warns
3493 about cases where the compiler implements some optimization. Thus
3494 this warning depends on the optimization level.
3496 An optimization which assumes that signed overflow does not occur is
3497 perfectly safe if the values of the variables involved are such that
3498 overflow never does, in fact, occur. Therefore this warning can
3499 easily give a false positive: a warning about code which is not
3500 actually a problem. To help focus on important issues, several
3501 warning levels are defined. No warnings are issued for the use of
3502 undefined signed overflow when estimating how many iterations a loop
3503 will require, in particular when determining whether a loop will be
3507 @item -Wstrict-overflow=1
3508 Warn about cases which are both questionable and easy to avoid. For
3509 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3510 compiler will simplify this to @code{1}. This level of
3511 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3512 are not, and must be explicitly requested.
3514 @item -Wstrict-overflow=2
3515 Also warn about other cases where a comparison is simplified to a
3516 constant. For example: @code{abs (x) >= 0}. This can only be
3517 simplified when @option{-fstrict-overflow} is in effect, because
3518 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3519 zero. @option{-Wstrict-overflow} (with no level) is the same as
3520 @option{-Wstrict-overflow=2}.
3522 @item -Wstrict-overflow=3
3523 Also warn about other cases where a comparison is simplified. For
3524 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3526 @item -Wstrict-overflow=4
3527 Also warn about other simplifications not covered by the above cases.
3528 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3530 @item -Wstrict-overflow=5
3531 Also warn about cases where the compiler reduces the magnitude of a
3532 constant involved in a comparison. For example: @code{x + 2 > y} will
3533 be simplified to @code{x + 1 >= y}. This is reported only at the
3534 highest warning level because this simplification applies to many
3535 comparisons, so this warning level will give a very large number of
3539 @item -Warray-bounds
3540 @opindex Wno-array-bounds
3541 @opindex Warray-bounds
3542 This option is only active when @option{-ftree-vrp} is active
3543 (default for -O2 and above). It warns about subscripts to arrays
3544 that are always out of bounds. This warning is enabled by @option{-Wall}.
3546 @item -Wno-div-by-zero
3547 @opindex Wno-div-by-zero
3548 @opindex Wdiv-by-zero
3549 Do not warn about compile-time integer division by zero. Floating point
3550 division by zero is not warned about, as it can be a legitimate way of
3551 obtaining infinities and NaNs.
3553 @item -Wsystem-headers
3554 @opindex Wsystem-headers
3555 @opindex Wno-system-headers
3556 @cindex warnings from system headers
3557 @cindex system headers, warnings from
3558 Print warning messages for constructs found in system header files.
3559 Warnings from system headers are normally suppressed, on the assumption
3560 that they usually do not indicate real problems and would only make the
3561 compiler output harder to read. Using this command line option tells
3562 GCC to emit warnings from system headers as if they occurred in user
3563 code. However, note that using @option{-Wall} in conjunction with this
3564 option will @emph{not} warn about unknown pragmas in system
3565 headers---for that, @option{-Wunknown-pragmas} must also be used.
3568 @opindex Wfloat-equal
3569 @opindex Wno-float-equal
3570 Warn if floating point values are used in equality comparisons.
3572 The idea behind this is that sometimes it is convenient (for the
3573 programmer) to consider floating-point values as approximations to
3574 infinitely precise real numbers. If you are doing this, then you need
3575 to compute (by analyzing the code, or in some other way) the maximum or
3576 likely maximum error that the computation introduces, and allow for it
3577 when performing comparisons (and when producing output, but that's a
3578 different problem). In particular, instead of testing for equality, you
3579 would check to see whether the two values have ranges that overlap; and
3580 this is done with the relational operators, so equality comparisons are
3583 @item -Wtraditional @r{(C and Objective-C only)}
3584 @opindex Wtraditional
3585 @opindex Wno-traditional
3586 Warn about certain constructs that behave differently in traditional and
3587 ISO C@. Also warn about ISO C constructs that have no traditional C
3588 equivalent, and/or problematic constructs which should be avoided.
3592 Macro parameters that appear within string literals in the macro body.
3593 In traditional C macro replacement takes place within string literals,
3594 but does not in ISO C@.
3597 In traditional C, some preprocessor directives did not exist.
3598 Traditional preprocessors would only consider a line to be a directive
3599 if the @samp{#} appeared in column 1 on the line. Therefore
3600 @option{-Wtraditional} warns about directives that traditional C
3601 understands but would ignore because the @samp{#} does not appear as the
3602 first character on the line. It also suggests you hide directives like
3603 @samp{#pragma} not understood by traditional C by indenting them. Some
3604 traditional implementations would not recognize @samp{#elif}, so it
3605 suggests avoiding it altogether.
3608 A function-like macro that appears without arguments.
3611 The unary plus operator.
3614 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3615 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3616 constants.) Note, these suffixes appear in macros defined in the system
3617 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3618 Use of these macros in user code might normally lead to spurious
3619 warnings, however GCC's integrated preprocessor has enough context to
3620 avoid warning in these cases.
3623 A function declared external in one block and then used after the end of
3627 A @code{switch} statement has an operand of type @code{long}.
3630 A non-@code{static} function declaration follows a @code{static} one.
3631 This construct is not accepted by some traditional C compilers.
3634 The ISO type of an integer constant has a different width or
3635 signedness from its traditional type. This warning is only issued if
3636 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3637 typically represent bit patterns, are not warned about.
3640 Usage of ISO string concatenation is detected.
3643 Initialization of automatic aggregates.
3646 Identifier conflicts with labels. Traditional C lacks a separate
3647 namespace for labels.
3650 Initialization of unions. If the initializer is zero, the warning is
3651 omitted. This is done under the assumption that the zero initializer in
3652 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3653 initializer warnings and relies on default initialization to zero in the
3657 Conversions by prototypes between fixed/floating point values and vice
3658 versa. The absence of these prototypes when compiling with traditional
3659 C would cause serious problems. This is a subset of the possible
3660 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3663 Use of ISO C style function definitions. This warning intentionally is
3664 @emph{not} issued for prototype declarations or variadic functions
3665 because these ISO C features will appear in your code when using
3666 libiberty's traditional C compatibility macros, @code{PARAMS} and
3667 @code{VPARAMS}. This warning is also bypassed for nested functions
3668 because that feature is already a GCC extension and thus not relevant to
3669 traditional C compatibility.
3672 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3673 @opindex Wtraditional-conversion
3674 @opindex Wno-traditional-conversion
3675 Warn if a prototype causes a type conversion that is different from what
3676 would happen to the same argument in the absence of a prototype. This
3677 includes conversions of fixed point to floating and vice versa, and
3678 conversions changing the width or signedness of a fixed point argument
3679 except when the same as the default promotion.
3681 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3682 @opindex Wdeclaration-after-statement
3683 @opindex Wno-declaration-after-statement
3684 Warn when a declaration is found after a statement in a block. This
3685 construct, known from C++, was introduced with ISO C99 and is by default
3686 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3687 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3692 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3694 @item -Wno-endif-labels
3695 @opindex Wno-endif-labels
3696 @opindex Wendif-labels
3697 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3702 Warn whenever a local variable shadows another local variable, parameter or
3703 global variable or whenever a built-in function is shadowed.
3705 @item -Wlarger-than=@var{len}
3706 @opindex Wlarger-than=@var{len}
3707 @opindex Wlarger-than-@var{len}
3708 Warn whenever an object of larger than @var{len} bytes is defined.
3710 @item -Wframe-larger-than=@var{len}
3711 @opindex Wframe-larger-than
3712 Warn if the size of a function frame is larger than @var{len} bytes.
3713 The computation done to determine the stack frame size is approximate
3714 and not conservative.
3715 The actual requirements may be somewhat greater than @var{len}
3716 even if you do not get a warning. In addition, any space allocated
3717 via @code{alloca}, variable-length arrays, or related constructs
3718 is not included by the compiler when determining
3719 whether or not to issue a warning.
3721 @item -Wunsafe-loop-optimizations
3722 @opindex Wunsafe-loop-optimizations
3723 @opindex Wno-unsafe-loop-optimizations
3724 Warn if the loop cannot be optimized because the compiler could not
3725 assume anything on the bounds of the loop indices. With
3726 @option{-funsafe-loop-optimizations} warn if the compiler made
3729 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
3730 @opindex Wno-pedantic-ms-format
3731 @opindex Wpedantic-ms-format
3732 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
3733 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
3734 depending on the MS runtime, when you are using the options @option{-Wformat}
3735 and @option{-pedantic} without gnu-extensions.
3737 @item -Wpointer-arith
3738 @opindex Wpointer-arith
3739 @opindex Wno-pointer-arith
3740 Warn about anything that depends on the ``size of'' a function type or
3741 of @code{void}. GNU C assigns these types a size of 1, for
3742 convenience in calculations with @code{void *} pointers and pointers
3743 to functions. In C++, warn also when an arithmetic operation involves
3744 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3747 @opindex Wtype-limits
3748 @opindex Wno-type-limits
3749 Warn if a comparison is always true or always false due to the limited
3750 range of the data type, but do not warn for constant expressions. For
3751 example, warn if an unsigned variable is compared against zero with
3752 @samp{<} or @samp{>=}. This warning is also enabled by
3755 @item -Wbad-function-cast @r{(C and Objective-C only)}
3756 @opindex Wbad-function-cast
3757 @opindex Wno-bad-function-cast
3758 Warn whenever a function call is cast to a non-matching type.
3759 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3761 @item -Wc++-compat @r{(C and Objective-C only)}
3762 Warn about ISO C constructs that are outside of the common subset of
3763 ISO C and ISO C++, e.g.@: request for implicit conversion from
3764 @code{void *} to a pointer to non-@code{void} type.
3766 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3767 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3768 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3769 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3773 @opindex Wno-cast-qual
3774 Warn whenever a pointer is cast so as to remove a type qualifier from
3775 the target type. For example, warn if a @code{const char *} is cast
3776 to an ordinary @code{char *}.
3778 Also warn when making a cast which introduces a type qualifier in an
3779 unsafe way. For example, casting @code{char **} to @code{const char **}
3780 is unsafe, as in this example:
3783 /* p is char ** value. */
3784 const char **q = (const char **) p;
3785 /* Assignment of readonly string to const char * is OK. */
3787 /* Now char** pointer points to read-only memory. */
3792 @opindex Wcast-align
3793 @opindex Wno-cast-align
3794 Warn whenever a pointer is cast such that the required alignment of the
3795 target is increased. For example, warn if a @code{char *} is cast to
3796 an @code{int *} on machines where integers can only be accessed at
3797 two- or four-byte boundaries.
3799 @item -Wwrite-strings
3800 @opindex Wwrite-strings
3801 @opindex Wno-write-strings
3802 When compiling C, give string constants the type @code{const
3803 char[@var{length}]} so that copying the address of one into a
3804 non-@code{const} @code{char *} pointer will get a warning. These
3805 warnings will help you find at compile time code that can try to write
3806 into a string constant, but only if you have been very careful about
3807 using @code{const} in declarations and prototypes. Otherwise, it will
3808 just be a nuisance. This is why we did not make @option{-Wall} request
3811 When compiling C++, warn about the deprecated conversion from string
3812 literals to @code{char *}. This warning is enabled by default for C++
3817 @opindex Wno-clobbered
3818 Warn for variables that might be changed by @samp{longjmp} or
3819 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
3822 @opindex Wconversion
3823 @opindex Wno-conversion
3824 Warn for implicit conversions that may alter a value. This includes
3825 conversions between real and integer, like @code{abs (x)} when
3826 @code{x} is @code{double}; conversions between signed and unsigned,
3827 like @code{unsigned ui = -1}; and conversions to smaller types, like
3828 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
3829 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
3830 changed by the conversion like in @code{abs (2.0)}. Warnings about
3831 conversions between signed and unsigned integers can be disabled by
3832 using @option{-Wno-sign-conversion}.
3834 For C++, also warn for conversions between @code{NULL} and non-pointer
3835 types; confusing overload resolution for user-defined conversions; and
3836 conversions that will never use a type conversion operator:
3837 conversions to @code{void}, the same type, a base class or a reference
3838 to them. Warnings about conversions between signed and unsigned
3839 integers are disabled by default in C++ unless
3840 @option{-Wsign-conversion} is explicitly enabled.
3843 @opindex Wempty-body
3844 @opindex Wno-empty-body
3845 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
3846 while} statement. This warning is also enabled by @option{-Wextra}.
3848 @item -Wenum-compare
3849 @opindex Wenum-compare
3850 @opindex Wno-enum-compare
3851 Warn about a comparison between values of different enum types. In C++
3852 this warning is enabled by default. In C this warning is enabled by
3855 @item -Wjump-misses-init @r{(C, Objective-C only)}
3856 @opindex Wjump-misses-init
3857 @opindex Wno-jump-misses-init
3858 Warn if a @code{goto} statement or a @code{switch} statement jumps
3859 forward across the initialization of a variable, or jumps backward to a
3860 label after the variable has been initialized. This only warns about
3861 variables which are initialized when they are declared. This warning is
3862 only supported for C and Objective C; in C++ this sort of branch is an
3865 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
3866 can be disabled with the @option{-Wno-jump-misses-init} option.
3868 @item -Wsign-compare
3869 @opindex Wsign-compare
3870 @opindex Wno-sign-compare
3871 @cindex warning for comparison of signed and unsigned values
3872 @cindex comparison of signed and unsigned values, warning
3873 @cindex signed and unsigned values, comparison warning
3874 Warn when a comparison between signed and unsigned values could produce
3875 an incorrect result when the signed value is converted to unsigned.
3876 This warning is also enabled by @option{-Wextra}; to get the other warnings
3877 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3879 @item -Wsign-conversion
3880 @opindex Wsign-conversion
3881 @opindex Wno-sign-conversion
3882 Warn for implicit conversions that may change the sign of an integer
3883 value, like assigning a signed integer expression to an unsigned
3884 integer variable. An explicit cast silences the warning. In C, this
3885 option is enabled also by @option{-Wconversion}.
3889 @opindex Wno-address
3890 Warn about suspicious uses of memory addresses. These include using
3891 the address of a function in a conditional expression, such as
3892 @code{void func(void); if (func)}, and comparisons against the memory
3893 address of a string literal, such as @code{if (x == "abc")}. Such
3894 uses typically indicate a programmer error: the address of a function
3895 always evaluates to true, so their use in a conditional usually
3896 indicate that the programmer forgot the parentheses in a function
3897 call; and comparisons against string literals result in unspecified
3898 behavior and are not portable in C, so they usually indicate that the
3899 programmer intended to use @code{strcmp}. This warning is enabled by
3903 @opindex Wlogical-op
3904 @opindex Wno-logical-op
3905 Warn about suspicious uses of logical operators in expressions.
3906 This includes using logical operators in contexts where a
3907 bit-wise operator is likely to be expected.
3909 @item -Waggregate-return
3910 @opindex Waggregate-return
3911 @opindex Wno-aggregate-return
3912 Warn if any functions that return structures or unions are defined or
3913 called. (In languages where you can return an array, this also elicits
3916 @item -Wno-attributes
3917 @opindex Wno-attributes
3918 @opindex Wattributes
3919 Do not warn if an unexpected @code{__attribute__} is used, such as
3920 unrecognized attributes, function attributes applied to variables,
3921 etc. This will not stop errors for incorrect use of supported
3924 @item -Wno-builtin-macro-redefined
3925 @opindex Wno-builtin-macro-redefined
3926 @opindex Wbuiltin-macro-redefined
3927 Do not warn if certain built-in macros are redefined. This suppresses
3928 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
3929 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
3931 @item -Wstrict-prototypes @r{(C and Objective-C only)}
3932 @opindex Wstrict-prototypes
3933 @opindex Wno-strict-prototypes
3934 Warn if a function is declared or defined without specifying the
3935 argument types. (An old-style function definition is permitted without
3936 a warning if preceded by a declaration which specifies the argument
3939 @item -Wold-style-declaration @r{(C and Objective-C only)}
3940 @opindex Wold-style-declaration
3941 @opindex Wno-old-style-declaration
3942 Warn for obsolescent usages, according to the C Standard, in a
3943 declaration. For example, warn if storage-class specifiers like
3944 @code{static} are not the first things in a declaration. This warning
3945 is also enabled by @option{-Wextra}.
3947 @item -Wold-style-definition @r{(C and Objective-C only)}
3948 @opindex Wold-style-definition
3949 @opindex Wno-old-style-definition
3950 Warn if an old-style function definition is used. A warning is given
3951 even if there is a previous prototype.
3953 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
3954 @opindex Wmissing-parameter-type
3955 @opindex Wno-missing-parameter-type
3956 A function parameter is declared without a type specifier in K&R-style
3963 This warning is also enabled by @option{-Wextra}.
3965 @item -Wmissing-prototypes @r{(C and Objective-C only)}
3966 @opindex Wmissing-prototypes
3967 @opindex Wno-missing-prototypes
3968 Warn if a global function is defined without a previous prototype
3969 declaration. This warning is issued even if the definition itself
3970 provides a prototype. The aim is to detect global functions that fail
3971 to be declared in header files.
3973 @item -Wmissing-declarations
3974 @opindex Wmissing-declarations
3975 @opindex Wno-missing-declarations
3976 Warn if a global function is defined without a previous declaration.
3977 Do so even if the definition itself provides a prototype.
3978 Use this option to detect global functions that are not declared in
3979 header files. In C++, no warnings are issued for function templates,
3980 or for inline functions, or for functions in anonymous namespaces.
3982 @item -Wmissing-field-initializers
3983 @opindex Wmissing-field-initializers
3984 @opindex Wno-missing-field-initializers
3988 Warn if a structure's initializer has some fields missing. For
3989 example, the following code would cause such a warning, because
3990 @code{x.h} is implicitly zero:
3993 struct s @{ int f, g, h; @};
3994 struct s x = @{ 3, 4 @};
3997 This option does not warn about designated initializers, so the following
3998 modification would not trigger a warning:
4001 struct s @{ int f, g, h; @};
4002 struct s x = @{ .f = 3, .g = 4 @};
4005 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
4006 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
4008 @item -Wmissing-noreturn
4009 @opindex Wmissing-noreturn
4010 @opindex Wno-missing-noreturn
4011 Warn about functions which might be candidates for attribute @code{noreturn}.
4012 Note these are only possible candidates, not absolute ones. Care should
4013 be taken to manually verify functions actually do not ever return before
4014 adding the @code{noreturn} attribute, otherwise subtle code generation
4015 bugs could be introduced. You will not get a warning for @code{main} in
4016 hosted C environments.
4018 @item -Wmissing-format-attribute
4019 @opindex Wmissing-format-attribute
4020 @opindex Wno-missing-format-attribute
4023 Warn about function pointers which might be candidates for @code{format}
4024 attributes. Note these are only possible candidates, not absolute ones.
4025 GCC will guess that function pointers with @code{format} attributes that
4026 are used in assignment, initialization, parameter passing or return
4027 statements should have a corresponding @code{format} attribute in the
4028 resulting type. I.e.@: the left-hand side of the assignment or
4029 initialization, the type of the parameter variable, or the return type
4030 of the containing function respectively should also have a @code{format}
4031 attribute to avoid the warning.
4033 GCC will also warn about function definitions which might be
4034 candidates for @code{format} attributes. Again, these are only
4035 possible candidates. GCC will guess that @code{format} attributes
4036 might be appropriate for any function that calls a function like
4037 @code{vprintf} or @code{vscanf}, but this might not always be the
4038 case, and some functions for which @code{format} attributes are
4039 appropriate may not be detected.
4041 @item -Wno-multichar
4042 @opindex Wno-multichar
4044 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4045 Usually they indicate a typo in the user's code, as they have
4046 implementation-defined values, and should not be used in portable code.
4048 @item -Wnormalized=<none|id|nfc|nfkc>
4049 @opindex Wnormalized=
4052 @cindex character set, input normalization
4053 In ISO C and ISO C++, two identifiers are different if they are
4054 different sequences of characters. However, sometimes when characters
4055 outside the basic ASCII character set are used, you can have two
4056 different character sequences that look the same. To avoid confusion,
4057 the ISO 10646 standard sets out some @dfn{normalization rules} which
4058 when applied ensure that two sequences that look the same are turned into
4059 the same sequence. GCC can warn you if you are using identifiers which
4060 have not been normalized; this option controls that warning.
4062 There are four levels of warning that GCC supports. The default is
4063 @option{-Wnormalized=nfc}, which warns about any identifier which is
4064 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4065 recommended form for most uses.
4067 Unfortunately, there are some characters which ISO C and ISO C++ allow
4068 in identifiers that when turned into NFC aren't allowable as
4069 identifiers. That is, there's no way to use these symbols in portable
4070 ISO C or C++ and have all your identifiers in NFC@.
4071 @option{-Wnormalized=id} suppresses the warning for these characters.
4072 It is hoped that future versions of the standards involved will correct
4073 this, which is why this option is not the default.
4075 You can switch the warning off for all characters by writing
4076 @option{-Wnormalized=none}. You would only want to do this if you
4077 were using some other normalization scheme (like ``D''), because
4078 otherwise you can easily create bugs that are literally impossible to see.
4080 Some characters in ISO 10646 have distinct meanings but look identical
4081 in some fonts or display methodologies, especially once formatting has
4082 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4083 LETTER N'', will display just like a regular @code{n} which has been
4084 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4085 normalization scheme to convert all these into a standard form as
4086 well, and GCC will warn if your code is not in NFKC if you use
4087 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4088 about every identifier that contains the letter O because it might be
4089 confused with the digit 0, and so is not the default, but may be
4090 useful as a local coding convention if the programming environment is
4091 unable to be fixed to display these characters distinctly.
4093 @item -Wno-deprecated
4094 @opindex Wno-deprecated
4095 @opindex Wdeprecated
4096 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4098 @item -Wno-deprecated-declarations
4099 @opindex Wno-deprecated-declarations
4100 @opindex Wdeprecated-declarations
4101 Do not warn about uses of functions (@pxref{Function Attributes}),
4102 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4103 Attributes}) marked as deprecated by using the @code{deprecated}
4107 @opindex Wno-overflow
4109 Do not warn about compile-time overflow in constant expressions.
4111 @item -Woverride-init @r{(C and Objective-C only)}
4112 @opindex Woverride-init
4113 @opindex Wno-override-init
4117 Warn if an initialized field without side effects is overridden when
4118 using designated initializers (@pxref{Designated Inits, , Designated
4121 This warning is included in @option{-Wextra}. To get other
4122 @option{-Wextra} warnings without this one, use @samp{-Wextra
4123 -Wno-override-init}.
4128 Warn if a structure is given the packed attribute, but the packed
4129 attribute has no effect on the layout or size of the structure.
4130 Such structures may be mis-aligned for little benefit. For
4131 instance, in this code, the variable @code{f.x} in @code{struct bar}
4132 will be misaligned even though @code{struct bar} does not itself
4133 have the packed attribute:
4140 @} __attribute__((packed));
4148 @item -Wpacked-bitfield-compat
4149 @opindex Wpacked-bitfield-compat
4150 @opindex Wno-packed-bitfield-compat
4151 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4152 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4153 the change can lead to differences in the structure layout. GCC
4154 informs you when the offset of such a field has changed in GCC 4.4.
4155 For example there is no longer a 4-bit padding between field @code{a}
4156 and @code{b} in this structure:
4163 @} __attribute__ ((packed));
4166 This warning is enabled by default. Use
4167 @option{-Wno-packed-bitfield-compat} to disable this warning.
4172 Warn if padding is included in a structure, either to align an element
4173 of the structure or to align the whole structure. Sometimes when this
4174 happens it is possible to rearrange the fields of the structure to
4175 reduce the padding and so make the structure smaller.
4177 @item -Wredundant-decls
4178 @opindex Wredundant-decls
4179 @opindex Wno-redundant-decls
4180 Warn if anything is declared more than once in the same scope, even in
4181 cases where multiple declaration is valid and changes nothing.
4183 @item -Wnested-externs @r{(C and Objective-C only)}
4184 @opindex Wnested-externs
4185 @opindex Wno-nested-externs
4186 Warn if an @code{extern} declaration is encountered within a function.
4188 @item -Wunreachable-code
4189 @opindex Wunreachable-code
4190 @opindex Wno-unreachable-code
4191 Warn if the compiler detects that code will never be executed.
4193 This option is intended to warn when the compiler detects that at
4194 least a whole line of source code will never be executed, because
4195 some condition is never satisfied or because it is after a
4196 procedure that never returns.
4198 It is possible for this option to produce a warning even though there
4199 are circumstances under which part of the affected line can be executed,
4200 so care should be taken when removing apparently-unreachable code.
4202 For instance, when a function is inlined, a warning may mean that the
4203 line is unreachable in only one inlined copy of the function.
4205 This option is not made part of @option{-Wall} because in a debugging
4206 version of a program there is often substantial code which checks
4207 correct functioning of the program and is, hopefully, unreachable
4208 because the program does work. Another common use of unreachable
4209 code is to provide behavior which is selectable at compile-time.
4214 Warn if a function can not be inlined and it was declared as inline.
4215 Even with this option, the compiler will not warn about failures to
4216 inline functions declared in system headers.
4218 The compiler uses a variety of heuristics to determine whether or not
4219 to inline a function. For example, the compiler takes into account
4220 the size of the function being inlined and the amount of inlining
4221 that has already been done in the current function. Therefore,
4222 seemingly insignificant changes in the source program can cause the
4223 warnings produced by @option{-Winline} to appear or disappear.
4225 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4226 @opindex Wno-invalid-offsetof
4227 @opindex Winvalid-offsetof
4228 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4229 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4230 to a non-POD type is undefined. In existing C++ implementations,
4231 however, @samp{offsetof} typically gives meaningful results even when
4232 applied to certain kinds of non-POD types. (Such as a simple
4233 @samp{struct} that fails to be a POD type only by virtue of having a
4234 constructor.) This flag is for users who are aware that they are
4235 writing nonportable code and who have deliberately chosen to ignore the
4238 The restrictions on @samp{offsetof} may be relaxed in a future version
4239 of the C++ standard.
4241 @item -Wno-int-to-pointer-cast @r{(C and Objective-C only)}
4242 @opindex Wno-int-to-pointer-cast
4243 @opindex Wint-to-pointer-cast
4244 Suppress warnings from casts to pointer type of an integer of a
4247 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4248 @opindex Wno-pointer-to-int-cast
4249 @opindex Wpointer-to-int-cast
4250 Suppress warnings from casts from a pointer to an integer type of a
4254 @opindex Winvalid-pch
4255 @opindex Wno-invalid-pch
4256 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4257 the search path but can't be used.
4261 @opindex Wno-long-long
4262 Warn if @samp{long long} type is used. This is enabled by either
4263 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4264 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4266 @item -Wvariadic-macros
4267 @opindex Wvariadic-macros
4268 @opindex Wno-variadic-macros
4269 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4270 alternate syntax when in pedantic ISO C99 mode. This is default.
4271 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4276 Warn if variable length array is used in the code.
4277 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4278 the variable length array.
4280 @item -Wvolatile-register-var
4281 @opindex Wvolatile-register-var
4282 @opindex Wno-volatile-register-var
4283 Warn if a register variable is declared volatile. The volatile
4284 modifier does not inhibit all optimizations that may eliminate reads
4285 and/or writes to register variables. This warning is enabled by
4288 @item -Wdisabled-optimization
4289 @opindex Wdisabled-optimization
4290 @opindex Wno-disabled-optimization
4291 Warn if a requested optimization pass is disabled. This warning does
4292 not generally indicate that there is anything wrong with your code; it
4293 merely indicates that GCC's optimizers were unable to handle the code
4294 effectively. Often, the problem is that your code is too big or too
4295 complex; GCC will refuse to optimize programs when the optimization
4296 itself is likely to take inordinate amounts of time.
4298 @item -Wpointer-sign @r{(C and Objective-C only)}
4299 @opindex Wpointer-sign
4300 @opindex Wno-pointer-sign
4301 Warn for pointer argument passing or assignment with different signedness.
4302 This option is only supported for C and Objective-C@. It is implied by
4303 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4304 @option{-Wno-pointer-sign}.
4306 @item -Wstack-protector
4307 @opindex Wstack-protector
4308 @opindex Wno-stack-protector
4309 This option is only active when @option{-fstack-protector} is active. It
4310 warns about functions that will not be protected against stack smashing.
4313 @opindex Wno-mudflap
4314 Suppress warnings about constructs that cannot be instrumented by
4317 @item -Woverlength-strings
4318 @opindex Woverlength-strings
4319 @opindex Wno-overlength-strings
4320 Warn about string constants which are longer than the ``minimum
4321 maximum'' length specified in the C standard. Modern compilers
4322 generally allow string constants which are much longer than the
4323 standard's minimum limit, but very portable programs should avoid
4324 using longer strings.
4326 The limit applies @emph{after} string constant concatenation, and does
4327 not count the trailing NUL@. In C89, the limit was 509 characters; in
4328 C99, it was raised to 4095. C++98 does not specify a normative
4329 minimum maximum, so we do not diagnose overlength strings in C++@.
4331 This option is implied by @option{-pedantic}, and can be disabled with
4332 @option{-Wno-overlength-strings}.
4334 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
4335 @opindex Wunsuffixed-float-constants
4337 GCC will issue a warning for any floating constant that does not have
4338 a suffix. When used together with @option{-Wsystem-headers} it will
4339 warn about such constants in system header files. This can be useful
4340 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4341 from the decimal floating-point extension to C99.
4344 @node Debugging Options
4345 @section Options for Debugging Your Program or GCC
4346 @cindex options, debugging
4347 @cindex debugging information options
4349 GCC has various special options that are used for debugging
4350 either your program or GCC:
4355 Produce debugging information in the operating system's native format
4356 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4359 On most systems that use stabs format, @option{-g} enables use of extra
4360 debugging information that only GDB can use; this extra information
4361 makes debugging work better in GDB but will probably make other debuggers
4363 refuse to read the program. If you want to control for certain whether
4364 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4365 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4367 GCC allows you to use @option{-g} with
4368 @option{-O}. The shortcuts taken by optimized code may occasionally
4369 produce surprising results: some variables you declared may not exist
4370 at all; flow of control may briefly move where you did not expect it;
4371 some statements may not be executed because they compute constant
4372 results or their values were already at hand; some statements may
4373 execute in different places because they were moved out of loops.
4375 Nevertheless it proves possible to debug optimized output. This makes
4376 it reasonable to use the optimizer for programs that might have bugs.
4378 The following options are useful when GCC is generated with the
4379 capability for more than one debugging format.
4383 Produce debugging information for use by GDB@. This means to use the
4384 most expressive format available (DWARF 2, stabs, or the native format
4385 if neither of those are supported), including GDB extensions if at all
4390 Produce debugging information in stabs format (if that is supported),
4391 without GDB extensions. This is the format used by DBX on most BSD
4392 systems. On MIPS, Alpha and System V Release 4 systems this option
4393 produces stabs debugging output which is not understood by DBX or SDB@.
4394 On System V Release 4 systems this option requires the GNU assembler.
4396 @item -feliminate-unused-debug-symbols
4397 @opindex feliminate-unused-debug-symbols
4398 Produce debugging information in stabs format (if that is supported),
4399 for only symbols that are actually used.
4401 @item -femit-class-debug-always
4402 Instead of emitting debugging information for a C++ class in only one
4403 object file, emit it in all object files using the class. This option
4404 should be used only with debuggers that are unable to handle the way GCC
4405 normally emits debugging information for classes because using this
4406 option will increase the size of debugging information by as much as a
4411 Produce debugging information in stabs format (if that is supported),
4412 using GNU extensions understood only by the GNU debugger (GDB)@. The
4413 use of these extensions is likely to make other debuggers crash or
4414 refuse to read the program.
4418 Produce debugging information in COFF format (if that is supported).
4419 This is the format used by SDB on most System V systems prior to
4424 Produce debugging information in XCOFF format (if that is supported).
4425 This is the format used by the DBX debugger on IBM RS/6000 systems.
4429 Produce debugging information in XCOFF format (if that is supported),
4430 using GNU extensions understood only by the GNU debugger (GDB)@. The
4431 use of these extensions is likely to make other debuggers crash or
4432 refuse to read the program, and may cause assemblers other than the GNU
4433 assembler (GAS) to fail with an error.
4435 @item -gdwarf-@var{version}
4436 @opindex gdwarf-@var{version}
4437 Produce debugging information in DWARF format (if that is
4438 supported). This is the format used by DBX on IRIX 6. The value
4439 of @var{version} may be either 2, 3 or 4; the default version is 2.
4441 Note that with DWARF version 2 some ports require, and will always
4442 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4444 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
4445 for maximum benefit.
4447 @item -gstrict-dwarf
4448 @opindex gstrict-dwarf
4449 Disallow using extensions of later DWARF standard version than selected
4450 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
4451 DWARF extensions from later standard versions is allowed.
4453 @item -gno-strict-dwarf
4454 @opindex gno-strict-dwarf
4455 Allow using extensions of later DWARF standard version than selected with
4456 @option{-gdwarf-@var{version}}.
4460 Produce debugging information in VMS debug format (if that is
4461 supported). This is the format used by DEBUG on VMS systems.
4464 @itemx -ggdb@var{level}
4465 @itemx -gstabs@var{level}
4466 @itemx -gcoff@var{level}
4467 @itemx -gxcoff@var{level}
4468 @itemx -gvms@var{level}
4469 Request debugging information and also use @var{level} to specify how
4470 much information. The default level is 2.
4472 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4475 Level 1 produces minimal information, enough for making backtraces in
4476 parts of the program that you don't plan to debug. This includes
4477 descriptions of functions and external variables, but no information
4478 about local variables and no line numbers.
4480 Level 3 includes extra information, such as all the macro definitions
4481 present in the program. Some debuggers support macro expansion when
4482 you use @option{-g3}.
4484 @option{-gdwarf-2} does not accept a concatenated debug level, because
4485 GCC used to support an option @option{-gdwarf} that meant to generate
4486 debug information in version 1 of the DWARF format (which is very
4487 different from version 2), and it would have been too confusing. That
4488 debug format is long obsolete, but the option cannot be changed now.
4489 Instead use an additional @option{-g@var{level}} option to change the
4490 debug level for DWARF.
4494 Turn off generation of debug info, if leaving out this option would have
4495 generated it, or turn it on at level 2 otherwise. The position of this
4496 argument in the command line does not matter, it takes effect after all
4497 other options are processed, and it does so only once, no matter how
4498 many times it is given. This is mainly intended to be used with
4499 @option{-fcompare-debug}.
4501 @item -fdump-final-insns@r{[}=@var{file}@r{]}
4502 @opindex fdump-final-insns
4503 Dump the final internal representation (RTL) to @var{file}. If the
4504 optional argument is omitted (or if @var{file} is @code{.}), the name
4505 of the dump file will be determined by appending @code{.gkd} to the
4506 compilation output file name.
4508 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4509 @opindex fcompare-debug
4510 @opindex fno-compare-debug
4511 If no error occurs during compilation, run the compiler a second time,
4512 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4513 passed to the second compilation. Dump the final internal
4514 representation in both compilations, and print an error if they differ.
4516 If the equal sign is omitted, the default @option{-gtoggle} is used.
4518 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4519 and nonzero, implicitly enables @option{-fcompare-debug}. If
4520 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4521 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4524 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4525 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4526 of the final representation and the second compilation, preventing even
4527 @env{GCC_COMPARE_DEBUG} from taking effect.
4529 To verify full coverage during @option{-fcompare-debug} testing, set
4530 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4531 which GCC will reject as an invalid option in any actual compilation
4532 (rather than preprocessing, assembly or linking). To get just a
4533 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4534 not overridden} will do.
4536 @item -fcompare-debug-second
4537 @opindex fcompare-debug-second
4538 This option is implicitly passed to the compiler for the second
4539 compilation requested by @option{-fcompare-debug}, along with options to
4540 silence warnings, and omitting other options that would cause
4541 side-effect compiler outputs to files or to the standard output. Dump
4542 files and preserved temporary files are renamed so as to contain the
4543 @code{.gk} additional extension during the second compilation, to avoid
4544 overwriting those generated by the first.
4546 When this option is passed to the compiler driver, it causes the
4547 @emph{first} compilation to be skipped, which makes it useful for little
4548 other than debugging the compiler proper.
4550 @item -feliminate-dwarf2-dups
4551 @opindex feliminate-dwarf2-dups
4552 Compress DWARF2 debugging information by eliminating duplicated
4553 information about each symbol. This option only makes sense when
4554 generating DWARF2 debugging information with @option{-gdwarf-2}.
4556 @item -femit-struct-debug-baseonly
4557 Emit debug information for struct-like types
4558 only when the base name of the compilation source file
4559 matches the base name of file in which the struct was defined.
4561 This option substantially reduces the size of debugging information,
4562 but at significant potential loss in type information to the debugger.
4563 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4564 See @option{-femit-struct-debug-detailed} for more detailed control.
4566 This option works only with DWARF 2.
4568 @item -femit-struct-debug-reduced
4569 Emit debug information for struct-like types
4570 only when the base name of the compilation source file
4571 matches the base name of file in which the type was defined,
4572 unless the struct is a template or defined in a system header.
4574 This option significantly reduces the size of debugging information,
4575 with some potential loss in type information to the debugger.
4576 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4577 See @option{-femit-struct-debug-detailed} for more detailed control.
4579 This option works only with DWARF 2.
4581 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4582 Specify the struct-like types
4583 for which the compiler will generate debug information.
4584 The intent is to reduce duplicate struct debug information
4585 between different object files within the same program.
4587 This option is a detailed version of
4588 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4589 which will serve for most needs.
4591 A specification has the syntax
4592 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4594 The optional first word limits the specification to
4595 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4596 A struct type is used directly when it is the type of a variable, member.
4597 Indirect uses arise through pointers to structs.
4598 That is, when use of an incomplete struct would be legal, the use is indirect.
4600 @samp{struct one direct; struct two * indirect;}.
4602 The optional second word limits the specification to
4603 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4604 Generic structs are a bit complicated to explain.
4605 For C++, these are non-explicit specializations of template classes,
4606 or non-template classes within the above.
4607 Other programming languages have generics,
4608 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4610 The third word specifies the source files for those
4611 structs for which the compiler will emit debug information.
4612 The values @samp{none} and @samp{any} have the normal meaning.
4613 The value @samp{base} means that
4614 the base of name of the file in which the type declaration appears
4615 must match the base of the name of the main compilation file.
4616 In practice, this means that
4617 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4618 but types declared in other header will not.
4619 The value @samp{sys} means those types satisfying @samp{base}
4620 or declared in system or compiler headers.
4622 You may need to experiment to determine the best settings for your application.
4624 The default is @samp{-femit-struct-debug-detailed=all}.
4626 This option works only with DWARF 2.
4628 @item -fenable-icf-debug
4629 @opindex fenable-icf-debug
4630 Generate additional debug information to support identical code folding (ICF).
4631 This option only works with DWARF version 2 or higher.
4633 @item -fno-merge-debug-strings
4634 @opindex fmerge-debug-strings
4635 @opindex fno-merge-debug-strings
4636 Direct the linker to not merge together strings in the debugging
4637 information which are identical in different object files. Merging is
4638 not supported by all assemblers or linkers. Merging decreases the size
4639 of the debug information in the output file at the cost of increasing
4640 link processing time. Merging is enabled by default.
4642 @item -fdebug-prefix-map=@var{old}=@var{new}
4643 @opindex fdebug-prefix-map
4644 When compiling files in directory @file{@var{old}}, record debugging
4645 information describing them as in @file{@var{new}} instead.
4647 @item -fno-dwarf2-cfi-asm
4648 @opindex fdwarf2-cfi-asm
4649 @opindex fno-dwarf2-cfi-asm
4650 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4651 instead of using GAS @code{.cfi_*} directives.
4653 @cindex @command{prof}
4656 Generate extra code to write profile information suitable for the
4657 analysis program @command{prof}. You must use this option when compiling
4658 the source files you want data about, and you must also use it when
4661 @cindex @command{gprof}
4664 Generate extra code to write profile information suitable for the
4665 analysis program @command{gprof}. You must use this option when compiling
4666 the source files you want data about, and you must also use it when
4671 Makes the compiler print out each function name as it is compiled, and
4672 print some statistics about each pass when it finishes.
4675 @opindex ftime-report
4676 Makes the compiler print some statistics about the time consumed by each
4677 pass when it finishes.
4680 @opindex fmem-report
4681 Makes the compiler print some statistics about permanent memory
4682 allocation when it finishes.
4684 @item -fpre-ipa-mem-report
4685 @opindex fpre-ipa-mem-report
4686 @item -fpost-ipa-mem-report
4687 @opindex fpost-ipa-mem-report
4688 Makes the compiler print some statistics about permanent memory
4689 allocation before or after interprocedural optimization.
4691 @item -fprofile-arcs
4692 @opindex fprofile-arcs
4693 Add code so that program flow @dfn{arcs} are instrumented. During
4694 execution the program records how many times each branch and call is
4695 executed and how many times it is taken or returns. When the compiled
4696 program exits it saves this data to a file called
4697 @file{@var{auxname}.gcda} for each source file. The data may be used for
4698 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4699 test coverage analysis (@option{-ftest-coverage}). Each object file's
4700 @var{auxname} is generated from the name of the output file, if
4701 explicitly specified and it is not the final executable, otherwise it is
4702 the basename of the source file. In both cases any suffix is removed
4703 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4704 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4705 @xref{Cross-profiling}.
4707 @cindex @command{gcov}
4711 This option is used to compile and link code instrumented for coverage
4712 analysis. The option is a synonym for @option{-fprofile-arcs}
4713 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4714 linking). See the documentation for those options for more details.
4719 Compile the source files with @option{-fprofile-arcs} plus optimization
4720 and code generation options. For test coverage analysis, use the
4721 additional @option{-ftest-coverage} option. You do not need to profile
4722 every source file in a program.
4725 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4726 (the latter implies the former).
4729 Run the program on a representative workload to generate the arc profile
4730 information. This may be repeated any number of times. You can run
4731 concurrent instances of your program, and provided that the file system
4732 supports locking, the data files will be correctly updated. Also
4733 @code{fork} calls are detected and correctly handled (double counting
4737 For profile-directed optimizations, compile the source files again with
4738 the same optimization and code generation options plus
4739 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4740 Control Optimization}).
4743 For test coverage analysis, use @command{gcov} to produce human readable
4744 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4745 @command{gcov} documentation for further information.
4749 With @option{-fprofile-arcs}, for each function of your program GCC
4750 creates a program flow graph, then finds a spanning tree for the graph.
4751 Only arcs that are not on the spanning tree have to be instrumented: the
4752 compiler adds code to count the number of times that these arcs are
4753 executed. When an arc is the only exit or only entrance to a block, the
4754 instrumentation code can be added to the block; otherwise, a new basic
4755 block must be created to hold the instrumentation code.
4758 @item -ftest-coverage
4759 @opindex ftest-coverage
4760 Produce a notes file that the @command{gcov} code-coverage utility
4761 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4762 show program coverage. Each source file's note file is called
4763 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4764 above for a description of @var{auxname} and instructions on how to
4765 generate test coverage data. Coverage data will match the source files
4766 more closely, if you do not optimize.
4768 @item -fdbg-cnt-list
4769 @opindex fdbg-cnt-list
4770 Print the name and the counter upperbound for all debug counters.
4772 @item -fdbg-cnt=@var{counter-value-list}
4774 Set the internal debug counter upperbound. @var{counter-value-list}
4775 is a comma-separated list of @var{name}:@var{value} pairs
4776 which sets the upperbound of each debug counter @var{name} to @var{value}.
4777 All debug counters have the initial upperbound of @var{UINT_MAX},
4778 thus dbg_cnt() returns true always unless the upperbound is set by this option.
4779 e.g. With -fdbg-cnt=dce:10,tail_call:0
4780 dbg_cnt(dce) will return true only for first 10 invocations
4781 and dbg_cnt(tail_call) will return false always.
4783 @item -d@var{letters}
4784 @itemx -fdump-rtl-@var{pass}
4786 Says to make debugging dumps during compilation at times specified by
4787 @var{letters}. This is used for debugging the RTL-based passes of the
4788 compiler. The file names for most of the dumps are made by appending
4789 a pass number and a word to the @var{dumpname}, and the files are
4790 created in the directory of the output file. @var{dumpname} is
4791 generated from the name of the output file, if explicitly specified
4792 and it is not an executable, otherwise it is the basename of the
4793 source file. These switches may have different effects when
4794 @option{-E} is used for preprocessing.
4796 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
4797 @option{-d} option @var{letters}. Here are the possible
4798 letters for use in @var{pass} and @var{letters}, and their meanings:
4802 @item -fdump-rtl-alignments
4803 @opindex fdump-rtl-alignments
4804 Dump after branch alignments have been computed.
4806 @item -fdump-rtl-asmcons
4807 @opindex fdump-rtl-asmcons
4808 Dump after fixing rtl statements that have unsatisfied in/out constraints.
4810 @item -fdump-rtl-auto_inc_dec
4811 @opindex fdump-rtl-auto_inc_dec
4812 Dump after auto-inc-dec discovery. This pass is only run on
4813 architectures that have auto inc or auto dec instructions.
4815 @item -fdump-rtl-barriers
4816 @opindex fdump-rtl-barriers
4817 Dump after cleaning up the barrier instructions.
4819 @item -fdump-rtl-bbpart
4820 @opindex fdump-rtl-bbpart
4821 Dump after partitioning hot and cold basic blocks.
4823 @item -fdump-rtl-bbro
4824 @opindex fdump-rtl-bbro
4825 Dump after block reordering.
4827 @item -fdump-rtl-btl1
4828 @itemx -fdump-rtl-btl2
4829 @opindex fdump-rtl-btl2
4830 @opindex fdump-rtl-btl2
4831 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
4832 after the two branch
4833 target load optimization passes.
4835 @item -fdump-rtl-bypass
4836 @opindex fdump-rtl-bypass
4837 Dump after jump bypassing and control flow optimizations.
4839 @item -fdump-rtl-combine
4840 @opindex fdump-rtl-combine
4841 Dump after the RTL instruction combination pass.
4843 @item -fdump-rtl-compgotos
4844 @opindex fdump-rtl-compgotos
4845 Dump after duplicating the computed gotos.
4847 @item -fdump-rtl-ce1
4848 @itemx -fdump-rtl-ce2
4849 @itemx -fdump-rtl-ce3
4850 @opindex fdump-rtl-ce1
4851 @opindex fdump-rtl-ce2
4852 @opindex fdump-rtl-ce3
4853 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
4854 @option{-fdump-rtl-ce3} enable dumping after the three
4855 if conversion passes.
4857 @itemx -fdump-rtl-cprop_hardreg
4858 @opindex fdump-rtl-cprop_hardreg
4859 Dump after hard register copy propagation.
4861 @itemx -fdump-rtl-csa
4862 @opindex fdump-rtl-csa
4863 Dump after combining stack adjustments.
4865 @item -fdump-rtl-cse1
4866 @itemx -fdump-rtl-cse2
4867 @opindex fdump-rtl-cse1
4868 @opindex fdump-rtl-cse2
4869 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
4870 the two common sub-expression elimination passes.
4872 @itemx -fdump-rtl-dce
4873 @opindex fdump-rtl-dce
4874 Dump after the standalone dead code elimination passes.
4876 @itemx -fdump-rtl-dbr
4877 @opindex fdump-rtl-dbr
4878 Dump after delayed branch scheduling.
4880 @item -fdump-rtl-dce1
4881 @itemx -fdump-rtl-dce2
4882 @opindex fdump-rtl-dce1
4883 @opindex fdump-rtl-dce2
4884 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
4885 the two dead store elimination passes.
4888 @opindex fdump-rtl-eh
4889 Dump after finalization of EH handling code.
4891 @item -fdump-rtl-eh_ranges
4892 @opindex fdump-rtl-eh_ranges
4893 Dump after conversion of EH handling range regions.
4895 @item -fdump-rtl-expand
4896 @opindex fdump-rtl-expand
4897 Dump after RTL generation.
4899 @item -fdump-rtl-fwprop1
4900 @itemx -fdump-rtl-fwprop2
4901 @opindex fdump-rtl-fwprop1
4902 @opindex fdump-rtl-fwprop2
4903 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
4904 dumping after the two forward propagation passes.
4906 @item -fdump-rtl-gcse1
4907 @itemx -fdump-rtl-gcse2
4908 @opindex fdump-rtl-gcse1
4909 @opindex fdump-rtl-gcse2
4910 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
4911 after global common subexpression elimination.
4913 @item -fdump-rtl-init-regs
4914 @opindex fdump-rtl-init-regs
4915 Dump after the initialization of the registers.
4917 @item -fdump-rtl-initvals
4918 @opindex fdump-rtl-initvals
4919 Dump after the computation of the initial value sets.
4921 @itemx -fdump-rtl-into_cfglayout
4922 @opindex fdump-rtl-into_cfglayout
4923 Dump after converting to cfglayout mode.
4925 @item -fdump-rtl-ira
4926 @opindex fdump-rtl-ira
4927 Dump after iterated register allocation.
4929 @item -fdump-rtl-jump
4930 @opindex fdump-rtl-jump
4931 Dump after the second jump optimization.
4933 @item -fdump-rtl-loop2
4934 @opindex fdump-rtl-loop2
4935 @option{-fdump-rtl-loop2} enables dumping after the rtl
4936 loop optimization passes.
4938 @item -fdump-rtl-mach
4939 @opindex fdump-rtl-mach
4940 Dump after performing the machine dependent reorganization pass, if that
4943 @item -fdump-rtl-mode_sw
4944 @opindex fdump-rtl-mode_sw
4945 Dump after removing redundant mode switches.
4947 @item -fdump-rtl-rnreg
4948 @opindex fdump-rtl-rnreg
4949 Dump after register renumbering.
4951 @itemx -fdump-rtl-outof_cfglayout
4952 @opindex fdump-rtl-outof_cfglayout
4953 Dump after converting from cfglayout mode.
4955 @item -fdump-rtl-peephole2
4956 @opindex fdump-rtl-peephole2
4957 Dump after the peephole pass.
4959 @item -fdump-rtl-postreload
4960 @opindex fdump-rtl-postreload
4961 Dump after post-reload optimizations.
4963 @itemx -fdump-rtl-pro_and_epilogue
4964 @opindex fdump-rtl-pro_and_epilogue
4965 Dump after generating the function pro and epilogues.
4967 @item -fdump-rtl-regmove
4968 @opindex fdump-rtl-regmove
4969 Dump after the register move pass.
4971 @item -fdump-rtl-sched1
4972 @itemx -fdump-rtl-sched2
4973 @opindex fdump-rtl-sched1
4974 @opindex fdump-rtl-sched2
4975 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
4976 after the basic block scheduling passes.
4978 @item -fdump-rtl-see
4979 @opindex fdump-rtl-see
4980 Dump after sign extension elimination.
4982 @item -fdump-rtl-seqabstr
4983 @opindex fdump-rtl-seqabstr
4984 Dump after common sequence discovery.
4986 @item -fdump-rtl-shorten
4987 @opindex fdump-rtl-shorten
4988 Dump after shortening branches.
4990 @item -fdump-rtl-sibling
4991 @opindex fdump-rtl-sibling
4992 Dump after sibling call optimizations.
4994 @item -fdump-rtl-split1
4995 @itemx -fdump-rtl-split2
4996 @itemx -fdump-rtl-split3
4997 @itemx -fdump-rtl-split4
4998 @itemx -fdump-rtl-split5
4999 @opindex fdump-rtl-split1
5000 @opindex fdump-rtl-split2
5001 @opindex fdump-rtl-split3
5002 @opindex fdump-rtl-split4
5003 @opindex fdump-rtl-split5
5004 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
5005 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
5006 @option{-fdump-rtl-split5} enable dumping after five rounds of
5007 instruction splitting.
5009 @item -fdump-rtl-sms
5010 @opindex fdump-rtl-sms
5011 Dump after modulo scheduling. This pass is only run on some
5014 @item -fdump-rtl-stack
5015 @opindex fdump-rtl-stack
5016 Dump after conversion from GCC's "flat register file" registers to the
5017 x87's stack-like registers. This pass is only run on x86 variants.
5019 @item -fdump-rtl-subreg1
5020 @itemx -fdump-rtl-subreg2
5021 @opindex fdump-rtl-subreg1
5022 @opindex fdump-rtl-subreg2
5023 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
5024 the two subreg expansion passes.
5026 @item -fdump-rtl-unshare
5027 @opindex fdump-rtl-unshare
5028 Dump after all rtl has been unshared.
5030 @item -fdump-rtl-vartrack
5031 @opindex fdump-rtl-vartrack
5032 Dump after variable tracking.
5034 @item -fdump-rtl-vregs
5035 @opindex fdump-rtl-vregs
5036 Dump after converting virtual registers to hard registers.
5038 @item -fdump-rtl-web
5039 @opindex fdump-rtl-web
5040 Dump after live range splitting.
5042 @item -fdump-rtl-regclass
5043 @itemx -fdump-rtl-subregs_of_mode_init
5044 @itemx -fdump-rtl-subregs_of_mode_finish
5045 @itemx -fdump-rtl-dfinit
5046 @itemx -fdump-rtl-dfinish
5047 @opindex fdump-rtl-regclass
5048 @opindex fdump-rtl-subregs_of_mode_init
5049 @opindex fdump-rtl-subregs_of_mode_finish
5050 @opindex fdump-rtl-dfinit
5051 @opindex fdump-rtl-dfinish
5052 These dumps are defined but always produce empty files.
5054 @item -fdump-rtl-all
5055 @opindex fdump-rtl-all
5056 Produce all the dumps listed above.
5060 Annotate the assembler output with miscellaneous debugging information.
5064 Dump all macro definitions, at the end of preprocessing, in addition to
5069 Produce a core dump whenever an error occurs.
5073 Print statistics on memory usage, at the end of the run, to
5078 Annotate the assembler output with a comment indicating which
5079 pattern and alternative was used. The length of each instruction is
5084 Dump the RTL in the assembler output as a comment before each instruction.
5085 Also turns on @option{-dp} annotation.
5089 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
5090 dump a representation of the control flow graph suitable for viewing with VCG
5091 to @file{@var{file}.@var{pass}.vcg}.
5095 Just generate RTL for a function instead of compiling it. Usually used
5096 with @option{-fdump-rtl-expand}.
5100 Dump debugging information during parsing, to standard error.
5104 @opindex fdump-noaddr
5105 When doing debugging dumps, suppress address output. This makes it more
5106 feasible to use diff on debugging dumps for compiler invocations with
5107 different compiler binaries and/or different
5108 text / bss / data / heap / stack / dso start locations.
5110 @item -fdump-unnumbered
5111 @opindex fdump-unnumbered
5112 When doing debugging dumps, suppress instruction numbers and address output.
5113 This makes it more feasible to use diff on debugging dumps for compiler
5114 invocations with different options, in particular with and without
5117 @item -fdump-unnumbered-links
5118 @opindex fdump-unnumbered-links
5119 When doing debugging dumps (see @option{-d} option above), suppress
5120 instruction numbers for the links to the previous and next instructions
5123 @item -fdump-translation-unit @r{(C++ only)}
5124 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5125 @opindex fdump-translation-unit
5126 Dump a representation of the tree structure for the entire translation
5127 unit to a file. The file name is made by appending @file{.tu} to the
5128 source file name, and the file is created in the same directory as the
5129 output file. If the @samp{-@var{options}} form is used, @var{options}
5130 controls the details of the dump as described for the
5131 @option{-fdump-tree} options.
5133 @item -fdump-class-hierarchy @r{(C++ only)}
5134 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5135 @opindex fdump-class-hierarchy
5136 Dump a representation of each class's hierarchy and virtual function
5137 table layout to a file. The file name is made by appending
5138 @file{.class} to the source file name, and the file is created in the
5139 same directory as the output file. If the @samp{-@var{options}} form
5140 is used, @var{options} controls the details of the dump as described
5141 for the @option{-fdump-tree} options.
5143 @item -fdump-ipa-@var{switch}
5145 Control the dumping at various stages of inter-procedural analysis
5146 language tree to a file. The file name is generated by appending a
5147 switch specific suffix to the source file name, and the file is created
5148 in the same directory as the output file. The following dumps are
5153 Enables all inter-procedural analysis dumps.
5156 Dumps information about call-graph optimization, unused function removal,
5157 and inlining decisions.
5160 Dump after function inlining.
5164 @item -fdump-statistics-@var{option}
5165 @opindex fdump-statistics
5166 Enable and control dumping of pass statistics in a separate file. The
5167 file name is generated by appending a suffix ending in
5168 @samp{.statistics} to the source file name, and the file is created in
5169 the same directory as the output file. If the @samp{-@var{option}}
5170 form is used, @samp{-stats} will cause counters to be summed over the
5171 whole compilation unit while @samp{-details} will dump every event as
5172 the passes generate them. The default with no option is to sum
5173 counters for each function compiled.
