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 -fuse-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 -mxop -mlwp @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}
635 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
636 -msign-extend-enabled -muser-enabled}
638 @emph{M32R/D Options}
639 @gccoptlist{-m32r2 -m32rx -m32r @gol
641 -malign-loops -mno-align-loops @gol
642 -missue-rate=@var{number} @gol
643 -mbranch-cost=@var{number} @gol
644 -mmodel=@var{code-size-model-type} @gol
645 -msdata=@var{sdata-type} @gol
646 -mno-flush-func -mflush-func=@var{name} @gol
647 -mno-flush-trap -mflush-trap=@var{number} @gol
651 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
653 @emph{M680x0 Options}
654 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
655 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
656 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
657 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
658 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
659 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
660 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
661 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
664 @emph{M68hc1x Options}
665 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
666 -mauto-incdec -minmax -mlong-calls -mshort @gol
667 -msoft-reg-count=@var{count}}
670 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
671 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
672 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
673 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
674 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
677 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
678 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
679 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
680 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
684 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
685 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
686 -mips64 -mips64r2 @gol
687 -mips16 -mno-mips16 -mflip-mips16 @gol
688 -minterlink-mips16 -mno-interlink-mips16 @gol
689 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
690 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
691 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
692 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
693 -mfpu=@var{fpu-type} @gol
694 -msmartmips -mno-smartmips @gol
695 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
696 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
697 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
698 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
699 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
700 -membedded-data -mno-embedded-data @gol
701 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
702 -mcode-readable=@var{setting} @gol
703 -msplit-addresses -mno-split-addresses @gol
704 -mexplicit-relocs -mno-explicit-relocs @gol
705 -mcheck-zero-division -mno-check-zero-division @gol
706 -mdivide-traps -mdivide-breaks @gol
707 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
708 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
709 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
710 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
711 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
712 -mflush-func=@var{func} -mno-flush-func @gol
713 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
714 -mfp-exceptions -mno-fp-exceptions @gol
715 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
716 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
719 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
720 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
721 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
722 -mno-base-addresses -msingle-exit -mno-single-exit}
724 @emph{MN10300 Options}
725 @gccoptlist{-mmult-bug -mno-mult-bug @gol
726 -mam33 -mno-am33 @gol
727 -mam33-2 -mno-am33-2 @gol
728 -mreturn-pointer-on-d0 @gol
731 @emph{PDP-11 Options}
732 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
733 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
734 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
735 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
736 -mbranch-expensive -mbranch-cheap @gol
737 -msplit -mno-split -munix-asm -mdec-asm}
739 @emph{picoChip Options}
740 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N}
741 -msymbol-as-address -mno-inefficient-warnings}
743 @emph{PowerPC Options}
744 See RS/6000 and PowerPC Options.
746 @emph{RS/6000 and PowerPC Options}
747 @gccoptlist{-mcpu=@var{cpu-type} @gol
748 -mtune=@var{cpu-type} @gol
749 -mpower -mno-power -mpower2 -mno-power2 @gol
750 -mpowerpc -mpowerpc64 -mno-powerpc @gol
751 -maltivec -mno-altivec @gol
752 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
753 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
754 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
755 -mfprnd -mno-fprnd @gol
756 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
757 -mnew-mnemonics -mold-mnemonics @gol
758 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
759 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
760 -malign-power -malign-natural @gol
761 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
762 -msingle-float -mdouble-float -msimple-fpu @gol
763 -mstring -mno-string -mupdate -mno-update @gol
764 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
765 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
766 -mstrict-align -mno-strict-align -mrelocatable @gol
767 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
768 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
769 -mdynamic-no-pic -maltivec -mswdiv @gol
770 -mprioritize-restricted-insns=@var{priority} @gol
771 -msched-costly-dep=@var{dependence_type} @gol
772 -minsert-sched-nops=@var{scheme} @gol
773 -mcall-sysv -mcall-netbsd @gol
774 -maix-struct-return -msvr4-struct-return @gol
775 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
776 -misel -mno-isel @gol
777 -misel=yes -misel=no @gol
779 -mspe=yes -mspe=no @gol
781 -mgen-cell-microcode -mwarn-cell-microcode @gol
782 -mvrsave -mno-vrsave @gol
783 -mmulhw -mno-mulhw @gol
784 -mdlmzb -mno-dlmzb @gol
785 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
786 -mprototype -mno-prototype @gol
787 -msim -mmvme -mads -myellowknife -memb -msdata @gol
788 -msdata=@var{opt} -mvxworks -G @var{num} -pthread}
791 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
793 -mbig-endian-data -mlittle-endian-data @gol
796 -mas100-syntax -mno-as100-syntax@gol
798 -mmax-constant-size=@gol
800 -msave-acc-in-interrupts}
802 @emph{S/390 and zSeries Options}
803 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
804 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
805 -mlong-double-64 -mlong-double-128 @gol
806 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
807 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
808 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
809 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
810 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
813 @gccoptlist{-meb -mel @gol
817 -mscore5 -mscore5u -mscore7 -mscore7d}
820 @gccoptlist{-m1 -m2 -m2e @gol
821 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
823 -m4-nofpu -m4-single-only -m4-single -m4 @gol
824 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
825 -m5-64media -m5-64media-nofpu @gol
826 -m5-32media -m5-32media-nofpu @gol
827 -m5-compact -m5-compact-nofpu @gol
828 -mb -ml -mdalign -mrelax @gol
829 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
830 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
831 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
832 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
833 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
837 @gccoptlist{-mcpu=@var{cpu-type} @gol
838 -mtune=@var{cpu-type} @gol
839 -mcmodel=@var{code-model} @gol
840 -m32 -m64 -mapp-regs -mno-app-regs @gol
841 -mfaster-structs -mno-faster-structs @gol
842 -mfpu -mno-fpu -mhard-float -msoft-float @gol
843 -mhard-quad-float -msoft-quad-float @gol
844 -mimpure-text -mno-impure-text -mlittle-endian @gol
845 -mstack-bias -mno-stack-bias @gol
846 -munaligned-doubles -mno-unaligned-doubles @gol
847 -mv8plus -mno-v8plus -mvis -mno-vis
848 -threads -pthreads -pthread}
851 @gccoptlist{-mwarn-reloc -merror-reloc @gol
852 -msafe-dma -munsafe-dma @gol
854 -msmall-mem -mlarge-mem -mstdmain @gol
855 -mfixed-range=@var{register-range} @gol
857 -maddress-space-conversion -mno-address-space-conversion @gol
858 -mcache-size=@var{cache-size} @gol
859 -matomic-updates -mno-atomic-updates}
861 @emph{System V Options}
862 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
865 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
866 -mprolog-function -mno-prolog-function -mspace @gol
867 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
868 -mapp-regs -mno-app-regs @gol
869 -mdisable-callt -mno-disable-callt @gol
875 @gccoptlist{-mg -mgnu -munix}
877 @emph{VxWorks Options}
878 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
879 -Xbind-lazy -Xbind-now}
881 @emph{x86-64 Options}
882 See i386 and x86-64 Options.
884 @emph{i386 and x86-64 Windows Options}
885 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll
886 -mnop-fun-dllimport -mthread -municode -mwin32 -mwindows
887 -fno-set-stack-executable}
889 @emph{Xstormy16 Options}
892 @emph{Xtensa Options}
893 @gccoptlist{-mconst16 -mno-const16 @gol
894 -mfused-madd -mno-fused-madd @gol
895 -mserialize-volatile -mno-serialize-volatile @gol
896 -mtext-section-literals -mno-text-section-literals @gol
897 -mtarget-align -mno-target-align @gol
898 -mlongcalls -mno-longcalls}
900 @emph{zSeries Options}
901 See S/390 and zSeries Options.
903 @item Code Generation Options
904 @xref{Code Gen Options,,Options for Code Generation Conventions}.
905 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
906 -ffixed-@var{reg} -fexceptions @gol
907 -fnon-call-exceptions -funwind-tables @gol
908 -fasynchronous-unwind-tables @gol
909 -finhibit-size-directive -finstrument-functions @gol
910 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
911 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
912 -fno-common -fno-ident @gol
913 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
914 -fno-jump-tables @gol
915 -frecord-gcc-switches @gol
916 -freg-struct-return -fshort-enums @gol
917 -fshort-double -fshort-wchar @gol
918 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
919 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
920 -fno-stack-limit -fargument-alias -fargument-noalias @gol
921 -fargument-noalias-global -fargument-noalias-anything @gol
922 -fleading-underscore -ftls-model=@var{model} @gol
923 -ftrapv -fwrapv -fbounds-check @gol
928 * Overall Options:: Controlling the kind of output:
929 an executable, object files, assembler files,
930 or preprocessed source.
931 * C Dialect Options:: Controlling the variant of C language compiled.
932 * C++ Dialect Options:: Variations on C++.
933 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
935 * Language Independent Options:: Controlling how diagnostics should be
937 * Warning Options:: How picky should the compiler be?
938 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
939 * Optimize Options:: How much optimization?
940 * Preprocessor Options:: Controlling header files and macro definitions.
941 Also, getting dependency information for Make.
942 * Assembler Options:: Passing options to the assembler.
943 * Link Options:: Specifying libraries and so on.
944 * Directory Options:: Where to find header files and libraries.
945 Where to find the compiler executable files.
946 * Spec Files:: How to pass switches to sub-processes.
947 * Target Options:: Running a cross-compiler, or an old version of GCC.
950 @node Overall Options
951 @section Options Controlling the Kind of Output
953 Compilation can involve up to four stages: preprocessing, compilation
954 proper, assembly and linking, always in that order. GCC is capable of
955 preprocessing and compiling several files either into several
956 assembler input files, or into one assembler input file; then each
957 assembler input file produces an object file, and linking combines all
958 the object files (those newly compiled, and those specified as input)
959 into an executable file.
961 @cindex file name suffix
962 For any given input file, the file name suffix determines what kind of
967 C source code which must be preprocessed.
970 C source code which should not be preprocessed.
973 C++ source code which should not be preprocessed.
976 Objective-C source code. Note that you must link with the @file{libobjc}
977 library to make an Objective-C program work.
980 Objective-C source code which should not be preprocessed.
984 Objective-C++ source code. Note that you must link with the @file{libobjc}
985 library to make an Objective-C++ program work. Note that @samp{.M} refers
986 to a literal capital M@.
989 Objective-C++ source code which should not be preprocessed.
992 C, C++, Objective-C or Objective-C++ header file to be turned into a
997 @itemx @var{file}.cxx
998 @itemx @var{file}.cpp
999 @itemx @var{file}.CPP
1000 @itemx @var{file}.c++
1002 C++ source code which must be preprocessed. Note that in @samp{.cxx},
1003 the last two letters must both be literally @samp{x}. Likewise,
1004 @samp{.C} refers to a literal capital C@.
1008 Objective-C++ source code which must be preprocessed.
1010 @item @var{file}.mii
1011 Objective-C++ source code which should not be preprocessed.
1015 @itemx @var{file}.hp
1016 @itemx @var{file}.hxx
1017 @itemx @var{file}.hpp
1018 @itemx @var{file}.HPP
1019 @itemx @var{file}.h++
1020 @itemx @var{file}.tcc
1021 C++ header file to be turned into a precompiled header.
1024 @itemx @var{file}.for
1025 @itemx @var{file}.ftn
1026 Fixed form Fortran source code which should not be preprocessed.
1029 @itemx @var{file}.FOR
1030 @itemx @var{file}.fpp
1031 @itemx @var{file}.FPP
1032 @itemx @var{file}.FTN
1033 Fixed form Fortran source code which must be preprocessed (with the traditional
1036 @item @var{file}.f90
1037 @itemx @var{file}.f95
1038 @itemx @var{file}.f03
1039 @itemx @var{file}.f08
1040 Free form Fortran source code which should not be preprocessed.
1042 @item @var{file}.F90
1043 @itemx @var{file}.F95
1044 @itemx @var{file}.F03
1045 @itemx @var{file}.F08
1046 Free form Fortran source code which must be preprocessed (with the
1047 traditional preprocessor).
1049 @c FIXME: Descriptions of Java file types.
1055 @item @var{file}.ads
1056 Ada source code file which contains a library unit declaration (a
1057 declaration of a package, subprogram, or generic, or a generic
1058 instantiation), or a library unit renaming declaration (a package,
1059 generic, or subprogram renaming declaration). Such files are also
1062 @item @var{file}.adb
1063 Ada source code file containing a library unit body (a subprogram or
1064 package body). Such files are also called @dfn{bodies}.
1066 @c GCC also knows about some suffixes for languages not yet included:
1077 @itemx @var{file}.sx
1078 Assembler code which must be preprocessed.
1081 An object file to be fed straight into linking.
1082 Any file name with no recognized suffix is treated this way.
1086 You can specify the input language explicitly with the @option{-x} option:
1089 @item -x @var{language}
1090 Specify explicitly the @var{language} for the following input files
1091 (rather than letting the compiler choose a default based on the file
1092 name suffix). This option applies to all following input files until
1093 the next @option{-x} option. Possible values for @var{language} are:
1095 c c-header c-cpp-output
1096 c++ c++-header c++-cpp-output
1097 objective-c objective-c-header objective-c-cpp-output
1098 objective-c++ objective-c++-header objective-c++-cpp-output
1099 assembler assembler-with-cpp
1101 f77 f77-cpp-input f95 f95-cpp-input
1106 Turn off any specification of a language, so that subsequent files are
1107 handled according to their file name suffixes (as they are if @option{-x}
1108 has not been used at all).
1110 @item -pass-exit-codes
1111 @opindex pass-exit-codes
1112 Normally the @command{gcc} program will exit with the code of 1 if any
1113 phase of the compiler returns a non-success return code. If you specify
1114 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1115 numerically highest error produced by any phase that returned an error
1116 indication. The C, C++, and Fortran frontends return 4, if an internal
1117 compiler error is encountered.
1120 If you only want some of the stages of compilation, you can use
1121 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1122 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1123 @command{gcc} is to stop. Note that some combinations (for example,
1124 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1129 Compile or assemble the source files, but do not link. The linking
1130 stage simply is not done. The ultimate output is in the form of an
1131 object file for each source file.
1133 By default, the object file name for a source file is made by replacing
1134 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1136 Unrecognized input files, not requiring compilation or assembly, are
1141 Stop after the stage of compilation proper; do not assemble. The output
1142 is in the form of an assembler code file for each non-assembler input
1145 By default, the assembler file name for a source file is made by
1146 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1148 Input files that don't require compilation are ignored.
1152 Stop after the preprocessing stage; do not run the compiler proper. The
1153 output is in the form of preprocessed source code, which is sent to the
1156 Input files which don't require preprocessing are ignored.
1158 @cindex output file option
1161 Place output in file @var{file}. This applies regardless to whatever
1162 sort of output is being produced, whether it be an executable file,
1163 an object file, an assembler file or preprocessed C code.
1165 If @option{-o} is not specified, the default is to put an executable
1166 file in @file{a.out}, the object file for
1167 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1168 assembler file in @file{@var{source}.s}, a precompiled header file in
1169 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1174 Print (on standard error output) the commands executed to run the stages
1175 of compilation. Also print the version number of the compiler driver
1176 program and of the preprocessor and the compiler proper.
1180 Like @option{-v} except the commands are not executed and all command
1181 arguments are quoted. This is useful for shell scripts to capture the
1182 driver-generated command lines.
1186 Use pipes rather than temporary files for communication between the
1187 various stages of compilation. This fails to work on some systems where
1188 the assembler is unable to read from a pipe; but the GNU assembler has
1193 If you are compiling multiple source files, this option tells the driver
1194 to pass all the source files to the compiler at once (for those
1195 languages for which the compiler can handle this). This will allow
1196 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1197 language for which this is supported is C@. If you pass source files for
1198 multiple languages to the driver, using this option, the driver will invoke
1199 the compiler(s) that support IMA once each, passing each compiler all the
1200 source files appropriate for it. For those languages that do not support
1201 IMA this option will be ignored, and the compiler will be invoked once for
1202 each source file in that language. If you use this option in conjunction
1203 with @option{-save-temps}, the compiler will generate multiple
1205 (one for each source file), but only one (combined) @file{.o} or
1210 Print (on the standard output) a description of the command line options
1211 understood by @command{gcc}. If the @option{-v} option is also specified
1212 then @option{--help} will also be passed on to the various processes
1213 invoked by @command{gcc}, so that they can display the command line options
1214 they accept. If the @option{-Wextra} option has also been specified
1215 (prior to the @option{--help} option), then command line options which
1216 have no documentation associated with them will also be displayed.
1219 @opindex target-help
1220 Print (on the standard output) a description of target-specific command
1221 line options for each tool. For some targets extra target-specific
1222 information may also be printed.
1224 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1225 Print (on the standard output) a description of the command line
1226 options understood by the compiler that fit into all specified classes
1227 and qualifiers. These are the supported classes:
1230 @item @samp{optimizers}
1231 This will display all of the optimization options supported by the
1234 @item @samp{warnings}
1235 This will display all of the options controlling warning messages
1236 produced by the compiler.
1239 This will display target-specific options. Unlike the
1240 @option{--target-help} option however, target-specific options of the
1241 linker and assembler will not be displayed. This is because those
1242 tools do not currently support the extended @option{--help=} syntax.
1245 This will display the values recognized by the @option{--param}
1248 @item @var{language}
1249 This will display the options supported for @var{language}, where
1250 @var{language} is the name of one of the languages supported in this
1254 This will display the options that are common to all languages.
1257 These are the supported qualifiers:
1260 @item @samp{undocumented}
1261 Display only those options which are undocumented.
1264 Display options which take an argument that appears after an equal
1265 sign in the same continuous piece of text, such as:
1266 @samp{--help=target}.
1268 @item @samp{separate}
1269 Display options which take an argument that appears as a separate word
1270 following the original option, such as: @samp{-o output-file}.
1273 Thus for example to display all the undocumented target-specific
1274 switches supported by the compiler the following can be used:
1277 --help=target,undocumented
1280 The sense of a qualifier can be inverted by prefixing it with the
1281 @samp{^} character, so for example to display all binary warning
1282 options (i.e., ones that are either on or off and that do not take an
1283 argument), which have a description the following can be used:
1286 --help=warnings,^joined,^undocumented
1289 The argument to @option{--help=} should not consist solely of inverted
1292 Combining several classes is possible, although this usually
1293 restricts the output by so much that there is nothing to display. One
1294 case where it does work however is when one of the classes is
1295 @var{target}. So for example to display all the target-specific
1296 optimization options the following can be used:
1299 --help=target,optimizers
1302 The @option{--help=} option can be repeated on the command line. Each
1303 successive use will display its requested class of options, skipping
1304 those that have already been displayed.
1306 If the @option{-Q} option appears on the command line before the
1307 @option{--help=} option, then the descriptive text displayed by
1308 @option{--help=} is changed. Instead of describing the displayed
1309 options, an indication is given as to whether the option is enabled,
1310 disabled or set to a specific value (assuming that the compiler
1311 knows this at the point where the @option{--help=} option is used).
1313 Here is a truncated example from the ARM port of @command{gcc}:
1316 % gcc -Q -mabi=2 --help=target -c
1317 The following options are target specific:
1319 -mabort-on-noreturn [disabled]
1323 The output is sensitive to the effects of previous command line
1324 options, so for example it is possible to find out which optimizations
1325 are enabled at @option{-O2} by using:
1328 -Q -O2 --help=optimizers
1331 Alternatively you can discover which binary optimizations are enabled
1332 by @option{-O3} by using:
1335 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1336 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1337 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1340 @item -no-canonical-prefixes
1341 @opindex no-canonical-prefixes
1342 Do not expand any symbolic links, resolve references to @samp{/../}
1343 or @samp{/./}, or make the path absolute when generating a relative
1348 Display the version number and copyrights of the invoked GCC@.
1352 Invoke all subcommands under a wrapper program. It takes a single
1353 comma separated list as an argument, which will be used to invoke
1357 gcc -c t.c -wrapper gdb,--args
1360 This will invoke all subprograms of gcc under "gdb --args",
1361 thus cc1 invocation will be "gdb --args cc1 ...".
1363 @item -fplugin=@var{name}.so
1364 Load the plugin code in file @var{name}.so, assumed to be a
1365 shared object to be dlopen'd by the compiler. The base name of
1366 the shared object file is used to identify the plugin for the
1367 purposes of argument parsing (See
1368 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1369 Each plugin should define the callback functions specified in the
1372 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1373 Define an argument called @var{key} with a value of @var{value}
1374 for the plugin called @var{name}.
1376 @include @value{srcdir}/../libiberty/at-file.texi
1380 @section Compiling C++ Programs
1382 @cindex suffixes for C++ source
1383 @cindex C++ source file suffixes
1384 C++ source files conventionally use one of the suffixes @samp{.C},
1385 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1386 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1387 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1388 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1389 files with these names and compiles them as C++ programs even if you
1390 call the compiler the same way as for compiling C programs (usually
1391 with the name @command{gcc}).
1395 However, the use of @command{gcc} does not add the C++ library.
1396 @command{g++} is a program that calls GCC and treats @samp{.c},
1397 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1398 files unless @option{-x} is used, and automatically specifies linking
1399 against the C++ library. This program is also useful when
1400 precompiling a C header file with a @samp{.h} extension for use in C++
1401 compilations. On many systems, @command{g++} is also installed with
1402 the name @command{c++}.
1404 @cindex invoking @command{g++}
1405 When you compile C++ programs, you may specify many of the same
1406 command-line options that you use for compiling programs in any
1407 language; or command-line options meaningful for C and related
1408 languages; or options that are meaningful only for C++ programs.
1409 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1410 explanations of options for languages related to C@.
1411 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1412 explanations of options that are meaningful only for C++ programs.
1414 @node C Dialect Options
1415 @section Options Controlling C Dialect
1416 @cindex dialect options
1417 @cindex language dialect options
1418 @cindex options, dialect
1420 The following options control the dialect of C (or languages derived
1421 from C, such as C++, Objective-C and Objective-C++) that the compiler
1425 @cindex ANSI support
1429 In C mode, this is equivalent to @samp{-std=c89}. In C++ mode, it is
1430 equivalent to @samp{-std=c++98}.
1432 This turns off certain features of GCC that are incompatible with ISO
1433 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1434 such as the @code{asm} and @code{typeof} keywords, and
1435 predefined macros such as @code{unix} and @code{vax} that identify the
1436 type of system you are using. It also enables the undesirable and
1437 rarely used ISO trigraph feature. For the C compiler,
1438 it disables recognition of C++ style @samp{//} comments as well as
1439 the @code{inline} keyword.
1441 The alternate keywords @code{__asm__}, @code{__extension__},
1442 @code{__inline__} and @code{__typeof__} continue to work despite
1443 @option{-ansi}. You would not want to use them in an ISO C program, of
1444 course, but it is useful to put them in header files that might be included
1445 in compilations done with @option{-ansi}. Alternate predefined macros
1446 such as @code{__unix__} and @code{__vax__} are also available, with or
1447 without @option{-ansi}.
1449 The @option{-ansi} option does not cause non-ISO programs to be
1450 rejected gratuitously. For that, @option{-pedantic} is required in
1451 addition to @option{-ansi}. @xref{Warning Options}.
1453 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1454 option is used. Some header files may notice this macro and refrain
1455 from declaring certain functions or defining certain macros that the
1456 ISO standard doesn't call for; this is to avoid interfering with any
1457 programs that might use these names for other things.
1459 Functions that would normally be built in but do not have semantics
1460 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1461 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1462 built-in functions provided by GCC}, for details of the functions
1467 Determine the language standard. @xref{Standards,,Language Standards
1468 Supported by GCC}, for details of these standard versions. This option
1469 is currently only supported when compiling C or C++.
1471 The compiler can accept several base standards, such as @samp{c89} or
1472 @samp{c++98}, and GNU dialects of those standards, such as
1473 @samp{gnu89} or @samp{gnu++98}. By specifying a base standard, the
1474 compiler will accept all programs following that standard and those
1475 using GNU extensions that do not contradict it. For example,
1476 @samp{-std=c89} turns off certain features of GCC that are
1477 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1478 keywords, but not other GNU extensions that do not have a meaning in
1479 ISO C90, such as omitting the middle term of a @code{?:}
1480 expression. On the other hand, by specifying a GNU dialect of a
1481 standard, all features the compiler support are enabled, even when
1482 those features change the meaning of the base standard and some
1483 strict-conforming programs may be rejected. The particular standard
1484 is used by @option{-pedantic} to identify which features are GNU
1485 extensions given that version of the standard. For example
1486 @samp{-std=gnu89 -pedantic} would warn about C++ style @samp{//}
1487 comments, while @samp{-std=gnu99 -pedantic} would not.
1489 A value for this option must be provided; possible values are
1494 Support all ISO C90 programs (certain GNU extensions that conflict
1495 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1497 @item iso9899:199409
1498 ISO C90 as modified in amendment 1.
1504 ISO C99. Note that this standard is not yet fully supported; see
1505 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1506 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1509 GNU dialect of ISO C90 (including some C99 features). This
1510 is the default for C code.
1514 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1515 this will become the default. The name @samp{gnu9x} is deprecated.
1518 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1522 GNU dialect of @option{-std=c++98}. This is the default for
1526 The working draft of the upcoming ISO C++0x standard. This option
1527 enables experimental features that are likely to be included in
1528 C++0x. The working draft is constantly changing, and any feature that is
1529 enabled by this flag may be removed from future versions of GCC if it is
1530 not part of the C++0x standard.
1533 GNU dialect of @option{-std=c++0x}. This option enables
1534 experimental features that may be removed in future versions of GCC.
1537 @item -fgnu89-inline
1538 @opindex fgnu89-inline
1539 The option @option{-fgnu89-inline} tells GCC to use the traditional
1540 GNU semantics for @code{inline} functions when in C99 mode.
1541 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1542 is accepted and ignored by GCC versions 4.1.3 up to but not including
1543 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1544 C99 mode. Using this option is roughly equivalent to adding the
1545 @code{gnu_inline} function attribute to all inline functions
1546 (@pxref{Function Attributes}).
1548 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1549 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1550 specifies the default behavior). This option was first supported in
1551 GCC 4.3. This option is not supported in C89 or gnu89 mode.
1553 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1554 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1555 in effect for @code{inline} functions. @xref{Common Predefined
1556 Macros,,,cpp,The C Preprocessor}.
1558 @item -aux-info @var{filename}
1560 Output to the given filename prototyped declarations for all functions
1561 declared and/or defined in a translation unit, including those in header
1562 files. This option is silently ignored in any language other than C@.
1564 Besides declarations, the file indicates, in comments, the origin of
1565 each declaration (source file and line), whether the declaration was
1566 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1567 @samp{O} for old, respectively, in the first character after the line
1568 number and the colon), and whether it came from a declaration or a
1569 definition (@samp{C} or @samp{F}, respectively, in the following
1570 character). In the case of function definitions, a K&R-style list of
1571 arguments followed by their declarations is also provided, inside
1572 comments, after the declaration.
1576 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1577 keyword, so that code can use these words as identifiers. You can use
1578 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1579 instead. @option{-ansi} implies @option{-fno-asm}.
1581 In C++, this switch only affects the @code{typeof} keyword, since
1582 @code{asm} and @code{inline} are standard keywords. You may want to
1583 use the @option{-fno-gnu-keywords} flag instead, which has the same
1584 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1585 switch only affects the @code{asm} and @code{typeof} keywords, since
1586 @code{inline} is a standard keyword in ISO C99.
1589 @itemx -fno-builtin-@var{function}
1590 @opindex fno-builtin
1591 @cindex built-in functions
1592 Don't recognize built-in functions that do not begin with
1593 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1594 functions provided by GCC}, for details of the functions affected,
1595 including those which are not built-in functions when @option{-ansi} or
1596 @option{-std} options for strict ISO C conformance are used because they
1597 do not have an ISO standard meaning.
1599 GCC normally generates special code to handle certain built-in functions
1600 more efficiently; for instance, calls to @code{alloca} may become single
1601 instructions that adjust the stack directly, and calls to @code{memcpy}
1602 may become inline copy loops. The resulting code is often both smaller
1603 and faster, but since the function calls no longer appear as such, you
1604 cannot set a breakpoint on those calls, nor can you change the behavior
1605 of the functions by linking with a different library. In addition,
1606 when a function is recognized as a built-in function, GCC may use
1607 information about that function to warn about problems with calls to
1608 that function, or to generate more efficient code, even if the
1609 resulting code still contains calls to that function. For example,
1610 warnings are given with @option{-Wformat} for bad calls to
1611 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1612 known not to modify global memory.
1614 With the @option{-fno-builtin-@var{function}} option
1615 only the built-in function @var{function} is
1616 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1617 function is named that is not built-in in this version of GCC, this
1618 option is ignored. There is no corresponding
1619 @option{-fbuiltin-@var{function}} option; if you wish to enable
1620 built-in functions selectively when using @option{-fno-builtin} or
1621 @option{-ffreestanding}, you may define macros such as:
1624 #define abs(n) __builtin_abs ((n))
1625 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1630 @cindex hosted environment
1632 Assert that compilation takes place in a hosted environment. This implies
1633 @option{-fbuiltin}. A hosted environment is one in which the
1634 entire standard library is available, and in which @code{main} has a return
1635 type of @code{int}. Examples are nearly everything except a kernel.
1636 This is equivalent to @option{-fno-freestanding}.
1638 @item -ffreestanding
1639 @opindex ffreestanding
1640 @cindex hosted environment
1642 Assert that compilation takes place in a freestanding environment. This
1643 implies @option{-fno-builtin}. A freestanding environment
1644 is one in which the standard library may not exist, and program startup may
1645 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1646 This is equivalent to @option{-fno-hosted}.
1648 @xref{Standards,,Language Standards Supported by GCC}, for details of
1649 freestanding and hosted environments.
1653 @cindex openmp parallel
1654 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1655 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1656 compiler generates parallel code according to the OpenMP Application
1657 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1658 implies @option{-pthread}, and thus is only supported on targets that
1659 have support for @option{-pthread}.
1661 @item -fms-extensions
1662 @opindex fms-extensions
1663 Accept some non-standard constructs used in Microsoft header files.
1665 Some cases of unnamed fields in structures and unions are only
1666 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1667 fields within structs/unions}, for details.
1671 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1672 options for strict ISO C conformance) implies @option{-trigraphs}.
1674 @item -no-integrated-cpp
1675 @opindex no-integrated-cpp
1676 Performs a compilation in two passes: preprocessing and compiling. This
1677 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1678 @option{-B} option. The user supplied compilation step can then add in
1679 an additional preprocessing step after normal preprocessing but before
1680 compiling. The default is to use the integrated cpp (internal cpp)
1682 The semantics of this option will change if "cc1", "cc1plus", and
1683 "cc1obj" are merged.
1685 @cindex traditional C language
1686 @cindex C language, traditional
1688 @itemx -traditional-cpp
1689 @opindex traditional-cpp
1690 @opindex traditional
1691 Formerly, these options caused GCC to attempt to emulate a pre-standard
1692 C compiler. They are now only supported with the @option{-E} switch.
1693 The preprocessor continues to support a pre-standard mode. See the GNU
1694 CPP manual for details.
1696 @item -fcond-mismatch
1697 @opindex fcond-mismatch
1698 Allow conditional expressions with mismatched types in the second and
1699 third arguments. The value of such an expression is void. This option
1700 is not supported for C++.