5175 @item -fdump-tree-@var{switch}
5176 @itemx -fdump-tree-@var{switch}-@var{options}
5178 Control the dumping at various stages of processing the intermediate
5179 language tree to a file. The file name is generated by appending a
5180 switch specific suffix to the source file name, and the file is
5181 created in the same directory as the output file. If the
5182 @samp{-@var{options}} form is used, @var{options} is a list of
5183 @samp{-} separated options that control the details of the dump. Not
5184 all options are applicable to all dumps, those which are not
5185 meaningful will be ignored. The following options are available
5189 Print the address of each node. Usually this is not meaningful as it
5190 changes according to the environment and source file. Its primary use
5191 is for tying up a dump file with a debug environment.
5193 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
5194 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
5195 use working backward from mangled names in the assembly file.
5197 Inhibit dumping of members of a scope or body of a function merely
5198 because that scope has been reached. Only dump such items when they
5199 are directly reachable by some other path. When dumping pretty-printed
5200 trees, this option inhibits dumping the bodies of control structures.
5202 Print a raw representation of the tree. By default, trees are
5203 pretty-printed into a C-like representation.
5205 Enable more detailed dumps (not honored by every dump option).
5207 Enable dumping various statistics about the pass (not honored by every dump
5210 Enable showing basic block boundaries (disabled in raw dumps).
5212 Enable showing virtual operands for every statement.
5214 Enable showing line numbers for statements.
5216 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5218 Enable showing the tree dump for each statement.
5220 Enable showing the EH region number holding each statement.
5222 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5223 and @option{lineno}.
5226 The following tree dumps are possible:
5230 @opindex fdump-tree-original
5231 Dump before any tree based optimization, to @file{@var{file}.original}.
5234 @opindex fdump-tree-optimized
5235 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5238 @opindex fdump-tree-gimple
5239 Dump each function before and after the gimplification pass to a file. The
5240 file name is made by appending @file{.gimple} to the source file name.
5243 @opindex fdump-tree-cfg
5244 Dump the control flow graph of each function to a file. The file name is
5245 made by appending @file{.cfg} to the source file name.
5248 @opindex fdump-tree-vcg
5249 Dump the control flow graph of each function to a file in VCG format. The
5250 file name is made by appending @file{.vcg} to the source file name. Note
5251 that if the file contains more than one function, the generated file cannot
5252 be used directly by VCG@. You will need to cut and paste each function's
5253 graph into its own separate file first.
5256 @opindex fdump-tree-ch
5257 Dump each function after copying loop headers. The file name is made by
5258 appending @file{.ch} to the source file name.
5261 @opindex fdump-tree-ssa
5262 Dump SSA related information to a file. The file name is made by appending
5263 @file{.ssa} to the source file name.
5266 @opindex fdump-tree-alias
5267 Dump aliasing information for each function. The file name is made by
5268 appending @file{.alias} to the source file name.
5271 @opindex fdump-tree-ccp
5272 Dump each function after CCP@. The file name is made by appending
5273 @file{.ccp} to the source file name.
5276 @opindex fdump-tree-storeccp
5277 Dump each function after STORE-CCP@. The file name is made by appending
5278 @file{.storeccp} to the source file name.
5281 @opindex fdump-tree-pre
5282 Dump trees after partial redundancy elimination. The file name is made
5283 by appending @file{.pre} to the source file name.
5286 @opindex fdump-tree-fre
5287 Dump trees after full redundancy elimination. The file name is made
5288 by appending @file{.fre} to the source file name.
5291 @opindex fdump-tree-copyprop
5292 Dump trees after copy propagation. The file name is made
5293 by appending @file{.copyprop} to the source file name.
5295 @item store_copyprop
5296 @opindex fdump-tree-store_copyprop
5297 Dump trees after store copy-propagation. The file name is made
5298 by appending @file{.store_copyprop} to the source file name.
5301 @opindex fdump-tree-dce
5302 Dump each function after dead code elimination. The file name is made by
5303 appending @file{.dce} to the source file name.
5306 @opindex fdump-tree-mudflap
5307 Dump each function after adding mudflap instrumentation. The file name is
5308 made by appending @file{.mudflap} to the source file name.
5311 @opindex fdump-tree-sra
5312 Dump each function after performing scalar replacement of aggregates. The
5313 file name is made by appending @file{.sra} to the source file name.
5316 @opindex fdump-tree-sink
5317 Dump each function after performing code sinking. The file name is made
5318 by appending @file{.sink} to the source file name.
5321 @opindex fdump-tree-dom
5322 Dump each function after applying dominator tree optimizations. The file
5323 name is made by appending @file{.dom} to the source file name.
5326 @opindex fdump-tree-dse
5327 Dump each function after applying dead store elimination. The file
5328 name is made by appending @file{.dse} to the source file name.
5331 @opindex fdump-tree-phiopt
5332 Dump each function after optimizing PHI nodes into straightline code. The file
5333 name is made by appending @file{.phiopt} to the source file name.
5336 @opindex fdump-tree-forwprop
5337 Dump each function after forward propagating single use variables. The file
5338 name is made by appending @file{.forwprop} to the source file name.
5341 @opindex fdump-tree-copyrename
5342 Dump each function after applying the copy rename optimization. The file
5343 name is made by appending @file{.copyrename} to the source file name.
5346 @opindex fdump-tree-nrv
5347 Dump each function after applying the named return value optimization on
5348 generic trees. The file name is made by appending @file{.nrv} to the source
5352 @opindex fdump-tree-vect
5353 Dump each function after applying vectorization of loops. The file name is
5354 made by appending @file{.vect} to the source file name.
5357 @opindex fdump-tree-vrp
5358 Dump each function after Value Range Propagation (VRP). The file name
5359 is made by appending @file{.vrp} to the source file name.
5362 @opindex fdump-tree-all
5363 Enable all the available tree dumps with the flags provided in this option.
5366 @item -ftree-vectorizer-verbose=@var{n}
5367 @opindex ftree-vectorizer-verbose
5368 This option controls the amount of debugging output the vectorizer prints.
5369 This information is written to standard error, unless
5370 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5371 in which case it is output to the usual dump listing file, @file{.vect}.
5372 For @var{n}=0 no diagnostic information is reported.
5373 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5374 and the total number of loops that got vectorized.
5375 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5376 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5377 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5378 level that @option{-fdump-tree-vect-stats} uses.
5379 Higher verbosity levels mean either more information dumped for each
5380 reported loop, or same amount of information reported for more loops:
5381 If @var{n}=3, alignment related information is added to the reports.
5382 If @var{n}=4, data-references related information (e.g.@: memory dependences,
5383 memory access-patterns) is added to the reports.
5384 If @var{n}=5, the vectorizer reports also non-vectorized inner-most loops
5385 that did not pass the first analysis phase (i.e., may not be countable, or
5386 may have complicated control-flow).
5387 If @var{n}=6, the vectorizer reports also non-vectorized nested loops.
5388 For @var{n}=7, all the information the vectorizer generates during its
5389 analysis and transformation is reported. This is the same verbosity level
5390 that @option{-fdump-tree-vect-details} uses.
5392 @item -frandom-seed=@var{string}
5393 @opindex frandom-seed
5394 This option provides a seed that GCC uses when it would otherwise use
5395 random numbers. It is used to generate certain symbol names
5396 that have to be different in every compiled file. It is also used to
5397 place unique stamps in coverage data files and the object files that
5398 produce them. You can use the @option{-frandom-seed} option to produce
5399 reproducibly identical object files.
5401 The @var{string} should be different for every file you compile.
5403 @item -fsched-verbose=@var{n}
5404 @opindex fsched-verbose
5405 On targets that use instruction scheduling, this option controls the
5406 amount of debugging output the scheduler prints. This information is
5407 written to standard error, unless @option{-fdump-rtl-sched1} or
5408 @option{-fdump-rtl-sched2} is specified, in which case it is output
5409 to the usual dump listing file, @file{.sched} or @file{.sched2}
5410 respectively. However for @var{n} greater than nine, the output is
5411 always printed to standard error.
5413 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5414 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5415 For @var{n} greater than one, it also output basic block probabilities,
5416 detailed ready list information and unit/insn info. For @var{n} greater
5417 than two, it includes RTL at abort point, control-flow and regions info.
5418 And for @var{n} over four, @option{-fsched-verbose} also includes
5422 @itemx -save-temps=cwd
5424 Store the usual ``temporary'' intermediate files permanently; place them
5425 in the current directory and name them based on the source file. Thus,
5426 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5427 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5428 preprocessed @file{foo.i} output file even though the compiler now
5429 normally uses an integrated preprocessor.
5431 When used in combination with the @option{-x} command line option,
5432 @option{-save-temps} is sensible enough to avoid over writing an
5433 input source file with the same extension as an intermediate file.
5434 The corresponding intermediate file may be obtained by renaming the
5435 source file before using @option{-save-temps}.
5437 If you invoke GCC in parallel, compiling several different source
5438 files that share a common base name in different subdirectories or the
5439 same source file compiled for multiple output destinations, it is
5440 likely that the different parallel compilers will interfere with each
5441 other, and overwrite the temporary files. For instance:
5444 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5445 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5448 may result in @file{foo.i} and @file{foo.o} being written to
5449 simultaneously by both compilers.
5451 @item -save-temps=obj
5452 @opindex save-temps=obj
5453 Store the usual ``temporary'' intermediate files permanently. If the
5454 @option{-o} option is used, the temporary files are based on the
5455 object file. If the @option{-o} option is not used, the
5456 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5461 gcc -save-temps=obj -c foo.c
5462 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5463 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5466 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5467 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5468 @file{dir2/yfoobar.o}.
5470 @item -time@r{[}=@var{file}@r{]}
5472 Report the CPU time taken by each subprocess in the compilation
5473 sequence. For C source files, this is the compiler proper and assembler
5474 (plus the linker if linking is done).
5476 Without the specification of an output file, the output looks like this:
5483 The first number on each line is the ``user time'', that is time spent
5484 executing the program itself. The second number is ``system time'',
5485 time spent executing operating system routines on behalf of the program.
5486 Both numbers are in seconds.
5488 With the specification of an output file, the output is appended to the
5489 named file, and it looks like this:
5492 0.12 0.01 cc1 @var{options}
5493 0.00 0.01 as @var{options}
5496 The ``user time'' and the ``system time'' are moved before the program
5497 name, and the options passed to the program are displayed, so that one
5498 can later tell what file was being compiled, and with which options.
5500 @item -fvar-tracking
5501 @opindex fvar-tracking
5502 Run variable tracking pass. It computes where variables are stored at each
5503 position in code. Better debugging information is then generated
5504 (if the debugging information format supports this information).
5506 It is enabled by default when compiling with optimization (@option{-Os},
5507 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5508 the debug info format supports it.
5510 @item -fvar-tracking-assignments
5511 @opindex fvar-tracking-assignments
5512 @opindex fno-var-tracking-assignments
5513 Annotate assignments to user variables early in the compilation and
5514 attempt to carry the annotations over throughout the compilation all the
5515 way to the end, in an attempt to improve debug information while
5516 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
5518 It can be enabled even if var-tracking is disabled, in which case
5519 annotations will be created and maintained, but discarded at the end.
5521 @item -fvar-tracking-assignments-toggle
5522 @opindex fvar-tracking-assignments-toggle
5523 @opindex fno-var-tracking-assignments-toggle
5524 Toggle @option{-fvar-tracking-assignments}, in the same way that
5525 @option{-gtoggle} toggles @option{-g}.
5527 @item -print-file-name=@var{library}
5528 @opindex print-file-name
5529 Print the full absolute name of the library file @var{library} that
5530 would be used when linking---and don't do anything else. With this
5531 option, GCC does not compile or link anything; it just prints the
5534 @item -print-multi-directory
5535 @opindex print-multi-directory
5536 Print the directory name corresponding to the multilib selected by any
5537 other switches present in the command line. This directory is supposed
5538 to exist in @env{GCC_EXEC_PREFIX}.
5540 @item -print-multi-lib
5541 @opindex print-multi-lib
5542 Print the mapping from multilib directory names to compiler switches
5543 that enable them. The directory name is separated from the switches by
5544 @samp{;}, and each switch starts with an @samp{@@} instead of the
5545 @samp{-}, without spaces between multiple switches. This is supposed to
5546 ease shell-processing.
5548 @item -print-multi-os-directory
5549 @opindex print-multi-os-directory
5550 Print the path to OS libraries for the selected
5551 multilib, relative to some @file{lib} subdirectory. If OS libraries are
5552 present in the @file{lib} subdirectory and no multilibs are used, this is
5553 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
5554 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
5555 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
5556 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
5558 @item -print-prog-name=@var{program}
5559 @opindex print-prog-name
5560 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5562 @item -print-libgcc-file-name
5563 @opindex print-libgcc-file-name
5564 Same as @option{-print-file-name=libgcc.a}.
5566 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5567 but you do want to link with @file{libgcc.a}. You can do
5570 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5573 @item -print-search-dirs
5574 @opindex print-search-dirs
5575 Print the name of the configured installation directory and a list of
5576 program and library directories @command{gcc} will search---and don't do anything else.
5578 This is useful when @command{gcc} prints the error message
5579 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5580 To resolve this you either need to put @file{cpp0} and the other compiler
5581 components where @command{gcc} expects to find them, or you can set the environment
5582 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5583 Don't forget the trailing @samp{/}.
5584 @xref{Environment Variables}.
5586 @item -print-sysroot
5587 @opindex print-sysroot
5588 Print the target sysroot directory that will be used during
5589 compilation. This is the target sysroot specified either at configure
5590 time or using the @option{--sysroot} option, possibly with an extra
5591 suffix that depends on compilation options. If no target sysroot is
5592 specified, the option prints nothing.
5594 @item -print-sysroot-headers-suffix
5595 @opindex print-sysroot-headers-suffix
5596 Print the suffix added to the target sysroot when searching for
5597 headers, or give an error if the compiler is not configured with such
5598 a suffix---and don't do anything else.
5601 @opindex dumpmachine
5602 Print the compiler's target machine (for example,
5603 @samp{i686-pc-linux-gnu})---and don't do anything else.
5606 @opindex dumpversion
5607 Print the compiler version (for example, @samp{3.0})---and don't do
5612 Print the compiler's built-in specs---and don't do anything else. (This
5613 is used when GCC itself is being built.) @xref{Spec Files}.
5615 @item -feliminate-unused-debug-types
5616 @opindex feliminate-unused-debug-types
5617 Normally, when producing DWARF2 output, GCC will emit debugging
5618 information for all types declared in a compilation
5619 unit, regardless of whether or not they are actually used
5620 in that compilation unit. Sometimes this is useful, such as
5621 if, in the debugger, you want to cast a value to a type that is
5622 not actually used in your program (but is declared). More often,
5623 however, this results in a significant amount of wasted space.
5624 With this option, GCC will avoid producing debug symbol output
5625 for types that are nowhere used in the source file being compiled.
5628 @node Optimize Options
5629 @section Options That Control Optimization
5630 @cindex optimize options
5631 @cindex options, optimization
5633 These options control various sorts of optimizations.
5635 Without any optimization option, the compiler's goal is to reduce the
5636 cost of compilation and to make debugging produce the expected
5637 results. Statements are independent: if you stop the program with a
5638 breakpoint between statements, you can then assign a new value to any
5639 variable or change the program counter to any other statement in the
5640 function and get exactly the results you would expect from the source
5643 Turning on optimization flags makes the compiler attempt to improve
5644 the performance and/or code size at the expense of compilation time
5645 and possibly the ability to debug the program.
5647 The compiler performs optimization based on the knowledge it has of the
5648 program. Compiling multiple files at once to a single output file mode allows
5649 the compiler to use information gained from all of the files when compiling
5652 Not all optimizations are controlled directly by a flag. Only
5653 optimizations that have a flag are listed in this section.
5655 Most of the optimizations are not enabled if a @option{-O} level is not set on
5656 the command line, even if individual optimization flags are specified.
5658 Depending on the target and how GCC was configured, a slightly different
5659 set of optimizations may be enabled at each @option{-O} level than
5660 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
5661 to find out the exact set of optimizations that are enabled at each level.
5662 @xref{Overall Options}, for examples.
5669 Optimize. Optimizing compilation takes somewhat more time, and a lot
5670 more memory for a large function.
5672 With @option{-O}, the compiler tries to reduce code size and execution
5673 time, without performing any optimizations that take a great deal of
5676 @option{-O} turns on the following optimization flags:
5679 -fcprop-registers @gol
5682 -fdelayed-branch @gol
5684 -fguess-branch-probability @gol
5685 -fif-conversion2 @gol
5686 -fif-conversion @gol
5687 -fipa-pure-const @gol
5688 -fipa-reference @gol
5690 -fsplit-wide-types @gol
5691 -ftree-builtin-call-dce @gol
5694 -ftree-copyrename @gol
5696 -ftree-dominator-opts @gol
5698 -ftree-forwprop @gol
5706 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5707 where doing so does not interfere with debugging.
5711 Optimize even more. GCC performs nearly all supported optimizations
5712 that do not involve a space-speed tradeoff.
5713 As compared to @option{-O}, this option increases both compilation time
5714 and the performance of the generated code.
5716 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5717 also turns on the following optimization flags:
5718 @gccoptlist{-fthread-jumps @gol
5719 -falign-functions -falign-jumps @gol
5720 -falign-loops -falign-labels @gol
5723 -fcse-follow-jumps -fcse-skip-blocks @gol
5724 -fdelete-null-pointer-checks @gol
5725 -fexpensive-optimizations @gol
5726 -fgcse -fgcse-lm @gol
5727 -finline-small-functions @gol
5728 -findirect-inlining @gol
5730 -foptimize-sibling-calls @gol
5733 -freorder-blocks -freorder-functions @gol
5734 -frerun-cse-after-loop @gol
5735 -fsched-interblock -fsched-spec @gol
5736 -fschedule-insns -fschedule-insns2 @gol
5737 -fstrict-aliasing -fstrict-overflow @gol
5738 -ftree-switch-conversion @gol
5742 Please note the warning under @option{-fgcse} about
5743 invoking @option{-O2} on programs that use computed gotos.
5747 Optimize yet more. @option{-O3} turns on all optimizations specified
5748 by @option{-O2} and also turns on the @option{-finline-functions},
5749 @option{-funswitch-loops}, @option{-fpredictive-commoning},
5750 @option{-fgcse-after-reload} and @option{-ftree-vectorize} options.
5754 Reduce compilation time and make debugging produce the expected
5755 results. This is the default.
5759 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5760 do not typically increase code size. It also performs further
5761 optimizations designed to reduce code size.
5763 @option{-Os} disables the following optimization flags:
5764 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5765 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5766 -fprefetch-loop-arrays -ftree-vect-loop-version}
5768 If you use multiple @option{-O} options, with or without level numbers,
5769 the last such option is the one that is effective.
5772 Options of the form @option{-f@var{flag}} specify machine-independent
5773 flags. Most flags have both positive and negative forms; the negative
5774 form of @option{-ffoo} would be @option{-fno-foo}. In the table
5775 below, only one of the forms is listed---the one you typically will
5776 use. You can figure out the other form by either removing @samp{no-}
5779 The following options control specific optimizations. They are either
5780 activated by @option{-O} options or are related to ones that are. You
5781 can use the following flags in the rare cases when ``fine-tuning'' of
5782 optimizations to be performed is desired.
5785 @item -fno-default-inline
5786 @opindex fno-default-inline
5787 Do not make member functions inline by default merely because they are
5788 defined inside the class scope (C++ only). Otherwise, when you specify
5789 @w{@option{-O}}, member functions defined inside class scope are compiled
5790 inline by default; i.e., you don't need to add @samp{inline} in front of
5791 the member function name.
5793 @item -fno-defer-pop
5794 @opindex fno-defer-pop
5795 Always pop the arguments to each function call as soon as that function
5796 returns. For machines which must pop arguments after a function call,
5797 the compiler normally lets arguments accumulate on the stack for several
5798 function calls and pops them all at once.
5800 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5802 @item -fforward-propagate
5803 @opindex fforward-propagate
5804 Perform a forward propagation pass on RTL@. The pass tries to combine two
5805 instructions and checks if the result can be simplified. If loop unrolling
5806 is active, two passes are performed and the second is scheduled after
5809 This option is enabled by default at optimization levels @option{-O},
5810 @option{-O2}, @option{-O3}, @option{-Os}.
5812 @item -fomit-frame-pointer
5813 @opindex fomit-frame-pointer
5814 Don't keep the frame pointer in a register for functions that
5815 don't need one. This avoids the instructions to save, set up and
5816 restore frame pointers; it also makes an extra register available
5817 in many functions. @strong{It also makes debugging impossible on
5820 On some machines, such as the VAX, this flag has no effect, because
5821 the standard calling sequence automatically handles the frame pointer
5822 and nothing is saved by pretending it doesn't exist. The
5823 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
5824 whether a target machine supports this flag. @xref{Registers,,Register
5825 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
5827 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5829 @item -foptimize-sibling-calls
5830 @opindex foptimize-sibling-calls
5831 Optimize sibling and tail recursive calls.
5833 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5837 Don't pay attention to the @code{inline} keyword. Normally this option
5838 is used to keep the compiler from expanding any functions inline.
5839 Note that if you are not optimizing, no functions can be expanded inline.
5841 @item -finline-small-functions
5842 @opindex finline-small-functions
5843 Integrate functions into their callers when their body is smaller than expected
5844 function call code (so overall size of program gets smaller). The compiler
5845 heuristically decides which functions are simple enough to be worth integrating
5848 Enabled at level @option{-O2}.
5850 @item -findirect-inlining
5851 @opindex findirect-inlining
5852 Inline also indirect calls that are discovered to be known at compile
5853 time thanks to previous inlining. This option has any effect only
5854 when inlining itself is turned on by the @option{-finline-functions}
5855 or @option{-finline-small-functions} options.
5857 Enabled at level @option{-O2}.
5859 @item -finline-functions
5860 @opindex finline-functions
5861 Integrate all simple functions into their callers. The compiler
5862 heuristically decides which functions are simple enough to be worth
5863 integrating in this way.
5865 If all calls to a given function are integrated, and the function is
5866 declared @code{static}, then the function is normally not output as
5867 assembler code in its own right.
5869 Enabled at level @option{-O3}.
5871 @item -finline-functions-called-once
5872 @opindex finline-functions-called-once
5873 Consider all @code{static} functions called once for inlining into their
5874 caller even if they are not marked @code{inline}. If a call to a given
5875 function is integrated, then the function is not output as assembler code
5878 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
5880 @item -fearly-inlining
5881 @opindex fearly-inlining
5882 Inline functions marked by @code{always_inline} and functions whose body seems
5883 smaller than the function call overhead early before doing
5884 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
5885 makes profiling significantly cheaper and usually inlining faster on programs
5886 having large chains of nested wrapper functions.
5892 Perform interprocedural scalar replacement of aggregates, removal of
5893 unused parameters and replacement of parameters passed by reference
5894 by parameters passed by value.
5896 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
5898 @item -finline-limit=@var{n}
5899 @opindex finline-limit
5900 By default, GCC limits the size of functions that can be inlined. This flag
5901 allows coarse control of this limit. @var{n} is the size of functions that
5902 can be inlined in number of pseudo instructions.
5904 Inlining is actually controlled by a number of parameters, which may be
5905 specified individually by using @option{--param @var{name}=@var{value}}.
5906 The @option{-finline-limit=@var{n}} option sets some of these parameters
5910 @item max-inline-insns-single
5911 is set to @var{n}/2.
5912 @item max-inline-insns-auto
5913 is set to @var{n}/2.
5916 See below for a documentation of the individual
5917 parameters controlling inlining and for the defaults of these parameters.
5919 @emph{Note:} there may be no value to @option{-finline-limit} that results
5920 in default behavior.
5922 @emph{Note:} pseudo instruction represents, in this particular context, an
5923 abstract measurement of function's size. In no way does it represent a count
5924 of assembly instructions and as such its exact meaning might change from one
5925 release to an another.
5927 @item -fkeep-inline-functions
5928 @opindex fkeep-inline-functions
5929 In C, emit @code{static} functions that are declared @code{inline}
5930 into the object file, even if the function has been inlined into all
5931 of its callers. This switch does not affect functions using the
5932 @code{extern inline} extension in GNU C89@. In C++, emit any and all
5933 inline functions into the object file.
5935 @item -fkeep-static-consts
5936 @opindex fkeep-static-consts
5937 Emit variables declared @code{static const} when optimization isn't turned
5938 on, even if the variables aren't referenced.
5940 GCC enables this option by default. If you want to force the compiler to
5941 check if the variable was referenced, regardless of whether or not
5942 optimization is turned on, use the @option{-fno-keep-static-consts} option.
5944 @item -fmerge-constants
5945 @opindex fmerge-constants
5946 Attempt to merge identical constants (string constants and floating point
5947 constants) across compilation units.
5949 This option is the default for optimized compilation if the assembler and
5950 linker support it. Use @option{-fno-merge-constants} to inhibit this
5953 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5955 @item -fmerge-all-constants
5956 @opindex fmerge-all-constants
5957 Attempt to merge identical constants and identical variables.
5959 This option implies @option{-fmerge-constants}. In addition to
5960 @option{-fmerge-constants} this considers e.g.@: even constant initialized
5961 arrays or initialized constant variables with integral or floating point
5962 types. Languages like C or C++ require each variable, including multiple
5963 instances of the same variable in recursive calls, to have distinct locations,
5964 so using this option will result in non-conforming
5967 @item -fmodulo-sched
5968 @opindex fmodulo-sched
5969 Perform swing modulo scheduling immediately before the first scheduling
5970 pass. This pass looks at innermost loops and reorders their
5971 instructions by overlapping different iterations.
5973 @item -fmodulo-sched-allow-regmoves
5974 @opindex fmodulo-sched-allow-regmoves
5975 Perform more aggressive SMS based modulo scheduling with register moves
5976 allowed. By setting this flag certain anti-dependences edges will be
5977 deleted which will trigger the generation of reg-moves based on the
5978 life-range analysis. This option is effective only with
5979 @option{-fmodulo-sched} enabled.
5981 @item -fno-branch-count-reg
5982 @opindex fno-branch-count-reg
5983 Do not use ``decrement and branch'' instructions on a count register,
5984 but instead generate a sequence of instructions that decrement a
5985 register, compare it against zero, then branch based upon the result.
5986 This option is only meaningful on architectures that support such
5987 instructions, which include x86, PowerPC, IA-64 and S/390.
5989 The default is @option{-fbranch-count-reg}.
5991 @item -fno-function-cse
5992 @opindex fno-function-cse
5993 Do not put function addresses in registers; make each instruction that
5994 calls a constant function contain the function's address explicitly.
5996 This option results in less efficient code, but some strange hacks
5997 that alter the assembler output may be confused by the optimizations
5998 performed when this option is not used.
6000 The default is @option{-ffunction-cse}
6002 @item -fno-zero-initialized-in-bss
6003 @opindex fno-zero-initialized-in-bss
6004 If the target supports a BSS section, GCC by default puts variables that
6005 are initialized to zero into BSS@. This can save space in the resulting
6008 This option turns off this behavior because some programs explicitly
6009 rely on variables going to the data section. E.g., so that the
6010 resulting executable can find the beginning of that section and/or make
6011 assumptions based on that.
6013 The default is @option{-fzero-initialized-in-bss}.
6015 @item -fmudflap -fmudflapth -fmudflapir
6019 @cindex bounds checking
6021 For front-ends that support it (C and C++), instrument all risky
6022 pointer/array dereferencing operations, some standard library
6023 string/heap functions, and some other associated constructs with
6024 range/validity tests. Modules so instrumented should be immune to
6025 buffer overflows, invalid heap use, and some other classes of C/C++
6026 programming errors. The instrumentation relies on a separate runtime
6027 library (@file{libmudflap}), which will be linked into a program if
6028 @option{-fmudflap} is given at link time. Run-time behavior of the
6029 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
6030 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
6033 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
6034 link if your program is multi-threaded. Use @option{-fmudflapir}, in
6035 addition to @option{-fmudflap} or @option{-fmudflapth}, if
6036 instrumentation should ignore pointer reads. This produces less
6037 instrumentation (and therefore faster execution) and still provides
6038 some protection against outright memory corrupting writes, but allows
6039 erroneously read data to propagate within a program.
6041 @item -fthread-jumps
6042 @opindex fthread-jumps
6043 Perform optimizations where we check to see if a jump branches to a
6044 location where another comparison subsumed by the first is found. If
6045 so, the first branch is redirected to either the destination of the
6046 second branch or a point immediately following it, depending on whether
6047 the condition is known to be true or false.
6049 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6051 @item -fsplit-wide-types
6052 @opindex fsplit-wide-types
6053 When using a type that occupies multiple registers, such as @code{long
6054 long} on a 32-bit system, split the registers apart and allocate them
6055 independently. This normally generates better code for those types,
6056 but may make debugging more difficult.
6058 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
6061 @item -fcse-follow-jumps
6062 @opindex fcse-follow-jumps
6063 In common subexpression elimination (CSE), scan through jump instructions
6064 when the target of the jump is not reached by any other path. For
6065 example, when CSE encounters an @code{if} statement with an
6066 @code{else} clause, CSE will follow the jump when the condition
6069 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6071 @item -fcse-skip-blocks
6072 @opindex fcse-skip-blocks
6073 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
6074 follow jumps which conditionally skip over blocks. When CSE
6075 encounters a simple @code{if} statement with no else clause,
6076 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6077 body of the @code{if}.
6079 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6081 @item -frerun-cse-after-loop
6082 @opindex frerun-cse-after-loop
6083 Re-run common subexpression elimination after loop optimizations has been
6086 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6090 Perform a global common subexpression elimination pass.
6091 This pass also performs global constant and copy propagation.
6093 @emph{Note:} When compiling a program using computed gotos, a GCC
6094 extension, you may get better runtime performance if you disable
6095 the global common subexpression elimination pass by adding
6096 @option{-fno-gcse} to the command line.
6098 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6102 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
6103 attempt to move loads which are only killed by stores into themselves. This
6104 allows a loop containing a load/store sequence to be changed to a load outside
6105 the loop, and a copy/store within the loop.
6107 Enabled by default when gcse is enabled.
6111 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6112 global common subexpression elimination. This pass will attempt to move
6113 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6114 loops containing a load/store sequence can be changed to a load before
6115 the loop and a store after the loop.
6117 Not enabled at any optimization level.
6121 When @option{-fgcse-las} is enabled, the global common subexpression
6122 elimination pass eliminates redundant loads that come after stores to the
6123 same memory location (both partial and full redundancies).
6125 Not enabled at any optimization level.
6127 @item -fgcse-after-reload
6128 @opindex fgcse-after-reload
6129 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6130 pass is performed after reload. The purpose of this pass is to cleanup
6133 @item -funsafe-loop-optimizations
6134 @opindex funsafe-loop-optimizations
6135 If given, the loop optimizer will assume that loop indices do not
6136 overflow, and that the loops with nontrivial exit condition are not
6137 infinite. This enables a wider range of loop optimizations even if
6138 the loop optimizer itself cannot prove that these assumptions are valid.
6139 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
6140 if it finds this kind of loop.
6142 @item -fcrossjumping
6143 @opindex fcrossjumping
6144 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6145 resulting code may or may not perform better than without cross-jumping.
6147 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6149 @item -fauto-inc-dec
6150 @opindex fauto-inc-dec
6151 Combine increments or decrements of addresses with memory accesses.
6152 This pass is always skipped on architectures that do not have
6153 instructions to support this. Enabled by default at @option{-O} and
6154 higher on architectures that support this.
6158 Perform dead code elimination (DCE) on RTL@.
6159 Enabled by default at @option{-O} and higher.
6163 Perform dead store elimination (DSE) on RTL@.
6164 Enabled by default at @option{-O} and higher.
6166 @item -fif-conversion
6167 @opindex fif-conversion
6168 Attempt to transform conditional jumps into branch-less equivalents. This
6169 include use of conditional moves, min, max, set flags and abs instructions, and
6170 some tricks doable by standard arithmetics. The use of conditional execution
6171 on chips where it is available is controlled by @code{if-conversion2}.
6173 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6175 @item -fif-conversion2
6176 @opindex fif-conversion2
6177 Use conditional execution (where available) to transform conditional jumps into
6178 branch-less equivalents.
6180 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6182 @item -fdelete-null-pointer-checks
6183 @opindex fdelete-null-pointer-checks
6184 Assume that programs cannot safely dereference null pointers, and that
6185 no code or data element resides there. This enables simple constant
6186 folding optimizations at all optimization levels. In addition, other
6187 optimization passes in GCC use this flag to control global dataflow
6188 analyses that eliminate useless checks for null pointers; these assume
6189 that if a pointer is checked after it has already been dereferenced,
6192 Note however that in some environments this assumption is not true.
6193 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6194 for programs which depend on that behavior.
6196 Some targets, especially embedded ones, disable this option at all levels.
6197 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6198 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6199 are enabled independently at different optimization levels.
6201 @item -fexpensive-optimizations
6202 @opindex fexpensive-optimizations
6203 Perform a number of minor optimizations that are relatively expensive.
6205 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6207 @item -foptimize-register-move
6209 @opindex foptimize-register-move
6211 Attempt to reassign register numbers in move instructions and as
6212 operands of other simple instructions in order to maximize the amount of
6213 register tying. This is especially helpful on machines with two-operand
6216 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6219 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6221 @item -fira-algorithm=@var{algorithm}
6222 Use specified coloring algorithm for the integrated register
6223 allocator. The @var{algorithm} argument should be @code{priority} or
6224 @code{CB}. The first algorithm specifies Chow's priority coloring,
6225 the second one specifies Chaitin-Briggs coloring. The second
6226 algorithm can be unimplemented for some architectures. If it is
6227 implemented, it is the default because Chaitin-Briggs coloring as a
6228 rule generates a better code.
6230 @item -fira-region=@var{region}
6231 Use specified regions for the integrated register allocator. The
6232 @var{region} argument should be one of @code{all}, @code{mixed}, or
6233 @code{one}. The first value means using all loops as register
6234 allocation regions, the second value which is the default means using
6235 all loops except for loops with small register pressure as the
6236 regions, and third one means using all function as a single region.
6237 The first value can give best result for machines with small size and
6238 irregular register set, the third one results in faster and generates
6239 decent code and the smallest size code, and the default value usually
6240 give the best results in most cases and for most architectures.
6242 @item -fira-coalesce
6243 @opindex fira-coalesce
6244 Do optimistic register coalescing. This option might be profitable for
6245 architectures with big regular register files.
6247 @item -fira-loop-pressure
6248 @opindex fira-loop-pressure
6249 Use IRA to evaluate register pressure in loops for decision to move
6250 loop invariants. Usage of this option usually results in generation
6251 of faster and smaller code on machines with big register files (>= 32
6252 registers) but it can slow compiler down.
6254 This option is enabled at level @option{-O3} for some targets.
6256 @item -fno-ira-share-save-slots
6257 @opindex fno-ira-share-save-slots
6258 Switch off sharing stack slots used for saving call used hard
6259 registers living through a call. Each hard register will get a
6260 separate stack slot and as a result function stack frame will be
6263 @item -fno-ira-share-spill-slots
6264 @opindex fno-ira-share-spill-slots
6265 Switch off sharing stack slots allocated for pseudo-registers. Each
6266 pseudo-register which did not get a hard register will get a separate
6267 stack slot and as a result function stack frame will be bigger.
6269 @item -fira-verbose=@var{n}
6270 @opindex fira-verbose
6271 Set up how verbose dump file for the integrated register allocator
6272 will be. Default value is 5. If the value is greater or equal to 10,
6273 the dump file will be stderr as if the value were @var{n} minus 10.
6275 @item -fdelayed-branch
6276 @opindex fdelayed-branch
6277 If supported for the target machine, attempt to reorder instructions
6278 to exploit instruction slots available after delayed branch
6281 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6283 @item -fschedule-insns
6284 @opindex fschedule-insns
6285 If supported for the target machine, attempt to reorder instructions to
6286 eliminate execution stalls due to required data being unavailable. This
6287 helps machines that have slow floating point or memory load instructions
6288 by allowing other instructions to be issued until the result of the load
6289 or floating point instruction is required.
6291 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6293 @item -fschedule-insns2
6294 @opindex fschedule-insns2
6295 Similar to @option{-fschedule-insns}, but requests an additional pass of
6296 instruction scheduling after register allocation has been done. This is
6297 especially useful on machines with a relatively small number of
6298 registers and where memory load instructions take more than one cycle.
6300 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6302 @item -fno-sched-interblock
6303 @opindex fno-sched-interblock
6304 Don't schedule instructions across basic blocks. This is normally
6305 enabled by default when scheduling before register allocation, i.e.@:
6306 with @option{-fschedule-insns} or at @option{-O2} or higher.
6308 @item -fno-sched-spec
6309 @opindex fno-sched-spec
6310 Don't allow speculative motion of non-load instructions. This is normally
6311 enabled by default when scheduling before register allocation, i.e.@:
6312 with @option{-fschedule-insns} or at @option{-O2} or higher.
6314 @item -fsched-pressure
6315 @opindex fsched-pressure
6316 Enable register pressure sensitive insn scheduling before the register
6317 allocation. This only makes sense when scheduling before register
6318 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
6319 @option{-O2} or higher. Usage of this option can improve the
6320 generated code and decrease its size by preventing register pressure
6321 increase above the number of available hard registers and as a
6322 consequence register spills in the register allocation.
6324 @item -fsched-spec-load
6325 @opindex fsched-spec-load
6326 Allow speculative motion of some load instructions. This only makes
6327 sense when scheduling before register allocation, i.e.@: with
6328 @option{-fschedule-insns} or at @option{-O2} or higher.
6330 @item -fsched-spec-load-dangerous
6331 @opindex fsched-spec-load-dangerous
6332 Allow speculative motion of more load instructions. This only makes
6333 sense when scheduling before register allocation, i.e.@: with
6334 @option{-fschedule-insns} or at @option{-O2} or higher.
6336 @item -fsched-stalled-insns
6337 @itemx -fsched-stalled-insns=@var{n}
6338 @opindex fsched-stalled-insns
6339 Define how many insns (if any) can be moved prematurely from the queue
6340 of stalled insns into the ready list, during the second scheduling pass.
6341 @option{-fno-sched-stalled-insns} means that no insns will be moved
6342 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6343 on how many queued insns can be moved prematurely.
6344 @option{-fsched-stalled-insns} without a value is equivalent to
6345 @option{-fsched-stalled-insns=1}.
6347 @item -fsched-stalled-insns-dep
6348 @itemx -fsched-stalled-insns-dep=@var{n}
6349 @opindex fsched-stalled-insns-dep
6350 Define how many insn groups (cycles) will be examined for a dependency
6351 on a stalled insn that is candidate for premature removal from the queue
6352 of stalled insns. This has an effect only during the second scheduling pass,
6353 and only if @option{-fsched-stalled-insns} is used.
6354 @option{-fno-sched-stalled-insns-dep} is equivalent to
6355 @option{-fsched-stalled-insns-dep=0}.
6356 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6357 @option{-fsched-stalled-insns-dep=1}.
6359 @item -fsched2-use-superblocks
6360 @opindex fsched2-use-superblocks
6361 When scheduling after register allocation, do use superblock scheduling
6362 algorithm. Superblock scheduling allows motion across basic block boundaries
6363 resulting on faster schedules. This option is experimental, as not all machine
6364 descriptions used by GCC model the CPU closely enough to avoid unreliable
6365 results from the algorithm.
6367 This only makes sense when scheduling after register allocation, i.e.@: with
6368 @option{-fschedule-insns2} or at @option{-O2} or higher.
6370 @item -fsched-group-heuristic
6371 @opindex fsched-group-heuristic
6372 Enable the group heuristic in the scheduler. This heuristic favors
6373 the instruction that belongs to a schedule group. This is enabled
6374 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6375 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6377 @item -fsched-critical-path-heuristic
6378 @opindex fsched-critical-path-heuristic
6379 Enable the critical-path heuristic in the scheduler. This heuristic favors
6380 instructions on the critical path. This is enabled by default when
6381 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6382 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6384 @item -fsched-spec-insn-heuristic
6385 @opindex fsched-spec-insn-heuristic
6386 Enable the speculative instruction heuristic in the scheduler. This
6387 heuristic favors speculative instructions with greater dependency weakness.
6388 This is enabled by default when scheduling is enabled, i.e.@:
6389 with @option{-fschedule-insns} or @option{-fschedule-insns2}
6390 or at @option{-O2} or higher.
6392 @item -fsched-rank-heuristic
6393 @opindex fsched-rank-heuristic
6394 Enable the rank heuristic in the scheduler. This heuristic favors
6395 the instruction belonging to a basic block with greater size or frequency.
6396 This is enabled by default when scheduling is enabled, i.e.@:
6397 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6398 at @option{-O2} or higher.
6400 @item -fsched-last-insn-heuristic
6401 @opindex fsched-last-insn-heuristic
6402 Enable the last-instruction heuristic in the scheduler. This heuristic
6403 favors the instruction that is less dependent on the last instruction
6404 scheduled. This is enabled by default when scheduling is enabled,
6405 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6406 at @option{-O2} or higher.
6408 @item -fsched-dep-count-heuristic
6409 @opindex fsched-dep-count-heuristic
6410 Enable the dependent-count heuristic in the scheduler. This heuristic
6411 favors the instruction that has more instructions depending on it.
6412 This is enabled by default when scheduling is enabled, i.e.@:
6413 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6414 at @option{-O2} or higher.
6416 @item -fsched2-use-traces
6417 @opindex fsched2-use-traces
6418 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
6419 allocation and additionally perform code duplication in order to increase the
6420 size of superblocks using tracer pass. See @option{-ftracer} for details on
6423 This mode should produce faster but significantly longer programs. Also
6424 without @option{-fbranch-probabilities} the traces constructed may not
6425 match the reality and hurt the performance. This only makes
6426 sense when scheduling after register allocation, i.e.@: with
6427 @option{-fschedule-insns2} or at @option{-O2} or higher.
6429 @item -freschedule-modulo-scheduled-loops
6430 @opindex freschedule-modulo-scheduled-loops
6431 The modulo scheduling comes before the traditional scheduling, if a loop
6432 was modulo scheduled we may want to prevent the later scheduling passes
6433 from changing its schedule, we use this option to control that.
6435 @item -fselective-scheduling
6436 @opindex fselective-scheduling
6437 Schedule instructions using selective scheduling algorithm. Selective
6438 scheduling runs instead of the first scheduler pass.
6440 @item -fselective-scheduling2
6441 @opindex fselective-scheduling2
6442 Schedule instructions using selective scheduling algorithm. Selective
6443 scheduling runs instead of the second scheduler pass.
6445 @item -fsel-sched-pipelining
6446 @opindex fsel-sched-pipelining
6447 Enable software pipelining of innermost loops during selective scheduling.
6448 This option has no effect until one of @option{-fselective-scheduling} or
6449 @option{-fselective-scheduling2} is turned on.
6451 @item -fsel-sched-pipelining-outer-loops
6452 @opindex fsel-sched-pipelining-outer-loops
6453 When pipelining loops during selective scheduling, also pipeline outer loops.
6454 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6456 @item -fcaller-saves
6457 @opindex fcaller-saves
6458 Enable values to be allocated in registers that will be clobbered by
6459 function calls, by emitting extra instructions to save and restore the
6460 registers around such calls. Such allocation is done only when it
6461 seems to result in better code than would otherwise be produced.
6463 This option is always enabled by default on certain machines, usually
6464 those which have no call-preserved registers to use instead.
6466 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6468 @item -fconserve-stack
6469 @opindex fconserve-stack
6470 Attempt to minimize stack usage. The compiler will attempt to use less
6471 stack space, even if that makes the program slower. This option
6472 implies setting the @option{large-stack-frame} parameter to 100
6473 and the @option{large-stack-frame-growth} parameter to 400.
6475 @item -ftree-reassoc
6476 @opindex ftree-reassoc
6477 Perform reassociation on trees. This flag is enabled by default
6478 at @option{-O} and higher.
6482 Perform partial redundancy elimination (PRE) on trees. This flag is
6483 enabled by default at @option{-O2} and @option{-O3}.
6485 @item -ftree-forwprop
6486 @opindex ftree-forwprop
6487 Perform forward propagation on trees. This flag is enabled by default
6488 at @option{-O} and higher.
6492 Perform full redundancy elimination (FRE) on trees. The difference
6493 between FRE and PRE is that FRE only considers expressions
6494 that are computed on all paths leading to the redundant computation.
6495 This analysis is faster than PRE, though it exposes fewer redundancies.
6496 This flag is enabled by default at @option{-O} and higher.
6498 @item -ftree-phiprop
6499 @opindex ftree-phiprop
6500 Perform hoisting of loads from conditional pointers on trees. This
6501 pass is enabled by default at @option{-O} and higher.
6503 @item -ftree-copy-prop
6504 @opindex ftree-copy-prop
6505 Perform copy propagation on trees. This pass eliminates unnecessary
6506 copy operations. This flag is enabled by default at @option{-O} and
6509 @item -fipa-pure-const
6510 @opindex fipa-pure-const
6511 Discover which functions are pure or constant.
6512 Enabled by default at @option{-O} and higher.
6514 @item -fipa-reference
6515 @opindex fipa-reference
6516 Discover which static variables do not escape cannot escape the
6518 Enabled by default at @option{-O} and higher.
6520 @item -fipa-struct-reorg
6521 @opindex fipa-struct-reorg
6522 Perform structure reorganization optimization, that change C-like structures
6523 layout in order to better utilize spatial locality. This transformation is
6524 affective for programs containing arrays of structures. Available in two
6525 compilation modes: profile-based (enabled with @option{-fprofile-generate})
6526 or static (which uses built-in heuristics). Require @option{-fipa-type-escape}
6527 to provide the safety of this transformation. It works only in whole program
6528 mode, so it requires @option{-fwhole-program} and @option{-combine} to be
6529 enabled. Structures considered @samp{cold} by this transformation are not
6530 affected (see @option{--param struct-reorg-cold-struct-ratio=@var{value}}).
6532 With this flag, the program debug info reflects a new structure layout.
6536 Perform interprocedural pointer analysis. This option is experimental
6537 and does not affect generated code.
6541 Perform interprocedural constant propagation.
6542 This optimization analyzes the program to determine when values passed
6543 to functions are constants and then optimizes accordingly.
6544 This optimization can substantially increase performance
6545 if the application has constants passed to functions.
6546 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
6548 @item -fipa-cp-clone
6549 @opindex fipa-cp-clone
6550 Perform function cloning to make interprocedural constant propagation stronger.
6551 When enabled, interprocedural constant propagation will perform function cloning
6552 when externally visible function can be called with constant arguments.
6553 Because this optimization can create multiple copies of functions,
6554 it may significantly increase code size
6555 (see @option{--param ipcp-unit-growth=@var{value}}).
6556 This flag is enabled by default at @option{-O3}.
6558 @item -fipa-matrix-reorg
6559 @opindex fipa-matrix-reorg
6560 Perform matrix flattening and transposing.
6561 Matrix flattening tries to replace an @math{m}-dimensional matrix
6562 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
6563 This reduces the level of indirection needed for accessing the elements
6564 of the matrix. The second optimization is matrix transposing that
6565 attempts to change the order of the matrix's dimensions in order to
6566 improve cache locality.
6567 Both optimizations need the @option{-fwhole-program} flag.
6568 Transposing is enabled only if profiling information is available.
6572 Perform forward store motion on trees. This flag is
6573 enabled by default at @option{-O} and higher.
6577 Perform sparse conditional constant propagation (CCP) on trees. This
6578 pass only operates on local scalar variables and is enabled by default
6579 at @option{-O} and higher.
6581 @item -ftree-switch-conversion
6582 Perform conversion of simple initializations in a switch to
6583 initializations from a scalar array. This flag is enabled by default
6584 at @option{-O2} and higher.
6588 Perform dead code elimination (DCE) on trees. This flag is enabled by
6589 default at @option{-O} and higher.
6591 @item -ftree-builtin-call-dce
6592 @opindex ftree-builtin-call-dce
6593 Perform conditional dead code elimination (DCE) for calls to builtin functions
6594 that may set @code{errno} but are otherwise side-effect free. This flag is
6595 enabled by default at @option{-O2} and higher if @option{-Os} is not also
6598 @item -ftree-dominator-opts
6599 @opindex ftree-dominator-opts
6600 Perform a variety of simple scalar cleanups (constant/copy
6601 propagation, redundancy elimination, range propagation and expression
6602 simplification) based on a dominator tree traversal. This also
6603 performs jump threading (to reduce jumps to jumps). This flag is
6604 enabled by default at @option{-O} and higher.
6608 Perform dead store elimination (DSE) on trees. A dead store is a store into
6609 a memory location which will later be overwritten by another store without
6610 any intervening loads. In this case the earlier store can be deleted. This
6611 flag is enabled by default at @option{-O} and higher.
6615 Perform loop header copying on trees. This is beneficial since it increases
6616 effectiveness of code motion optimizations. It also saves one jump. This flag
6617 is enabled by default at @option{-O} and higher. It is not enabled
6618 for @option{-Os}, since it usually increases code size.
6620 @item -ftree-loop-optimize
6621 @opindex ftree-loop-optimize
6622 Perform loop optimizations on trees. This flag is enabled by default
6623 at @option{-O} and higher.
6625 @item -ftree-loop-linear
6626 @opindex ftree-loop-linear
6627 Perform linear loop transformations on tree. This flag can improve cache
6628 performance and allow further loop optimizations to take place.
6630 @item -floop-interchange
6631 Perform loop interchange transformations on loops. Interchanging two
6632 nested loops switches the inner and outer loops. For example, given a
6637 A(J, I) = A(J, I) * C
6641 loop interchange will transform the loop as if the user had written:
6645 A(J, I) = A(J, I) * C
6649 which can be beneficial when @code{N} is larger than the caches,
6650 because in Fortran, the elements of an array are stored in memory
6651 contiguously by column, and the original loop iterates over rows,
6652 potentially creating at each access a cache miss. This optimization
6653 applies to all the languages supported by GCC and is not limited to
6654 Fortran. To use this code transformation, GCC has to be configured
6655 with @option{--with-ppl} and @option{--with-cloog} to enable the
6656 Graphite loop transformation infrastructure.
6658 @item -floop-strip-mine
6659 Perform loop strip mining transformations on loops. Strip mining
6660 splits a loop into two nested loops. The outer loop has strides
6661 equal to the strip size and the inner loop has strides of the
6662 original loop within a strip. For example, given a loop like:
6668 loop strip mining will transform the loop as if the user had written:
6671 DO I = II, min (II + 3, N)
6676 This optimization applies to all the languages supported by GCC and is
6677 not limited to Fortran. To use this code transformation, GCC has to
6678 be configured with @option{--with-ppl} and @option{--with-cloog} to
6679 enable the Graphite loop transformation infrastructure.
6682 Perform loop blocking transformations on loops. Blocking strip mines
6683 each loop in the loop nest such that the memory accesses of the
6684 element loops fit inside caches. For example, given a loop like:
6688 A(J, I) = B(I) + C(J)
6692 loop blocking will transform the loop as if the user had written:
6696 DO I = II, min (II + 63, N)
6697 DO J = JJ, min (JJ + 63, M)
6698 A(J, I) = B(I) + C(J)
6704 which can be beneficial when @code{M} is larger than the caches,
6705 because the innermost loop will iterate over a smaller amount of data
6706 that can be kept in the caches. This optimization applies to all the
6707 languages supported by GCC and is not limited to Fortran. To use this
6708 code transformation, GCC has to be configured with @option{--with-ppl}
6709 and @option{--with-cloog} to enable the Graphite loop transformation
6712 @item -fgraphite-identity
6713 @opindex fgraphite-identity
6714 Enable the identity transformation for graphite. For every SCoP we generate
6715 the polyhedral representation and transform it back to gimple. Using
6716 @option{-fgraphite-identity} we can check the costs or benefits of the
6717 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
6718 are also performed by the code generator CLooG, like index splitting and
6719 dead code elimination in loops.
6721 @item -floop-parallelize-all
6722 Use the Graphite data dependence analysis to identify loops that can
6723 be parallelized. Parallelize all the loops that can be analyzed to
6724 not contain loop carried dependences without checking that it is
6725 profitable to parallelize the loops.
6727 @item -fcheck-data-deps
6728 @opindex fcheck-data-deps
6729 Compare the results of several data dependence analyzers. This option
6730 is used for debugging the data dependence analyzers.
6732 @item -ftree-loop-distribution
6733 Perform loop distribution. This flag can improve cache performance on
6734 big loop bodies and allow further loop optimizations, like
6735 parallelization or vectorization, to take place. For example, the loop
6752 @item -ftree-loop-im
6753 @opindex ftree-loop-im
6754 Perform loop invariant motion on trees. This pass moves only invariants that
6755 would be hard to handle at RTL level (function calls, operations that expand to
6756 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
6757 operands of conditions that are invariant out of the loop, so that we can use
6758 just trivial invariantness analysis in loop unswitching. The pass also includes
6761 @item -ftree-loop-ivcanon
6762 @opindex ftree-loop-ivcanon
6763 Create a canonical counter for number of iterations in the loop for that
6764 determining number of iterations requires complicated analysis. Later
6765 optimizations then may determine the number easily. Useful especially
6766 in connection with unrolling.
6770 Perform induction variable optimizations (strength reduction, induction
6771 variable merging and induction variable elimination) on trees.
6773 @item -ftree-parallelize-loops=n
6774 @opindex ftree-parallelize-loops
6775 Parallelize loops, i.e., split their iteration space to run in n threads.
6776 This is only possible for loops whose iterations are independent
6777 and can be arbitrarily reordered. The optimization is only
6778 profitable on multiprocessor machines, for loops that are CPU-intensive,
6779 rather than constrained e.g.@: by memory bandwidth. This option
6780 implies @option{-pthread}, and thus is only supported on targets
6781 that have support for @option{-pthread}.
6785 Perform function-local points-to analysis on trees. This flag is
6786 enabled by default at @option{-O} and higher.
6790 Perform scalar replacement of aggregates. This pass replaces structure
6791 references with scalars to prevent committing structures to memory too
6792 early. This flag is enabled by default at @option{-O} and higher.
6794 @item -ftree-copyrename
6795 @opindex ftree-copyrename
6796 Perform copy renaming on trees. This pass attempts to rename compiler
6797 temporaries to other variables at copy locations, usually resulting in
6798 variable names which more closely resemble the original variables. This flag
6799 is enabled by default at @option{-O} and higher.
6803 Perform temporary expression replacement during the SSA->normal phase. Single
6804 use/single def temporaries are replaced at their use location with their
6805 defining expression. This results in non-GIMPLE code, but gives the expanders
6806 much more complex trees to work on resulting in better RTL generation. This is
6807 enabled by default at @option{-O} and higher.
6809 @item -ftree-vectorize
6810 @opindex ftree-vectorize
6811 Perform loop vectorization on trees. This flag is enabled by default at
6814 @item -ftree-vect-loop-version
6815 @opindex ftree-vect-loop-version
6816 Perform loop versioning when doing loop vectorization on trees. When a loop
6817 appears to be vectorizable except that data alignment or data dependence cannot
6818 be determined at compile time then vectorized and non-vectorized versions of
6819 the loop are generated along with runtime checks for alignment or dependence
6820 to control which version is executed. This option is enabled by default
6821 except at level @option{-Os} where it is disabled.
6823 @item -fvect-cost-model
6824 @opindex fvect-cost-model
6825 Enable cost model for vectorization.
6829 Perform Value Range Propagation on trees. This is similar to the
6830 constant propagation pass, but instead of values, ranges of values are
6831 propagated. This allows the optimizers to remove unnecessary range
6832 checks like array bound checks and null pointer checks. This is
6833 enabled by default at @option{-O2} and higher. Null pointer check
6834 elimination is only done if @option{-fdelete-null-pointer-checks} is
6839 Perform tail duplication to enlarge superblock size. This transformation
6840 simplifies the control flow of the function allowing other optimizations to do
6843 @item -funroll-loops
6844 @opindex funroll-loops
6845 Unroll loops whose number of iterations can be determined at compile
6846 time or upon entry to the loop. @option{-funroll-loops} implies
6847 @option{-frerun-cse-after-loop}. This option makes code larger,
6848 and may or may not make it run faster.