1702 @item -flax-vector-conversions
1703 @opindex flax-vector-conversions
1704 Allow implicit conversions between vectors with differing numbers of
1705 elements and/or incompatible element types. This option should not be
1708 @item -funsigned-char
1709 @opindex funsigned-char
1710 Let the type @code{char} be unsigned, like @code{unsigned char}.
1712 Each kind of machine has a default for what @code{char} should
1713 be. It is either like @code{unsigned char} by default or like
1714 @code{signed char} by default.
1716 Ideally, a portable program should always use @code{signed char} or
1717 @code{unsigned char} when it depends on the signedness of an object.
1718 But many programs have been written to use plain @code{char} and
1719 expect it to be signed, or expect it to be unsigned, depending on the
1720 machines they were written for. This option, and its inverse, let you
1721 make such a program work with the opposite default.
1723 The type @code{char} is always a distinct type from each of
1724 @code{signed char} or @code{unsigned char}, even though its behavior
1725 is always just like one of those two.
1728 @opindex fsigned-char
1729 Let the type @code{char} be signed, like @code{signed char}.
1731 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1732 the negative form of @option{-funsigned-char}. Likewise, the option
1733 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1735 @item -fsigned-bitfields
1736 @itemx -funsigned-bitfields
1737 @itemx -fno-signed-bitfields
1738 @itemx -fno-unsigned-bitfields
1739 @opindex fsigned-bitfields
1740 @opindex funsigned-bitfields
1741 @opindex fno-signed-bitfields
1742 @opindex fno-unsigned-bitfields
1743 These options control whether a bit-field is signed or unsigned, when the
1744 declaration does not use either @code{signed} or @code{unsigned}. By
1745 default, such a bit-field is signed, because this is consistent: the
1746 basic integer types such as @code{int} are signed types.
1749 @node C++ Dialect Options
1750 @section Options Controlling C++ Dialect
1752 @cindex compiler options, C++
1753 @cindex C++ options, command line
1754 @cindex options, C++
1755 This section describes the command-line options that are only meaningful
1756 for C++ programs; but you can also use most of the GNU compiler options
1757 regardless of what language your program is in. For example, you
1758 might compile a file @code{firstClass.C} like this:
1761 g++ -g -frepo -O -c firstClass.C
1765 In this example, only @option{-frepo} is an option meant
1766 only for C++ programs; you can use the other options with any
1767 language supported by GCC@.
1769 Here is a list of options that are @emph{only} for compiling C++ programs:
1773 @item -fabi-version=@var{n}
1774 @opindex fabi-version
1775 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1776 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1777 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1778 the version that conforms most closely to the C++ ABI specification.
1779 Therefore, the ABI obtained using version 0 will change as ABI bugs
1782 The default is version 2.
1784 @item -fno-access-control
1785 @opindex fno-access-control
1786 Turn off all access checking. This switch is mainly useful for working
1787 around bugs in the access control code.
1791 Check that the pointer returned by @code{operator new} is non-null
1792 before attempting to modify the storage allocated. This check is
1793 normally unnecessary because the C++ standard specifies that
1794 @code{operator new} will only return @code{0} if it is declared
1795 @samp{throw()}, in which case the compiler will always check the
1796 return value even without this option. In all other cases, when
1797 @code{operator new} has a non-empty exception specification, memory
1798 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1799 @samp{new (nothrow)}.
1801 @item -fconserve-space
1802 @opindex fconserve-space
1803 Put uninitialized or runtime-initialized global variables into the
1804 common segment, as C does. This saves space in the executable at the
1805 cost of not diagnosing duplicate definitions. If you compile with this
1806 flag and your program mysteriously crashes after @code{main()} has
1807 completed, you may have an object that is being destroyed twice because
1808 two definitions were merged.
1810 This option is no longer useful on most targets, now that support has
1811 been added for putting variables into BSS without making them common.
1813 @item -fno-deduce-init-list
1814 @opindex fno-deduce-init-list
1815 Disable deduction of a template type parameter as
1816 std::initializer_list from a brace-enclosed initializer list, i.e.
1819 template <class T> auto forward(T t) -> decltype (realfn (t))
1826 forward(@{1,2@}); // call forward<std::initializer_list<int>>
1830 This option is present because this deduction is an extension to the
1831 current specification in the C++0x working draft, and there was
1832 some concern about potential overload resolution problems.
1834 @item -ffriend-injection
1835 @opindex ffriend-injection
1836 Inject friend functions into the enclosing namespace, so that they are
1837 visible outside the scope of the class in which they are declared.
1838 Friend functions were documented to work this way in the old Annotated
1839 C++ Reference Manual, and versions of G++ before 4.1 always worked
1840 that way. However, in ISO C++ a friend function which is not declared
1841 in an enclosing scope can only be found using argument dependent
1842 lookup. This option causes friends to be injected as they were in
1845 This option is for compatibility, and may be removed in a future
1848 @item -fno-elide-constructors
1849 @opindex fno-elide-constructors
1850 The C++ standard allows an implementation to omit creating a temporary
1851 which is only used to initialize another object of the same type.
1852 Specifying this option disables that optimization, and forces G++ to
1853 call the copy constructor in all cases.
1855 @item -fno-enforce-eh-specs
1856 @opindex fno-enforce-eh-specs
1857 Don't generate code to check for violation of exception specifications
1858 at runtime. This option violates the C++ standard, but may be useful
1859 for reducing code size in production builds, much like defining
1860 @samp{NDEBUG}. This does not give user code permission to throw
1861 exceptions in violation of the exception specifications; the compiler
1862 will still optimize based on the specifications, so throwing an
1863 unexpected exception will result in undefined behavior.
1866 @itemx -fno-for-scope
1868 @opindex fno-for-scope
1869 If @option{-ffor-scope} is specified, the scope of variables declared in
1870 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1871 as specified by the C++ standard.
1872 If @option{-fno-for-scope} is specified, the scope of variables declared in
1873 a @i{for-init-statement} extends to the end of the enclosing scope,
1874 as was the case in old versions of G++, and other (traditional)
1875 implementations of C++.
1877 The default if neither flag is given to follow the standard,
1878 but to allow and give a warning for old-style code that would
1879 otherwise be invalid, or have different behavior.
1881 @item -fno-gnu-keywords
1882 @opindex fno-gnu-keywords
1883 Do not recognize @code{typeof} as a keyword, so that code can use this
1884 word as an identifier. You can use the keyword @code{__typeof__} instead.
1885 @option{-ansi} implies @option{-fno-gnu-keywords}.
1887 @item -fno-implicit-templates
1888 @opindex fno-implicit-templates
1889 Never emit code for non-inline templates which are instantiated
1890 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1891 @xref{Template Instantiation}, for more information.
1893 @item -fno-implicit-inline-templates
1894 @opindex fno-implicit-inline-templates
1895 Don't emit code for implicit instantiations of inline templates, either.
1896 The default is to handle inlines differently so that compiles with and
1897 without optimization will need the same set of explicit instantiations.
1899 @item -fno-implement-inlines
1900 @opindex fno-implement-inlines
1901 To save space, do not emit out-of-line copies of inline functions
1902 controlled by @samp{#pragma implementation}. This will cause linker
1903 errors if these functions are not inlined everywhere they are called.
1905 @item -fms-extensions
1906 @opindex fms-extensions
1907 Disable pedantic warnings about constructs used in MFC, such as implicit
1908 int and getting a pointer to member function via non-standard syntax.
1910 @item -fno-nonansi-builtins
1911 @opindex fno-nonansi-builtins
1912 Disable built-in declarations of functions that are not mandated by
1913 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1914 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1916 @item -fno-operator-names
1917 @opindex fno-operator-names
1918 Do not treat the operator name keywords @code{and}, @code{bitand},
1919 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1920 synonyms as keywords.
1922 @item -fno-optional-diags
1923 @opindex fno-optional-diags
1924 Disable diagnostics that the standard says a compiler does not need to
1925 issue. Currently, the only such diagnostic issued by G++ is the one for
1926 a name having multiple meanings within a class.
1929 @opindex fpermissive
1930 Downgrade some diagnostics about nonconformant code from errors to
1931 warnings. Thus, using @option{-fpermissive} will allow some
1932 nonconforming code to compile.
1934 @item -fno-pretty-templates
1935 @opindex fno-pretty-templates
1936 When an error message refers to a specialization of a function
1937 template, the compiler will normally print the signature of the
1938 template followed by the template arguments and any typedefs or
1939 typenames in the signature (e.g. @code{void f(T) [with T = int]}
1940 rather than @code{void f(int)}) so that it's clear which template is
1941 involved. When an error message refers to a specialization of a class
1942 template, the compiler will omit any template arguments which match
1943 the default template arguments for that template. If either of these
1944 behaviors make it harder to understand the error message rather than
1945 easier, using @option{-fno-pretty-templates} will disable them.
1949 Enable automatic template instantiation at link time. This option also
1950 implies @option{-fno-implicit-templates}. @xref{Template
1951 Instantiation}, for more information.
1955 Disable generation of information about every class with virtual
1956 functions for use by the C++ runtime type identification features
1957 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1958 of the language, you can save some space by using this flag. Note that
1959 exception handling uses the same information, but it will generate it as
1960 needed. The @samp{dynamic_cast} operator can still be used for casts that
1961 do not require runtime type information, i.e.@: casts to @code{void *} or to
1962 unambiguous base classes.
1966 Emit statistics about front-end processing at the end of the compilation.
1967 This information is generally only useful to the G++ development team.
1969 @item -ftemplate-depth-@var{n}
1970 @opindex ftemplate-depth
1971 Set the maximum instantiation depth for template classes to @var{n}.
1972 A limit on the template instantiation depth is needed to detect
1973 endless recursions during template class instantiation. ANSI/ISO C++
1974 conforming programs must not rely on a maximum depth greater than 17
1975 (changed to 1024 in C++0x).
1977 @item -fno-threadsafe-statics
1978 @opindex fno-threadsafe-statics
1979 Do not emit the extra code to use the routines specified in the C++
1980 ABI for thread-safe initialization of local statics. You can use this
1981 option to reduce code size slightly in code that doesn't need to be
1984 @item -fuse-cxa-atexit
1985 @opindex fuse-cxa-atexit
1986 Register destructors for objects with static storage duration with the
1987 @code{__cxa_atexit} function rather than the @code{atexit} function.
1988 This option is required for fully standards-compliant handling of static
1989 destructors, but will only work if your C library supports
1990 @code{__cxa_atexit}.
1992 @item -fno-use-cxa-get-exception-ptr
1993 @opindex fno-use-cxa-get-exception-ptr
1994 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
1995 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
1996 if the runtime routine is not available.
1998 @item -fvisibility-inlines-hidden
1999 @opindex fvisibility-inlines-hidden
2000 This switch declares that the user does not attempt to compare
2001 pointers to inline methods where the addresses of the two functions
2002 were taken in different shared objects.
2004 The effect of this is that GCC may, effectively, mark inline methods with
2005 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2006 appear in the export table of a DSO and do not require a PLT indirection
2007 when used within the DSO@. Enabling this option can have a dramatic effect
2008 on load and link times of a DSO as it massively reduces the size of the
2009 dynamic export table when the library makes heavy use of templates.
2011 The behavior of this switch is not quite the same as marking the
2012 methods as hidden directly, because it does not affect static variables
2013 local to the function or cause the compiler to deduce that
2014 the function is defined in only one shared object.
2016 You may mark a method as having a visibility explicitly to negate the
2017 effect of the switch for that method. For example, if you do want to
2018 compare pointers to a particular inline method, you might mark it as
2019 having default visibility. Marking the enclosing class with explicit
2020 visibility will have no effect.
2022 Explicitly instantiated inline methods are unaffected by this option
2023 as their linkage might otherwise cross a shared library boundary.
2024 @xref{Template Instantiation}.
2026 @item -fvisibility-ms-compat
2027 @opindex fvisibility-ms-compat
2028 This flag attempts to use visibility settings to make GCC's C++
2029 linkage model compatible with that of Microsoft Visual Studio.
2031 The flag makes these changes to GCC's linkage model:
2035 It sets the default visibility to @code{hidden}, like
2036 @option{-fvisibility=hidden}.
2039 Types, but not their members, are not hidden by default.
2042 The One Definition Rule is relaxed for types without explicit
2043 visibility specifications which are defined in more than one different
2044 shared object: those declarations are permitted if they would have
2045 been permitted when this option was not used.
2048 In new code it is better to use @option{-fvisibility=hidden} and
2049 export those classes which are intended to be externally visible.
2050 Unfortunately it is possible for code to rely, perhaps accidentally,
2051 on the Visual Studio behavior.
2053 Among the consequences of these changes are that static data members
2054 of the same type with the same name but defined in different shared
2055 objects will be different, so changing one will not change the other;
2056 and that pointers to function members defined in different shared
2057 objects may not compare equal. When this flag is given, it is a
2058 violation of the ODR to define types with the same name differently.
2062 Do not use weak symbol support, even if it is provided by the linker.
2063 By default, G++ will use weak symbols if they are available. This
2064 option exists only for testing, and should not be used by end-users;
2065 it will result in inferior code and has no benefits. This option may
2066 be removed in a future release of G++.
2070 Do not search for header files in the standard directories specific to
2071 C++, but do still search the other standard directories. (This option
2072 is used when building the C++ library.)
2075 In addition, these optimization, warning, and code generation options
2076 have meanings only for C++ programs:
2079 @item -fno-default-inline
2080 @opindex fno-default-inline
2081 Do not assume @samp{inline} for functions defined inside a class scope.
2082 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2083 functions will have linkage like inline functions; they just won't be
2086 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2089 Warn when G++ generates code that is probably not compatible with the
2090 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2091 all such cases, there are probably some cases that are not warned about,
2092 even though G++ is generating incompatible code. There may also be
2093 cases where warnings are emitted even though the code that is generated
2096 You should rewrite your code to avoid these warnings if you are
2097 concerned about the fact that code generated by G++ may not be binary
2098 compatible with code generated by other compilers.
2100 The known incompatibilities at this point include:
2105 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2106 pack data into the same byte as a base class. For example:
2109 struct A @{ virtual void f(); int f1 : 1; @};
2110 struct B : public A @{ int f2 : 1; @};
2114 In this case, G++ will place @code{B::f2} into the same byte
2115 as@code{A::f1}; other compilers will not. You can avoid this problem
2116 by explicitly padding @code{A} so that its size is a multiple of the
2117 byte size on your platform; that will cause G++ and other compilers to
2118 layout @code{B} identically.
2121 Incorrect handling of tail-padding for virtual bases. G++ does not use
2122 tail padding when laying out virtual bases. For example:
2125 struct A @{ virtual void f(); char c1; @};
2126 struct B @{ B(); char c2; @};
2127 struct C : public A, public virtual B @{@};
2131 In this case, G++ will not place @code{B} into the tail-padding for
2132 @code{A}; other compilers will. You can avoid this problem by
2133 explicitly padding @code{A} so that its size is a multiple of its
2134 alignment (ignoring virtual base classes); that will cause G++ and other
2135 compilers to layout @code{C} identically.
2138 Incorrect handling of bit-fields with declared widths greater than that
2139 of their underlying types, when the bit-fields appear in a union. For
2143 union U @{ int i : 4096; @};
2147 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2148 union too small by the number of bits in an @code{int}.
2151 Empty classes can be placed at incorrect offsets. For example:
2161 struct C : public B, public A @{@};
2165 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2166 it should be placed at offset zero. G++ mistakenly believes that the
2167 @code{A} data member of @code{B} is already at offset zero.
2170 Names of template functions whose types involve @code{typename} or
2171 template template parameters can be mangled incorrectly.
2174 template <typename Q>
2175 void f(typename Q::X) @{@}
2177 template <template <typename> class Q>
2178 void f(typename Q<int>::X) @{@}
2182 Instantiations of these templates may be mangled incorrectly.
2186 It also warns psABI related changes. The known psABI changes at this
2192 For SYSV/x86-64, when passing union with long double, it is changed to
2193 pass in memory as specified in psABI. For example:
2203 @code{union U} will always be passed in memory.
2207 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2208 @opindex Wctor-dtor-privacy
2209 @opindex Wno-ctor-dtor-privacy
2210 Warn when a class seems unusable because all the constructors or
2211 destructors in that class are private, and it has neither friends nor
2212 public static member functions.
2214 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2215 @opindex Wnon-virtual-dtor
2216 @opindex Wno-non-virtual-dtor
2217 Warn when a class has virtual functions and accessible non-virtual
2218 destructor, in which case it would be possible but unsafe to delete
2219 an instance of a derived class through a pointer to the base class.
2220 This warning is also enabled if -Weffc++ is specified.
2222 @item -Wreorder @r{(C++ and Objective-C++ only)}
2224 @opindex Wno-reorder
2225 @cindex reordering, warning
2226 @cindex warning for reordering of member initializers
2227 Warn when the order of member initializers given in the code does not
2228 match the order in which they must be executed. For instance:
2234 A(): j (0), i (1) @{ @}
2238 The compiler will rearrange the member initializers for @samp{i}
2239 and @samp{j} to match the declaration order of the members, emitting
2240 a warning to that effect. This warning is enabled by @option{-Wall}.
2243 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2246 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2249 Warn about violations of the following style guidelines from Scott Meyers'
2250 @cite{Effective C++} book:
2254 Item 11: Define a copy constructor and an assignment operator for classes
2255 with dynamically allocated memory.
2258 Item 12: Prefer initialization to assignment in constructors.
2261 Item 14: Make destructors virtual in base classes.
2264 Item 15: Have @code{operator=} return a reference to @code{*this}.
2267 Item 23: Don't try to return a reference when you must return an object.
2271 Also warn about violations of the following style guidelines from
2272 Scott Meyers' @cite{More Effective C++} book:
2276 Item 6: Distinguish between prefix and postfix forms of increment and
2277 decrement operators.
2280 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2284 When selecting this option, be aware that the standard library
2285 headers do not obey all of these guidelines; use @samp{grep -v}
2286 to filter out those warnings.
2288 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2289 @opindex Wstrict-null-sentinel
2290 @opindex Wno-strict-null-sentinel
2291 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2292 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2293 to @code{__null}. Although it is a null pointer constant not a null pointer,
2294 it is guaranteed to be of the same size as a pointer. But this use is
2295 not portable across different compilers.
2297 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2298 @opindex Wno-non-template-friend
2299 @opindex Wnon-template-friend
2300 Disable warnings when non-templatized friend functions are declared
2301 within a template. Since the advent of explicit template specification
2302 support in G++, if the name of the friend is an unqualified-id (i.e.,
2303 @samp{friend foo(int)}), the C++ language specification demands that the
2304 friend declare or define an ordinary, nontemplate function. (Section
2305 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2306 could be interpreted as a particular specialization of a templatized
2307 function. Because this non-conforming behavior is no longer the default
2308 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2309 check existing code for potential trouble spots and is on by default.
2310 This new compiler behavior can be turned off with
2311 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2312 but disables the helpful warning.
2314 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2315 @opindex Wold-style-cast
2316 @opindex Wno-old-style-cast
2317 Warn if an old-style (C-style) cast to a non-void type is used within
2318 a C++ program. The new-style casts (@samp{dynamic_cast},
2319 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2320 less vulnerable to unintended effects and much easier to search for.
2322 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2323 @opindex Woverloaded-virtual
2324 @opindex Wno-overloaded-virtual
2325 @cindex overloaded virtual fn, warning
2326 @cindex warning for overloaded virtual fn
2327 Warn when a function declaration hides virtual functions from a
2328 base class. For example, in:
2335 struct B: public A @{
2340 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2348 will fail to compile.
2350 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2351 @opindex Wno-pmf-conversions
2352 @opindex Wpmf-conversions
2353 Disable the diagnostic for converting a bound pointer to member function
2356 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2357 @opindex Wsign-promo
2358 @opindex Wno-sign-promo
2359 Warn when overload resolution chooses a promotion from unsigned or
2360 enumerated type to a signed type, over a conversion to an unsigned type of
2361 the same size. Previous versions of G++ would try to preserve
2362 unsignedness, but the standard mandates the current behavior.
2367 A& operator = (int);
2377 In this example, G++ will synthesize a default @samp{A& operator =
2378 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2381 @node Objective-C and Objective-C++ Dialect Options
2382 @section Options Controlling Objective-C and Objective-C++ Dialects
2384 @cindex compiler options, Objective-C and Objective-C++
2385 @cindex Objective-C and Objective-C++ options, command line
2386 @cindex options, Objective-C and Objective-C++
2387 (NOTE: This manual does not describe the Objective-C and Objective-C++
2388 languages themselves. See @xref{Standards,,Language Standards
2389 Supported by GCC}, for references.)
2391 This section describes the command-line options that are only meaningful
2392 for Objective-C and Objective-C++ programs, but you can also use most of
2393 the language-independent GNU compiler options.
2394 For example, you might compile a file @code{some_class.m} like this:
2397 gcc -g -fgnu-runtime -O -c some_class.m
2401 In this example, @option{-fgnu-runtime} is an option meant only for
2402 Objective-C and Objective-C++ programs; you can use the other options with
2403 any language supported by GCC@.
2405 Note that since Objective-C is an extension of the C language, Objective-C
2406 compilations may also use options specific to the C front-end (e.g.,
2407 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2408 C++-specific options (e.g., @option{-Wabi}).
2410 Here is a list of options that are @emph{only} for compiling Objective-C
2411 and Objective-C++ programs:
2414 @item -fconstant-string-class=@var{class-name}
2415 @opindex fconstant-string-class
2416 Use @var{class-name} as the name of the class to instantiate for each
2417 literal string specified with the syntax @code{@@"@dots{}"}. The default
2418 class name is @code{NXConstantString} if the GNU runtime is being used, and
2419 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2420 @option{-fconstant-cfstrings} option, if also present, will override the
2421 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2422 to be laid out as constant CoreFoundation strings.
2425 @opindex fgnu-runtime
2426 Generate object code compatible with the standard GNU Objective-C
2427 runtime. This is the default for most types of systems.
2429 @item -fnext-runtime
2430 @opindex fnext-runtime
2431 Generate output compatible with the NeXT runtime. This is the default
2432 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2433 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2436 @item -fno-nil-receivers
2437 @opindex fno-nil-receivers
2438 Assume that all Objective-C message dispatches (e.g.,
2439 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
2440 is not @code{nil}. This allows for more efficient entry points in the runtime
2441 to be used. Currently, this option is only available in conjunction with
2442 the NeXT runtime on Mac OS X 10.3 and later.
2444 @item -fobjc-call-cxx-cdtors
2445 @opindex fobjc-call-cxx-cdtors
2446 For each Objective-C class, check if any of its instance variables is a
2447 C++ object with a non-trivial default constructor. If so, synthesize a
2448 special @code{- (id) .cxx_construct} instance method that will run
2449 non-trivial default constructors on any such instance variables, in order,
2450 and then return @code{self}. Similarly, check if any instance variable
2451 is a C++ object with a non-trivial destructor, and if so, synthesize a
2452 special @code{- (void) .cxx_destruct} method that will run
2453 all such default destructors, in reverse order.
2455 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
2456 thusly generated will only operate on instance variables declared in the
2457 current Objective-C class, and not those inherited from superclasses. It
2458 is the responsibility of the Objective-C runtime to invoke all such methods
2459 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
2460 will be invoked by the runtime immediately after a new object
2461 instance is allocated; the @code{- (void) .cxx_destruct} methods will
2462 be invoked immediately before the runtime deallocates an object instance.
2464 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2465 support for invoking the @code{- (id) .cxx_construct} and
2466 @code{- (void) .cxx_destruct} methods.
2468 @item -fobjc-direct-dispatch
2469 @opindex fobjc-direct-dispatch
2470 Allow fast jumps to the message dispatcher. On Darwin this is
2471 accomplished via the comm page.
2473 @item -fobjc-exceptions
2474 @opindex fobjc-exceptions
2475 Enable syntactic support for structured exception handling in Objective-C,
2476 similar to what is offered by C++ and Java. This option is
2477 unavailable in conjunction with the NeXT runtime on Mac OS X 10.2 and
2486 @@catch (AnObjCClass *exc) @{
2493 @@catch (AnotherClass *exc) @{
2496 @@catch (id allOthers) @{
2506 The @code{@@throw} statement may appear anywhere in an Objective-C or
2507 Objective-C++ program; when used inside of a @code{@@catch} block, the
2508 @code{@@throw} may appear without an argument (as shown above), in which case
2509 the object caught by the @code{@@catch} will be rethrown.
2511 Note that only (pointers to) Objective-C objects may be thrown and
2512 caught using this scheme. When an object is thrown, it will be caught
2513 by the nearest @code{@@catch} clause capable of handling objects of that type,
2514 analogously to how @code{catch} blocks work in C++ and Java. A
2515 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2516 any and all Objective-C exceptions not caught by previous @code{@@catch}
2519 The @code{@@finally} clause, if present, will be executed upon exit from the
2520 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2521 regardless of whether any exceptions are thrown, caught or rethrown
2522 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2523 of the @code{finally} clause in Java.
2525 There are several caveats to using the new exception mechanism:
2529 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2530 idioms provided by the @code{NSException} class, the new
2531 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2532 systems, due to additional functionality needed in the (NeXT) Objective-C
2536 As mentioned above, the new exceptions do not support handling
2537 types other than Objective-C objects. Furthermore, when used from
2538 Objective-C++, the Objective-C exception model does not interoperate with C++
2539 exceptions at this time. This means you cannot @code{@@throw} an exception
2540 from Objective-C and @code{catch} it in C++, or vice versa
2541 (i.e., @code{throw @dots{} @@catch}).
2544 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2545 blocks for thread-safe execution:
2548 @@synchronized (ObjCClass *guard) @{
2553 Upon entering the @code{@@synchronized} block, a thread of execution shall
2554 first check whether a lock has been placed on the corresponding @code{guard}
2555 object by another thread. If it has, the current thread shall wait until
2556 the other thread relinquishes its lock. Once @code{guard} becomes available,
2557 the current thread will place its own lock on it, execute the code contained in
2558 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2559 making @code{guard} available to other threads).
2561 Unlike Java, Objective-C does not allow for entire methods to be marked
2562 @code{@@synchronized}. Note that throwing exceptions out of
2563 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2564 to be unlocked properly.
2568 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2570 @item -freplace-objc-classes
2571 @opindex freplace-objc-classes
2572 Emit a special marker instructing @command{ld(1)} not to statically link in
2573 the resulting object file, and allow @command{dyld(1)} to load it in at
2574 run time instead. This is used in conjunction with the Fix-and-Continue
2575 debugging mode, where the object file in question may be recompiled and
2576 dynamically reloaded in the course of program execution, without the need
2577 to restart the program itself. Currently, Fix-and-Continue functionality
2578 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2583 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2584 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2585 compile time) with static class references that get initialized at load time,
2586 which improves run-time performance. Specifying the @option{-fzero-link} flag
2587 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2588 to be retained. This is useful in Zero-Link debugging mode, since it allows
2589 for individual class implementations to be modified during program execution.
2593 Dump interface declarations for all classes seen in the source file to a
2594 file named @file{@var{sourcename}.decl}.
2596 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2597 @opindex Wassign-intercept
2598 @opindex Wno-assign-intercept
2599 Warn whenever an Objective-C assignment is being intercepted by the
2602 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2603 @opindex Wno-protocol
2605 If a class is declared to implement a protocol, a warning is issued for
2606 every method in the protocol that is not implemented by the class. The
2607 default behavior is to issue a warning for every method not explicitly
2608 implemented in the class, even if a method implementation is inherited
2609 from the superclass. If you use the @option{-Wno-protocol} option, then
2610 methods inherited from the superclass are considered to be implemented,
2611 and no warning is issued for them.
2613 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2615 @opindex Wno-selector
2616 Warn if multiple methods of different types for the same selector are
2617 found during compilation. The check is performed on the list of methods
2618 in the final stage of compilation. Additionally, a check is performed
2619 for each selector appearing in a @code{@@selector(@dots{})}
2620 expression, and a corresponding method for that selector has been found
2621 during compilation. Because these checks scan the method table only at
2622 the end of compilation, these warnings are not produced if the final
2623 stage of compilation is not reached, for example because an error is
2624 found during compilation, or because the @option{-fsyntax-only} option is
2627 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2628 @opindex Wstrict-selector-match
2629 @opindex Wno-strict-selector-match
2630 Warn if multiple methods with differing argument and/or return types are
2631 found for a given selector when attempting to send a message using this
2632 selector to a receiver of type @code{id} or @code{Class}. When this flag
2633 is off (which is the default behavior), the compiler will omit such warnings
2634 if any differences found are confined to types which share the same size
2637 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2638 @opindex Wundeclared-selector
2639 @opindex Wno-undeclared-selector
2640 Warn if a @code{@@selector(@dots{})} expression referring to an
2641 undeclared selector is found. A selector is considered undeclared if no
2642 method with that name has been declared before the
2643 @code{@@selector(@dots{})} expression, either explicitly in an
2644 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2645 an @code{@@implementation} section. This option always performs its
2646 checks as soon as a @code{@@selector(@dots{})} expression is found,
2647 while @option{-Wselector} only performs its checks in the final stage of
2648 compilation. This also enforces the coding style convention
2649 that methods and selectors must be declared before being used.
2651 @item -print-objc-runtime-info
2652 @opindex print-objc-runtime-info
2653 Generate C header describing the largest structure that is passed by
2658 @node Language Independent Options
2659 @section Options to Control Diagnostic Messages Formatting
2660 @cindex options to control diagnostics formatting
2661 @cindex diagnostic messages
2662 @cindex message formatting
2664 Traditionally, diagnostic messages have been formatted irrespective of
2665 the output device's aspect (e.g.@: its width, @dots{}). The options described
2666 below can be used to control the diagnostic messages formatting
2667 algorithm, e.g.@: how many characters per line, how often source location
2668 information should be reported. Right now, only the C++ front end can
2669 honor these options. However it is expected, in the near future, that
2670 the remaining front ends would be able to digest them correctly.
2673 @item -fmessage-length=@var{n}
2674 @opindex fmessage-length
2675 Try to format error messages so that they fit on lines of about @var{n}
2676 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2677 the front ends supported by GCC@. If @var{n} is zero, then no
2678 line-wrapping will be done; each error message will appear on a single
2681 @opindex fdiagnostics-show-location
2682 @item -fdiagnostics-show-location=once
2683 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2684 reporter to emit @emph{once} source location information; that is, in
2685 case the message is too long to fit on a single physical line and has to
2686 be wrapped, the source location won't be emitted (as prefix) again,
2687 over and over, in subsequent continuation lines. This is the default
2690 @item -fdiagnostics-show-location=every-line
2691 Only meaningful in line-wrapping mode. Instructs the diagnostic
2692 messages reporter to emit the same source location information (as
2693 prefix) for physical lines that result from the process of breaking
2694 a message which is too long to fit on a single line.
2696 @item -fdiagnostics-show-option
2697 @opindex fdiagnostics-show-option
2698 This option instructs the diagnostic machinery to add text to each
2699 diagnostic emitted, which indicates which command line option directly
2700 controls that diagnostic, when such an option is known to the
2701 diagnostic machinery.