6850 @item -funroll-all-loops
6851 @opindex funroll-all-loops
6852 Unroll all loops, even if their number of iterations is uncertain when
6853 the loop is entered. This usually makes programs run more slowly.
6854 @option{-funroll-all-loops} implies the same options as
6855 @option{-funroll-loops},
6857 @item -fsplit-ivs-in-unroller
6858 @opindex fsplit-ivs-in-unroller
6859 Enables expressing of values of induction variables in later iterations
6860 of the unrolled loop using the value in the first iteration. This breaks
6861 long dependency chains, thus improving efficiency of the scheduling passes.
6863 Combination of @option{-fweb} and CSE is often sufficient to obtain the
6864 same effect. However in cases the loop body is more complicated than
6865 a single basic block, this is not reliable. It also does not work at all
6866 on some of the architectures due to restrictions in the CSE pass.
6868 This optimization is enabled by default.
6870 @item -fvariable-expansion-in-unroller
6871 @opindex fvariable-expansion-in-unroller
6872 With this option, the compiler will create multiple copies of some
6873 local variables when unrolling a loop which can result in superior code.
6875 @item -fpredictive-commoning
6876 @opindex fpredictive-commoning
6877 Perform predictive commoning optimization, i.e., reusing computations
6878 (especially memory loads and stores) performed in previous
6879 iterations of loops.
6881 This option is enabled at level @option{-O3}.
6883 @item -fprefetch-loop-arrays
6884 @opindex fprefetch-loop-arrays
6885 If supported by the target machine, generate instructions to prefetch
6886 memory to improve the performance of loops that access large arrays.
6888 This option may generate better or worse code; results are highly
6889 dependent on the structure of loops within the source code.
6891 Disabled at level @option{-Os}.
6894 @itemx -fno-peephole2
6895 @opindex fno-peephole
6896 @opindex fno-peephole2
6897 Disable any machine-specific peephole optimizations. The difference
6898 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
6899 are implemented in the compiler; some targets use one, some use the
6900 other, a few use both.
6902 @option{-fpeephole} is enabled by default.
6903 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6905 @item -fno-guess-branch-probability
6906 @opindex fno-guess-branch-probability
6907 Do not guess branch probabilities using heuristics.
6909 GCC will use heuristics to guess branch probabilities if they are
6910 not provided by profiling feedback (@option{-fprofile-arcs}). These
6911 heuristics are based on the control flow graph. If some branch probabilities
6912 are specified by @samp{__builtin_expect}, then the heuristics will be
6913 used to guess branch probabilities for the rest of the control flow graph,
6914 taking the @samp{__builtin_expect} info into account. The interactions
6915 between the heuristics and @samp{__builtin_expect} can be complex, and in
6916 some cases, it may be useful to disable the heuristics so that the effects
6917 of @samp{__builtin_expect} are easier to understand.
6919 The default is @option{-fguess-branch-probability} at levels
6920 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6922 @item -freorder-blocks
6923 @opindex freorder-blocks
6924 Reorder basic blocks in the compiled function in order to reduce number of
6925 taken branches and improve code locality.
6927 Enabled at levels @option{-O2}, @option{-O3}.
6929 @item -freorder-blocks-and-partition
6930 @opindex freorder-blocks-and-partition
6931 In addition to reordering basic blocks in the compiled function, in order
6932 to reduce number of taken branches, partitions hot and cold basic blocks
6933 into separate sections of the assembly and .o files, to improve
6934 paging and cache locality performance.
6936 This optimization is automatically turned off in the presence of
6937 exception handling, for linkonce sections, for functions with a user-defined
6938 section attribute and on any architecture that does not support named
6941 @item -freorder-functions
6942 @opindex freorder-functions
6943 Reorder functions in the object file in order to
6944 improve code locality. This is implemented by using special
6945 subsections @code{.text.hot} for most frequently executed functions and
6946 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
6947 the linker so object file format must support named sections and linker must
6948 place them in a reasonable way.
6950 Also profile feedback must be available in to make this option effective. See
6951 @option{-fprofile-arcs} for details.
6953 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6955 @item -fstrict-aliasing
6956 @opindex fstrict-aliasing
6957 Allow the compiler to assume the strictest aliasing rules applicable to
6958 the language being compiled. For C (and C++), this activates
6959 optimizations based on the type of expressions. In particular, an
6960 object of one type is assumed never to reside at the same address as an
6961 object of a different type, unless the types are almost the same. For
6962 example, an @code{unsigned int} can alias an @code{int}, but not a
6963 @code{void*} or a @code{double}. A character type may alias any other
6966 @anchor{Type-punning}Pay special attention to code like this:
6979 The practice of reading from a different union member than the one most
6980 recently written to (called ``type-punning'') is common. Even with
6981 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
6982 is accessed through the union type. So, the code above will work as
6983 expected. @xref{Structures unions enumerations and bit-fields
6984 implementation}. However, this code might not:
6995 Similarly, access by taking the address, casting the resulting pointer
6996 and dereferencing the result has undefined behavior, even if the cast
6997 uses a union type, e.g.:
7001 return ((union a_union *) &d)->i;
7005 The @option{-fstrict-aliasing} option is enabled at levels
7006 @option{-O2}, @option{-O3}, @option{-Os}.
7008 @item -fstrict-overflow
7009 @opindex fstrict-overflow
7010 Allow the compiler to assume strict signed overflow rules, depending
7011 on the language being compiled. For C (and C++) this means that
7012 overflow when doing arithmetic with signed numbers is undefined, which
7013 means that the compiler may assume that it will not happen. This
7014 permits various optimizations. For example, the compiler will assume
7015 that an expression like @code{i + 10 > i} will always be true for
7016 signed @code{i}. This assumption is only valid if signed overflow is
7017 undefined, as the expression is false if @code{i + 10} overflows when
7018 using twos complement arithmetic. When this option is in effect any
7019 attempt to determine whether an operation on signed numbers will
7020 overflow must be written carefully to not actually involve overflow.
7022 This option also allows the compiler to assume strict pointer
7023 semantics: given a pointer to an object, if adding an offset to that
7024 pointer does not produce a pointer to the same object, the addition is
7025 undefined. This permits the compiler to conclude that @code{p + u >
7026 p} is always true for a pointer @code{p} and unsigned integer
7027 @code{u}. This assumption is only valid because pointer wraparound is
7028 undefined, as the expression is false if @code{p + u} overflows using
7029 twos complement arithmetic.
7031 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
7032 that integer signed overflow is fully defined: it wraps. When
7033 @option{-fwrapv} is used, there is no difference between
7034 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
7035 integers. With @option{-fwrapv} certain types of overflow are
7036 permitted. For example, if the compiler gets an overflow when doing
7037 arithmetic on constants, the overflowed value can still be used with
7038 @option{-fwrapv}, but not otherwise.
7040 The @option{-fstrict-overflow} option is enabled at levels
7041 @option{-O2}, @option{-O3}, @option{-Os}.
7043 @item -falign-functions
7044 @itemx -falign-functions=@var{n}
7045 @opindex falign-functions
7046 Align the start of functions to the next power-of-two greater than
7047 @var{n}, skipping up to @var{n} bytes. For instance,
7048 @option{-falign-functions=32} aligns functions to the next 32-byte
7049 boundary, but @option{-falign-functions=24} would align to the next
7050 32-byte boundary only if this can be done by skipping 23 bytes or less.
7052 @option{-fno-align-functions} and @option{-falign-functions=1} are
7053 equivalent and mean that functions will not be aligned.
7055 Some assemblers only support this flag when @var{n} is a power of two;
7056 in that case, it is rounded up.
7058 If @var{n} is not specified or is zero, use a machine-dependent default.
7060 Enabled at levels @option{-O2}, @option{-O3}.
7062 @item -falign-labels
7063 @itemx -falign-labels=@var{n}
7064 @opindex falign-labels
7065 Align all branch targets to a power-of-two boundary, skipping up to
7066 @var{n} bytes like @option{-falign-functions}. This option can easily
7067 make code slower, because it must insert dummy operations for when the
7068 branch target is reached in the usual flow of the code.
7070 @option{-fno-align-labels} and @option{-falign-labels=1} are
7071 equivalent and mean that labels will not be aligned.
7073 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
7074 are greater than this value, then their values are used instead.
7076 If @var{n} is not specified or is zero, use a machine-dependent default
7077 which is very likely to be @samp{1}, meaning no alignment.
7079 Enabled at levels @option{-O2}, @option{-O3}.
7082 @itemx -falign-loops=@var{n}
7083 @opindex falign-loops
7084 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7085 like @option{-falign-functions}. The hope is that the loop will be
7086 executed many times, which will make up for any execution of the dummy
7089 @option{-fno-align-loops} and @option{-falign-loops=1} are
7090 equivalent and mean that loops will not be aligned.
7092 If @var{n} is not specified or is zero, use a machine-dependent default.
7094 Enabled at levels @option{-O2}, @option{-O3}.
7097 @itemx -falign-jumps=@var{n}
7098 @opindex falign-jumps
7099 Align branch targets to a power-of-two boundary, for branch targets
7100 where the targets can only be reached by jumping, skipping up to @var{n}
7101 bytes like @option{-falign-functions}. In this case, no dummy operations
7104 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7105 equivalent and mean that loops will not be aligned.
7107 If @var{n} is not specified or is zero, use a machine-dependent default.
7109 Enabled at levels @option{-O2}, @option{-O3}.
7111 @item -funit-at-a-time
7112 @opindex funit-at-a-time
7113 This option is left for compatibility reasons. @option{-funit-at-a-time}
7114 has no effect, while @option{-fno-unit-at-a-time} implies
7115 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7119 @item -fno-toplevel-reorder
7120 @opindex fno-toplevel-reorder
7121 Do not reorder top-level functions, variables, and @code{asm}
7122 statements. Output them in the same order that they appear in the
7123 input file. When this option is used, unreferenced static variables
7124 will not be removed. This option is intended to support existing code
7125 which relies on a particular ordering. For new code, it is better to
7128 Enabled at level @option{-O0}. When disabled explicitly, it also imply
7129 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
7134 Constructs webs as commonly used for register allocation purposes and assign
7135 each web individual pseudo register. This allows the register allocation pass
7136 to operate on pseudos directly, but also strengthens several other optimization
7137 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7138 however, make debugging impossible, since variables will no longer stay in a
7141 Enabled by default with @option{-funroll-loops}.
7143 @item -fwhole-program
7144 @opindex fwhole-program
7145 Assume that the current compilation unit represents the whole program being
7146 compiled. All public functions and variables with the exception of @code{main}
7147 and those merged by attribute @code{externally_visible} become static functions
7148 and in effect are optimized more aggressively by interprocedural optimizers.
7149 While this option is equivalent to proper use of the @code{static} keyword for
7150 programs consisting of a single file, in combination with option
7151 @option{-combine}, @option{-flto} or @option{-fwhopr} this flag can be used to
7152 compile many smaller scale programs since the functions and variables become
7153 local for the whole combined compilation unit, not for the single source file
7156 This option implies @option{-fwhole-file} for Fortran programs.
7160 This option runs the standard link-time optimizer. When invoked
7161 with source code, it generates GIMPLE (one of GCC's internal
7162 representations) and writes it to special ELF sections in the object
7163 file. When the object files are linked together, all the function
7164 bodies are read from these ELF sections and instantiated as if they
7165 had been part of the same translation unit.
7167 To use the link-timer optimizer, @option{-flto} needs to be specified at
7168 compile time and during the final link. For example,
7171 gcc -c -O2 -flto foo.c
7172 gcc -c -O2 -flto bar.c
7173 gcc -o myprog -flto -O2 foo.o bar.o
7176 The first two invocations to GCC will save a bytecode representation
7177 of GIMPLE into special ELF sections inside @file{foo.o} and
7178 @file{bar.o}. The final invocation will read the GIMPLE bytecode from
7179 @file{foo.o} and @file{bar.o}, merge the two files into a single
7180 internal image, and compile the result as usual. Since both
7181 @file{foo.o} and @file{bar.o} are merged into a single image, this
7182 causes all the inter-procedural analyses and optimizations in GCC to
7183 work across the two files as if they were a single one. This means,
7184 for example, that the inliner will be able to inline functions in
7185 @file{bar.o} into functions in @file{foo.o} and vice-versa.
7187 Another (simpler) way to enable link-time optimization is,
7190 gcc -o myprog -flto -O2 foo.c bar.c
7193 The above will generate bytecode for @file{foo.c} and @file{bar.c},
7194 merge them together into a single GIMPLE representation and optimize
7195 them as usual to produce @file{myprog}.
7197 The only important thing to keep in mind is that to enable link-time
7198 optimizations the @option{-flto} flag needs to be passed to both the
7199 compile and the link commands.
7201 Note that when a file is compiled with @option{-flto}, the generated
7202 object file will be larger than a regular object file because it will
7203 contain GIMPLE bytecodes and the usual final code. This means that
7204 object files with LTO information can be linked as a normal object
7205 file. So, in the previous example, if the final link is done with
7208 gcc -o myprog foo.o bar.o
7211 The only difference will be that no inter-procedural optimizations
7212 will be applied to produce @file{myprog}. The two object files
7213 @file{foo.o} and @file{bar.o} will be simply sent to the regular
7216 Additionally, the optimization flags used to compile individual files
7217 are not necessarily related to those used at link-time. For instance,
7220 gcc -c -O0 -flto foo.c
7221 gcc -c -O0 -flto bar.c
7222 gcc -o myprog -flto -O3 foo.o bar.o
7225 This will produce individual object files with unoptimized assembler
7226 code, but the resulting binary @file{myprog} will be optimized at
7227 @option{-O3}. Now, if the final binary is generated without
7228 @option{-flto}, then @file{myprog} will not be optimized.
7230 When producing the final binary with @option{-flto}, GCC will only
7231 apply link-time optimizations to those files that contain bytecode.
7232 Therefore, you can mix and match object files and libraries with
7233 GIMPLE bytecodes and final object code. GCC will automatically select
7234 which files to optimize in LTO mode and which files to link without
7237 There are some code generation flags that GCC will preserve when
7238 generating bytecodes, as they need to be used during the final link
7239 stage. Currently, the following options are saved into the GIMPLE
7240 bytecode files: @option{-fPIC}, @option{-fcommon} and all the
7241 @option{-m} target flags.
7243 At link time, these options are read-in and reapplied. Note that the
7244 current implementation makes no attempt at recognizing conflicting
7245 values for these options. If two or more files have a conflicting
7246 value (e.g., one file is compiled with @option{-fPIC} and another
7247 isn't), the compiler will simply use the last value read from the
7248 bytecode files. It is recommended, then, that all the files
7249 participating in the same link be compiled with the same options.
7251 Another feature of LTO is that it is possible to apply interprocedural
7252 optimizations on files written in different languages. This requires
7253 some support in the language front end. Currently, the C, C++ and
7254 Fortran front ends are capable of emitting GIMPLE bytecodes, so
7255 something like this should work
7260 gfortran -c -flto baz.f90
7261 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
7264 Notice that the final link is done with @command{g++} to get the C++
7265 runtime libraries and @option{-lgfortran} is added to get the Fortran
7266 runtime libraries. In general, when mixing languages in LTO mode, you
7267 should use the same link command used when mixing languages in a
7268 regular (non-LTO) compilation. This means that if your build process
7269 was mixing languages before, all you need to add is @option{-flto} to
7270 all the compile and link commands.
7272 If object files containing GIMPLE bytecode are stored in a library
7273 archive, say @file{libfoo.a}, it is possible to extract and use them
7274 in an LTO link if you are using @command{gold} as the linker (which,
7275 in turn requires GCC to be configured with @option{--enable-gold}).
7276 To enable this feature, use the flag @option{-use-linker-plugin} at
7280 gcc -o myprog -O2 -flto -use-linker-plugin a.o b.o -lfoo
7283 With the linker plugin enabled, @command{gold} will extract the needed
7284 GIMPLE files from @file{libfoo.a} and pass them on to the running GCC
7285 to make them part of the aggregated GIMPLE image to be optimized.
7287 If you are not using @command{gold} and/or do not specify
7288 @option{-use-linker-plugin} then the objects inside @file{libfoo.a}
7289 will be extracted and linked as usual, but they will not participate
7290 in the LTO optimization process.
7292 Link time optimizations do not require the presence of the whole
7293 program to operate. If the program does not require any symbols to
7294 be exported, it is possible to combine @option{-flto} and
7295 @option{-fwhopr} with @option{-fwhole-program} to allow the
7296 interprocedural optimizers to use more aggressive assumptions which
7297 may lead to improved optimization opportunities.
7299 Regarding portability: the current implementation of LTO makes no
7300 attempt at generating bytecode that can be ported between different
7301 types of hosts. The bytecode files are versioned and there is a
7302 strict version check, so bytecode files generated in one version of
7303 GCC will not work with an older/newer version of GCC.
7305 This option is disabled by default.
7309 This option is identical in functionality to @option{-flto} but it
7310 differs in how the final link stage is executed. Instead of loading
7311 all the function bodies in memory, the callgraph is analyzed and
7312 optimization decisions are made (whole program analysis or WPA). Once
7313 optimization decisions are made, the callgraph is partitioned and the
7314 different sections are compiled separately (local transformations or
7315 LTRANS)@. This process allows optimizations on very large programs
7316 that otherwise would not fit in memory. This option enables
7317 @option{-fwpa} and @option{-fltrans} automatically.
7319 Disabled by default.
7323 This is an internal option used by GCC when compiling with
7324 @option{-fwhopr}. You should never need to use it.
7326 This option runs the link-time optimizer in the whole-program-analysis
7327 (WPA) mode, which reads in summary information from all inputs and
7328 performs a whole-program analysis based on summary information only.
7329 It generates object files for subsequent runs of the link-time
7330 optimizer where individual object files are optimized using both
7331 summary information from the WPA mode and the actual function bodies.
7332 It then drives the LTRANS phase.
7334 Disabled by default.
7338 This is an internal option used by GCC when compiling with
7339 @option{-fwhopr}. You should never need to use it.
7341 This option runs the link-time optimizer in the local-transformation (LTRANS)
7342 mode, which reads in output from a previous run of the LTO in WPA mode.
7343 In the LTRANS mode, LTO optimizes an object and produces the final assembly.
7345 Disabled by default.
7347 @item -fltrans-output-list=@var{file}
7348 @opindex fltrans-output-list
7349 This is an internal option used by GCC when compiling with
7350 @option{-fwhopr}. You should never need to use it.
7352 This option specifies a file to which the names of LTRANS output files are
7353 written. This option is only meaningful in conjunction with @option{-fwpa}.
7355 Disabled by default.
7357 @item -flto-compression-level=@var{n}
7358 This option specifies the level of compression used for intermediate
7359 language written to LTO object files, and is only meaningful in
7360 conjunction with LTO mode (@option{-fwhopr}, @option{-flto}). Valid
7361 values are 0 (no compression) to 9 (maximum compression). Values
7362 outside this range are clamped to either 0 or 9. If the option is not
7363 given, a default balanced compression setting is used.
7366 Prints a report with internal details on the workings of the link-time
7367 optimizer. The contents of this report vary from version to version,
7368 it is meant to be useful to GCC developers when processing object
7369 files in LTO mode (via @option{-fwhopr} or @option{-flto}).
7371 Disabled by default.
7373 @item -use-linker-plugin
7374 Enables the extraction of objects with GIMPLE bytecode information
7375 from library archives. This option relies on features available only
7376 in @command{gold}, so to use this you must configure GCC with
7377 @option{--enable-gold}. See @option{-flto} for a description on the
7378 effect of this flag and how to use it.
7380 Disabled by default.
7382 @item -fcprop-registers
7383 @opindex fcprop-registers
7384 After register allocation and post-register allocation instruction splitting,
7385 we perform a copy-propagation pass to try to reduce scheduling dependencies
7386 and occasionally eliminate the copy.
7388 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7390 @item -fprofile-correction
7391 @opindex fprofile-correction
7392 Profiles collected using an instrumented binary for multi-threaded programs may
7393 be inconsistent due to missed counter updates. When this option is specified,
7394 GCC will use heuristics to correct or smooth out such inconsistencies. By
7395 default, GCC will emit an error message when an inconsistent profile is detected.
7397 @item -fprofile-dir=@var{path}
7398 @opindex fprofile-dir
7400 Set the directory to search the profile data files in to @var{path}.
7401 This option affects only the profile data generated by
7402 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
7403 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
7404 and its related options.
7405 By default, GCC will use the current directory as @var{path}
7406 thus the profile data file will appear in the same directory as the object file.
7408 @item -fprofile-generate
7409 @itemx -fprofile-generate=@var{path}
7410 @opindex fprofile-generate
7412 Enable options usually used for instrumenting application to produce
7413 profile useful for later recompilation with profile feedback based
7414 optimization. You must use @option{-fprofile-generate} both when
7415 compiling and when linking your program.
7417 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
7419 If @var{path} is specified, GCC will look at the @var{path} to find
7420 the profile feedback data files. See @option{-fprofile-dir}.
7423 @itemx -fprofile-use=@var{path}
7424 @opindex fprofile-use
7425 Enable profile feedback directed optimizations, and optimizations
7426 generally profitable only with profile feedback available.
7428 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
7429 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
7431 By default, GCC emits an error message if the feedback profiles do not
7432 match the source code. This error can be turned into a warning by using
7433 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
7436 If @var{path} is specified, GCC will look at the @var{path} to find
7437 the profile feedback data files. See @option{-fprofile-dir}.
7440 The following options control compiler behavior regarding floating
7441 point arithmetic. These options trade off between speed and
7442 correctness. All must be specifically enabled.
7446 @opindex ffloat-store
7447 Do not store floating point variables in registers, and inhibit other
7448 options that might change whether a floating point value is taken from a
7451 @cindex floating point precision
7452 This option prevents undesirable excess precision on machines such as
7453 the 68000 where the floating registers (of the 68881) keep more
7454 precision than a @code{double} is supposed to have. Similarly for the
7455 x86 architecture. For most programs, the excess precision does only
7456 good, but a few programs rely on the precise definition of IEEE floating
7457 point. Use @option{-ffloat-store} for such programs, after modifying
7458 them to store all pertinent intermediate computations into variables.
7460 @item -fexcess-precision=@var{style}
7461 @opindex fexcess-precision
7462 This option allows further control over excess precision on machines
7463 where floating-point registers have more precision than the IEEE
7464 @code{float} and @code{double} types and the processor does not
7465 support operations rounding to those types. By default,
7466 @option{-fexcess-precision=fast} is in effect; this means that
7467 operations are carried out in the precision of the registers and that
7468 it is unpredictable when rounding to the types specified in the source
7469 code takes place. When compiling C, if
7470 @option{-fexcess-precision=standard} is specified then excess
7471 precision will follow the rules specified in ISO C99; in particular,
7472 both casts and assignments cause values to be rounded to their
7473 semantic types (whereas @option{-ffloat-store} only affects
7474 assignments). This option is enabled by default for C if a strict
7475 conformance option such as @option{-std=c99} is used.
7478 @option{-fexcess-precision=standard} is not implemented for languages
7479 other than C, and has no effect if
7480 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
7481 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
7482 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
7483 semantics apply without excess precision, and in the latter, rounding
7488 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
7489 @option{-ffinite-math-only}, @option{-fno-rounding-math},
7490 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
7492 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
7494 This option is not turned on by any @option{-O} option since
7495 it can result in incorrect output for programs which depend on
7496 an exact implementation of IEEE or ISO rules/specifications for
7497 math functions. It may, however, yield faster code for programs
7498 that do not require the guarantees of these specifications.
7500 @item -fno-math-errno
7501 @opindex fno-math-errno
7502 Do not set ERRNO after calling math functions that are executed
7503 with a single instruction, e.g., sqrt. A program that relies on
7504 IEEE exceptions for math error handling may want to use this flag
7505 for speed while maintaining IEEE arithmetic compatibility.
7507 This option is not turned on by any @option{-O} option since
7508 it can result in incorrect output for programs which depend on
7509 an exact implementation of IEEE or ISO rules/specifications for
7510 math functions. It may, however, yield faster code for programs
7511 that do not require the guarantees of these specifications.
7513 The default is @option{-fmath-errno}.
7515 On Darwin systems, the math library never sets @code{errno}. There is
7516 therefore no reason for the compiler to consider the possibility that
7517 it might, and @option{-fno-math-errno} is the default.
7519 @item -funsafe-math-optimizations
7520 @opindex funsafe-math-optimizations
7522 Allow optimizations for floating-point arithmetic that (a) assume
7523 that arguments and results are valid and (b) may violate IEEE or
7524 ANSI standards. When used at link-time, it may include libraries
7525 or startup files that change the default FPU control word or other
7526 similar optimizations.
7528 This option is not turned on by any @option{-O} option since
7529 it can result in incorrect output for programs which depend on
7530 an exact implementation of IEEE or ISO rules/specifications for
7531 math functions. It may, however, yield faster code for programs
7532 that do not require the guarantees of these specifications.
7533 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
7534 @option{-fassociative-math} and @option{-freciprocal-math}.
7536 The default is @option{-fno-unsafe-math-optimizations}.
7538 @item -fassociative-math
7539 @opindex fassociative-math
7541 Allow re-association of operands in series of floating-point operations.
7542 This violates the ISO C and C++ language standard by possibly changing
7543 computation result. NOTE: re-ordering may change the sign of zero as
7544 well as ignore NaNs and inhibit or create underflow or overflow (and
7545 thus cannot be used on a code which relies on rounding behavior like
7546 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
7547 and thus may not be used when ordered comparisons are required.
7548 This option requires that both @option{-fno-signed-zeros} and
7549 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
7550 much sense with @option{-frounding-math}.
7552 The default is @option{-fno-associative-math}.
7554 @item -freciprocal-math
7555 @opindex freciprocal-math
7557 Allow the reciprocal of a value to be used instead of dividing by
7558 the value if this enables optimizations. For example @code{x / y}
7559 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
7560 is subject to common subexpression elimination. Note that this loses
7561 precision and increases the number of flops operating on the value.
7563 The default is @option{-fno-reciprocal-math}.
7565 @item -ffinite-math-only
7566 @opindex ffinite-math-only
7567 Allow optimizations for floating-point arithmetic that assume
7568 that arguments and results are not NaNs or +-Infs.
7570 This option is not turned on by any @option{-O} option since
7571 it can result in incorrect output for programs which depend on
7572 an exact implementation of IEEE or ISO rules/specifications for
7573 math functions. It may, however, yield faster code for programs
7574 that do not require the guarantees of these specifications.
7576 The default is @option{-fno-finite-math-only}.
7578 @item -fno-signed-zeros
7579 @opindex fno-signed-zeros
7580 Allow optimizations for floating point arithmetic that ignore the
7581 signedness of zero. IEEE arithmetic specifies the behavior of
7582 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
7583 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
7584 This option implies that the sign of a zero result isn't significant.
7586 The default is @option{-fsigned-zeros}.
7588 @item -fno-trapping-math
7589 @opindex fno-trapping-math
7590 Compile code assuming that floating-point operations cannot generate
7591 user-visible traps. These traps include division by zero, overflow,
7592 underflow, inexact result and invalid operation. This option requires
7593 that @option{-fno-signaling-nans} be in effect. Setting this option may
7594 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
7596 This option should never be turned on by any @option{-O} option since
7597 it can result in incorrect output for programs which depend on
7598 an exact implementation of IEEE or ISO rules/specifications for
7601 The default is @option{-ftrapping-math}.
7603 @item -frounding-math
7604 @opindex frounding-math
7605 Disable transformations and optimizations that assume default floating
7606 point rounding behavior. This is round-to-zero for all floating point
7607 to integer conversions, and round-to-nearest for all other arithmetic
7608 truncations. This option should be specified for programs that change
7609 the FP rounding mode dynamically, or that may be executed with a
7610 non-default rounding mode. This option disables constant folding of
7611 floating point expressions at compile-time (which may be affected by
7612 rounding mode) and arithmetic transformations that are unsafe in the
7613 presence of sign-dependent rounding modes.
7615 The default is @option{-fno-rounding-math}.
7617 This option is experimental and does not currently guarantee to
7618 disable all GCC optimizations that are affected by rounding mode.
7619 Future versions of GCC may provide finer control of this setting
7620 using C99's @code{FENV_ACCESS} pragma. This command line option
7621 will be used to specify the default state for @code{FENV_ACCESS}.
7623 @item -fsignaling-nans
7624 @opindex fsignaling-nans
7625 Compile code assuming that IEEE signaling NaNs may generate user-visible
7626 traps during floating-point operations. Setting this option disables
7627 optimizations that may change the number of exceptions visible with
7628 signaling NaNs. This option implies @option{-ftrapping-math}.
7630 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
7633 The default is @option{-fno-signaling-nans}.
7635 This option is experimental and does not currently guarantee to
7636 disable all GCC optimizations that affect signaling NaN behavior.
7638 @item -fsingle-precision-constant
7639 @opindex fsingle-precision-constant
7640 Treat floating point constant as single precision constant instead of
7641 implicitly converting it to double precision constant.
7643 @item -fcx-limited-range
7644 @opindex fcx-limited-range
7645 When enabled, this option states that a range reduction step is not
7646 needed when performing complex division. Also, there is no checking
7647 whether the result of a complex multiplication or division is @code{NaN
7648 + I*NaN}, with an attempt to rescue the situation in that case. The
7649 default is @option{-fno-cx-limited-range}, but is enabled by
7650 @option{-ffast-math}.
7652 This option controls the default setting of the ISO C99
7653 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
7656 @item -fcx-fortran-rules
7657 @opindex fcx-fortran-rules
7658 Complex multiplication and division follow Fortran rules. Range
7659 reduction is done as part of complex division, but there is no checking
7660 whether the result of a complex multiplication or division is @code{NaN
7661 + I*NaN}, with an attempt to rescue the situation in that case.
7663 The default is @option{-fno-cx-fortran-rules}.
7667 The following options control optimizations that may improve
7668 performance, but are not enabled by any @option{-O} options. This
7669 section includes experimental options that may produce broken code.
7672 @item -fbranch-probabilities
7673 @opindex fbranch-probabilities
7674 After running a program compiled with @option{-fprofile-arcs}
7675 (@pxref{Debugging Options,, Options for Debugging Your Program or
7676 @command{gcc}}), you can compile it a second time using
7677 @option{-fbranch-probabilities}, to improve optimizations based on
7678 the number of times each branch was taken. When the program
7679 compiled with @option{-fprofile-arcs} exits it saves arc execution
7680 counts to a file called @file{@var{sourcename}.gcda} for each source
7681 file. The information in this data file is very dependent on the
7682 structure of the generated code, so you must use the same source code
7683 and the same optimization options for both compilations.
7685 With @option{-fbranch-probabilities}, GCC puts a
7686 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
7687 These can be used to improve optimization. Currently, they are only
7688 used in one place: in @file{reorg.c}, instead of guessing which path a
7689 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
7690 exactly determine which path is taken more often.
7692 @item -fprofile-values
7693 @opindex fprofile-values
7694 If combined with @option{-fprofile-arcs}, it adds code so that some
7695 data about values of expressions in the program is gathered.
7697 With @option{-fbranch-probabilities}, it reads back the data gathered
7698 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
7699 notes to instructions for their later usage in optimizations.
7701 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
7705 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
7706 a code to gather information about values of expressions.
7708 With @option{-fbranch-probabilities}, it reads back the data gathered
7709 and actually performs the optimizations based on them.
7710 Currently the optimizations include specialization of division operation
7711 using the knowledge about the value of the denominator.
7713 @item -frename-registers
7714 @opindex frename-registers
7715 Attempt to avoid false dependencies in scheduled code by making use
7716 of registers left over after register allocation. This optimization
7717 will most benefit processors with lots of registers. Depending on the
7718 debug information format adopted by the target, however, it can
7719 make debugging impossible, since variables will no longer stay in
7720 a ``home register''.
7722 Enabled by default with @option{-funroll-loops}.
7726 Perform tail duplication to enlarge superblock size. This transformation
7727 simplifies the control flow of the function allowing other optimizations to do
7730 Enabled with @option{-fprofile-use}.
7732 @item -funroll-loops
7733 @opindex funroll-loops
7734 Unroll loops whose number of iterations can be determined at compile time or
7735 upon entry to the loop. @option{-funroll-loops} implies
7736 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
7737 It also turns on complete loop peeling (i.e.@: complete removal of loops with
7738 small constant number of iterations). This option makes code larger, and may
7739 or may not make it run faster.
7741 Enabled with @option{-fprofile-use}.
7743 @item -funroll-all-loops
7744 @opindex funroll-all-loops
7745 Unroll all loops, even if their number of iterations is uncertain when
7746 the loop is entered. This usually makes programs run more slowly.
7747 @option{-funroll-all-loops} implies the same options as
7748 @option{-funroll-loops}.
7751 @opindex fpeel-loops
7752 Peels the loops for that there is enough information that they do not
7753 roll much (from profile feedback). It also turns on complete loop peeling
7754 (i.e.@: complete removal of loops with small constant number of iterations).
7756 Enabled with @option{-fprofile-use}.
7758 @item -fmove-loop-invariants
7759 @opindex fmove-loop-invariants
7760 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
7761 at level @option{-O1}
7763 @item -funswitch-loops
7764 @opindex funswitch-loops
7765 Move branches with loop invariant conditions out of the loop, with duplicates
7766 of the loop on both branches (modified according to result of the condition).
7768 @item -ffunction-sections
7769 @itemx -fdata-sections
7770 @opindex ffunction-sections
7771 @opindex fdata-sections
7772 Place each function or data item into its own section in the output
7773 file if the target supports arbitrary sections. The name of the
7774 function or the name of the data item determines the section's name
7777 Use these options on systems where the linker can perform optimizations
7778 to improve locality of reference in the instruction space. Most systems
7779 using the ELF object format and SPARC processors running Solaris 2 have
7780 linkers with such optimizations. AIX may have these optimizations in
7783 Only use these options when there are significant benefits from doing
7784 so. When you specify these options, the assembler and linker will
7785 create larger object and executable files and will also be slower.
7786 You will not be able to use @code{gprof} on all systems if you
7787 specify this option and you may have problems with debugging if
7788 you specify both this option and @option{-g}.
7790 @item -fbranch-target-load-optimize
7791 @opindex fbranch-target-load-optimize
7792 Perform branch target register load optimization before prologue / epilogue
7794 The use of target registers can typically be exposed only during reload,
7795 thus hoisting loads out of loops and doing inter-block scheduling needs
7796 a separate optimization pass.
7798 @item -fbranch-target-load-optimize2
7799 @opindex fbranch-target-load-optimize2
7800 Perform branch target register load optimization after prologue / epilogue
7803 @item -fbtr-bb-exclusive
7804 @opindex fbtr-bb-exclusive
7805 When performing branch target register load optimization, don't reuse
7806 branch target registers in within any basic block.
7808 @item -fstack-protector
7809 @opindex fstack-protector
7810 Emit extra code to check for buffer overflows, such as stack smashing
7811 attacks. This is done by adding a guard variable to functions with
7812 vulnerable objects. This includes functions that call alloca, and
7813 functions with buffers larger than 8 bytes. The guards are initialized
7814 when a function is entered and then checked when the function exits.
7815 If a guard check fails, an error message is printed and the program exits.
7817 @item -fstack-protector-all
7818 @opindex fstack-protector-all
7819 Like @option{-fstack-protector} except that all functions are protected.
7821 @item -fsection-anchors
7822 @opindex fsection-anchors
7823 Try to reduce the number of symbolic address calculations by using
7824 shared ``anchor'' symbols to address nearby objects. This transformation
7825 can help to reduce the number of GOT entries and GOT accesses on some
7828 For example, the implementation of the following function @code{foo}:
7832 int foo (void) @{ return a + b + c; @}
7835 would usually calculate the addresses of all three variables, but if you
7836 compile it with @option{-fsection-anchors}, it will access the variables
7837 from a common anchor point instead. The effect is similar to the
7838 following pseudocode (which isn't valid C):
7843 register int *xr = &x;
7844 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
7848 Not all targets support this option.
7850 @item --param @var{name}=@var{value}
7852 In some places, GCC uses various constants to control the amount of
7853 optimization that is done. For example, GCC will not inline functions
7854 that contain more that a certain number of instructions. You can
7855 control some of these constants on the command-line using the
7856 @option{--param} option.
7858 The names of specific parameters, and the meaning of the values, are
7859 tied to the internals of the compiler, and are subject to change
7860 without notice in future releases.
7862 In each case, the @var{value} is an integer. The allowable choices for
7863 @var{name} are given in the following table:
7866 @item struct-reorg-cold-struct-ratio
7867 The threshold ratio (as a percentage) between a structure frequency
7868 and the frequency of the hottest structure in the program. This parameter
7869 is used by struct-reorg optimization enabled by @option{-fipa-struct-reorg}.
7870 We say that if the ratio of a structure frequency, calculated by profiling,
7871 to the hottest structure frequency in the program is less than this
7872 parameter, then structure reorganization is not applied to this structure.
7875 @item predictable-branch-cost-outcome
7876 When branch is predicted to be taken with probability lower than this threshold
7877 (in percent), then it is considered well predictable. The default is 10.
7879 @item max-crossjump-edges
7880 The maximum number of incoming edges to consider for crossjumping.
7881 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
7882 the number of edges incoming to each block. Increasing values mean
7883 more aggressive optimization, making the compile time increase with
7884 probably small improvement in executable size.
7886 @item min-crossjump-insns
7887 The minimum number of instructions which must be matched at the end
7888 of two blocks before crossjumping will be performed on them. This
7889 value is ignored in the case where all instructions in the block being
7890 crossjumped from are matched. The default value is 5.
7892 @item max-grow-copy-bb-insns
7893 The maximum code size expansion factor when copying basic blocks
7894 instead of jumping. The expansion is relative to a jump instruction.
7895 The default value is 8.
7897 @item max-goto-duplication-insns
7898 The maximum number of instructions to duplicate to a block that jumps
7899 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
7900 passes, GCC factors computed gotos early in the compilation process,
7901 and unfactors them as late as possible. Only computed jumps at the
7902 end of a basic blocks with no more than max-goto-duplication-insns are
7903 unfactored. The default value is 8.
7905 @item max-delay-slot-insn-search
7906 The maximum number of instructions to consider when looking for an
7907 instruction to fill a delay slot. If more than this arbitrary number of
7908 instructions is searched, the time savings from filling the delay slot
7909 will be minimal so stop searching. Increasing values mean more
7910 aggressive optimization, making the compile time increase with probably
7911 small improvement in executable run time.
7913 @item max-delay-slot-live-search
7914 When trying to fill delay slots, the maximum number of instructions to
7915 consider when searching for a block with valid live register
7916 information. Increasing this arbitrarily chosen value means more
7917 aggressive optimization, increasing the compile time. This parameter
7918 should be removed when the delay slot code is rewritten to maintain the
7921 @item max-gcse-memory
7922 The approximate maximum amount of memory that will be allocated in
7923 order to perform the global common subexpression elimination
7924 optimization. If more memory than specified is required, the
7925 optimization will not be done.
7927 @item max-pending-list-length
7928 The maximum number of pending dependencies scheduling will allow
7929 before flushing the current state and starting over. Large functions
7930 with few branches or calls can create excessively large lists which
7931 needlessly consume memory and resources.
7933 @item max-inline-insns-single
7934 Several parameters control the tree inliner used in gcc.
7935 This number sets the maximum number of instructions (counted in GCC's
7936 internal representation) in a single function that the tree inliner
7937 will consider for inlining. This only affects functions declared
7938 inline and methods implemented in a class declaration (C++).
7939 The default value is 300.
7941 @item max-inline-insns-auto
7942 When you use @option{-finline-functions} (included in @option{-O3}),
7943 a lot of functions that would otherwise not be considered for inlining
7944 by the compiler will be investigated. To those functions, a different
7945 (more restrictive) limit compared to functions declared inline can
7947 The default value is 50.
7949 @item large-function-insns
7950 The limit specifying really large functions. For functions larger than this
7951 limit after inlining, inlining is constrained by
7952 @option{--param large-function-growth}. This parameter is useful primarily
7953 to avoid extreme compilation time caused by non-linear algorithms used by the
7955 The default value is 2700.
7957 @item large-function-growth
7958 Specifies maximal growth of large function caused by inlining in percents.
7959 The default value is 100 which limits large function growth to 2.0 times
7962 @item large-unit-insns
7963 The limit specifying large translation unit. Growth caused by inlining of
7964 units larger than this limit is limited by @option{--param inline-unit-growth}.
7965 For small units this might be too tight (consider unit consisting of function A
7966 that is inline and B that just calls A three time. If B is small relative to
7967 A, the growth of unit is 300\% and yet such inlining is very sane. For very
7968 large units consisting of small inlineable functions however the overall unit
7969 growth limit is needed to avoid exponential explosion of code size. Thus for
7970 smaller units, the size is increased to @option{--param large-unit-insns}
7971 before applying @option{--param inline-unit-growth}. The default is 10000
7973 @item inline-unit-growth
7974 Specifies maximal overall growth of the compilation unit caused by inlining.
7975 The default value is 30 which limits unit growth to 1.3 times the original
7978 @item ipcp-unit-growth
7979 Specifies maximal overall growth of the compilation unit caused by
7980 interprocedural constant propagation. The default value is 10 which limits
7981 unit growth to 1.1 times the original size.
7983 @item large-stack-frame
7984 The limit specifying large stack frames. While inlining the algorithm is trying
7985 to not grow past this limit too much. Default value is 256 bytes.
7987 @item large-stack-frame-growth
7988 Specifies maximal growth of large stack frames caused by inlining in percents.
7989 The default value is 1000 which limits large stack frame growth to 11 times
7992 @item max-inline-insns-recursive
7993 @itemx max-inline-insns-recursive-auto
7994 Specifies maximum number of instructions out-of-line copy of self recursive inline
7995 function can grow into by performing recursive inlining.
7997 For functions declared inline @option{--param max-inline-insns-recursive} is
7998 taken into account. For function not declared inline, recursive inlining
7999 happens only when @option{-finline-functions} (included in @option{-O3}) is
8000 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
8001 default value is 450.
8003 @item max-inline-recursive-depth
8004 @itemx max-inline-recursive-depth-auto
8005 Specifies maximum recursion depth used by the recursive inlining.
8007 For functions declared inline @option{--param max-inline-recursive-depth} is
8008 taken into account. For function not declared inline, recursive inlining
8009 happens only when @option{-finline-functions} (included in @option{-O3}) is
8010 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
8013 @item min-inline-recursive-probability
8014 Recursive inlining is profitable only for function having deep recursion
8015 in average and can hurt for function having little recursion depth by
8016 increasing the prologue size or complexity of function body to other
8019 When profile feedback is available (see @option{-fprofile-generate}) the actual
8020 recursion depth can be guessed from probability that function will recurse via
8021 given call expression. This parameter limits inlining only to call expression
8022 whose probability exceeds given threshold (in percents). The default value is
8025 @item early-inlining-insns
8026 Specify growth that early inliner can make. In effect it increases amount of
8027 inlining for code having large abstraction penalty. The default value is 8.
8029 @item max-early-inliner-iterations
8030 @itemx max-early-inliner-iterations
8031 Limit of iterations of early inliner. This basically bounds number of nested
8032 indirect calls early inliner can resolve. Deeper chains are still handled by
8035 @item min-vect-loop-bound
8036 The minimum number of iterations under which a loop will not get vectorized
8037 when @option{-ftree-vectorize} is used. The number of iterations after
8038 vectorization needs to be greater than the value specified by this option
8039 to allow vectorization. The default value is 0.
8041 @item max-unrolled-insns
8042 The maximum number of instructions that a loop should have if that loop
8043 is unrolled, and if the loop is unrolled, it determines how many times
8044 the loop code is unrolled.
8046 @item max-average-unrolled-insns
8047 The maximum number of instructions biased by probabilities of their execution
8048 that a loop should have if that loop is unrolled, and if the loop is unrolled,
8049 it determines how many times the loop code is unrolled.
8051 @item max-unroll-times
8052 The maximum number of unrollings of a single loop.
8054 @item max-peeled-insns
8055 The maximum number of instructions that a loop should have if that loop
8056 is peeled, and if the loop is peeled, it determines how many times
8057 the loop code is peeled.
8059 @item max-peel-times
8060 The maximum number of peelings of a single loop.
8062 @item max-completely-peeled-insns
8063 The maximum number of insns of a completely peeled loop.
8065 @item max-completely-peel-times
8066 The maximum number of iterations of a loop to be suitable for complete peeling.
8068 @item max-unswitch-insns
8069 The maximum number of insns of an unswitched loop.
8071 @item max-unswitch-level
8072 The maximum number of branches unswitched in a single loop.
8075 The minimum cost of an expensive expression in the loop invariant motion.
8077 @item iv-consider-all-candidates-bound
8078 Bound on number of candidates for induction variables below that
8079 all candidates are considered for each use in induction variable
8080 optimizations. Only the most relevant candidates are considered
8081 if there are more candidates, to avoid quadratic time complexity.
8083 @item iv-max-considered-uses
8084 The induction variable optimizations give up on loops that contain more
8085 induction variable uses.
8087 @item iv-always-prune-cand-set-bound
8088 If number of candidates in the set is smaller than this value,
8089 we always try to remove unnecessary ivs from the set during its
8090 optimization when a new iv is added to the set.
8092 @item scev-max-expr-size
8093 Bound on size of expressions used in the scalar evolutions analyzer.
8094 Large expressions slow the analyzer.
8096 @item omega-max-vars
8097 The maximum number of variables in an Omega constraint system.
8098 The default value is 128.
8100 @item omega-max-geqs
8101 The maximum number of inequalities in an Omega constraint system.
8102 The default value is 256.
8105 The maximum number of equalities in an Omega constraint system.
8106 The default value is 128.
8108 @item omega-max-wild-cards
8109 The maximum number of wildcard variables that the Omega solver will
8110 be able to insert. The default value is 18.
8112 @item omega-hash-table-size
8113 The size of the hash table in the Omega solver. The default value is
8116 @item omega-max-keys
8117 The maximal number of keys used by the Omega solver. The default
8120 @item omega-eliminate-redundant-constraints
8121 When set to 1, use expensive methods to eliminate all redundant
8122 constraints. The default value is 0.
8124 @item vect-max-version-for-alignment-checks
8125 The maximum number of runtime checks that can be performed when
8126 doing loop versioning for alignment in the vectorizer. See option
8127 ftree-vect-loop-version for more information.
8129 @item vect-max-version-for-alias-checks
8130 The maximum number of runtime checks that can be performed when
8131 doing loop versioning for alias in the vectorizer. See option
8132 ftree-vect-loop-version for more information.
8134 @item max-iterations-to-track
8136 The maximum number of iterations of a loop the brute force algorithm
8137 for analysis of # of iterations of the loop tries to evaluate.
8139 @item hot-bb-count-fraction
8140 Select fraction of the maximal count of repetitions of basic block in program
8141 given basic block needs to have to be considered hot.
8143 @item hot-bb-frequency-fraction
8144 Select fraction of the maximal frequency of executions of basic block in
8145 function given basic block needs to have to be considered hot
8147 @item max-predicted-iterations
8148 The maximum number of loop iterations we predict statically. This is useful
8149 in cases where function contain single loop with known bound and other loop
8150 with unknown. We predict the known number of iterations correctly, while
8151 the unknown number of iterations average to roughly 10. This means that the
8152 loop without bounds would appear artificially cold relative to the other one.
8154 @item align-threshold
8156 Select fraction of the maximal frequency of executions of basic block in
8157 function given basic block will get aligned.
8159 @item align-loop-iterations
8161 A loop expected to iterate at lest the selected number of iterations will get
8164 @item tracer-dynamic-coverage
8165 @itemx tracer-dynamic-coverage-feedback
8167 This value is used to limit superblock formation once the given percentage of
8168 executed instructions is covered. This limits unnecessary code size
8171 The @option{tracer-dynamic-coverage-feedback} is used only when profile
8172 feedback is available. The real profiles (as opposed to statically estimated
8173 ones) are much less balanced allowing the threshold to be larger value.
8175 @item tracer-max-code-growth
8176 Stop tail duplication once code growth has reached given percentage. This is
8177 rather hokey argument, as most of the duplicates will be eliminated later in
8178 cross jumping, so it may be set to much higher values than is the desired code
8181 @item tracer-min-branch-ratio
8183 Stop reverse growth when the reverse probability of best edge is less than this
8184 threshold (in percent).
8186 @item tracer-min-branch-ratio
8187 @itemx tracer-min-branch-ratio-feedback
8189 Stop forward growth if the best edge do have probability lower than this
8192 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
8193 compilation for profile feedback and one for compilation without. The value
8194 for compilation with profile feedback needs to be more conservative (higher) in
8195 order to make tracer effective.
8197 @item max-cse-path-length
8199 Maximum number of basic blocks on path that cse considers. The default is 10.
8202 The maximum instructions CSE process before flushing. The default is 1000.
8204 @item ggc-min-expand
8206 GCC uses a garbage collector to manage its own memory allocation. This
8207 parameter specifies the minimum percentage by which the garbage
8208 collector's heap should be allowed to expand between collections.
8209 Tuning this may improve compilation speed; it has no effect on code
8212 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
8213 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
8214 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
8215 GCC is not able to calculate RAM on a particular platform, the lower
8216 bound of 30% is used. Setting this parameter and
8217 @option{ggc-min-heapsize} to zero causes a full collection to occur at
8218 every opportunity. This is extremely slow, but can be useful for
8221 @item ggc-min-heapsize
8223 Minimum size of the garbage collector's heap before it begins bothering
8224 to collect garbage. The first collection occurs after the heap expands
8225 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
8226 tuning this may improve compilation speed, and has no effect on code
8229 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
8230 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
8231 with a lower bound of 4096 (four megabytes) and an upper bound of
8232 131072 (128 megabytes). If GCC is not able to calculate RAM on a
8233 particular platform, the lower bound is used. Setting this parameter
8234 very large effectively disables garbage collection. Setting this
8235 parameter and @option{ggc-min-expand} to zero causes a full collection
8236 to occur at every opportunity.
8238 @item max-reload-search-insns
8239 The maximum number of instruction reload should look backward for equivalent
8240 register. Increasing values mean more aggressive optimization, making the
8241 compile time increase with probably slightly better performance. The default
8244 @item max-cselib-memory-locations
8245 The maximum number of memory locations cselib should take into account.
8246 Increasing values mean more aggressive optimization, making the compile time
8247 increase with probably slightly better performance. The default value is 500.
8249 @item reorder-blocks-duplicate
8250 @itemx reorder-blocks-duplicate-feedback
8252 Used by basic block reordering pass to decide whether to use unconditional
8253 branch or duplicate the code on its destination. Code is duplicated when its
8254 estimated size is smaller than this value multiplied by the estimated size of
8255 unconditional jump in the hot spots of the program.
8257 The @option{reorder-block-duplicate-feedback} is used only when profile
8258 feedback is available and may be set to higher values than
8259 @option{reorder-block-duplicate} since information about the hot spots is more
8262 @item max-sched-ready-insns
8263 The maximum number of instructions ready to be issued the scheduler should
8264 consider at any given time during the first scheduling pass. Increasing
8265 values mean more thorough searches, making the compilation time increase
8266 with probably little benefit. The default value is 100.
8268 @item max-sched-region-blocks
8269 The maximum number of blocks in a region to be considered for
8270 interblock scheduling. The default value is 10.
8272 @item max-pipeline-region-blocks
8273 The maximum number of blocks in a region to be considered for
8274 pipelining in the selective scheduler. The default value is 15.
8276 @item max-sched-region-insns
8277 The maximum number of insns in a region to be considered for
8278 interblock scheduling. The default value is 100.
8280 @item max-pipeline-region-insns
8281 The maximum number of insns in a region to be considered for
8282 pipelining in the selective scheduler. The default value is 200.
8285 The minimum probability (in percents) of reaching a source block
8286 for interblock speculative scheduling. The default value is 40.
8288 @item max-sched-extend-regions-iters
8289 The maximum number of iterations through CFG to extend regions.
8290 0 - disable region extension,
8291 N - do at most N iterations.
8292 The default value is 0.
8294 @item max-sched-insn-conflict-delay
8295 The maximum conflict delay for an insn to be considered for speculative motion.
8296 The default value is 3.
8298 @item sched-spec-prob-cutoff
8299 The minimal probability of speculation success (in percents), so that
8300 speculative insn will be scheduled.
8301 The default value is 40.
8303 @item sched-mem-true-dep-cost
8304 Minimal distance (in CPU cycles) between store and load targeting same
8305 memory locations. The default value is 1.
8307 @item selsched-max-lookahead
8308 The maximum size of the lookahead window of selective scheduling. It is a
8309 depth of search for available instructions.
8310 The default value is 50.
8312 @item selsched-max-sched-times
8313 The maximum number of times that an instruction will be scheduled during
8314 selective scheduling. This is the limit on the number of iterations
8315 through which the instruction may be pipelined. The default value is 2.
8317 @item selsched-max-insns-to-rename
8318 The maximum number of best instructions in the ready list that are considered
8319 for renaming in the selective scheduler. The default value is 2.
8321 @item max-last-value-rtl
8322 The maximum size measured as number of RTLs that can be recorded in an expression
8323 in combiner for a pseudo register as last known value of that register. The default
8326 @item integer-share-limit
8327 Small integer constants can use a shared data structure, reducing the
8328 compiler's memory usage and increasing its speed. This sets the maximum
8329 value of a shared integer constant. The default value is 256.
8331 @item min-virtual-mappings
8332 Specifies the minimum number of virtual mappings in the incremental
8333 SSA updater that should be registered to trigger the virtual mappings
8334 heuristic defined by virtual-mappings-ratio. The default value is
8337 @item virtual-mappings-ratio
8338 If the number of virtual mappings is virtual-mappings-ratio bigger
8339 than the number of virtual symbols to be updated, then the incremental
8340 SSA updater switches to a full update for those symbols. The default
8343 @item ssp-buffer-size
8344 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
8345 protection when @option{-fstack-protection} is used.
8347 @item max-jump-thread-duplication-stmts
8348 Maximum number of statements allowed in a block that needs to be
8349 duplicated when threading jumps.
8351 @item max-fields-for-field-sensitive
8352 Maximum number of fields in a structure we will treat in
8353 a field sensitive manner during pointer analysis. The default is zero
8354 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
8356 @item prefetch-latency
8357 Estimate on average number of instructions that are executed before
8358 prefetch finishes. The distance we prefetch ahead is proportional
8359 to this constant. Increasing this number may also lead to less
8360 streams being prefetched (see @option{simultaneous-prefetches}).
8362 @item simultaneous-prefetches
8363 Maximum number of prefetches that can run at the same time.
8365 @item l1-cache-line-size
8366 The size of cache line in L1 cache, in bytes.
8369 The size of L1 cache, in kilobytes.
8372 The size of L2 cache, in kilobytes.
8374 @item min-insn-to-prefetch-ratio
8375 The minimum ratio between the number of instructions and the
8376 number of prefetches to enable prefetching in a loop with an
8379 @item prefetch-min-insn-to-mem-ratio
8380 The minimum ratio between the number of instructions and the
8381 number of memory references to enable prefetching in a loop.
8383 @item use-canonical-types
8384 Whether the compiler should use the ``canonical'' type system. By
8385 default, this should always be 1, which uses a more efficient internal
8386 mechanism for comparing types in C++ and Objective-C++. However, if
8387 bugs in the canonical type system are causing compilation failures,
8388 set this value to 0 to disable canonical types.
8390 @item switch-conversion-max-branch-ratio
8391 Switch initialization conversion will refuse to create arrays that are
8392 bigger than @option{switch-conversion-max-branch-ratio} times the number of
8393 branches in the switch.
8395 @item max-partial-antic-length
8396 Maximum length of the partial antic set computed during the tree
8397 partial redundancy elimination optimization (@option{-ftree-pre}) when
8398 optimizing at @option{-O3} and above. For some sorts of source code
8399 the enhanced partial redundancy elimination optimization can run away,
8400 consuming all of the memory available on the host machine. This
8401 parameter sets a limit on the length of the sets that are computed,
8402 which prevents the runaway behavior. Setting a value of 0 for
8403 this parameter will allow an unlimited set length.
8405 @item sccvn-max-scc-size
8406 Maximum size of a strongly connected component (SCC) during SCCVN
8407 processing. If this limit is hit, SCCVN processing for the whole
8408 function will not be done and optimizations depending on it will
8409 be disabled. The default maximum SCC size is 10000.
8411 @item ira-max-loops-num
8412 IRA uses a regional register allocation by default. If a function
8413 contains loops more than number given by the parameter, only at most
8414 given number of the most frequently executed loops will form regions
8415 for the regional register allocation. The default value of the
8418 @item ira-max-conflict-table-size
8419 Although IRA uses a sophisticated algorithm of compression conflict
8420 table, the table can be still big for huge functions. If the conflict
8421 table for a function could be more than size in MB given by the
8422 parameter, the conflict table is not built and faster, simpler, and
8423 lower quality register allocation algorithm will be used. The
8424 algorithm do not use pseudo-register conflicts. The default value of
8425 the parameter is 2000.
8427 @item ira-loop-reserved-regs
8428 IRA can be used to evaluate more accurate register pressure in loops
8429 for decision to move loop invariants (see @option{-O3}). The number
8430 of available registers reserved for some other purposes is described
8431 by this parameter. The default value of the parameter is 2 which is
8432 minimal number of registers needed for execution of typical
8433 instruction. This value is the best found from numerous experiments.
8435 @item loop-invariant-max-bbs-in-loop
8436 Loop invariant motion can be very expensive, both in compile time and
8437 in amount of needed compile time memory, with very large loops. Loops
8438 with more basic blocks than this parameter won't have loop invariant
8439 motion optimization performed on them. The default value of the
8440 parameter is 1000 for -O1 and 10000 for -O2 and above.
8442 @item min-nondebug-insn-uid
8443 Use uids starting at this parameter for nondebug insns. The range below
8444 the parameter is reserved exclusively for debug insns created by
8445 @option{-fvar-tracking-assignments}, but debug insns may get
8446 (non-overlapping) uids above it if the reserved range is exhausted.
8448 @item ipa-sra-ptr-growth-factor
8449 IPA-SRA will replace a pointer to an aggregate with one or more new
8450 parameters only when their cumulative size is less or equal to
8451 @option{ipa-sra-ptr-growth-factor} times the size of the original
8457 @node Preprocessor Options
8458 @section Options Controlling the Preprocessor
8459 @cindex preprocessor options
8460 @cindex options, preprocessor
8462 These options control the C preprocessor, which is run on each C source
8463 file before actual compilation.
8465 If you use the @option{-E} option, nothing is done except preprocessing.
8466 Some of these options make sense only together with @option{-E} because
8467 they cause the preprocessor output to be unsuitable for actual
8471 @item -Wp,@var{option}
8473 You can use @option{-Wp,@var{option}} to bypass the compiler driver
8474 and pass @var{option} directly through to the preprocessor. If
8475 @var{option} contains commas, it is split into multiple options at the
8476 commas. However, many options are modified, translated or interpreted
8477 by the compiler driver before being passed to the preprocessor, and
8478 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
8479 interface is undocumented and subject to change, so whenever possible
8480 you should avoid using @option{-Wp} and let the driver handle the
8483 @item -Xpreprocessor @var{option}
8484 @opindex Xpreprocessor
8485 Pass @var{option} as an option to the preprocessor. You can use this to
8486 supply system-specific preprocessor options which GCC does not know how to
8489 If you want to pass an option that takes an argument, you must use
8490 @option{-Xpreprocessor} twice, once for the option and once for the argument.
8493 @include cppopts.texi
8495 @node Assembler Options
8496 @section Passing Options to the Assembler
8498 @c prevent bad page break with this line
8499 You can pass options to the assembler.
8502 @item -Wa,@var{option}
8504 Pass @var{option} as an option to the assembler. If @var{option}
8505 contains commas, it is split into multiple options at the commas.
8507 @item -Xassembler @var{option}
8509 Pass @var{option} as an option to the assembler. You can use this to
8510 supply system-specific assembler options which GCC does not know how to
8513 If you want to pass an option that takes an argument, you must use
8514 @option{-Xassembler} twice, once for the option and once for the argument.
8519 @section Options for Linking
8520 @cindex link options
8521 @cindex options, linking
8523 These options come into play when the compiler links object files into
8524 an executable output file. They are meaningless if the compiler is
8525 not doing a link step.
8529 @item @var{object-file-name}
8530 A file name that does not end in a special recognized suffix is
8531 considered to name an object file or library. (Object files are
8532 distinguished from libraries by the linker according to the file
8533 contents.) If linking is done, these object files are used as input
8542 If any of these options is used, then the linker is not run, and
8543 object file names should not be used as arguments. @xref{Overall
8547 @item -l@var{library}
8548 @itemx -l @var{library}
8550 Search the library named @var{library} when linking. (The second
8551 alternative with the library as a separate argument is only for
8552 POSIX compliance and is not recommended.)
8554 It makes a difference where in the command you write this option; the
8555 linker searches and processes libraries and object files in the order they
8556 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
8557 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
8558 to functions in @samp{z}, those functions may not be loaded.
8560 The linker searches a standard list of directories for the library,
8561 which is actually a file named @file{lib@var{library}.a}. The linker
8562 then uses this file as if it had been specified precisely by name.
8564 The directories searched include several standard system directories
8565 plus any that you specify with @option{-L}.
8567 Normally the files found this way are library files---archive files
8568 whose members are object files. The linker handles an archive file by
8569 scanning through it for members which define symbols that have so far
8570 been referenced but not defined. But if the file that is found is an
8571 ordinary object file, it is linked in the usual fashion. The only
8572 difference between using an @option{-l} option and specifying a file name
8573 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
8574 and searches several directories.
8578 You need this special case of the @option{-l} option in order to
8579 link an Objective-C or Objective-C++ program.
8582 @opindex nostartfiles
8583 Do not use the standard system startup files when linking.
8584 The standard system libraries are used normally, unless @option{-nostdlib}
8585 or @option{-nodefaultlibs} is used.
8587 @item -nodefaultlibs
8588 @opindex nodefaultlibs
8589 Do not use the standard system libraries when linking.
8590 Only the libraries you specify will be passed to the linker, options
8591 specifying linkage of the system libraries, such as @code{-static-libgcc}
8592 or @code{-shared-libgcc}, will be ignored.
8593 The standard startup files are used normally, unless @option{-nostartfiles}
8594 is used. The compiler may generate calls to @code{memcmp},
8595 @code{memset}, @code{memcpy} and @code{memmove}.
8596 These entries are usually resolved by entries in
8597 libc. These entry points should be supplied through some other
8598 mechanism when this option is specified.
8602 Do not use the standard system startup files or libraries when linking.
8603 No startup files and only the libraries you specify will be passed to
8604 the linker, options specifying linkage of the system libraries, such as
8605 @code{-static-libgcc} or @code{-shared-libgcc}, will be ignored.
8606 The compiler may generate calls to @code{memcmp}, @code{memset},
8607 @code{memcpy} and @code{memmove}.
8608 These entries are usually resolved by entries in
8609 libc. These entry points should be supplied through some other
8610 mechanism when this option is specified.
8612 @cindex @option{-lgcc}, use with @option{-nostdlib}
8613 @cindex @option{-nostdlib} and unresolved references
8614 @cindex unresolved references and @option{-nostdlib}
8615 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
8616 @cindex @option{-nodefaultlibs} and unresolved references
8617 @cindex unresolved references and @option{-nodefaultlibs}
8618 One of the standard libraries bypassed by @option{-nostdlib} and
8619 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
8620 that GCC uses to overcome shortcomings of particular machines, or special
8621 needs for some languages.
8622 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
8623 Collection (GCC) Internals},
8624 for more discussion of @file{libgcc.a}.)
8625 In most cases, you need @file{libgcc.a} even when you want to avoid
8626 other standard libraries. In other words, when you specify @option{-nostdlib}
8627 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
8628 This ensures that you have no unresolved references to internal GCC
8629 library subroutines. (For example, @samp{__main}, used to ensure C++
8630 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
8631 GNU Compiler Collection (GCC) Internals}.)
8635 Produce a position independent executable on targets which support it.
8636 For predictable results, you must also specify the same set of options
8637 that were used to generate code (@option{-fpie}, @option{-fPIE},
8638 or model suboptions) when you specify this option.
8642 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
8643 that support it. This instructs the linker to add all symbols, not
8644 only used ones, to the dynamic symbol table. This option is needed
8645 for some uses of @code{dlopen} or to allow obtaining backtraces
8646 from within a program.
8650 Remove all symbol table and relocation information from the executable.
8654 On systems that support dynamic linking, this prevents linking with the shared
8655 libraries. On other systems, this option has no effect.
8659 Produce a shared object which can then be linked with other objects to
8660 form an executable. Not all systems support this option. For predictable
8661 results, you must also specify the same set of options that were used to
8662 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
8663 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
8664 needs to build supplementary stub code for constructors to work. On
8665 multi-libbed systems, @samp{gcc -shared} must select the correct support
8666 libraries to link against. Failing to supply the correct flags may lead
8667 to subtle defects. Supplying them in cases where they are not necessary
8670 @item -shared-libgcc
8671 @itemx -static-libgcc
8672 @opindex shared-libgcc
8673 @opindex static-libgcc
8674 On systems that provide @file{libgcc} as a shared library, these options
8675 force the use of either the shared or static version respectively.
8676 If no shared version of @file{libgcc} was built when the compiler was
8677 configured, these options have no effect.
8679 There are several situations in which an application should use the
8680 shared @file{libgcc} instead of the static version. The most common
8681 of these is when the application wishes to throw and catch exceptions
8682 across different shared libraries. In that case, each of the libraries
8683 as well as the application itself should use the shared @file{libgcc}.
8685 Therefore, the G++ and GCJ drivers automatically add
8686 @option{-shared-libgcc} whenever you build a shared library or a main
8687 executable, because C++ and Java programs typically use exceptions, so
8688 this is the right thing to do.
8690 If, instead, you use the GCC driver to create shared libraries, you may
8691 find that they will not always be linked with the shared @file{libgcc}.
8692 If GCC finds, at its configuration time, that you have a non-GNU linker
8693 or a GNU linker that does not support option @option{--eh-frame-hdr},
8694 it will link the shared version of @file{libgcc} into shared libraries
8695 by default. Otherwise, it will take advantage of the linker and optimize
8696 away the linking with the shared version of @file{libgcc}, linking with
8697 the static version of libgcc by default. This allows exceptions to
8698 propagate through such shared libraries, without incurring relocation
8699 costs at library load time.
8701 However, if a library or main executable is supposed to throw or catch
8702 exceptions, you must link it using the G++ or GCJ driver, as appropriate
8703 for the languages used in the program, or using the option
8704 @option{-shared-libgcc}, such that it is linked with the shared
8707 @item -static-libstdc++
8708 When the @command{g++} program is used to link a C++ program, it will
8709 normally automatically link against @option{libstdc++}. If
8710 @file{libstdc++} is available as a shared library, and the
8711 @option{-static} option is not used, then this will link against the
8712 shared version of @file{libstdc++}. That is normally fine. However, it
8713 is sometimes useful to freeze the version of @file{libstdc++} used by
8714 the program without going all the way to a fully static link. The
8715 @option{-static-libstdc++} option directs the @command{g++} driver to
8716 link @file{libstdc++} statically, without necessarily linking other
8717 libraries statically.
8721 Bind references to global symbols when building a shared object. Warn
8722 about any unresolved references (unless overridden by the link editor
8723 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
8726 @item -T @var{script}
8728 @cindex linker script
8729 Use @var{script} as the linker script. This option is supported by most
8730 systems using the GNU linker. On some targets, such as bare-board
8731 targets without an operating system, the @option{-T} option may be required
8732 when linking to avoid references to undefined symbols.
8734 @item -Xlinker @var{option}
8736 Pass @var{option} as an option to the linker. You can use this to
8737 supply system-specific linker options which GCC does not know how to
8740 If you want to pass an option that takes a separate argument, you must use
8741 @option{-Xlinker} twice, once for the option and once for the argument.
8742 For example, to pass @option{-assert definitions}, you must write
8743 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
8744 @option{-Xlinker "-assert definitions"}, because this passes the entire
8745 string as a single argument, which is not what the linker expects.
8747 When using the GNU linker, it is usually more convenient to pass
8748 arguments to linker options using the @option{@var{option}=@var{value}}
8749 syntax than as separate arguments. For example, you can specify
8750 @samp{-Xlinker -Map=output.map} rather than
8751 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
8752 this syntax for command-line options.
8754 @item -Wl,@var{option}
8756 Pass @var{option} as an option to the linker. If @var{option} contains
8757 commas, it is split into multiple options at the commas. You can use this
8758 syntax to pass an argument to the option.
8759 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
8760 linker. When using the GNU linker, you can also get the same effect with
8761 @samp{-Wl,-Map=output.map}.
8763 @item -u @var{symbol}
8765 Pretend the symbol @var{symbol} is undefined, to force linking of
8766 library modules to define it. You can use @option{-u} multiple times with
8767 different symbols to force loading of additional library modules.
8770 @node Directory Options
8771 @section Options for Directory Search
8772 @cindex directory options
8773 @cindex options, directory search
8776 These options specify directories to search for header files, for
8777 libraries and for parts of the compiler:
8782 Add the directory @var{dir} to the head of the list of directories to be
8783 searched for header files. This can be used to override a system header
8784 file, substituting your own version, since these directories are
8785 searched before the system header file directories. However, you should
8786 not use this option to add directories that contain vendor-supplied
8787 system header files (use @option{-isystem} for that). If you use more than
8788 one @option{-I} option, the directories are scanned in left-to-right
8789 order; the standard system directories come after.
8791 If a standard system include directory, or a directory specified with
8792 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
8793 option will be ignored. The directory will still be searched but as a
8794 system directory at its normal position in the system include chain.
8795 This is to ensure that GCC's procedure to fix buggy system headers and
8796 the ordering for the include_next directive are not inadvertently changed.
8797 If you really need to change the search order for system directories,
8798 use the @option{-nostdinc} and/or @option{-isystem} options.
8800 @item -iquote@var{dir}
8802 Add the directory @var{dir} to the head of the list of directories to
8803 be searched for header files only for the case of @samp{#include
8804 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
8805 otherwise just like @option{-I}.
8809 Add directory @var{dir} to the list of directories to be searched
8812 @item -B@var{prefix}
8814 This option specifies where to find the executables, libraries,
8815 include files, and data files of the compiler itself.
8817 The compiler driver program runs one or more of the subprograms
8818 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
8819 @var{prefix} as a prefix for each program it tries to run, both with and
8820 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
8822 For each subprogram to be run, the compiler driver first tries the
8823 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
8824 was not specified, the driver tries two standard prefixes, which are
8825 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
8826 those results in a file name that is found, the unmodified program
8827 name is searched for using the directories specified in your
8828 @env{PATH} environment variable.
8830 The compiler will check to see if the path provided by the @option{-B}
8831 refers to a directory, and if necessary it will add a directory
8832 separator character at the end of the path.
8834 @option{-B} prefixes that effectively specify directory names also apply
8835 to libraries in the linker, because the compiler translates these
8836 options into @option{-L} options for the linker. They also apply to
8837 includes files in the preprocessor, because the compiler translates these
8838 options into @option{-isystem} options for the preprocessor. In this case,
8839 the compiler appends @samp{include} to the prefix.
8841 The run-time support file @file{libgcc.a} can also be searched for using
8842 the @option{-B} prefix, if needed. If it is not found there, the two
8843 standard prefixes above are tried, and that is all. The file is left
8844 out of the link if it is not found by those means.
8846 Another way to specify a prefix much like the @option{-B} prefix is to use
8847 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
8850 As a special kludge, if the path provided by @option{-B} is
8851 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
8852 9, then it will be replaced by @file{[dir/]include}. This is to help
8853 with boot-strapping the compiler.
8855 @item -specs=@var{file}
8857 Process @var{file} after the compiler reads in the standard @file{specs}
8858 file, in order to override the defaults that the @file{gcc} driver
8859 program uses when determining what switches to pass to @file{cc1},
8860 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
8861 @option{-specs=@var{file}} can be specified on the command line, and they
8862 are processed in order, from left to right.
8864 @item --sysroot=@var{dir}
8866 Use @var{dir} as the logical root directory for headers and libraries.
8867 For example, if the compiler would normally search for headers in
8868 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
8869 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
8871 If you use both this option and the @option{-isysroot} option, then
8872 the @option{--sysroot} option will apply to libraries, but the
8873 @option{-isysroot} option will apply to header files.
8875 The GNU linker (beginning with version 2.16) has the necessary support
8876 for this option. If your linker does not support this option, the
8877 header file aspect of @option{--sysroot} will still work, but the
8878 library aspect will not.
8882 This option has been deprecated. Please use @option{-iquote} instead for
8883 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
8884 Any directories you specify with @option{-I} options before the @option{-I-}
8885 option are searched only for the case of @samp{#include "@var{file}"};
8886 they are not searched for @samp{#include <@var{file}>}.
8888 If additional directories are specified with @option{-I} options after
8889 the @option{-I-}, these directories are searched for all @samp{#include}
8890 directives. (Ordinarily @emph{all} @option{-I} directories are used
8893 In addition, the @option{-I-} option inhibits the use of the current
8894 directory (where the current input file came from) as the first search
8895 directory for @samp{#include "@var{file}"}. There is no way to
8896 override this effect of @option{-I-}. With @option{-I.} you can specify
8897 searching the directory which was current when the compiler was
8898 invoked. That is not exactly the same as what the preprocessor does
8899 by default, but it is often satisfactory.
8901 @option{-I-} does not inhibit the use of the standard system directories
8902 for header files. Thus, @option{-I-} and @option{-nostdinc} are
8909 @section Specifying subprocesses and the switches to pass to them
8912 @command{gcc} is a driver program. It performs its job by invoking a
8913 sequence of other programs to do the work of compiling, assembling and
8914 linking. GCC interprets its command-line parameters and uses these to
8915 deduce which programs it should invoke, and which command-line options
8916 it ought to place on their command lines. This behavior is controlled
8917 by @dfn{spec strings}. In most cases there is one spec string for each
8918 program that GCC can invoke, but a few programs have multiple spec
8919 strings to control their behavior. The spec strings built into GCC can
8920 be overridden by using the @option{-specs=} command-line switch to specify
8923 @dfn{Spec files} are plaintext files that are used to construct spec
8924 strings. They consist of a sequence of directives separated by blank
8925 lines. The type of directive is determined by the first non-whitespace
8926 character on the line and it can be one of the following:
8929 @item %@var{command}
8930 Issues a @var{command} to the spec file processor. The commands that can
8934 @item %include <@var{file}>
8936 Search for @var{file} and insert its text at the current point in the
8939 @item %include_noerr <@var{file}>
8940 @cindex %include_noerr
8941 Just like @samp{%include}, but do not generate an error message if the include
8942 file cannot be found.
8944 @item %rename @var{old_name} @var{new_name}
8946 Rename the spec string @var{old_name} to @var{new_name}.
8950 @item *[@var{spec_name}]:
8951 This tells the compiler to create, override or delete the named spec
8952 string. All lines after this directive up to the next directive or
8953 blank line are considered to be the text for the spec string. If this
8954 results in an empty string then the spec will be deleted. (Or, if the
8955 spec did not exist, then nothing will happened.) Otherwise, if the spec
8956 does not currently exist a new spec will be created. If the spec does
8957 exist then its contents will be overridden by the text of this
8958 directive, unless the first character of that text is the @samp{+}
8959 character, in which case the text will be appended to the spec.
8961 @item [@var{suffix}]:
8962 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
8963 and up to the next directive or blank line are considered to make up the
8964 spec string for the indicated suffix. When the compiler encounters an
8965 input file with the named suffix, it will processes the spec string in
8966 order to work out how to compile that file. For example:
8973 This says that any input file whose name ends in @samp{.ZZ} should be
8974 passed to the program @samp{z-compile}, which should be invoked with the
8975 command-line switch @option{-input} and with the result of performing the
8976 @samp{%i} substitution. (See below.)
8978 As an alternative to providing a spec string, the text that follows a
8979 suffix directive can be one of the following:
8982 @item @@@var{language}
8983 This says that the suffix is an alias for a known @var{language}. This is
8984 similar to using the @option{-x} command-line switch to GCC to specify a
8985 language explicitly. For example:
8992 Says that .ZZ files are, in fact, C++ source files.
8995 This causes an error messages saying:
8998 @var{name} compiler not installed on this system.
9002 GCC already has an extensive list of suffixes built into it.
9003 This directive will add an entry to the end of the list of suffixes, but
9004 since the list is searched from the end backwards, it is effectively
9005 possible to override earlier entries using this technique.
9009 GCC has the following spec strings built into it. Spec files can
9010 override these strings or create their own. Note that individual
9011 targets can also add their own spec strings to this list.
9014 asm Options to pass to the assembler
9015 asm_final Options to pass to the assembler post-processor
9016 cpp Options to pass to the C preprocessor
9017 cc1 Options to pass to the C compiler
9018 cc1plus Options to pass to the C++ compiler
9019 endfile Object files to include at the end of the link
9020 link Options to pass to the linker
9021 lib Libraries to include on the command line to the linker
9022 libgcc Decides which GCC support library to pass to the linker
9023 linker Sets the name of the linker
9024 predefines Defines to be passed to the C preprocessor
9025 signed_char Defines to pass to CPP to say whether @code{char} is signed
9027 startfile Object files to include at the start of the link
9030 Here is a small example of a spec file:
9036 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
9039 This example renames the spec called @samp{lib} to @samp{old_lib} and
9040 then overrides the previous definition of @samp{lib} with a new one.
9041 The new definition adds in some extra command-line options before
9042 including the text of the old definition.
9044 @dfn{Spec strings} are a list of command-line options to be passed to their
9045 corresponding program. In addition, the spec strings can contain
9046 @samp{%}-prefixed sequences to substitute variable text or to
9047 conditionally insert text into the command line. Using these constructs
9048 it is possible to generate quite complex command lines.
9050 Here is a table of all defined @samp{%}-sequences for spec
9051 strings. Note that spaces are not generated automatically around the
9052 results of expanding these sequences. Therefore you can concatenate them
9053 together or combine them with constant text in a single argument.
9057 Substitute one @samp{%} into the program name or argument.
9060 Substitute the name of the input file being processed.
9063 Substitute the basename of the input file being processed.
9064 This is the substring up to (and not including) the last period
9065 and not including the directory.
9068 This is the same as @samp{%b}, but include the file suffix (text after
9072 Marks the argument containing or following the @samp{%d} as a
9073 temporary file name, so that that file will be deleted if GCC exits
9074 successfully. Unlike @samp{%g}, this contributes no text to the
9077 @item %g@var{suffix}
9078 Substitute a file name that has suffix @var{suffix} and is chosen
9079 once per compilation, and mark the argument in the same way as
9080 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
9081 name is now chosen in a way that is hard to predict even when previously
9082 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
9083 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
9084 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
9085 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
9086 was simply substituted with a file name chosen once per compilation,
9087 without regard to any appended suffix (which was therefore treated
9088 just like ordinary text), making such attacks more likely to succeed.
9090 @item %u@var{suffix}
9091 Like @samp{%g}, but generates a new temporary file name even if
9092 @samp{%u@var{suffix}} was already seen.
9094 @item %U@var{suffix}
9095 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
9096 new one if there is no such last file name. In the absence of any
9097 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
9098 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
9099 would involve the generation of two distinct file names, one
9100 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
9101 simply substituted with a file name chosen for the previous @samp{%u},
9102 without regard to any appended suffix.
9104 @item %j@var{suffix}
9105 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
9106 writable, and if save-temps is off; otherwise, substitute the name
9107 of a temporary file, just like @samp{%u}. This temporary file is not
9108 meant for communication between processes, but rather as a junk
9111 @item %|@var{suffix}
9112 @itemx %m@var{suffix}
9113 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
9114 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
9115 all. These are the two most common ways to instruct a program that it
9116 should read from standard input or write to standard output. If you
9117 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
9118 construct: see for example @file{f/lang-specs.h}.
9120 @item %.@var{SUFFIX}
9121 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
9122 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
9123 terminated by the next space or %.
9126 Marks the argument containing or following the @samp{%w} as the
9127 designated output file of this compilation. This puts the argument
9128 into the sequence of arguments that @samp{%o} will substitute later.
9131 Substitutes the names of all the output files, with spaces
9132 automatically placed around them. You should write spaces
9133 around the @samp{%o} as well or the results are undefined.
9134 @samp{%o} is for use in the specs for running the linker.
9135 Input files whose names have no recognized suffix are not compiled
9136 at all, but they are included among the output files, so they will
9140 Substitutes the suffix for object files. Note that this is
9141 handled specially when it immediately follows @samp{%g, %u, or %U},
9142 because of the need for those to form complete file names. The
9143 handling is such that @samp{%O} is treated exactly as if it had already
9144 been substituted, except that @samp{%g, %u, and %U} do not currently
9145 support additional @var{suffix} characters following @samp{%O} as they would
9146 following, for example, @samp{.o}.
9149 Substitutes the standard macro predefinitions for the
9150 current target machine. Use this when running @code{cpp}.
9153 Like @samp{%p}, but puts @samp{__} before and after the name of each
9154 predefined macro, except for macros that start with @samp{__} or with
9155 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
9159 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
9160 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
9161 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
9162 and @option{-imultilib} as necessary.
9165 Current argument is the name of a library or startup file of some sort.
9166 Search for that file in a standard list of directories and substitute
9167 the full name found. The current working directory is included in the
9168 list of directories scanned.
9171 Current argument is the name of a linker script. Search for that file
9172 in the current list of directories to scan for libraries. If the file
9173 is located insert a @option{--script} option into the command line
9174 followed by the full path name found. If the file is not found then
9175 generate an error message. Note: the current working directory is not
9179 Print @var{str} as an error message. @var{str} is terminated by a newline.
9180 Use this when inconsistent options are detected.
9183 Substitute the contents of spec string @var{name} at this point.
9186 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
9188 @item %x@{@var{option}@}
9189 Accumulate an option for @samp{%X}.
9192 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
9196 Output the accumulated assembler options specified by @option{-Wa}.
9199 Output the accumulated preprocessor options specified by @option{-Wp}.
9202 Process the @code{asm} spec. This is used to compute the
9203 switches to be passed to the assembler.
9206 Process the @code{asm_final} spec. This is a spec string for
9207 passing switches to an assembler post-processor, if such a program is
9211 Process the @code{link} spec. This is the spec for computing the
9212 command line passed to the linker. Typically it will make use of the
9213 @samp{%L %G %S %D and %E} sequences.
9216 Dump out a @option{-L} option for each directory that GCC believes might
9217 contain startup files. If the target supports multilibs then the
9218 current multilib directory will be prepended to each of these paths.
9221 Process the @code{lib} spec. This is a spec string for deciding which
9222 libraries should be included on the command line to the linker.
9225 Process the @code{libgcc} spec. This is a spec string for deciding
9226 which GCC support library should be included on the command line to the linker.
9229 Process the @code{startfile} spec. This is a spec for deciding which
9230 object files should be the first ones passed to the linker. Typically
9231 this might be a file named @file{crt0.o}.
9234 Process the @code{endfile} spec. This is a spec string that specifies
9235 the last object files that will be passed to the linker.
9238 Process the @code{cpp} spec. This is used to construct the arguments
9239 to be passed to the C preprocessor.
9242 Process the @code{cc1} spec. This is used to construct the options to be
9243 passed to the actual C compiler (@samp{cc1}).
9246 Process the @code{cc1plus} spec. This is used to construct the options to be
9247 passed to the actual C++ compiler (@samp{cc1plus}).
9250 Substitute the variable part of a matched option. See below.
9251 Note that each comma in the substituted string is replaced by
9255 Remove all occurrences of @code{-S} from the command line. Note---this
9256 command is position dependent. @samp{%} commands in the spec string
9257 before this one will see @code{-S}, @samp{%} commands in the spec string
9258 after this one will not.
9260 @item %:@var{function}(@var{args})
9261 Call the named function @var{function}, passing it @var{args}.
9262 @var{args} is first processed as a nested spec string, then split
9263 into an argument vector in the usual fashion. The function returns
9264 a string which is processed as if it had appeared literally as part
9265 of the current spec.
9267 The following built-in spec functions are provided:
9271 The @code{getenv} spec function takes two arguments: an environment
9272 variable name and a string. If the environment variable is not
9273 defined, a fatal error is issued. Otherwise, the return value is the
9274 value of the environment variable concatenated with the string. For
9275 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
9278 %:getenv(TOPDIR /include)
9281 expands to @file{/path/to/top/include}.
9283 @item @code{if-exists}
9284 The @code{if-exists} spec function takes one argument, an absolute
9285 pathname to a file. If the file exists, @code{if-exists} returns the
9286 pathname. Here is a small example of its usage:
9290 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
9293 @item @code{if-exists-else}
9294 The @code{if-exists-else} spec function is similar to the @code{if-exists}
9295 spec function, except that it takes two arguments. The first argument is
9296 an absolute pathname to a file. If the file exists, @code{if-exists-else}
9297 returns the pathname. If it does not exist, it returns the second argument.
9298 This way, @code{if-exists-else} can be used to select one file or another,
9299 based on the existence of the first. Here is a small example of its usage:
9303 crt0%O%s %:if-exists(crti%O%s) \
9304 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
9307 @item @code{replace-outfile}
9308 The @code{replace-outfile} spec function takes two arguments. It looks for the
9309 first argument in the outfiles array and replaces it with the second argument. Here
9310 is a small example of its usage:
9313 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
9316 @item @code{print-asm-header}
9317 The @code{print-asm-header} function takes no arguments and simply
9318 prints a banner like:
9324 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
9327 It is used to separate compiler options from assembler options
9328 in the @option{--target-help} output.
9332 Substitutes the @code{-S} switch, if that switch was given to GCC@.
9333 If that switch was not specified, this substitutes nothing. Note that
9334 the leading dash is omitted when specifying this option, and it is
9335 automatically inserted if the substitution is performed. Thus the spec
9336 string @samp{%@{foo@}} would match the command-line option @option{-foo}
9337 and would output the command line option @option{-foo}.
9339 @item %W@{@code{S}@}
9340 Like %@{@code{S}@} but mark last argument supplied within as a file to be
9343 @item %@{@code{S}*@}
9344 Substitutes all the switches specified to GCC whose names start
9345 with @code{-S}, but which also take an argument. This is used for
9346 switches like @option{-o}, @option{-D}, @option{-I}, etc.
9347 GCC considers @option{-o foo} as being
9348 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
9349 text, including the space. Thus two arguments would be generated.
9351 @item %@{@code{S}*&@code{T}*@}
9352 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
9353 (the order of @code{S} and @code{T} in the spec is not significant).
9354 There can be any number of ampersand-separated variables; for each the
9355 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
9357 @item %@{@code{S}:@code{X}@}
9358 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
9360 @item %@{!@code{S}:@code{X}@}
9361 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
9363 @item %@{@code{S}*:@code{X}@}
9364 Substitutes @code{X} if one or more switches whose names start with
9365 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
9366 once, no matter how many such switches appeared. However, if @code{%*}
9367 appears somewhere in @code{X}, then @code{X} will be substituted once
9368 for each matching switch, with the @code{%*} replaced by the part of
9369 that switch that matched the @code{*}.
9371 @item %@{.@code{S}:@code{X}@}
9372 Substitutes @code{X}, if processing a file with suffix @code{S}.
9374 @item %@{!.@code{S}:@code{X}@}
9375 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
9377 @item %@{,@code{S}:@code{X}@}
9378 Substitutes @code{X}, if processing a file for language @code{S}.
9380 @item %@{!,@code{S}:@code{X}@}
9381 Substitutes @code{X}, if not processing a file for language @code{S}.
9383 @item %@{@code{S}|@code{P}:@code{X}@}
9384 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
9385 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
9386 @code{*} sequences as well, although they have a stronger binding than
9387 the @samp{|}. If @code{%*} appears in @code{X}, all of the
9388 alternatives must be starred, and only the first matching alternative
9391 For example, a spec string like this:
9394 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
9397 will output the following command-line options from the following input
9398 command-line options:
9403 -d fred.c -foo -baz -boggle
9404 -d jim.d -bar -baz -boggle
9407 @item %@{S:X; T:Y; :D@}
9409 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
9410 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
9411 be as many clauses as you need. This may be combined with @code{.},
9412 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
9417 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
9418 construct may contain other nested @samp{%} constructs or spaces, or
9419 even newlines. They are processed as usual, as described above.
9420 Trailing white space in @code{X} is ignored. White space may also
9421 appear anywhere on the left side of the colon in these constructs,
9422 except between @code{.} or @code{*} and the corresponding word.
9424 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
9425 handled specifically in these constructs. If another value of
9426 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
9427 @option{-W} switch is found later in the command line, the earlier
9428 switch value is ignored, except with @{@code{S}*@} where @code{S} is
9429 just one letter, which passes all matching options.
9431 The character @samp{|} at the beginning of the predicate text is used to
9432 indicate that a command should be piped to the following command, but
9433 only if @option{-pipe} is specified.
9435 It is built into GCC which switches take arguments and which do not.
9436 (You might think it would be useful to generalize this to allow each
9437 compiler's spec to say which switches take arguments. But this cannot
9438 be done in a consistent fashion. GCC cannot even decide which input
9439 files have been specified without knowing which switches take arguments,
9440 and it must know which input files to compile in order to tell which
9443 GCC also knows implicitly that arguments starting in @option{-l} are to be
9444 treated as compiler output files, and passed to the linker in their
9445 proper position among the other output files.
9447 @c man begin OPTIONS
9449 @node Target Options
9450 @section Specifying Target Machine and Compiler Version
9451 @cindex target options
9452 @cindex cross compiling
9453 @cindex specifying machine version
9454 @cindex specifying compiler version and target machine
9455 @cindex compiler version, specifying
9456 @cindex target machine, specifying
9458 The usual way to run GCC is to run the executable called @file{gcc}, or
9459 @file{<machine>-gcc} when cross-compiling, or
9460 @file{<machine>-gcc-<version>} to run a version other than the one that
9461 was installed last. Sometimes this is inconvenient, so GCC provides
9462 options that will switch to another cross-compiler or version.
9465 @item -b @var{machine}
9467 The argument @var{machine} specifies the target machine for compilation.
9469 The value to use for @var{machine} is the same as was specified as the
9470 machine type when configuring GCC as a cross-compiler. For
9471 example, if a cross-compiler was configured with @samp{configure
9472 arm-elf}, meaning to compile for an arm processor with elf binaries,
9473 then you would specify @option{-b arm-elf} to run that cross compiler.
9474 Because there are other options beginning with @option{-b}, the
9475 configuration must contain a hyphen, or @option{-b} alone should be one
9476 argument followed by the configuration in the next argument.
9478 @item -V @var{version}
9480 The argument @var{version} specifies which version of GCC to run.
9481 This is useful when multiple versions are installed. For example,
9482 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
9485 The @option{-V} and @option{-b} options work by running the
9486 @file{<machine>-gcc-<version>} executable, so there's no real reason to
9487 use them if you can just run that directly.
9489 @node Submodel Options
9490 @section Hardware Models and Configurations
9491 @cindex submodel options
9492 @cindex specifying hardware config
9493 @cindex hardware models and configurations, specifying
9494 @cindex machine dependent options
9496 Earlier we discussed the standard option @option{-b} which chooses among
9497 different installed compilers for completely different target
9498 machines, such as VAX vs.@: 68000 vs.@: 80386.
9500 In addition, each of these target machine types can have its own
9501 special options, starting with @samp{-m}, to choose among various
9502 hardware models or configurations---for example, 68010 vs 68020,
9503 floating coprocessor or none. A single installed version of the
9504 compiler can compile for any model or configuration, according to the
9507 Some configurations of the compiler also support additional special
9508 options, usually for compatibility with other compilers on the same
9511 @c This list is ordered alphanumerically by subsection name.
9512 @c It should be the same order and spelling as these options are listed
9513 @c in Machine Dependent Options
9519 * Blackfin Options::
9523 * DEC Alpha Options::
9524 * DEC Alpha/VMS Options::
9527 * GNU/Linux Options::
9530 * i386 and x86-64 Options::
9531 * i386 and x86-64 Windows Options::
9533 * IA-64/VMS Options::
9544 * picoChip Options::
9546 * RS/6000 and PowerPC Options::
9548 * S/390 and zSeries Options::
9553 * System V Options::
9558 * Xstormy16 Options::
9564 @subsection ARC Options
9567 These options are defined for ARC implementations:
9572 Compile code for little endian mode. This is the default.
9576 Compile code for big endian mode.
9579 @opindex mmangle-cpu
9580 Prepend the name of the cpu to all public symbol names.
9581 In multiple-processor systems, there are many ARC variants with different
9582 instruction and register set characteristics. This flag prevents code
9583 compiled for one cpu to be linked with code compiled for another.
9584 No facility exists for handling variants that are ``almost identical''.
9585 This is an all or nothing option.
9587 @item -mcpu=@var{cpu}
9589 Compile code for ARC variant @var{cpu}.
9590 Which variants are supported depend on the configuration.
9591 All variants support @option{-mcpu=base}, this is the default.
9593 @item -mtext=@var{text-section}
9594 @itemx -mdata=@var{data-section}
9595 @itemx -mrodata=@var{readonly-data-section}
9599 Put functions, data, and readonly data in @var{text-section},
9600 @var{data-section}, and @var{readonly-data-section} respectively
9601 by default. This can be overridden with the @code{section} attribute.
9602 @xref{Variable Attributes}.
9604 @item -mfix-cortex-m3-ldrd
9605 @opindex mfix-cortex-m3-ldrd
9606 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
9607 with overlapping destination and base registers are used. This option avoids
9608 generating these instructions. This option is enabled by default when
9609 @option{-mcpu=cortex-m3} is specified.
9614 @subsection ARM Options
9617 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
9621 @item -mabi=@var{name}
9623 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
9624 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
9627 @opindex mapcs-frame
9628 Generate a stack frame that is compliant with the ARM Procedure Call
9629 Standard for all functions, even if this is not strictly necessary for
9630 correct execution of the code. Specifying @option{-fomit-frame-pointer}
9631 with this option will cause the stack frames not to be generated for
9632 leaf functions. The default is @option{-mno-apcs-frame}.
9636 This is a synonym for @option{-mapcs-frame}.
9639 @c not currently implemented
9640 @item -mapcs-stack-check
9641 @opindex mapcs-stack-check
9642 Generate code to check the amount of stack space available upon entry to
9643 every function (that actually uses some stack space). If there is
9644 insufficient space available then either the function
9645 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
9646 called, depending upon the amount of stack space required. The run time
9647 system is required to provide these functions. The default is
9648 @option{-mno-apcs-stack-check}, since this produces smaller code.
9650 @c not currently implemented
9652 @opindex mapcs-float
9653 Pass floating point arguments using the float point registers. This is
9654 one of the variants of the APCS@. This option is recommended if the
9655 target hardware has a floating point unit or if a lot of floating point
9656 arithmetic is going to be performed by the code. The default is
9657 @option{-mno-apcs-float}, since integer only code is slightly increased in
9658 size if @option{-mapcs-float} is used.
9660 @c not currently implemented
9661 @item -mapcs-reentrant
9662 @opindex mapcs-reentrant
9663 Generate reentrant, position independent code. The default is
9664 @option{-mno-apcs-reentrant}.
9667 @item -mthumb-interwork
9668 @opindex mthumb-interwork
9669 Generate code which supports calling between the ARM and Thumb
9670 instruction sets. Without this option the two instruction sets cannot
9671 be reliably used inside one program. The default is
9672 @option{-mno-thumb-interwork}, since slightly larger code is generated
9673 when @option{-mthumb-interwork} is specified.
9675 @item -mno-sched-prolog
9676 @opindex mno-sched-prolog
9677 Prevent the reordering of instructions in the function prolog, or the
9678 merging of those instruction with the instructions in the function's
9679 body. This means that all functions will start with a recognizable set
9680 of instructions (or in fact one of a choice from a small set of
9681 different function prologues), and this information can be used to
9682 locate the start if functions inside an executable piece of code. The
9683 default is @option{-msched-prolog}.
9685 @item -mfloat-abi=@var{name}
9687 Specifies which floating-point ABI to use. Permissible values
9688 are: @samp{soft}, @samp{softfp} and @samp{hard}.
9690 Specifying @samp{soft} causes GCC to generate output containing
9691 library calls for floating-point operations.
9692 @samp{softfp} allows the generation of code using hardware floating-point
9693 instructions, but still uses the soft-float calling conventions.
9694 @samp{hard} allows generation of floating-point instructions
9695 and uses FPU-specific calling conventions.
9697 The default depends on the specific target configuration. Note that
9698 the hard-float and soft-float ABIs are not link-compatible; you must
9699 compile your entire program with the same ABI, and link with a
9700 compatible set of libraries.
9703 @opindex mhard-float
9704 Equivalent to @option{-mfloat-abi=hard}.
9707 @opindex msoft-float
9708 Equivalent to @option{-mfloat-abi=soft}.
9710 @item -mlittle-endian
9711 @opindex mlittle-endian
9712 Generate code for a processor running in little-endian mode. This is
9713 the default for all standard configurations.
9716 @opindex mbig-endian
9717 Generate code for a processor running in big-endian mode; the default is
9718 to compile code for a little-endian processor.
9720 @item -mwords-little-endian
9721 @opindex mwords-little-endian
9722 This option only applies when generating code for big-endian processors.
9723 Generate code for a little-endian word order but a big-endian byte
9724 order. That is, a byte order of the form @samp{32107654}. Note: this
9725 option should only be used if you require compatibility with code for
9726 big-endian ARM processors generated by versions of the compiler prior to
9729 @item -mcpu=@var{name}
9731 This specifies the name of the target ARM processor. GCC uses this name
9732 to determine what kind of instructions it can emit when generating
9733 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
9734 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
9735 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
9736 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
9737 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
9739 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
9740 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
9741 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
9742 @samp{strongarm1110},
9743 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
9744 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
9745 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
9746 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
9747 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
9748 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
9749 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
9750 @samp{cortex-a8}, @samp{cortex-a9},
9751 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-m3},
9754 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9756 @item -mtune=@var{name}
9758 This option is very similar to the @option{-mcpu=} option, except that
9759 instead of specifying the actual target processor type, and hence
9760 restricting which instructions can be used, it specifies that GCC should
9761 tune the performance of the code as if the target were of the type
9762 specified in this option, but still choosing the instructions that it
9763 will generate based on the cpu specified by a @option{-mcpu=} option.
9764 For some ARM implementations better performance can be obtained by using
9767 @item -march=@var{name}
9769 This specifies the name of the target ARM architecture. GCC uses this
9770 name to determine what kind of instructions it can emit when generating
9771 assembly code. This option can be used in conjunction with or instead
9772 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
9773 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
9774 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
9775 @samp{armv6}, @samp{armv6j},
9776 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
9777 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
9778 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9780 @item -mfpu=@var{name}
9781 @itemx -mfpe=@var{number}
9782 @itemx -mfp=@var{number}
9786 This specifies what floating point hardware (or hardware emulation) is
9787 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
9788 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-d16},
9789 @samp{neon}, and @samp{neon-fp16}. @option{-mfp} and @option{-mfpe}
9790 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
9791 with older versions of GCC@.
9793 If @option{-msoft-float} is specified this specifies the format of
9794 floating point values.
9796 @item -mfp16-format=@var{name}
9797 @opindex mfp16-format
9798 Specify the format of the @code{__fp16} half-precision floating-point type.
9799 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
9800 the default is @samp{none}, in which case the @code{__fp16} type is not
9801 defined. @xref{Half-Precision}, for more information.
9803 @item -mstructure-size-boundary=@var{n}
9804 @opindex mstructure-size-boundary
9805 The size of all structures and unions will be rounded up to a multiple
9806 of the number of bits set by this option. Permissible values are 8, 32
9807 and 64. The default value varies for different toolchains. For the COFF
9808 targeted toolchain the default value is 8. A value of 64 is only allowed
9809 if the underlying ABI supports it.
9811 Specifying the larger number can produce faster, more efficient code, but
9812 can also increase the size of the program. Different values are potentially
9813 incompatible. Code compiled with one value cannot necessarily expect to
9814 work with code or libraries compiled with another value, if they exchange
9815 information using structures or unions.
9817 @item -mabort-on-noreturn
9818 @opindex mabort-on-noreturn
9819 Generate a call to the function @code{abort} at the end of a
9820 @code{noreturn} function. It will be executed if the function tries to
9824 @itemx -mno-long-calls
9825 @opindex mlong-calls
9826 @opindex mno-long-calls
9827 Tells the compiler to perform function calls by first loading the
9828 address of the function into a register and then performing a subroutine
9829 call on this register. This switch is needed if the target function
9830 will lie outside of the 64 megabyte addressing range of the offset based
9831 version of subroutine call instruction.
9833 Even if this switch is enabled, not all function calls will be turned
9834 into long calls. The heuristic is that static functions, functions
9835 which have the @samp{short-call} attribute, functions that are inside
9836 the scope of a @samp{#pragma no_long_calls} directive and functions whose
9837 definitions have already been compiled within the current compilation
9838 unit, will not be turned into long calls. The exception to this rule is
9839 that weak function definitions, functions with the @samp{long-call}
9840 attribute or the @samp{section} attribute, and functions that are within
9841 the scope of a @samp{#pragma long_calls} directive, will always be
9842 turned into long calls.
9844 This feature is not enabled by default. Specifying
9845 @option{-mno-long-calls} will restore the default behavior, as will
9846 placing the function calls within the scope of a @samp{#pragma
9847 long_calls_off} directive. Note these switches have no effect on how
9848 the compiler generates code to handle function calls via function
9851 @item -msingle-pic-base
9852 @opindex msingle-pic-base
9853 Treat the register used for PIC addressing as read-only, rather than
9854 loading it in the prologue for each function. The run-time system is
9855 responsible for initializing this register with an appropriate value
9856 before execution begins.
9858 @item -mpic-register=@var{reg}
9859 @opindex mpic-register
9860 Specify the register to be used for PIC addressing. The default is R10
9861 unless stack-checking is enabled, when R9 is used.
9863 @item -mcirrus-fix-invalid-insns
9864 @opindex mcirrus-fix-invalid-insns
9865 @opindex mno-cirrus-fix-invalid-insns
9866 Insert NOPs into the instruction stream to in order to work around
9867 problems with invalid Maverick instruction combinations. This option
9868 is only valid if the @option{-mcpu=ep9312} option has been used to
9869 enable generation of instructions for the Cirrus Maverick floating
9870 point co-processor. This option is not enabled by default, since the
9871 problem is only present in older Maverick implementations. The default
9872 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
9875 @item -mpoke-function-name
9876 @opindex mpoke-function-name
9877 Write the name of each function into the text section, directly
9878 preceding the function prologue. The generated code is similar to this:
9882 .ascii "arm_poke_function_name", 0
9885 .word 0xff000000 + (t1 - t0)
9886 arm_poke_function_name
9888 stmfd sp!, @{fp, ip, lr, pc@}
9892 When performing a stack backtrace, code can inspect the value of
9893 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
9894 location @code{pc - 12} and the top 8 bits are set, then we know that
9895 there is a function name embedded immediately preceding this location
9896 and has length @code{((pc[-3]) & 0xff000000)}.
9900 Generate code for the Thumb instruction set. The default is to
9901 use the 32-bit ARM instruction set.
9902 This option automatically enables either 16-bit Thumb-1 or
9903 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
9904 and @option{-march=@var{name}} options. This option is not passed to the
9905 assembler. If you want to force assembler files to be interpreted as Thumb code,
9906 either add a @samp{.thumb} directive to the source or pass the @option{-mthumb}
9907 option directly to the assembler by prefixing it with @option{-Wa}.
9910 @opindex mtpcs-frame
9911 Generate a stack frame that is compliant with the Thumb Procedure Call
9912 Standard for all non-leaf functions. (A leaf function is one that does
9913 not call any other functions.) The default is @option{-mno-tpcs-frame}.
9915 @item -mtpcs-leaf-frame
9916 @opindex mtpcs-leaf-frame
9917 Generate a stack frame that is compliant with the Thumb Procedure Call
9918 Standard for all leaf functions. (A leaf function is one that does
9919 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
9921 @item -mcallee-super-interworking
9922 @opindex mcallee-super-interworking
9923 Gives all externally visible functions in the file being compiled an ARM
9924 instruction set header which switches to Thumb mode before executing the
9925 rest of the function. This allows these functions to be called from
9926 non-interworking code. This option is not valid in AAPCS configurations
9927 because interworking is enabled by default.
9929 @item -mcaller-super-interworking
9930 @opindex mcaller-super-interworking
9931 Allows calls via function pointers (including virtual functions) to
9932 execute correctly regardless of whether the target code has been
9933 compiled for interworking or not. There is a small overhead in the cost
9934 of executing a function pointer if this option is enabled. This option
9935 is not valid in AAPCS configurations because interworking is enabled
9938 @item -mtp=@var{name}
9940 Specify the access model for the thread local storage pointer. The valid
9941 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
9942 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
9943 (supported in the arm6k architecture), and @option{auto}, which uses the
9944 best available method for the selected processor. The default setting is
9947 @item -mword-relocations
9948 @opindex mword-relocations
9949 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
9950 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
9951 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
9957 @subsection AVR Options
9960 These options are defined for AVR implementations:
9963 @item -mmcu=@var{mcu}
9965 Specify ATMEL AVR instruction set or MCU type.
9967 Instruction set avr1 is for the minimal AVR core, not supported by the C
9968 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
9969 attiny11, attiny12, attiny15, attiny28).
9971 Instruction set avr2 (default) is for the classic AVR core with up to
9972 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
9973 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
9974 at90c8534, at90s8535).
9976 Instruction set avr3 is for the classic AVR core with up to 128K program
9977 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
9979 Instruction set avr4 is for the enhanced AVR core with up to 8K program
9980 memory space (MCU types: atmega8, atmega83, atmega85).
9982 Instruction set avr5 is for the enhanced AVR core with up to 128K program
9983 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
9984 atmega64, atmega128, at43usb355, at94k).
9986 @item -mno-interrupts
9987 @opindex mno-interrupts
9988 Generated code is not compatible with hardware interrupts.
9989 Code size will be smaller.
9991 @item -mcall-prologues
9992 @opindex mcall-prologues
9993 Functions prologues/epilogues expanded as call to appropriate
9994 subroutines. Code size will be smaller.
9997 @opindex mtiny-stack
9998 Change only the low 8 bits of the stack pointer.
10002 Assume int to be 8 bit integer. This affects the sizes of all types: A
10003 char will be 1 byte, an int will be 1 byte, a long will be 2 bytes
10004 and long long will be 4 bytes. Please note that this option does not
10005 comply to the C standards, but it will provide you with smaller code
10009 @node Blackfin Options
10010 @subsection Blackfin Options
10011 @cindex Blackfin Options
10014 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
10016 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
10017 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
10018 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
10019 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
10020 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
10021 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
10022 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
10024 The optional @var{sirevision} specifies the silicon revision of the target
10025 Blackfin processor. Any workarounds available for the targeted silicon revision
10026 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
10027 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
10028 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
10029 hexadecimal digits representing the major and minor numbers in the silicon
10030 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
10031 is not defined. If @var{sirevision} is @samp{any}, the
10032 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
10033 If this optional @var{sirevision} is not used, GCC assumes the latest known
10034 silicon revision of the targeted Blackfin processor.
10036 Support for @samp{bf561} is incomplete. For @samp{bf561},
10037 Only the processor macro is defined.
10038 Without this option, @samp{bf532} is used as the processor by default.
10039 The corresponding predefined processor macros for @var{cpu} is to
10040 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
10041 provided by libgloss to be linked in if @option{-msim} is not given.
10045 Specifies that the program will be run on the simulator. This causes
10046 the simulator BSP provided by libgloss to be linked in. This option
10047 has effect only for @samp{bfin-elf} toolchain.
10048 Certain other options, such as @option{-mid-shared-library} and
10049 @option{-mfdpic}, imply @option{-msim}.
10051 @item -momit-leaf-frame-pointer
10052 @opindex momit-leaf-frame-pointer
10053 Don't keep the frame pointer in a register for leaf functions. This
10054 avoids the instructions to save, set up and restore frame pointers and
10055 makes an extra register available in leaf functions. The option
10056 @option{-fomit-frame-pointer} removes the frame pointer for all functions
10057 which might make debugging harder.
10059 @item -mspecld-anomaly
10060 @opindex mspecld-anomaly
10061 When enabled, the compiler will ensure that the generated code does not
10062 contain speculative loads after jump instructions. If this option is used,
10063 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
10065 @item -mno-specld-anomaly
10066 @opindex mno-specld-anomaly
10067 Don't generate extra code to prevent speculative loads from occurring.
10069 @item -mcsync-anomaly
10070 @opindex mcsync-anomaly
10071 When enabled, the compiler will ensure that the generated code does not
10072 contain CSYNC or SSYNC instructions too soon after conditional branches.
10073 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
10075 @item -mno-csync-anomaly
10076 @opindex mno-csync-anomaly
10077 Don't generate extra code to prevent CSYNC or SSYNC instructions from
10078 occurring too soon after a conditional branch.
10082 When enabled, the compiler is free to take advantage of the knowledge that
10083 the entire program fits into the low 64k of memory.
10086 @opindex mno-low-64k
10087 Assume that the program is arbitrarily large. This is the default.
10089 @item -mstack-check-l1
10090 @opindex mstack-check-l1
10091 Do stack checking using information placed into L1 scratchpad memory by the
10094 @item -mid-shared-library
10095 @opindex mid-shared-library
10096 Generate code that supports shared libraries via the library ID method.
10097 This allows for execute in place and shared libraries in an environment
10098 without virtual memory management. This option implies @option{-fPIC}.
10099 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10101 @item -mno-id-shared-library
10102 @opindex mno-id-shared-library
10103 Generate code that doesn't assume ID based shared libraries are being used.
10104 This is the default.
10106 @item -mleaf-id-shared-library
10107 @opindex mleaf-id-shared-library
10108 Generate code that supports shared libraries via the library ID method,
10109 but assumes that this library or executable won't link against any other
10110 ID shared libraries. That allows the compiler to use faster code for jumps
10113 @item -mno-leaf-id-shared-library
10114 @opindex mno-leaf-id-shared-library
10115 Do not assume that the code being compiled won't link against any ID shared
10116 libraries. Slower code will be generated for jump and call insns.
10118 @item -mshared-library-id=n
10119 @opindex mshared-library-id
10120 Specified the identification number of the ID based shared library being
10121 compiled. Specifying a value of 0 will generate more compact code, specifying
10122 other values will force the allocation of that number to the current
10123 library but is no more space or time efficient than omitting this option.
10127 Generate code that allows the data segment to be located in a different
10128 area of memory from the text segment. This allows for execute in place in
10129 an environment without virtual memory management by eliminating relocations
10130 against the text section.
10132 @item -mno-sep-data
10133 @opindex mno-sep-data
10134 Generate code that assumes that the data segment follows the text segment.
10135 This is the default.
10138 @itemx -mno-long-calls
10139 @opindex mlong-calls
10140 @opindex mno-long-calls
10141 Tells the compiler to perform function calls by first loading the
10142 address of the function into a register and then performing a subroutine
10143 call on this register. This switch is needed if the target function
10144 will lie outside of the 24 bit addressing range of the offset based
10145 version of subroutine call instruction.
10147 This feature is not enabled by default. Specifying
10148 @option{-mno-long-calls} will restore the default behavior. Note these
10149 switches have no effect on how the compiler generates code to handle
10150 function calls via function pointers.
10154 Link with the fast floating-point library. This library relaxes some of
10155 the IEEE floating-point standard's rules for checking inputs against
10156 Not-a-Number (NAN), in the interest of performance.
10159 @opindex minline-plt
10160 Enable inlining of PLT entries in function calls to functions that are
10161 not known to bind locally. It has no effect without @option{-mfdpic}.
10164 @opindex mmulticore
10165 Build standalone application for multicore Blackfin processor. Proper
10166 start files and link scripts will be used to support multicore.
10167 This option defines @code{__BFIN_MULTICORE}. It can only be used with
10168 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
10169 @option{-mcorea} or @option{-mcoreb}. If it's used without
10170 @option{-mcorea} or @option{-mcoreb}, single application/dual core
10171 programming model is used. In this model, the main function of Core B
10172 should be named as coreb_main. If it's used with @option{-mcorea} or
10173 @option{-mcoreb}, one application per core programming model is used.
10174 If this option is not used, single core application programming
10179 Build standalone application for Core A of BF561 when using
10180 one application per core programming model. Proper start files
10181 and link scripts will be used to support Core A. This option
10182 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
10186 Build standalone application for Core B of BF561 when using
10187 one application per core programming model. Proper start files
10188 and link scripts will be used to support Core B. This option
10189 defines @code{__BFIN_COREB}. When this option is used, coreb_main
10190 should be used instead of main. It must be used with
10191 @option{-mmulticore}.
10195 Build standalone application for SDRAM. Proper start files and
10196 link scripts will be used to put the application into SDRAM.