2703 @item -Wcoverage-mismatch
2704 @opindex Wcoverage-mismatch
2705 Warn if feedback profiles do not match when using the
2706 @option{-fprofile-use} option.
2707 If a source file was changed between @option{-fprofile-gen} and
2708 @option{-fprofile-use}, the files with the profile feedback can fail
2709 to match the source file and GCC can not use the profile feedback
2710 information. By default, GCC emits an error message in this case.
2711 The option @option{-Wcoverage-mismatch} emits a warning instead of an
2712 error. GCC does not use appropriate feedback profiles, so using this
2713 option can result in poorly optimized code. This option is useful
2714 only in the case of very minor changes such as bug fixes to an
2719 @node Warning Options
2720 @section Options to Request or Suppress Warnings
2721 @cindex options to control warnings
2722 @cindex warning messages
2723 @cindex messages, warning
2724 @cindex suppressing warnings
2726 Warnings are diagnostic messages that report constructions which
2727 are not inherently erroneous but which are risky or suggest there
2728 may have been an error.
2730 The following language-independent options do not enable specific
2731 warnings but control the kinds of diagnostics produced by GCC.
2734 @cindex syntax checking
2736 @opindex fsyntax-only
2737 Check the code for syntax errors, but don't do anything beyond that.
2741 Inhibit all warning messages.
2746 Make all warnings into errors.
2751 Make the specified warning into an error. The specifier for a warning
2752 is appended, for example @option{-Werror=switch} turns the warnings
2753 controlled by @option{-Wswitch} into errors. This switch takes a
2754 negative form, to be used to negate @option{-Werror} for specific
2755 warnings, for example @option{-Wno-error=switch} makes
2756 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2757 is in effect. You can use the @option{-fdiagnostics-show-option}
2758 option to have each controllable warning amended with the option which
2759 controls it, to determine what to use with this option.
2761 Note that specifying @option{-Werror=}@var{foo} automatically implies
2762 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2765 @item -Wfatal-errors
2766 @opindex Wfatal-errors
2767 @opindex Wno-fatal-errors
2768 This option causes the compiler to abort compilation on the first error
2769 occurred rather than trying to keep going and printing further error
2774 You can request many specific warnings with options beginning
2775 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2776 implicit declarations. Each of these specific warning options also
2777 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2778 example, @option{-Wno-implicit}. This manual lists only one of the
2779 two forms, whichever is not the default. For further,
2780 language-specific options also refer to @ref{C++ Dialect Options} and
2781 @ref{Objective-C and Objective-C++ Dialect Options}.
2786 Issue all the warnings demanded by strict ISO C and ISO C++;
2787 reject all programs that use forbidden extensions, and some other
2788 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2789 version of the ISO C standard specified by any @option{-std} option used.
2791 Valid ISO C and ISO C++ programs should compile properly with or without
2792 this option (though a rare few will require @option{-ansi} or a
2793 @option{-std} option specifying the required version of ISO C)@. However,
2794 without this option, certain GNU extensions and traditional C and C++
2795 features are supported as well. With this option, they are rejected.
2797 @option{-pedantic} does not cause warning messages for use of the
2798 alternate keywords whose names begin and end with @samp{__}. Pedantic
2799 warnings are also disabled in the expression that follows
2800 @code{__extension__}. However, only system header files should use
2801 these escape routes; application programs should avoid them.
2802 @xref{Alternate Keywords}.
2804 Some users try to use @option{-pedantic} to check programs for strict ISO
2805 C conformance. They soon find that it does not do quite what they want:
2806 it finds some non-ISO practices, but not all---only those for which
2807 ISO C @emph{requires} a diagnostic, and some others for which
2808 diagnostics have been added.
2810 A feature to report any failure to conform to ISO C might be useful in
2811 some instances, but would require considerable additional work and would
2812 be quite different from @option{-pedantic}. We don't have plans to
2813 support such a feature in the near future.
2815 Where the standard specified with @option{-std} represents a GNU
2816 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2817 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2818 extended dialect is based. Warnings from @option{-pedantic} are given
2819 where they are required by the base standard. (It would not make sense
2820 for such warnings to be given only for features not in the specified GNU
2821 C dialect, since by definition the GNU dialects of C include all
2822 features the compiler supports with the given option, and there would be
2823 nothing to warn about.)
2825 @item -pedantic-errors
2826 @opindex pedantic-errors
2827 Like @option{-pedantic}, except that errors are produced rather than
2833 This enables all the warnings about constructions that some users
2834 consider questionable, and that are easy to avoid (or modify to
2835 prevent the warning), even in conjunction with macros. This also
2836 enables some language-specific warnings described in @ref{C++ Dialect
2837 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2839 @option{-Wall} turns on the following warning flags:
2841 @gccoptlist{-Waddress @gol
2842 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2844 -Wchar-subscripts @gol
2845 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
2847 -Wimplicit-function-declaration @gol
2850 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2851 -Wmissing-braces @gol
2857 -Wsequence-point @gol
2858 -Wsign-compare @r{(only in C++)} @gol
2859 -Wstrict-aliasing @gol
2860 -Wstrict-overflow=1 @gol
2863 -Wuninitialized @gol
2864 -Wunknown-pragmas @gol
2865 -Wunused-function @gol
2868 -Wunused-variable @gol
2869 -Wvolatile-register-var @gol
2872 Note that some warning flags are not implied by @option{-Wall}. Some of
2873 them warn about constructions that users generally do not consider
2874 questionable, but which occasionally you might wish to check for;
2875 others warn about constructions that are necessary or hard to avoid in
2876 some cases, and there is no simple way to modify the code to suppress
2877 the warning. Some of them are enabled by @option{-Wextra} but many of
2878 them must be enabled individually.
2884 This enables some extra warning flags that are not enabled by
2885 @option{-Wall}. (This option used to be called @option{-W}. The older
2886 name is still supported, but the newer name is more descriptive.)
2888 @gccoptlist{-Wclobbered @gol
2890 -Wignored-qualifiers @gol
2891 -Wmissing-field-initializers @gol
2892 -Wmissing-parameter-type @r{(C only)} @gol
2893 -Wold-style-declaration @r{(C only)} @gol
2894 -Woverride-init @gol
2897 -Wuninitialized @gol
2898 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2901 The option @option{-Wextra} also prints warning messages for the
2907 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2908 @samp{>}, or @samp{>=}.
2911 (C++ only) An enumerator and a non-enumerator both appear in a
2912 conditional expression.
2915 (C++ only) Ambiguous virtual bases.
2918 (C++ only) Subscripting an array which has been declared @samp{register}.
2921 (C++ only) Taking the address of a variable which has been declared
2925 (C++ only) A base class is not initialized in a derived class' copy
2930 @item -Wchar-subscripts
2931 @opindex Wchar-subscripts
2932 @opindex Wno-char-subscripts
2933 Warn if an array subscript has type @code{char}. This is a common cause
2934 of error, as programmers often forget that this type is signed on some
2936 This warning is enabled by @option{-Wall}.
2940 @opindex Wno-comment
2941 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2942 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2943 This warning is enabled by @option{-Wall}.
2948 @opindex ffreestanding
2949 @opindex fno-builtin
2950 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2951 the arguments supplied have types appropriate to the format string
2952 specified, and that the conversions specified in the format string make
2953 sense. This includes standard functions, and others specified by format
2954 attributes (@pxref{Function Attributes}), in the @code{printf},
2955 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2956 not in the C standard) families (or other target-specific families).
2957 Which functions are checked without format attributes having been
2958 specified depends on the standard version selected, and such checks of
2959 functions without the attribute specified are disabled by
2960 @option{-ffreestanding} or @option{-fno-builtin}.
2962 The formats are checked against the format features supported by GNU
2963 libc version 2.2. These include all ISO C90 and C99 features, as well
2964 as features from the Single Unix Specification and some BSD and GNU
2965 extensions. Other library implementations may not support all these
2966 features; GCC does not support warning about features that go beyond a
2967 particular library's limitations. However, if @option{-pedantic} is used
2968 with @option{-Wformat}, warnings will be given about format features not
2969 in the selected standard version (but not for @code{strfmon} formats,
2970 since those are not in any version of the C standard). @xref{C Dialect
2971 Options,,Options Controlling C Dialect}.
2973 Since @option{-Wformat} also checks for null format arguments for
2974 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2976 @option{-Wformat} is included in @option{-Wall}. For more control over some
2977 aspects of format checking, the options @option{-Wformat-y2k},
2978 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2979 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2980 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2983 @opindex Wformat-y2k
2984 @opindex Wno-format-y2k
2985 If @option{-Wformat} is specified, also warn about @code{strftime}
2986 formats which may yield only a two-digit year.
2988 @item -Wno-format-contains-nul
2989 @opindex Wno-format-contains-nul
2990 @opindex Wformat-contains-nul
2991 If @option{-Wformat} is specified, do not warn about format strings that
2994 @item -Wno-format-extra-args
2995 @opindex Wno-format-extra-args
2996 @opindex Wformat-extra-args
2997 If @option{-Wformat} is specified, do not warn about excess arguments to a
2998 @code{printf} or @code{scanf} format function. The C standard specifies
2999 that such arguments are ignored.
3001 Where the unused arguments lie between used arguments that are
3002 specified with @samp{$} operand number specifications, normally
3003 warnings are still given, since the implementation could not know what
3004 type to pass to @code{va_arg} to skip the unused arguments. However,
3005 in the case of @code{scanf} formats, this option will suppress the
3006 warning if the unused arguments are all pointers, since the Single
3007 Unix Specification says that such unused arguments are allowed.
3009 @item -Wno-format-zero-length @r{(C and Objective-C only)}
3010 @opindex Wno-format-zero-length
3011 @opindex Wformat-zero-length
3012 If @option{-Wformat} is specified, do not warn about zero-length formats.
3013 The C standard specifies that zero-length formats are allowed.
3015 @item -Wformat-nonliteral
3016 @opindex Wformat-nonliteral
3017 @opindex Wno-format-nonliteral
3018 If @option{-Wformat} is specified, also warn if the format string is not a
3019 string literal and so cannot be checked, unless the format function
3020 takes its format arguments as a @code{va_list}.
3022 @item -Wformat-security
3023 @opindex Wformat-security
3024 @opindex Wno-format-security
3025 If @option{-Wformat} is specified, also warn about uses of format
3026 functions that represent possible security problems. At present, this
3027 warns about calls to @code{printf} and @code{scanf} functions where the
3028 format string is not a string literal and there are no format arguments,
3029 as in @code{printf (foo);}. This may be a security hole if the format
3030 string came from untrusted input and contains @samp{%n}. (This is
3031 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3032 in future warnings may be added to @option{-Wformat-security} that are not
3033 included in @option{-Wformat-nonliteral}.)
3037 @opindex Wno-format=2
3038 Enable @option{-Wformat} plus format checks not included in
3039 @option{-Wformat}. Currently equivalent to @samp{-Wformat
3040 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
3042 @item -Wnonnull @r{(C and Objective-C only)}
3044 @opindex Wno-nonnull
3045 Warn about passing a null pointer for arguments marked as
3046 requiring a non-null value by the @code{nonnull} function attribute.
3048 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3049 can be disabled with the @option{-Wno-nonnull} option.
3051 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3053 @opindex Wno-init-self
3054 Warn about uninitialized variables which are initialized with themselves.
3055 Note this option can only be used with the @option{-Wuninitialized} option.
3057 For example, GCC will warn about @code{i} being uninitialized in the
3058 following snippet only when @option{-Winit-self} has been specified:
3069 @item -Wimplicit-int @r{(C and Objective-C only)}
3070 @opindex Wimplicit-int
3071 @opindex Wno-implicit-int
3072 Warn when a declaration does not specify a type.
3073 This warning is enabled by @option{-Wall}.
3075 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3076 @opindex Wimplicit-function-declaration
3077 @opindex Wno-implicit-function-declaration
3078 Give a warning whenever a function is used before being declared. In
3079 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3080 enabled by default and it is made into an error by
3081 @option{-pedantic-errors}. This warning is also enabled by
3086 @opindex Wno-implicit
3087 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3088 This warning is enabled by @option{-Wall}.
3090 @item -Wignored-qualifiers @r{(C and C++ only)}
3091 @opindex Wignored-qualifiers
3092 @opindex Wno-ignored-qualifiers
3093 Warn if the return type of a function has a type qualifier
3094 such as @code{const}. For ISO C such a type qualifier has no effect,
3095 since the value returned by a function is not an lvalue.
3096 For C++, the warning is only emitted for scalar types or @code{void}.
3097 ISO C prohibits qualified @code{void} return types on function
3098 definitions, so such return types always receive a warning
3099 even without this option.
3101 This warning is also enabled by @option{-Wextra}.
3106 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3107 a function with external linkage, returning int, taking either zero
3108 arguments, two, or three arguments of appropriate types. This warning
3109 is enabled by default in C++ and is enabled by either @option{-Wall}
3110 or @option{-pedantic}.
3112 @item -Wmissing-braces
3113 @opindex Wmissing-braces
3114 @opindex Wno-missing-braces
3115 Warn if an aggregate or union initializer is not fully bracketed. In
3116 the following example, the initializer for @samp{a} is not fully
3117 bracketed, but that for @samp{b} is fully bracketed.
3120 int a[2][2] = @{ 0, 1, 2, 3 @};
3121 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3124 This warning is enabled by @option{-Wall}.
3126 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3127 @opindex Wmissing-include-dirs
3128 @opindex Wno-missing-include-dirs
3129 Warn if a user-supplied include directory does not exist.
3132 @opindex Wparentheses
3133 @opindex Wno-parentheses
3134 Warn if parentheses are omitted in certain contexts, such
3135 as when there is an assignment in a context where a truth value
3136 is expected, or when operators are nested whose precedence people
3137 often get confused about.
3139 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3140 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3141 interpretation from that of ordinary mathematical notation.
3143 Also warn about constructions where there may be confusion to which
3144 @code{if} statement an @code{else} branch belongs. Here is an example of
3159 In C/C++, every @code{else} branch belongs to the innermost possible
3160 @code{if} statement, which in this example is @code{if (b)}. This is
3161 often not what the programmer expected, as illustrated in the above
3162 example by indentation the programmer chose. When there is the
3163 potential for this confusion, GCC will issue a warning when this flag
3164 is specified. To eliminate the warning, add explicit braces around
3165 the innermost @code{if} statement so there is no way the @code{else}
3166 could belong to the enclosing @code{if}. The resulting code would
3183 This warning is enabled by @option{-Wall}.
3185 @item -Wsequence-point
3186 @opindex Wsequence-point
3187 @opindex Wno-sequence-point
3188 Warn about code that may have undefined semantics because of violations
3189 of sequence point rules in the C and C++ standards.
3191 The C and C++ standards defines the order in which expressions in a C/C++
3192 program are evaluated in terms of @dfn{sequence points}, which represent
3193 a partial ordering between the execution of parts of the program: those
3194 executed before the sequence point, and those executed after it. These
3195 occur after the evaluation of a full expression (one which is not part
3196 of a larger expression), after the evaluation of the first operand of a
3197 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3198 function is called (but after the evaluation of its arguments and the
3199 expression denoting the called function), and in certain other places.
3200 Other than as expressed by the sequence point rules, the order of
3201 evaluation of subexpressions of an expression is not specified. All
3202 these rules describe only a partial order rather than a total order,
3203 since, for example, if two functions are called within one expression
3204 with no sequence point between them, the order in which the functions
3205 are called is not specified. However, the standards committee have
3206 ruled that function calls do not overlap.
3208 It is not specified when between sequence points modifications to the
3209 values of objects take effect. Programs whose behavior depends on this
3210 have undefined behavior; the C and C++ standards specify that ``Between
3211 the previous and next sequence point an object shall have its stored
3212 value modified at most once by the evaluation of an expression.
3213 Furthermore, the prior value shall be read only to determine the value
3214 to be stored.''. If a program breaks these rules, the results on any
3215 particular implementation are entirely unpredictable.
3217 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3218 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3219 diagnosed by this option, and it may give an occasional false positive
3220 result, but in general it has been found fairly effective at detecting
3221 this sort of problem in programs.
3223 The standard is worded confusingly, therefore there is some debate
3224 over the precise meaning of the sequence point rules in subtle cases.
3225 Links to discussions of the problem, including proposed formal
3226 definitions, may be found on the GCC readings page, at
3227 @w{@uref{http://gcc.gnu.org/readings.html}}.
3229 This warning is enabled by @option{-Wall} for C and C++.
3232 @opindex Wreturn-type
3233 @opindex Wno-return-type
3234 Warn whenever a function is defined with a return-type that defaults
3235 to @code{int}. Also warn about any @code{return} statement with no
3236 return-value in a function whose return-type is not @code{void}
3237 (falling off the end of the function body is considered returning
3238 without a value), and about a @code{return} statement with an
3239 expression in a function whose return-type is @code{void}.
3241 For C++, a function without return type always produces a diagnostic
3242 message, even when @option{-Wno-return-type} is specified. The only
3243 exceptions are @samp{main} and functions defined in system headers.
3245 This warning is enabled by @option{-Wall}.
3250 Warn whenever a @code{switch} statement has an index of enumerated type
3251 and lacks a @code{case} for one or more of the named codes of that
3252 enumeration. (The presence of a @code{default} label prevents this
3253 warning.) @code{case} labels outside the enumeration range also
3254 provoke warnings when this option is used (even if there is a
3255 @code{default} label).
3256 This warning is enabled by @option{-Wall}.
3258 @item -Wswitch-default
3259 @opindex Wswitch-default
3260 @opindex Wno-switch-default
3261 Warn whenever a @code{switch} statement does not have a @code{default}
3265 @opindex Wswitch-enum
3266 @opindex Wno-switch-enum
3267 Warn whenever a @code{switch} statement has an index of enumerated type
3268 and lacks a @code{case} for one or more of the named codes of that
3269 enumeration. @code{case} labels outside the enumeration range also
3270 provoke warnings when this option is used. The only difference
3271 between @option{-Wswitch} and this option is that this option gives a
3272 warning about an omitted enumeration code even if there is a
3273 @code{default} label.
3275 @item -Wsync-nand @r{(C and C++ only)}
3277 @opindex Wno-sync-nand
3278 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3279 built-in functions are used. These functions changed semantics in GCC 4.4.
3283 @opindex Wno-trigraphs
3284 Warn if any trigraphs are encountered that might change the meaning of
3285 the program (trigraphs within comments are not warned about).
3286 This warning is enabled by @option{-Wall}.
3288 @item -Wunused-function
3289 @opindex Wunused-function
3290 @opindex Wno-unused-function
3291 Warn whenever a static function is declared but not defined or a
3292 non-inline static function is unused.
3293 This warning is enabled by @option{-Wall}.
3295 @item -Wunused-label
3296 @opindex Wunused-label
3297 @opindex Wno-unused-label
3298 Warn whenever a label is declared but not used.
3299 This warning is enabled by @option{-Wall}.
3301 To suppress this warning use the @samp{unused} attribute
3302 (@pxref{Variable Attributes}).
3304 @item -Wunused-parameter
3305 @opindex Wunused-parameter
3306 @opindex Wno-unused-parameter
3307 Warn whenever a function parameter is unused aside from its declaration.
3309 To suppress this warning use the @samp{unused} attribute
3310 (@pxref{Variable Attributes}).
3312 @item -Wno-unused-result
3313 @opindex Wunused-result
3314 @opindex Wno-unused-result
3315 Do not warn if a caller of a function marked with attribute
3316 @code{warn_unused_result} (@pxref{Variable Attributes}) does not use
3317 its return value. The default is @option{-Wunused-result}.
3319 @item -Wunused-variable
3320 @opindex Wunused-variable
3321 @opindex Wno-unused-variable
3322 Warn whenever a local variable or non-constant static variable is unused
3323 aside from its declaration.
3324 This warning is enabled by @option{-Wall}.
3326 To suppress this warning use the @samp{unused} attribute
3327 (@pxref{Variable Attributes}).
3329 @item -Wunused-value
3330 @opindex Wunused-value
3331 @opindex Wno-unused-value
3332 Warn whenever a statement computes a result that is explicitly not
3333 used. To suppress this warning cast the unused expression to
3334 @samp{void}. This includes an expression-statement or the left-hand
3335 side of a comma expression that contains no side effects. For example,
3336 an expression such as @samp{x[i,j]} will cause a warning, while
3337 @samp{x[(void)i,j]} will not.
3339 This warning is enabled by @option{-Wall}.
3344 All the above @option{-Wunused} options combined.
3346 In order to get a warning about an unused function parameter, you must
3347 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3348 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3350 @item -Wuninitialized
3351 @opindex Wuninitialized
3352 @opindex Wno-uninitialized
3353 Warn if an automatic variable is used without first being initialized
3354 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3355 warn if a non-static reference or non-static @samp{const} member
3356 appears in a class without constructors.
3358 If you want to warn about code which uses the uninitialized value of the
3359 variable in its own initializer, use the @option{-Winit-self} option.
3361 These warnings occur for individual uninitialized or clobbered
3362 elements of structure, union or array variables as well as for
3363 variables which are uninitialized or clobbered as a whole. They do
3364 not occur for variables or elements declared @code{volatile}. Because
3365 these warnings depend on optimization, the exact variables or elements
3366 for which there are warnings will depend on the precise optimization
3367 options and version of GCC used.
3369 Note that there may be no warning about a variable that is used only
3370 to compute a value that itself is never used, because such
3371 computations may be deleted by data flow analysis before the warnings
3374 These warnings are made optional because GCC is not smart
3375 enough to see all the reasons why the code might be correct
3376 despite appearing to have an error. Here is one example of how
3397 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3398 always initialized, but GCC doesn't know this. Here is
3399 another common case:
3404 if (change_y) save_y = y, y = new_y;
3406 if (change_y) y = save_y;
3411 This has no bug because @code{save_y} is used only if it is set.
3413 @cindex @code{longjmp} warnings
3414 This option also warns when a non-volatile automatic variable might be
3415 changed by a call to @code{longjmp}. These warnings as well are possible
3416 only in optimizing compilation.
3418 The compiler sees only the calls to @code{setjmp}. It cannot know
3419 where @code{longjmp} will be called; in fact, a signal handler could
3420 call it at any point in the code. As a result, you may get a warning
3421 even when there is in fact no problem because @code{longjmp} cannot
3422 in fact be called at the place which would cause a problem.
3424 Some spurious warnings can be avoided if you declare all the functions
3425 you use that never return as @code{noreturn}. @xref{Function
3428 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3430 @item -Wunknown-pragmas
3431 @opindex Wunknown-pragmas
3432 @opindex Wno-unknown-pragmas
3433 @cindex warning for unknown pragmas
3434 @cindex unknown pragmas, warning
3435 @cindex pragmas, warning of unknown
3436 Warn when a #pragma directive is encountered which is not understood by
3437 GCC@. If this command line option is used, warnings will even be issued
3438 for unknown pragmas in system header files. This is not the case if
3439 the warnings were only enabled by the @option{-Wall} command line option.
3442 @opindex Wno-pragmas
3444 Do not warn about misuses of pragmas, such as incorrect parameters,
3445 invalid syntax, or conflicts between pragmas. See also
3446 @samp{-Wunknown-pragmas}.
3448 @item -Wstrict-aliasing
3449 @opindex Wstrict-aliasing
3450 @opindex Wno-strict-aliasing
3451 This option is only active when @option{-fstrict-aliasing} is active.
3452 It warns about code which might break the strict aliasing rules that the
3453 compiler is using for optimization. The warning does not catch all
3454 cases, but does attempt to catch the more common pitfalls. It is
3455 included in @option{-Wall}.
3456 It is equivalent to @option{-Wstrict-aliasing=3}
3458 @item -Wstrict-aliasing=n
3459 @opindex Wstrict-aliasing=n
3460 @opindex Wno-strict-aliasing=n
3461 This option is only active when @option{-fstrict-aliasing} is active.
3462 It warns about code which might break the strict aliasing rules that the
3463 compiler is using for optimization.
3464 Higher levels correspond to higher accuracy (fewer false positives).
3465 Higher levels also correspond to more effort, similar to the way -O works.
3466 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3469 Level 1: Most aggressive, quick, least accurate.
3470 Possibly useful when higher levels
3471 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3472 false negatives. However, it has many false positives.
3473 Warns for all pointer conversions between possibly incompatible types,
3474 even if never dereferenced. Runs in the frontend only.
3476 Level 2: Aggressive, quick, not too precise.
3477 May still have many false positives (not as many as level 1 though),
3478 and few false negatives (but possibly more than level 1).
3479 Unlike level 1, it only warns when an address is taken. Warns about
3480 incomplete types. Runs in the frontend only.
3482 Level 3 (default for @option{-Wstrict-aliasing}):
3483 Should have very few false positives and few false
3484 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3485 Takes care of the common punn+dereference pattern in the frontend:
3486 @code{*(int*)&some_float}.
3487 If optimization is enabled, it also runs in the backend, where it deals
3488 with multiple statement cases using flow-sensitive points-to information.
3489 Only warns when the converted pointer is dereferenced.
3490 Does not warn about incomplete types.
3492 @item -Wstrict-overflow
3493 @itemx -Wstrict-overflow=@var{n}
3494 @opindex Wstrict-overflow
3495 @opindex Wno-strict-overflow
3496 This option is only active when @option{-fstrict-overflow} is active.
3497 It warns about cases where the compiler optimizes based on the
3498 assumption that signed overflow does not occur. Note that it does not
3499 warn about all cases where the code might overflow: it only warns
3500 about cases where the compiler implements some optimization. Thus
3501 this warning depends on the optimization level.
3503 An optimization which assumes that signed overflow does not occur is
3504 perfectly safe if the values of the variables involved are such that
3505 overflow never does, in fact, occur. Therefore this warning can
3506 easily give a false positive: a warning about code which is not
3507 actually a problem. To help focus on important issues, several
3508 warning levels are defined. No warnings are issued for the use of
3509 undefined signed overflow when estimating how many iterations a loop
3510 will require, in particular when determining whether a loop will be
3514 @item -Wstrict-overflow=1
3515 Warn about cases which are both questionable and easy to avoid. For
3516 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3517 compiler will simplify this to @code{1}. This level of
3518 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3519 are not, and must be explicitly requested.
3521 @item -Wstrict-overflow=2
3522 Also warn about other cases where a comparison is simplified to a
3523 constant. For example: @code{abs (x) >= 0}. This can only be
3524 simplified when @option{-fstrict-overflow} is in effect, because
3525 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3526 zero. @option{-Wstrict-overflow} (with no level) is the same as
3527 @option{-Wstrict-overflow=2}.
3529 @item -Wstrict-overflow=3
3530 Also warn about other cases where a comparison is simplified. For
3531 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3533 @item -Wstrict-overflow=4
3534 Also warn about other simplifications not covered by the above cases.
3535 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3537 @item -Wstrict-overflow=5
3538 Also warn about cases where the compiler reduces the magnitude of a
3539 constant involved in a comparison. For example: @code{x + 2 > y} will
3540 be simplified to @code{x + 1 >= y}. This is reported only at the
3541 highest warning level because this simplification applies to many
3542 comparisons, so this warning level will give a very large number of
3546 @item -Warray-bounds
3547 @opindex Wno-array-bounds
3548 @opindex Warray-bounds
3549 This option is only active when @option{-ftree-vrp} is active
3550 (default for -O2 and above). It warns about subscripts to arrays
3551 that are always out of bounds. This warning is enabled by @option{-Wall}.
3553 @item -Wno-div-by-zero
3554 @opindex Wno-div-by-zero
3555 @opindex Wdiv-by-zero
3556 Do not warn about compile-time integer division by zero. Floating point
3557 division by zero is not warned about, as it can be a legitimate way of
3558 obtaining infinities and NaNs.
3560 @item -Wsystem-headers
3561 @opindex Wsystem-headers
3562 @opindex Wno-system-headers
3563 @cindex warnings from system headers
3564 @cindex system headers, warnings from
3565 Print warning messages for constructs found in system header files.
3566 Warnings from system headers are normally suppressed, on the assumption
3567 that they usually do not indicate real problems and would only make the
3568 compiler output harder to read. Using this command line option tells
3569 GCC to emit warnings from system headers as if they occurred in user
3570 code. However, note that using @option{-Wall} in conjunction with this
3571 option will @emph{not} warn about unknown pragmas in system
3572 headers---for that, @option{-Wunknown-pragmas} must also be used.
3575 @opindex Wfloat-equal
3576 @opindex Wno-float-equal
3577 Warn if floating point values are used in equality comparisons.
3579 The idea behind this is that sometimes it is convenient (for the
3580 programmer) to consider floating-point values as approximations to
3581 infinitely precise real numbers. If you are doing this, then you need
3582 to compute (by analyzing the code, or in some other way) the maximum or
3583 likely maximum error that the computation introduces, and allow for it
3584 when performing comparisons (and when producing output, but that's a
3585 different problem). In particular, instead of testing for equality, you
3586 would check to see whether the two values have ranges that overlap; and
3587 this is done with the relational operators, so equality comparisons are
3590 @item -Wtraditional @r{(C and Objective-C only)}
3591 @opindex Wtraditional
3592 @opindex Wno-traditional
3593 Warn about certain constructs that behave differently in traditional and
3594 ISO C@. Also warn about ISO C constructs that have no traditional C
3595 equivalent, and/or problematic constructs which should be avoided.
3599 Macro parameters that appear within string literals in the macro body.
3600 In traditional C macro replacement takes place within string literals,
3601 but does not in ISO C@.
3604 In traditional C, some preprocessor directives did not exist.
3605 Traditional preprocessors would only consider a line to be a directive
3606 if the @samp{#} appeared in column 1 on the line. Therefore
3607 @option{-Wtraditional} warns about directives that traditional C
3608 understands but would ignore because the @samp{#} does not appear as the
3609 first character on the line. It also suggests you hide directives like
3610 @samp{#pragma} not understood by traditional C by indenting them. Some
3611 traditional implementations would not recognize @samp{#elif}, so it
3612 suggests avoiding it altogether.
3615 A function-like macro that appears without arguments.
3618 The unary plus operator.
3621 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3622 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3623 constants.) Note, these suffixes appear in macros defined in the system
3624 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3625 Use of these macros in user code might normally lead to spurious
3626 warnings, however GCC's integrated preprocessor has enough context to
3627 avoid warning in these cases.
3630 A function declared external in one block and then used after the end of
3634 A @code{switch} statement has an operand of type @code{long}.
3637 A non-@code{static} function declaration follows a @code{static} one.
3638 This construct is not accepted by some traditional C compilers.
3641 The ISO type of an integer constant has a different width or
3642 signedness from its traditional type. This warning is only issued if
3643 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3644 typically represent bit patterns, are not warned about.
3647 Usage of ISO string concatenation is detected.
3650 Initialization of automatic aggregates.
3653 Identifier conflicts with labels. Traditional C lacks a separate
3654 namespace for labels.