10197 Loader should initialize SDRAM before loading the application
10198 into SDRAM. This option defines @code{__BFIN_SDRAM}.
10202 Assume that ICPLBs are enabled at runtime. This has an effect on certain
10203 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
10204 are enabled; for standalone applications the default is off.
10208 @subsection CRIS Options
10209 @cindex CRIS Options
10211 These options are defined specifically for the CRIS ports.
10214 @item -march=@var{architecture-type}
10215 @itemx -mcpu=@var{architecture-type}
10218 Generate code for the specified architecture. The choices for
10219 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
10220 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
10221 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
10224 @item -mtune=@var{architecture-type}
10226 Tune to @var{architecture-type} everything applicable about the generated
10227 code, except for the ABI and the set of available instructions. The
10228 choices for @var{architecture-type} are the same as for
10229 @option{-march=@var{architecture-type}}.
10231 @item -mmax-stack-frame=@var{n}
10232 @opindex mmax-stack-frame
10233 Warn when the stack frame of a function exceeds @var{n} bytes.
10239 The options @option{-metrax4} and @option{-metrax100} are synonyms for
10240 @option{-march=v3} and @option{-march=v8} respectively.
10242 @item -mmul-bug-workaround
10243 @itemx -mno-mul-bug-workaround
10244 @opindex mmul-bug-workaround
10245 @opindex mno-mul-bug-workaround
10246 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
10247 models where it applies. This option is active by default.
10251 Enable CRIS-specific verbose debug-related information in the assembly
10252 code. This option also has the effect to turn off the @samp{#NO_APP}
10253 formatted-code indicator to the assembler at the beginning of the
10258 Do not use condition-code results from previous instruction; always emit
10259 compare and test instructions before use of condition codes.
10261 @item -mno-side-effects
10262 @opindex mno-side-effects
10263 Do not emit instructions with side-effects in addressing modes other than
10266 @item -mstack-align
10267 @itemx -mno-stack-align
10268 @itemx -mdata-align
10269 @itemx -mno-data-align
10270 @itemx -mconst-align
10271 @itemx -mno-const-align
10272 @opindex mstack-align
10273 @opindex mno-stack-align
10274 @opindex mdata-align
10275 @opindex mno-data-align
10276 @opindex mconst-align
10277 @opindex mno-const-align
10278 These options (no-options) arranges (eliminate arrangements) for the
10279 stack-frame, individual data and constants to be aligned for the maximum
10280 single data access size for the chosen CPU model. The default is to
10281 arrange for 32-bit alignment. ABI details such as structure layout are
10282 not affected by these options.
10290 Similar to the stack- data- and const-align options above, these options
10291 arrange for stack-frame, writable data and constants to all be 32-bit,
10292 16-bit or 8-bit aligned. The default is 32-bit alignment.
10294 @item -mno-prologue-epilogue
10295 @itemx -mprologue-epilogue
10296 @opindex mno-prologue-epilogue
10297 @opindex mprologue-epilogue
10298 With @option{-mno-prologue-epilogue}, the normal function prologue and
10299 epilogue that sets up the stack-frame are omitted and no return
10300 instructions or return sequences are generated in the code. Use this
10301 option only together with visual inspection of the compiled code: no
10302 warnings or errors are generated when call-saved registers must be saved,
10303 or storage for local variable needs to be allocated.
10307 @opindex mno-gotplt
10309 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
10310 instruction sequences that load addresses for functions from the PLT part
10311 of the GOT rather than (traditional on other architectures) calls to the
10312 PLT@. The default is @option{-mgotplt}.
10316 Legacy no-op option only recognized with the cris-axis-elf and
10317 cris-axis-linux-gnu targets.
10321 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
10325 This option, recognized for the cris-axis-elf arranges
10326 to link with input-output functions from a simulator library. Code,
10327 initialized data and zero-initialized data are allocated consecutively.
10331 Like @option{-sim}, but pass linker options to locate initialized data at
10332 0x40000000 and zero-initialized data at 0x80000000.
10336 @subsection CRX Options
10337 @cindex CRX Options
10339 These options are defined specifically for the CRX ports.
10345 Enable the use of multiply-accumulate instructions. Disabled by default.
10348 @opindex mpush-args
10349 Push instructions will be used to pass outgoing arguments when functions
10350 are called. Enabled by default.
10353 @node Darwin Options
10354 @subsection Darwin Options
10355 @cindex Darwin options
10357 These options are defined for all architectures running the Darwin operating
10360 FSF GCC on Darwin does not create ``fat'' object files; it will create
10361 an object file for the single architecture that it was built to
10362 target. Apple's GCC on Darwin does create ``fat'' files if multiple
10363 @option{-arch} options are used; it does so by running the compiler or
10364 linker multiple times and joining the results together with
10367 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
10368 @samp{i686}) is determined by the flags that specify the ISA
10369 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
10370 @option{-force_cpusubtype_ALL} option can be used to override this.
10372 The Darwin tools vary in their behavior when presented with an ISA
10373 mismatch. The assembler, @file{as}, will only permit instructions to
10374 be used that are valid for the subtype of the file it is generating,
10375 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
10376 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
10377 and print an error if asked to create a shared library with a less
10378 restrictive subtype than its input files (for instance, trying to put
10379 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
10380 for executables, @file{ld}, will quietly give the executable the most
10381 restrictive subtype of any of its input files.
10386 Add the framework directory @var{dir} to the head of the list of
10387 directories to be searched for header files. These directories are
10388 interleaved with those specified by @option{-I} options and are
10389 scanned in a left-to-right order.
10391 A framework directory is a directory with frameworks in it. A
10392 framework is a directory with a @samp{"Headers"} and/or
10393 @samp{"PrivateHeaders"} directory contained directly in it that ends
10394 in @samp{".framework"}. The name of a framework is the name of this
10395 directory excluding the @samp{".framework"}. Headers associated with
10396 the framework are found in one of those two directories, with
10397 @samp{"Headers"} being searched first. A subframework is a framework
10398 directory that is in a framework's @samp{"Frameworks"} directory.
10399 Includes of subframework headers can only appear in a header of a
10400 framework that contains the subframework, or in a sibling subframework
10401 header. Two subframeworks are siblings if they occur in the same
10402 framework. A subframework should not have the same name as a
10403 framework, a warning will be issued if this is violated. Currently a
10404 subframework cannot have subframeworks, in the future, the mechanism
10405 may be extended to support this. The standard frameworks can be found
10406 in @samp{"/System/Library/Frameworks"} and
10407 @samp{"/Library/Frameworks"}. An example include looks like
10408 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
10409 the name of the framework and header.h is found in the
10410 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
10412 @item -iframework@var{dir}
10413 @opindex iframework
10414 Like @option{-F} except the directory is a treated as a system
10415 directory. The main difference between this @option{-iframework} and
10416 @option{-F} is that with @option{-iframework} the compiler does not
10417 warn about constructs contained within header files found via
10418 @var{dir}. This option is valid only for the C family of languages.
10422 Emit debugging information for symbols that are used. For STABS
10423 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
10424 This is by default ON@.
10428 Emit debugging information for all symbols and types.
10430 @item -mmacosx-version-min=@var{version}
10431 The earliest version of MacOS X that this executable will run on
10432 is @var{version}. Typical values of @var{version} include @code{10.1},
10433 @code{10.2}, and @code{10.3.9}.
10435 If the compiler was built to use the system's headers by default,
10436 then the default for this option is the system version on which the
10437 compiler is running, otherwise the default is to make choices which
10438 are compatible with as many systems and code bases as possible.
10442 Enable kernel development mode. The @option{-mkernel} option sets
10443 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
10444 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
10445 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
10446 applicable. This mode also sets @option{-mno-altivec},
10447 @option{-msoft-float}, @option{-fno-builtin} and
10448 @option{-mlong-branch} for PowerPC targets.
10450 @item -mone-byte-bool
10451 @opindex mone-byte-bool
10452 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
10453 By default @samp{sizeof(bool)} is @samp{4} when compiling for
10454 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
10455 option has no effect on x86.
10457 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
10458 to generate code that is not binary compatible with code generated
10459 without that switch. Using this switch may require recompiling all
10460 other modules in a program, including system libraries. Use this
10461 switch to conform to a non-default data model.
10463 @item -mfix-and-continue
10464 @itemx -ffix-and-continue
10465 @itemx -findirect-data
10466 @opindex mfix-and-continue
10467 @opindex ffix-and-continue
10468 @opindex findirect-data
10469 Generate code suitable for fast turn around development. Needed to
10470 enable gdb to dynamically load @code{.o} files into already running
10471 programs. @option{-findirect-data} and @option{-ffix-and-continue}
10472 are provided for backwards compatibility.
10476 Loads all members of static archive libraries.
10477 See man ld(1) for more information.
10479 @item -arch_errors_fatal
10480 @opindex arch_errors_fatal
10481 Cause the errors having to do with files that have the wrong architecture
10484 @item -bind_at_load
10485 @opindex bind_at_load
10486 Causes the output file to be marked such that the dynamic linker will
10487 bind all undefined references when the file is loaded or launched.
10491 Produce a Mach-o bundle format file.
10492 See man ld(1) for more information.
10494 @item -bundle_loader @var{executable}
10495 @opindex bundle_loader
10496 This option specifies the @var{executable} that will be loading the build
10497 output file being linked. See man ld(1) for more information.
10500 @opindex dynamiclib
10501 When passed this option, GCC will produce a dynamic library instead of
10502 an executable when linking, using the Darwin @file{libtool} command.
10504 @item -force_cpusubtype_ALL
10505 @opindex force_cpusubtype_ALL
10506 This causes GCC's output file to have the @var{ALL} subtype, instead of
10507 one controlled by the @option{-mcpu} or @option{-march} option.
10509 @item -allowable_client @var{client_name}
10510 @itemx -client_name
10511 @itemx -compatibility_version
10512 @itemx -current_version
10514 @itemx -dependency-file
10516 @itemx -dylinker_install_name
10518 @itemx -exported_symbols_list
10520 @itemx -flat_namespace
10521 @itemx -force_flat_namespace
10522 @itemx -headerpad_max_install_names
10525 @itemx -install_name
10526 @itemx -keep_private_externs
10527 @itemx -multi_module
10528 @itemx -multiply_defined
10529 @itemx -multiply_defined_unused
10531 @itemx -no_dead_strip_inits_and_terms
10532 @itemx -nofixprebinding
10533 @itemx -nomultidefs
10535 @itemx -noseglinkedit
10536 @itemx -pagezero_size
10538 @itemx -prebind_all_twolevel_modules
10539 @itemx -private_bundle
10540 @itemx -read_only_relocs
10542 @itemx -sectobjectsymbols
10546 @itemx -sectobjectsymbols
10549 @itemx -segs_read_only_addr
10550 @itemx -segs_read_write_addr
10551 @itemx -seg_addr_table
10552 @itemx -seg_addr_table_filename
10553 @itemx -seglinkedit
10555 @itemx -segs_read_only_addr
10556 @itemx -segs_read_write_addr
10557 @itemx -single_module
10559 @itemx -sub_library
10560 @itemx -sub_umbrella
10561 @itemx -twolevel_namespace
10564 @itemx -unexported_symbols_list
10565 @itemx -weak_reference_mismatches
10566 @itemx -whatsloaded
10567 @opindex allowable_client
10568 @opindex client_name
10569 @opindex compatibility_version
10570 @opindex current_version
10571 @opindex dead_strip
10572 @opindex dependency-file
10573 @opindex dylib_file
10574 @opindex dylinker_install_name
10576 @opindex exported_symbols_list
10578 @opindex flat_namespace
10579 @opindex force_flat_namespace
10580 @opindex headerpad_max_install_names
10581 @opindex image_base
10583 @opindex install_name
10584 @opindex keep_private_externs
10585 @opindex multi_module
10586 @opindex multiply_defined
10587 @opindex multiply_defined_unused
10588 @opindex noall_load
10589 @opindex no_dead_strip_inits_and_terms
10590 @opindex nofixprebinding
10591 @opindex nomultidefs
10593 @opindex noseglinkedit
10594 @opindex pagezero_size
10596 @opindex prebind_all_twolevel_modules
10597 @opindex private_bundle
10598 @opindex read_only_relocs
10600 @opindex sectobjectsymbols
10603 @opindex sectcreate
10604 @opindex sectobjectsymbols
10607 @opindex segs_read_only_addr
10608 @opindex segs_read_write_addr
10609 @opindex seg_addr_table
10610 @opindex seg_addr_table_filename
10611 @opindex seglinkedit
10613 @opindex segs_read_only_addr
10614 @opindex segs_read_write_addr
10615 @opindex single_module
10617 @opindex sub_library
10618 @opindex sub_umbrella
10619 @opindex twolevel_namespace
10622 @opindex unexported_symbols_list
10623 @opindex weak_reference_mismatches
10624 @opindex whatsloaded
10625 These options are passed to the Darwin linker. The Darwin linker man page
10626 describes them in detail.
10629 @node DEC Alpha Options
10630 @subsection DEC Alpha Options
10632 These @samp{-m} options are defined for the DEC Alpha implementations:
10635 @item -mno-soft-float
10636 @itemx -msoft-float
10637 @opindex mno-soft-float
10638 @opindex msoft-float
10639 Use (do not use) the hardware floating-point instructions for
10640 floating-point operations. When @option{-msoft-float} is specified,
10641 functions in @file{libgcc.a} will be used to perform floating-point
10642 operations. Unless they are replaced by routines that emulate the
10643 floating-point operations, or compiled in such a way as to call such
10644 emulations routines, these routines will issue floating-point
10645 operations. If you are compiling for an Alpha without floating-point
10646 operations, you must ensure that the library is built so as not to call
10649 Note that Alpha implementations without floating-point operations are
10650 required to have floating-point registers.
10653 @itemx -mno-fp-regs
10655 @opindex mno-fp-regs
10656 Generate code that uses (does not use) the floating-point register set.
10657 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
10658 register set is not used, floating point operands are passed in integer
10659 registers as if they were integers and floating-point results are passed
10660 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
10661 so any function with a floating-point argument or return value called by code
10662 compiled with @option{-mno-fp-regs} must also be compiled with that
10665 A typical use of this option is building a kernel that does not use,
10666 and hence need not save and restore, any floating-point registers.
10670 The Alpha architecture implements floating-point hardware optimized for
10671 maximum performance. It is mostly compliant with the IEEE floating
10672 point standard. However, for full compliance, software assistance is
10673 required. This option generates code fully IEEE compliant code
10674 @emph{except} that the @var{inexact-flag} is not maintained (see below).
10675 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
10676 defined during compilation. The resulting code is less efficient but is
10677 able to correctly support denormalized numbers and exceptional IEEE
10678 values such as not-a-number and plus/minus infinity. Other Alpha
10679 compilers call this option @option{-ieee_with_no_inexact}.
10681 @item -mieee-with-inexact
10682 @opindex mieee-with-inexact
10683 This is like @option{-mieee} except the generated code also maintains
10684 the IEEE @var{inexact-flag}. Turning on this option causes the
10685 generated code to implement fully-compliant IEEE math. In addition to
10686 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
10687 macro. On some Alpha implementations the resulting code may execute
10688 significantly slower than the code generated by default. Since there is
10689 very little code that depends on the @var{inexact-flag}, you should
10690 normally not specify this option. Other Alpha compilers call this
10691 option @option{-ieee_with_inexact}.
10693 @item -mfp-trap-mode=@var{trap-mode}
10694 @opindex mfp-trap-mode
10695 This option controls what floating-point related traps are enabled.
10696 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
10697 The trap mode can be set to one of four values:
10701 This is the default (normal) setting. The only traps that are enabled
10702 are the ones that cannot be disabled in software (e.g., division by zero
10706 In addition to the traps enabled by @samp{n}, underflow traps are enabled
10710 Like @samp{u}, but the instructions are marked to be safe for software
10711 completion (see Alpha architecture manual for details).
10714 Like @samp{su}, but inexact traps are enabled as well.
10717 @item -mfp-rounding-mode=@var{rounding-mode}
10718 @opindex mfp-rounding-mode
10719 Selects the IEEE rounding mode. Other Alpha compilers call this option
10720 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
10725 Normal IEEE rounding mode. Floating point numbers are rounded towards
10726 the nearest machine number or towards the even machine number in case
10730 Round towards minus infinity.
10733 Chopped rounding mode. Floating point numbers are rounded towards zero.
10736 Dynamic rounding mode. A field in the floating point control register
10737 (@var{fpcr}, see Alpha architecture reference manual) controls the
10738 rounding mode in effect. The C library initializes this register for
10739 rounding towards plus infinity. Thus, unless your program modifies the
10740 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
10743 @item -mtrap-precision=@var{trap-precision}
10744 @opindex mtrap-precision
10745 In the Alpha architecture, floating point traps are imprecise. This
10746 means without software assistance it is impossible to recover from a
10747 floating trap and program execution normally needs to be terminated.
10748 GCC can generate code that can assist operating system trap handlers
10749 in determining the exact location that caused a floating point trap.
10750 Depending on the requirements of an application, different levels of
10751 precisions can be selected:
10755 Program precision. This option is the default and means a trap handler
10756 can only identify which program caused a floating point exception.
10759 Function precision. The trap handler can determine the function that
10760 caused a floating point exception.
10763 Instruction precision. The trap handler can determine the exact
10764 instruction that caused a floating point exception.
10767 Other Alpha compilers provide the equivalent options called
10768 @option{-scope_safe} and @option{-resumption_safe}.
10770 @item -mieee-conformant
10771 @opindex mieee-conformant
10772 This option marks the generated code as IEEE conformant. You must not
10773 use this option unless you also specify @option{-mtrap-precision=i} and either
10774 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
10775 is to emit the line @samp{.eflag 48} in the function prologue of the
10776 generated assembly file. Under DEC Unix, this has the effect that
10777 IEEE-conformant math library routines will be linked in.
10779 @item -mbuild-constants
10780 @opindex mbuild-constants
10781 Normally GCC examines a 32- or 64-bit integer constant to
10782 see if it can construct it from smaller constants in two or three
10783 instructions. If it cannot, it will output the constant as a literal and
10784 generate code to load it from the data segment at runtime.
10786 Use this option to require GCC to construct @emph{all} integer constants
10787 using code, even if it takes more instructions (the maximum is six).
10789 You would typically use this option to build a shared library dynamic
10790 loader. Itself a shared library, it must relocate itself in memory
10791 before it can find the variables and constants in its own data segment.
10797 Select whether to generate code to be assembled by the vendor-supplied
10798 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
10816 Indicate whether GCC should generate code to use the optional BWX,
10817 CIX, FIX and MAX instruction sets. The default is to use the instruction
10818 sets supported by the CPU type specified via @option{-mcpu=} option or that
10819 of the CPU on which GCC was built if none was specified.
10822 @itemx -mfloat-ieee
10823 @opindex mfloat-vax
10824 @opindex mfloat-ieee
10825 Generate code that uses (does not use) VAX F and G floating point
10826 arithmetic instead of IEEE single and double precision.
10828 @item -mexplicit-relocs
10829 @itemx -mno-explicit-relocs
10830 @opindex mexplicit-relocs
10831 @opindex mno-explicit-relocs
10832 Older Alpha assemblers provided no way to generate symbol relocations
10833 except via assembler macros. Use of these macros does not allow
10834 optimal instruction scheduling. GNU binutils as of version 2.12
10835 supports a new syntax that allows the compiler to explicitly mark
10836 which relocations should apply to which instructions. This option
10837 is mostly useful for debugging, as GCC detects the capabilities of
10838 the assembler when it is built and sets the default accordingly.
10841 @itemx -mlarge-data
10842 @opindex msmall-data
10843 @opindex mlarge-data
10844 When @option{-mexplicit-relocs} is in effect, static data is
10845 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
10846 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
10847 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
10848 16-bit relocations off of the @code{$gp} register. This limits the
10849 size of the small data area to 64KB, but allows the variables to be
10850 directly accessed via a single instruction.
10852 The default is @option{-mlarge-data}. With this option the data area
10853 is limited to just below 2GB@. Programs that require more than 2GB of
10854 data must use @code{malloc} or @code{mmap} to allocate the data in the
10855 heap instead of in the program's data segment.
10857 When generating code for shared libraries, @option{-fpic} implies
10858 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
10861 @itemx -mlarge-text
10862 @opindex msmall-text
10863 @opindex mlarge-text
10864 When @option{-msmall-text} is used, the compiler assumes that the
10865 code of the entire program (or shared library) fits in 4MB, and is
10866 thus reachable with a branch instruction. When @option{-msmall-data}
10867 is used, the compiler can assume that all local symbols share the
10868 same @code{$gp} value, and thus reduce the number of instructions
10869 required for a function call from 4 to 1.
10871 The default is @option{-mlarge-text}.
10873 @item -mcpu=@var{cpu_type}
10875 Set the instruction set and instruction scheduling parameters for
10876 machine type @var{cpu_type}. You can specify either the @samp{EV}
10877 style name or the corresponding chip number. GCC supports scheduling
10878 parameters for the EV4, EV5 and EV6 family of processors and will
10879 choose the default values for the instruction set from the processor
10880 you specify. If you do not specify a processor type, GCC will default
10881 to the processor on which the compiler was built.
10883 Supported values for @var{cpu_type} are
10889 Schedules as an EV4 and has no instruction set extensions.
10893 Schedules as an EV5 and has no instruction set extensions.
10897 Schedules as an EV5 and supports the BWX extension.
10902 Schedules as an EV5 and supports the BWX and MAX extensions.
10906 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
10910 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
10913 Native Linux/GNU toolchains also support the value @samp{native},
10914 which selects the best architecture option for the host processor.
10915 @option{-mcpu=native} has no effect if GCC does not recognize
10918 @item -mtune=@var{cpu_type}
10920 Set only the instruction scheduling parameters for machine type
10921 @var{cpu_type}. The instruction set is not changed.
10923 Native Linux/GNU toolchains also support the value @samp{native},
10924 which selects the best architecture option for the host processor.
10925 @option{-mtune=native} has no effect if GCC does not recognize
10928 @item -mmemory-latency=@var{time}
10929 @opindex mmemory-latency
10930 Sets the latency the scheduler should assume for typical memory
10931 references as seen by the application. This number is highly
10932 dependent on the memory access patterns used by the application
10933 and the size of the external cache on the machine.
10935 Valid options for @var{time} are
10939 A decimal number representing clock cycles.
10945 The compiler contains estimates of the number of clock cycles for
10946 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
10947 (also called Dcache, Scache, and Bcache), as well as to main memory.
10948 Note that L3 is only valid for EV5.
10953 @node DEC Alpha/VMS Options
10954 @subsection DEC Alpha/VMS Options
10956 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
10959 @item -mvms-return-codes
10960 @opindex mvms-return-codes
10961 Return VMS condition codes from main. The default is to return POSIX
10962 style condition (e.g.@: error) codes.
10964 @item -mdebug-main=@var{prefix}
10965 @opindex mdebug-main=@var{prefix}
10966 Flag the first routine whose name starts with @var{prefix} as the main
10967 routine for the debugger.
10971 Default to 64bit memory allocation routines.
10975 @subsection FR30 Options
10976 @cindex FR30 Options
10978 These options are defined specifically for the FR30 port.
10982 @item -msmall-model
10983 @opindex msmall-model
10984 Use the small address space model. This can produce smaller code, but
10985 it does assume that all symbolic values and addresses will fit into a
10990 Assume that run-time support has been provided and so there is no need
10991 to include the simulator library (@file{libsim.a}) on the linker
10997 @subsection FRV Options
10998 @cindex FRV Options
11004 Only use the first 32 general purpose registers.
11009 Use all 64 general purpose registers.
11014 Use only the first 32 floating point registers.
11019 Use all 64 floating point registers
11022 @opindex mhard-float
11024 Use hardware instructions for floating point operations.
11027 @opindex msoft-float
11029 Use library routines for floating point operations.
11034 Dynamically allocate condition code registers.
11039 Do not try to dynamically allocate condition code registers, only
11040 use @code{icc0} and @code{fcc0}.
11045 Change ABI to use double word insns.
11050 Do not use double word instructions.
11055 Use floating point double instructions.
11058 @opindex mno-double
11060 Do not use floating point double instructions.
11065 Use media instructions.
11070 Do not use media instructions.
11075 Use multiply and add/subtract instructions.
11078 @opindex mno-muladd
11080 Do not use multiply and add/subtract instructions.
11085 Select the FDPIC ABI, that uses function descriptors to represent
11086 pointers to functions. Without any PIC/PIE-related options, it
11087 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
11088 assumes GOT entries and small data are within a 12-bit range from the
11089 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
11090 are computed with 32 bits.
11091 With a @samp{bfin-elf} target, this option implies @option{-msim}.
11094 @opindex minline-plt
11096 Enable inlining of PLT entries in function calls to functions that are
11097 not known to bind locally. It has no effect without @option{-mfdpic}.
11098 It's enabled by default if optimizing for speed and compiling for
11099 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
11100 optimization option such as @option{-O3} or above is present in the
11106 Assume a large TLS segment when generating thread-local code.
11111 Do not assume a large TLS segment when generating thread-local code.
11116 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
11117 that is known to be in read-only sections. It's enabled by default,
11118 except for @option{-fpic} or @option{-fpie}: even though it may help
11119 make the global offset table smaller, it trades 1 instruction for 4.
11120 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
11121 one of which may be shared by multiple symbols, and it avoids the need
11122 for a GOT entry for the referenced symbol, so it's more likely to be a
11123 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
11125 @item -multilib-library-pic
11126 @opindex multilib-library-pic
11128 Link with the (library, not FD) pic libraries. It's implied by
11129 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
11130 @option{-fpic} without @option{-mfdpic}. You should never have to use
11134 @opindex mlinked-fp
11136 Follow the EABI requirement of always creating a frame pointer whenever
11137 a stack frame is allocated. This option is enabled by default and can
11138 be disabled with @option{-mno-linked-fp}.
11141 @opindex mlong-calls
11143 Use indirect addressing to call functions outside the current
11144 compilation unit. This allows the functions to be placed anywhere
11145 within the 32-bit address space.
11147 @item -malign-labels
11148 @opindex malign-labels
11150 Try to align labels to an 8-byte boundary by inserting nops into the
11151 previous packet. This option only has an effect when VLIW packing
11152 is enabled. It doesn't create new packets; it merely adds nops to
11155 @item -mlibrary-pic
11156 @opindex mlibrary-pic
11158 Generate position-independent EABI code.
11163 Use only the first four media accumulator registers.
11168 Use all eight media accumulator registers.
11173 Pack VLIW instructions.
11178 Do not pack VLIW instructions.
11181 @opindex mno-eflags
11183 Do not mark ABI switches in e_flags.
11186 @opindex mcond-move
11188 Enable the use of conditional-move instructions (default).
11190 This switch is mainly for debugging the compiler and will likely be removed
11191 in a future version.
11193 @item -mno-cond-move
11194 @opindex mno-cond-move
11196 Disable the use of conditional-move instructions.
11198 This switch is mainly for debugging the compiler and will likely be removed
11199 in a future version.
11204 Enable the use of conditional set instructions (default).
11206 This switch is mainly for debugging the compiler and will likely be removed
11207 in a future version.
11212 Disable the use of conditional set instructions.
11214 This switch is mainly for debugging the compiler and will likely be removed
11215 in a future version.
11218 @opindex mcond-exec
11220 Enable the use of conditional execution (default).
11222 This switch is mainly for debugging the compiler and will likely be removed
11223 in a future version.
11225 @item -mno-cond-exec
11226 @opindex mno-cond-exec
11228 Disable the use of conditional execution.
11230 This switch is mainly for debugging the compiler and will likely be removed
11231 in a future version.
11233 @item -mvliw-branch
11234 @opindex mvliw-branch
11236 Run a pass to pack branches into VLIW instructions (default).
11238 This switch is mainly for debugging the compiler and will likely be removed
11239 in a future version.
11241 @item -mno-vliw-branch
11242 @opindex mno-vliw-branch
11244 Do not run a pass to pack branches into VLIW instructions.
11246 This switch is mainly for debugging the compiler and will likely be removed
11247 in a future version.
11249 @item -mmulti-cond-exec
11250 @opindex mmulti-cond-exec
11252 Enable optimization of @code{&&} and @code{||} in conditional execution
11255 This switch is mainly for debugging the compiler and will likely be removed
11256 in a future version.
11258 @item -mno-multi-cond-exec
11259 @opindex mno-multi-cond-exec
11261 Disable optimization of @code{&&} and @code{||} in conditional execution.
11263 This switch is mainly for debugging the compiler and will likely be removed
11264 in a future version.
11266 @item -mnested-cond-exec
11267 @opindex mnested-cond-exec
11269 Enable nested conditional execution optimizations (default).
11271 This switch is mainly for debugging the compiler and will likely be removed
11272 in a future version.
11274 @item -mno-nested-cond-exec
11275 @opindex mno-nested-cond-exec
11277 Disable nested conditional execution optimizations.
11279 This switch is mainly for debugging the compiler and will likely be removed
11280 in a future version.
11282 @item -moptimize-membar
11283 @opindex moptimize-membar
11285 This switch removes redundant @code{membar} instructions from the
11286 compiler generated code. It is enabled by default.
11288 @item -mno-optimize-membar
11289 @opindex mno-optimize-membar
11291 This switch disables the automatic removal of redundant @code{membar}
11292 instructions from the generated code.
11294 @item -mtomcat-stats
11295 @opindex mtomcat-stats
11297 Cause gas to print out tomcat statistics.
11299 @item -mcpu=@var{cpu}
11302 Select the processor type for which to generate code. Possible values are
11303 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
11304 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
11308 @node GNU/Linux Options
11309 @subsection GNU/Linux Options
11311 These @samp{-m} options are defined for GNU/Linux targets:
11316 Use the GNU C library instead of uClibc. This is the default except
11317 on @samp{*-*-linux-*uclibc*} targets.
11321 Use uClibc instead of the GNU C library. This is the default on
11322 @samp{*-*-linux-*uclibc*} targets.
11325 @node H8/300 Options
11326 @subsection H8/300 Options
11328 These @samp{-m} options are defined for the H8/300 implementations:
11333 Shorten some address references at link time, when possible; uses the
11334 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
11335 ld, Using ld}, for a fuller description.
11339 Generate code for the H8/300H@.
11343 Generate code for the H8S@.
11347 Generate code for the H8S and H8/300H in the normal mode. This switch
11348 must be used either with @option{-mh} or @option{-ms}.
11352 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
11356 Make @code{int} data 32 bits by default.
11359 @opindex malign-300
11360 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
11361 The default for the H8/300H and H8S is to align longs and floats on 4
11363 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
11364 This option has no effect on the H8/300.
11368 @subsection HPPA Options
11369 @cindex HPPA Options
11371 These @samp{-m} options are defined for the HPPA family of computers:
11374 @item -march=@var{architecture-type}
11376 Generate code for the specified architecture. The choices for
11377 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
11378 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
11379 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
11380 architecture option for your machine. Code compiled for lower numbered
11381 architectures will run on higher numbered architectures, but not the
11384 @item -mpa-risc-1-0
11385 @itemx -mpa-risc-1-1
11386 @itemx -mpa-risc-2-0
11387 @opindex mpa-risc-1-0
11388 @opindex mpa-risc-1-1
11389 @opindex mpa-risc-2-0
11390 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
11393 @opindex mbig-switch
11394 Generate code suitable for big switch tables. Use this option only if
11395 the assembler/linker complain about out of range branches within a switch
11398 @item -mjump-in-delay
11399 @opindex mjump-in-delay
11400 Fill delay slots of function calls with unconditional jump instructions
11401 by modifying the return pointer for the function call to be the target
11402 of the conditional jump.
11404 @item -mdisable-fpregs
11405 @opindex mdisable-fpregs
11406 Prevent floating point registers from being used in any manner. This is
11407 necessary for compiling kernels which perform lazy context switching of
11408 floating point registers. If you use this option and attempt to perform
11409 floating point operations, the compiler will abort.
11411 @item -mdisable-indexing
11412 @opindex mdisable-indexing
11413 Prevent the compiler from using indexing address modes. This avoids some
11414 rather obscure problems when compiling MIG generated code under MACH@.
11416 @item -mno-space-regs
11417 @opindex mno-space-regs
11418 Generate code that assumes the target has no space registers. This allows
11419 GCC to generate faster indirect calls and use unscaled index address modes.
11421 Such code is suitable for level 0 PA systems and kernels.
11423 @item -mfast-indirect-calls
11424 @opindex mfast-indirect-calls
11425 Generate code that assumes calls never cross space boundaries. This
11426 allows GCC to emit code which performs faster indirect calls.
11428 This option will not work in the presence of shared libraries or nested
11431 @item -mfixed-range=@var{register-range}
11432 @opindex mfixed-range
11433 Generate code treating the given register range as fixed registers.
11434 A fixed register is one that the register allocator can not use. This is
11435 useful when compiling kernel code. A register range is specified as
11436 two registers separated by a dash. Multiple register ranges can be
11437 specified separated by a comma.
11439 @item -mlong-load-store
11440 @opindex mlong-load-store
11441 Generate 3-instruction load and store sequences as sometimes required by
11442 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
11445 @item -mportable-runtime
11446 @opindex mportable-runtime
11447 Use the portable calling conventions proposed by HP for ELF systems.
11451 Enable the use of assembler directives only GAS understands.
11453 @item -mschedule=@var{cpu-type}
11455 Schedule code according to the constraints for the machine type
11456 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
11457 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
11458 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
11459 proper scheduling option for your machine. The default scheduling is
11463 @opindex mlinker-opt
11464 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
11465 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
11466 linkers in which they give bogus error messages when linking some programs.
11469 @opindex msoft-float
11470 Generate output containing library calls for floating point.
11471 @strong{Warning:} the requisite libraries are not available for all HPPA
11472 targets. Normally the facilities of the machine's usual C compiler are
11473 used, but this cannot be done directly in cross-compilation. You must make
11474 your own arrangements to provide suitable library functions for
11477 @option{-msoft-float} changes the calling convention in the output file;
11478 therefore, it is only useful if you compile @emph{all} of a program with
11479 this option. In particular, you need to compile @file{libgcc.a}, the
11480 library that comes with GCC, with @option{-msoft-float} in order for
11485 Generate the predefine, @code{_SIO}, for server IO@. The default is
11486 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
11487 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
11488 options are available under HP-UX and HI-UX@.
11492 Use GNU ld specific options. This passes @option{-shared} to ld when
11493 building a shared library. It is the default when GCC is configured,
11494 explicitly or implicitly, with the GNU linker. This option does not
11495 have any affect on which ld is called, it only changes what parameters
11496 are passed to that ld. The ld that is called is determined by the
11497 @option{--with-ld} configure option, GCC's program search path, and
11498 finally by the user's @env{PATH}. The linker used by GCC can be printed
11499 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
11500 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11504 Use HP ld specific options. This passes @option{-b} to ld when building
11505 a shared library and passes @option{+Accept TypeMismatch} to ld on all
11506 links. It is the default when GCC is configured, explicitly or
11507 implicitly, with the HP linker. This option does not have any affect on
11508 which ld is called, it only changes what parameters are passed to that
11509 ld. The ld that is called is determined by the @option{--with-ld}
11510 configure option, GCC's program search path, and finally by the user's
11511 @env{PATH}. The linker used by GCC can be printed using @samp{which
11512 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
11513 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11516 @opindex mno-long-calls
11517 Generate code that uses long call sequences. This ensures that a call
11518 is always able to reach linker generated stubs. The default is to generate
11519 long calls only when the distance from the call site to the beginning
11520 of the function or translation unit, as the case may be, exceeds a
11521 predefined limit set by the branch type being used. The limits for
11522 normal calls are 7,600,000 and 240,000 bytes, respectively for the
11523 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
11526 Distances are measured from the beginning of functions when using the
11527 @option{-ffunction-sections} option, or when using the @option{-mgas}
11528 and @option{-mno-portable-runtime} options together under HP-UX with
11531 It is normally not desirable to use this option as it will degrade
11532 performance. However, it may be useful in large applications,
11533 particularly when partial linking is used to build the application.
11535 The types of long calls used depends on the capabilities of the
11536 assembler and linker, and the type of code being generated. The
11537 impact on systems that support long absolute calls, and long pic
11538 symbol-difference or pc-relative calls should be relatively small.
11539 However, an indirect call is used on 32-bit ELF systems in pic code
11540 and it is quite long.
11542 @item -munix=@var{unix-std}
11544 Generate compiler predefines and select a startfile for the specified
11545 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
11546 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
11547 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
11548 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
11549 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
11552 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
11553 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
11554 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
11555 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
11556 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
11557 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
11559 It is @emph{important} to note that this option changes the interfaces
11560 for various library routines. It also affects the operational behavior
11561 of the C library. Thus, @emph{extreme} care is needed in using this
11564 Library code that is intended to operate with more than one UNIX
11565 standard must test, set and restore the variable @var{__xpg4_extended_mask}
11566 as appropriate. Most GNU software doesn't provide this capability.
11570 Suppress the generation of link options to search libdld.sl when the
11571 @option{-static} option is specified on HP-UX 10 and later.
11575 The HP-UX implementation of setlocale in libc has a dependency on
11576 libdld.sl. There isn't an archive version of libdld.sl. Thus,
11577 when the @option{-static} option is specified, special link options
11578 are needed to resolve this dependency.
11580 On HP-UX 10 and later, the GCC driver adds the necessary options to
11581 link with libdld.sl when the @option{-static} option is specified.
11582 This causes the resulting binary to be dynamic. On the 64-bit port,
11583 the linkers generate dynamic binaries by default in any case. The
11584 @option{-nolibdld} option can be used to prevent the GCC driver from
11585 adding these link options.
11589 Add support for multithreading with the @dfn{dce thread} library
11590 under HP-UX@. This option sets flags for both the preprocessor and
11594 @node i386 and x86-64 Options
11595 @subsection Intel 386 and AMD x86-64 Options
11596 @cindex i386 Options
11597 @cindex x86-64 Options
11598 @cindex Intel 386 Options
11599 @cindex AMD x86-64 Options
11601 These @samp{-m} options are defined for the i386 and x86-64 family of
11605 @item -mtune=@var{cpu-type}
11607 Tune to @var{cpu-type} everything applicable about the generated code, except
11608 for the ABI and the set of available instructions. The choices for
11609 @var{cpu-type} are:
11612 Produce code optimized for the most common IA32/AMD64/EM64T processors.
11613 If you know the CPU on which your code will run, then you should use
11614 the corresponding @option{-mtune} option instead of
11615 @option{-mtune=generic}. But, if you do not know exactly what CPU users
11616 of your application will have, then you should use this option.
11618 As new processors are deployed in the marketplace, the behavior of this
11619 option will change. Therefore, if you upgrade to a newer version of
11620 GCC, the code generated option will change to reflect the processors
11621 that were most common when that version of GCC was released.
11623 There is no @option{-march=generic} option because @option{-march}
11624 indicates the instruction set the compiler can use, and there is no
11625 generic instruction set applicable to all processors. In contrast,
11626 @option{-mtune} indicates the processor (or, in this case, collection of
11627 processors) for which the code is optimized.
11629 This selects the CPU to tune for at compilation time by determining
11630 the processor type of the compiling machine. Using @option{-mtune=native}
11631 will produce code optimized for the local machine under the constraints
11632 of the selected instruction set. Using @option{-march=native} will
11633 enable all instruction subsets supported by the local machine (hence
11634 the result might not run on different machines).
11636 Original Intel's i386 CPU@.
11638 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
11639 @item i586, pentium
11640 Intel Pentium CPU with no MMX support.
11642 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
11644 Intel PentiumPro CPU@.
11646 Same as @code{generic}, but when used as @code{march} option, PentiumPro
11647 instruction set will be used, so the code will run on all i686 family chips.
11649 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
11650 @item pentium3, pentium3m
11651 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
11654 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
11655 support. Used by Centrino notebooks.
11656 @item pentium4, pentium4m
11657 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
11659 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
11662 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
11663 SSE2 and SSE3 instruction set support.
11665 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
11666 instruction set support.
11668 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
11669 instruction set support.
11671 AMD K6 CPU with MMX instruction set support.
11673 Improved versions of AMD K6 CPU with MMX and 3dNOW!@: instruction set support.
11674 @item athlon, athlon-tbird
11675 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and SSE prefetch instructions
11677 @item athlon-4, athlon-xp, athlon-mp
11678 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and full SSE
11679 instruction set support.
11680 @item k8, opteron, athlon64, athlon-fx
11681 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
11682 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW!@: and 64-bit instruction set extensions.)
11683 @item k8-sse3, opteron-sse3, athlon64-sse3
11684 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
11685 @item amdfam10, barcelona
11686 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
11687 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3dNOW!, enhanced 3dNOW!, ABM and 64-bit
11688 instruction set extensions.)
11690 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
11693 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!@:
11694 instruction set support.
11696 Via C3 CPU with MMX and 3dNOW!@: instruction set support. (No scheduling is
11697 implemented for this chip.)
11699 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
11700 implemented for this chip.)
11702 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
11705 While picking a specific @var{cpu-type} will schedule things appropriately
11706 for that particular chip, the compiler will not generate any code that
11707 does not run on the i386 without the @option{-march=@var{cpu-type}} option
11710 @item -march=@var{cpu-type}
11712 Generate instructions for the machine type @var{cpu-type}. The choices
11713 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
11714 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
11716 @item -mcpu=@var{cpu-type}
11718 A deprecated synonym for @option{-mtune}.
11720 @item -mfpmath=@var{unit}
11722 Generate floating point arithmetics for selected unit @var{unit}. The choices
11723 for @var{unit} are:
11727 Use the standard 387 floating point coprocessor present majority of chips and
11728 emulated otherwise. Code compiled with this option will run almost everywhere.
11729 The temporary results are computed in 80bit precision instead of precision
11730 specified by the type resulting in slightly different results compared to most
11731 of other chips. See @option{-ffloat-store} for more detailed description.
11733 This is the default choice for i386 compiler.
11736 Use scalar floating point instructions present in the SSE instruction set.
11737 This instruction set is supported by Pentium3 and newer chips, in the AMD line
11738 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
11739 instruction set supports only single precision arithmetics, thus the double and
11740 extended precision arithmetics is still done using 387. Later version, present
11741 only in Pentium4 and the future AMD x86-64 chips supports double precision
11744 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
11745 or @option{-msse2} switches to enable SSE extensions and make this option
11746 effective. For the x86-64 compiler, these extensions are enabled by default.
11748 The resulting code should be considerably faster in the majority of cases and avoid
11749 the numerical instability problems of 387 code, but may break some existing
11750 code that expects temporaries to be 80bit.
11752 This is the default choice for the x86-64 compiler.
11757 Attempt to utilize both instruction sets at once. This effectively double the
11758 amount of available registers and on chips with separate execution units for
11759 387 and SSE the execution resources too. Use this option with care, as it is
11760 still experimental, because the GCC register allocator does not model separate
11761 functional units well resulting in instable performance.
11764 @item -masm=@var{dialect}
11765 @opindex masm=@var{dialect}
11766 Output asm instructions using selected @var{dialect}. Supported
11767 choices are @samp{intel} or @samp{att} (the default one). Darwin does
11768 not support @samp{intel}.
11771 @itemx -mno-ieee-fp
11773 @opindex mno-ieee-fp
11774 Control whether or not the compiler uses IEEE floating point
11775 comparisons. These handle correctly the case where the result of a
11776 comparison is unordered.
11779 @opindex msoft-float
11780 Generate output containing library calls for floating point.
11781 @strong{Warning:} the requisite libraries are not part of GCC@.
11782 Normally the facilities of the machine's usual C compiler are used, but
11783 this can't be done directly in cross-compilation. You must make your
11784 own arrangements to provide suitable library functions for
11787 On machines where a function returns floating point results in the 80387
11788 register stack, some floating point opcodes may be emitted even if
11789 @option{-msoft-float} is used.
11791 @item -mno-fp-ret-in-387
11792 @opindex mno-fp-ret-in-387
11793 Do not use the FPU registers for return values of functions.
11795 The usual calling convention has functions return values of types
11796 @code{float} and @code{double} in an FPU register, even if there
11797 is no FPU@. The idea is that the operating system should emulate
11800 The option @option{-mno-fp-ret-in-387} causes such values to be returned
11801 in ordinary CPU registers instead.
11803 @item -mno-fancy-math-387
11804 @opindex mno-fancy-math-387
11805 Some 387 emulators do not support the @code{sin}, @code{cos} and
11806 @code{sqrt} instructions for the 387. Specify this option to avoid
11807 generating those instructions. This option is the default on FreeBSD,
11808 OpenBSD and NetBSD@. This option is overridden when @option{-march}
11809 indicates that the target cpu will always have an FPU and so the
11810 instruction will not need emulation. As of revision 2.6.1, these
11811 instructions are not generated unless you also use the
11812 @option{-funsafe-math-optimizations} switch.
11814 @item -malign-double
11815 @itemx -mno-align-double
11816 @opindex malign-double
11817 @opindex mno-align-double
11818 Control whether GCC aligns @code{double}, @code{long double}, and
11819 @code{long long} variables on a two word boundary or a one word
11820 boundary. Aligning @code{double} variables on a two word boundary will
11821 produce code that runs somewhat faster on a @samp{Pentium} at the
11822 expense of more memory.
11824 On x86-64, @option{-malign-double} is enabled by default.
11826 @strong{Warning:} if you use the @option{-malign-double} switch,
11827 structures containing the above types will be aligned differently than
11828 the published application binary interface specifications for the 386
11829 and will not be binary compatible with structures in code compiled
11830 without that switch.
11832 @item -m96bit-long-double
11833 @itemx -m128bit-long-double
11834 @opindex m96bit-long-double
11835 @opindex m128bit-long-double
11836 These switches control the size of @code{long double} type. The i386
11837 application binary interface specifies the size to be 96 bits,
11838 so @option{-m96bit-long-double} is the default in 32 bit mode.
11840 Modern architectures (Pentium and newer) would prefer @code{long double}
11841 to be aligned to an 8 or 16 byte boundary. In arrays or structures
11842 conforming to the ABI, this would not be possible. So specifying a
11843 @option{-m128bit-long-double} will align @code{long double}
11844 to a 16 byte boundary by padding the @code{long double} with an additional
11847 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
11848 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
11850 Notice that neither of these options enable any extra precision over the x87
11851 standard of 80 bits for a @code{long double}.
11853 @strong{Warning:} if you override the default value for your target ABI, the
11854 structures and arrays containing @code{long double} variables will change
11855 their size as well as function calling convention for function taking
11856 @code{long double} will be modified. Hence they will not be binary
11857 compatible with arrays or structures in code compiled without that switch.
11859 @item -mlarge-data-threshold=@var{number}
11860 @opindex mlarge-data-threshold=@var{number}
11861 When @option{-mcmodel=medium} is specified, the data greater than
11862 @var{threshold} are placed in large data section. This value must be the
11863 same across all object linked into the binary and defaults to 65535.
11867 Use a different function-calling convention, in which functions that
11868 take a fixed number of arguments return with the @code{ret} @var{num}
11869 instruction, which pops their arguments while returning. This saves one
11870 instruction in the caller since there is no need to pop the arguments
11873 You can specify that an individual function is called with this calling
11874 sequence with the function attribute @samp{stdcall}. You can also
11875 override the @option{-mrtd} option by using the function attribute
11876 @samp{cdecl}. @xref{Function Attributes}.
11878 @strong{Warning:} this calling convention is incompatible with the one
11879 normally used on Unix, so you cannot use it if you need to call
11880 libraries compiled with the Unix compiler.
11882 Also, you must provide function prototypes for all functions that
11883 take variable numbers of arguments (including @code{printf});
11884 otherwise incorrect code will be generated for calls to those
11887 In addition, seriously incorrect code will result if you call a
11888 function with too many arguments. (Normally, extra arguments are
11889 harmlessly ignored.)
11891 @item -mregparm=@var{num}
11893 Control how many registers are used to pass integer arguments. By
11894 default, no registers are used to pass arguments, and at most 3
11895 registers can be used. You can control this behavior for a specific
11896 function by using the function attribute @samp{regparm}.
11897 @xref{Function Attributes}.
11899 @strong{Warning:} if you use this switch, and
11900 @var{num} is nonzero, then you must build all modules with the same
11901 value, including any libraries. This includes the system libraries and
11905 @opindex msseregparm
11906 Use SSE register passing conventions for float and double arguments
11907 and return values. You can control this behavior for a specific
11908 function by using the function attribute @samp{sseregparm}.
11909 @xref{Function Attributes}.
11911 @strong{Warning:} if you use this switch then you must build all
11912 modules with the same value, including any libraries. This includes
11913 the system libraries and startup modules.
11922 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
11923 is specified, the significands of results of floating-point operations are
11924 rounded to 24 bits (single precision); @option{-mpc64} rounds the
11925 significands of results of floating-point operations to 53 bits (double
11926 precision) and @option{-mpc80} rounds the significands of results of
11927 floating-point operations to 64 bits (extended double precision), which is
11928 the default. When this option is used, floating-point operations in higher
11929 precisions are not available to the programmer without setting the FPU
11930 control word explicitly.
11932 Setting the rounding of floating-point operations to less than the default
11933 80 bits can speed some programs by 2% or more. Note that some mathematical
11934 libraries assume that extended precision (80 bit) floating-point operations
11935 are enabled by default; routines in such libraries could suffer significant
11936 loss of accuracy, typically through so-called "catastrophic cancellation",
11937 when this option is used to set the precision to less than extended precision.
11939 @item -mstackrealign
11940 @opindex mstackrealign
11941 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
11942 option will generate an alternate prologue and epilogue that realigns the
11943 runtime stack if necessary. This supports mixing legacy codes that keep
11944 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
11945 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
11946 applicable to individual functions.
11948 @item -mpreferred-stack-boundary=@var{num}
11949 @opindex mpreferred-stack-boundary
11950 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
11951 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
11952 the default is 4 (16 bytes or 128 bits).
11954 @item -mincoming-stack-boundary=@var{num}
11955 @opindex mincoming-stack-boundary
11956 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
11957 boundary. If @option{-mincoming-stack-boundary} is not specified,
11958 the one specified by @option{-mpreferred-stack-boundary} will be used.
11960 On Pentium and PentiumPro, @code{double} and @code{long double} values
11961 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
11962 suffer significant run time performance penalties. On Pentium III, the
11963 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
11964 properly if it is not 16 byte aligned.
11966 To ensure proper alignment of this values on the stack, the stack boundary
11967 must be as aligned as that required by any value stored on the stack.
11968 Further, every function must be generated such that it keeps the stack
11969 aligned. Thus calling a function compiled with a higher preferred
11970 stack boundary from a function compiled with a lower preferred stack
11971 boundary will most likely misalign the stack. It is recommended that
11972 libraries that use callbacks always use the default setting.
11974 This extra alignment does consume extra stack space, and generally
11975 increases code size. Code that is sensitive to stack space usage, such
11976 as embedded systems and operating system kernels, may want to reduce the
11977 preferred alignment to @option{-mpreferred-stack-boundary=2}.
12017 These switches enable or disable the use of instructions in the MMX,
12018 SSE, SSE2, SSE3, SSSE3, SSE4.1, AVX, AES, PCLMUL, SSE4A, FMA4, ABM or
12019 3DNow!@: extended instruction sets.
12020 These extensions are also available as built-in functions: see
12021 @ref{X86 Built-in Functions}, for details of the functions enabled and
12022 disabled by these switches.
12024 To have SSE/SSE2 instructions generated automatically from floating-point
12025 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
12027 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
12028 generates new AVX instructions or AVX equivalence for all SSEx instructions
12031 These options will enable GCC to use these extended instructions in
12032 generated code, even without @option{-mfpmath=sse}. Applications which
12033 perform runtime CPU detection must compile separate files for each
12034 supported architecture, using the appropriate flags. In particular,
12035 the file containing the CPU detection code should be compiled without
12040 This option instructs GCC to emit a @code{cld} instruction in the prologue
12041 of functions that use string instructions. String instructions depend on
12042 the DF flag to select between autoincrement or autodecrement mode. While the
12043 ABI specifies the DF flag to be cleared on function entry, some operating
12044 systems violate this specification by not clearing the DF flag in their
12045 exception dispatchers. The exception handler can be invoked with the DF flag
12046 set which leads to wrong direction mode, when string instructions are used.
12047 This option can be enabled by default on 32-bit x86 targets by configuring
12048 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
12049 instructions can be suppressed with the @option{-mno-cld} compiler option
12054 This option will enable GCC to use CMPXCHG16B instruction in generated code.
12055 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
12056 data types. This is useful for high resolution counters that could be updated
12057 by multiple processors (or cores). This instruction is generated as part of
12058 atomic built-in functions: see @ref{Atomic Builtins} for details.
12062 This option will enable GCC to use SAHF instruction in generated 64-bit code.
12063 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
12064 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
12065 SAHF are load and store instructions, respectively, for certain status flags.
12066 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
12067 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
12071 This option will enable GCC to use movbe instruction to implement
12072 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
12076 This option will enable built-in functions, @code{__builtin_ia32_crc32qi},
12077 @code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and
12078 @code{__builtin_ia32_crc32di} to generate the crc32 machine instruction.
12082 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
12083 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
12084 to increase precision instead of DIVSS and SQRTSS (and their vectorized
12085 variants) for single precision floating point arguments. These instructions
12086 are generated only when @option{-funsafe-math-optimizations} is enabled
12087 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
12088 Note that while the throughput of the sequence is higher than the throughput
12089 of the non-reciprocal instruction, the precision of the sequence can be
12090 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
12092 @item -mveclibabi=@var{type}
12093 @opindex mveclibabi
12094 Specifies the ABI type to use for vectorizing intrinsics using an
12095 external library. Supported types are @code{svml} for the Intel short
12096 vector math library and @code{acml} for the AMD math core library style
12097 of interfacing. GCC will currently emit calls to @code{vmldExp2},
12098 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
12099 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
12100 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
12101 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
12102 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
12103 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
12104 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
12105 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
12106 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
12107 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
12108 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
12109 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
12110 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
12111 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
12112 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
12113 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
12114 compatible library will have to be specified at link time.
12116 @item -mabi=@var{name}
12118 Generate code for the specified calling convention. Permissible values
12119 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
12120 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
12121 ABI when targeting Windows. On all other systems, the default is the
12122 SYSV ABI. You can control this behavior for a specific function by
12123 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
12124 @xref{Function Attributes}.
12127 @itemx -mno-push-args
12128 @opindex mpush-args
12129 @opindex mno-push-args
12130 Use PUSH operations to store outgoing parameters. This method is shorter
12131 and usually equally fast as method using SUB/MOV operations and is enabled
12132 by default. In some cases disabling it may improve performance because of
12133 improved scheduling and reduced dependencies.
12135 @item -maccumulate-outgoing-args
12136 @opindex maccumulate-outgoing-args
12137 If enabled, the maximum amount of space required for outgoing arguments will be
12138 computed in the function prologue. This is faster on most modern CPUs
12139 because of reduced dependencies, improved scheduling and reduced stack usage
12140 when preferred stack boundary is not equal to 2. The drawback is a notable
12141 increase in code size. This switch implies @option{-mno-push-args}.
12145 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
12146 on thread-safe exception handling must compile and link all code with the
12147 @option{-mthreads} option. When compiling, @option{-mthreads} defines
12148 @option{-D_MT}; when linking, it links in a special thread helper library
12149 @option{-lmingwthrd} which cleans up per thread exception handling data.
12151 @item -mno-align-stringops
12152 @opindex mno-align-stringops
12153 Do not align destination of inlined string operations. This switch reduces
12154 code size and improves performance in case the destination is already aligned,
12155 but GCC doesn't know about it.
12157 @item -minline-all-stringops
12158 @opindex minline-all-stringops
12159 By default GCC inlines string operations only when destination is known to be
12160 aligned at least to 4 byte boundary. This enables more inlining, increase code
12161 size, but may improve performance of code that depends on fast memcpy, strlen
12162 and memset for short lengths.
12164 @item -minline-stringops-dynamically
12165 @opindex minline-stringops-dynamically
12166 For string operation of unknown size, inline runtime checks so for small
12167 blocks inline code is used, while for large blocks library call is used.
12169 @item -mstringop-strategy=@var{alg}
12170 @opindex mstringop-strategy=@var{alg}
12171 Overwrite internal decision heuristic about particular algorithm to inline
12172 string operation with. The allowed values are @code{rep_byte},
12173 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
12174 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
12175 expanding inline loop, @code{libcall} for always expanding library call.
12177 @item -momit-leaf-frame-pointer
12178 @opindex momit-leaf-frame-pointer
12179 Don't keep the frame pointer in a register for leaf functions. This
12180 avoids the instructions to save, set up and restore frame pointers and
12181 makes an extra register available in leaf functions. The option
12182 @option{-fomit-frame-pointer} removes the frame pointer for all functions
12183 which might make debugging harder.