3657 Initialization of unions. If the initializer is zero, the warning is
3658 omitted. This is done under the assumption that the zero initializer in
3659 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3660 initializer warnings and relies on default initialization to zero in the
3664 Conversions by prototypes between fixed/floating point values and vice
3665 versa. The absence of these prototypes when compiling with traditional
3666 C would cause serious problems. This is a subset of the possible
3667 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3670 Use of ISO C style function definitions. This warning intentionally is
3671 @emph{not} issued for prototype declarations or variadic functions
3672 because these ISO C features will appear in your code when using
3673 libiberty's traditional C compatibility macros, @code{PARAMS} and
3674 @code{VPARAMS}. This warning is also bypassed for nested functions
3675 because that feature is already a GCC extension and thus not relevant to
3676 traditional C compatibility.
3679 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3680 @opindex Wtraditional-conversion
3681 @opindex Wno-traditional-conversion
3682 Warn if a prototype causes a type conversion that is different from what
3683 would happen to the same argument in the absence of a prototype. This
3684 includes conversions of fixed point to floating and vice versa, and
3685 conversions changing the width or signedness of a fixed point argument
3686 except when the same as the default promotion.
3688 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3689 @opindex Wdeclaration-after-statement
3690 @opindex Wno-declaration-after-statement
3691 Warn when a declaration is found after a statement in a block. This
3692 construct, known from C++, was introduced with ISO C99 and is by default
3693 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3694 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3699 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3701 @item -Wno-endif-labels
3702 @opindex Wno-endif-labels
3703 @opindex Wendif-labels
3704 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3709 Warn whenever a local variable shadows another local variable, parameter or
3710 global variable or whenever a built-in function is shadowed.
3712 @item -Wlarger-than=@var{len}
3713 @opindex Wlarger-than=@var{len}
3714 @opindex Wlarger-than-@var{len}
3715 Warn whenever an object of larger than @var{len} bytes is defined.
3717 @item -Wframe-larger-than=@var{len}
3718 @opindex Wframe-larger-than
3719 Warn if the size of a function frame is larger than @var{len} bytes.
3720 The computation done to determine the stack frame size is approximate
3721 and not conservative.
3722 The actual requirements may be somewhat greater than @var{len}
3723 even if you do not get a warning. In addition, any space allocated
3724 via @code{alloca}, variable-length arrays, or related constructs
3725 is not included by the compiler when determining
3726 whether or not to issue a warning.
3728 @item -Wunsafe-loop-optimizations
3729 @opindex Wunsafe-loop-optimizations
3730 @opindex Wno-unsafe-loop-optimizations
3731 Warn if the loop cannot be optimized because the compiler could not
3732 assume anything on the bounds of the loop indices. With
3733 @option{-funsafe-loop-optimizations} warn if the compiler made
3736 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
3737 @opindex Wno-pedantic-ms-format
3738 @opindex Wpedantic-ms-format
3739 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
3740 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
3741 depending on the MS runtime, when you are using the options @option{-Wformat}
3742 and @option{-pedantic} without gnu-extensions.
3744 @item -Wpointer-arith
3745 @opindex Wpointer-arith
3746 @opindex Wno-pointer-arith
3747 Warn about anything that depends on the ``size of'' a function type or
3748 of @code{void}. GNU C assigns these types a size of 1, for
3749 convenience in calculations with @code{void *} pointers and pointers
3750 to functions. In C++, warn also when an arithmetic operation involves
3751 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3754 @opindex Wtype-limits
3755 @opindex Wno-type-limits
3756 Warn if a comparison is always true or always false due to the limited
3757 range of the data type, but do not warn for constant expressions. For
3758 example, warn if an unsigned variable is compared against zero with
3759 @samp{<} or @samp{>=}. This warning is also enabled by
3762 @item -Wbad-function-cast @r{(C and Objective-C only)}
3763 @opindex Wbad-function-cast
3764 @opindex Wno-bad-function-cast
3765 Warn whenever a function call is cast to a non-matching type.
3766 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3768 @item -Wc++-compat @r{(C and Objective-C only)}
3769 Warn about ISO C constructs that are outside of the common subset of
3770 ISO C and ISO C++, e.g.@: request for implicit conversion from
3771 @code{void *} to a pointer to non-@code{void} type.
3773 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3774 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3775 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3776 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3780 @opindex Wno-cast-qual
3781 Warn whenever a pointer is cast so as to remove a type qualifier from
3782 the target type. For example, warn if a @code{const char *} is cast
3783 to an ordinary @code{char *}.
3785 Also warn when making a cast which introduces a type qualifier in an
3786 unsafe way. For example, casting @code{char **} to @code{const char **}
3787 is unsafe, as in this example:
3790 /* p is char ** value. */
3791 const char **q = (const char **) p;
3792 /* Assignment of readonly string to const char * is OK. */
3794 /* Now char** pointer points to read-only memory. */
3799 @opindex Wcast-align
3800 @opindex Wno-cast-align
3801 Warn whenever a pointer is cast such that the required alignment of the
3802 target is increased. For example, warn if a @code{char *} is cast to
3803 an @code{int *} on machines where integers can only be accessed at
3804 two- or four-byte boundaries.
3806 @item -Wwrite-strings
3807 @opindex Wwrite-strings
3808 @opindex Wno-write-strings
3809 When compiling C, give string constants the type @code{const
3810 char[@var{length}]} so that copying the address of one into a
3811 non-@code{const} @code{char *} pointer will get a warning. These
3812 warnings will help you find at compile time code that can try to write
3813 into a string constant, but only if you have been very careful about
3814 using @code{const} in declarations and prototypes. Otherwise, it will
3815 just be a nuisance. This is why we did not make @option{-Wall} request
3818 When compiling C++, warn about the deprecated conversion from string
3819 literals to @code{char *}. This warning is enabled by default for C++
3824 @opindex Wno-clobbered
3825 Warn for variables that might be changed by @samp{longjmp} or
3826 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
3829 @opindex Wconversion
3830 @opindex Wno-conversion
3831 Warn for implicit conversions that may alter a value. This includes
3832 conversions between real and integer, like @code{abs (x)} when
3833 @code{x} is @code{double}; conversions between signed and unsigned,
3834 like @code{unsigned ui = -1}; and conversions to smaller types, like
3835 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
3836 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
3837 changed by the conversion like in @code{abs (2.0)}. Warnings about
3838 conversions between signed and unsigned integers can be disabled by
3839 using @option{-Wno-sign-conversion}.
3841 For C++, also warn for conversions between @code{NULL} and non-pointer
3842 types; confusing overload resolution for user-defined conversions; and
3843 conversions that will never use a type conversion operator:
3844 conversions to @code{void}, the same type, a base class or a reference
3845 to them. Warnings about conversions between signed and unsigned
3846 integers are disabled by default in C++ unless
3847 @option{-Wsign-conversion} is explicitly enabled.
3850 @opindex Wempty-body
3851 @opindex Wno-empty-body
3852 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
3853 while} statement. This warning is also enabled by @option{-Wextra}.
3855 @item -Wenum-compare
3856 @opindex Wenum-compare
3857 @opindex Wno-enum-compare
3858 Warn about a comparison between values of different enum types. In C++
3859 this warning is enabled by default. In C this warning is enabled by
3862 @item -Wjump-misses-init @r{(C, Objective-C only)}
3863 @opindex Wjump-misses-init
3864 @opindex Wno-jump-misses-init
3865 Warn if a @code{goto} statement or a @code{switch} statement jumps
3866 forward across the initialization of a variable, or jumps backward to a
3867 label after the variable has been initialized. This only warns about
3868 variables which are initialized when they are declared. This warning is
3869 only supported for C and Objective C; in C++ this sort of branch is an
3872 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
3873 can be disabled with the @option{-Wno-jump-misses-init} option.
3875 @item -Wsign-compare
3876 @opindex Wsign-compare
3877 @opindex Wno-sign-compare
3878 @cindex warning for comparison of signed and unsigned values
3879 @cindex comparison of signed and unsigned values, warning
3880 @cindex signed and unsigned values, comparison warning
3881 Warn when a comparison between signed and unsigned values could produce
3882 an incorrect result when the signed value is converted to unsigned.
3883 This warning is also enabled by @option{-Wextra}; to get the other warnings
3884 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3886 @item -Wsign-conversion
3887 @opindex Wsign-conversion
3888 @opindex Wno-sign-conversion
3889 Warn for implicit conversions that may change the sign of an integer
3890 value, like assigning a signed integer expression to an unsigned
3891 integer variable. An explicit cast silences the warning. In C, this
3892 option is enabled also by @option{-Wconversion}.
3896 @opindex Wno-address
3897 Warn about suspicious uses of memory addresses. These include using
3898 the address of a function in a conditional expression, such as
3899 @code{void func(void); if (func)}, and comparisons against the memory
3900 address of a string literal, such as @code{if (x == "abc")}. Such
3901 uses typically indicate a programmer error: the address of a function
3902 always evaluates to true, so their use in a conditional usually
3903 indicate that the programmer forgot the parentheses in a function
3904 call; and comparisons against string literals result in unspecified
3905 behavior and are not portable in C, so they usually indicate that the
3906 programmer intended to use @code{strcmp}. This warning is enabled by
3910 @opindex Wlogical-op
3911 @opindex Wno-logical-op
3912 Warn about suspicious uses of logical operators in expressions.
3913 This includes using logical operators in contexts where a
3914 bit-wise operator is likely to be expected.
3916 @item -Waggregate-return
3917 @opindex Waggregate-return
3918 @opindex Wno-aggregate-return
3919 Warn if any functions that return structures or unions are defined or
3920 called. (In languages where you can return an array, this also elicits
3923 @item -Wno-attributes
3924 @opindex Wno-attributes
3925 @opindex Wattributes
3926 Do not warn if an unexpected @code{__attribute__} is used, such as
3927 unrecognized attributes, function attributes applied to variables,
3928 etc. This will not stop errors for incorrect use of supported
3931 @item -Wno-builtin-macro-redefined
3932 @opindex Wno-builtin-macro-redefined
3933 @opindex Wbuiltin-macro-redefined
3934 Do not warn if certain built-in macros are redefined. This suppresses
3935 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
3936 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
3938 @item -Wstrict-prototypes @r{(C and Objective-C only)}
3939 @opindex Wstrict-prototypes
3940 @opindex Wno-strict-prototypes
3941 Warn if a function is declared or defined without specifying the
3942 argument types. (An old-style function definition is permitted without
3943 a warning if preceded by a declaration which specifies the argument
3946 @item -Wold-style-declaration @r{(C and Objective-C only)}
3947 @opindex Wold-style-declaration
3948 @opindex Wno-old-style-declaration
3949 Warn for obsolescent usages, according to the C Standard, in a
3950 declaration. For example, warn if storage-class specifiers like
3951 @code{static} are not the first things in a declaration. This warning
3952 is also enabled by @option{-Wextra}.
3954 @item -Wold-style-definition @r{(C and Objective-C only)}
3955 @opindex Wold-style-definition
3956 @opindex Wno-old-style-definition
3957 Warn if an old-style function definition is used. A warning is given
3958 even if there is a previous prototype.
3960 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
3961 @opindex Wmissing-parameter-type
3962 @opindex Wno-missing-parameter-type
3963 A function parameter is declared without a type specifier in K&R-style
3970 This warning is also enabled by @option{-Wextra}.
3972 @item -Wmissing-prototypes @r{(C and Objective-C only)}
3973 @opindex Wmissing-prototypes
3974 @opindex Wno-missing-prototypes
3975 Warn if a global function is defined without a previous prototype
3976 declaration. This warning is issued even if the definition itself
3977 provides a prototype. The aim is to detect global functions that fail
3978 to be declared in header files.
3980 @item -Wmissing-declarations
3981 @opindex Wmissing-declarations
3982 @opindex Wno-missing-declarations
3983 Warn if a global function is defined without a previous declaration.
3984 Do so even if the definition itself provides a prototype.
3985 Use this option to detect global functions that are not declared in
3986 header files. In C++, no warnings are issued for function templates,
3987 or for inline functions, or for functions in anonymous namespaces.
3989 @item -Wmissing-field-initializers
3990 @opindex Wmissing-field-initializers
3991 @opindex Wno-missing-field-initializers
3995 Warn if a structure's initializer has some fields missing. For
3996 example, the following code would cause such a warning, because
3997 @code{x.h} is implicitly zero:
4000 struct s @{ int f, g, h; @};
4001 struct s x = @{ 3, 4 @};
4004 This option does not warn about designated initializers, so the following
4005 modification would not trigger a warning:
4008 struct s @{ int f, g, h; @};
4009 struct s x = @{ .f = 3, .g = 4 @};
4012 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
4013 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
4015 @item -Wmissing-noreturn
4016 @opindex Wmissing-noreturn
4017 @opindex Wno-missing-noreturn
4018 Warn about functions which might be candidates for attribute @code{noreturn}.
4019 Note these are only possible candidates, not absolute ones. Care should
4020 be taken to manually verify functions actually do not ever return before
4021 adding the @code{noreturn} attribute, otherwise subtle code generation
4022 bugs could be introduced. You will not get a warning for @code{main} in
4023 hosted C environments.
4025 @item -Wmissing-format-attribute
4026 @opindex Wmissing-format-attribute
4027 @opindex Wno-missing-format-attribute
4030 Warn about function pointers which might be candidates for @code{format}
4031 attributes. Note these are only possible candidates, not absolute ones.
4032 GCC will guess that function pointers with @code{format} attributes that
4033 are used in assignment, initialization, parameter passing or return
4034 statements should have a corresponding @code{format} attribute in the
4035 resulting type. I.e.@: the left-hand side of the assignment or
4036 initialization, the type of the parameter variable, or the return type
4037 of the containing function respectively should also have a @code{format}
4038 attribute to avoid the warning.
4040 GCC will also warn about function definitions which might be
4041 candidates for @code{format} attributes. Again, these are only
4042 possible candidates. GCC will guess that @code{format} attributes
4043 might be appropriate for any function that calls a function like
4044 @code{vprintf} or @code{vscanf}, but this might not always be the
4045 case, and some functions for which @code{format} attributes are
4046 appropriate may not be detected.
4048 @item -Wno-multichar
4049 @opindex Wno-multichar
4051 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4052 Usually they indicate a typo in the user's code, as they have
4053 implementation-defined values, and should not be used in portable code.
4055 @item -Wnormalized=<none|id|nfc|nfkc>
4056 @opindex Wnormalized=
4059 @cindex character set, input normalization
4060 In ISO C and ISO C++, two identifiers are different if they are
4061 different sequences of characters. However, sometimes when characters
4062 outside the basic ASCII character set are used, you can have two
4063 different character sequences that look the same. To avoid confusion,
4064 the ISO 10646 standard sets out some @dfn{normalization rules} which
4065 when applied ensure that two sequences that look the same are turned into
4066 the same sequence. GCC can warn you if you are using identifiers which
4067 have not been normalized; this option controls that warning.
4069 There are four levels of warning that GCC supports. The default is
4070 @option{-Wnormalized=nfc}, which warns about any identifier which is
4071 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4072 recommended form for most uses.
4074 Unfortunately, there are some characters which ISO C and ISO C++ allow
4075 in identifiers that when turned into NFC aren't allowable as
4076 identifiers. That is, there's no way to use these symbols in portable
4077 ISO C or C++ and have all your identifiers in NFC@.
4078 @option{-Wnormalized=id} suppresses the warning for these characters.
4079 It is hoped that future versions of the standards involved will correct
4080 this, which is why this option is not the default.
4082 You can switch the warning off for all characters by writing
4083 @option{-Wnormalized=none}. You would only want to do this if you
4084 were using some other normalization scheme (like ``D''), because
4085 otherwise you can easily create bugs that are literally impossible to see.
4087 Some characters in ISO 10646 have distinct meanings but look identical
4088 in some fonts or display methodologies, especially once formatting has
4089 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4090 LETTER N'', will display just like a regular @code{n} which has been
4091 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4092 normalization scheme to convert all these into a standard form as
4093 well, and GCC will warn if your code is not in NFKC if you use
4094 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4095 about every identifier that contains the letter O because it might be
4096 confused with the digit 0, and so is not the default, but may be
4097 useful as a local coding convention if the programming environment is
4098 unable to be fixed to display these characters distinctly.
4100 @item -Wno-deprecated
4101 @opindex Wno-deprecated
4102 @opindex Wdeprecated
4103 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4105 @item -Wno-deprecated-declarations
4106 @opindex Wno-deprecated-declarations
4107 @opindex Wdeprecated-declarations
4108 Do not warn about uses of functions (@pxref{Function Attributes}),
4109 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4110 Attributes}) marked as deprecated by using the @code{deprecated}
4114 @opindex Wno-overflow
4116 Do not warn about compile-time overflow in constant expressions.
4118 @item -Woverride-init @r{(C and Objective-C only)}
4119 @opindex Woverride-init
4120 @opindex Wno-override-init
4124 Warn if an initialized field without side effects is overridden when
4125 using designated initializers (@pxref{Designated Inits, , Designated
4128 This warning is included in @option{-Wextra}. To get other
4129 @option{-Wextra} warnings without this one, use @samp{-Wextra
4130 -Wno-override-init}.
4135 Warn if a structure is given the packed attribute, but the packed
4136 attribute has no effect on the layout or size of the structure.
4137 Such structures may be mis-aligned for little benefit. For
4138 instance, in this code, the variable @code{f.x} in @code{struct bar}
4139 will be misaligned even though @code{struct bar} does not itself
4140 have the packed attribute:
4147 @} __attribute__((packed));
4155 @item -Wpacked-bitfield-compat
4156 @opindex Wpacked-bitfield-compat
4157 @opindex Wno-packed-bitfield-compat
4158 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4159 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4160 the change can lead to differences in the structure layout. GCC
4161 informs you when the offset of such a field has changed in GCC 4.4.
4162 For example there is no longer a 4-bit padding between field @code{a}
4163 and @code{b} in this structure:
4170 @} __attribute__ ((packed));
4173 This warning is enabled by default. Use
4174 @option{-Wno-packed-bitfield-compat} to disable this warning.
4179 Warn if padding is included in a structure, either to align an element
4180 of the structure or to align the whole structure. Sometimes when this
4181 happens it is possible to rearrange the fields of the structure to
4182 reduce the padding and so make the structure smaller.
4184 @item -Wredundant-decls
4185 @opindex Wredundant-decls
4186 @opindex Wno-redundant-decls
4187 Warn if anything is declared more than once in the same scope, even in
4188 cases where multiple declaration is valid and changes nothing.
4190 @item -Wnested-externs @r{(C and Objective-C only)}
4191 @opindex Wnested-externs
4192 @opindex Wno-nested-externs
4193 Warn if an @code{extern} declaration is encountered within a function.
4195 @item -Wunreachable-code
4196 @opindex Wunreachable-code
4197 @opindex Wno-unreachable-code
4198 Warn if the compiler detects that code will never be executed.
4200 This option is intended to warn when the compiler detects that at
4201 least a whole line of source code will never be executed, because
4202 some condition is never satisfied or because it is after a
4203 procedure that never returns.
4205 It is possible for this option to produce a warning even though there
4206 are circumstances under which part of the affected line can be executed,
4207 so care should be taken when removing apparently-unreachable code.
4209 For instance, when a function is inlined, a warning may mean that the
4210 line is unreachable in only one inlined copy of the function.
4212 This option is not made part of @option{-Wall} because in a debugging
4213 version of a program there is often substantial code which checks
4214 correct functioning of the program and is, hopefully, unreachable
4215 because the program does work. Another common use of unreachable
4216 code is to provide behavior which is selectable at compile-time.
4221 Warn if a function can not be inlined and it was declared as inline.
4222 Even with this option, the compiler will not warn about failures to
4223 inline functions declared in system headers.
4225 The compiler uses a variety of heuristics to determine whether or not
4226 to inline a function. For example, the compiler takes into account
4227 the size of the function being inlined and the amount of inlining
4228 that has already been done in the current function. Therefore,
4229 seemingly insignificant changes in the source program can cause the
4230 warnings produced by @option{-Winline} to appear or disappear.
4232 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4233 @opindex Wno-invalid-offsetof
4234 @opindex Winvalid-offsetof
4235 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4236 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4237 to a non-POD type is undefined. In existing C++ implementations,
4238 however, @samp{offsetof} typically gives meaningful results even when
4239 applied to certain kinds of non-POD types. (Such as a simple
4240 @samp{struct} that fails to be a POD type only by virtue of having a
4241 constructor.) This flag is for users who are aware that they are
4242 writing nonportable code and who have deliberately chosen to ignore the
4245 The restrictions on @samp{offsetof} may be relaxed in a future version
4246 of the C++ standard.
4248 @item -Wno-int-to-pointer-cast @r{(C and Objective-C only)}
4249 @opindex Wno-int-to-pointer-cast
4250 @opindex Wint-to-pointer-cast
4251 Suppress warnings from casts to pointer type of an integer of a
4254 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4255 @opindex Wno-pointer-to-int-cast
4256 @opindex Wpointer-to-int-cast
4257 Suppress warnings from casts from a pointer to an integer type of a
4261 @opindex Winvalid-pch
4262 @opindex Wno-invalid-pch
4263 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4264 the search path but can't be used.
4268 @opindex Wno-long-long
4269 Warn if @samp{long long} type is used. This is enabled by either
4270 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4271 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4273 @item -Wvariadic-macros
4274 @opindex Wvariadic-macros
4275 @opindex Wno-variadic-macros
4276 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4277 alternate syntax when in pedantic ISO C99 mode. This is default.
4278 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4283 Warn if variable length array is used in the code.
4284 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4285 the variable length array.
4287 @item -Wvolatile-register-var
4288 @opindex Wvolatile-register-var
4289 @opindex Wno-volatile-register-var
4290 Warn if a register variable is declared volatile. The volatile
4291 modifier does not inhibit all optimizations that may eliminate reads
4292 and/or writes to register variables. This warning is enabled by
4295 @item -Wdisabled-optimization
4296 @opindex Wdisabled-optimization
4297 @opindex Wno-disabled-optimization
4298 Warn if a requested optimization pass is disabled. This warning does
4299 not generally indicate that there is anything wrong with your code; it
4300 merely indicates that GCC's optimizers were unable to handle the code
4301 effectively. Often, the problem is that your code is too big or too
4302 complex; GCC will refuse to optimize programs when the optimization
4303 itself is likely to take inordinate amounts of time.
4305 @item -Wpointer-sign @r{(C and Objective-C only)}
4306 @opindex Wpointer-sign
4307 @opindex Wno-pointer-sign
4308 Warn for pointer argument passing or assignment with different signedness.
4309 This option is only supported for C and Objective-C@. It is implied by
4310 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4311 @option{-Wno-pointer-sign}.
4313 @item -Wstack-protector
4314 @opindex Wstack-protector
4315 @opindex Wno-stack-protector
4316 This option is only active when @option{-fstack-protector} is active. It
4317 warns about functions that will not be protected against stack smashing.
4320 @opindex Wno-mudflap
4321 Suppress warnings about constructs that cannot be instrumented by
4324 @item -Woverlength-strings
4325 @opindex Woverlength-strings
4326 @opindex Wno-overlength-strings
4327 Warn about string constants which are longer than the ``minimum
4328 maximum'' length specified in the C standard. Modern compilers
4329 generally allow string constants which are much longer than the
4330 standard's minimum limit, but very portable programs should avoid
4331 using longer strings.
4333 The limit applies @emph{after} string constant concatenation, and does
4334 not count the trailing NUL@. In C89, the limit was 509 characters; in
4335 C99, it was raised to 4095. C++98 does not specify a normative
4336 minimum maximum, so we do not diagnose overlength strings in C++@.
4338 This option is implied by @option{-pedantic}, and can be disabled with
4339 @option{-Wno-overlength-strings}.
4341 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
4342 @opindex Wunsuffixed-float-constants
4344 GCC will issue a warning for any floating constant that does not have
4345 a suffix. When used together with @option{-Wsystem-headers} it will
4346 warn about such constants in system header files. This can be useful
4347 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4348 from the decimal floating-point extension to C99.
4351 @node Debugging Options
4352 @section Options for Debugging Your Program or GCC
4353 @cindex options, debugging
4354 @cindex debugging information options
4356 GCC has various special options that are used for debugging
4357 either your program or GCC:
4362 Produce debugging information in the operating system's native format
4363 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4366 On most systems that use stabs format, @option{-g} enables use of extra
4367 debugging information that only GDB can use; this extra information
4368 makes debugging work better in GDB but will probably make other debuggers
4370 refuse to read the program. If you want to control for certain whether
4371 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4372 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4374 GCC allows you to use @option{-g} with
4375 @option{-O}. The shortcuts taken by optimized code may occasionally
4376 produce surprising results: some variables you declared may not exist
4377 at all; flow of control may briefly move where you did not expect it;
4378 some statements may not be executed because they compute constant
4379 results or their values were already at hand; some statements may
4380 execute in different places because they were moved out of loops.
4382 Nevertheless it proves possible to debug optimized output. This makes
4383 it reasonable to use the optimizer for programs that might have bugs.
4385 The following options are useful when GCC is generated with the
4386 capability for more than one debugging format.
4390 Produce debugging information for use by GDB@. This means to use the
4391 most expressive format available (DWARF 2, stabs, or the native format
4392 if neither of those are supported), including GDB extensions if at all
4397 Produce debugging information in stabs format (if that is supported),
4398 without GDB extensions. This is the format used by DBX on most BSD
4399 systems. On MIPS, Alpha and System V Release 4 systems this option
4400 produces stabs debugging output which is not understood by DBX or SDB@.
4401 On System V Release 4 systems this option requires the GNU assembler.
4403 @item -feliminate-unused-debug-symbols
4404 @opindex feliminate-unused-debug-symbols
4405 Produce debugging information in stabs format (if that is supported),
4406 for only symbols that are actually used.
4408 @item -femit-class-debug-always
4409 Instead of emitting debugging information for a C++ class in only one
4410 object file, emit it in all object files using the class. This option
4411 should be used only with debuggers that are unable to handle the way GCC
4412 normally emits debugging information for classes because using this
4413 option will increase the size of debugging information by as much as a
4418 Produce debugging information in stabs format (if that is supported),
4419 using GNU extensions understood only by the GNU debugger (GDB)@. The
4420 use of these extensions is likely to make other debuggers crash or
4421 refuse to read the program.
4425 Produce debugging information in COFF format (if that is supported).
4426 This is the format used by SDB on most System V systems prior to
4431 Produce debugging information in XCOFF format (if that is supported).
4432 This is the format used by the DBX debugger on IBM RS/6000 systems.
4436 Produce debugging information in XCOFF format (if that is supported),
4437 using GNU extensions understood only by the GNU debugger (GDB)@. The
4438 use of these extensions is likely to make other debuggers crash or
4439 refuse to read the program, and may cause assemblers other than the GNU
4440 assembler (GAS) to fail with an error.
4442 @item -gdwarf-@var{version}
4443 @opindex gdwarf-@var{version}
4444 Produce debugging information in DWARF format (if that is
4445 supported). This is the format used by DBX on IRIX 6. The value
4446 of @var{version} may be either 2, 3 or 4; the default version is 2.
4448 Note that with DWARF version 2 some ports require, and will always
4449 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4451 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
4452 for maximum benefit.
4454 @item -gstrict-dwarf
4455 @opindex gstrict-dwarf
4456 Disallow using extensions of later DWARF standard version than selected
4457 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
4458 DWARF extensions from later standard versions is allowed.
4460 @item -gno-strict-dwarf
4461 @opindex gno-strict-dwarf
4462 Allow using extensions of later DWARF standard version than selected with
4463 @option{-gdwarf-@var{version}}.
4467 Produce debugging information in VMS debug format (if that is
4468 supported). This is the format used by DEBUG on VMS systems.
4471 @itemx -ggdb@var{level}
4472 @itemx -gstabs@var{level}
4473 @itemx -gcoff@var{level}
4474 @itemx -gxcoff@var{level}
4475 @itemx -gvms@var{level}
4476 Request debugging information and also use @var{level} to specify how
4477 much information. The default level is 2.
4479 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4482 Level 1 produces minimal information, enough for making backtraces in
4483 parts of the program that you don't plan to debug. This includes
4484 descriptions of functions and external variables, but no information
4485 about local variables and no line numbers.
4487 Level 3 includes extra information, such as all the macro definitions
4488 present in the program. Some debuggers support macro expansion when
4489 you use @option{-g3}.
4491 @option{-gdwarf-2} does not accept a concatenated debug level, because
4492 GCC used to support an option @option{-gdwarf} that meant to generate
4493 debug information in version 1 of the DWARF format (which is very
4494 different from version 2), and it would have been too confusing. That
4495 debug format is long obsolete, but the option cannot be changed now.
4496 Instead use an additional @option{-g@var{level}} option to change the
4497 debug level for DWARF.
4501 Turn off generation of debug info, if leaving out this option would have
4502 generated it, or turn it on at level 2 otherwise. The position of this
4503 argument in the command line does not matter, it takes effect after all
4504 other options are processed, and it does so only once, no matter how
4505 many times it is given. This is mainly intended to be used with
4506 @option{-fcompare-debug}.
4508 @item -fdump-final-insns@r{[}=@var{file}@r{]}
4509 @opindex fdump-final-insns
4510 Dump the final internal representation (RTL) to @var{file}. If the
4511 optional argument is omitted (or if @var{file} is @code{.}), the name
4512 of the dump file will be determined by appending @code{.gkd} to the
4513 compilation output file name.
4515 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4516 @opindex fcompare-debug
4517 @opindex fno-compare-debug
4518 If no error occurs during compilation, run the compiler a second time,
4519 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4520 passed to the second compilation. Dump the final internal
4521 representation in both compilations, and print an error if they differ.
4523 If the equal sign is omitted, the default @option{-gtoggle} is used.
4525 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4526 and nonzero, implicitly enables @option{-fcompare-debug}. If
4527 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4528 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4531 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4532 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4533 of the final representation and the second compilation, preventing even
4534 @env{GCC_COMPARE_DEBUG} from taking effect.
4536 To verify full coverage during @option{-fcompare-debug} testing, set
4537 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4538 which GCC will reject as an invalid option in any actual compilation
4539 (rather than preprocessing, assembly or linking). To get just a
4540 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4541 not overridden} will do.
4543 @item -fcompare-debug-second
4544 @opindex fcompare-debug-second
4545 This option is implicitly passed to the compiler for the second
4546 compilation requested by @option{-fcompare-debug}, along with options to
4547 silence warnings, and omitting other options that would cause
4548 side-effect compiler outputs to files or to the standard output. Dump
4549 files and preserved temporary files are renamed so as to contain the
4550 @code{.gk} additional extension during the second compilation, to avoid
4551 overwriting those generated by the first.
4553 When this option is passed to the compiler driver, it causes the
4554 @emph{first} compilation to be skipped, which makes it useful for little
4555 other than debugging the compiler proper.
4557 @item -feliminate-dwarf2-dups
4558 @opindex feliminate-dwarf2-dups
4559 Compress DWARF2 debugging information by eliminating duplicated
4560 information about each symbol. This option only makes sense when
4561 generating DWARF2 debugging information with @option{-gdwarf-2}.
4563 @item -femit-struct-debug-baseonly
4564 Emit debug information for struct-like types
4565 only when the base name of the compilation source file
4566 matches the base name of file in which the struct was defined.
4568 This option substantially reduces the size of debugging information,
4569 but at significant potential loss in type information to the debugger.
4570 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4571 See @option{-femit-struct-debug-detailed} for more detailed control.
4573 This option works only with DWARF 2.
4575 @item -femit-struct-debug-reduced
4576 Emit debug information for struct-like types
4577 only when the base name of the compilation source file
4578 matches the base name of file in which the type was defined,
4579 unless the struct is a template or defined in a system header.
4581 This option significantly reduces the size of debugging information,
4582 with some potential loss in type information to the debugger.
4583 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4584 See @option{-femit-struct-debug-detailed} for more detailed control.
4586 This option works only with DWARF 2.
4588 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4589 Specify the struct-like types
4590 for which the compiler will generate debug information.
4591 The intent is to reduce duplicate struct debug information
4592 between different object files within the same program.
4594 This option is a detailed version of
4595 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4596 which will serve for most needs.
4598 A specification has the syntax
4599 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4601 The optional first word limits the specification to
4602 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4603 A struct type is used directly when it is the type of a variable, member.
4604 Indirect uses arise through pointers to structs.
4605 That is, when use of an incomplete struct would be legal, the use is indirect.
4607 @samp{struct one direct; struct two * indirect;}.
4609 The optional second word limits the specification to
4610 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4611 Generic structs are a bit complicated to explain.
4612 For C++, these are non-explicit specializations of template classes,
4613 or non-template classes within the above.
4614 Other programming languages have generics,
4615 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4617 The third word specifies the source files for those
4618 structs for which the compiler will emit debug information.
4619 The values @samp{none} and @samp{any} have the normal meaning.
4620 The value @samp{base} means that
4621 the base of name of the file in which the type declaration appears
4622 must match the base of the name of the main compilation file.
4623 In practice, this means that
4624 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4625 but types declared in other header will not.
4626 The value @samp{sys} means those types satisfying @samp{base}
4627 or declared in system or compiler headers.
4629 You may need to experiment to determine the best settings for your application.
4631 The default is @samp{-femit-struct-debug-detailed=all}.
4633 This option works only with DWARF 2.
4635 @item -fenable-icf-debug
4636 @opindex fenable-icf-debug
4637 Generate additional debug information to support identical code folding (ICF).
4638 This option only works with DWARF version 2 or higher.
4640 @item -fno-merge-debug-strings
4641 @opindex fmerge-debug-strings
4642 @opindex fno-merge-debug-strings
4643 Direct the linker to not merge together strings in the debugging
4644 information which are identical in different object files. Merging is
4645 not supported by all assemblers or linkers. Merging decreases the size
4646 of the debug information in the output file at the cost of increasing
4647 link processing time. Merging is enabled by default.
4649 @item -fdebug-prefix-map=@var{old}=@var{new}
4650 @opindex fdebug-prefix-map
4651 When compiling files in directory @file{@var{old}}, record debugging
4652 information describing them as in @file{@var{new}} instead.
4654 @item -fno-dwarf2-cfi-asm
4655 @opindex fdwarf2-cfi-asm
4656 @opindex fno-dwarf2-cfi-asm
4657 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4658 instead of using GAS @code{.cfi_*} directives.
4660 @cindex @command{prof}
4663 Generate extra code to write profile information suitable for the
4664 analysis program @command{prof}. You must use this option when compiling
4665 the source files you want data about, and you must also use it when
4668 @cindex @command{gprof}
4671 Generate extra code to write profile information suitable for the
4672 analysis program @command{gprof}. You must use this option when compiling
4673 the source files you want data about, and you must also use it when
4678 Makes the compiler print out each function name as it is compiled, and
4679 print some statistics about each pass when it finishes.
4682 @opindex ftime-report
4683 Makes the compiler print some statistics about the time consumed by each
4684 pass when it finishes.
4687 @opindex fmem-report
4688 Makes the compiler print some statistics about permanent memory
4689 allocation when it finishes.
4691 @item -fpre-ipa-mem-report
4692 @opindex fpre-ipa-mem-report
4693 @item -fpost-ipa-mem-report
4694 @opindex fpost-ipa-mem-report
4695 Makes the compiler print some statistics about permanent memory
4696 allocation before or after interprocedural optimization.
4698 @item -fprofile-arcs
4699 @opindex fprofile-arcs
4700 Add code so that program flow @dfn{arcs} are instrumented. During
4701 execution the program records how many times each branch and call is
4702 executed and how many times it is taken or returns. When the compiled
4703 program exits it saves this data to a file called
4704 @file{@var{auxname}.gcda} for each source file. The data may be used for
4705 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4706 test coverage analysis (@option{-ftest-coverage}). Each object file's
4707 @var{auxname} is generated from the name of the output file, if
4708 explicitly specified and it is not the final executable, otherwise it is
4709 the basename of the source file. In both cases any suffix is removed
4710 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4711 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4712 @xref{Cross-profiling}.
4714 @cindex @command{gcov}
4718 This option is used to compile and link code instrumented for coverage
4719 analysis. The option is a synonym for @option{-fprofile-arcs}
4720 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4721 linking). See the documentation for those options for more details.
4726 Compile the source files with @option{-fprofile-arcs} plus optimization
4727 and code generation options. For test coverage analysis, use the
4728 additional @option{-ftest-coverage} option. You do not need to profile
4729 every source file in a program.
4732 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4733 (the latter implies the former).
4736 Run the program on a representative workload to generate the arc profile
4737 information. This may be repeated any number of times. You can run
4738 concurrent instances of your program, and provided that the file system
4739 supports locking, the data files will be correctly updated. Also
4740 @code{fork} calls are detected and correctly handled (double counting
4744 For profile-directed optimizations, compile the source files again with
4745 the same optimization and code generation options plus
4746 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4747 Control Optimization}).
4750 For test coverage analysis, use @command{gcov} to produce human readable
4751 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4752 @command{gcov} documentation for further information.
4756 With @option{-fprofile-arcs}, for each function of your program GCC
4757 creates a program flow graph, then finds a spanning tree for the graph.
4758 Only arcs that are not on the spanning tree have to be instrumented: the
4759 compiler adds code to count the number of times that these arcs are
4760 executed. When an arc is the only exit or only entrance to a block, the
4761 instrumentation code can be added to the block; otherwise, a new basic
4762 block must be created to hold the instrumentation code.
4765 @item -ftest-coverage
4766 @opindex ftest-coverage
4767 Produce a notes file that the @command{gcov} code-coverage utility
4768 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4769 show program coverage. Each source file's note file is called
4770 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4771 above for a description of @var{auxname} and instructions on how to
4772 generate test coverage data. Coverage data will match the source files
4773 more closely, if you do not optimize.
4775 @item -fdbg-cnt-list
4776 @opindex fdbg-cnt-list
4777 Print the name and the counter upperbound for all debug counters.
4779 @item -fdbg-cnt=@var{counter-value-list}
4781 Set the internal debug counter upperbound. @var{counter-value-list}
4782 is a comma-separated list of @var{name}:@var{value} pairs
4783 which sets the upperbound of each debug counter @var{name} to @var{value}.
4784 All debug counters have the initial upperbound of @var{UINT_MAX},
4785 thus dbg_cnt() returns true always unless the upperbound is set by this option.
4786 e.g. With -fdbg-cnt=dce:10,tail_call:0
4787 dbg_cnt(dce) will return true only for first 10 invocations
4788 and dbg_cnt(tail_call) will return false always.
4790 @item -d@var{letters}
4791 @itemx -fdump-rtl-@var{pass}
4793 Says to make debugging dumps during compilation at times specified by
4794 @var{letters}. This is used for debugging the RTL-based passes of the
4795 compiler. The file names for most of the dumps are made by appending
4796 a pass number and a word to the @var{dumpname}, and the files are
4797 created in the directory of the output file. @var{dumpname} is
4798 generated from the name of the output file, if explicitly specified
4799 and it is not an executable, otherwise it is the basename of the
4800 source file. These switches may have different effects when
4801 @option{-E} is used for preprocessing.
4803 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
4804 @option{-d} option @var{letters}. Here are the possible
4805 letters for use in @var{pass} and @var{letters}, and their meanings:
4809 @item -fdump-rtl-alignments
4810 @opindex fdump-rtl-alignments
4811 Dump after branch alignments have been computed.
4813 @item -fdump-rtl-asmcons
4814 @opindex fdump-rtl-asmcons
4815 Dump after fixing rtl statements that have unsatisfied in/out constraints.
4817 @item -fdump-rtl-auto_inc_dec
4818 @opindex fdump-rtl-auto_inc_dec
4819 Dump after auto-inc-dec discovery. This pass is only run on
4820 architectures that have auto inc or auto dec instructions.
4822 @item -fdump-rtl-barriers
4823 @opindex fdump-rtl-barriers
4824 Dump after cleaning up the barrier instructions.
4826 @item -fdump-rtl-bbpart
4827 @opindex fdump-rtl-bbpart
4828 Dump after partitioning hot and cold basic blocks.
4830 @item -fdump-rtl-bbro
4831 @opindex fdump-rtl-bbro
4832 Dump after block reordering.
4834 @item -fdump-rtl-btl1
4835 @itemx -fdump-rtl-btl2
4836 @opindex fdump-rtl-btl2
4837 @opindex fdump-rtl-btl2
4838 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
4839 after the two branch
4840 target load optimization passes.
4842 @item -fdump-rtl-bypass
4843 @opindex fdump-rtl-bypass
4844 Dump after jump bypassing and control flow optimizations.
4846 @item -fdump-rtl-combine
4847 @opindex fdump-rtl-combine
4848 Dump after the RTL instruction combination pass.
4850 @item -fdump-rtl-compgotos
4851 @opindex fdump-rtl-compgotos
4852 Dump after duplicating the computed gotos.
4854 @item -fdump-rtl-ce1
4855 @itemx -fdump-rtl-ce2
4856 @itemx -fdump-rtl-ce3
4857 @opindex fdump-rtl-ce1
4858 @opindex fdump-rtl-ce2
4859 @opindex fdump-rtl-ce3
4860 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
4861 @option{-fdump-rtl-ce3} enable dumping after the three
4862 if conversion passes.
4864 @itemx -fdump-rtl-cprop_hardreg
4865 @opindex fdump-rtl-cprop_hardreg
4866 Dump after hard register copy propagation.
4868 @itemx -fdump-rtl-csa
4869 @opindex fdump-rtl-csa
4870 Dump after combining stack adjustments.
4872 @item -fdump-rtl-cse1
4873 @itemx -fdump-rtl-cse2
4874 @opindex fdump-rtl-cse1
4875 @opindex fdump-rtl-cse2
4876 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
4877 the two common sub-expression elimination passes.
4879 @itemx -fdump-rtl-dce
4880 @opindex fdump-rtl-dce
4881 Dump after the standalone dead code elimination passes.
4883 @itemx -fdump-rtl-dbr
4884 @opindex fdump-rtl-dbr
4885 Dump after delayed branch scheduling.
4887 @item -fdump-rtl-dce1
4888 @itemx -fdump-rtl-dce2
4889 @opindex fdump-rtl-dce1
4890 @opindex fdump-rtl-dce2
4891 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
4892 the two dead store elimination passes.
4895 @opindex fdump-rtl-eh
4896 Dump after finalization of EH handling code.
4898 @item -fdump-rtl-eh_ranges
4899 @opindex fdump-rtl-eh_ranges
4900 Dump after conversion of EH handling range regions.
4902 @item -fdump-rtl-expand
4903 @opindex fdump-rtl-expand
4904 Dump after RTL generation.
4906 @item -fdump-rtl-fwprop1
4907 @itemx -fdump-rtl-fwprop2
4908 @opindex fdump-rtl-fwprop1
4909 @opindex fdump-rtl-fwprop2
4910 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
4911 dumping after the two forward propagation passes.
4913 @item -fdump-rtl-gcse1
4914 @itemx -fdump-rtl-gcse2
4915 @opindex fdump-rtl-gcse1
4916 @opindex fdump-rtl-gcse2
4917 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
4918 after global common subexpression elimination.
4920 @item -fdump-rtl-init-regs
4921 @opindex fdump-rtl-init-regs
4922 Dump after the initialization of the registers.
4924 @item -fdump-rtl-initvals
4925 @opindex fdump-rtl-initvals
4926 Dump after the computation of the initial value sets.
4928 @itemx -fdump-rtl-into_cfglayout
4929 @opindex fdump-rtl-into_cfglayout
4930 Dump after converting to cfglayout mode.
4932 @item -fdump-rtl-ira
4933 @opindex fdump-rtl-ira
4934 Dump after iterated register allocation.
4936 @item -fdump-rtl-jump
4937 @opindex fdump-rtl-jump
4938 Dump after the second jump optimization.
4940 @item -fdump-rtl-loop2
4941 @opindex fdump-rtl-loop2
4942 @option{-fdump-rtl-loop2} enables dumping after the rtl
4943 loop optimization passes.
4945 @item -fdump-rtl-mach
4946 @opindex fdump-rtl-mach
4947 Dump after performing the machine dependent reorganization pass, if that
4950 @item -fdump-rtl-mode_sw
4951 @opindex fdump-rtl-mode_sw
4952 Dump after removing redundant mode switches.
4954 @item -fdump-rtl-rnreg
4955 @opindex fdump-rtl-rnreg
4956 Dump after register renumbering.
4958 @itemx -fdump-rtl-outof_cfglayout
4959 @opindex fdump-rtl-outof_cfglayout
4960 Dump after converting from cfglayout mode.
4962 @item -fdump-rtl-peephole2
4963 @opindex fdump-rtl-peephole2
4964 Dump after the peephole pass.
4966 @item -fdump-rtl-postreload
4967 @opindex fdump-rtl-postreload
4968 Dump after post-reload optimizations.
4970 @itemx -fdump-rtl-pro_and_epilogue
4971 @opindex fdump-rtl-pro_and_epilogue
4972 Dump after generating the function pro and epilogues.
4974 @item -fdump-rtl-regmove
4975 @opindex fdump-rtl-regmove
4976 Dump after the register move pass.
4978 @item -fdump-rtl-sched1
4979 @itemx -fdump-rtl-sched2
4980 @opindex fdump-rtl-sched1
4981 @opindex fdump-rtl-sched2
4982 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
4983 after the basic block scheduling passes.
4985 @item -fdump-rtl-see
4986 @opindex fdump-rtl-see
4987 Dump after sign extension elimination.
4989 @item -fdump-rtl-seqabstr
4990 @opindex fdump-rtl-seqabstr
4991 Dump after common sequence discovery.
4993 @item -fdump-rtl-shorten
4994 @opindex fdump-rtl-shorten
4995 Dump after shortening branches.
4997 @item -fdump-rtl-sibling
4998 @opindex fdump-rtl-sibling
4999 Dump after sibling call optimizations.
5001 @item -fdump-rtl-split1
5002 @itemx -fdump-rtl-split2
5003 @itemx -fdump-rtl-split3
5004 @itemx -fdump-rtl-split4
5005 @itemx -fdump-rtl-split5
5006 @opindex fdump-rtl-split1
5007 @opindex fdump-rtl-split2
5008 @opindex fdump-rtl-split3
5009 @opindex fdump-rtl-split4
5010 @opindex fdump-rtl-split5
5011 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
5012 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
5013 @option{-fdump-rtl-split5} enable dumping after five rounds of
5014 instruction splitting.
5016 @item -fdump-rtl-sms
5017 @opindex fdump-rtl-sms
5018 Dump after modulo scheduling. This pass is only run on some
5021 @item -fdump-rtl-stack
5022 @opindex fdump-rtl-stack
5023 Dump after conversion from GCC's "flat register file" registers to the
5024 x87's stack-like registers. This pass is only run on x86 variants.
5026 @item -fdump-rtl-subreg1
5027 @itemx -fdump-rtl-subreg2
5028 @opindex fdump-rtl-subreg1
5029 @opindex fdump-rtl-subreg2
5030 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
5031 the two subreg expansion passes.
5033 @item -fdump-rtl-unshare
5034 @opindex fdump-rtl-unshare
5035 Dump after all rtl has been unshared.
5037 @item -fdump-rtl-vartrack
5038 @opindex fdump-rtl-vartrack
5039 Dump after variable tracking.
5041 @item -fdump-rtl-vregs
5042 @opindex fdump-rtl-vregs
5043 Dump after converting virtual registers to hard registers.
5045 @item -fdump-rtl-web
5046 @opindex fdump-rtl-web
5047 Dump after live range splitting.
5049 @item -fdump-rtl-regclass
5050 @itemx -fdump-rtl-subregs_of_mode_init
5051 @itemx -fdump-rtl-subregs_of_mode_finish
5052 @itemx -fdump-rtl-dfinit
5053 @itemx -fdump-rtl-dfinish
5054 @opindex fdump-rtl-regclass
5055 @opindex fdump-rtl-subregs_of_mode_init
5056 @opindex fdump-rtl-subregs_of_mode_finish
5057 @opindex fdump-rtl-dfinit
5058 @opindex fdump-rtl-dfinish
5059 These dumps are defined but always produce empty files.
5061 @item -fdump-rtl-all
5062 @opindex fdump-rtl-all
5063 Produce all the dumps listed above.
5067 Annotate the assembler output with miscellaneous debugging information.
5071 Dump all macro definitions, at the end of preprocessing, in addition to
5076 Produce a core dump whenever an error occurs.
5080 Print statistics on memory usage, at the end of the run, to
5085 Annotate the assembler output with a comment indicating which
5086 pattern and alternative was used. The length of each instruction is
5091 Dump the RTL in the assembler output as a comment before each instruction.
5092 Also turns on @option{-dp} annotation.
5096 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
5097 dump a representation of the control flow graph suitable for viewing with VCG
5098 to @file{@var{file}.@var{pass}.vcg}.
5102 Just generate RTL for a function instead of compiling it. Usually used
5103 with @option{-fdump-rtl-expand}.
5107 Dump debugging information during parsing, to standard error.
5111 @opindex fdump-noaddr
5112 When doing debugging dumps, suppress address output. This makes it more
5113 feasible to use diff on debugging dumps for compiler invocations with
5114 different compiler binaries and/or different
5115 text / bss / data / heap / stack / dso start locations.
5117 @item -fdump-unnumbered
5118 @opindex fdump-unnumbered
5119 When doing debugging dumps, suppress instruction numbers and address output.
5120 This makes it more feasible to use diff on debugging dumps for compiler
5121 invocations with different options, in particular with and without
5124 @item -fdump-unnumbered-links
5125 @opindex fdump-unnumbered-links
5126 When doing debugging dumps (see @option{-d} option above), suppress
5127 instruction numbers for the links to the previous and next instructions
5130 @item -fdump-translation-unit @r{(C++ only)}
5131 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5132 @opindex fdump-translation-unit
5133 Dump a representation of the tree structure for the entire translation
5134 unit to a file. The file name is made by appending @file{.tu} to the
5135 source file name, and the file is created in the same directory as the
5136 output file. If the @samp{-@var{options}} form is used, @var{options}
5137 controls the details of the dump as described for the
5138 @option{-fdump-tree} options.
5140 @item -fdump-class-hierarchy @r{(C++ only)}
5141 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5142 @opindex fdump-class-hierarchy
5143 Dump a representation of each class's hierarchy and virtual function
5144 table layout to a file. The file name is made by appending
5145 @file{.class} to the source file name, and the file is created in the
5146 same directory as the output file. If the @samp{-@var{options}} form
5147 is used, @var{options} controls the details of the dump as described
5148 for the @option{-fdump-tree} options.
5150 @item -fdump-ipa-@var{switch}
5152 Control the dumping at various stages of inter-procedural analysis
5153 language tree to a file. The file name is generated by appending a
5154 switch specific suffix to the source file name, and the file is created
5155 in the same directory as the output file. The following dumps are
5160 Enables all inter-procedural analysis dumps.
5163 Dumps information about call-graph optimization, unused function removal,
5164 and inlining decisions.
5167 Dump after function inlining.
5171 @item -fdump-statistics-@var{option}
5172 @opindex fdump-statistics
5173 Enable and control dumping of pass statistics in a separate file. The
5174 file name is generated by appending a suffix ending in
5175 @samp{.statistics} to the source file name, and the file is created in
5176 the same directory as the output file. If the @samp{-@var{option}}
5177 form is used, @samp{-stats} will cause counters to be summed over the
5178 whole compilation unit while @samp{-details} will dump every event as
5179 the passes generate them. The default with no option is to sum
5180 counters for each function compiled.
5182 @item -fdump-tree-@var{switch}
5183 @itemx -fdump-tree-@var{switch}-@var{options}
5185 Control the dumping at various stages of processing the intermediate
5186 language tree to a file. The file name is generated by appending a
5187 switch specific suffix to the source file name, and the file is
5188 created in the same directory as the output file. If the
5189 @samp{-@var{options}} form is used, @var{options} is a list of
5190 @samp{-} separated options that control the details of the dump. Not
5191 all options are applicable to all dumps, those which are not
5192 meaningful will be ignored. The following options are available
5196 Print the address of each node. Usually this is not meaningful as it
5197 changes according to the environment and source file. Its primary use
5198 is for tying up a dump file with a debug environment.
5200 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
5201 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
5202 use working backward from mangled names in the assembly file.
5204 Inhibit dumping of members of a scope or body of a function merely
5205 because that scope has been reached. Only dump such items when they
5206 are directly reachable by some other path. When dumping pretty-printed
5207 trees, this option inhibits dumping the bodies of control structures.
5209 Print a raw representation of the tree. By default, trees are
5210 pretty-printed into a C-like representation.
5212 Enable more detailed dumps (not honored by every dump option).
5214 Enable dumping various statistics about the pass (not honored by every dump
5217 Enable showing basic block boundaries (disabled in raw dumps).
5219 Enable showing virtual operands for every statement.
5221 Enable showing line numbers for statements.
5223 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5225 Enable showing the tree dump for each statement.
5227 Enable showing the EH region number holding each statement.
5229 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5230 and @option{lineno}.
5233 The following tree dumps are possible:
5237 @opindex fdump-tree-original
5238 Dump before any tree based optimization, to @file{@var{file}.original}.
5241 @opindex fdump-tree-optimized
5242 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5245 @opindex fdump-tree-gimple
5246 Dump each function before and after the gimplification pass to a file. The
5247 file name is made by appending @file{.gimple} to the source file name.
5250 @opindex fdump-tree-cfg
5251 Dump the control flow graph of each function to a file. The file name is
5252 made by appending @file{.cfg} to the source file name.
5255 @opindex fdump-tree-vcg
5256 Dump the control flow graph of each function to a file in VCG format. The
5257 file name is made by appending @file{.vcg} to the source file name. Note
5258 that if the file contains more than one function, the generated file cannot
5259 be used directly by VCG@. You will need to cut and paste each function's
5260 graph into its own separate file first.
5263 @opindex fdump-tree-ch
5264 Dump each function after copying loop headers. The file name is made by
5265 appending @file{.ch} to the source file name.
5268 @opindex fdump-tree-ssa
5269 Dump SSA related information to a file. The file name is made by appending
5270 @file{.ssa} to the source file name.
5273 @opindex fdump-tree-alias
5274 Dump aliasing information for each function. The file name is made by
5275 appending @file{.alias} to the source file name.
5278 @opindex fdump-tree-ccp
5279 Dump each function after CCP@. The file name is made by appending
5280 @file{.ccp} to the source file name.
5283 @opindex fdump-tree-storeccp
5284 Dump each function after STORE-CCP@. The file name is made by appending
5285 @file{.storeccp} to the source file name.
5288 @opindex fdump-tree-pre
5289 Dump trees after partial redundancy elimination. The file name is made
5290 by appending @file{.pre} to the source file name.
5293 @opindex fdump-tree-fre
5294 Dump trees after full redundancy elimination. The file name is made
5295 by appending @file{.fre} to the source file name.
5298 @opindex fdump-tree-copyprop
5299 Dump trees after copy propagation. The file name is made
5300 by appending @file{.copyprop} to the source file name.
5302 @item store_copyprop
5303 @opindex fdump-tree-store_copyprop
5304 Dump trees after store copy-propagation. The file name is made
5305 by appending @file{.store_copyprop} to the source file name.
5308 @opindex fdump-tree-dce
5309 Dump each function after dead code elimination. The file name is made by
5310 appending @file{.dce} to the source file name.
5313 @opindex fdump-tree-mudflap
5314 Dump each function after adding mudflap instrumentation. The file name is
5315 made by appending @file{.mudflap} to the source file name.
5318 @opindex fdump-tree-sra
5319 Dump each function after performing scalar replacement of aggregates. The
5320 file name is made by appending @file{.sra} to the source file name.
5323 @opindex fdump-tree-sink
5324 Dump each function after performing code sinking. The file name is made
5325 by appending @file{.sink} to the source file name.
5328 @opindex fdump-tree-dom
5329 Dump each function after applying dominator tree optimizations. The file
5330 name is made by appending @file{.dom} to the source file name.
5333 @opindex fdump-tree-dse
5334 Dump each function after applying dead store elimination. The file
5335 name is made by appending @file{.dse} to the source file name.
5338 @opindex fdump-tree-phiopt
5339 Dump each function after optimizing PHI nodes into straightline code. The file
5340 name is made by appending @file{.phiopt} to the source file name.
5343 @opindex fdump-tree-forwprop
5344 Dump each function after forward propagating single use variables. The file
5345 name is made by appending @file{.forwprop} to the source file name.
5348 @opindex fdump-tree-copyrename
5349 Dump each function after applying the copy rename optimization. The file
5350 name is made by appending @file{.copyrename} to the source file name.
5353 @opindex fdump-tree-nrv
5354 Dump each function after applying the named return value optimization on
5355 generic trees. The file name is made by appending @file{.nrv} to the source
5359 @opindex fdump-tree-vect
5360 Dump each function after applying vectorization of loops. The file name is
5361 made by appending @file{.vect} to the source file name.
5364 @opindex fdump-tree-vrp
5365 Dump each function after Value Range Propagation (VRP). The file name
5366 is made by appending @file{.vrp} to the source file name.
5369 @opindex fdump-tree-all
5370 Enable all the available tree dumps with the flags provided in this option.
5373 @item -ftree-vectorizer-verbose=@var{n}
5374 @opindex ftree-vectorizer-verbose
5375 This option controls the amount of debugging output the vectorizer prints.
5376 This information is written to standard error, unless
5377 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5378 in which case it is output to the usual dump listing file, @file{.vect}.
5379 For @var{n}=0 no diagnostic information is reported.
5380 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5381 and the total number of loops that got vectorized.
5382 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5383 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5384 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5385 level that @option{-fdump-tree-vect-stats} uses.
5386 Higher verbosity levels mean either more information dumped for each
5387 reported loop, or same amount of information reported for more loops:
5388 If @var{n}=3, alignment related information is added to the reports.
5389 If @var{n}=4, data-references related information (e.g.@: memory dependences,
5390 memory access-patterns) is added to the reports.
5391 If @var{n}=5, the vectorizer reports also non-vectorized inner-most loops
5392 that did not pass the first analysis phase (i.e., may not be countable, or
5393 may have complicated control-flow).
5394 If @var{n}=6, the vectorizer reports also non-vectorized nested loops.
5395 For @var{n}=7, all the information the vectorizer generates during its
5396 analysis and transformation is reported. This is the same verbosity level
5397 that @option{-fdump-tree-vect-details} uses.
5399 @item -frandom-seed=@var{string}
5400 @opindex frandom-seed
5401 This option provides a seed that GCC uses when it would otherwise use
5402 random numbers. It is used to generate certain symbol names
5403 that have to be different in every compiled file. It is also used to
5404 place unique stamps in coverage data files and the object files that
5405 produce them. You can use the @option{-frandom-seed} option to produce
5406 reproducibly identical object files.
5408 The @var{string} should be different for every file you compile.
5410 @item -fsched-verbose=@var{n}
5411 @opindex fsched-verbose
5412 On targets that use instruction scheduling, this option controls the
5413 amount of debugging output the scheduler prints. This information is
5414 written to standard error, unless @option{-fdump-rtl-sched1} or
5415 @option{-fdump-rtl-sched2} is specified, in which case it is output
5416 to the usual dump listing file, @file{.sched} or @file{.sched2}
5417 respectively. However for @var{n} greater than nine, the output is
5418 always printed to standard error.
5420 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5421 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5422 For @var{n} greater than one, it also output basic block probabilities,
5423 detailed ready list information and unit/insn info. For @var{n} greater
5424 than two, it includes RTL at abort point, control-flow and regions info.
5425 And for @var{n} over four, @option{-fsched-verbose} also includes
5429 @itemx -save-temps=cwd
5431 Store the usual ``temporary'' intermediate files permanently; place them
5432 in the current directory and name them based on the source file. Thus,
5433 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5434 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5435 preprocessed @file{foo.i} output file even though the compiler now
5436 normally uses an integrated preprocessor.
5438 When used in combination with the @option{-x} command line option,
5439 @option{-save-temps} is sensible enough to avoid over writing an
5440 input source file with the same extension as an intermediate file.
5441 The corresponding intermediate file may be obtained by renaming the
5442 source file before using @option{-save-temps}.
5444 If you invoke GCC in parallel, compiling several different source
5445 files that share a common base name in different subdirectories or the
5446 same source file compiled for multiple output destinations, it is
5447 likely that the different parallel compilers will interfere with each
5448 other, and overwrite the temporary files. For instance:
5451 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5452 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5455 may result in @file{foo.i} and @file{foo.o} being written to
5456 simultaneously by both compilers.
5458 @item -save-temps=obj
5459 @opindex save-temps=obj
5460 Store the usual ``temporary'' intermediate files permanently. If the
5461 @option{-o} option is used, the temporary files are based on the
5462 object file. If the @option{-o} option is not used, the
5463 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5468 gcc -save-temps=obj -c foo.c
5469 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5470 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5473 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5474 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5475 @file{dir2/yfoobar.o}.
5477 @item -time@r{[}=@var{file}@r{]}
5479 Report the CPU time taken by each subprocess in the compilation
5480 sequence. For C source files, this is the compiler proper and assembler
5481 (plus the linker if linking is done).
5483 Without the specification of an output file, the output looks like this:
5490 The first number on each line is the ``user time'', that is time spent
5491 executing the program itself. The second number is ``system time'',
5492 time spent executing operating system routines on behalf of the program.
5493 Both numbers are in seconds.
5495 With the specification of an output file, the output is appended to the
5496 named file, and it looks like this:
5499 0.12 0.01 cc1 @var{options}
5500 0.00 0.01 as @var{options}
5503 The ``user time'' and the ``system time'' are moved before the program
5504 name, and the options passed to the program are displayed, so that one
5505 can later tell what file was being compiled, and with which options.
5507 @item -fvar-tracking
5508 @opindex fvar-tracking
5509 Run variable tracking pass. It computes where variables are stored at each
5510 position in code. Better debugging information is then generated
5511 (if the debugging information format supports this information).
5513 It is enabled by default when compiling with optimization (@option{-Os},
5514 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5515 the debug info format supports it.
5517 @item -fvar-tracking-assignments
5518 @opindex fvar-tracking-assignments
5519 @opindex fno-var-tracking-assignments
5520 Annotate assignments to user variables early in the compilation and
5521 attempt to carry the annotations over throughout the compilation all the
5522 way to the end, in an attempt to improve debug information while
5523 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
5525 It can be enabled even if var-tracking is disabled, in which case
5526 annotations will be created and maintained, but discarded at the end.