12185 @item -mtls-direct-seg-refs
12186 @itemx -mno-tls-direct-seg-refs
12187 @opindex mtls-direct-seg-refs
12188 Controls whether TLS variables may be accessed with offsets from the
12189 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
12190 or whether the thread base pointer must be added. Whether or not this
12191 is legal depends on the operating system, and whether it maps the
12192 segment to cover the entire TLS area.
12194 For systems that use GNU libc, the default is on.
12197 @itemx -mno-sse2avx
12199 Specify that the assembler should encode SSE instructions with VEX
12200 prefix. The option @option{-mavx} turns this on by default.
12203 These @samp{-m} switches are supported in addition to the above
12204 on AMD x86-64 processors in 64-bit environments.
12211 Generate code for a 32-bit or 64-bit environment.
12212 The 32-bit environment sets int, long and pointer to 32 bits and
12213 generates code that runs on any i386 system.
12214 The 64-bit environment sets int to 32 bits and long and pointer
12215 to 64 bits and generates code for AMD's x86-64 architecture. For
12216 darwin only the -m64 option turns off the @option{-fno-pic} and
12217 @option{-mdynamic-no-pic} options.
12219 @item -mno-red-zone
12220 @opindex mno-red-zone
12221 Do not use a so called red zone for x86-64 code. The red zone is mandated
12222 by the x86-64 ABI, it is a 128-byte area beyond the location of the
12223 stack pointer that will not be modified by signal or interrupt handlers
12224 and therefore can be used for temporary data without adjusting the stack
12225 pointer. The flag @option{-mno-red-zone} disables this red zone.
12227 @item -mcmodel=small
12228 @opindex mcmodel=small
12229 Generate code for the small code model: the program and its symbols must
12230 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
12231 Programs can be statically or dynamically linked. This is the default
12234 @item -mcmodel=kernel
12235 @opindex mcmodel=kernel
12236 Generate code for the kernel code model. The kernel runs in the
12237 negative 2 GB of the address space.
12238 This model has to be used for Linux kernel code.
12240 @item -mcmodel=medium
12241 @opindex mcmodel=medium
12242 Generate code for the medium model: The program is linked in the lower 2
12243 GB of the address space. Small symbols are also placed there. Symbols
12244 with sizes larger than @option{-mlarge-data-threshold} are put into
12245 large data or bss sections and can be located above 2GB. Programs can
12246 be statically or dynamically linked.
12248 @item -mcmodel=large
12249 @opindex mcmodel=large
12250 Generate code for the large model: This model makes no assumptions
12251 about addresses and sizes of sections.
12254 @node IA-64 Options
12255 @subsection IA-64 Options
12256 @cindex IA-64 Options
12258 These are the @samp{-m} options defined for the Intel IA-64 architecture.
12262 @opindex mbig-endian
12263 Generate code for a big endian target. This is the default for HP-UX@.
12265 @item -mlittle-endian
12266 @opindex mlittle-endian
12267 Generate code for a little endian target. This is the default for AIX5
12273 @opindex mno-gnu-as
12274 Generate (or don't) code for the GNU assembler. This is the default.
12275 @c Also, this is the default if the configure option @option{--with-gnu-as}
12281 @opindex mno-gnu-ld
12282 Generate (or don't) code for the GNU linker. This is the default.
12283 @c Also, this is the default if the configure option @option{--with-gnu-ld}
12288 Generate code that does not use a global pointer register. The result
12289 is not position independent code, and violates the IA-64 ABI@.
12291 @item -mvolatile-asm-stop
12292 @itemx -mno-volatile-asm-stop
12293 @opindex mvolatile-asm-stop
12294 @opindex mno-volatile-asm-stop
12295 Generate (or don't) a stop bit immediately before and after volatile asm
12298 @item -mregister-names
12299 @itemx -mno-register-names
12300 @opindex mregister-names
12301 @opindex mno-register-names
12302 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
12303 the stacked registers. This may make assembler output more readable.
12309 Disable (or enable) optimizations that use the small data section. This may
12310 be useful for working around optimizer bugs.
12312 @item -mconstant-gp
12313 @opindex mconstant-gp
12314 Generate code that uses a single constant global pointer value. This is
12315 useful when compiling kernel code.
12319 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
12320 This is useful when compiling firmware code.
12322 @item -minline-float-divide-min-latency
12323 @opindex minline-float-divide-min-latency
12324 Generate code for inline divides of floating point values
12325 using the minimum latency algorithm.
12327 @item -minline-float-divide-max-throughput
12328 @opindex minline-float-divide-max-throughput
12329 Generate code for inline divides of floating point values
12330 using the maximum throughput algorithm.
12332 @item -mno-inline-float-divide
12333 @opindex mno-inline-float-divide
12334 Do not generate inline code for divides of floating point values.
12336 @item -minline-int-divide-min-latency
12337 @opindex minline-int-divide-min-latency
12338 Generate code for inline divides of integer values
12339 using the minimum latency algorithm.
12341 @item -minline-int-divide-max-throughput
12342 @opindex minline-int-divide-max-throughput
12343 Generate code for inline divides of integer values
12344 using the maximum throughput algorithm.
12346 @item -mno-inline-int-divide
12347 @opindex mno-inline-int-divide
12348 Do not generate inline code for divides of integer values.
12350 @item -minline-sqrt-min-latency
12351 @opindex minline-sqrt-min-latency
12352 Generate code for inline square roots
12353 using the minimum latency algorithm.
12355 @item -minline-sqrt-max-throughput
12356 @opindex minline-sqrt-max-throughput
12357 Generate code for inline square roots
12358 using the maximum throughput algorithm.
12360 @item -mno-inline-sqrt
12361 @opindex mno-inline-sqrt
12362 Do not generate inline code for sqrt.
12365 @itemx -mno-fused-madd
12366 @opindex mfused-madd
12367 @opindex mno-fused-madd
12368 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
12369 instructions. The default is to use these instructions.
12371 @item -mno-dwarf2-asm
12372 @itemx -mdwarf2-asm
12373 @opindex mno-dwarf2-asm
12374 @opindex mdwarf2-asm
12375 Don't (or do) generate assembler code for the DWARF2 line number debugging
12376 info. This may be useful when not using the GNU assembler.
12378 @item -mearly-stop-bits
12379 @itemx -mno-early-stop-bits
12380 @opindex mearly-stop-bits
12381 @opindex mno-early-stop-bits
12382 Allow stop bits to be placed earlier than immediately preceding the
12383 instruction that triggered the stop bit. This can improve instruction
12384 scheduling, but does not always do so.
12386 @item -mfixed-range=@var{register-range}
12387 @opindex mfixed-range
12388 Generate code treating the given register range as fixed registers.
12389 A fixed register is one that the register allocator can not use. This is
12390 useful when compiling kernel code. A register range is specified as
12391 two registers separated by a dash. Multiple register ranges can be
12392 specified separated by a comma.
12394 @item -mtls-size=@var{tls-size}
12396 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
12399 @item -mtune=@var{cpu-type}
12401 Tune the instruction scheduling for a particular CPU, Valid values are
12402 itanium, itanium1, merced, itanium2, and mckinley.
12408 Generate code for a 32-bit or 64-bit environment.
12409 The 32-bit environment sets int, long and pointer to 32 bits.
12410 The 64-bit environment sets int to 32 bits and long and pointer
12411 to 64 bits. These are HP-UX specific flags.
12413 @item -mno-sched-br-data-spec
12414 @itemx -msched-br-data-spec
12415 @opindex mno-sched-br-data-spec
12416 @opindex msched-br-data-spec
12417 (Dis/En)able data speculative scheduling before reload.
12418 This will result in generation of the ld.a instructions and
12419 the corresponding check instructions (ld.c / chk.a).
12420 The default is 'disable'.
12422 @item -msched-ar-data-spec
12423 @itemx -mno-sched-ar-data-spec
12424 @opindex msched-ar-data-spec
12425 @opindex mno-sched-ar-data-spec
12426 (En/Dis)able data speculative scheduling after reload.
12427 This will result in generation of the ld.a instructions and
12428 the corresponding check instructions (ld.c / chk.a).
12429 The default is 'enable'.
12431 @item -mno-sched-control-spec
12432 @itemx -msched-control-spec
12433 @opindex mno-sched-control-spec
12434 @opindex msched-control-spec
12435 (Dis/En)able control speculative scheduling. This feature is
12436 available only during region scheduling (i.e.@: before reload).
12437 This will result in generation of the ld.s instructions and
12438 the corresponding check instructions chk.s .
12439 The default is 'disable'.
12441 @item -msched-br-in-data-spec
12442 @itemx -mno-sched-br-in-data-spec
12443 @opindex msched-br-in-data-spec
12444 @opindex mno-sched-br-in-data-spec
12445 (En/Dis)able speculative scheduling of the instructions that
12446 are dependent on the data speculative loads before reload.
12447 This is effective only with @option{-msched-br-data-spec} enabled.
12448 The default is 'enable'.
12450 @item -msched-ar-in-data-spec
12451 @itemx -mno-sched-ar-in-data-spec
12452 @opindex msched-ar-in-data-spec
12453 @opindex mno-sched-ar-in-data-spec
12454 (En/Dis)able speculative scheduling of the instructions that
12455 are dependent on the data speculative loads after reload.
12456 This is effective only with @option{-msched-ar-data-spec} enabled.
12457 The default is 'enable'.
12459 @item -msched-in-control-spec
12460 @itemx -mno-sched-in-control-spec
12461 @opindex msched-in-control-spec
12462 @opindex mno-sched-in-control-spec
12463 (En/Dis)able speculative scheduling of the instructions that
12464 are dependent on the control speculative loads.
12465 This is effective only with @option{-msched-control-spec} enabled.
12466 The default is 'enable'.
12468 @item -mno-sched-prefer-non-data-spec-insns
12469 @itemx -msched-prefer-non-data-spec-insns
12470 @opindex mno-sched-prefer-non-data-spec-insns
12471 @opindex msched-prefer-non-data-spec-insns
12472 If enabled, data speculative instructions will be chosen for schedule
12473 only if there are no other choices at the moment. This will make
12474 the use of the data speculation much more conservative.
12475 The default is 'disable'.
12477 @item -mno-sched-prefer-non-control-spec-insns
12478 @itemx -msched-prefer-non-control-spec-insns
12479 @opindex mno-sched-prefer-non-control-spec-insns
12480 @opindex msched-prefer-non-control-spec-insns
12481 If enabled, control speculative instructions will be chosen for schedule
12482 only if there are no other choices at the moment. This will make
12483 the use of the control speculation much more conservative.
12484 The default is 'disable'.
12486 @item -mno-sched-count-spec-in-critical-path
12487 @itemx -msched-count-spec-in-critical-path
12488 @opindex mno-sched-count-spec-in-critical-path
12489 @opindex msched-count-spec-in-critical-path
12490 If enabled, speculative dependencies will be considered during
12491 computation of the instructions priorities. This will make the use of the
12492 speculation a bit more conservative.
12493 The default is 'disable'.
12495 @item -msched-spec-ldc
12496 @opindex msched-spec-ldc
12497 Use a simple data speculation check. This option is on by default.
12499 @item -msched-control-spec-ldc
12500 @opindex msched-spec-ldc
12501 Use a simple check for control speculation. This option is on by default.
12503 @item -msched-stop-bits-after-every-cycle
12504 @opindex msched-stop-bits-after-every-cycle
12505 Place a stop bit after every cycle when scheduling. This option is on
12508 @item -msched-fp-mem-deps-zero-cost
12509 @opindex msched-fp-mem-deps-zero-cost
12510 Assume that floating-point stores and loads are not likely to cause a conflict
12511 when placed into the same instruction group. This option is disabled by
12514 @item -msel-sched-dont-check-control-spec
12515 @opindex msel-sched-dont-check-control-spec
12516 Generate checks for control speculation in selective scheduling.
12517 This flag is disabled by default.
12519 @item -msched-max-memory-insns=@var{max-insns}
12520 @opindex msched-max-memory-insns
12521 Limit on the number of memory insns per instruction group, giving lower
12522 priority to subsequent memory insns attempting to schedule in the same
12523 instruction group. Frequently useful to prevent cache bank conflicts.
12524 The default value is 1.
12526 @item -msched-max-memory-insns-hard-limit
12527 @opindex msched-max-memory-insns-hard-limit
12528 Disallow more than `msched-max-memory-insns' in instruction group.
12529 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
12530 when limit is reached but may still schedule memory operations.
12534 @node IA-64/VMS Options
12535 @subsection IA-64/VMS Options
12537 These @samp{-m} options are defined for the IA-64/VMS implementations:
12540 @item -mvms-return-codes
12541 @opindex mvms-return-codes
12542 Return VMS condition codes from main. The default is to return POSIX
12543 style condition (e.g.@ error) codes.
12545 @item -mdebug-main=@var{prefix}
12546 @opindex mdebug-main=@var{prefix}
12547 Flag the first routine whose name starts with @var{prefix} as the main
12548 routine for the debugger.
12552 Default to 64bit memory allocation routines.
12556 @subsection M32C Options
12557 @cindex M32C options
12560 @item -mcpu=@var{name}
12562 Select the CPU for which code is generated. @var{name} may be one of
12563 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
12564 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
12565 the M32C/80 series.
12569 Specifies that the program will be run on the simulator. This causes
12570 an alternate runtime library to be linked in which supports, for
12571 example, file I/O@. You must not use this option when generating
12572 programs that will run on real hardware; you must provide your own
12573 runtime library for whatever I/O functions are needed.
12575 @item -memregs=@var{number}
12577 Specifies the number of memory-based pseudo-registers GCC will use
12578 during code generation. These pseudo-registers will be used like real
12579 registers, so there is a tradeoff between GCC's ability to fit the
12580 code into available registers, and the performance penalty of using
12581 memory instead of registers. Note that all modules in a program must
12582 be compiled with the same value for this option. Because of that, you
12583 must not use this option with the default runtime libraries gcc
12588 @node M32R/D Options
12589 @subsection M32R/D Options
12590 @cindex M32R/D options
12592 These @option{-m} options are defined for Renesas M32R/D architectures:
12597 Generate code for the M32R/2@.
12601 Generate code for the M32R/X@.
12605 Generate code for the M32R@. This is the default.
12607 @item -mmodel=small
12608 @opindex mmodel=small
12609 Assume all objects live in the lower 16MB of memory (so that their addresses
12610 can be loaded with the @code{ld24} instruction), and assume all subroutines
12611 are reachable with the @code{bl} instruction.
12612 This is the default.
12614 The addressability of a particular object can be set with the
12615 @code{model} attribute.
12617 @item -mmodel=medium
12618 @opindex mmodel=medium
12619 Assume objects may be anywhere in the 32-bit address space (the compiler
12620 will generate @code{seth/add3} instructions to load their addresses), and
12621 assume all subroutines are reachable with the @code{bl} instruction.
12623 @item -mmodel=large
12624 @opindex mmodel=large
12625 Assume objects may be anywhere in the 32-bit address space (the compiler
12626 will generate @code{seth/add3} instructions to load their addresses), and
12627 assume subroutines may not be reachable with the @code{bl} instruction
12628 (the compiler will generate the much slower @code{seth/add3/jl}
12629 instruction sequence).
12632 @opindex msdata=none
12633 Disable use of the small data area. Variables will be put into
12634 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
12635 @code{section} attribute has been specified).
12636 This is the default.
12638 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
12639 Objects may be explicitly put in the small data area with the
12640 @code{section} attribute using one of these sections.
12642 @item -msdata=sdata
12643 @opindex msdata=sdata
12644 Put small global and static data in the small data area, but do not
12645 generate special code to reference them.
12648 @opindex msdata=use
12649 Put small global and static data in the small data area, and generate
12650 special instructions to reference them.
12654 @cindex smaller data references
12655 Put global and static objects less than or equal to @var{num} bytes
12656 into the small data or bss sections instead of the normal data or bss
12657 sections. The default value of @var{num} is 8.
12658 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
12659 for this option to have any effect.
12661 All modules should be compiled with the same @option{-G @var{num}} value.
12662 Compiling with different values of @var{num} may or may not work; if it
12663 doesn't the linker will give an error message---incorrect code will not be
12668 Makes the M32R specific code in the compiler display some statistics
12669 that might help in debugging programs.
12671 @item -malign-loops
12672 @opindex malign-loops
12673 Align all loops to a 32-byte boundary.
12675 @item -mno-align-loops
12676 @opindex mno-align-loops
12677 Do not enforce a 32-byte alignment for loops. This is the default.
12679 @item -missue-rate=@var{number}
12680 @opindex missue-rate=@var{number}
12681 Issue @var{number} instructions per cycle. @var{number} can only be 1
12684 @item -mbranch-cost=@var{number}
12685 @opindex mbranch-cost=@var{number}
12686 @var{number} can only be 1 or 2. If it is 1 then branches will be
12687 preferred over conditional code, if it is 2, then the opposite will
12690 @item -mflush-trap=@var{number}
12691 @opindex mflush-trap=@var{number}
12692 Specifies the trap number to use to flush the cache. The default is
12693 12. Valid numbers are between 0 and 15 inclusive.
12695 @item -mno-flush-trap
12696 @opindex mno-flush-trap
12697 Specifies that the cache cannot be flushed by using a trap.
12699 @item -mflush-func=@var{name}
12700 @opindex mflush-func=@var{name}
12701 Specifies the name of the operating system function to call to flush
12702 the cache. The default is @emph{_flush_cache}, but a function call
12703 will only be used if a trap is not available.
12705 @item -mno-flush-func
12706 @opindex mno-flush-func
12707 Indicates that there is no OS function for flushing the cache.
12711 @node M680x0 Options
12712 @subsection M680x0 Options
12713 @cindex M680x0 options
12715 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
12716 The default settings depend on which architecture was selected when
12717 the compiler was configured; the defaults for the most common choices
12721 @item -march=@var{arch}
12723 Generate code for a specific M680x0 or ColdFire instruction set
12724 architecture. Permissible values of @var{arch} for M680x0
12725 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
12726 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
12727 architectures are selected according to Freescale's ISA classification
12728 and the permissible values are: @samp{isaa}, @samp{isaaplus},
12729 @samp{isab} and @samp{isac}.
12731 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
12732 code for a ColdFire target. The @var{arch} in this macro is one of the
12733 @option{-march} arguments given above.
12735 When used together, @option{-march} and @option{-mtune} select code
12736 that runs on a family of similar processors but that is optimized
12737 for a particular microarchitecture.
12739 @item -mcpu=@var{cpu}
12741 Generate code for a specific M680x0 or ColdFire processor.
12742 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
12743 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
12744 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
12745 below, which also classifies the CPUs into families:
12747 @multitable @columnfractions 0.20 0.80
12748 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
12749 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
12750 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
12751 @item @samp{5206e} @tab @samp{5206e}
12752 @item @samp{5208} @tab @samp{5207} @samp{5208}
12753 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
12754 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
12755 @item @samp{5216} @tab @samp{5214} @samp{5216}
12756 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
12757 @item @samp{5225} @tab @samp{5224} @samp{5225}
12758 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
12759 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
12760 @item @samp{5249} @tab @samp{5249}
12761 @item @samp{5250} @tab @samp{5250}
12762 @item @samp{5271} @tab @samp{5270} @samp{5271}
12763 @item @samp{5272} @tab @samp{5272}
12764 @item @samp{5275} @tab @samp{5274} @samp{5275}
12765 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
12766 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
12767 @item @samp{5307} @tab @samp{5307}
12768 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
12769 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
12770 @item @samp{5407} @tab @samp{5407}
12771 @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}
12774 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
12775 @var{arch} is compatible with @var{cpu}. Other combinations of
12776 @option{-mcpu} and @option{-march} are rejected.
12778 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
12779 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
12780 where the value of @var{family} is given by the table above.
12782 @item -mtune=@var{tune}
12784 Tune the code for a particular microarchitecture, within the
12785 constraints set by @option{-march} and @option{-mcpu}.
12786 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
12787 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
12788 and @samp{cpu32}. The ColdFire microarchitectures
12789 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
12791 You can also use @option{-mtune=68020-40} for code that needs
12792 to run relatively well on 68020, 68030 and 68040 targets.
12793 @option{-mtune=68020-60} is similar but includes 68060 targets
12794 as well. These two options select the same tuning decisions as
12795 @option{-m68020-40} and @option{-m68020-60} respectively.
12797 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
12798 when tuning for 680x0 architecture @var{arch}. It also defines
12799 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
12800 option is used. If gcc is tuning for a range of architectures,
12801 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
12802 it defines the macros for every architecture in the range.
12804 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
12805 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
12806 of the arguments given above.
12812 Generate output for a 68000. This is the default
12813 when the compiler is configured for 68000-based systems.
12814 It is equivalent to @option{-march=68000}.
12816 Use this option for microcontrollers with a 68000 or EC000 core,
12817 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
12821 Generate output for a 68010. This is the default
12822 when the compiler is configured for 68010-based systems.
12823 It is equivalent to @option{-march=68010}.
12829 Generate output for a 68020. This is the default
12830 when the compiler is configured for 68020-based systems.
12831 It is equivalent to @option{-march=68020}.
12835 Generate output for a 68030. This is the default when the compiler is
12836 configured for 68030-based systems. It is equivalent to
12837 @option{-march=68030}.
12841 Generate output for a 68040. This is the default when the compiler is
12842 configured for 68040-based systems. It is equivalent to
12843 @option{-march=68040}.
12845 This option inhibits the use of 68881/68882 instructions that have to be
12846 emulated by software on the 68040. Use this option if your 68040 does not
12847 have code to emulate those instructions.
12851 Generate output for a 68060. This is the default when the compiler is
12852 configured for 68060-based systems. It is equivalent to
12853 @option{-march=68060}.
12855 This option inhibits the use of 68020 and 68881/68882 instructions that
12856 have to be emulated by software on the 68060. Use this option if your 68060
12857 does not have code to emulate those instructions.
12861 Generate output for a CPU32. This is the default
12862 when the compiler is configured for CPU32-based systems.
12863 It is equivalent to @option{-march=cpu32}.
12865 Use this option for microcontrollers with a
12866 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
12867 68336, 68340, 68341, 68349 and 68360.
12871 Generate output for a 520X ColdFire CPU@. This is the default
12872 when the compiler is configured for 520X-based systems.
12873 It is equivalent to @option{-mcpu=5206}, and is now deprecated
12874 in favor of that option.
12876 Use this option for microcontroller with a 5200 core, including
12877 the MCF5202, MCF5203, MCF5204 and MCF5206.
12881 Generate output for a 5206e ColdFire CPU@. The option is now
12882 deprecated in favor of the equivalent @option{-mcpu=5206e}.
12886 Generate output for a member of the ColdFire 528X family.
12887 The option is now deprecated in favor of the equivalent
12888 @option{-mcpu=528x}.
12892 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
12893 in favor of the equivalent @option{-mcpu=5307}.
12897 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
12898 in favor of the equivalent @option{-mcpu=5407}.
12902 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
12903 This includes use of hardware floating point instructions.
12904 The option is equivalent to @option{-mcpu=547x}, and is now
12905 deprecated in favor of that option.
12909 Generate output for a 68040, without using any of the new instructions.
12910 This results in code which can run relatively efficiently on either a
12911 68020/68881 or a 68030 or a 68040. The generated code does use the
12912 68881 instructions that are emulated on the 68040.
12914 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
12918 Generate output for a 68060, without using any of the new instructions.
12919 This results in code which can run relatively efficiently on either a
12920 68020/68881 or a 68030 or a 68040. The generated code does use the
12921 68881 instructions that are emulated on the 68060.
12923 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
12927 @opindex mhard-float
12929 Generate floating-point instructions. This is the default for 68020
12930 and above, and for ColdFire devices that have an FPU@. It defines the
12931 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
12932 on ColdFire targets.
12935 @opindex msoft-float
12936 Do not generate floating-point instructions; use library calls instead.
12937 This is the default for 68000, 68010, and 68832 targets. It is also
12938 the default for ColdFire devices that have no FPU.
12944 Generate (do not generate) ColdFire hardware divide and remainder
12945 instructions. If @option{-march} is used without @option{-mcpu},
12946 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
12947 architectures. Otherwise, the default is taken from the target CPU
12948 (either the default CPU, or the one specified by @option{-mcpu}). For
12949 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
12950 @option{-mcpu=5206e}.
12952 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
12956 Consider type @code{int} to be 16 bits wide, like @code{short int}.
12957 Additionally, parameters passed on the stack are also aligned to a
12958 16-bit boundary even on targets whose API mandates promotion to 32-bit.
12962 Do not consider type @code{int} to be 16 bits wide. This is the default.
12965 @itemx -mno-bitfield
12966 @opindex mnobitfield
12967 @opindex mno-bitfield
12968 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
12969 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
12973 Do use the bit-field instructions. The @option{-m68020} option implies
12974 @option{-mbitfield}. This is the default if you use a configuration
12975 designed for a 68020.
12979 Use a different function-calling convention, in which functions
12980 that take a fixed number of arguments return with the @code{rtd}
12981 instruction, which pops their arguments while returning. This
12982 saves one instruction in the caller since there is no need to pop
12983 the arguments there.
12985 This calling convention is incompatible with the one normally
12986 used on Unix, so you cannot use it if you need to call libraries
12987 compiled with the Unix compiler.
12989 Also, you must provide function prototypes for all functions that
12990 take variable numbers of arguments (including @code{printf});
12991 otherwise incorrect code will be generated for calls to those
12994 In addition, seriously incorrect code will result if you call a
12995 function with too many arguments. (Normally, extra arguments are
12996 harmlessly ignored.)
12998 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
12999 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
13003 Do not use the calling conventions selected by @option{-mrtd}.
13004 This is the default.
13007 @itemx -mno-align-int
13008 @opindex malign-int
13009 @opindex mno-align-int
13010 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
13011 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
13012 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
13013 Aligning variables on 32-bit boundaries produces code that runs somewhat
13014 faster on processors with 32-bit busses at the expense of more memory.
13016 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
13017 align structures containing the above types differently than
13018 most published application binary interface specifications for the m68k.
13022 Use the pc-relative addressing mode of the 68000 directly, instead of
13023 using a global offset table. At present, this option implies @option{-fpic},
13024 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
13025 not presently supported with @option{-mpcrel}, though this could be supported for
13026 68020 and higher processors.
13028 @item -mno-strict-align
13029 @itemx -mstrict-align
13030 @opindex mno-strict-align
13031 @opindex mstrict-align
13032 Do not (do) assume that unaligned memory references will be handled by
13036 Generate code that allows the data segment to be located in a different
13037 area of memory from the text segment. This allows for execute in place in
13038 an environment without virtual memory management. This option implies
13041 @item -mno-sep-data
13042 Generate code that assumes that the data segment follows the text segment.
13043 This is the default.
13045 @item -mid-shared-library
13046 Generate code that supports shared libraries via the library ID method.
13047 This allows for execute in place and shared libraries in an environment
13048 without virtual memory management. This option implies @option{-fPIC}.
13050 @item -mno-id-shared-library
13051 Generate code that doesn't assume ID based shared libraries are being used.
13052 This is the default.
13054 @item -mshared-library-id=n
13055 Specified the identification number of the ID based shared library being
13056 compiled. Specifying a value of 0 will generate more compact code, specifying
13057 other values will force the allocation of that number to the current
13058 library but is no more space or time efficient than omitting this option.
13064 When generating position-independent code for ColdFire, generate code
13065 that works if the GOT has more than 8192 entries. This code is
13066 larger and slower than code generated without this option. On M680x0
13067 processors, this option is not needed; @option{-fPIC} suffices.
13069 GCC normally uses a single instruction to load values from the GOT@.
13070 While this is relatively efficient, it only works if the GOT
13071 is smaller than about 64k. Anything larger causes the linker
13072 to report an error such as:
13074 @cindex relocation truncated to fit (ColdFire)
13076 relocation truncated to fit: R_68K_GOT16O foobar
13079 If this happens, you should recompile your code with @option{-mxgot}.
13080 It should then work with very large GOTs. However, code generated with
13081 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
13082 the value of a global symbol.
13084 Note that some linkers, including newer versions of the GNU linker,
13085 can create multiple GOTs and sort GOT entries. If you have such a linker,
13086 you should only need to use @option{-mxgot} when compiling a single
13087 object file that accesses more than 8192 GOT entries. Very few do.
13089 These options have no effect unless GCC is generating
13090 position-independent code.
13094 @node M68hc1x Options
13095 @subsection M68hc1x Options
13096 @cindex M68hc1x options
13098 These are the @samp{-m} options defined for the 68hc11 and 68hc12
13099 microcontrollers. The default values for these options depends on
13100 which style of microcontroller was selected when the compiler was configured;
13101 the defaults for the most common choices are given below.
13108 Generate output for a 68HC11. This is the default
13109 when the compiler is configured for 68HC11-based systems.
13115 Generate output for a 68HC12. This is the default
13116 when the compiler is configured for 68HC12-based systems.
13122 Generate output for a 68HCS12.
13124 @item -mauto-incdec
13125 @opindex mauto-incdec
13126 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
13133 Enable the use of 68HC12 min and max instructions.
13136 @itemx -mno-long-calls
13137 @opindex mlong-calls
13138 @opindex mno-long-calls
13139 Treat all calls as being far away (near). If calls are assumed to be
13140 far away, the compiler will use the @code{call} instruction to
13141 call a function and the @code{rtc} instruction for returning.
13145 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13147 @item -msoft-reg-count=@var{count}
13148 @opindex msoft-reg-count
13149 Specify the number of pseudo-soft registers which are used for the
13150 code generation. The maximum number is 32. Using more pseudo-soft
13151 register may or may not result in better code depending on the program.
13152 The default is 4 for 68HC11 and 2 for 68HC12.
13156 @node MCore Options
13157 @subsection MCore Options
13158 @cindex MCore options
13160 These are the @samp{-m} options defined for the Motorola M*Core
13166 @itemx -mno-hardlit
13168 @opindex mno-hardlit
13169 Inline constants into the code stream if it can be done in two
13170 instructions or less.
13176 Use the divide instruction. (Enabled by default).
13178 @item -mrelax-immediate
13179 @itemx -mno-relax-immediate
13180 @opindex mrelax-immediate
13181 @opindex mno-relax-immediate
13182 Allow arbitrary sized immediates in bit operations.
13184 @item -mwide-bitfields
13185 @itemx -mno-wide-bitfields
13186 @opindex mwide-bitfields
13187 @opindex mno-wide-bitfields
13188 Always treat bit-fields as int-sized.
13190 @item -m4byte-functions
13191 @itemx -mno-4byte-functions
13192 @opindex m4byte-functions
13193 @opindex mno-4byte-functions
13194 Force all functions to be aligned to a four byte boundary.
13196 @item -mcallgraph-data
13197 @itemx -mno-callgraph-data
13198 @opindex mcallgraph-data
13199 @opindex mno-callgraph-data
13200 Emit callgraph information.
13203 @itemx -mno-slow-bytes
13204 @opindex mslow-bytes
13205 @opindex mno-slow-bytes
13206 Prefer word access when reading byte quantities.
13208 @item -mlittle-endian
13209 @itemx -mbig-endian
13210 @opindex mlittle-endian
13211 @opindex mbig-endian
13212 Generate code for a little endian target.
13218 Generate code for the 210 processor.
13222 Assume that run-time support has been provided and so omit the
13223 simulator library (@file{libsim.a)} from the linker command line.
13225 @item -mstack-increment=@var{size}
13226 @opindex mstack-increment
13227 Set the maximum amount for a single stack increment operation. Large
13228 values can increase the speed of programs which contain functions
13229 that need a large amount of stack space, but they can also trigger a
13230 segmentation fault if the stack is extended too much. The default
13236 @subsection MeP Options
13237 @cindex MeP options
13243 Enables the @code{abs} instruction, which is the absolute difference
13244 between two registers.
13248 Enables all the optional instructions - average, multiply, divide, bit
13249 operations, leading zero, absolute difference, min/max, clip, and
13255 Enables the @code{ave} instruction, which computes the average of two
13258 @item -mbased=@var{n}
13260 Variables of size @var{n} bytes or smaller will be placed in the
13261 @code{.based} section by default. Based variables use the @code{$tp}
13262 register as a base register, and there is a 128 byte limit to the
13263 @code{.based} section.
13267 Enables the bit operation instructions - bit test (@code{btstm}), set
13268 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
13269 test-and-set (@code{tas}).
13271 @item -mc=@var{name}
13273 Selects which section constant data will be placed in. @var{name} may
13274 be @code{tiny}, @code{near}, or @code{far}.
13278 Enables the @code{clip} instruction. Note that @code{-mclip} is not
13279 useful unless you also provide @code{-mminmax}.
13281 @item -mconfig=@var{name}
13283 Selects one of the build-in core configurations. Each MeP chip has
13284 one or more modules in it; each module has a core CPU and a variety of
13285 coprocessors, optional instructions, and peripherals. The
13286 @code{MeP-Integrator} tool, not part of GCC, provides these
13287 configurations through this option; using this option is the same as
13288 using all the corresponding command line options. The default
13289 configuration is @code{default}.
13293 Enables the coprocessor instructions. By default, this is a 32-bit
13294 coprocessor. Note that the coprocessor is normally enabled via the
13295 @code{-mconfig=} option.
13299 Enables the 32-bit coprocessor's instructions.
13303 Enables the 64-bit coprocessor's instructions.
13307 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
13311 Causes constant variables to be placed in the @code{.near} section.
13315 Enables the @code{div} and @code{divu} instructions.
13319 Generate big-endian code.
13323 Generate little-endian code.
13325 @item -mio-volatile
13326 @opindex mio-volatile
13327 Tells the compiler that any variable marked with the @code{io}
13328 attribute is to be considered volatile.
13332 Causes variables to be assigned to the @code{.far} section by default.
13336 Enables the @code{leadz} (leading zero) instruction.
13340 Causes variables to be assigned to the @code{.near} section by default.
13344 Enables the @code{min} and @code{max} instructions.
13348 Enables the multiplication and multiply-accumulate instructions.
13352 Disables all the optional instructions enabled by @code{-mall-opts}.
13356 Enables the @code{repeat} and @code{erepeat} instructions, used for
13357 low-overhead looping.
13361 Causes all variables to default to the @code{.tiny} section. Note
13362 that there is a 65536 byte limit to this section. Accesses to these
13363 variables use the @code{%gp} base register.
13367 Enables the saturation instructions. Note that the compiler does not
13368 currently generate these itself, but this option is included for
13369 compatibility with other tools, like @code{as}.
13373 Link the SDRAM-based runtime instead of the default ROM-based runtime.
13377 Link the simulator runtime libraries.
13381 Link the simulator runtime libraries, excluding built-in support
13382 for reset and exception vectors and tables.
13386 Causes all functions to default to the @code{.far} section. Without
13387 this option, functions default to the @code{.near} section.
13389 @item -mtiny=@var{n}
13391 Variables that are @var{n} bytes or smaller will be allocated to the
13392 @code{.tiny} section. These variables use the @code{$gp} base
13393 register. The default for this option is 4, but note that there's a
13394 65536 byte limit to the @code{.tiny} section.
13399 @subsection MIPS Options
13400 @cindex MIPS options
13406 Generate big-endian code.
13410 Generate little-endian code. This is the default for @samp{mips*el-*-*}
13413 @item -march=@var{arch}
13415 Generate code that will run on @var{arch}, which can be the name of a
13416 generic MIPS ISA, or the name of a particular processor.
13418 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
13419 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
13420 The processor names are:
13421 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
13422 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
13423 @samp{5kc}, @samp{5kf},
13425 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
13426 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
13427 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
13428 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
13429 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
13430 @samp{loongson2e}, @samp{loongson2f},
13434 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
13435 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
13436 @samp{rm7000}, @samp{rm9000},
13437 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
13440 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
13441 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
13443 The special value @samp{from-abi} selects the
13444 most compatible architecture for the selected ABI (that is,
13445 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
13447 Native Linux/GNU toolchains also support the value @samp{native},
13448 which selects the best architecture option for the host processor.
13449 @option{-march=native} has no effect if GCC does not recognize
13452 In processor names, a final @samp{000} can be abbreviated as @samp{k}
13453 (for example, @samp{-march=r2k}). Prefixes are optional, and
13454 @samp{vr} may be written @samp{r}.
13456 Names of the form @samp{@var{n}f2_1} refer to processors with
13457 FPUs clocked at half the rate of the core, names of the form
13458 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
13459 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
13460 processors with FPUs clocked a ratio of 3:2 with respect to the core.
13461 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
13462 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
13463 accepted as synonyms for @samp{@var{n}f1_1}.
13465 GCC defines two macros based on the value of this option. The first
13466 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
13467 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
13468 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
13469 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
13470 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
13472 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
13473 above. In other words, it will have the full prefix and will not
13474 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
13475 the macro names the resolved architecture (either @samp{"mips1"} or
13476 @samp{"mips3"}). It names the default architecture when no
13477 @option{-march} option is given.
13479 @item -mtune=@var{arch}
13481 Optimize for @var{arch}. Among other things, this option controls
13482 the way instructions are scheduled, and the perceived cost of arithmetic
13483 operations. The list of @var{arch} values is the same as for
13486 When this option is not used, GCC will optimize for the processor
13487 specified by @option{-march}. By using @option{-march} and
13488 @option{-mtune} together, it is possible to generate code that will
13489 run on a family of processors, but optimize the code for one
13490 particular member of that family.
13492 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
13493 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
13494 @samp{-march} ones described above.
13498 Equivalent to @samp{-march=mips1}.
13502 Equivalent to @samp{-march=mips2}.
13506 Equivalent to @samp{-march=mips3}.
13510 Equivalent to @samp{-march=mips4}.
13514 Equivalent to @samp{-march=mips32}.
13518 Equivalent to @samp{-march=mips32r2}.
13522 Equivalent to @samp{-march=mips64}.
13526 Equivalent to @samp{-march=mips64r2}.
13531 @opindex mno-mips16
13532 Generate (do not generate) MIPS16 code. If GCC is targetting a
13533 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
13535 MIPS16 code generation can also be controlled on a per-function basis
13536 by means of @code{mips16} and @code{nomips16} attributes.
13537 @xref{Function Attributes}, for more information.
13539 @item -mflip-mips16
13540 @opindex mflip-mips16
13541 Generate MIPS16 code on alternating functions. This option is provided
13542 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
13543 not intended for ordinary use in compiling user code.
13545 @item -minterlink-mips16
13546 @itemx -mno-interlink-mips16
13547 @opindex minterlink-mips16
13548 @opindex mno-interlink-mips16
13549 Require (do not require) that non-MIPS16 code be link-compatible with
13552 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
13553 it must either use a call or an indirect jump. @option{-minterlink-mips16}
13554 therefore disables direct jumps unless GCC knows that the target of the
13555 jump is not MIPS16.
13567 Generate code for the given ABI@.
13569 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
13570 generates 64-bit code when you select a 64-bit architecture, but you
13571 can use @option{-mgp32} to get 32-bit code instead.
13573 For information about the O64 ABI, see
13574 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
13576 GCC supports a variant of the o32 ABI in which floating-point registers
13577 are 64 rather than 32 bits wide. You can select this combination with
13578 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
13579 and @samp{mfhc1} instructions and is therefore only supported for
13580 MIPS32R2 processors.
13582 The register assignments for arguments and return values remain the
13583 same, but each scalar value is passed in a single 64-bit register
13584 rather than a pair of 32-bit registers. For example, scalar
13585 floating-point values are returned in @samp{$f0} only, not a
13586 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
13587 remains the same, but all 64 bits are saved.
13590 @itemx -mno-abicalls
13592 @opindex mno-abicalls
13593 Generate (do not generate) code that is suitable for SVR4-style
13594 dynamic objects. @option{-mabicalls} is the default for SVR4-based
13599 Generate (do not generate) code that is fully position-independent,
13600 and that can therefore be linked into shared libraries. This option
13601 only affects @option{-mabicalls}.
13603 All @option{-mabicalls} code has traditionally been position-independent,
13604 regardless of options like @option{-fPIC} and @option{-fpic}. However,
13605 as an extension, the GNU toolchain allows executables to use absolute
13606 accesses for locally-binding symbols. It can also use shorter GP
13607 initialization sequences and generate direct calls to locally-defined
13608 functions. This mode is selected by @option{-mno-shared}.
13610 @option{-mno-shared} depends on binutils 2.16 or higher and generates
13611 objects that can only be linked by the GNU linker. However, the option
13612 does not affect the ABI of the final executable; it only affects the ABI
13613 of relocatable objects. Using @option{-mno-shared} will generally make
13614 executables both smaller and quicker.
13616 @option{-mshared} is the default.
13622 Assume (do not assume) that the static and dynamic linkers
13623 support PLTs and copy relocations. This option only affects
13624 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
13625 has no effect without @samp{-msym32}.
13627 You can make @option{-mplt} the default by configuring
13628 GCC with @option{--with-mips-plt}. The default is
13629 @option{-mno-plt} otherwise.
13635 Lift (do not lift) the usual restrictions on the size of the global
13638 GCC normally uses a single instruction to load values from the GOT@.
13639 While this is relatively efficient, it will only work if the GOT
13640 is smaller than about 64k. Anything larger will cause the linker
13641 to report an error such as:
13643 @cindex relocation truncated to fit (MIPS)
13645 relocation truncated to fit: R_MIPS_GOT16 foobar
13648 If this happens, you should recompile your code with @option{-mxgot}.
13649 It should then work with very large GOTs, although it will also be
13650 less efficient, since it will take three instructions to fetch the
13651 value of a global symbol.
13653 Note that some linkers can create multiple GOTs. If you have such a
13654 linker, you should only need to use @option{-mxgot} when a single object
13655 file accesses more than 64k's worth of GOT entries. Very few do.
13657 These options have no effect unless GCC is generating position
13662 Assume that general-purpose registers are 32 bits wide.
13666 Assume that general-purpose registers are 64 bits wide.
13670 Assume that floating-point registers are 32 bits wide.
13674 Assume that floating-point registers are 64 bits wide.
13677 @opindex mhard-float
13678 Use floating-point coprocessor instructions.
13681 @opindex msoft-float
13682 Do not use floating-point coprocessor instructions. Implement
13683 floating-point calculations using library calls instead.
13685 @item -msingle-float
13686 @opindex msingle-float
13687 Assume that the floating-point coprocessor only supports single-precision
13690 @item -mdouble-float
13691 @opindex mdouble-float
13692 Assume that the floating-point coprocessor supports double-precision
13693 operations. This is the default.
13699 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
13700 implement atomic memory built-in functions. When neither option is
13701 specified, GCC will use the instructions if the target architecture
13704 @option{-mllsc} is useful if the runtime environment can emulate the
13705 instructions and @option{-mno-llsc} can be useful when compiling for
13706 nonstandard ISAs. You can make either option the default by
13707 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
13708 respectively. @option{--with-llsc} is the default for some
13709 configurations; see the installation documentation for details.
13715 Use (do not use) revision 1 of the MIPS DSP ASE@.
13716 @xref{MIPS DSP Built-in Functions}. This option defines the
13717 preprocessor macro @samp{__mips_dsp}. It also defines
13718 @samp{__mips_dsp_rev} to 1.
13724 Use (do not use) revision 2 of the MIPS DSP ASE@.
13725 @xref{MIPS DSP Built-in Functions}. This option defines the
13726 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
13727 It also defines @samp{__mips_dsp_rev} to 2.
13730 @itemx -mno-smartmips
13731 @opindex msmartmips
13732 @opindex mno-smartmips
13733 Use (do not use) the MIPS SmartMIPS ASE.
13735 @item -mpaired-single
13736 @itemx -mno-paired-single
13737 @opindex mpaired-single
13738 @opindex mno-paired-single
13739 Use (do not use) paired-single floating-point instructions.
13740 @xref{MIPS Paired-Single Support}. This option requires
13741 hardware floating-point support to be enabled.
13747 Use (do not use) MIPS Digital Media Extension instructions.
13748 This option can only be used when generating 64-bit code and requires
13749 hardware floating-point support to be enabled.
13754 @opindex mno-mips3d
13755 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
13756 The option @option{-mips3d} implies @option{-mpaired-single}.
13762 Use (do not use) MT Multithreading instructions.
13766 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
13767 an explanation of the default and the way that the pointer size is
13772 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
13774 The default size of @code{int}s, @code{long}s and pointers depends on
13775 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
13776 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
13777 32-bit @code{long}s. Pointers are the same size as @code{long}s,
13778 or the same size as integer registers, whichever is smaller.
13784 Assume (do not assume) that all symbols have 32-bit values, regardless
13785 of the selected ABI@. This option is useful in combination with
13786 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
13787 to generate shorter and faster references to symbolic addresses.
13791 Put definitions of externally-visible data in a small data section
13792 if that data is no bigger than @var{num} bytes. GCC can then access
13793 the data more efficiently; see @option{-mgpopt} for details.
13795 The default @option{-G} option depends on the configuration.
13797 @item -mlocal-sdata
13798 @itemx -mno-local-sdata
13799 @opindex mlocal-sdata
13800 @opindex mno-local-sdata
13801 Extend (do not extend) the @option{-G} behavior to local data too,
13802 such as to static variables in C@. @option{-mlocal-sdata} is the
13803 default for all configurations.
13805 If the linker complains that an application is using too much small data,
13806 you might want to try rebuilding the less performance-critical parts with
13807 @option{-mno-local-sdata}. You might also want to build large
13808 libraries with @option{-mno-local-sdata}, so that the libraries leave
13809 more room for the main program.
13811 @item -mextern-sdata
13812 @itemx -mno-extern-sdata
13813 @opindex mextern-sdata
13814 @opindex mno-extern-sdata
13815 Assume (do not assume) that externally-defined data will be in
13816 a small data section if that data is within the @option{-G} limit.
13817 @option{-mextern-sdata} is the default for all configurations.
13819 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
13820 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
13821 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
13822 is placed in a small data section. If @var{Var} is defined by another
13823 module, you must either compile that module with a high-enough
13824 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
13825 definition. If @var{Var} is common, you must link the application
13826 with a high-enough @option{-G} setting.
13828 The easiest way of satisfying these restrictions is to compile
13829 and link every module with the same @option{-G} option. However,
13830 you may wish to build a library that supports several different
13831 small data limits. You can do this by compiling the library with
13832 the highest supported @option{-G} setting and additionally using
13833 @option{-mno-extern-sdata} to stop the library from making assumptions
13834 about externally-defined data.
13840 Use (do not use) GP-relative accesses for symbols that are known to be
13841 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
13842 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
13845 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
13846 might not hold the value of @code{_gp}. For example, if the code is
13847 part of a library that might be used in a boot monitor, programs that
13848 call boot monitor routines will pass an unknown value in @code{$gp}.
13849 (In such situations, the boot monitor itself would usually be compiled
13850 with @option{-G0}.)
13852 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
13853 @option{-mno-extern-sdata}.
13855 @item -membedded-data
13856 @itemx -mno-embedded-data
13857 @opindex membedded-data
13858 @opindex mno-embedded-data
13859 Allocate variables to the read-only data section first if possible, then
13860 next in the small data section if possible, otherwise in data. This gives
13861 slightly slower code than the default, but reduces the amount of RAM required
13862 when executing, and thus may be preferred for some embedded systems.
13864 @item -muninit-const-in-rodata
13865 @itemx -mno-uninit-const-in-rodata
13866 @opindex muninit-const-in-rodata
13867 @opindex mno-uninit-const-in-rodata
13868 Put uninitialized @code{const} variables in the read-only data section.
13869 This option is only meaningful in conjunction with @option{-membedded-data}.
13871 @item -mcode-readable=@var{setting}
13872 @opindex mcode-readable
13873 Specify whether GCC may generate code that reads from executable sections.
13874 There are three possible settings:
13877 @item -mcode-readable=yes
13878 Instructions may freely access executable sections. This is the
13881 @item -mcode-readable=pcrel
13882 MIPS16 PC-relative load instructions can access executable sections,
13883 but other instructions must not do so. This option is useful on 4KSc
13884 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
13885 It is also useful on processors that can be configured to have a dual
13886 instruction/data SRAM interface and that, like the M4K, automatically
13887 redirect PC-relative loads to the instruction RAM.
13889 @item -mcode-readable=no
13890 Instructions must not access executable sections. This option can be
13891 useful on targets that are configured to have a dual instruction/data
13892 SRAM interface but that (unlike the M4K) do not automatically redirect
13893 PC-relative loads to the instruction RAM.
13896 @item -msplit-addresses
13897 @itemx -mno-split-addresses
13898 @opindex msplit-addresses
13899 @opindex mno-split-addresses
13900 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
13901 relocation operators. This option has been superseded by
13902 @option{-mexplicit-relocs} but is retained for backwards compatibility.
13904 @item -mexplicit-relocs
13905 @itemx -mno-explicit-relocs
13906 @opindex mexplicit-relocs
13907 @opindex mno-explicit-relocs
13908 Use (do not use) assembler relocation operators when dealing with symbolic
13909 addresses. The alternative, selected by @option{-mno-explicit-relocs},
13910 is to use assembler macros instead.
13912 @option{-mexplicit-relocs} is the default if GCC was configured
13913 to use an assembler that supports relocation operators.
13915 @item -mcheck-zero-division
13916 @itemx -mno-check-zero-division
13917 @opindex mcheck-zero-division
13918 @opindex mno-check-zero-division
13919 Trap (do not trap) on integer division by zero.
13921 The default is @option{-mcheck-zero-division}.
13923 @item -mdivide-traps
13924 @itemx -mdivide-breaks
13925 @opindex mdivide-traps
13926 @opindex mdivide-breaks
13927 MIPS systems check for division by zero by generating either a
13928 conditional trap or a break instruction. Using traps results in
13929 smaller code, but is only supported on MIPS II and later. Also, some
13930 versions of the Linux kernel have a bug that prevents trap from
13931 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
13932 allow conditional traps on architectures that support them and
13933 @option{-mdivide-breaks} to force the use of breaks.
13935 The default is usually @option{-mdivide-traps}, but this can be
13936 overridden at configure time using @option{--with-divide=breaks}.
13937 Divide-by-zero checks can be completely disabled using
13938 @option{-mno-check-zero-division}.
13943 @opindex mno-memcpy
13944 Force (do not force) the use of @code{memcpy()} for non-trivial block
13945 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
13946 most constant-sized copies.
13949 @itemx -mno-long-calls
13950 @opindex mlong-calls
13951 @opindex mno-long-calls
13952 Disable (do not disable) use of the @code{jal} instruction. Calling
13953 functions using @code{jal} is more efficient but requires the caller
13954 and callee to be in the same 256 megabyte segment.
13956 This option has no effect on abicalls code. The default is
13957 @option{-mno-long-calls}.
13963 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
13964 instructions, as provided by the R4650 ISA@.
13967 @itemx -mno-fused-madd
13968 @opindex mfused-madd
13969 @opindex mno-fused-madd
13970 Enable (disable) use of the floating point multiply-accumulate
13971 instructions, when they are available. The default is
13972 @option{-mfused-madd}.
13974 When multiply-accumulate instructions are used, the intermediate
13975 product is calculated to infinite precision and is not subject to
13976 the FCSR Flush to Zero bit. This may be undesirable in some
13981 Tell the MIPS assembler to not run its preprocessor over user
13982 assembler files (with a @samp{.s} suffix) when assembling them.
13985 @itemx -mno-fix-r4000
13986 @opindex mfix-r4000
13987 @opindex mno-fix-r4000
13988 Work around certain R4000 CPU errata:
13991 A double-word or a variable shift may give an incorrect result if executed
13992 immediately after starting an integer division.
13994 A double-word or a variable shift may give an incorrect result if executed
13995 while an integer multiplication is in progress.
13997 An integer division may give an incorrect result if started in a delay slot
13998 of a taken branch or a jump.
14002 @itemx -mno-fix-r4400
14003 @opindex mfix-r4400
14004 @opindex mno-fix-r4400
14005 Work around certain R4400 CPU errata:
14008 A double-word or a variable shift may give an incorrect result if executed
14009 immediately after starting an integer division.
14013 @itemx -mno-fix-r10000
14014 @opindex mfix-r10000
14015 @opindex mno-fix-r10000
14016 Work around certain R10000 errata:
14019 @code{ll}/@code{sc} sequences may not behave atomically on revisions
14020 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
14023 This option can only be used if the target architecture supports
14024 branch-likely instructions. @option{-mfix-r10000} is the default when
14025 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
14029 @itemx -mno-fix-vr4120
14030 @opindex mfix-vr4120
14031 Work around certain VR4120 errata:
14034 @code{dmultu} does not always produce the correct result.
14036 @code{div} and @code{ddiv} do not always produce the correct result if one
14037 of the operands is negative.
14039 The workarounds for the division errata rely on special functions in
14040 @file{libgcc.a}. At present, these functions are only provided by
14041 the @code{mips64vr*-elf} configurations.
14043 Other VR4120 errata require a nop to be inserted between certain pairs of
14044 instructions. These errata are handled by the assembler, not by GCC itself.
14047 @opindex mfix-vr4130
14048 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
14049 workarounds are implemented by the assembler rather than by GCC,
14050 although GCC will avoid using @code{mflo} and @code{mfhi} if the
14051 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
14052 instructions are available instead.
14055 @itemx -mno-fix-sb1
14057 Work around certain SB-1 CPU core errata.
14058 (This flag currently works around the SB-1 revision 2
14059 ``F1'' and ``F2'' floating point errata.)
14061 @item -mr10k-cache-barrier=@var{setting}
14062 @opindex mr10k-cache-barrier
14063 Specify whether GCC should insert cache barriers to avoid the
14064 side-effects of speculation on R10K processors.
14066 In common with many processors, the R10K tries to predict the outcome
14067 of a conditional branch and speculatively executes instructions from
14068 the ``taken'' branch. It later aborts these instructions if the
14069 predicted outcome was wrong. However, on the R10K, even aborted
14070 instructions can have side effects.
14072 This problem only affects kernel stores and, depending on the system,
14073 kernel loads. As an example, a speculatively-executed store may load
14074 the target memory into cache and mark the cache line as dirty, even if
14075 the store itself is later aborted. If a DMA operation writes to the
14076 same area of memory before the ``dirty'' line is flushed, the cached
14077 data will overwrite the DMA-ed data. See the R10K processor manual
14078 for a full description, including other potential problems.
14080 One workaround is to insert cache barrier instructions before every memory
14081 access that might be speculatively executed and that might have side
14082 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
14083 controls GCC's implementation of this workaround. It assumes that
14084 aborted accesses to any byte in the following regions will not have
14089 the memory occupied by the current function's stack frame;
14092 the memory occupied by an incoming stack argument;
14095 the memory occupied by an object with a link-time-constant address.
14098 It is the kernel's responsibility to ensure that speculative
14099 accesses to these regions are indeed safe.
14101 If the input program contains a function declaration such as:
14107 then the implementation of @code{foo} must allow @code{j foo} and
14108 @code{jal foo} to be executed speculatively. GCC honors this
14109 restriction for functions it compiles itself. It expects non-GCC
14110 functions (such as hand-written assembly code) to do the same.
14112 The option has three forms:
14115 @item -mr10k-cache-barrier=load-store
14116 Insert a cache barrier before a load or store that might be
14117 speculatively executed and that might have side effects even
14120 @item -mr10k-cache-barrier=store
14121 Insert a cache barrier before a store that might be speculatively
14122 executed and that might have side effects even if aborted.
14124 @item -mr10k-cache-barrier=none
14125 Disable the insertion of cache barriers. This is the default setting.
14128 @item -mflush-func=@var{func}
14129 @itemx -mno-flush-func
14130 @opindex mflush-func
14131 Specifies the function to call to flush the I and D caches, or to not
14132 call any such function. If called, the function must take the same
14133 arguments as the common @code{_flush_func()}, that is, the address of the
14134 memory range for which the cache is being flushed, the size of the
14135 memory range, and the number 3 (to flush both caches). The default
14136 depends on the target GCC was configured for, but commonly is either
14137 @samp{_flush_func} or @samp{__cpu_flush}.
14139 @item mbranch-cost=@var{num}
14140 @opindex mbranch-cost
14141 Set the cost of branches to roughly @var{num} ``simple'' instructions.
14142 This cost is only a heuristic and is not guaranteed to produce
14143 consistent results across releases. A zero cost redundantly selects
14144 the default, which is based on the @option{-mtune} setting.
14146 @item -mbranch-likely
14147 @itemx -mno-branch-likely
14148 @opindex mbranch-likely
14149 @opindex mno-branch-likely
14150 Enable or disable use of Branch Likely instructions, regardless of the
14151 default for the selected architecture. By default, Branch Likely
14152 instructions may be generated if they are supported by the selected
14153 architecture. An exception is for the MIPS32 and MIPS64 architectures
14154 and processors which implement those architectures; for those, Branch
14155 Likely instructions will not be generated by default because the MIPS32
14156 and MIPS64 architectures specifically deprecate their use.
14158 @item -mfp-exceptions
14159 @itemx -mno-fp-exceptions
14160 @opindex mfp-exceptions
14161 Specifies whether FP exceptions are enabled. This affects how we schedule
14162 FP instructions for some processors. The default is that FP exceptions are
14165 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
14166 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
14169 @item -mvr4130-align
14170 @itemx -mno-vr4130-align
14171 @opindex mvr4130-align
14172 The VR4130 pipeline is two-way superscalar, but can only issue two
14173 instructions together if the first one is 8-byte aligned. When this
14174 option is enabled, GCC will align pairs of instructions that it
14175 thinks should execute in parallel.
14177 This option only has an effect when optimizing for the VR4130.
14178 It normally makes code faster, but at the expense of making it bigger.
14179 It is enabled by default at optimization level @option{-O3}.
14184 Enable (disable) generation of @code{synci} instructions on
14185 architectures that support it. The @code{synci} instructions (if
14186 enabled) will be generated when @code{__builtin___clear_cache()} is
14189 This option defaults to @code{-mno-synci}, but the default can be
14190 overridden by configuring with @code{--with-synci}.
14192 When compiling code for single processor systems, it is generally safe
14193 to use @code{synci}. However, on many multi-core (SMP) systems, it
14194 will not invalidate the instruction caches on all cores and may lead
14195 to undefined behavior.
14197 @item -mrelax-pic-calls
14198 @itemx -mno-relax-pic-calls
14199 @opindex mrelax-pic-calls
14200 Try to turn PIC calls that are normally dispatched via register
14201 @code{$25} into direct calls. This is only possible if the linker can
14202 resolve the destination at link-time and if the destination is within
14203 range for a direct call.
14205 @option{-mrelax-pic-calls} is the default if GCC was configured to use
14206 an assembler and a linker that supports the @code{.reloc} assembly
14207 directive and @code{-mexplicit-relocs} is in effect. With
14208 @code{-mno-explicit-relocs}, this optimization can be performed by the
14209 assembler and the linker alone without help from the compiler.
14213 @subsection MMIX Options
14214 @cindex MMIX Options
14216 These options are defined for the MMIX:
14220 @itemx -mno-libfuncs
14222 @opindex mno-libfuncs
14223 Specify that intrinsic library functions are being compiled, passing all
14224 values in registers, no matter the size.
14227 @itemx -mno-epsilon
14229 @opindex mno-epsilon
14230 Generate floating-point comparison instructions that compare with respect
14231 to the @code{rE} epsilon register.
14233 @item -mabi=mmixware
14235 @opindex mabi=mmixware
14237 Generate code that passes function parameters and return values that (in
14238 the called function) are seen as registers @code{$0} and up, as opposed to
14239 the GNU ABI which uses global registers @code{$231} and up.
14241 @item -mzero-extend
14242 @itemx -mno-zero-extend
14243 @opindex mzero-extend
14244 @opindex mno-zero-extend
14245 When reading data from memory in sizes shorter than 64 bits, use (do not
14246 use) zero-extending load instructions by default, rather than
14247 sign-extending ones.
14250 @itemx -mno-knuthdiv
14252 @opindex mno-knuthdiv
14253 Make the result of a division yielding a remainder have the same sign as
14254 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
14255 remainder follows the sign of the dividend. Both methods are
14256 arithmetically valid, the latter being almost exclusively used.
14258 @item -mtoplevel-symbols
14259 @itemx -mno-toplevel-symbols
14260 @opindex mtoplevel-symbols
14261 @opindex mno-toplevel-symbols
14262 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
14263 code can be used with the @code{PREFIX} assembly directive.
14267 Generate an executable in the ELF format, rather than the default
14268 @samp{mmo} format used by the @command{mmix} simulator.
14270 @item -mbranch-predict
14271 @itemx -mno-branch-predict
14272 @opindex mbranch-predict
14273 @opindex mno-branch-predict
14274 Use (do not use) the probable-branch instructions, when static branch
14275 prediction indicates a probable branch.
14277 @item -mbase-addresses
14278 @itemx -mno-base-addresses
14279 @opindex mbase-addresses
14280 @opindex mno-base-addresses
14281 Generate (do not generate) code that uses @emph{base addresses}. Using a
14282 base address automatically generates a request (handled by the assembler
14283 and the linker) for a constant to be set up in a global register. The
14284 register is used for one or more base address requests within the range 0
14285 to 255 from the value held in the register. The generally leads to short
14286 and fast code, but the number of different data items that can be
14287 addressed is limited. This means that a program that uses lots of static
14288 data may require @option{-mno-base-addresses}.
14290 @item -msingle-exit
14291 @itemx -mno-single-exit
14292 @opindex msingle-exit
14293 @opindex mno-single-exit
14294 Force (do not force) generated code to have a single exit point in each
14298 @node MN10300 Options
14299 @subsection MN10300 Options
14300 @cindex MN10300 options
14302 These @option{-m} options are defined for Matsushita MN10300 architectures:
14307 Generate code to avoid bugs in the multiply instructions for the MN10300
14308 processors. This is the default.
14310 @item -mno-mult-bug
14311 @opindex mno-mult-bug
14312 Do not generate code to avoid bugs in the multiply instructions for the
14313 MN10300 processors.
14317 Generate code which uses features specific to the AM33 processor.
14321 Do not generate code which uses features specific to the AM33 processor. This
14324 @item -mreturn-pointer-on-d0
14325 @opindex mreturn-pointer-on-d0
14326 When generating a function which returns a pointer, return the pointer
14327 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
14328 only in a0, and attempts to call such functions without a prototype
14329 would result in errors. Note that this option is on by default; use
14330 @option{-mno-return-pointer-on-d0} to disable it.
14334 Do not link in the C run-time initialization object file.
14338 Indicate to the linker that it should perform a relaxation optimization pass
14339 to shorten branches, calls and absolute memory addresses. This option only
14340 has an effect when used on the command line for the final link step.
14342 This option makes symbolic debugging impossible.
14345 @node PDP-11 Options
14346 @subsection PDP-11 Options
14347 @cindex PDP-11 Options
14349 These options are defined for the PDP-11:
14354 Use hardware FPP floating point. This is the default. (FIS floating
14355 point on the PDP-11/40 is not supported.)
14358 @opindex msoft-float
14359 Do not use hardware floating point.
14363 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
14367 Return floating-point results in memory. This is the default.
14371 Generate code for a PDP-11/40.
14375 Generate code for a PDP-11/45. This is the default.
14379 Generate code for a PDP-11/10.
14381 @item -mbcopy-builtin
14382 @opindex mbcopy-builtin
14383 Use inline @code{movmemhi} patterns for copying memory. This is the
14388 Do not use inline @code{movmemhi} patterns for copying memory.
14394 Use 16-bit @code{int}. This is the default.
14400 Use 32-bit @code{int}.
14403 @itemx -mno-float32
14405 @opindex mno-float32
14406 Use 64-bit @code{float}. This is the default.
14409 @itemx -mno-float64
14411 @opindex mno-float64
14412 Use 32-bit @code{float}.
14416 Use @code{abshi2} pattern. This is the default.
14420 Do not use @code{abshi2} pattern.
14422 @item -mbranch-expensive
14423 @opindex mbranch-expensive
14424 Pretend that branches are expensive. This is for experimenting with
14425 code generation only.
14427 @item -mbranch-cheap
14428 @opindex mbranch-cheap
14429 Do not pretend that branches are expensive. This is the default.
14433 Generate code for a system with split I&D@.
14437 Generate code for a system without split I&D@. This is the default.
14441 Use Unix assembler syntax. This is the default when configured for
14442 @samp{pdp11-*-bsd}.
14446 Use DEC assembler syntax. This is the default when configured for any
14447 PDP-11 target other than @samp{pdp11-*-bsd}.
14450 @node picoChip Options
14451 @subsection picoChip Options
14452 @cindex picoChip options
14454 These @samp{-m} options are defined for picoChip implementations:
14458 @item -mae=@var{ae_type}
14460 Set the instruction set, register set, and instruction scheduling
14461 parameters for array element type @var{ae_type}. Supported values
14462 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
14464 @option{-mae=ANY} selects a completely generic AE type. Code
14465 generated with this option will run on any of the other AE types. The
14466 code will not be as efficient as it would be if compiled for a specific
14467 AE type, and some types of operation (e.g., multiplication) will not
14468 work properly on all types of AE.
14470 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
14471 for compiled code, and is the default.
14473 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
14474 option may suffer from poor performance of byte (char) manipulation,
14475 since the DSP AE does not provide hardware support for byte load/stores.
14477 @item -msymbol-as-address
14478 Enable the compiler to directly use a symbol name as an address in a
14479 load/store instruction, without first loading it into a
14480 register. Typically, the use of this option will generate larger
14481 programs, which run faster than when the option isn't used. However, the
14482 results vary from program to program, so it is left as a user option,
14483 rather than being permanently enabled.
14485 @item -mno-inefficient-warnings
14486 Disables warnings about the generation of inefficient code. These
14487 warnings can be generated, for example, when compiling code which
14488 performs byte-level memory operations on the MAC AE type. The MAC AE has
14489 no hardware support for byte-level memory operations, so all byte
14490 load/stores must be synthesized from word load/store operations. This is
14491 inefficient and a warning will be generated indicating to the programmer
14492 that they should rewrite the code to avoid byte operations, or to target
14493 an AE type which has the necessary hardware support. This option enables
14494 the warning to be turned off.
14498 @node PowerPC Options
14499 @subsection PowerPC Options
14500 @cindex PowerPC options
14502 These are listed under @xref{RS/6000 and PowerPC Options}.
14504 @node RS/6000 and PowerPC Options
14505 @subsection IBM RS/6000 and PowerPC Options
14506 @cindex RS/6000 and PowerPC Options
14507 @cindex IBM RS/6000 and PowerPC Options
14509 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
14516 @itemx -mno-powerpc
14517 @itemx -mpowerpc-gpopt
14518 @itemx -mno-powerpc-gpopt
14519 @itemx -mpowerpc-gfxopt
14520 @itemx -mno-powerpc-gfxopt
14522 @itemx -mno-powerpc64
14526 @itemx -mno-popcntb
14528 @itemx -mno-popcntd
14536 @itemx -mno-hard-dfp
14540 @opindex mno-power2
14542 @opindex mno-powerpc
14543 @opindex mpowerpc-gpopt
14544 @opindex mno-powerpc-gpopt
14545 @opindex mpowerpc-gfxopt
14546 @opindex mno-powerpc-gfxopt
14547 @opindex mpowerpc64
14548 @opindex mno-powerpc64
14552 @opindex mno-popcntb
14554 @opindex mno-popcntd
14560 @opindex mno-mfpgpr
14562 @opindex mno-hard-dfp
14563 GCC supports two related instruction set architectures for the
14564 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
14565 instructions supported by the @samp{rios} chip set used in the original
14566 RS/6000 systems and the @dfn{PowerPC} instruction set is the
14567 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
14568 the IBM 4xx, 6xx, and follow-on microprocessors.
14570 Neither architecture is a subset of the other. However there is a
14571 large common subset of instructions supported by both. An MQ
14572 register is included in processors supporting the POWER architecture.
14574 You use these options to specify which instructions are available on the
14575 processor you are using. The default value of these options is
14576 determined when configuring GCC@. Specifying the
14577 @option{-mcpu=@var{cpu_type}} overrides the specification of these
14578 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
14579 rather than the options listed above.
14581 The @option{-mpower} option allows GCC to generate instructions that
14582 are found only in the POWER architecture and to use the MQ register.
14583 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
14584 to generate instructions that are present in the POWER2 architecture but
14585 not the original POWER architecture.
14587 The @option{-mpowerpc} option allows GCC to generate instructions that
14588 are found only in the 32-bit subset of the PowerPC architecture.
14589 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
14590 GCC to use the optional PowerPC architecture instructions in the
14591 General Purpose group, including floating-point square root. Specifying
14592 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
14593 use the optional PowerPC architecture instructions in the Graphics
14594 group, including floating-point select.
14596 The @option{-mmfcrf} option allows GCC to generate the move from
14597 condition register field instruction implemented on the POWER4
14598 processor and other processors that support the PowerPC V2.01
14600 The @option{-mpopcntb} option allows GCC to generate the popcount and
14601 double precision FP reciprocal estimate instruction implemented on the
14602 POWER5 processor and other processors that support the PowerPC V2.02
14604 The @option{-mpopcntd} option allows GCC to generate the popcount
14605 instruction implemented on the POWER7 processor and other processors
14606 that support the PowerPC V2.06 architecture.
14607 The @option{-mfprnd} option allows GCC to generate the FP round to
14608 integer instructions implemented on the POWER5+ processor and other
14609 processors that support the PowerPC V2.03 architecture.
14610 The @option{-mcmpb} option allows GCC to generate the compare bytes
14611 instruction implemented on the POWER6 processor and other processors
14612 that support the PowerPC V2.05 architecture.
14613 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
14614 general purpose register instructions implemented on the POWER6X
14615 processor and other processors that support the extended PowerPC V2.05
14617 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
14618 point instructions implemented on some POWER processors.
14620 The @option{-mpowerpc64} option allows GCC to generate the additional
14621 64-bit instructions that are found in the full PowerPC64 architecture
14622 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
14623 @option{-mno-powerpc64}.
14625 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
14626 will use only the instructions in the common subset of both
14627 architectures plus some special AIX common-mode calls, and will not use
14628 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
14629 permits GCC to use any instruction from either architecture and to
14630 allow use of the MQ register; specify this for the Motorola MPC601.
14632 @item -mnew-mnemonics
14633 @itemx -mold-mnemonics
14634 @opindex mnew-mnemonics
14635 @opindex mold-mnemonics
14636 Select which mnemonics to use in the generated assembler code. With
14637 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
14638 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
14639 assembler mnemonics defined for the POWER architecture. Instructions
14640 defined in only one architecture have only one mnemonic; GCC uses that
14641 mnemonic irrespective of which of these options is specified.
14643 GCC defaults to the mnemonics appropriate for the architecture in
14644 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
14645 value of these option. Unless you are building a cross-compiler, you
14646 should normally not specify either @option{-mnew-mnemonics} or
14647 @option{-mold-mnemonics}, but should instead accept the default.
14649 @item -mcpu=@var{cpu_type}
14651 Set architecture type, register usage, choice of mnemonics, and
14652 instruction scheduling parameters for machine type @var{cpu_type}.
14653 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
14654 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
14655 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
14656 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
14657 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
14658 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
14659 @samp{e300c3}, @samp{e500mc}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
14660 @samp{power}, @samp{power2}, @samp{power3}, @samp{power4},
14661 @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x}, @samp{power7},
14662 @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
14663 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
14665 @option{-mcpu=common} selects a completely generic processor. Code
14666 generated under this option will run on any POWER or PowerPC processor.
14667 GCC will use only the instructions in the common subset of both
14668 architectures, and will not use the MQ register. GCC assumes a generic
14669 processor model for scheduling purposes.
14671 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
14672 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
14673 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
14674 types, with an appropriate, generic processor model assumed for
14675 scheduling purposes.
14677 The other options specify a specific processor. Code generated under
14678 those options will run best on that processor, and may not run at all on
14681 The @option{-mcpu} options automatically enable or disable the
14684 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
14685 -mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol
14686 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
14687 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx}
14689 The particular options set for any particular CPU will vary between
14690 compiler versions, depending on what setting seems to produce optimal
14691 code for that CPU; it doesn't necessarily reflect the actual hardware's
14692 capabilities. If you wish to set an individual option to a particular
14693 value, you may specify it after the @option{-mcpu} option, like
14694 @samp{-mcpu=970 -mno-altivec}.
14696 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
14697 not enabled or disabled by the @option{-mcpu} option at present because
14698 AIX does not have full support for these options. You may still
14699 enable or disable them individually if you're sure it'll work in your
14702 @item -mtune=@var{cpu_type}
14704 Set the instruction scheduling parameters for machine type
14705 @var{cpu_type}, but do not set the architecture type, register usage, or
14706 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
14707 values for @var{cpu_type} are used for @option{-mtune} as for
14708 @option{-mcpu}. If both are specified, the code generated will use the
14709 architecture, registers, and mnemonics set by @option{-mcpu}, but the
14710 scheduling parameters set by @option{-mtune}.
14716 Generate code to compute division as reciprocal estimate and iterative
14717 refinement, creating opportunities for increased throughput. This
14718 feature requires: optional PowerPC Graphics instruction set for single
14719 precision and FRE instruction for double precision, assuming divides
14720 cannot generate user-visible traps, and the domain values not include
14721 Infinities, denormals or zero denominator.
14724 @itemx -mno-altivec
14726 @opindex mno-altivec
14727 Generate code that uses (does not use) AltiVec instructions, and also
14728 enable the use of built-in functions that allow more direct access to
14729 the AltiVec instruction set. You may also need to set
14730 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
14736 @opindex mno-vrsave
14737 Generate VRSAVE instructions when generating AltiVec code.
14739 @item -mgen-cell-microcode
14740 @opindex mgen-cell-microcode
14741 Generate Cell microcode instructions
14743 @item -mwarn-cell-microcode
14744 @opindex mwarn-cell-microcode
14745 Warning when a Cell microcode instruction is going to emitted. An example
14746 of a Cell microcode instruction is a variable shift.
14749 @opindex msecure-plt
14750 Generate code that allows ld and ld.so to build executables and shared
14751 libraries with non-exec .plt and .got sections. This is a PowerPC
14752 32-bit SYSV ABI option.
14756 Generate code that uses a BSS .plt section that ld.so fills in, and
14757 requires .plt and .got sections that are both writable and executable.
14758 This is a PowerPC 32-bit SYSV ABI option.
14764 This switch enables or disables the generation of ISEL instructions.
14766 @item -misel=@var{yes/no}
14767 This switch has been deprecated. Use @option{-misel} and
14768 @option{-mno-isel} instead.
14774 This switch enables or disables the generation of SPE simd
14780 @opindex mno-paired
14781 This switch enables or disables the generation of PAIRED simd
14784 @item -mspe=@var{yes/no}
14785 This option has been deprecated. Use @option{-mspe} and
14786 @option{-mno-spe} instead.
14792 Generate code that uses (does not use) vector/scalar (VSX)
14793 instructions, and also enable the use of built-in functions that allow
14794 more direct access to the VSX instruction set.
14796 @item -mfloat-gprs=@var{yes/single/double/no}
14797 @itemx -mfloat-gprs
14798 @opindex mfloat-gprs
14799 This switch enables or disables the generation of floating point
14800 operations on the general purpose registers for architectures that
14803 The argument @var{yes} or @var{single} enables the use of
14804 single-precision floating point operations.
14806 The argument @var{double} enables the use of single and
14807 double-precision floating point operations.
14809 The argument @var{no} disables floating point operations on the
14810 general purpose registers.
14812 This option is currently only available on the MPC854x.
14818 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
14819 targets (including GNU/Linux). The 32-bit environment sets int, long
14820 and pointer to 32 bits and generates code that runs on any PowerPC
14821 variant. The 64-bit environment sets int to 32 bits and long and
14822 pointer to 64 bits, and generates code for PowerPC64, as for
14823 @option{-mpowerpc64}.
14826 @itemx -mno-fp-in-toc
14827 @itemx -mno-sum-in-toc
14828 @itemx -mminimal-toc
14830 @opindex mno-fp-in-toc
14831 @opindex mno-sum-in-toc
14832 @opindex mminimal-toc
14833 Modify generation of the TOC (Table Of Contents), which is created for
14834 every executable file. The @option{-mfull-toc} option is selected by
14835 default. In that case, GCC will allocate at least one TOC entry for
14836 each unique non-automatic variable reference in your program. GCC
14837 will also place floating-point constants in the TOC@. However, only
14838 16,384 entries are available in the TOC@.
14840 If you receive a linker error message that saying you have overflowed
14841 the available TOC space, you can reduce the amount of TOC space used
14842 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
14843 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
14844 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
14845 generate code to calculate the sum of an address and a constant at
14846 run-time instead of putting that sum into the TOC@. You may specify one
14847 or both of these options. Each causes GCC to produce very slightly
14848 slower and larger code at the expense of conserving TOC space.
14850 If you still run out of space in the TOC even when you specify both of
14851 these options, specify @option{-mminimal-toc} instead. This option causes
14852 GCC to make only one TOC entry for every file. When you specify this
14853 option, GCC will produce code that is slower and larger but which
14854 uses extremely little TOC space. You may wish to use this option
14855 only on files that contain less frequently executed code.
14861 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
14862 @code{long} type, and the infrastructure needed to support them.
14863 Specifying @option{-maix64} implies @option{-mpowerpc64} and
14864 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
14865 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
14868 @itemx -mno-xl-compat
14869 @opindex mxl-compat
14870 @opindex mno-xl-compat
14871 Produce code that conforms more closely to IBM XL compiler semantics
14872 when using AIX-compatible ABI@. Pass floating-point arguments to
14873 prototyped functions beyond the register save area (RSA) on the stack
14874 in addition to argument FPRs. Do not assume that most significant
14875 double in 128-bit long double value is properly rounded when comparing
14876 values and converting to double. Use XL symbol names for long double
14879 The AIX calling convention was extended but not initially documented to
14880 handle an obscure K&R C case of calling a function that takes the
14881 address of its arguments with fewer arguments than declared. IBM XL
14882 compilers access floating point arguments which do not fit in the
14883 RSA from the stack when a subroutine is compiled without
14884 optimization. Because always storing floating-point arguments on the
14885 stack is inefficient and rarely needed, this option is not enabled by
14886 default and only is necessary when calling subroutines compiled by IBM
14887 XL compilers without optimization.
14891 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
14892 application written to use message passing with special startup code to
14893 enable the application to run. The system must have PE installed in the
14894 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
14895 must be overridden with the @option{-specs=} option to specify the
14896 appropriate directory location. The Parallel Environment does not
14897 support threads, so the @option{-mpe} option and the @option{-pthread}
14898 option are incompatible.
14900 @item -malign-natural
14901 @itemx -malign-power
14902 @opindex malign-natural
14903 @opindex malign-power
14904 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
14905 @option{-malign-natural} overrides the ABI-defined alignment of larger
14906 types, such as floating-point doubles, on their natural size-based boundary.
14907 The option @option{-malign-power} instructs GCC to follow the ABI-specified
14908 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
14910 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
14914 @itemx -mhard-float
14915 @opindex msoft-float
14916 @opindex mhard-float
14917 Generate code that does not use (uses) the floating-point register set.
14918 Software floating point emulation is provided if you use the
14919 @option{-msoft-float} option, and pass the option to GCC when linking.
14921 @item -msingle-float
14922 @itemx -mdouble-float
14923 @opindex msingle-float
14924 @opindex mdouble-float
14925 Generate code for single or double-precision floating point operations.
14926 @option{-mdouble-float} implies @option{-msingle-float}.
14929 @opindex msimple-fpu
14930 Do not generate sqrt and div instructions for hardware floating point unit.
14934 Specify type of floating point unit. Valid values are @var{sp_lite}
14935 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
14936 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
14937 and @var{dp_full} (equivalent to -mdouble-float).
14940 @opindex mxilinx-fpu
14941 Perform optimizations for floating point unit on Xilinx PPC 405/440.
14944 @itemx -mno-multiple
14946 @opindex mno-multiple
14947 Generate code that uses (does not use) the load multiple word
14948 instructions and the store multiple word instructions. These
14949 instructions are generated by default on POWER systems, and not
14950 generated on PowerPC systems. Do not use @option{-mmultiple} on little
14951 endian PowerPC systems, since those instructions do not work when the
14952 processor is in little endian mode. The exceptions are PPC740 and
14953 PPC750 which permit the instructions usage in little endian mode.
14958 @opindex mno-string
14959 Generate code that uses (does not use) the load string instructions
14960 and the store string word instructions to save multiple registers and
14961 do small block moves. These instructions are generated by default on
14962 POWER systems, and not generated on PowerPC systems. Do not use
14963 @option{-mstring} on little endian PowerPC systems, since those
14964 instructions do not work when the processor is in little endian mode.
14965 The exceptions are PPC740 and PPC750 which permit the instructions
14966 usage in little endian mode.
14971 @opindex mno-update
14972 Generate code that uses (does not use) the load or store instructions
14973 that update the base register to the address of the calculated memory
14974 location. These instructions are generated by default. If you use
14975 @option{-mno-update}, there is a small window between the time that the
14976 stack pointer is updated and the address of the previous frame is
14977 stored, which means code that walks the stack frame across interrupts or
14978 signals may get corrupted data.
14980 @item -mavoid-indexed-addresses
14981 @item -mno-avoid-indexed-addresses
14982 @opindex mavoid-indexed-addresses
14983 @opindex mno-avoid-indexed-addresses
14984 Generate code that tries to avoid (not avoid) the use of indexed load
14985 or store instructions. These instructions can incur a performance
14986 penalty on Power6 processors in certain situations, such as when
14987 stepping through large arrays that cross a 16M boundary. This option
14988 is enabled by default when targetting Power6 and disabled otherwise.
14991 @itemx -mno-fused-madd
14992 @opindex mfused-madd
14993 @opindex mno-fused-madd
14994 Generate code that uses (does not use) the floating point multiply and
14995 accumulate instructions. These instructions are generated by default if
14996 hardware floating is used.
15002 Generate code that uses (does not use) the half-word multiply and
15003 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
15004 These instructions are generated by default when targetting those
15011 Generate code that uses (does not use) the string-search @samp{dlmzb}
15012 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
15013 generated by default when targetting those processors.
15015 @item -mno-bit-align
15017 @opindex mno-bit-align
15018 @opindex mbit-align
15019 On System V.4 and embedded PowerPC systems do not (do) force structures
15020 and unions that contain bit-fields to be aligned to the base type of the
15023 For example, by default a structure containing nothing but 8
15024 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
15025 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
15026 the structure would be aligned to a 1 byte boundary and be one byte in
15029 @item -mno-strict-align
15030 @itemx -mstrict-align
15031 @opindex mno-strict-align
15032 @opindex mstrict-align
15033 On System V.4 and embedded PowerPC systems do not (do) assume that
15034 unaligned memory references will be handled by the system.
15036 @item -mrelocatable
15037 @itemx -mno-relocatable
15038 @opindex mrelocatable
15039 @opindex mno-relocatable
15040 On embedded PowerPC systems generate code that allows (does not allow)
15041 the program to be relocated to a different address at runtime. If you
15042 use @option{-mrelocatable} on any module, all objects linked together must
15043 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
15045 @item -mrelocatable-lib
15046 @itemx -mno-relocatable-lib
15047 @opindex mrelocatable-lib
15048 @opindex mno-relocatable-lib
15049 On embedded PowerPC systems generate code that allows (does not allow)
15050 the program to be relocated to a different address at runtime. Modules
15051 compiled with @option{-mrelocatable-lib} can be linked with either modules
15052 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
15053 with modules compiled with the @option{-mrelocatable} options.
15059 On System V.4 and embedded PowerPC systems do not (do) assume that
15060 register 2 contains a pointer to a global area pointing to the addresses
15061 used in the program.
15064 @itemx -mlittle-endian
15066 @opindex mlittle-endian
15067 On System V.4 and embedded PowerPC systems compile code for the
15068 processor in little endian mode. The @option{-mlittle-endian} option is
15069 the same as @option{-mlittle}.
15072 @itemx -mbig-endian
15074 @opindex mbig-endian
15075 On System V.4 and embedded PowerPC systems compile code for the
15076 processor in big endian mode. The @option{-mbig-endian} option is
15077 the same as @option{-mbig}.
15079 @item -mdynamic-no-pic
15080 @opindex mdynamic-no-pic
15081 On Darwin and Mac OS X systems, compile code so that it is not
15082 relocatable, but that its external references are relocatable. The
15083 resulting code is suitable for applications, but not shared
15086 @item -mprioritize-restricted-insns=@var{priority}
15087 @opindex mprioritize-restricted-insns
15088 This option controls the priority that is assigned to
15089 dispatch-slot restricted instructions during the second scheduling
15090 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
15091 @var{no/highest/second-highest} priority to dispatch slot restricted
15094 @item -msched-costly-dep=@var{dependence_type}
15095 @opindex msched-costly-dep
15096 This option controls which dependences are considered costly
15097 by the target during instruction scheduling. The argument
15098 @var{dependence_type} takes one of the following values:
15099 @var{no}: no dependence is costly,
15100 @var{all}: all dependences are costly,
15101 @var{true_store_to_load}: a true dependence from store to load is costly,
15102 @var{store_to_load}: any dependence from store to load is costly,
15103 @var{number}: any dependence which latency >= @var{number} is costly.
15105 @item -minsert-sched-nops=@var{scheme}
15106 @opindex minsert-sched-nops
15107 This option controls which nop insertion scheme will be used during
15108 the second scheduling pass. The argument @var{scheme} takes one of the
15110 @var{no}: Don't insert nops.
15111 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
15112 according to the scheduler's grouping.
15113 @var{regroup_exact}: Insert nops to force costly dependent insns into
15114 separate groups. Insert exactly as many nops as needed to force an insn
15115 to a new group, according to the estimated processor grouping.
15116 @var{number}: Insert nops to force costly dependent insns into
15117 separate groups. Insert @var{number} nops to force an insn to a new group.
15120 @opindex mcall-sysv
15121 On System V.4 and embedded PowerPC systems compile code using calling
15122 conventions that adheres to the March 1995 draft of the System V
15123 Application Binary Interface, PowerPC processor supplement. This is the
15124 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
15126 @item -mcall-sysv-eabi
15128 @opindex mcall-sysv-eabi
15129 @opindex mcall-eabi
15130 Specify both @option{-mcall-sysv} and @option{-meabi} options.
15132 @item -mcall-sysv-noeabi
15133 @opindex mcall-sysv-noeabi
15134 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
15136 @item -mcall-aixdesc
15138 On System V.4 and embedded PowerPC systems compile code for the AIX
15142 @opindex mcall-linux
15143 On System V.4 and embedded PowerPC systems compile code for the
15144 Linux-based GNU system.
15148 On System V.4 and embedded PowerPC systems compile code for the
15149 Hurd-based GNU system.
15151 @item -mcall-freebsd
15152 @opindex mcall-freebsd
15153 On System V.4 and embedded PowerPC systems compile code for the
15154 FreeBSD operating system.
15156 @item -mcall-netbsd
15157 @opindex mcall-netbsd
15158 On System V.4 and embedded PowerPC systems compile code for the
15159 NetBSD operating system.
15161 @item -mcall-openbsd
15162 @opindex mcall-netbsd
15163 On System V.4 and embedded PowerPC systems compile code for the
15164 OpenBSD operating system.
15166 @item -maix-struct-return
15167 @opindex maix-struct-return
15168 Return all structures in memory (as specified by the AIX ABI)@.
15170 @item -msvr4-struct-return
15171 @opindex msvr4-struct-return
15172 Return structures smaller than 8 bytes in registers (as specified by the
15175 @item -mabi=@var{abi-type}
15177 Extend the current ABI with a particular extension, or remove such extension.
15178 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
15179 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
15183 Extend the current ABI with SPE ABI extensions. This does not change
15184 the default ABI, instead it adds the SPE ABI extensions to the current
15188 @opindex mabi=no-spe
15189 Disable Booke SPE ABI extensions for the current ABI@.
15191 @item -mabi=ibmlongdouble
15192 @opindex mabi=ibmlongdouble
15193 Change the current ABI to use IBM extended precision long double.
15194 This is a PowerPC 32-bit SYSV ABI option.
15196 @item -mabi=ieeelongdouble
15197 @opindex mabi=ieeelongdouble
15198 Change the current ABI to use IEEE extended precision long double.
15199 This is a PowerPC 32-bit Linux ABI option.
15202 @itemx -mno-prototype
15203 @opindex mprototype
15204 @opindex mno-prototype
15205 On System V.4 and embedded PowerPC systems assume that all calls to
15206 variable argument functions are properly prototyped. Otherwise, the
15207 compiler must insert an instruction before every non prototyped call to
15208 set or clear bit 6 of the condition code register (@var{CR}) to
15209 indicate whether floating point values were passed in the floating point
15210 registers in case the function takes a variable arguments. With
15211 @option{-mprototype}, only calls to prototyped variable argument functions
15212 will set or clear the bit.
15216 On embedded PowerPC systems, assume that the startup module is called
15217 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
15218 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
15223 On embedded PowerPC systems, assume that the startup module is called
15224 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
15229 On embedded PowerPC systems, assume that the startup module is called
15230 @file{crt0.o} and the standard C libraries are @file{libads.a} and
15233 @item -myellowknife
15234 @opindex myellowknife
15235 On embedded PowerPC systems, assume that the startup module is called
15236 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
15241 On System V.4 and embedded PowerPC systems, specify that you are
15242 compiling for a VxWorks system.
15246 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
15247 header to indicate that @samp{eabi} extended relocations are used.
15253 On System V.4 and embedded PowerPC systems do (do not) adhere to the
15254 Embedded Applications Binary Interface (eabi) which is a set of
15255 modifications to the System V.4 specifications. Selecting @option{-meabi}
15256 means that the stack is aligned to an 8 byte boundary, a function
15257 @code{__eabi} is called to from @code{main} to set up the eabi
15258 environment, and the @option{-msdata} option can use both @code{r2} and
15259 @code{r13} to point to two separate small data areas. Selecting
15260 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
15261 do not call an initialization function from @code{main}, and the
15262 @option{-msdata} option will only use @code{r13} to point to a single
15263 small data area. The @option{-meabi} option is on by default if you
15264 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
15267 @opindex msdata=eabi
15268 On System V.4 and embedded PowerPC systems, put small initialized
15269 @code{const} global and static data in the @samp{.sdata2} section, which
15270 is pointed to by register @code{r2}. Put small initialized
15271 non-@code{const} global and static data in the @samp{.sdata} section,
15272 which is pointed to by register @code{r13}. Put small uninitialized
15273 global and static data in the @samp{.sbss} section, which is adjacent to
15274 the @samp{.sdata} section. The @option{-msdata=eabi} option is
15275 incompatible with the @option{-mrelocatable} option. The
15276 @option{-msdata=eabi} option also sets the @option{-memb} option.
15279 @opindex msdata=sysv
15280 On System V.4 and embedded PowerPC systems, put small global and static
15281 data in the @samp{.sdata} section, which is pointed to by register
15282 @code{r13}. Put small uninitialized global and static data in the
15283 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
15284 The @option{-msdata=sysv} option is incompatible with the
15285 @option{-mrelocatable} option.
15287 @item -msdata=default
15289 @opindex msdata=default
15291 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
15292 compile code the same as @option{-msdata=eabi}, otherwise compile code the
15293 same as @option{-msdata=sysv}.
15296 @opindex msdata=data
15297 On System V.4 and embedded PowerPC systems, put small global
15298 data in the @samp{.sdata} section. Put small uninitialized global
15299 data in the @samp{.sbss} section. Do not use register @code{r13}
15300 to address small data however. This is the default behavior unless
15301 other @option{-msdata} options are used.
15305 @opindex msdata=none
15307 On embedded PowerPC systems, put all initialized global and static data
15308 in the @samp{.data} section, and all uninitialized data in the
15309 @samp{.bss} section.
15313 @cindex smaller data references (PowerPC)
15314 @cindex .sdata/.sdata2 references (PowerPC)
15315 On embedded PowerPC systems, put global and static items less than or
15316 equal to @var{num} bytes into the small data or bss sections instead of
15317 the normal data or bss section. By default, @var{num} is 8. The
15318 @option{-G @var{num}} switch is also passed to the linker.
15319 All modules should be compiled with the same @option{-G @var{num}} value.
15322 @itemx -mno-regnames
15324 @opindex mno-regnames
15325 On System V.4 and embedded PowerPC systems do (do not) emit register
15326 names in the assembly language output using symbolic forms.
15329 @itemx -mno-longcall
15331 @opindex mno-longcall
15332 By default assume that all calls are far away so that a longer more
15333 expensive calling sequence is required. This is required for calls
15334 further than 32 megabytes (33,554,432 bytes) from the current location.
15335 A short call will be generated if the compiler knows
15336 the call cannot be that far away. This setting can be overridden by
15337 the @code{shortcall} function attribute, or by @code{#pragma
15340 Some linkers are capable of detecting out-of-range calls and generating
15341 glue code on the fly. On these systems, long calls are unnecessary and
15342 generate slower code. As of this writing, the AIX linker can do this,
15343 as can the GNU linker for PowerPC/64. It is planned to add this feature
15344 to the GNU linker for 32-bit PowerPC systems as well.
15346 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
15347 callee, L42'', plus a ``branch island'' (glue code). The two target
15348 addresses represent the callee and the ``branch island''. The
15349 Darwin/PPC linker will prefer the first address and generate a ``bl
15350 callee'' if the PPC ``bl'' instruction will reach the callee directly;
15351 otherwise, the linker will generate ``bl L42'' to call the ``branch
15352 island''. The ``branch island'' is appended to the body of the
15353 calling function; it computes the full 32-bit address of the callee
15356 On Mach-O (Darwin) systems, this option directs the compiler emit to
15357 the glue for every direct call, and the Darwin linker decides whether
15358 to use or discard it.
15360 In the future, we may cause GCC to ignore all longcall specifications
15361 when the linker is known to generate glue.
15363 @item -mtls-markers
15364 @itemx -mno-tls-markers
15365 @opindex mtls-markers
15366 @opindex mno-tls-markers
15367 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
15368 specifying the function argument. The relocation allows ld to
15369 reliably associate function call with argument setup instructions for
15370 TLS optimization, which in turn allows gcc to better schedule the
15375 Adds support for multithreading with the @dfn{pthreads} library.
15376 This option sets flags for both the preprocessor and linker.
15381 @subsection RX Options
15384 These @option{-m} options are defined for RX implementations:
15387 @item -m64bit-doubles
15388 @itemx -m32bit-doubles
15389 @opindex m64bit-doubles
15390 @opindex m32bit-doubles
15391 Make the @code{double} data type be 64-bits (@option{-m64bit-doubles})
15392 or 32-bits (@option{-m32bit-doubles}) in size. The default is
15393 @option{-m32bit-doubles}. @emph{Note} the RX's hardware floating
15394 point instructions are only used for 32-bit floating point values, and
15395 then only if @option{-ffast-math} has been specified on the command
15396 line. This is because the RX FPU instructions do not properly support
15397 denormal (or sub-normal) values.
15399 @item -mbig-endian-data
15400 @itemx -mlittle-endian-data
15401 @opindex mbig-endian-data
15402 @opindex mlittle-endian-data
15403 Store data (but not code) in the big-endian format. The default is
15404 @option{-mlittle-endian-data}, ie to store data in the little endian
15407 @item -msmall-data-limit=@var{N}
15408 @opindex msmall-data-limit
15409 Specifies the maximum size in bytes of global and static variables
15410 which can be placed into the small data area. Using the small data
15411 area can lead to smaller and faster code, but the size of area is
15412 limited and it is up to the programmer to ensure that the area does
15413 not overflow. Also when the small data area is used one of the RX's
15414 registers (@code{r13}) is reserved for use pointing to this area, so
15415 it is no longer available for use by the compiler. This could result
15416 in slower and/or larger code if variables which once could have been
15417 held in @code{r13} are now pushed onto the stack.
15419 Note, common variables (variables which have not been initialised) and
15420 constants are not placed into the small data area as they are assigned
15421 to other sections in the output executeable.
15423 The default value is zero, which disables this feature. Note, this
15424 feature is not enabled by default with higher optimization levels
15425 (@option{-O2} etc) because of the potentially deterimental effects of
15426 reserving register @code{r13}. It is up to the programmer to
15427 experiment and discover whether this feature is of benefit to their
15434 Use the simulator runtime. The default is to use the libgloss board
15437 @item -mas100-syntax
15438 @item -mno-as100-syntax
15439 @opindex mas100-syntax
15440 @opindex mno-as100-syntax
15441 When generating assembler output use a syntax that is compatible with
15442 Renesas's AS100 assembler. This syntax can also be handled by the GAS
15443 assembler but it has some restrictions so generating it is not the
15446 @item -mmax-constant-size=@var{N}
15447 @opindex mmax-constant-size
15448 Specifies the maxium size, in bytes, of a constant that can be used as
15449 an operand in a RX instruction. Although the RX instruction set does
15450 allow consants of up to 4 bytes in length to be used in instructions,
15451 a longer value equates to a longer instruction. Thus in some
15452 circumstances it can be beneficial to restrict the size of constants
15453 that are used in instructions. Constants that are too big are instead
15454 placed into a constant pool and referenced via register indirection.
15456 The value @var{N} can be between 0 and 3. A value of 0, the default,
15457 means that constants of any size are allowed.
15461 Enable linker relaxation. Linker relaxation is a process whereby the
15462 linker will attempt to reduce the size of a program by finding shorter
15463 versions of various instructions. Disabled by default.
15465 @item -mint-register=@var{N}
15466 @opindex mint-register
15467 Specify the number of registers to reserve for fast interrupt handler
15468 functions. The value @var{N} can be between 0 and 4. A value of 1
15469 means that register @code{r13} will be reserved for ther exclusive use
15470 of fast interrupt handlers. A value of 2 reserves @code{r13} and
15471 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
15472 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
15473 A value of 0, the default, does not reserve any registers.
15476 @emph{Note:} The generic GCC command line @option{-ffixed-@var{reg}}
15477 has special significance to the RX port when used with the
15478 @code{interrupt} function attribute. This attribute indicates a
15479 function intended to process fast interrupts. GCC will will ensure
15480 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
15481 and/or @code{r13} and only provided that the normal use of the
15482 corresponding registers have been restricted via the
15483 @option{-ffixed-@var{reg}} or @option{-mint-register} command line
15486 @node S/390 and zSeries Options
15487 @subsection S/390 and zSeries Options
15488 @cindex S/390 and zSeries Options
15490 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
15494 @itemx -msoft-float
15495 @opindex mhard-float
15496 @opindex msoft-float
15497 Use (do not use) the hardware floating-point instructions and registers
15498 for floating-point operations. When @option{-msoft-float} is specified,
15499 functions in @file{libgcc.a} will be used to perform floating-point
15500 operations. When @option{-mhard-float} is specified, the compiler
15501 generates IEEE floating-point instructions. This is the default.
15504 @itemx -mno-hard-dfp
15506 @opindex mno-hard-dfp
15507 Use (do not use) the hardware decimal-floating-point instructions for
15508 decimal-floating-point operations. When @option{-mno-hard-dfp} is
15509 specified, functions in @file{libgcc.a} will be used to perform
15510 decimal-floating-point operations. When @option{-mhard-dfp} is
15511 specified, the compiler generates decimal-floating-point hardware
15512 instructions. This is the default for @option{-march=z9-ec} or higher.
15514 @item -mlong-double-64
15515 @itemx -mlong-double-128
15516 @opindex mlong-double-64
15517 @opindex mlong-double-128
15518 These switches control the size of @code{long double} type. A size
15519 of 64bit makes the @code{long double} type equivalent to the @code{double}
15520 type. This is the default.
15523 @itemx -mno-backchain
15524 @opindex mbackchain
15525 @opindex mno-backchain
15526 Store (do not store) the address of the caller's frame as backchain pointer
15527 into the callee's stack frame.
15528 A backchain may be needed to allow debugging using tools that do not understand
15529 DWARF-2 call frame information.
15530 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
15531 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
15532 the backchain is placed into the topmost word of the 96/160 byte register
15535 In general, code compiled with @option{-mbackchain} is call-compatible with
15536 code compiled with @option{-mmo-backchain}; however, use of the backchain
15537 for debugging purposes usually requires that the whole binary is built with
15538 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
15539 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
15540 to build a linux kernel use @option{-msoft-float}.
15542 The default is to not maintain the backchain.
15544 @item -mpacked-stack
15545 @itemx -mno-packed-stack
15546 @opindex mpacked-stack
15547 @opindex mno-packed-stack
15548 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
15549 specified, the compiler uses the all fields of the 96/160 byte register save
15550 area only for their default purpose; unused fields still take up stack space.
15551 When @option{-mpacked-stack} is specified, register save slots are densely
15552 packed at the top of the register save area; unused space is reused for other
15553 purposes, allowing for more efficient use of the available stack space.
15554 However, when @option{-mbackchain} is also in effect, the topmost word of
15555 the save area is always used to store the backchain, and the return address
15556 register is always saved two words below the backchain.
15558 As long as the stack frame backchain is not used, code generated with
15559 @option{-mpacked-stack} is call-compatible with code generated with
15560 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
15561 S/390 or zSeries generated code that uses the stack frame backchain at run
15562 time, not just for debugging purposes. Such code is not call-compatible
15563 with code compiled with @option{-mpacked-stack}. Also, note that the
15564 combination of @option{-mbackchain},
15565 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
15566 to build a linux kernel use @option{-msoft-float}.
15568 The default is to not use the packed stack layout.
15571 @itemx -mno-small-exec
15572 @opindex msmall-exec
15573 @opindex mno-small-exec
15574 Generate (or do not generate) code using the @code{bras} instruction
15575 to do subroutine calls.
15576 This only works reliably if the total executable size does not
15577 exceed 64k. The default is to use the @code{basr} instruction instead,
15578 which does not have this limitation.
15584 When @option{-m31} is specified, generate code compliant to the
15585 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
15586 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
15587 particular to generate 64-bit instructions. For the @samp{s390}
15588 targets, the default is @option{-m31}, while the @samp{s390x}
15589 targets default to @option{-m64}.
15595 When @option{-mzarch} is specified, generate code using the
15596 instructions available on z/Architecture.
15597 When @option{-mesa} is specified, generate code using the
15598 instructions available on ESA/390. Note that @option{-mesa} is
15599 not possible with @option{-m64}.
15600 When generating code compliant to the GNU/Linux for S/390 ABI,
15601 the default is @option{-mesa}. When generating code compliant
15602 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
15608 Generate (or do not generate) code using the @code{mvcle} instruction
15609 to perform block moves. When @option{-mno-mvcle} is specified,
15610 use a @code{mvc} loop instead. This is the default unless optimizing for
15617 Print (or do not print) additional debug information when compiling.
15618 The default is to not print debug information.
15620 @item -march=@var{cpu-type}
15622 Generate code that will run on @var{cpu-type}, which is the name of a system
15623 representing a certain processor type. Possible values for
15624 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
15625 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
15626 When generating code using the instructions available on z/Architecture,
15627 the default is @option{-march=z900}. Otherwise, the default is
15628 @option{-march=g5}.
15630 @item -mtune=@var{cpu-type}
15632 Tune to @var{cpu-type} everything applicable about the generated code,
15633 except for the ABI and the set of available instructions.
15634 The list of @var{cpu-type} values is the same as for @option{-march}.
15635 The default is the value used for @option{-march}.
15638 @itemx -mno-tpf-trace
15639 @opindex mtpf-trace
15640 @opindex mno-tpf-trace
15641 Generate code that adds (does not add) in TPF OS specific branches to trace
15642 routines in the operating system. This option is off by default, even
15643 when compiling for the TPF OS@.
15646 @itemx -mno-fused-madd
15647 @opindex mfused-madd
15648 @opindex mno-fused-madd
15649 Generate code that uses (does not use) the floating point multiply and
15650 accumulate instructions. These instructions are generated by default if
15651 hardware floating point is used.
15653 @item -mwarn-framesize=@var{framesize}
15654 @opindex mwarn-framesize
15655 Emit a warning if the current function exceeds the given frame size. Because
15656 this is a compile time check it doesn't need to be a real problem when the program
15657 runs. It is intended to identify functions which most probably cause
15658 a stack overflow. It is useful to be used in an environment with limited stack
15659 size e.g.@: the linux kernel.
15661 @item -mwarn-dynamicstack
15662 @opindex mwarn-dynamicstack
15663 Emit a warning if the function calls alloca or uses dynamically
15664 sized arrays. This is generally a bad idea with a limited stack size.
15666 @item -mstack-guard=@var{stack-guard}
15667 @itemx -mstack-size=@var{stack-size}
15668 @opindex mstack-guard
15669 @opindex mstack-size
15670 If these options are provided the s390 back end emits additional instructions in
15671 the function prologue which trigger a trap if the stack size is @var{stack-guard}
15672 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
15673 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
15674 the frame size of the compiled function is chosen.
15675 These options are intended to be used to help debugging stack overflow problems.
15676 The additionally emitted code causes only little overhead and hence can also be
15677 used in production like systems without greater performance degradation. The given
15678 values have to be exact powers of 2 and @var{stack-size} has to be greater than
15679 @var{stack-guard} without exceeding 64k.
15680 In order to be efficient the extra code makes the assumption that the stack starts
15681 at an address aligned to the value given by @var{stack-size}.
15682 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
15685 @node Score Options
15686 @subsection Score Options
15687 @cindex Score Options
15689 These options are defined for Score implementations:
15694 Compile code for big endian mode. This is the default.
15698 Compile code for little endian mode.
15702 Disable generate bcnz instruction.
15706 Enable generate unaligned load and store instruction.
15710 Enable the use of multiply-accumulate instructions. Disabled by default.
15714 Specify the SCORE5 as the target architecture.
15718 Specify the SCORE5U of the target architecture.
15722 Specify the SCORE7 as the target architecture. This is the default.
15726 Specify the SCORE7D as the target architecture.
15730 @subsection SH Options
15732 These @samp{-m} options are defined for the SH implementations:
15737 Generate code for the SH1.
15741 Generate code for the SH2.
15744 Generate code for the SH2e.
15748 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
15749 that the floating-point unit is not used.
15751 @item -m2a-single-only
15752 @opindex m2a-single-only
15753 Generate code for the SH2a-FPU, in such a way that no double-precision
15754 floating point operations are used.
15757 @opindex m2a-single
15758 Generate code for the SH2a-FPU assuming the floating-point unit is in
15759 single-precision mode by default.
15763 Generate code for the SH2a-FPU assuming the floating-point unit is in
15764 double-precision mode by default.
15768 Generate code for the SH3.
15772 Generate code for the SH3e.
15776 Generate code for the SH4 without a floating-point unit.
15778 @item -m4-single-only
15779 @opindex m4-single-only
15780 Generate code for the SH4 with a floating-point unit that only
15781 supports single-precision arithmetic.
15785 Generate code for the SH4 assuming the floating-point unit is in
15786 single-precision mode by default.
15790 Generate code for the SH4.
15794 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
15795 floating-point unit is not used.
15797 @item -m4a-single-only
15798 @opindex m4a-single-only
15799 Generate code for the SH4a, in such a way that no double-precision
15800 floating point operations are used.
15803 @opindex m4a-single
15804 Generate code for the SH4a assuming the floating-point unit is in
15805 single-precision mode by default.
15809 Generate code for the SH4a.
15813 Same as @option{-m4a-nofpu}, except that it implicitly passes
15814 @option{-dsp} to the assembler. GCC doesn't generate any DSP
15815 instructions at the moment.
15819 Compile code for the processor in big endian mode.
15823 Compile code for the processor in little endian mode.
15827 Align doubles at 64-bit boundaries. Note that this changes the calling
15828 conventions, and thus some functions from the standard C library will
15829 not work unless you recompile it first with @option{-mdalign}.
15833 Shorten some address references at link time, when possible; uses the
15834 linker option @option{-relax}.
15838 Use 32-bit offsets in @code{switch} tables. The default is to use
15843 Enable the use of bit manipulation instructions on SH2A.
15847 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
15848 alignment constraints.
15852 Comply with the calling conventions defined by Renesas.
15856 Comply with the calling conventions defined by Renesas.
15860 Comply with the calling conventions defined for GCC before the Renesas
15861 conventions were available. This option is the default for all
15862 targets of the SH toolchain except for @samp{sh-symbianelf}.
15865 @opindex mnomacsave
15866 Mark the @code{MAC} register as call-clobbered, even if
15867 @option{-mhitachi} is given.
15871 Increase IEEE-compliance of floating-point code.
15872 At the moment, this is equivalent to @option{-fno-finite-math-only}.
15873 When generating 16 bit SH opcodes, getting IEEE-conforming results for
15874 comparisons of NANs / infinities incurs extra overhead in every
15875 floating point comparison, therefore the default is set to
15876 @option{-ffinite-math-only}.
15878 @item -minline-ic_invalidate
15879 @opindex minline-ic_invalidate
15880 Inline code to invalidate instruction cache entries after setting up
15881 nested function trampolines.
15882 This option has no effect if -musermode is in effect and the selected
15883 code generation option (e.g. -m4) does not allow the use of the icbi
15885 If the selected code generation option does not allow the use of the icbi
15886 instruction, and -musermode is not in effect, the inlined code will
15887 manipulate the instruction cache address array directly with an associative
15888 write. This not only requires privileged mode, but it will also
15889 fail if the cache line had been mapped via the TLB and has become unmapped.
15893 Dump instruction size and location in the assembly code.
15896 @opindex mpadstruct
15897 This option is deprecated. It pads structures to multiple of 4 bytes,
15898 which is incompatible with the SH ABI@.
15902 Optimize for space instead of speed. Implied by @option{-Os}.
15905 @opindex mprefergot
15906 When generating position-independent code, emit function calls using
15907 the Global Offset Table instead of the Procedure Linkage Table.
15911 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
15912 if the inlined code would not work in user mode.
15913 This is the default when the target is @code{sh-*-linux*}.
15915 @item -multcost=@var{number}
15916 @opindex multcost=@var{number}
15917 Set the cost to assume for a multiply insn.
15919 @item -mdiv=@var{strategy}
15920 @opindex mdiv=@var{strategy}
15921 Set the division strategy to use for SHmedia code. @var{strategy} must be
15922 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
15923 inv:call2, inv:fp .
15924 "fp" performs the operation in floating point. This has a very high latency,
15925 but needs only a few instructions, so it might be a good choice if
15926 your code has enough easily exploitable ILP to allow the compiler to
15927 schedule the floating point instructions together with other instructions.
15928 Division by zero causes a floating point exception.
15929 "inv" uses integer operations to calculate the inverse of the divisor,
15930 and then multiplies the dividend with the inverse. This strategy allows
15931 cse and hoisting of the inverse calculation. Division by zero calculates
15932 an unspecified result, but does not trap.
15933 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
15934 have been found, or if the entire operation has been hoisted to the same
15935 place, the last stages of the inverse calculation are intertwined with the
15936 final multiply to reduce the overall latency, at the expense of using a few
15937 more instructions, and thus offering fewer scheduling opportunities with
15939 "call" calls a library function that usually implements the inv:minlat
15941 This gives high code density for m5-*media-nofpu compilations.
15942 "call2" uses a different entry point of the same library function, where it
15943 assumes that a pointer to a lookup table has already been set up, which
15944 exposes the pointer load to cse / code hoisting optimizations.
15945 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
15946 code generation, but if the code stays unoptimized, revert to the "call",
15947 "call2", or "fp" strategies, respectively. Note that the
15948 potentially-trapping side effect of division by zero is carried by a
15949 separate instruction, so it is possible that all the integer instructions
15950 are hoisted out, but the marker for the side effect stays where it is.
15951 A recombination to fp operations or a call is not possible in that case.
15952 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
15953 that the inverse calculation was nor separated from the multiply, they speed
15954 up division where the dividend fits into 20 bits (plus sign where applicable),
15955 by inserting a test to skip a number of operations in this case; this test
15956 slows down the case of larger dividends. inv20u assumes the case of a such
15957 a small dividend to be unlikely, and inv20l assumes it to be likely.
15959 @item -mdivsi3_libfunc=@var{name}
15960 @opindex mdivsi3_libfunc=@var{name}
15961 Set the name of the library function used for 32 bit signed division to
15962 @var{name}. This only affect the name used in the call and inv:call
15963 division strategies, and the compiler will still expect the same
15964 sets of input/output/clobbered registers as if this option was not present.
15966 @item -mfixed-range=@var{register-range}
15967 @opindex mfixed-range
15968 Generate code treating the given register range as fixed registers.
15969 A fixed register is one that the register allocator can not use. This is
15970 useful when compiling kernel code. A register range is specified as
15971 two registers separated by a dash. Multiple register ranges can be
15972 specified separated by a comma.