5528 @item -fvar-tracking-assignments-toggle
5529 @opindex fvar-tracking-assignments-toggle
5530 @opindex fno-var-tracking-assignments-toggle
5531 Toggle @option{-fvar-tracking-assignments}, in the same way that
5532 @option{-gtoggle} toggles @option{-g}.
5534 @item -print-file-name=@var{library}
5535 @opindex print-file-name
5536 Print the full absolute name of the library file @var{library} that
5537 would be used when linking---and don't do anything else. With this
5538 option, GCC does not compile or link anything; it just prints the
5541 @item -print-multi-directory
5542 @opindex print-multi-directory
5543 Print the directory name corresponding to the multilib selected by any
5544 other switches present in the command line. This directory is supposed
5545 to exist in @env{GCC_EXEC_PREFIX}.
5547 @item -print-multi-lib
5548 @opindex print-multi-lib
5549 Print the mapping from multilib directory names to compiler switches
5550 that enable them. The directory name is separated from the switches by
5551 @samp{;}, and each switch starts with an @samp{@@} instead of the
5552 @samp{-}, without spaces between multiple switches. This is supposed to
5553 ease shell-processing.
5555 @item -print-multi-os-directory
5556 @opindex print-multi-os-directory
5557 Print the path to OS libraries for the selected
5558 multilib, relative to some @file{lib} subdirectory. If OS libraries are
5559 present in the @file{lib} subdirectory and no multilibs are used, this is
5560 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
5561 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
5562 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
5563 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
5565 @item -print-prog-name=@var{program}
5566 @opindex print-prog-name
5567 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5569 @item -print-libgcc-file-name
5570 @opindex print-libgcc-file-name
5571 Same as @option{-print-file-name=libgcc.a}.
5573 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5574 but you do want to link with @file{libgcc.a}. You can do
5577 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5580 @item -print-search-dirs
5581 @opindex print-search-dirs
5582 Print the name of the configured installation directory and a list of
5583 program and library directories @command{gcc} will search---and don't do anything else.
5585 This is useful when @command{gcc} prints the error message
5586 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5587 To resolve this you either need to put @file{cpp0} and the other compiler
5588 components where @command{gcc} expects to find them, or you can set the environment
5589 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5590 Don't forget the trailing @samp{/}.
5591 @xref{Environment Variables}.
5593 @item -print-sysroot
5594 @opindex print-sysroot
5595 Print the target sysroot directory that will be used during
5596 compilation. This is the target sysroot specified either at configure
5597 time or using the @option{--sysroot} option, possibly with an extra
5598 suffix that depends on compilation options. If no target sysroot is
5599 specified, the option prints nothing.
5601 @item -print-sysroot-headers-suffix
5602 @opindex print-sysroot-headers-suffix
5603 Print the suffix added to the target sysroot when searching for
5604 headers, or give an error if the compiler is not configured with such
5605 a suffix---and don't do anything else.
5608 @opindex dumpmachine
5609 Print the compiler's target machine (for example,
5610 @samp{i686-pc-linux-gnu})---and don't do anything else.
5613 @opindex dumpversion
5614 Print the compiler version (for example, @samp{3.0})---and don't do
5619 Print the compiler's built-in specs---and don't do anything else. (This
5620 is used when GCC itself is being built.) @xref{Spec Files}.
5622 @item -feliminate-unused-debug-types
5623 @opindex feliminate-unused-debug-types
5624 Normally, when producing DWARF2 output, GCC will emit debugging
5625 information for all types declared in a compilation
5626 unit, regardless of whether or not they are actually used
5627 in that compilation unit. Sometimes this is useful, such as
5628 if, in the debugger, you want to cast a value to a type that is
5629 not actually used in your program (but is declared). More often,
5630 however, this results in a significant amount of wasted space.
5631 With this option, GCC will avoid producing debug symbol output
5632 for types that are nowhere used in the source file being compiled.
5635 @node Optimize Options
5636 @section Options That Control Optimization
5637 @cindex optimize options
5638 @cindex options, optimization
5640 These options control various sorts of optimizations.
5642 Without any optimization option, the compiler's goal is to reduce the
5643 cost of compilation and to make debugging produce the expected
5644 results. Statements are independent: if you stop the program with a
5645 breakpoint between statements, you can then assign a new value to any
5646 variable or change the program counter to any other statement in the
5647 function and get exactly the results you would expect from the source
5650 Turning on optimization flags makes the compiler attempt to improve
5651 the performance and/or code size at the expense of compilation time
5652 and possibly the ability to debug the program.
5654 The compiler performs optimization based on the knowledge it has of the
5655 program. Compiling multiple files at once to a single output file mode allows
5656 the compiler to use information gained from all of the files when compiling
5659 Not all optimizations are controlled directly by a flag. Only
5660 optimizations that have a flag are listed in this section.
5662 Most of the optimizations are not enabled if a @option{-O} level is not set on
5663 the command line, even if individual optimization flags are specified.
5665 Depending on the target and how GCC was configured, a slightly different
5666 set of optimizations may be enabled at each @option{-O} level than
5667 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
5668 to find out the exact set of optimizations that are enabled at each level.
5669 @xref{Overall Options}, for examples.
5676 Optimize. Optimizing compilation takes somewhat more time, and a lot
5677 more memory for a large function.
5679 With @option{-O}, the compiler tries to reduce code size and execution
5680 time, without performing any optimizations that take a great deal of
5683 @option{-O} turns on the following optimization flags:
5686 -fcprop-registers @gol
5689 -fdelayed-branch @gol
5691 -fguess-branch-probability @gol
5692 -fif-conversion2 @gol
5693 -fif-conversion @gol
5694 -fipa-pure-const @gol
5695 -fipa-reference @gol
5697 -fsplit-wide-types @gol
5698 -ftree-builtin-call-dce @gol
5701 -ftree-copyrename @gol
5703 -ftree-dominator-opts @gol
5705 -ftree-forwprop @gol
5713 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5714 where doing so does not interfere with debugging.
5718 Optimize even more. GCC performs nearly all supported optimizations
5719 that do not involve a space-speed tradeoff.
5720 As compared to @option{-O}, this option increases both compilation time
5721 and the performance of the generated code.
5723 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5724 also turns on the following optimization flags:
5725 @gccoptlist{-fthread-jumps @gol
5726 -falign-functions -falign-jumps @gol
5727 -falign-loops -falign-labels @gol
5730 -fcse-follow-jumps -fcse-skip-blocks @gol
5731 -fdelete-null-pointer-checks @gol
5732 -fexpensive-optimizations @gol
5733 -fgcse -fgcse-lm @gol
5734 -finline-small-functions @gol
5735 -findirect-inlining @gol
5737 -foptimize-sibling-calls @gol
5740 -freorder-blocks -freorder-functions @gol
5741 -frerun-cse-after-loop @gol
5742 -fsched-interblock -fsched-spec @gol
5743 -fschedule-insns -fschedule-insns2 @gol
5744 -fstrict-aliasing -fstrict-overflow @gol
5745 -ftree-switch-conversion @gol
5749 Please note the warning under @option{-fgcse} about
5750 invoking @option{-O2} on programs that use computed gotos.
5754 Optimize yet more. @option{-O3} turns on all optimizations specified
5755 by @option{-O2} and also turns on the @option{-finline-functions},
5756 @option{-funswitch-loops}, @option{-fpredictive-commoning},
5757 @option{-fgcse-after-reload} and @option{-ftree-vectorize} options.
5761 Reduce compilation time and make debugging produce the expected
5762 results. This is the default.
5766 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5767 do not typically increase code size. It also performs further
5768 optimizations designed to reduce code size.
5770 @option{-Os} disables the following optimization flags:
5771 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5772 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5773 -fprefetch-loop-arrays -ftree-vect-loop-version}
5775 If you use multiple @option{-O} options, with or without level numbers,
5776 the last such option is the one that is effective.
5779 Options of the form @option{-f@var{flag}} specify machine-independent
5780 flags. Most flags have both positive and negative forms; the negative
5781 form of @option{-ffoo} would be @option{-fno-foo}. In the table
5782 below, only one of the forms is listed---the one you typically will
5783 use. You can figure out the other form by either removing @samp{no-}
5786 The following options control specific optimizations. They are either
5787 activated by @option{-O} options or are related to ones that are. You
5788 can use the following flags in the rare cases when ``fine-tuning'' of
5789 optimizations to be performed is desired.
5792 @item -fno-default-inline
5793 @opindex fno-default-inline
5794 Do not make member functions inline by default merely because they are
5795 defined inside the class scope (C++ only). Otherwise, when you specify
5796 @w{@option{-O}}, member functions defined inside class scope are compiled
5797 inline by default; i.e., you don't need to add @samp{inline} in front of
5798 the member function name.
5800 @item -fno-defer-pop
5801 @opindex fno-defer-pop
5802 Always pop the arguments to each function call as soon as that function
5803 returns. For machines which must pop arguments after a function call,
5804 the compiler normally lets arguments accumulate on the stack for several
5805 function calls and pops them all at once.
5807 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5809 @item -fforward-propagate
5810 @opindex fforward-propagate
5811 Perform a forward propagation pass on RTL@. The pass tries to combine two
5812 instructions and checks if the result can be simplified. If loop unrolling
5813 is active, two passes are performed and the second is scheduled after
5816 This option is enabled by default at optimization levels @option{-O},
5817 @option{-O2}, @option{-O3}, @option{-Os}.
5819 @item -fomit-frame-pointer
5820 @opindex fomit-frame-pointer
5821 Don't keep the frame pointer in a register for functions that
5822 don't need one. This avoids the instructions to save, set up and
5823 restore frame pointers; it also makes an extra register available
5824 in many functions. @strong{It also makes debugging impossible on
5827 On some machines, such as the VAX, this flag has no effect, because
5828 the standard calling sequence automatically handles the frame pointer
5829 and nothing is saved by pretending it doesn't exist. The
5830 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
5831 whether a target machine supports this flag. @xref{Registers,,Register
5832 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
5834 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5836 @item -foptimize-sibling-calls
5837 @opindex foptimize-sibling-calls
5838 Optimize sibling and tail recursive calls.
5840 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5844 Don't pay attention to the @code{inline} keyword. Normally this option
5845 is used to keep the compiler from expanding any functions inline.
5846 Note that if you are not optimizing, no functions can be expanded inline.
5848 @item -finline-small-functions
5849 @opindex finline-small-functions
5850 Integrate functions into their callers when their body is smaller than expected
5851 function call code (so overall size of program gets smaller). The compiler
5852 heuristically decides which functions are simple enough to be worth integrating
5855 Enabled at level @option{-O2}.
5857 @item -findirect-inlining
5858 @opindex findirect-inlining
5859 Inline also indirect calls that are discovered to be known at compile
5860 time thanks to previous inlining. This option has any effect only
5861 when inlining itself is turned on by the @option{-finline-functions}
5862 or @option{-finline-small-functions} options.
5864 Enabled at level @option{-O2}.
5866 @item -finline-functions
5867 @opindex finline-functions
5868 Integrate all simple functions into their callers. The compiler
5869 heuristically decides which functions are simple enough to be worth
5870 integrating in this way.
5872 If all calls to a given function are integrated, and the function is
5873 declared @code{static}, then the function is normally not output as
5874 assembler code in its own right.
5876 Enabled at level @option{-O3}.
5878 @item -finline-functions-called-once
5879 @opindex finline-functions-called-once
5880 Consider all @code{static} functions called once for inlining into their
5881 caller even if they are not marked @code{inline}. If a call to a given
5882 function is integrated, then the function is not output as assembler code
5885 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
5887 @item -fearly-inlining
5888 @opindex fearly-inlining
5889 Inline functions marked by @code{always_inline} and functions whose body seems
5890 smaller than the function call overhead early before doing
5891 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
5892 makes profiling significantly cheaper and usually inlining faster on programs
5893 having large chains of nested wrapper functions.
5899 Perform interprocedural scalar replacement of aggregates, removal of
5900 unused parameters and replacement of parameters passed by reference
5901 by parameters passed by value.
5903 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
5905 @item -finline-limit=@var{n}
5906 @opindex finline-limit
5907 By default, GCC limits the size of functions that can be inlined. This flag
5908 allows coarse control of this limit. @var{n} is the size of functions that
5909 can be inlined in number of pseudo instructions.
5911 Inlining is actually controlled by a number of parameters, which may be
5912 specified individually by using @option{--param @var{name}=@var{value}}.
5913 The @option{-finline-limit=@var{n}} option sets some of these parameters
5917 @item max-inline-insns-single
5918 is set to @var{n}/2.
5919 @item max-inline-insns-auto
5920 is set to @var{n}/2.
5923 See below for a documentation of the individual
5924 parameters controlling inlining and for the defaults of these parameters.
5926 @emph{Note:} there may be no value to @option{-finline-limit} that results
5927 in default behavior.
5929 @emph{Note:} pseudo instruction represents, in this particular context, an
5930 abstract measurement of function's size. In no way does it represent a count
5931 of assembly instructions and as such its exact meaning might change from one
5932 release to an another.
5934 @item -fkeep-inline-functions
5935 @opindex fkeep-inline-functions
5936 In C, emit @code{static} functions that are declared @code{inline}
5937 into the object file, even if the function has been inlined into all
5938 of its callers. This switch does not affect functions using the
5939 @code{extern inline} extension in GNU C89@. In C++, emit any and all
5940 inline functions into the object file.
5942 @item -fkeep-static-consts
5943 @opindex fkeep-static-consts
5944 Emit variables declared @code{static const} when optimization isn't turned
5945 on, even if the variables aren't referenced.
5947 GCC enables this option by default. If you want to force the compiler to
5948 check if the variable was referenced, regardless of whether or not
5949 optimization is turned on, use the @option{-fno-keep-static-consts} option.
5951 @item -fmerge-constants
5952 @opindex fmerge-constants
5953 Attempt to merge identical constants (string constants and floating point
5954 constants) across compilation units.
5956 This option is the default for optimized compilation if the assembler and
5957 linker support it. Use @option{-fno-merge-constants} to inhibit this
5960 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5962 @item -fmerge-all-constants
5963 @opindex fmerge-all-constants
5964 Attempt to merge identical constants and identical variables.
5966 This option implies @option{-fmerge-constants}. In addition to
5967 @option{-fmerge-constants} this considers e.g.@: even constant initialized
5968 arrays or initialized constant variables with integral or floating point
5969 types. Languages like C or C++ require each variable, including multiple
5970 instances of the same variable in recursive calls, to have distinct locations,
5971 so using this option will result in non-conforming
5974 @item -fmodulo-sched
5975 @opindex fmodulo-sched
5976 Perform swing modulo scheduling immediately before the first scheduling
5977 pass. This pass looks at innermost loops and reorders their
5978 instructions by overlapping different iterations.
5980 @item -fmodulo-sched-allow-regmoves
5981 @opindex fmodulo-sched-allow-regmoves
5982 Perform more aggressive SMS based modulo scheduling with register moves
5983 allowed. By setting this flag certain anti-dependences edges will be
5984 deleted which will trigger the generation of reg-moves based on the
5985 life-range analysis. This option is effective only with
5986 @option{-fmodulo-sched} enabled.
5988 @item -fno-branch-count-reg
5989 @opindex fno-branch-count-reg
5990 Do not use ``decrement and branch'' instructions on a count register,
5991 but instead generate a sequence of instructions that decrement a
5992 register, compare it against zero, then branch based upon the result.
5993 This option is only meaningful on architectures that support such
5994 instructions, which include x86, PowerPC, IA-64 and S/390.
5996 The default is @option{-fbranch-count-reg}.
5998 @item -fno-function-cse
5999 @opindex fno-function-cse
6000 Do not put function addresses in registers; make each instruction that
6001 calls a constant function contain the function's address explicitly.
6003 This option results in less efficient code, but some strange hacks
6004 that alter the assembler output may be confused by the optimizations
6005 performed when this option is not used.
6007 The default is @option{-ffunction-cse}
6009 @item -fno-zero-initialized-in-bss
6010 @opindex fno-zero-initialized-in-bss
6011 If the target supports a BSS section, GCC by default puts variables that
6012 are initialized to zero into BSS@. This can save space in the resulting
6015 This option turns off this behavior because some programs explicitly
6016 rely on variables going to the data section. E.g., so that the
6017 resulting executable can find the beginning of that section and/or make
6018 assumptions based on that.
6020 The default is @option{-fzero-initialized-in-bss}.
6022 @item -fmudflap -fmudflapth -fmudflapir
6026 @cindex bounds checking
6028 For front-ends that support it (C and C++), instrument all risky
6029 pointer/array dereferencing operations, some standard library
6030 string/heap functions, and some other associated constructs with
6031 range/validity tests. Modules so instrumented should be immune to
6032 buffer overflows, invalid heap use, and some other classes of C/C++
6033 programming errors. The instrumentation relies on a separate runtime
6034 library (@file{libmudflap}), which will be linked into a program if
6035 @option{-fmudflap} is given at link time. Run-time behavior of the
6036 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
6037 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
6040 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
6041 link if your program is multi-threaded. Use @option{-fmudflapir}, in
6042 addition to @option{-fmudflap} or @option{-fmudflapth}, if
6043 instrumentation should ignore pointer reads. This produces less
6044 instrumentation (and therefore faster execution) and still provides
6045 some protection against outright memory corrupting writes, but allows
6046 erroneously read data to propagate within a program.
6048 @item -fthread-jumps
6049 @opindex fthread-jumps
6050 Perform optimizations where we check to see if a jump branches to a
6051 location where another comparison subsumed by the first is found. If
6052 so, the first branch is redirected to either the destination of the
6053 second branch or a point immediately following it, depending on whether
6054 the condition is known to be true or false.
6056 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6058 @item -fsplit-wide-types
6059 @opindex fsplit-wide-types
6060 When using a type that occupies multiple registers, such as @code{long
6061 long} on a 32-bit system, split the registers apart and allocate them
6062 independently. This normally generates better code for those types,
6063 but may make debugging more difficult.
6065 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
6068 @item -fcse-follow-jumps
6069 @opindex fcse-follow-jumps
6070 In common subexpression elimination (CSE), scan through jump instructions
6071 when the target of the jump is not reached by any other path. For
6072 example, when CSE encounters an @code{if} statement with an
6073 @code{else} clause, CSE will follow the jump when the condition
6076 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6078 @item -fcse-skip-blocks
6079 @opindex fcse-skip-blocks
6080 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
6081 follow jumps which conditionally skip over blocks. When CSE
6082 encounters a simple @code{if} statement with no else clause,
6083 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6084 body of the @code{if}.
6086 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6088 @item -frerun-cse-after-loop
6089 @opindex frerun-cse-after-loop
6090 Re-run common subexpression elimination after loop optimizations has been
6093 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6097 Perform a global common subexpression elimination pass.
6098 This pass also performs global constant and copy propagation.
6100 @emph{Note:} When compiling a program using computed gotos, a GCC
6101 extension, you may get better runtime performance if you disable
6102 the global common subexpression elimination pass by adding
6103 @option{-fno-gcse} to the command line.
6105 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6109 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
6110 attempt to move loads which are only killed by stores into themselves. This
6111 allows a loop containing a load/store sequence to be changed to a load outside
6112 the loop, and a copy/store within the loop.
6114 Enabled by default when gcse is enabled.
6118 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6119 global common subexpression elimination. This pass will attempt to move
6120 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6121 loops containing a load/store sequence can be changed to a load before
6122 the loop and a store after the loop.
6124 Not enabled at any optimization level.
6128 When @option{-fgcse-las} is enabled, the global common subexpression
6129 elimination pass eliminates redundant loads that come after stores to the
6130 same memory location (both partial and full redundancies).
6132 Not enabled at any optimization level.
6134 @item -fgcse-after-reload
6135 @opindex fgcse-after-reload
6136 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6137 pass is performed after reload. The purpose of this pass is to cleanup
6140 @item -funsafe-loop-optimizations
6141 @opindex funsafe-loop-optimizations
6142 If given, the loop optimizer will assume that loop indices do not
6143 overflow, and that the loops with nontrivial exit condition are not
6144 infinite. This enables a wider range of loop optimizations even if
6145 the loop optimizer itself cannot prove that these assumptions are valid.
6146 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
6147 if it finds this kind of loop.
6149 @item -fcrossjumping
6150 @opindex fcrossjumping
6151 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6152 resulting code may or may not perform better than without cross-jumping.
6154 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6156 @item -fauto-inc-dec
6157 @opindex fauto-inc-dec
6158 Combine increments or decrements of addresses with memory accesses.
6159 This pass is always skipped on architectures that do not have
6160 instructions to support this. Enabled by default at @option{-O} and
6161 higher on architectures that support this.
6165 Perform dead code elimination (DCE) on RTL@.
6166 Enabled by default at @option{-O} and higher.
6170 Perform dead store elimination (DSE) on RTL@.
6171 Enabled by default at @option{-O} and higher.
6173 @item -fif-conversion
6174 @opindex fif-conversion
6175 Attempt to transform conditional jumps into branch-less equivalents. This
6176 include use of conditional moves, min, max, set flags and abs instructions, and
6177 some tricks doable by standard arithmetics. The use of conditional execution
6178 on chips where it is available is controlled by @code{if-conversion2}.
6180 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6182 @item -fif-conversion2
6183 @opindex fif-conversion2
6184 Use conditional execution (where available) to transform conditional jumps into
6185 branch-less equivalents.
6187 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6189 @item -fdelete-null-pointer-checks
6190 @opindex fdelete-null-pointer-checks
6191 Assume that programs cannot safely dereference null pointers, and that
6192 no code or data element resides there. This enables simple constant
6193 folding optimizations at all optimization levels. In addition, other
6194 optimization passes in GCC use this flag to control global dataflow
6195 analyses that eliminate useless checks for null pointers; these assume
6196 that if a pointer is checked after it has already been dereferenced,
6199 Note however that in some environments this assumption is not true.
6200 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6201 for programs which depend on that behavior.
6203 Some targets, especially embedded ones, disable this option at all levels.
6204 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6205 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6206 are enabled independently at different optimization levels.
6208 @item -fexpensive-optimizations
6209 @opindex fexpensive-optimizations
6210 Perform a number of minor optimizations that are relatively expensive.
6212 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6214 @item -foptimize-register-move
6216 @opindex foptimize-register-move
6218 Attempt to reassign register numbers in move instructions and as
6219 operands of other simple instructions in order to maximize the amount of
6220 register tying. This is especially helpful on machines with two-operand
6223 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6226 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6228 @item -fira-algorithm=@var{algorithm}
6229 Use specified coloring algorithm for the integrated register
6230 allocator. The @var{algorithm} argument should be @code{priority} or
6231 @code{CB}. The first algorithm specifies Chow's priority coloring,
6232 the second one specifies Chaitin-Briggs coloring. The second
6233 algorithm can be unimplemented for some architectures. If it is
6234 implemented, it is the default because Chaitin-Briggs coloring as a
6235 rule generates a better code.
6237 @item -fira-region=@var{region}
6238 Use specified regions for the integrated register allocator. The
6239 @var{region} argument should be one of @code{all}, @code{mixed}, or
6240 @code{one}. The first value means using all loops as register
6241 allocation regions, the second value which is the default means using
6242 all loops except for loops with small register pressure as the
6243 regions, and third one means using all function as a single region.
6244 The first value can give best result for machines with small size and
6245 irregular register set, the third one results in faster and generates
6246 decent code and the smallest size code, and the default value usually
6247 give the best results in most cases and for most architectures.
6249 @item -fira-coalesce
6250 @opindex fira-coalesce
6251 Do optimistic register coalescing. This option might be profitable for
6252 architectures with big regular register files.
6254 @item -fira-loop-pressure
6255 @opindex fira-loop-pressure
6256 Use IRA to evaluate register pressure in loops for decision to move
6257 loop invariants. Usage of this option usually results in generation
6258 of faster and smaller code on machines with big register files (>= 32
6259 registers) but it can slow compiler down.
6261 This option is enabled at level @option{-O3} for some targets.
6263 @item -fno-ira-share-save-slots
6264 @opindex fno-ira-share-save-slots
6265 Switch off sharing stack slots used for saving call used hard
6266 registers living through a call. Each hard register will get a
6267 separate stack slot and as a result function stack frame will be
6270 @item -fno-ira-share-spill-slots
6271 @opindex fno-ira-share-spill-slots
6272 Switch off sharing stack slots allocated for pseudo-registers. Each
6273 pseudo-register which did not get a hard register will get a separate
6274 stack slot and as a result function stack frame will be bigger.
6276 @item -fira-verbose=@var{n}
6277 @opindex fira-verbose
6278 Set up how verbose dump file for the integrated register allocator
6279 will be. Default value is 5. If the value is greater or equal to 10,
6280 the dump file will be stderr as if the value were @var{n} minus 10.
6282 @item -fdelayed-branch
6283 @opindex fdelayed-branch
6284 If supported for the target machine, attempt to reorder instructions
6285 to exploit instruction slots available after delayed branch
6288 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6290 @item -fschedule-insns
6291 @opindex fschedule-insns
6292 If supported for the target machine, attempt to reorder instructions to
6293 eliminate execution stalls due to required data being unavailable. This
6294 helps machines that have slow floating point or memory load instructions
6295 by allowing other instructions to be issued until the result of the load
6296 or floating point instruction is required.
6298 Enabled at levels @option{-O2}, @option{-O3}.
6300 @item -fschedule-insns2
6301 @opindex fschedule-insns2
6302 Similar to @option{-fschedule-insns}, but requests an additional pass of
6303 instruction scheduling after register allocation has been done. This is
6304 especially useful on machines with a relatively small number of
6305 registers and where memory load instructions take more than one cycle.
6307 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6309 @item -fno-sched-interblock
6310 @opindex fno-sched-interblock
6311 Don't schedule instructions across basic blocks. This is normally
6312 enabled by default when scheduling before register allocation, i.e.@:
6313 with @option{-fschedule-insns} or at @option{-O2} or higher.
6315 @item -fno-sched-spec
6316 @opindex fno-sched-spec
6317 Don't allow speculative motion of non-load instructions. This is normally
6318 enabled by default when scheduling before register allocation, i.e.@:
6319 with @option{-fschedule-insns} or at @option{-O2} or higher.
6321 @item -fsched-pressure
6322 @opindex fsched-pressure
6323 Enable register pressure sensitive insn scheduling before the register
6324 allocation. This only makes sense when scheduling before register
6325 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
6326 @option{-O2} or higher. Usage of this option can improve the
6327 generated code and decrease its size by preventing register pressure
6328 increase above the number of available hard registers and as a
6329 consequence register spills in the register allocation.
6331 @item -fsched-spec-load
6332 @opindex fsched-spec-load
6333 Allow speculative motion of some load instructions. This only makes
6334 sense when scheduling before register allocation, i.e.@: with
6335 @option{-fschedule-insns} or at @option{-O2} or higher.
6337 @item -fsched-spec-load-dangerous
6338 @opindex fsched-spec-load-dangerous
6339 Allow speculative motion of more load instructions. This only makes
6340 sense when scheduling before register allocation, i.e.@: with
6341 @option{-fschedule-insns} or at @option{-O2} or higher.
6343 @item -fsched-stalled-insns
6344 @itemx -fsched-stalled-insns=@var{n}
6345 @opindex fsched-stalled-insns
6346 Define how many insns (if any) can be moved prematurely from the queue
6347 of stalled insns into the ready list, during the second scheduling pass.
6348 @option{-fno-sched-stalled-insns} means that no insns will be moved
6349 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6350 on how many queued insns can be moved prematurely.
6351 @option{-fsched-stalled-insns} without a value is equivalent to
6352 @option{-fsched-stalled-insns=1}.
6354 @item -fsched-stalled-insns-dep
6355 @itemx -fsched-stalled-insns-dep=@var{n}
6356 @opindex fsched-stalled-insns-dep
6357 Define how many insn groups (cycles) will be examined for a dependency
6358 on a stalled insn that is candidate for premature removal from the queue
6359 of stalled insns. This has an effect only during the second scheduling pass,
6360 and only if @option{-fsched-stalled-insns} is used.
6361 @option{-fno-sched-stalled-insns-dep} is equivalent to
6362 @option{-fsched-stalled-insns-dep=0}.
6363 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6364 @option{-fsched-stalled-insns-dep=1}.
6366 @item -fsched2-use-superblocks
6367 @opindex fsched2-use-superblocks
6368 When scheduling after register allocation, do use superblock scheduling
6369 algorithm. Superblock scheduling allows motion across basic block boundaries
6370 resulting on faster schedules. This option is experimental, as not all machine
6371 descriptions used by GCC model the CPU closely enough to avoid unreliable
6372 results from the algorithm.
6374 This only makes sense when scheduling after register allocation, i.e.@: with
6375 @option{-fschedule-insns2} or at @option{-O2} or higher.
6377 @item -fsched-group-heuristic
6378 @opindex fsched-group-heuristic
6379 Enable the group heuristic in the scheduler. This heuristic favors
6380 the instruction that belongs to a schedule group. This is enabled
6381 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6382 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6384 @item -fsched-critical-path-heuristic
6385 @opindex fsched-critical-path-heuristic
6386 Enable the critical-path heuristic in the scheduler. This heuristic favors
6387 instructions on the critical path. This is enabled by default when
6388 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6389 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6391 @item -fsched-spec-insn-heuristic
6392 @opindex fsched-spec-insn-heuristic
6393 Enable the speculative instruction heuristic in the scheduler. This
6394 heuristic favors speculative instructions with greater dependency weakness.
6395 This is enabled by default when scheduling is enabled, i.e.@:
6396 with @option{-fschedule-insns} or @option{-fschedule-insns2}
6397 or at @option{-O2} or higher.
6399 @item -fsched-rank-heuristic
6400 @opindex fsched-rank-heuristic
6401 Enable the rank heuristic in the scheduler. This heuristic favors
6402 the instruction belonging to a basic block with greater size or frequency.
6403 This is enabled by default when scheduling is enabled, i.e.@:
6404 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6405 at @option{-O2} or higher.
6407 @item -fsched-last-insn-heuristic
6408 @opindex fsched-last-insn-heuristic
6409 Enable the last-instruction heuristic in the scheduler. This heuristic
6410 favors the instruction that is less dependent on the last instruction
6411 scheduled. This is enabled by default when scheduling is enabled,
6412 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6413 at @option{-O2} or higher.
6415 @item -fsched-dep-count-heuristic
6416 @opindex fsched-dep-count-heuristic
6417 Enable the dependent-count heuristic in the scheduler. This heuristic
6418 favors the instruction that has more instructions depending on it.
6419 This is enabled by default when scheduling is enabled, i.e.@:
6420 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6421 at @option{-O2} or higher.
6423 @item -fsched2-use-traces
6424 @opindex fsched2-use-traces
6425 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
6426 allocation and additionally perform code duplication in order to increase the
6427 size of superblocks using tracer pass. See @option{-ftracer} for details on
6430 This mode should produce faster but significantly longer programs. Also
6431 without @option{-fbranch-probabilities} the traces constructed may not
6432 match the reality and hurt the performance. This only makes
6433 sense when scheduling after register allocation, i.e.@: with
6434 @option{-fschedule-insns2} or at @option{-O2} or higher.
6436 @item -freschedule-modulo-scheduled-loops
6437 @opindex freschedule-modulo-scheduled-loops
6438 The modulo scheduling comes before the traditional scheduling, if a loop
6439 was modulo scheduled we may want to prevent the later scheduling passes
6440 from changing its schedule, we use this option to control that.
6442 @item -fselective-scheduling
6443 @opindex fselective-scheduling
6444 Schedule instructions using selective scheduling algorithm. Selective
6445 scheduling runs instead of the first scheduler pass.
6447 @item -fselective-scheduling2
6448 @opindex fselective-scheduling2
6449 Schedule instructions using selective scheduling algorithm. Selective
6450 scheduling runs instead of the second scheduler pass.
6452 @item -fsel-sched-pipelining
6453 @opindex fsel-sched-pipelining
6454 Enable software pipelining of innermost loops during selective scheduling.
6455 This option has no effect until one of @option{-fselective-scheduling} or
6456 @option{-fselective-scheduling2} is turned on.
6458 @item -fsel-sched-pipelining-outer-loops
6459 @opindex fsel-sched-pipelining-outer-loops
6460 When pipelining loops during selective scheduling, also pipeline outer loops.
6461 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6463 @item -fcaller-saves
6464 @opindex fcaller-saves
6465 Enable values to be allocated in registers that will be clobbered by
6466 function calls, by emitting extra instructions to save and restore the
6467 registers around such calls. Such allocation is done only when it
6468 seems to result in better code than would otherwise be produced.
6470 This option is always enabled by default on certain machines, usually
6471 those which have no call-preserved registers to use instead.
6473 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6475 @item -fconserve-stack
6476 @opindex fconserve-stack
6477 Attempt to minimize stack usage. The compiler will attempt to use less
6478 stack space, even if that makes the program slower. This option
6479 implies setting the @option{large-stack-frame} parameter to 100
6480 and the @option{large-stack-frame-growth} parameter to 400.
6482 @item -ftree-reassoc
6483 @opindex ftree-reassoc
6484 Perform reassociation on trees. This flag is enabled by default
6485 at @option{-O} and higher.
6489 Perform partial redundancy elimination (PRE) on trees. This flag is
6490 enabled by default at @option{-O2} and @option{-O3}.
6492 @item -ftree-forwprop
6493 @opindex ftree-forwprop
6494 Perform forward propagation on trees. This flag is enabled by default
6495 at @option{-O} and higher.
6499 Perform full redundancy elimination (FRE) on trees. The difference
6500 between FRE and PRE is that FRE only considers expressions
6501 that are computed on all paths leading to the redundant computation.
6502 This analysis is faster than PRE, though it exposes fewer redundancies.
6503 This flag is enabled by default at @option{-O} and higher.
6505 @item -ftree-phiprop
6506 @opindex ftree-phiprop
6507 Perform hoisting of loads from conditional pointers on trees. This
6508 pass is enabled by default at @option{-O} and higher.
6510 @item -ftree-copy-prop
6511 @opindex ftree-copy-prop
6512 Perform copy propagation on trees. This pass eliminates unnecessary
6513 copy operations. This flag is enabled by default at @option{-O} and
6516 @item -fipa-pure-const
6517 @opindex fipa-pure-const
6518 Discover which functions are pure or constant.
6519 Enabled by default at @option{-O} and higher.
6521 @item -fipa-reference
6522 @opindex fipa-reference
6523 Discover which static variables do not escape cannot escape the
6525 Enabled by default at @option{-O} and higher.
6527 @item -fipa-struct-reorg
6528 @opindex fipa-struct-reorg
6529 Perform structure reorganization optimization, that change C-like structures
6530 layout in order to better utilize spatial locality. This transformation is
6531 affective for programs containing arrays of structures. Available in two
6532 compilation modes: profile-based (enabled with @option{-fprofile-generate})
6533 or static (which uses built-in heuristics). Require @option{-fipa-type-escape}
6534 to provide the safety of this transformation. It works only in whole program
6535 mode, so it requires @option{-fwhole-program} and @option{-combine} to be
6536 enabled. Structures considered @samp{cold} by this transformation are not
6537 affected (see @option{--param struct-reorg-cold-struct-ratio=@var{value}}).
6539 With this flag, the program debug info reflects a new structure layout.
6543 Perform interprocedural pointer analysis. This option is experimental
6544 and does not affect generated code.
6548 Perform interprocedural constant propagation.
6549 This optimization analyzes the program to determine when values passed
6550 to functions are constants and then optimizes accordingly.
6551 This optimization can substantially increase performance
6552 if the application has constants passed to functions.
6553 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
6555 @item -fipa-cp-clone
6556 @opindex fipa-cp-clone
6557 Perform function cloning to make interprocedural constant propagation stronger.
6558 When enabled, interprocedural constant propagation will perform function cloning
6559 when externally visible function can be called with constant arguments.
6560 Because this optimization can create multiple copies of functions,
6561 it may significantly increase code size
6562 (see @option{--param ipcp-unit-growth=@var{value}}).
6563 This flag is enabled by default at @option{-O3}.
6565 @item -fipa-matrix-reorg
6566 @opindex fipa-matrix-reorg
6567 Perform matrix flattening and transposing.
6568 Matrix flattening tries to replace an @math{m}-dimensional matrix
6569 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
6570 This reduces the level of indirection needed for accessing the elements
6571 of the matrix. The second optimization is matrix transposing that
6572 attempts to change the order of the matrix's dimensions in order to
6573 improve cache locality.
6574 Both optimizations need the @option{-fwhole-program} flag.
6575 Transposing is enabled only if profiling information is available.
6579 Perform forward store motion on trees. This flag is
6580 enabled by default at @option{-O} and higher.
6584 Perform sparse conditional constant propagation (CCP) on trees. This
6585 pass only operates on local scalar variables and is enabled by default
6586 at @option{-O} and higher.
6588 @item -ftree-switch-conversion
6589 Perform conversion of simple initializations in a switch to
6590 initializations from a scalar array. This flag is enabled by default
6591 at @option{-O2} and higher.
6595 Perform dead code elimination (DCE) on trees. This flag is enabled by
6596 default at @option{-O} and higher.
6598 @item -ftree-builtin-call-dce
6599 @opindex ftree-builtin-call-dce
6600 Perform conditional dead code elimination (DCE) for calls to builtin functions
6601 that may set @code{errno} but are otherwise side-effect free. This flag is
6602 enabled by default at @option{-O2} and higher if @option{-Os} is not also
6605 @item -ftree-dominator-opts
6606 @opindex ftree-dominator-opts
6607 Perform a variety of simple scalar cleanups (constant/copy
6608 propagation, redundancy elimination, range propagation and expression
6609 simplification) based on a dominator tree traversal. This also
6610 performs jump threading (to reduce jumps to jumps). This flag is
6611 enabled by default at @option{-O} and higher.
6615 Perform dead store elimination (DSE) on trees. A dead store is a store into
6616 a memory location which will later be overwritten by another store without
6617 any intervening loads. In this case the earlier store can be deleted. This
6618 flag is enabled by default at @option{-O} and higher.
6622 Perform loop header copying on trees. This is beneficial since it increases
6623 effectiveness of code motion optimizations. It also saves one jump. This flag
6624 is enabled by default at @option{-O} and higher. It is not enabled
6625 for @option{-Os}, since it usually increases code size.
6627 @item -ftree-loop-optimize
6628 @opindex ftree-loop-optimize
6629 Perform loop optimizations on trees. This flag is enabled by default
6630 at @option{-O} and higher.
6632 @item -ftree-loop-linear
6633 @opindex ftree-loop-linear
6634 Perform linear loop transformations on tree. This flag can improve cache
6635 performance and allow further loop optimizations to take place.
6637 @item -floop-interchange
6638 Perform loop interchange transformations on loops. Interchanging two
6639 nested loops switches the inner and outer loops. For example, given a
6644 A(J, I) = A(J, I) * C
6648 loop interchange will transform the loop as if the user had written:
6652 A(J, I) = A(J, I) * C
6656 which can be beneficial when @code{N} is larger than the caches,
6657 because in Fortran, the elements of an array are stored in memory
6658 contiguously by column, and the original loop iterates over rows,
6659 potentially creating at each access a cache miss. This optimization
6660 applies to all the languages supported by GCC and is not limited to
6661 Fortran. To use this code transformation, GCC has to be configured
6662 with @option{--with-ppl} and @option{--with-cloog} to enable the
6663 Graphite loop transformation infrastructure.
6665 @item -floop-strip-mine
6666 Perform loop strip mining transformations on loops. Strip mining
6667 splits a loop into two nested loops. The outer loop has strides
6668 equal to the strip size and the inner loop has strides of the
6669 original loop within a strip. For example, given a loop like:
6675 loop strip mining will transform the loop as if the user had written:
6678 DO I = II, min (II + 3, N)
6683 This optimization applies to all the languages supported by GCC and is
6684 not limited to Fortran. To use this code transformation, GCC has to
6685 be configured with @option{--with-ppl} and @option{--with-cloog} to
6686 enable the Graphite loop transformation infrastructure.
6689 Perform loop blocking transformations on loops. Blocking strip mines
6690 each loop in the loop nest such that the memory accesses of the
6691 element loops fit inside caches. For example, given a loop like:
6695 A(J, I) = B(I) + C(J)
6699 loop blocking will transform the loop as if the user had written:
6703 DO I = II, min (II + 63, N)
6704 DO J = JJ, min (JJ + 63, M)
6705 A(J, I) = B(I) + C(J)
6711 which can be beneficial when @code{M} is larger than the caches,
6712 because the innermost loop will iterate over a smaller amount of data
6713 that can be kept in the caches. This optimization applies to all the
6714 languages supported by GCC and is not limited to Fortran. To use this
6715 code transformation, GCC has to be configured with @option{--with-ppl}
6716 and @option{--with-cloog} to enable the Graphite loop transformation
6719 @item -fgraphite-identity
6720 @opindex fgraphite-identity
6721 Enable the identity transformation for graphite. For every SCoP we generate
6722 the polyhedral representation and transform it back to gimple. Using
6723 @option{-fgraphite-identity} we can check the costs or benefits of the
6724 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
6725 are also performed by the code generator CLooG, like index splitting and
6726 dead code elimination in loops.
6728 @item -floop-parallelize-all
6729 Use the Graphite data dependence analysis to identify loops that can
6730 be parallelized. Parallelize all the loops that can be analyzed to
6731 not contain loop carried dependences without checking that it is
6732 profitable to parallelize the loops.
6734 @item -fcheck-data-deps
6735 @opindex fcheck-data-deps
6736 Compare the results of several data dependence analyzers. This option
6737 is used for debugging the data dependence analyzers.
6739 @item -ftree-loop-distribution
6740 Perform loop distribution. This flag can improve cache performance on
6741 big loop bodies and allow further loop optimizations, like
6742 parallelization or vectorization, to take place. For example, the loop
6759 @item -ftree-loop-im
6760 @opindex ftree-loop-im
6761 Perform loop invariant motion on trees. This pass moves only invariants that
6762 would be hard to handle at RTL level (function calls, operations that expand to
6763 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
6764 operands of conditions that are invariant out of the loop, so that we can use
6765 just trivial invariantness analysis in loop unswitching. The pass also includes
6768 @item -ftree-loop-ivcanon
6769 @opindex ftree-loop-ivcanon
6770 Create a canonical counter for number of iterations in the loop for that
6771 determining number of iterations requires complicated analysis. Later
6772 optimizations then may determine the number easily. Useful especially
6773 in connection with unrolling.
6777 Perform induction variable optimizations (strength reduction, induction
6778 variable merging and induction variable elimination) on trees.
6780 @item -ftree-parallelize-loops=n
6781 @opindex ftree-parallelize-loops
6782 Parallelize loops, i.e., split their iteration space to run in n threads.
6783 This is only possible for loops whose iterations are independent
6784 and can be arbitrarily reordered. The optimization is only
6785 profitable on multiprocessor machines, for loops that are CPU-intensive,
6786 rather than constrained e.g.@: by memory bandwidth. This option
6787 implies @option{-pthread}, and thus is only supported on targets
6788 that have support for @option{-pthread}.
6792 Perform function-local points-to analysis on trees. This flag is
6793 enabled by default at @option{-O} and higher.
6797 Perform scalar replacement of aggregates. This pass replaces structure
6798 references with scalars to prevent committing structures to memory too
6799 early. This flag is enabled by default at @option{-O} and higher.
6801 @item -ftree-copyrename
6802 @opindex ftree-copyrename
6803 Perform copy renaming on trees. This pass attempts to rename compiler
6804 temporaries to other variables at copy locations, usually resulting in
6805 variable names which more closely resemble the original variables. This flag
6806 is enabled by default at @option{-O} and higher.
6810 Perform temporary expression replacement during the SSA->normal phase. Single
6811 use/single def temporaries are replaced at their use location with their
6812 defining expression. This results in non-GIMPLE code, but gives the expanders
6813 much more complex trees to work on resulting in better RTL generation. This is
6814 enabled by default at @option{-O} and higher.
6816 @item -ftree-vectorize
6817 @opindex ftree-vectorize
6818 Perform loop vectorization on trees. This flag is enabled by default at
6821 @item -ftree-vect-loop-version
6822 @opindex ftree-vect-loop-version
6823 Perform loop versioning when doing loop vectorization on trees. When a loop
6824 appears to be vectorizable except that data alignment or data dependence cannot
6825 be determined at compile time then vectorized and non-vectorized versions of
6826 the loop are generated along with runtime checks for alignment or dependence
6827 to control which version is executed. This option is enabled by default
6828 except at level @option{-Os} where it is disabled.
6830 @item -fvect-cost-model
6831 @opindex fvect-cost-model
6832 Enable cost model for vectorization.
6836 Perform Value Range Propagation on trees. This is similar to the
6837 constant propagation pass, but instead of values, ranges of values are
6838 propagated. This allows the optimizers to remove unnecessary range
6839 checks like array bound checks and null pointer checks. This is
6840 enabled by default at @option{-O2} and higher. Null pointer check
6841 elimination is only done if @option{-fdelete-null-pointer-checks} is
6846 Perform tail duplication to enlarge superblock size. This transformation
6847 simplifies the control flow of the function allowing other optimizations to do
6850 @item -funroll-loops
6851 @opindex funroll-loops
6852 Unroll loops whose number of iterations can be determined at compile
6853 time or upon entry to the loop. @option{-funroll-loops} implies
6854 @option{-frerun-cse-after-loop}. This option makes code larger,
6855 and may or may not make it run faster.
6857 @item -funroll-all-loops
6858 @opindex funroll-all-loops
6859 Unroll all loops, even if their number of iterations is uncertain when
6860 the loop is entered. This usually makes programs run more slowly.
6861 @option{-funroll-all-loops} implies the same options as
6862 @option{-funroll-loops},
6864 @item -fsplit-ivs-in-unroller
6865 @opindex fsplit-ivs-in-unroller
6866 Enables expressing of values of induction variables in later iterations
6867 of the unrolled loop using the value in the first iteration. This breaks
6868 long dependency chains, thus improving efficiency of the scheduling passes.
6870 Combination of @option{-fweb} and CSE is often sufficient to obtain the
6871 same effect. However in cases the loop body is more complicated than
6872 a single basic block, this is not reliable. It also does not work at all
6873 on some of the architectures due to restrictions in the CSE pass.
6875 This optimization is enabled by default.
6877 @item -fvariable-expansion-in-unroller
6878 @opindex fvariable-expansion-in-unroller
6879 With this option, the compiler will create multiple copies of some
6880 local variables when unrolling a loop which can result in superior code.
6882 @item -fpredictive-commoning
6883 @opindex fpredictive-commoning
6884 Perform predictive commoning optimization, i.e., reusing computations
6885 (especially memory loads and stores) performed in previous
6886 iterations of loops.
6888 This option is enabled at level @option{-O3}.
6890 @item -fprefetch-loop-arrays
6891 @opindex fprefetch-loop-arrays
6892 If supported by the target machine, generate instructions to prefetch
6893 memory to improve the performance of loops that access large arrays.
6895 This option may generate better or worse code; results are highly
6896 dependent on the structure of loops within the source code.
6898 Disabled at level @option{-Os}.
6901 @itemx -fno-peephole2
6902 @opindex fno-peephole
6903 @opindex fno-peephole2
6904 Disable any machine-specific peephole optimizations. The difference
6905 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
6906 are implemented in the compiler; some targets use one, some use the
6907 other, a few use both.
6909 @option{-fpeephole} is enabled by default.
6910 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6912 @item -fno-guess-branch-probability
6913 @opindex fno-guess-branch-probability
6914 Do not guess branch probabilities using heuristics.
6916 GCC will use heuristics to guess branch probabilities if they are
6917 not provided by profiling feedback (@option{-fprofile-arcs}). These
6918 heuristics are based on the control flow graph. If some branch probabilities
6919 are specified by @samp{__builtin_expect}, then the heuristics will be
6920 used to guess branch probabilities for the rest of the control flow graph,
6921 taking the @samp{__builtin_expect} info into account. The interactions
6922 between the heuristics and @samp{__builtin_expect} can be complex, and in
6923 some cases, it may be useful to disable the heuristics so that the effects
6924 of @samp{__builtin_expect} are easier to understand.
6926 The default is @option{-fguess-branch-probability} at levels
6927 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6929 @item -freorder-blocks
6930 @opindex freorder-blocks
6931 Reorder basic blocks in the compiled function in order to reduce number of
6932 taken branches and improve code locality.
6934 Enabled at levels @option{-O2}, @option{-O3}.
6936 @item -freorder-blocks-and-partition
6937 @opindex freorder-blocks-and-partition
6938 In addition to reordering basic blocks in the compiled function, in order
6939 to reduce number of taken branches, partitions hot and cold basic blocks
6940 into separate sections of the assembly and .o files, to improve
6941 paging and cache locality performance.
6943 This optimization is automatically turned off in the presence of
6944 exception handling, for linkonce sections, for functions with a user-defined
6945 section attribute and on any architecture that does not support named
6948 @item -freorder-functions
6949 @opindex freorder-functions
6950 Reorder functions in the object file in order to
6951 improve code locality. This is implemented by using special
6952 subsections @code{.text.hot} for most frequently executed functions and
6953 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
6954 the linker so object file format must support named sections and linker must
6955 place them in a reasonable way.
6957 Also profile feedback must be available in to make this option effective. See
6958 @option{-fprofile-arcs} for details.
6960 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6962 @item -fstrict-aliasing
6963 @opindex fstrict-aliasing
6964 Allow the compiler to assume the strictest aliasing rules applicable to
6965 the language being compiled. For C (and C++), this activates
6966 optimizations based on the type of expressions. In particular, an
6967 object of one type is assumed never to reside at the same address as an
6968 object of a different type, unless the types are almost the same. For
6969 example, an @code{unsigned int} can alias an @code{int}, but not a
6970 @code{void*} or a @code{double}. A character type may alias any other
6973 @anchor{Type-punning}Pay special attention to code like this:
6986 The practice of reading from a different union member than the one most
6987 recently written to (called ``type-punning'') is common. Even with
6988 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
6989 is accessed through the union type. So, the code above will work as
6990 expected. @xref{Structures unions enumerations and bit-fields
6991 implementation}. However, this code might not:
7002 Similarly, access by taking the address, casting the resulting pointer
7003 and dereferencing the result has undefined behavior, even if the cast
7004 uses a union type, e.g.:
7008 return ((union a_union *) &d)->i;
7012 The @option{-fstrict-aliasing} option is enabled at levels
7013 @option{-O2}, @option{-O3}, @option{-Os}.
7015 @item -fstrict-overflow
7016 @opindex fstrict-overflow
7017 Allow the compiler to assume strict signed overflow rules, depending
7018 on the language being compiled. For C (and C++) this means that
7019 overflow when doing arithmetic with signed numbers is undefined, which
7020 means that the compiler may assume that it will not happen. This
7021 permits various optimizations. For example, the compiler will assume
7022 that an expression like @code{i + 10 > i} will always be true for
7023 signed @code{i}. This assumption is only valid if signed overflow is
7024 undefined, as the expression is false if @code{i + 10} overflows when
7025 using twos complement arithmetic. When this option is in effect any
7026 attempt to determine whether an operation on signed numbers will
7027 overflow must be written carefully to not actually involve overflow.
7029 This option also allows the compiler to assume strict pointer
7030 semantics: given a pointer to an object, if adding an offset to that
7031 pointer does not produce a pointer to the same object, the addition is
7032 undefined. This permits the compiler to conclude that @code{p + u >
7033 p} is always true for a pointer @code{p} and unsigned integer
7034 @code{u}. This assumption is only valid because pointer wraparound is
7035 undefined, as the expression is false if @code{p + u} overflows using
7036 twos complement arithmetic.
7038 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
7039 that integer signed overflow is fully defined: it wraps. When
7040 @option{-fwrapv} is used, there is no difference between
7041 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
7042 integers. With @option{-fwrapv} certain types of overflow are
7043 permitted. For example, if the compiler gets an overflow when doing
7044 arithmetic on constants, the overflowed value can still be used with
7045 @option{-fwrapv}, but not otherwise.
7047 The @option{-fstrict-overflow} option is enabled at levels
7048 @option{-O2}, @option{-O3}, @option{-Os}.
7050 @item -falign-functions
7051 @itemx -falign-functions=@var{n}
7052 @opindex falign-functions
7053 Align the start of functions to the next power-of-two greater than
7054 @var{n}, skipping up to @var{n} bytes. For instance,
7055 @option{-falign-functions=32} aligns functions to the next 32-byte
7056 boundary, but @option{-falign-functions=24} would align to the next
7057 32-byte boundary only if this can be done by skipping 23 bytes or less.
7059 @option{-fno-align-functions} and @option{-falign-functions=1} are
7060 equivalent and mean that functions will not be aligned.
7062 Some assemblers only support this flag when @var{n} is a power of two;
7063 in that case, it is rounded up.
7065 If @var{n} is not specified or is zero, use a machine-dependent default.
7067 Enabled at levels @option{-O2}, @option{-O3}.
7069 @item -falign-labels
7070 @itemx -falign-labels=@var{n}
7071 @opindex falign-labels
7072 Align all branch targets to a power-of-two boundary, skipping up to
7073 @var{n} bytes like @option{-falign-functions}. This option can easily
7074 make code slower, because it must insert dummy operations for when the
7075 branch target is reached in the usual flow of the code.
7077 @option{-fno-align-labels} and @option{-falign-labels=1} are
7078 equivalent and mean that labels will not be aligned.
7080 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
7081 are greater than this value, then their values are used instead.
7083 If @var{n} is not specified or is zero, use a machine-dependent default
7084 which is very likely to be @samp{1}, meaning no alignment.
7086 Enabled at levels @option{-O2}, @option{-O3}.
7089 @itemx -falign-loops=@var{n}
7090 @opindex falign-loops
7091 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7092 like @option{-falign-functions}. The hope is that the loop will be
7093 executed many times, which will make up for any execution of the dummy
7096 @option{-fno-align-loops} and @option{-falign-loops=1} are
7097 equivalent and mean that loops will not be aligned.
7099 If @var{n} is not specified or is zero, use a machine-dependent default.
7101 Enabled at levels @option{-O2}, @option{-O3}.
7104 @itemx -falign-jumps=@var{n}
7105 @opindex falign-jumps
7106 Align branch targets to a power-of-two boundary, for branch targets
7107 where the targets can only be reached by jumping, skipping up to @var{n}
7108 bytes like @option{-falign-functions}. In this case, no dummy operations
7111 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7112 equivalent and mean that loops will not be aligned.
7114 If @var{n} is not specified or is zero, use a machine-dependent default.
7116 Enabled at levels @option{-O2}, @option{-O3}.
7118 @item -funit-at-a-time
7119 @opindex funit-at-a-time
7120 This option is left for compatibility reasons. @option{-funit-at-a-time}
7121 has no effect, while @option{-fno-unit-at-a-time} implies
7122 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7126 @item -fno-toplevel-reorder
7127 @opindex fno-toplevel-reorder
7128 Do not reorder top-level functions, variables, and @code{asm}
7129 statements. Output them in the same order that they appear in the
7130 input file. When this option is used, unreferenced static variables
7131 will not be removed. This option is intended to support existing code
7132 which relies on a particular ordering. For new code, it is better to
7135 Enabled at level @option{-O0}. When disabled explicitly, it also imply
7136 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
7141 Constructs webs as commonly used for register allocation purposes and assign
7142 each web individual pseudo register. This allows the register allocation pass
7143 to operate on pseudos directly, but also strengthens several other optimization
7144 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7145 however, make debugging impossible, since variables will no longer stay in a
7148 Enabled by default with @option{-funroll-loops}.
7150 @item -fwhole-program
7151 @opindex fwhole-program
7152 Assume that the current compilation unit represents the whole program being
7153 compiled. All public functions and variables with the exception of @code{main}
7154 and those merged by attribute @code{externally_visible} become static functions
7155 and in effect are optimized more aggressively by interprocedural optimizers.
7156 While this option is equivalent to proper use of the @code{static} keyword for
7157 programs consisting of a single file, in combination with option
7158 @option{-combine}, @option{-flto} or @option{-fwhopr} this flag can be used to
7159 compile many smaller scale programs since the functions and variables become
7160 local for the whole combined compilation unit, not for the single source file
7163 This option implies @option{-fwhole-file} for Fortran programs.
7167 This option runs the standard link-time optimizer. When invoked
7168 with source code, it generates GIMPLE (one of GCC's internal
7169 representations) and writes it to special ELF sections in the object
7170 file. When the object files are linked together, all the function
7171 bodies are read from these ELF sections and instantiated as if they
7172 had been part of the same translation unit.
7174 To use the link-timer optimizer, @option{-flto} needs to be specified at
7175 compile time and during the final link. For example,
7178 gcc -c -O2 -flto foo.c
7179 gcc -c -O2 -flto bar.c
7180 gcc -o myprog -flto -O2 foo.o bar.o
7183 The first two invocations to GCC will save a bytecode representation
7184 of GIMPLE into special ELF sections inside @file{foo.o} and
7185 @file{bar.o}. The final invocation will read the GIMPLE bytecode from
7186 @file{foo.o} and @file{bar.o}, merge the two files into a single
7187 internal image, and compile the result as usual. Since both
7188 @file{foo.o} and @file{bar.o} are merged into a single image, this
7189 causes all the inter-procedural analyses and optimizations in GCC to
7190 work across the two files as if they were a single one. This means,
7191 for example, that the inliner will be able to inline functions in
7192 @file{bar.o} into functions in @file{foo.o} and vice-versa.
7194 Another (simpler) way to enable link-time optimization is,
7197 gcc -o myprog -flto -O2 foo.c bar.c
7200 The above will generate bytecode for @file{foo.c} and @file{bar.c},
7201 merge them together into a single GIMPLE representation and optimize
7202 them as usual to produce @file{myprog}.
7204 The only important thing to keep in mind is that to enable link-time
7205 optimizations the @option{-flto} flag needs to be passed to both the
7206 compile and the link commands.
7208 Note that when a file is compiled with @option{-flto}, the generated
7209 object file will be larger than a regular object file because it will
7210 contain GIMPLE bytecodes and the usual final code. This means that
7211 object files with LTO information can be linked as a normal object
7212 file. So, in the previous example, if the final link is done with
7215 gcc -o myprog foo.o bar.o
7218 The only difference will be that no inter-procedural optimizations
7219 will be applied to produce @file{myprog}. The two object files
7220 @file{foo.o} and @file{bar.o} will be simply sent to the regular
7223 Additionally, the optimization flags used to compile individual files
7224 are not necessarily related to those used at link-time. For instance,
7227 gcc -c -O0 -flto foo.c
7228 gcc -c -O0 -flto bar.c
7229 gcc -o myprog -flto -O3 foo.o bar.o
7232 This will produce individual object files with unoptimized assembler
7233 code, but the resulting binary @file{myprog} will be optimized at
7234 @option{-O3}. Now, if the final binary is generated without
7235 @option{-flto}, then @file{myprog} will not be optimized.
7237 When producing the final binary with @option{-flto}, GCC will only
7238 apply link-time optimizations to those files that contain bytecode.
7239 Therefore, you can mix and match object files and libraries with
7240 GIMPLE bytecodes and final object code. GCC will automatically select
7241 which files to optimize in LTO mode and which files to link without
7244 There are some code generation flags that GCC will preserve when
7245 generating bytecodes, as they need to be used during the final link
7246 stage. Currently, the following options are saved into the GIMPLE
7247 bytecode files: @option{-fPIC}, @option{-fcommon} and all the
7248 @option{-m} target flags.
7250 At link time, these options are read-in and reapplied. Note that the
7251 current implementation makes no attempt at recognizing conflicting
7252 values for these options. If two or more files have a conflicting
7253 value (e.g., one file is compiled with @option{-fPIC} and another
7254 isn't), the compiler will simply use the last value read from the
7255 bytecode files. It is recommended, then, that all the files
7256 participating in the same link be compiled with the same options.
7258 Another feature of LTO is that it is possible to apply interprocedural
7259 optimizations on files written in different languages. This requires
7260 some support in the language front end. Currently, the C, C++ and
7261 Fortran front ends are capable of emitting GIMPLE bytecodes, so
7262 something like this should work
7267 gfortran -c -flto baz.f90
7268 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
7271 Notice that the final link is done with @command{g++} to get the C++
7272 runtime libraries and @option{-lgfortran} is added to get the Fortran
7273 runtime libraries. In general, when mixing languages in LTO mode, you
7274 should use the same link command used when mixing languages in a
7275 regular (non-LTO) compilation. This means that if your build process
7276 was mixing languages before, all you need to add is @option{-flto} to
7277 all the compile and link commands.
7279 If object files containing GIMPLE bytecode are stored in a library
7280 archive, say @file{libfoo.a}, it is possible to extract and use them
7281 in an LTO link if you are using @command{gold} as the linker (which,
7282 in turn requires GCC to be configured with @option{--enable-gold}).
7283 To enable this feature, use the flag @option{-fuse-linker-plugin} at
7287 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
7290 With the linker plugin enabled, @command{gold} will extract the needed
7291 GIMPLE files from @file{libfoo.a} and pass them on to the running GCC
7292 to make them part of the aggregated GIMPLE image to be optimized.
7294 If you are not using @command{gold} and/or do not specify
7295 @option{-fuse-linker-plugin} then the objects inside @file{libfoo.a}
7296 will be extracted and linked as usual, but they will not participate
7297 in the LTO optimization process.
7299 Link time optimizations do not require the presence of the whole
7300 program to operate. If the program does not require any symbols to
7301 be exported, it is possible to combine @option{-flto} and
7302 @option{-fwhopr} with @option{-fwhole-program} to allow the
7303 interprocedural optimizers to use more aggressive assumptions which
7304 may lead to improved optimization opportunities.
7306 Regarding portability: the current implementation of LTO makes no
7307 attempt at generating bytecode that can be ported between different
7308 types of hosts. The bytecode files are versioned and there is a
7309 strict version check, so bytecode files generated in one version of
7310 GCC will not work with an older/newer version of GCC.
7312 This option is disabled by default.
7316 This option is identical in functionality to @option{-flto} but it
7317 differs in how the final link stage is executed. Instead of loading
7318 all the function bodies in memory, the callgraph is analyzed and
7319 optimization decisions are made (whole program analysis or WPA). Once
7320 optimization decisions are made, the callgraph is partitioned and the
7321 different sections are compiled separately (local transformations or
7322 LTRANS)@. This process allows optimizations on very large programs
7323 that otherwise would not fit in memory. This option enables
7324 @option{-fwpa} and @option{-fltrans} automatically.