15974 @item -madjust-unroll
15975 @opindex madjust-unroll
15976 Throttle unrolling to avoid thrashing target registers.
15977 This option only has an effect if the gcc code base supports the
15978 TARGET_ADJUST_UNROLL_MAX target hook.
15980 @item -mindexed-addressing
15981 @opindex mindexed-addressing
15982 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
15983 This is only safe if the hardware and/or OS implement 32 bit wrap-around
15984 semantics for the indexed addressing mode. The architecture allows the
15985 implementation of processors with 64 bit MMU, which the OS could use to
15986 get 32 bit addressing, but since no current hardware implementation supports
15987 this or any other way to make the indexed addressing mode safe to use in
15988 the 32 bit ABI, the default is -mno-indexed-addressing.
15990 @item -mgettrcost=@var{number}
15991 @opindex mgettrcost=@var{number}
15992 Set the cost assumed for the gettr instruction to @var{number}.
15993 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
15997 Assume pt* instructions won't trap. This will generally generate better
15998 scheduled code, but is unsafe on current hardware. The current architecture
15999 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
16000 This has the unintentional effect of making it unsafe to schedule ptabs /
16001 ptrel before a branch, or hoist it out of a loop. For example,
16002 __do_global_ctors, a part of libgcc that runs constructors at program
16003 startup, calls functions in a list which is delimited by @minus{}1. With the
16004 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
16005 That means that all the constructors will be run a bit quicker, but when
16006 the loop comes to the end of the list, the program crashes because ptabs
16007 loads @minus{}1 into a target register. Since this option is unsafe for any
16008 hardware implementing the current architecture specification, the default
16009 is -mno-pt-fixed. Unless the user specifies a specific cost with
16010 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
16011 this deters register allocation using target registers for storing
16014 @item -minvalid-symbols
16015 @opindex minvalid-symbols
16016 Assume symbols might be invalid. Ordinary function symbols generated by
16017 the compiler will always be valid to load with movi/shori/ptabs or
16018 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
16019 to generate symbols that will cause ptabs / ptrel to trap.
16020 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
16021 It will then prevent cross-basic-block cse, hoisting and most scheduling
16022 of symbol loads. The default is @option{-mno-invalid-symbols}.
16025 @node SPARC Options
16026 @subsection SPARC Options
16027 @cindex SPARC options
16029 These @samp{-m} options are supported on the SPARC:
16032 @item -mno-app-regs
16034 @opindex mno-app-regs
16036 Specify @option{-mapp-regs} to generate output using the global registers
16037 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
16040 To be fully SVR4 ABI compliant at the cost of some performance loss,
16041 specify @option{-mno-app-regs}. You should compile libraries and system
16042 software with this option.
16045 @itemx -mhard-float
16047 @opindex mhard-float
16048 Generate output containing floating point instructions. This is the
16052 @itemx -msoft-float
16054 @opindex msoft-float
16055 Generate output containing library calls for floating point.
16056 @strong{Warning:} the requisite libraries are not available for all SPARC
16057 targets. Normally the facilities of the machine's usual C compiler are
16058 used, but this cannot be done directly in cross-compilation. You must make
16059 your own arrangements to provide suitable library functions for
16060 cross-compilation. The embedded targets @samp{sparc-*-aout} and
16061 @samp{sparclite-*-*} do provide software floating point support.
16063 @option{-msoft-float} changes the calling convention in the output file;
16064 therefore, it is only useful if you compile @emph{all} of a program with
16065 this option. In particular, you need to compile @file{libgcc.a}, the
16066 library that comes with GCC, with @option{-msoft-float} in order for
16069 @item -mhard-quad-float
16070 @opindex mhard-quad-float
16071 Generate output containing quad-word (long double) floating point
16074 @item -msoft-quad-float
16075 @opindex msoft-quad-float
16076 Generate output containing library calls for quad-word (long double)
16077 floating point instructions. The functions called are those specified
16078 in the SPARC ABI@. This is the default.
16080 As of this writing, there are no SPARC implementations that have hardware
16081 support for the quad-word floating point instructions. They all invoke
16082 a trap handler for one of these instructions, and then the trap handler
16083 emulates the effect of the instruction. Because of the trap handler overhead,
16084 this is much slower than calling the ABI library routines. Thus the
16085 @option{-msoft-quad-float} option is the default.
16087 @item -mno-unaligned-doubles
16088 @itemx -munaligned-doubles
16089 @opindex mno-unaligned-doubles
16090 @opindex munaligned-doubles
16091 Assume that doubles have 8 byte alignment. This is the default.
16093 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
16094 alignment only if they are contained in another type, or if they have an
16095 absolute address. Otherwise, it assumes they have 4 byte alignment.
16096 Specifying this option avoids some rare compatibility problems with code
16097 generated by other compilers. It is not the default because it results
16098 in a performance loss, especially for floating point code.
16100 @item -mno-faster-structs
16101 @itemx -mfaster-structs
16102 @opindex mno-faster-structs
16103 @opindex mfaster-structs
16104 With @option{-mfaster-structs}, the compiler assumes that structures
16105 should have 8 byte alignment. This enables the use of pairs of
16106 @code{ldd} and @code{std} instructions for copies in structure
16107 assignment, in place of twice as many @code{ld} and @code{st} pairs.
16108 However, the use of this changed alignment directly violates the SPARC
16109 ABI@. Thus, it's intended only for use on targets where the developer
16110 acknowledges that their resulting code will not be directly in line with
16111 the rules of the ABI@.
16113 @item -mimpure-text
16114 @opindex mimpure-text
16115 @option{-mimpure-text}, used in addition to @option{-shared}, tells
16116 the compiler to not pass @option{-z text} to the linker when linking a
16117 shared object. Using this option, you can link position-dependent
16118 code into a shared object.
16120 @option{-mimpure-text} suppresses the ``relocations remain against
16121 allocatable but non-writable sections'' linker error message.
16122 However, the necessary relocations will trigger copy-on-write, and the
16123 shared object is not actually shared across processes. Instead of
16124 using @option{-mimpure-text}, you should compile all source code with
16125 @option{-fpic} or @option{-fPIC}.
16127 This option is only available on SunOS and Solaris.
16129 @item -mcpu=@var{cpu_type}
16131 Set the instruction set, register set, and instruction scheduling parameters
16132 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
16133 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
16134 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
16135 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
16136 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
16138 Default instruction scheduling parameters are used for values that select
16139 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
16140 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
16142 Here is a list of each supported architecture and their supported
16147 v8: supersparc, hypersparc
16148 sparclite: f930, f934, sparclite86x
16150 v9: ultrasparc, ultrasparc3, niagara, niagara2
16153 By default (unless configured otherwise), GCC generates code for the V7
16154 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
16155 additionally optimizes it for the Cypress CY7C602 chip, as used in the
16156 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
16157 SPARCStation 1, 2, IPX etc.
16159 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
16160 architecture. The only difference from V7 code is that the compiler emits
16161 the integer multiply and integer divide instructions which exist in SPARC-V8
16162 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
16163 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
16166 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
16167 the SPARC architecture. This adds the integer multiply, integer divide step
16168 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
16169 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
16170 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
16171 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
16172 MB86934 chip, which is the more recent SPARClite with FPU@.
16174 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
16175 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
16176 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
16177 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
16178 optimizes it for the TEMIC SPARClet chip.
16180 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
16181 architecture. This adds 64-bit integer and floating-point move instructions,
16182 3 additional floating-point condition code registers and conditional move
16183 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
16184 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
16185 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
16186 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
16187 @option{-mcpu=niagara}, the compiler additionally optimizes it for
16188 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
16189 additionally optimizes it for Sun UltraSPARC T2 chips.
16191 @item -mtune=@var{cpu_type}
16193 Set the instruction scheduling parameters for machine type
16194 @var{cpu_type}, but do not set the instruction set or register set that the
16195 option @option{-mcpu=@var{cpu_type}} would.
16197 The same values for @option{-mcpu=@var{cpu_type}} can be used for
16198 @option{-mtune=@var{cpu_type}}, but the only useful values are those
16199 that select a particular cpu implementation. Those are @samp{cypress},
16200 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
16201 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
16202 @samp{ultrasparc3}, @samp{niagara}, and @samp{niagara2}.
16207 @opindex mno-v8plus
16208 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
16209 difference from the V8 ABI is that the global and out registers are
16210 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
16211 mode for all SPARC-V9 processors.
16217 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
16218 Visual Instruction Set extensions. The default is @option{-mno-vis}.
16221 These @samp{-m} options are supported in addition to the above
16222 on SPARC-V9 processors in 64-bit environments:
16225 @item -mlittle-endian
16226 @opindex mlittle-endian
16227 Generate code for a processor running in little-endian mode. It is only
16228 available for a few configurations and most notably not on Solaris and Linux.
16234 Generate code for a 32-bit or 64-bit environment.
16235 The 32-bit environment sets int, long and pointer to 32 bits.
16236 The 64-bit environment sets int to 32 bits and long and pointer
16239 @item -mcmodel=medlow
16240 @opindex mcmodel=medlow
16241 Generate code for the Medium/Low code model: 64-bit addresses, programs
16242 must be linked in the low 32 bits of memory. Programs can be statically
16243 or dynamically linked.
16245 @item -mcmodel=medmid
16246 @opindex mcmodel=medmid
16247 Generate code for the Medium/Middle code model: 64-bit addresses, programs
16248 must be linked in the low 44 bits of memory, the text and data segments must
16249 be less than 2GB in size and the data segment must be located within 2GB of
16252 @item -mcmodel=medany
16253 @opindex mcmodel=medany
16254 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
16255 may be linked anywhere in memory, the text and data segments must be less
16256 than 2GB in size and the data segment must be located within 2GB of the
16259 @item -mcmodel=embmedany
16260 @opindex mcmodel=embmedany
16261 Generate code for the Medium/Anywhere code model for embedded systems:
16262 64-bit addresses, the text and data segments must be less than 2GB in
16263 size, both starting anywhere in memory (determined at link time). The
16264 global register %g4 points to the base of the data segment. Programs
16265 are statically linked and PIC is not supported.
16268 @itemx -mno-stack-bias
16269 @opindex mstack-bias
16270 @opindex mno-stack-bias
16271 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
16272 frame pointer if present, are offset by @minus{}2047 which must be added back
16273 when making stack frame references. This is the default in 64-bit mode.
16274 Otherwise, assume no such offset is present.
16277 These switches are supported in addition to the above on Solaris:
16282 Add support for multithreading using the Solaris threads library. This
16283 option sets flags for both the preprocessor and linker. This option does
16284 not affect the thread safety of object code produced by the compiler or
16285 that of libraries supplied with it.
16289 Add support for multithreading using the POSIX threads library. This
16290 option sets flags for both the preprocessor and linker. This option does
16291 not affect the thread safety of object code produced by the compiler or
16292 that of libraries supplied with it.
16296 This is a synonym for @option{-pthreads}.
16300 @subsection SPU Options
16301 @cindex SPU options
16303 These @samp{-m} options are supported on the SPU:
16307 @itemx -merror-reloc
16308 @opindex mwarn-reloc
16309 @opindex merror-reloc
16311 The loader for SPU does not handle dynamic relocations. By default, GCC
16312 will give an error when it generates code that requires a dynamic
16313 relocation. @option{-mno-error-reloc} disables the error,
16314 @option{-mwarn-reloc} will generate a warning instead.
16317 @itemx -munsafe-dma
16319 @opindex munsafe-dma
16321 Instructions which initiate or test completion of DMA must not be
16322 reordered with respect to loads and stores of the memory which is being
16323 accessed. Users typically address this problem using the volatile
16324 keyword, but that can lead to inefficient code in places where the
16325 memory is known to not change. Rather than mark the memory as volatile
16326 we treat the DMA instructions as potentially effecting all memory. With
16327 @option{-munsafe-dma} users must use the volatile keyword to protect
16330 @item -mbranch-hints
16331 @opindex mbranch-hints
16333 By default, GCC will generate a branch hint instruction to avoid
16334 pipeline stalls for always taken or probably taken branches. A hint
16335 will not be generated closer than 8 instructions away from its branch.
16336 There is little reason to disable them, except for debugging purposes,
16337 or to make an object a little bit smaller.
16341 @opindex msmall-mem
16342 @opindex mlarge-mem
16344 By default, GCC generates code assuming that addresses are never larger
16345 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
16346 a full 32 bit address.
16351 By default, GCC links against startup code that assumes the SPU-style
16352 main function interface (which has an unconventional parameter list).
16353 With @option{-mstdmain}, GCC will link your program against startup
16354 code that assumes a C99-style interface to @code{main}, including a
16355 local copy of @code{argv} strings.
16357 @item -mfixed-range=@var{register-range}
16358 @opindex mfixed-range
16359 Generate code treating the given register range as fixed registers.
16360 A fixed register is one that the register allocator can not use. This is
16361 useful when compiling kernel code. A register range is specified as
16362 two registers separated by a dash. Multiple register ranges can be
16363 specified separated by a comma.
16369 Compile code assuming that pointers to the PPU address space accessed
16370 via the @code{__ea} named address space qualifier are either 32 or 64
16371 bits wide. The default is 32 bits. As this is an ABI changing option,
16372 all object code in an executable must be compiled with the same setting.
16374 @item -maddress-space-conversion
16375 @itemx -mno-address-space-conversion
16376 @opindex maddress-space-conversion
16377 @opindex mno-address-space-conversion
16378 Allow/disallow treating the @code{__ea} address space as superset
16379 of the generic address space. This enables explicit type casts
16380 between @code{__ea} and generic pointer as well as implicit
16381 conversions of generic pointers to @code{__ea} pointers. The
16382 default is to allow address space pointer conversions.
16384 @item -mcache-size=@var{cache-size}
16385 @opindex mcache-size
16386 This option controls the version of libgcc that the compiler links to an
16387 executable and selects a software-managed cache for accessing variables
16388 in the @code{__ea} address space with a particular cache size. Possible
16389 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
16390 and @samp{128}. The default cache size is 64KB.
16392 @item -matomic-updates
16393 @itemx -mno-atomic-updates
16394 @opindex matomic-updates
16395 @opindex mno-atomic-updates
16396 This option controls the version of libgcc that the compiler links to an
16397 executable and selects whether atomic updates to the software-managed
16398 cache of PPU-side variables are used. If you use atomic updates, changes
16399 to a PPU variable from SPU code using the @code{__ea} named address space
16400 qualifier will not interfere with changes to other PPU variables residing
16401 in the same cache line from PPU code. If you do not use atomic updates,
16402 such interference may occur; however, writing back cache lines will be
16403 more efficient. The default behavior is to use atomic updates.
16406 @itemx -mdual-nops=@var{n}
16407 @opindex mdual-nops
16408 By default, GCC will insert nops to increase dual issue when it expects
16409 it to increase performance. @var{n} can be a value from 0 to 10. A
16410 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
16411 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
16413 @item -mhint-max-nops=@var{n}
16414 @opindex mhint-max-nops
16415 Maximum number of nops to insert for a branch hint. A branch hint must
16416 be at least 8 instructions away from the branch it is effecting. GCC
16417 will insert up to @var{n} nops to enforce this, otherwise it will not
16418 generate the branch hint.
16420 @item -mhint-max-distance=@var{n}
16421 @opindex mhint-max-distance
16422 The encoding of the branch hint instruction limits the hint to be within
16423 256 instructions of the branch it is effecting. By default, GCC makes
16424 sure it is within 125.
16427 @opindex msafe-hints
16428 Work around a hardware bug which causes the SPU to stall indefinitely.
16429 By default, GCC will insert the @code{hbrp} instruction to make sure
16430 this stall won't happen.
16434 @node System V Options
16435 @subsection Options for System V
16437 These additional options are available on System V Release 4 for
16438 compatibility with other compilers on those systems:
16443 Create a shared object.
16444 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
16448 Identify the versions of each tool used by the compiler, in a
16449 @code{.ident} assembler directive in the output.
16453 Refrain from adding @code{.ident} directives to the output file (this is
16456 @item -YP,@var{dirs}
16458 Search the directories @var{dirs}, and no others, for libraries
16459 specified with @option{-l}.
16461 @item -Ym,@var{dir}
16463 Look in the directory @var{dir} to find the M4 preprocessor.
16464 The assembler uses this option.
16465 @c This is supposed to go with a -Yd for predefined M4 macro files, but
16466 @c the generic assembler that comes with Solaris takes just -Ym.
16470 @subsection V850 Options
16471 @cindex V850 Options
16473 These @samp{-m} options are defined for V850 implementations:
16477 @itemx -mno-long-calls
16478 @opindex mlong-calls
16479 @opindex mno-long-calls
16480 Treat all calls as being far away (near). If calls are assumed to be
16481 far away, the compiler will always load the functions address up into a
16482 register, and call indirect through the pointer.
16488 Do not optimize (do optimize) basic blocks that use the same index
16489 pointer 4 or more times to copy pointer into the @code{ep} register, and
16490 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
16491 option is on by default if you optimize.
16493 @item -mno-prolog-function
16494 @itemx -mprolog-function
16495 @opindex mno-prolog-function
16496 @opindex mprolog-function
16497 Do not use (do use) external functions to save and restore registers
16498 at the prologue and epilogue of a function. The external functions
16499 are slower, but use less code space if more than one function saves
16500 the same number of registers. The @option{-mprolog-function} option
16501 is on by default if you optimize.
16505 Try to make the code as small as possible. At present, this just turns
16506 on the @option{-mep} and @option{-mprolog-function} options.
16508 @item -mtda=@var{n}
16510 Put static or global variables whose size is @var{n} bytes or less into
16511 the tiny data area that register @code{ep} points to. The tiny data
16512 area can hold up to 256 bytes in total (128 bytes for byte references).
16514 @item -msda=@var{n}
16516 Put static or global variables whose size is @var{n} bytes or less into
16517 the small data area that register @code{gp} points to. The small data
16518 area can hold up to 64 kilobytes.
16520 @item -mzda=@var{n}
16522 Put static or global variables whose size is @var{n} bytes or less into
16523 the first 32 kilobytes of memory.
16527 Specify that the target processor is the V850.
16530 @opindex mbig-switch
16531 Generate code suitable for big switch tables. Use this option only if
16532 the assembler/linker complain about out of range branches within a switch
16537 This option will cause r2 and r5 to be used in the code generated by
16538 the compiler. This setting is the default.
16540 @item -mno-app-regs
16541 @opindex mno-app-regs
16542 This option will cause r2 and r5 to be treated as fixed registers.
16546 Specify that the target processor is the V850E1. The preprocessor
16547 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
16548 this option is used.
16552 Specify that the target processor is the V850E@. The preprocessor
16553 constant @samp{__v850e__} will be defined if this option is used.
16555 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
16556 are defined then a default target processor will be chosen and the
16557 relevant @samp{__v850*__} preprocessor constant will be defined.
16559 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
16560 defined, regardless of which processor variant is the target.
16562 @item -mdisable-callt
16563 @opindex mdisable-callt
16564 This option will suppress generation of the CALLT instruction for the
16565 v850e and v850e1 flavors of the v850 architecture. The default is
16566 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
16571 @subsection VAX Options
16572 @cindex VAX options
16574 These @samp{-m} options are defined for the VAX:
16579 Do not output certain jump instructions (@code{aobleq} and so on)
16580 that the Unix assembler for the VAX cannot handle across long
16585 Do output those jump instructions, on the assumption that you
16586 will assemble with the GNU assembler.
16590 Output code for g-format floating point numbers instead of d-format.
16593 @node VxWorks Options
16594 @subsection VxWorks Options
16595 @cindex VxWorks Options
16597 The options in this section are defined for all VxWorks targets.
16598 Options specific to the target hardware are listed with the other
16599 options for that target.
16604 GCC can generate code for both VxWorks kernels and real time processes
16605 (RTPs). This option switches from the former to the latter. It also
16606 defines the preprocessor macro @code{__RTP__}.
16609 @opindex non-static
16610 Link an RTP executable against shared libraries rather than static
16611 libraries. The options @option{-static} and @option{-shared} can
16612 also be used for RTPs (@pxref{Link Options}); @option{-static}
16619 These options are passed down to the linker. They are defined for
16620 compatibility with Diab.
16623 @opindex Xbind-lazy
16624 Enable lazy binding of function calls. This option is equivalent to
16625 @option{-Wl,-z,now} and is defined for compatibility with Diab.
16629 Disable lazy binding of function calls. This option is the default and
16630 is defined for compatibility with Diab.
16633 @node x86-64 Options
16634 @subsection x86-64 Options
16635 @cindex x86-64 options
16637 These are listed under @xref{i386 and x86-64 Options}.
16639 @node i386 and x86-64 Windows Options
16640 @subsection i386 and x86-64 Windows Options
16641 @cindex i386 and x86-64 Windows Options
16643 These additional options are available for Windows targets:
16648 This option is available for Cygwin and MinGW targets. It
16649 specifies that a console application is to be generated, by
16650 instructing the linker to set the PE header subsystem type
16651 required for console applications.
16652 This is the default behavior for Cygwin and MinGW targets.
16656 This option is available for Cygwin targets. It specifies that
16657 the Cygwin internal interface is to be used for predefined
16658 preprocessor macros, C runtime libraries and related linker
16659 paths and options. For Cygwin targets this is the default behavior.
16660 This option is deprecated and will be removed in a future release.
16663 @opindex mno-cygwin
16664 This option is available for Cygwin targets. It specifies that
16665 the MinGW internal interface is to be used instead of Cygwin's, by
16666 setting MinGW-related predefined macros and linker paths and default
16668 This option is deprecated and will be removed in a future release.
16672 This option is available for Cygwin and MinGW targets. It
16673 specifies that a DLL - a dynamic link library - is to be
16674 generated, enabling the selection of the required runtime
16675 startup object and entry point.
16677 @item -mnop-fun-dllimport
16678 @opindex mnop-fun-dllimport
16679 This option is available for Cygwin and MinGW targets. It
16680 specifies that the dllimport attribute should be ignored.
16684 This option is available for MinGW targets. It specifies
16685 that MinGW-specific thread support is to be used.
16689 This option is available for mingw-w64 targets. It specifies
16690 that the UNICODE macro is getting pre-defined and that the
16691 unicode capable runtime startup code is choosen.
16695 This option is available for Cygwin and MinGW targets. It
16696 specifies that the typical Windows pre-defined macros are to
16697 be set in the pre-processor, but does not influence the choice
16698 of runtime library/startup code.
16702 This option is available for Cygwin and MinGW targets. It
16703 specifies that a GUI application is to be generated by
16704 instructing the linker to set the PE header subsystem type
16707 @item -mpe-aligned-commons
16708 @opindex mpe-aligned-commons
16709 This option is available for Cygwin and MinGW targets. It
16710 specifies that the GNU extension to the PE file format that
16711 permits the correct alignment of COMMON variables should be
16712 used when generating code. It will be enabled by default if
16713 GCC detects that the target assembler found during configuration
16714 supports the feature.
16717 See also under @ref{i386 and x86-64 Options} for standard options.
16719 @node Xstormy16 Options
16720 @subsection Xstormy16 Options
16721 @cindex Xstormy16 Options
16723 These options are defined for Xstormy16:
16728 Choose startup files and linker script suitable for the simulator.
16731 @node Xtensa Options
16732 @subsection Xtensa Options
16733 @cindex Xtensa Options
16735 These options are supported for Xtensa targets:
16739 @itemx -mno-const16
16741 @opindex mno-const16
16742 Enable or disable use of @code{CONST16} instructions for loading
16743 constant values. The @code{CONST16} instruction is currently not a
16744 standard option from Tensilica. When enabled, @code{CONST16}
16745 instructions are always used in place of the standard @code{L32R}
16746 instructions. The use of @code{CONST16} is enabled by default only if
16747 the @code{L32R} instruction is not available.
16750 @itemx -mno-fused-madd
16751 @opindex mfused-madd
16752 @opindex mno-fused-madd
16753 Enable or disable use of fused multiply/add and multiply/subtract
16754 instructions in the floating-point option. This has no effect if the
16755 floating-point option is not also enabled. Disabling fused multiply/add
16756 and multiply/subtract instructions forces the compiler to use separate
16757 instructions for the multiply and add/subtract operations. This may be
16758 desirable in some cases where strict IEEE 754-compliant results are
16759 required: the fused multiply add/subtract instructions do not round the
16760 intermediate result, thereby producing results with @emph{more} bits of
16761 precision than specified by the IEEE standard. Disabling fused multiply
16762 add/subtract instructions also ensures that the program output is not
16763 sensitive to the compiler's ability to combine multiply and add/subtract
16766 @item -mserialize-volatile
16767 @itemx -mno-serialize-volatile
16768 @opindex mserialize-volatile
16769 @opindex mno-serialize-volatile
16770 When this option is enabled, GCC inserts @code{MEMW} instructions before
16771 @code{volatile} memory references to guarantee sequential consistency.
16772 The default is @option{-mserialize-volatile}. Use
16773 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
16775 @item -mtext-section-literals
16776 @itemx -mno-text-section-literals
16777 @opindex mtext-section-literals
16778 @opindex mno-text-section-literals
16779 Control the treatment of literal pools. The default is
16780 @option{-mno-text-section-literals}, which places literals in a separate
16781 section in the output file. This allows the literal pool to be placed
16782 in a data RAM/ROM, and it also allows the linker to combine literal
16783 pools from separate object files to remove redundant literals and
16784 improve code size. With @option{-mtext-section-literals}, the literals
16785 are interspersed in the text section in order to keep them as close as
16786 possible to their references. This may be necessary for large assembly
16789 @item -mtarget-align
16790 @itemx -mno-target-align
16791 @opindex mtarget-align
16792 @opindex mno-target-align
16793 When this option is enabled, GCC instructs the assembler to
16794 automatically align instructions to reduce branch penalties at the
16795 expense of some code density. The assembler attempts to widen density
16796 instructions to align branch targets and the instructions following call
16797 instructions. If there are not enough preceding safe density
16798 instructions to align a target, no widening will be performed. The
16799 default is @option{-mtarget-align}. These options do not affect the
16800 treatment of auto-aligned instructions like @code{LOOP}, which the
16801 assembler will always align, either by widening density instructions or
16802 by inserting no-op instructions.
16805 @itemx -mno-longcalls
16806 @opindex mlongcalls
16807 @opindex mno-longcalls
16808 When this option is enabled, GCC instructs the assembler to translate
16809 direct calls to indirect calls unless it can determine that the target
16810 of a direct call is in the range allowed by the call instruction. This
16811 translation typically occurs for calls to functions in other source
16812 files. Specifically, the assembler translates a direct @code{CALL}
16813 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
16814 The default is @option{-mno-longcalls}. This option should be used in
16815 programs where the call target can potentially be out of range. This
16816 option is implemented in the assembler, not the compiler, so the
16817 assembly code generated by GCC will still show direct call
16818 instructions---look at the disassembled object code to see the actual
16819 instructions. Note that the assembler will use an indirect call for
16820 every cross-file call, not just those that really will be out of range.
16823 @node zSeries Options
16824 @subsection zSeries Options
16825 @cindex zSeries options
16827 These are listed under @xref{S/390 and zSeries Options}.
16829 @node Code Gen Options
16830 @section Options for Code Generation Conventions
16831 @cindex code generation conventions
16832 @cindex options, code generation
16833 @cindex run-time options
16835 These machine-independent options control the interface conventions
16836 used in code generation.
16838 Most of them have both positive and negative forms; the negative form
16839 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
16840 one of the forms is listed---the one which is not the default. You
16841 can figure out the other form by either removing @samp{no-} or adding
16845 @item -fbounds-check
16846 @opindex fbounds-check
16847 For front-ends that support it, generate additional code to check that
16848 indices used to access arrays are within the declared range. This is
16849 currently only supported by the Java and Fortran front-ends, where
16850 this option defaults to true and false respectively.
16854 This option generates traps for signed overflow on addition, subtraction,
16855 multiplication operations.
16859 This option instructs the compiler to assume that signed arithmetic
16860 overflow of addition, subtraction and multiplication wraps around
16861 using twos-complement representation. This flag enables some optimizations
16862 and disables others. This option is enabled by default for the Java
16863 front-end, as required by the Java language specification.
16866 @opindex fexceptions
16867 Enable exception handling. Generates extra code needed to propagate
16868 exceptions. For some targets, this implies GCC will generate frame
16869 unwind information for all functions, which can produce significant data
16870 size overhead, although it does not affect execution. If you do not
16871 specify this option, GCC will enable it by default for languages like
16872 C++ which normally require exception handling, and disable it for
16873 languages like C that do not normally require it. However, you may need
16874 to enable this option when compiling C code that needs to interoperate
16875 properly with exception handlers written in C++. You may also wish to
16876 disable this option if you are compiling older C++ programs that don't
16877 use exception handling.
16879 @item -fnon-call-exceptions
16880 @opindex fnon-call-exceptions
16881 Generate code that allows trapping instructions to throw exceptions.
16882 Note that this requires platform-specific runtime support that does
16883 not exist everywhere. Moreover, it only allows @emph{trapping}
16884 instructions to throw exceptions, i.e.@: memory references or floating
16885 point instructions. It does not allow exceptions to be thrown from
16886 arbitrary signal handlers such as @code{SIGALRM}.
16888 @item -funwind-tables
16889 @opindex funwind-tables
16890 Similar to @option{-fexceptions}, except that it will just generate any needed
16891 static data, but will not affect the generated code in any other way.
16892 You will normally not enable this option; instead, a language processor
16893 that needs this handling would enable it on your behalf.
16895 @item -fasynchronous-unwind-tables
16896 @opindex fasynchronous-unwind-tables
16897 Generate unwind table in dwarf2 format, if supported by target machine. The
16898 table is exact at each instruction boundary, so it can be used for stack
16899 unwinding from asynchronous events (such as debugger or garbage collector).
16901 @item -fpcc-struct-return
16902 @opindex fpcc-struct-return
16903 Return ``short'' @code{struct} and @code{union} values in memory like
16904 longer ones, rather than in registers. This convention is less
16905 efficient, but it has the advantage of allowing intercallability between
16906 GCC-compiled files and files compiled with other compilers, particularly
16907 the Portable C Compiler (pcc).
16909 The precise convention for returning structures in memory depends
16910 on the target configuration macros.
16912 Short structures and unions are those whose size and alignment match
16913 that of some integer type.
16915 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
16916 switch is not binary compatible with code compiled with the
16917 @option{-freg-struct-return} switch.
16918 Use it to conform to a non-default application binary interface.
16920 @item -freg-struct-return
16921 @opindex freg-struct-return
16922 Return @code{struct} and @code{union} values in registers when possible.
16923 This is more efficient for small structures than
16924 @option{-fpcc-struct-return}.
16926 If you specify neither @option{-fpcc-struct-return} nor
16927 @option{-freg-struct-return}, GCC defaults to whichever convention is
16928 standard for the target. If there is no standard convention, GCC
16929 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
16930 the principal compiler. In those cases, we can choose the standard, and
16931 we chose the more efficient register return alternative.
16933 @strong{Warning:} code compiled with the @option{-freg-struct-return}
16934 switch is not binary compatible with code compiled with the
16935 @option{-fpcc-struct-return} switch.
16936 Use it to conform to a non-default application binary interface.
16938 @item -fshort-enums
16939 @opindex fshort-enums
16940 Allocate to an @code{enum} type only as many bytes as it needs for the
16941 declared range of possible values. Specifically, the @code{enum} type
16942 will be equivalent to the smallest integer type which has enough room.
16944 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
16945 code that is not binary compatible with code generated without that switch.
16946 Use it to conform to a non-default application binary interface.
16948 @item -fshort-double
16949 @opindex fshort-double
16950 Use the same size for @code{double} as for @code{float}.
16952 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
16953 code that is not binary compatible with code generated without that switch.
16954 Use it to conform to a non-default application binary interface.
16956 @item -fshort-wchar
16957 @opindex fshort-wchar
16958 Override the underlying type for @samp{wchar_t} to be @samp{short
16959 unsigned int} instead of the default for the target. This option is
16960 useful for building programs to run under WINE@.
16962 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
16963 code that is not binary compatible with code generated without that switch.
16964 Use it to conform to a non-default application binary interface.
16967 @opindex fno-common
16968 In C code, controls the placement of uninitialized global variables.
16969 Unix C compilers have traditionally permitted multiple definitions of
16970 such variables in different compilation units by placing the variables
16972 This is the behavior specified by @option{-fcommon}, and is the default
16973 for GCC on most targets.
16974 On the other hand, this behavior is not required by ISO C, and on some
16975 targets may carry a speed or code size penalty on variable references.
16976 The @option{-fno-common} option specifies that the compiler should place
16977 uninitialized global variables in the data section of the object file,
16978 rather than generating them as common blocks.
16979 This has the effect that if the same variable is declared
16980 (without @code{extern}) in two different compilations,
16981 you will get a multiple-definition error when you link them.
16982 In this case, you must compile with @option{-fcommon} instead.
16983 Compiling with @option{-fno-common} is useful on targets for which
16984 it provides better performance, or if you wish to verify that the
16985 program will work on other systems which always treat uninitialized
16986 variable declarations this way.
16990 Ignore the @samp{#ident} directive.
16992 @item -finhibit-size-directive
16993 @opindex finhibit-size-directive
16994 Don't output a @code{.size} assembler directive, or anything else that
16995 would cause trouble if the function is split in the middle, and the
16996 two halves are placed at locations far apart in memory. This option is
16997 used when compiling @file{crtstuff.c}; you should not need to use it
17000 @item -fverbose-asm
17001 @opindex fverbose-asm
17002 Put extra commentary information in the generated assembly code to
17003 make it more readable. This option is generally only of use to those
17004 who actually need to read the generated assembly code (perhaps while
17005 debugging the compiler itself).
17007 @option{-fno-verbose-asm}, the default, causes the
17008 extra information to be omitted and is useful when comparing two assembler
17011 @item -frecord-gcc-switches
17012 @opindex frecord-gcc-switches
17013 This switch causes the command line that was used to invoke the
17014 compiler to be recorded into the object file that is being created.
17015 This switch is only implemented on some targets and the exact format
17016 of the recording is target and binary file format dependent, but it
17017 usually takes the form of a section containing ASCII text. This
17018 switch is related to the @option{-fverbose-asm} switch, but that
17019 switch only records information in the assembler output file as
17020 comments, so it never reaches the object file.
17024 @cindex global offset table
17026 Generate position-independent code (PIC) suitable for use in a shared
17027 library, if supported for the target machine. Such code accesses all
17028 constant addresses through a global offset table (GOT)@. The dynamic
17029 loader resolves the GOT entries when the program starts (the dynamic
17030 loader is not part of GCC; it is part of the operating system). If
17031 the GOT size for the linked executable exceeds a machine-specific
17032 maximum size, you get an error message from the linker indicating that
17033 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
17034 instead. (These maximums are 8k on the SPARC and 32k
17035 on the m68k and RS/6000. The 386 has no such limit.)
17037 Position-independent code requires special support, and therefore works
17038 only on certain machines. For the 386, GCC supports PIC for System V
17039 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
17040 position-independent.
17042 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
17047 If supported for the target machine, emit position-independent code,
17048 suitable for dynamic linking and avoiding any limit on the size of the
17049 global offset table. This option makes a difference on the m68k,
17050 PowerPC and SPARC@.
17052 Position-independent code requires special support, and therefore works
17053 only on certain machines.
17055 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
17062 These options are similar to @option{-fpic} and @option{-fPIC}, but
17063 generated position independent code can be only linked into executables.
17064 Usually these options are used when @option{-pie} GCC option will be
17065 used during linking.
17067 @option{-fpie} and @option{-fPIE} both define the macros
17068 @code{__pie__} and @code{__PIE__}. The macros have the value 1
17069 for @option{-fpie} and 2 for @option{-fPIE}.
17071 @item -fno-jump-tables
17072 @opindex fno-jump-tables
17073 Do not use jump tables for switch statements even where it would be
17074 more efficient than other code generation strategies. This option is
17075 of use in conjunction with @option{-fpic} or @option{-fPIC} for
17076 building code which forms part of a dynamic linker and cannot
17077 reference the address of a jump table. On some targets, jump tables
17078 do not require a GOT and this option is not needed.
17080 @item -ffixed-@var{reg}
17082 Treat the register named @var{reg} as a fixed register; generated code
17083 should never refer to it (except perhaps as a stack pointer, frame
17084 pointer or in some other fixed role).
17086 @var{reg} must be the name of a register. The register names accepted
17087 are machine-specific and are defined in the @code{REGISTER_NAMES}
17088 macro in the machine description macro file.
17090 This flag does not have a negative form, because it specifies a
17093 @item -fcall-used-@var{reg}
17094 @opindex fcall-used
17095 Treat the register named @var{reg} as an allocable register that is
17096 clobbered by function calls. It may be allocated for temporaries or
17097 variables that do not live across a call. Functions compiled this way
17098 will not save and restore the register @var{reg}.
17100 It is an error to used this flag with the frame pointer or stack pointer.
17101 Use of this flag for other registers that have fixed pervasive roles in
17102 the machine's execution model will produce disastrous results.
17104 This flag does not have a negative form, because it specifies a
17107 @item -fcall-saved-@var{reg}
17108 @opindex fcall-saved
17109 Treat the register named @var{reg} as an allocable register saved by
17110 functions. It may be allocated even for temporaries or variables that
17111 live across a call. Functions compiled this way will save and restore
17112 the register @var{reg} if they use it.
17114 It is an error to used this flag with the frame pointer or stack pointer.
17115 Use of this flag for other registers that have fixed pervasive roles in
17116 the machine's execution model will produce disastrous results.
17118 A different sort of disaster will result from the use of this flag for
17119 a register in which function values may be returned.
17121 This flag does not have a negative form, because it specifies a
17124 @item -fpack-struct[=@var{n}]
17125 @opindex fpack-struct
17126 Without a value specified, pack all structure members together without
17127 holes. When a value is specified (which must be a small power of two), pack
17128 structure members according to this value, representing the maximum
17129 alignment (that is, objects with default alignment requirements larger than
17130 this will be output potentially unaligned at the next fitting location.
17132 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
17133 code that is not binary compatible with code generated without that switch.
17134 Additionally, it makes the code suboptimal.
17135 Use it to conform to a non-default application binary interface.
17137 @item -finstrument-functions
17138 @opindex finstrument-functions
17139 Generate instrumentation calls for entry and exit to functions. Just
17140 after function entry and just before function exit, the following
17141 profiling functions will be called with the address of the current
17142 function and its call site. (On some platforms,
17143 @code{__builtin_return_address} does not work beyond the current
17144 function, so the call site information may not be available to the
17145 profiling functions otherwise.)
17148 void __cyg_profile_func_enter (void *this_fn,
17150 void __cyg_profile_func_exit (void *this_fn,
17154 The first argument is the address of the start of the current function,
17155 which may be looked up exactly in the symbol table.
17157 This instrumentation is also done for functions expanded inline in other
17158 functions. The profiling calls will indicate where, conceptually, the
17159 inline function is entered and exited. This means that addressable
17160 versions of such functions must be available. If all your uses of a
17161 function are expanded inline, this may mean an additional expansion of
17162 code size. If you use @samp{extern inline} in your C code, an
17163 addressable version of such functions must be provided. (This is
17164 normally the case anyways, but if you get lucky and the optimizer always
17165 expands the functions inline, you might have gotten away without
17166 providing static copies.)
17168 A function may be given the attribute @code{no_instrument_function}, in
17169 which case this instrumentation will not be done. This can be used, for
17170 example, for the profiling functions listed above, high-priority
17171 interrupt routines, and any functions from which the profiling functions
17172 cannot safely be called (perhaps signal handlers, if the profiling
17173 routines generate output or allocate memory).
17175 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
17176 @opindex finstrument-functions-exclude-file-list
17178 Set the list of functions that are excluded from instrumentation (see
17179 the description of @code{-finstrument-functions}). If the file that
17180 contains a function definition matches with one of @var{file}, then
17181 that function is not instrumented. The match is done on substrings:
17182 if the @var{file} parameter is a substring of the file name, it is
17183 considered to be a match.
17186 @code{-finstrument-functions-exclude-file-list=/bits/stl,include/sys}
17187 will exclude any inline function defined in files whose pathnames
17188 contain @code{/bits/stl} or @code{include/sys}.
17190 If, for some reason, you want to include letter @code{','} in one of
17191 @var{sym}, write @code{'\,'}. For example,
17192 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
17193 (note the single quote surrounding the option).
17195 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
17196 @opindex finstrument-functions-exclude-function-list
17198 This is similar to @code{-finstrument-functions-exclude-file-list},
17199 but this option sets the list of function names to be excluded from
17200 instrumentation. The function name to be matched is its user-visible
17201 name, such as @code{vector<int> blah(const vector<int> &)}, not the
17202 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
17203 match is done on substrings: if the @var{sym} parameter is a substring
17204 of the function name, it is considered to be a match. For C99 and C++
17205 extended identifiers, the function name must be given in UTF-8, not
17206 using universal character names.
17208 @item -fstack-check
17209 @opindex fstack-check
17210 Generate code to verify that you do not go beyond the boundary of the
17211 stack. You should specify this flag if you are running in an
17212 environment with multiple threads, but only rarely need to specify it in
17213 a single-threaded environment since stack overflow is automatically
17214 detected on nearly all systems if there is only one stack.
17216 Note that this switch does not actually cause checking to be done; the
17217 operating system or the language runtime must do that. The switch causes
17218 generation of code to ensure that they see the stack being extended.
17220 You can additionally specify a string parameter: @code{no} means no
17221 checking, @code{generic} means force the use of old-style checking,
17222 @code{specific} means use the best checking method and is equivalent
17223 to bare @option{-fstack-check}.
17225 Old-style checking is a generic mechanism that requires no specific
17226 target support in the compiler but comes with the following drawbacks:
17230 Modified allocation strategy for large objects: they will always be
17231 allocated dynamically if their size exceeds a fixed threshold.
17234 Fixed limit on the size of the static frame of functions: when it is
17235 topped by a particular function, stack checking is not reliable and
17236 a warning is issued by the compiler.
17239 Inefficiency: because of both the modified allocation strategy and the
17240 generic implementation, the performances of the code are hampered.
17243 Note that old-style stack checking is also the fallback method for
17244 @code{specific} if no target support has been added in the compiler.
17246 @item -fstack-limit-register=@var{reg}
17247 @itemx -fstack-limit-symbol=@var{sym}
17248 @itemx -fno-stack-limit
17249 @opindex fstack-limit-register
17250 @opindex fstack-limit-symbol
17251 @opindex fno-stack-limit
17252 Generate code to ensure that the stack does not grow beyond a certain value,
17253 either the value of a register or the address of a symbol. If the stack
17254 would grow beyond the value, a signal is raised. For most targets,
17255 the signal is raised before the stack overruns the boundary, so
17256 it is possible to catch the signal without taking special precautions.
17258 For instance, if the stack starts at absolute address @samp{0x80000000}
17259 and grows downwards, you can use the flags
17260 @option{-fstack-limit-symbol=__stack_limit} and
17261 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
17262 of 128KB@. Note that this may only work with the GNU linker.
17264 @cindex aliasing of parameters
17265 @cindex parameters, aliased
17266 @item -fargument-alias
17267 @itemx -fargument-noalias
17268 @itemx -fargument-noalias-global
17269 @itemx -fargument-noalias-anything
17270 @opindex fargument-alias
17271 @opindex fargument-noalias
17272 @opindex fargument-noalias-global
17273 @opindex fargument-noalias-anything
17274 Specify the possible relationships among parameters and between
17275 parameters and global data.
17277 @option{-fargument-alias} specifies that arguments (parameters) may
17278 alias each other and may alias global storage.@*
17279 @option{-fargument-noalias} specifies that arguments do not alias
17280 each other, but may alias global storage.@*
17281 @option{-fargument-noalias-global} specifies that arguments do not
17282 alias each other and do not alias global storage.
17283 @option{-fargument-noalias-anything} specifies that arguments do not
17284 alias any other storage.
17286 Each language will automatically use whatever option is required by
17287 the language standard. You should not need to use these options yourself.
17289 @item -fleading-underscore
17290 @opindex fleading-underscore
17291 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
17292 change the way C symbols are represented in the object file. One use
17293 is to help link with legacy assembly code.
17295 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
17296 generate code that is not binary compatible with code generated without that
17297 switch. Use it to conform to a non-default application binary interface.
17298 Not all targets provide complete support for this switch.
17300 @item -ftls-model=@var{model}
17301 @opindex ftls-model
17302 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
17303 The @var{model} argument should be one of @code{global-dynamic},
17304 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
17306 The default without @option{-fpic} is @code{initial-exec}; with
17307 @option{-fpic} the default is @code{global-dynamic}.
17309 @item -fvisibility=@var{default|internal|hidden|protected}
17310 @opindex fvisibility
17311 Set the default ELF image symbol visibility to the specified option---all
17312 symbols will be marked with this unless overridden within the code.
17313 Using this feature can very substantially improve linking and
17314 load times of shared object libraries, produce more optimized
17315 code, provide near-perfect API export and prevent symbol clashes.
17316 It is @strong{strongly} recommended that you use this in any shared objects
17319 Despite the nomenclature, @code{default} always means public ie;
17320 available to be linked against from outside the shared object.
17321 @code{protected} and @code{internal} are pretty useless in real-world
17322 usage so the only other commonly used option will be @code{hidden}.
17323 The default if @option{-fvisibility} isn't specified is
17324 @code{default}, i.e., make every
17325 symbol public---this causes the same behavior as previous versions of
17328 A good explanation of the benefits offered by ensuring ELF
17329 symbols have the correct visibility is given by ``How To Write
17330 Shared Libraries'' by Ulrich Drepper (which can be found at
17331 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
17332 solution made possible by this option to marking things hidden when
17333 the default is public is to make the default hidden and mark things
17334 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
17335 and @code{__attribute__ ((visibility("default")))} instead of
17336 @code{__declspec(dllexport)} you get almost identical semantics with
17337 identical syntax. This is a great boon to those working with
17338 cross-platform projects.
17340 For those adding visibility support to existing code, you may find
17341 @samp{#pragma GCC visibility} of use. This works by you enclosing
17342 the declarations you wish to set visibility for with (for example)
17343 @samp{#pragma GCC visibility push(hidden)} and
17344 @samp{#pragma GCC visibility pop}.
17345 Bear in mind that symbol visibility should be viewed @strong{as
17346 part of the API interface contract} and thus all new code should
17347 always specify visibility when it is not the default ie; declarations
17348 only for use within the local DSO should @strong{always} be marked explicitly
17349 as hidden as so to avoid PLT indirection overheads---making this
17350 abundantly clear also aids readability and self-documentation of the code.
17351 Note that due to ISO C++ specification requirements, operator new and
17352 operator delete must always be of default visibility.
17354 Be aware that headers from outside your project, in particular system
17355 headers and headers from any other library you use, may not be
17356 expecting to be compiled with visibility other than the default. You
17357 may need to explicitly say @samp{#pragma GCC visibility push(default)}
17358 before including any such headers.
17360 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
17361 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
17362 no modifications. However, this means that calls to @samp{extern}
17363 functions with no explicit visibility will use the PLT, so it is more
17364 effective to use @samp{__attribute ((visibility))} and/or
17365 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
17366 declarations should be treated as hidden.
17368 Note that @samp{-fvisibility} does affect C++ vague linkage
17369 entities. This means that, for instance, an exception class that will
17370 be thrown between DSOs must be explicitly marked with default
17371 visibility so that the @samp{type_info} nodes will be unified between
17374 An overview of these techniques, their benefits and how to use them
17375 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
17381 @node Environment Variables
17382 @section Environment Variables Affecting GCC
17383 @cindex environment variables
17385 @c man begin ENVIRONMENT
17386 This section describes several environment variables that affect how GCC
17387 operates. Some of them work by specifying directories or prefixes to use
17388 when searching for various kinds of files. Some are used to specify other
17389 aspects of the compilation environment.
17391 Note that you can also specify places to search using options such as
17392 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
17393 take precedence over places specified using environment variables, which
17394 in turn take precedence over those specified by the configuration of GCC@.
17395 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
17396 GNU Compiler Collection (GCC) Internals}.
17401 @c @itemx LC_COLLATE
17403 @c @itemx LC_MONETARY
17404 @c @itemx LC_NUMERIC
17409 @c @findex LC_COLLATE
17410 @findex LC_MESSAGES
17411 @c @findex LC_MONETARY
17412 @c @findex LC_NUMERIC
17416 These environment variables control the way that GCC uses
17417 localization information that allow GCC to work with different
17418 national conventions. GCC inspects the locale categories
17419 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
17420 so. These locale categories can be set to any value supported by your
17421 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
17422 Kingdom encoded in UTF-8.
17424 The @env{LC_CTYPE} environment variable specifies character
17425 classification. GCC uses it to determine the character boundaries in
17426 a string; this is needed for some multibyte encodings that contain quote
17427 and escape characters that would otherwise be interpreted as a string
17430 The @env{LC_MESSAGES} environment variable specifies the language to
17431 use in diagnostic messages.
17433 If the @env{LC_ALL} environment variable is set, it overrides the value
17434 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
17435 and @env{LC_MESSAGES} default to the value of the @env{LANG}
17436 environment variable. If none of these variables are set, GCC
17437 defaults to traditional C English behavior.
17441 If @env{TMPDIR} is set, it specifies the directory to use for temporary
17442 files. GCC uses temporary files to hold the output of one stage of
17443 compilation which is to be used as input to the next stage: for example,
17444 the output of the preprocessor, which is the input to the compiler
17447 @item GCC_EXEC_PREFIX
17448 @findex GCC_EXEC_PREFIX
17449 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
17450 names of the subprograms executed by the compiler. No slash is added
17451 when this prefix is combined with the name of a subprogram, but you can
17452 specify a prefix that ends with a slash if you wish.
17454 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
17455 an appropriate prefix to use based on the pathname it was invoked with.
17457 If GCC cannot find the subprogram using the specified prefix, it
17458 tries looking in the usual places for the subprogram.
17460 The default value of @env{GCC_EXEC_PREFIX} is
17461 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
17462 the installed compiler. In many cases @var{prefix} is the value
17463 of @code{prefix} when you ran the @file{configure} script.
17465 Other prefixes specified with @option{-B} take precedence over this prefix.
17467 This prefix is also used for finding files such as @file{crt0.o} that are
17470 In addition, the prefix is used in an unusual way in finding the
17471 directories to search for header files. For each of the standard
17472 directories whose name normally begins with @samp{/usr/local/lib/gcc}
17473 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
17474 replacing that beginning with the specified prefix to produce an
17475 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
17476 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
17477 These alternate directories are searched first; the standard directories
17478 come next. If a standard directory begins with the configured
17479 @var{prefix} then the value of @var{prefix} is replaced by
17480 @env{GCC_EXEC_PREFIX} when looking for header files.
17482 @item COMPILER_PATH
17483 @findex COMPILER_PATH
17484 The value of @env{COMPILER_PATH} is a colon-separated list of
17485 directories, much like @env{PATH}. GCC tries the directories thus
17486 specified when searching for subprograms, if it can't find the
17487 subprograms using @env{GCC_EXEC_PREFIX}.
17490 @findex LIBRARY_PATH
17491 The value of @env{LIBRARY_PATH} is a colon-separated list of
17492 directories, much like @env{PATH}. When configured as a native compiler,
17493 GCC tries the directories thus specified when searching for special
17494 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
17495 using GCC also uses these directories when searching for ordinary
17496 libraries for the @option{-l} option (but directories specified with
17497 @option{-L} come first).
17501 @cindex locale definition
17502 This variable is used to pass locale information to the compiler. One way in
17503 which this information is used is to determine the character set to be used
17504 when character literals, string literals and comments are parsed in C and C++.
17505 When the compiler is configured to allow multibyte characters,
17506 the following values for @env{LANG} are recognized:
17510 Recognize JIS characters.
17512 Recognize SJIS characters.
17514 Recognize EUCJP characters.
17517 If @env{LANG} is not defined, or if it has some other value, then the
17518 compiler will use mblen and mbtowc as defined by the default locale to
17519 recognize and translate multibyte characters.
17523 Some additional environments variables affect the behavior of the
17526 @include cppenv.texi
17530 @node Precompiled Headers
17531 @section Using Precompiled Headers
17532 @cindex precompiled headers
17533 @cindex speed of compilation
17535 Often large projects have many header files that are included in every
17536 source file. The time the compiler takes to process these header files
17537 over and over again can account for nearly all of the time required to
17538 build the project. To make builds faster, GCC allows users to
17539 `precompile' a header file; then, if builds can use the precompiled
17540 header file they will be much faster.
17542 To create a precompiled header file, simply compile it as you would any
17543 other file, if necessary using the @option{-x} option to make the driver
17544 treat it as a C or C++ header file. You will probably want to use a
17545 tool like @command{make} to keep the precompiled header up-to-date when
17546 the headers it contains change.
17548 A precompiled header file will be searched for when @code{#include} is
17549 seen in the compilation. As it searches for the included file
17550 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
17551 compiler looks for a precompiled header in each directory just before it
17552 looks for the include file in that directory. The name searched for is
17553 the name specified in the @code{#include} with @samp{.gch} appended. If
17554 the precompiled header file can't be used, it is ignored.
17556 For instance, if you have @code{#include "all.h"}, and you have
17557 @file{all.h.gch} in the same directory as @file{all.h}, then the
17558 precompiled header file will be used if possible, and the original
17559 header will be used otherwise.
17561 Alternatively, you might decide to put the precompiled header file in a
17562 directory and use @option{-I} to ensure that directory is searched
17563 before (or instead of) the directory containing the original header.
17564 Then, if you want to check that the precompiled header file is always
17565 used, you can put a file of the same name as the original header in this
17566 directory containing an @code{#error} command.
17568 This also works with @option{-include}. So yet another way to use
17569 precompiled headers, good for projects not designed with precompiled
17570 header files in mind, is to simply take most of the header files used by
17571 a project, include them from another header file, precompile that header
17572 file, and @option{-include} the precompiled header. If the header files
17573 have guards against multiple inclusion, they will be skipped because
17574 they've already been included (in the precompiled header).
17576 If you need to precompile the same header file for different
17577 languages, targets, or compiler options, you can instead make a
17578 @emph{directory} named like @file{all.h.gch}, and put each precompiled
17579 header in the directory, perhaps using @option{-o}. It doesn't matter
17580 what you call the files in the directory, every precompiled header in
17581 the directory will be considered. The first precompiled header
17582 encountered in the directory that is valid for this compilation will
17583 be used; they're searched in no particular order.
17585 There are many other possibilities, limited only by your imagination,
17586 good sense, and the constraints of your build system.
17588 A precompiled header file can be used only when these conditions apply:
17592 Only one precompiled header can be used in a particular compilation.
17595 A precompiled header can't be used once the first C token is seen. You
17596 can have preprocessor directives before a precompiled header; you can
17597 even include a precompiled header from inside another header, so long as
17598 there are no C tokens before the @code{#include}.
17601 The precompiled header file must be produced for the same language as
17602 the current compilation. You can't use a C precompiled header for a C++
17606 The precompiled header file must have been produced by the same compiler
17607 binary as the current compilation is using.
17610 Any macros defined before the precompiled header is included must
17611 either be defined in the same way as when the precompiled header was
17612 generated, or must not affect the precompiled header, which usually
17613 means that they don't appear in the precompiled header at all.
17615 The @option{-D} option is one way to define a macro before a
17616 precompiled header is included; using a @code{#define} can also do it.
17617 There are also some options that define macros implicitly, like
17618 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
17621 @item If debugging information is output when using the precompiled
17622 header, using @option{-g} or similar, the same kind of debugging information
17623 must have been output when building the precompiled header. However,
17624 a precompiled header built using @option{-g} can be used in a compilation
17625 when no debugging information is being output.
17627 @item The same @option{-m} options must generally be used when building
17628 and using the precompiled header. @xref{Submodel Options},
17629 for any cases where this rule is relaxed.
17631 @item Each of the following options must be the same when building and using
17632 the precompiled header:
17634 @gccoptlist{-fexceptions}
17637 Some other command-line options starting with @option{-f},
17638 @option{-p}, or @option{-O} must be defined in the same way as when
17639 the precompiled header was generated. At present, it's not clear
17640 which options are safe to change and which are not; the safest choice
17641 is to use exactly the same options when generating and using the
17642 precompiled header. The following are known to be safe:
17644 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
17645 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
17646 -fsched-verbose=<number> -fschedule-insns -fvisibility= @gol
17651 For all of these except the last, the compiler will automatically
17652 ignore the precompiled header if the conditions aren't met. If you
17653 find an option combination that doesn't work and doesn't cause the
17654 precompiled header to be ignored, please consider filing a bug report,
17657 If you do use differing options when generating and using the
17658 precompiled header, the actual behavior will be a mixture of the
17659 behavior for the options. For instance, if you use @option{-g} to
17660 generate the precompiled header but not when using it, you may or may
17661 not get debugging information for routines in the precompiled header.