7326 Disabled by default.
7330 This is an internal option used by GCC when compiling with
7331 @option{-fwhopr}. You should never need to use it.
7333 This option runs the link-time optimizer in the whole-program-analysis
7334 (WPA) mode, which reads in summary information from all inputs and
7335 performs a whole-program analysis based on summary information only.
7336 It generates object files for subsequent runs of the link-time
7337 optimizer where individual object files are optimized using both
7338 summary information from the WPA mode and the actual function bodies.
7339 It then drives the LTRANS phase.
7341 Disabled by default.
7345 This is an internal option used by GCC when compiling with
7346 @option{-fwhopr}. You should never need to use it.
7348 This option runs the link-time optimizer in the local-transformation (LTRANS)
7349 mode, which reads in output from a previous run of the LTO in WPA mode.
7350 In the LTRANS mode, LTO optimizes an object and produces the final assembly.
7352 Disabled by default.
7354 @item -fltrans-output-list=@var{file}
7355 @opindex fltrans-output-list
7356 This is an internal option used by GCC when compiling with
7357 @option{-fwhopr}. You should never need to use it.
7359 This option specifies a file to which the names of LTRANS output files are
7360 written. This option is only meaningful in conjunction with @option{-fwpa}.
7362 Disabled by default.
7364 @item -flto-compression-level=@var{n}
7365 This option specifies the level of compression used for intermediate
7366 language written to LTO object files, and is only meaningful in
7367 conjunction with LTO mode (@option{-fwhopr}, @option{-flto}). Valid
7368 values are 0 (no compression) to 9 (maximum compression). Values
7369 outside this range are clamped to either 0 or 9. If the option is not
7370 given, a default balanced compression setting is used.
7373 Prints a report with internal details on the workings of the link-time
7374 optimizer. The contents of this report vary from version to version,
7375 it is meant to be useful to GCC developers when processing object
7376 files in LTO mode (via @option{-fwhopr} or @option{-flto}).
7378 Disabled by default.
7380 @item -fuse-linker-plugin
7381 Enables the extraction of objects with GIMPLE bytecode information
7382 from library archives. This option relies on features available only
7383 in @command{gold}, so to use this you must configure GCC with
7384 @option{--enable-gold}. See @option{-flto} for a description on the
7385 effect of this flag and how to use it.
7387 Disabled by default.
7389 @item -fcprop-registers
7390 @opindex fcprop-registers
7391 After register allocation and post-register allocation instruction splitting,
7392 we perform a copy-propagation pass to try to reduce scheduling dependencies
7393 and occasionally eliminate the copy.
7395 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7397 @item -fprofile-correction
7398 @opindex fprofile-correction
7399 Profiles collected using an instrumented binary for multi-threaded programs may
7400 be inconsistent due to missed counter updates. When this option is specified,
7401 GCC will use heuristics to correct or smooth out such inconsistencies. By
7402 default, GCC will emit an error message when an inconsistent profile is detected.
7404 @item -fprofile-dir=@var{path}
7405 @opindex fprofile-dir
7407 Set the directory to search the profile data files in to @var{path}.
7408 This option affects only the profile data generated by
7409 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
7410 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
7411 and its related options.
7412 By default, GCC will use the current directory as @var{path}
7413 thus the profile data file will appear in the same directory as the object file.
7415 @item -fprofile-generate
7416 @itemx -fprofile-generate=@var{path}
7417 @opindex fprofile-generate
7419 Enable options usually used for instrumenting application to produce
7420 profile useful for later recompilation with profile feedback based
7421 optimization. You must use @option{-fprofile-generate} both when
7422 compiling and when linking your program.
7424 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
7426 If @var{path} is specified, GCC will look at the @var{path} to find
7427 the profile feedback data files. See @option{-fprofile-dir}.
7430 @itemx -fprofile-use=@var{path}
7431 @opindex fprofile-use
7432 Enable profile feedback directed optimizations, and optimizations
7433 generally profitable only with profile feedback available.
7435 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
7436 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
7438 By default, GCC emits an error message if the feedback profiles do not
7439 match the source code. This error can be turned into a warning by using
7440 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
7443 If @var{path} is specified, GCC will look at the @var{path} to find
7444 the profile feedback data files. See @option{-fprofile-dir}.
7447 The following options control compiler behavior regarding floating
7448 point arithmetic. These options trade off between speed and
7449 correctness. All must be specifically enabled.
7453 @opindex ffloat-store
7454 Do not store floating point variables in registers, and inhibit other
7455 options that might change whether a floating point value is taken from a
7458 @cindex floating point precision
7459 This option prevents undesirable excess precision on machines such as
7460 the 68000 where the floating registers (of the 68881) keep more
7461 precision than a @code{double} is supposed to have. Similarly for the
7462 x86 architecture. For most programs, the excess precision does only
7463 good, but a few programs rely on the precise definition of IEEE floating
7464 point. Use @option{-ffloat-store} for such programs, after modifying
7465 them to store all pertinent intermediate computations into variables.
7467 @item -fexcess-precision=@var{style}
7468 @opindex fexcess-precision
7469 This option allows further control over excess precision on machines
7470 where floating-point registers have more precision than the IEEE
7471 @code{float} and @code{double} types and the processor does not
7472 support operations rounding to those types. By default,
7473 @option{-fexcess-precision=fast} is in effect; this means that
7474 operations are carried out in the precision of the registers and that
7475 it is unpredictable when rounding to the types specified in the source
7476 code takes place. When compiling C, if
7477 @option{-fexcess-precision=standard} is specified then excess
7478 precision will follow the rules specified in ISO C99; in particular,
7479 both casts and assignments cause values to be rounded to their
7480 semantic types (whereas @option{-ffloat-store} only affects
7481 assignments). This option is enabled by default for C if a strict
7482 conformance option such as @option{-std=c99} is used.
7485 @option{-fexcess-precision=standard} is not implemented for languages
7486 other than C, and has no effect if
7487 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
7488 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
7489 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
7490 semantics apply without excess precision, and in the latter, rounding
7495 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
7496 @option{-ffinite-math-only}, @option{-fno-rounding-math},
7497 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
7499 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
7501 This option is not turned on by any @option{-O} option since
7502 it can result in incorrect output for programs which depend on
7503 an exact implementation of IEEE or ISO rules/specifications for
7504 math functions. It may, however, yield faster code for programs
7505 that do not require the guarantees of these specifications.
7507 @item -fno-math-errno
7508 @opindex fno-math-errno
7509 Do not set ERRNO after calling math functions that are executed
7510 with a single instruction, e.g., sqrt. A program that relies on
7511 IEEE exceptions for math error handling may want to use this flag
7512 for speed while maintaining IEEE arithmetic compatibility.
7514 This option is not turned on by any @option{-O} option since
7515 it can result in incorrect output for programs which depend on
7516 an exact implementation of IEEE or ISO rules/specifications for
7517 math functions. It may, however, yield faster code for programs
7518 that do not require the guarantees of these specifications.
7520 The default is @option{-fmath-errno}.
7522 On Darwin systems, the math library never sets @code{errno}. There is
7523 therefore no reason for the compiler to consider the possibility that
7524 it might, and @option{-fno-math-errno} is the default.
7526 @item -funsafe-math-optimizations
7527 @opindex funsafe-math-optimizations
7529 Allow optimizations for floating-point arithmetic that (a) assume
7530 that arguments and results are valid and (b) may violate IEEE or
7531 ANSI standards. When used at link-time, it may include libraries
7532 or startup files that change the default FPU control word or other
7533 similar optimizations.
7535 This option is not turned on by any @option{-O} option since
7536 it can result in incorrect output for programs which depend on
7537 an exact implementation of IEEE or ISO rules/specifications for
7538 math functions. It may, however, yield faster code for programs
7539 that do not require the guarantees of these specifications.
7540 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
7541 @option{-fassociative-math} and @option{-freciprocal-math}.
7543 The default is @option{-fno-unsafe-math-optimizations}.
7545 @item -fassociative-math
7546 @opindex fassociative-math
7548 Allow re-association of operands in series of floating-point operations.
7549 This violates the ISO C and C++ language standard by possibly changing
7550 computation result. NOTE: re-ordering may change the sign of zero as
7551 well as ignore NaNs and inhibit or create underflow or overflow (and
7552 thus cannot be used on a code which relies on rounding behavior like
7553 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
7554 and thus may not be used when ordered comparisons are required.
7555 This option requires that both @option{-fno-signed-zeros} and
7556 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
7557 much sense with @option{-frounding-math}.
7559 The default is @option{-fno-associative-math}.
7561 @item -freciprocal-math
7562 @opindex freciprocal-math
7564 Allow the reciprocal of a value to be used instead of dividing by
7565 the value if this enables optimizations. For example @code{x / y}
7566 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
7567 is subject to common subexpression elimination. Note that this loses
7568 precision and increases the number of flops operating on the value.
7570 The default is @option{-fno-reciprocal-math}.
7572 @item -ffinite-math-only
7573 @opindex ffinite-math-only
7574 Allow optimizations for floating-point arithmetic that assume
7575 that arguments and results are not NaNs or +-Infs.
7577 This option is not turned on by any @option{-O} option since
7578 it can result in incorrect output for programs which depend on
7579 an exact implementation of IEEE or ISO rules/specifications for
7580 math functions. It may, however, yield faster code for programs
7581 that do not require the guarantees of these specifications.
7583 The default is @option{-fno-finite-math-only}.
7585 @item -fno-signed-zeros
7586 @opindex fno-signed-zeros
7587 Allow optimizations for floating point arithmetic that ignore the
7588 signedness of zero. IEEE arithmetic specifies the behavior of
7589 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
7590 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
7591 This option implies that the sign of a zero result isn't significant.
7593 The default is @option{-fsigned-zeros}.
7595 @item -fno-trapping-math
7596 @opindex fno-trapping-math
7597 Compile code assuming that floating-point operations cannot generate
7598 user-visible traps. These traps include division by zero, overflow,
7599 underflow, inexact result and invalid operation. This option requires
7600 that @option{-fno-signaling-nans} be in effect. Setting this option may
7601 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
7603 This option should never be turned on by any @option{-O} option since
7604 it can result in incorrect output for programs which depend on
7605 an exact implementation of IEEE or ISO rules/specifications for
7608 The default is @option{-ftrapping-math}.
7610 @item -frounding-math
7611 @opindex frounding-math
7612 Disable transformations and optimizations that assume default floating
7613 point rounding behavior. This is round-to-zero for all floating point
7614 to integer conversions, and round-to-nearest for all other arithmetic
7615 truncations. This option should be specified for programs that change
7616 the FP rounding mode dynamically, or that may be executed with a
7617 non-default rounding mode. This option disables constant folding of
7618 floating point expressions at compile-time (which may be affected by
7619 rounding mode) and arithmetic transformations that are unsafe in the
7620 presence of sign-dependent rounding modes.
7622 The default is @option{-fno-rounding-math}.
7624 This option is experimental and does not currently guarantee to
7625 disable all GCC optimizations that are affected by rounding mode.
7626 Future versions of GCC may provide finer control of this setting
7627 using C99's @code{FENV_ACCESS} pragma. This command line option
7628 will be used to specify the default state for @code{FENV_ACCESS}.
7630 @item -fsignaling-nans
7631 @opindex fsignaling-nans
7632 Compile code assuming that IEEE signaling NaNs may generate user-visible
7633 traps during floating-point operations. Setting this option disables
7634 optimizations that may change the number of exceptions visible with
7635 signaling NaNs. This option implies @option{-ftrapping-math}.
7637 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
7640 The default is @option{-fno-signaling-nans}.
7642 This option is experimental and does not currently guarantee to
7643 disable all GCC optimizations that affect signaling NaN behavior.
7645 @item -fsingle-precision-constant
7646 @opindex fsingle-precision-constant
7647 Treat floating point constant as single precision constant instead of
7648 implicitly converting it to double precision constant.
7650 @item -fcx-limited-range
7651 @opindex fcx-limited-range
7652 When enabled, this option states that a range reduction step is not
7653 needed when performing complex division. Also, there is no checking
7654 whether the result of a complex multiplication or division is @code{NaN
7655 + I*NaN}, with an attempt to rescue the situation in that case. The
7656 default is @option{-fno-cx-limited-range}, but is enabled by
7657 @option{-ffast-math}.
7659 This option controls the default setting of the ISO C99
7660 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
7663 @item -fcx-fortran-rules
7664 @opindex fcx-fortran-rules
7665 Complex multiplication and division follow Fortran rules. Range
7666 reduction is done as part of complex division, but there is no checking
7667 whether the result of a complex multiplication or division is @code{NaN
7668 + I*NaN}, with an attempt to rescue the situation in that case.
7670 The default is @option{-fno-cx-fortran-rules}.
7674 The following options control optimizations that may improve
7675 performance, but are not enabled by any @option{-O} options. This
7676 section includes experimental options that may produce broken code.
7679 @item -fbranch-probabilities
7680 @opindex fbranch-probabilities
7681 After running a program compiled with @option{-fprofile-arcs}
7682 (@pxref{Debugging Options,, Options for Debugging Your Program or
7683 @command{gcc}}), you can compile it a second time using
7684 @option{-fbranch-probabilities}, to improve optimizations based on
7685 the number of times each branch was taken. When the program
7686 compiled with @option{-fprofile-arcs} exits it saves arc execution
7687 counts to a file called @file{@var{sourcename}.gcda} for each source
7688 file. The information in this data file is very dependent on the
7689 structure of the generated code, so you must use the same source code
7690 and the same optimization options for both compilations.
7692 With @option{-fbranch-probabilities}, GCC puts a
7693 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
7694 These can be used to improve optimization. Currently, they are only
7695 used in one place: in @file{reorg.c}, instead of guessing which path a
7696 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
7697 exactly determine which path is taken more often.
7699 @item -fprofile-values
7700 @opindex fprofile-values
7701 If combined with @option{-fprofile-arcs}, it adds code so that some
7702 data about values of expressions in the program is gathered.
7704 With @option{-fbranch-probabilities}, it reads back the data gathered
7705 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
7706 notes to instructions for their later usage in optimizations.
7708 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
7712 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
7713 a code to gather information about values of expressions.
7715 With @option{-fbranch-probabilities}, it reads back the data gathered
7716 and actually performs the optimizations based on them.
7717 Currently the optimizations include specialization of division operation
7718 using the knowledge about the value of the denominator.
7720 @item -frename-registers
7721 @opindex frename-registers
7722 Attempt to avoid false dependencies in scheduled code by making use
7723 of registers left over after register allocation. This optimization
7724 will most benefit processors with lots of registers. Depending on the
7725 debug information format adopted by the target, however, it can
7726 make debugging impossible, since variables will no longer stay in
7727 a ``home register''.
7729 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
7733 Perform tail duplication to enlarge superblock size. This transformation
7734 simplifies the control flow of the function allowing other optimizations to do
7737 Enabled with @option{-fprofile-use}.
7739 @item -funroll-loops
7740 @opindex funroll-loops
7741 Unroll loops whose number of iterations can be determined at compile time or
7742 upon entry to the loop. @option{-funroll-loops} implies
7743 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
7744 It also turns on complete loop peeling (i.e.@: complete removal of loops with
7745 small constant number of iterations). This option makes code larger, and may
7746 or may not make it run faster.
7748 Enabled with @option{-fprofile-use}.
7750 @item -funroll-all-loops
7751 @opindex funroll-all-loops
7752 Unroll all loops, even if their number of iterations is uncertain when
7753 the loop is entered. This usually makes programs run more slowly.
7754 @option{-funroll-all-loops} implies the same options as
7755 @option{-funroll-loops}.
7758 @opindex fpeel-loops
7759 Peels the loops for that there is enough information that they do not
7760 roll much (from profile feedback). It also turns on complete loop peeling
7761 (i.e.@: complete removal of loops with small constant number of iterations).
7763 Enabled with @option{-fprofile-use}.
7765 @item -fmove-loop-invariants
7766 @opindex fmove-loop-invariants
7767 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
7768 at level @option{-O1}
7770 @item -funswitch-loops
7771 @opindex funswitch-loops
7772 Move branches with loop invariant conditions out of the loop, with duplicates
7773 of the loop on both branches (modified according to result of the condition).
7775 @item -ffunction-sections
7776 @itemx -fdata-sections
7777 @opindex ffunction-sections
7778 @opindex fdata-sections
7779 Place each function or data item into its own section in the output
7780 file if the target supports arbitrary sections. The name of the
7781 function or the name of the data item determines the section's name
7784 Use these options on systems where the linker can perform optimizations
7785 to improve locality of reference in the instruction space. Most systems
7786 using the ELF object format and SPARC processors running Solaris 2 have
7787 linkers with such optimizations. AIX may have these optimizations in
7790 Only use these options when there are significant benefits from doing
7791 so. When you specify these options, the assembler and linker will
7792 create larger object and executable files and will also be slower.
7793 You will not be able to use @code{gprof} on all systems if you
7794 specify this option and you may have problems with debugging if
7795 you specify both this option and @option{-g}.
7797 @item -fbranch-target-load-optimize
7798 @opindex fbranch-target-load-optimize
7799 Perform branch target register load optimization before prologue / epilogue
7801 The use of target registers can typically be exposed only during reload,
7802 thus hoisting loads out of loops and doing inter-block scheduling needs
7803 a separate optimization pass.
7805 @item -fbranch-target-load-optimize2
7806 @opindex fbranch-target-load-optimize2
7807 Perform branch target register load optimization after prologue / epilogue
7810 @item -fbtr-bb-exclusive
7811 @opindex fbtr-bb-exclusive
7812 When performing branch target register load optimization, don't reuse
7813 branch target registers in within any basic block.
7815 @item -fstack-protector
7816 @opindex fstack-protector
7817 Emit extra code to check for buffer overflows, such as stack smashing
7818 attacks. This is done by adding a guard variable to functions with
7819 vulnerable objects. This includes functions that call alloca, and
7820 functions with buffers larger than 8 bytes. The guards are initialized
7821 when a function is entered and then checked when the function exits.
7822 If a guard check fails, an error message is printed and the program exits.
7824 @item -fstack-protector-all
7825 @opindex fstack-protector-all
7826 Like @option{-fstack-protector} except that all functions are protected.
7828 @item -fsection-anchors
7829 @opindex fsection-anchors
7830 Try to reduce the number of symbolic address calculations by using
7831 shared ``anchor'' symbols to address nearby objects. This transformation
7832 can help to reduce the number of GOT entries and GOT accesses on some
7835 For example, the implementation of the following function @code{foo}:
7839 int foo (void) @{ return a + b + c; @}
7842 would usually calculate the addresses of all three variables, but if you
7843 compile it with @option{-fsection-anchors}, it will access the variables
7844 from a common anchor point instead. The effect is similar to the
7845 following pseudocode (which isn't valid C):
7850 register int *xr = &x;
7851 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
7855 Not all targets support this option.
7857 @item --param @var{name}=@var{value}
7859 In some places, GCC uses various constants to control the amount of
7860 optimization that is done. For example, GCC will not inline functions
7861 that contain more that a certain number of instructions. You can
7862 control some of these constants on the command-line using the
7863 @option{--param} option.
7865 The names of specific parameters, and the meaning of the values, are
7866 tied to the internals of the compiler, and are subject to change
7867 without notice in future releases.
7869 In each case, the @var{value} is an integer. The allowable choices for
7870 @var{name} are given in the following table:
7873 @item struct-reorg-cold-struct-ratio
7874 The threshold ratio (as a percentage) between a structure frequency
7875 and the frequency of the hottest structure in the program. This parameter
7876 is used by struct-reorg optimization enabled by @option{-fipa-struct-reorg}.
7877 We say that if the ratio of a structure frequency, calculated by profiling,
7878 to the hottest structure frequency in the program is less than this
7879 parameter, then structure reorganization is not applied to this structure.
7882 @item predictable-branch-cost-outcome
7883 When branch is predicted to be taken with probability lower than this threshold
7884 (in percent), then it is considered well predictable. The default is 10.
7886 @item max-crossjump-edges
7887 The maximum number of incoming edges to consider for crossjumping.
7888 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
7889 the number of edges incoming to each block. Increasing values mean
7890 more aggressive optimization, making the compile time increase with
7891 probably small improvement in executable size.
7893 @item min-crossjump-insns
7894 The minimum number of instructions which must be matched at the end
7895 of two blocks before crossjumping will be performed on them. This
7896 value is ignored in the case where all instructions in the block being
7897 crossjumped from are matched. The default value is 5.
7899 @item max-grow-copy-bb-insns
7900 The maximum code size expansion factor when copying basic blocks
7901 instead of jumping. The expansion is relative to a jump instruction.
7902 The default value is 8.
7904 @item max-goto-duplication-insns
7905 The maximum number of instructions to duplicate to a block that jumps
7906 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
7907 passes, GCC factors computed gotos early in the compilation process,
7908 and unfactors them as late as possible. Only computed jumps at the
7909 end of a basic blocks with no more than max-goto-duplication-insns are
7910 unfactored. The default value is 8.
7912 @item max-delay-slot-insn-search
7913 The maximum number of instructions to consider when looking for an
7914 instruction to fill a delay slot. If more than this arbitrary number of
7915 instructions is searched, the time savings from filling the delay slot
7916 will be minimal so stop searching. Increasing values mean more
7917 aggressive optimization, making the compile time increase with probably
7918 small improvement in executable run time.
7920 @item max-delay-slot-live-search
7921 When trying to fill delay slots, the maximum number of instructions to
7922 consider when searching for a block with valid live register
7923 information. Increasing this arbitrarily chosen value means more
7924 aggressive optimization, increasing the compile time. This parameter
7925 should be removed when the delay slot code is rewritten to maintain the
7928 @item max-gcse-memory
7929 The approximate maximum amount of memory that will be allocated in
7930 order to perform the global common subexpression elimination
7931 optimization. If more memory than specified is required, the
7932 optimization will not be done.
7934 @item max-pending-list-length
7935 The maximum number of pending dependencies scheduling will allow
7936 before flushing the current state and starting over. Large functions
7937 with few branches or calls can create excessively large lists which
7938 needlessly consume memory and resources.
7940 @item max-inline-insns-single
7941 Several parameters control the tree inliner used in gcc.
7942 This number sets the maximum number of instructions (counted in GCC's
7943 internal representation) in a single function that the tree inliner
7944 will consider for inlining. This only affects functions declared
7945 inline and methods implemented in a class declaration (C++).
7946 The default value is 300.
7948 @item max-inline-insns-auto
7949 When you use @option{-finline-functions} (included in @option{-O3}),
7950 a lot of functions that would otherwise not be considered for inlining
7951 by the compiler will be investigated. To those functions, a different
7952 (more restrictive) limit compared to functions declared inline can
7954 The default value is 50.
7956 @item large-function-insns
7957 The limit specifying really large functions. For functions larger than this
7958 limit after inlining, inlining is constrained by
7959 @option{--param large-function-growth}. This parameter is useful primarily
7960 to avoid extreme compilation time caused by non-linear algorithms used by the
7962 The default value is 2700.
7964 @item large-function-growth
7965 Specifies maximal growth of large function caused by inlining in percents.
7966 The default value is 100 which limits large function growth to 2.0 times
7969 @item large-unit-insns
7970 The limit specifying large translation unit. Growth caused by inlining of
7971 units larger than this limit is limited by @option{--param inline-unit-growth}.
7972 For small units this might be too tight (consider unit consisting of function A
7973 that is inline and B that just calls A three time. If B is small relative to
7974 A, the growth of unit is 300\% and yet such inlining is very sane. For very
7975 large units consisting of small inlineable functions however the overall unit
7976 growth limit is needed to avoid exponential explosion of code size. Thus for
7977 smaller units, the size is increased to @option{--param large-unit-insns}
7978 before applying @option{--param inline-unit-growth}. The default is 10000
7980 @item inline-unit-growth
7981 Specifies maximal overall growth of the compilation unit caused by inlining.
7982 The default value is 30 which limits unit growth to 1.3 times the original
7985 @item ipcp-unit-growth
7986 Specifies maximal overall growth of the compilation unit caused by
7987 interprocedural constant propagation. The default value is 10 which limits
7988 unit growth to 1.1 times the original size.
7990 @item large-stack-frame
7991 The limit specifying large stack frames. While inlining the algorithm is trying
7992 to not grow past this limit too much. Default value is 256 bytes.
7994 @item large-stack-frame-growth
7995 Specifies maximal growth of large stack frames caused by inlining in percents.
7996 The default value is 1000 which limits large stack frame growth to 11 times
7999 @item max-inline-insns-recursive
8000 @itemx max-inline-insns-recursive-auto
8001 Specifies maximum number of instructions out-of-line copy of self recursive inline
8002 function can grow into by performing recursive inlining.
8004 For functions declared inline @option{--param max-inline-insns-recursive} is
8005 taken into account. For function not declared inline, recursive inlining
8006 happens only when @option{-finline-functions} (included in @option{-O3}) is
8007 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
8008 default value is 450.
8010 @item max-inline-recursive-depth
8011 @itemx max-inline-recursive-depth-auto
8012 Specifies maximum recursion depth used by the recursive inlining.
8014 For functions declared inline @option{--param max-inline-recursive-depth} is
8015 taken into account. For function not declared inline, recursive inlining
8016 happens only when @option{-finline-functions} (included in @option{-O3}) is
8017 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
8020 @item min-inline-recursive-probability
8021 Recursive inlining is profitable only for function having deep recursion
8022 in average and can hurt for function having little recursion depth by
8023 increasing the prologue size or complexity of function body to other
8026 When profile feedback is available (see @option{-fprofile-generate}) the actual
8027 recursion depth can be guessed from probability that function will recurse via
8028 given call expression. This parameter limits inlining only to call expression
8029 whose probability exceeds given threshold (in percents). The default value is
8032 @item early-inlining-insns
8033 Specify growth that early inliner can make. In effect it increases amount of
8034 inlining for code having large abstraction penalty. The default value is 8.
8036 @item max-early-inliner-iterations
8037 @itemx max-early-inliner-iterations
8038 Limit of iterations of early inliner. This basically bounds number of nested
8039 indirect calls early inliner can resolve. Deeper chains are still handled by
8042 @item min-vect-loop-bound
8043 The minimum number of iterations under which a loop will not get vectorized
8044 when @option{-ftree-vectorize} is used. The number of iterations after
8045 vectorization needs to be greater than the value specified by this option
8046 to allow vectorization. The default value is 0.
8048 @item max-unrolled-insns
8049 The maximum number of instructions that a loop should have if that loop
8050 is unrolled, and if the loop is unrolled, it determines how many times
8051 the loop code is unrolled.
8053 @item max-average-unrolled-insns
8054 The maximum number of instructions biased by probabilities of their execution
8055 that a loop should have if that loop is unrolled, and if the loop is unrolled,
8056 it determines how many times the loop code is unrolled.
8058 @item max-unroll-times
8059 The maximum number of unrollings of a single loop.
8061 @item max-peeled-insns
8062 The maximum number of instructions that a loop should have if that loop
8063 is peeled, and if the loop is peeled, it determines how many times
8064 the loop code is peeled.
8066 @item max-peel-times
8067 The maximum number of peelings of a single loop.
8069 @item max-completely-peeled-insns
8070 The maximum number of insns of a completely peeled loop.
8072 @item max-completely-peel-times
8073 The maximum number of iterations of a loop to be suitable for complete peeling.
8075 @item max-unswitch-insns
8076 The maximum number of insns of an unswitched loop.
8078 @item max-unswitch-level
8079 The maximum number of branches unswitched in a single loop.
8082 The minimum cost of an expensive expression in the loop invariant motion.
8084 @item iv-consider-all-candidates-bound
8085 Bound on number of candidates for induction variables below that
8086 all candidates are considered for each use in induction variable
8087 optimizations. Only the most relevant candidates are considered
8088 if there are more candidates, to avoid quadratic time complexity.
8090 @item iv-max-considered-uses
8091 The induction variable optimizations give up on loops that contain more
8092 induction variable uses.
8094 @item iv-always-prune-cand-set-bound
8095 If number of candidates in the set is smaller than this value,
8096 we always try to remove unnecessary ivs from the set during its
8097 optimization when a new iv is added to the set.
8099 @item scev-max-expr-size
8100 Bound on size of expressions used in the scalar evolutions analyzer.
8101 Large expressions slow the analyzer.
8103 @item omega-max-vars
8104 The maximum number of variables in an Omega constraint system.
8105 The default value is 128.
8107 @item omega-max-geqs
8108 The maximum number of inequalities in an Omega constraint system.
8109 The default value is 256.
8112 The maximum number of equalities in an Omega constraint system.
8113 The default value is 128.
8115 @item omega-max-wild-cards
8116 The maximum number of wildcard variables that the Omega solver will
8117 be able to insert. The default value is 18.
8119 @item omega-hash-table-size
8120 The size of the hash table in the Omega solver. The default value is
8123 @item omega-max-keys
8124 The maximal number of keys used by the Omega solver. The default
8127 @item omega-eliminate-redundant-constraints
8128 When set to 1, use expensive methods to eliminate all redundant
8129 constraints. The default value is 0.
8131 @item vect-max-version-for-alignment-checks
8132 The maximum number of runtime checks that can be performed when
8133 doing loop versioning for alignment in the vectorizer. See option
8134 ftree-vect-loop-version for more information.
8136 @item vect-max-version-for-alias-checks
8137 The maximum number of runtime checks that can be performed when
8138 doing loop versioning for alias in the vectorizer. See option
8139 ftree-vect-loop-version for more information.
8141 @item max-iterations-to-track
8143 The maximum number of iterations of a loop the brute force algorithm
8144 for analysis of # of iterations of the loop tries to evaluate.
8146 @item hot-bb-count-fraction
8147 Select fraction of the maximal count of repetitions of basic block in program
8148 given basic block needs to have to be considered hot.
8150 @item hot-bb-frequency-fraction
8151 Select fraction of the maximal frequency of executions of basic block in
8152 function given basic block needs to have to be considered hot
8154 @item max-predicted-iterations
8155 The maximum number of loop iterations we predict statically. This is useful
8156 in cases where function contain single loop with known bound and other loop
8157 with unknown. We predict the known number of iterations correctly, while
8158 the unknown number of iterations average to roughly 10. This means that the
8159 loop without bounds would appear artificially cold relative to the other one.
8161 @item align-threshold
8163 Select fraction of the maximal frequency of executions of basic block in
8164 function given basic block will get aligned.
8166 @item align-loop-iterations
8168 A loop expected to iterate at lest the selected number of iterations will get
8171 @item tracer-dynamic-coverage
8172 @itemx tracer-dynamic-coverage-feedback
8174 This value is used to limit superblock formation once the given percentage of
8175 executed instructions is covered. This limits unnecessary code size
8178 The @option{tracer-dynamic-coverage-feedback} is used only when profile
8179 feedback is available. The real profiles (as opposed to statically estimated
8180 ones) are much less balanced allowing the threshold to be larger value.
8182 @item tracer-max-code-growth
8183 Stop tail duplication once code growth has reached given percentage. This is
8184 rather hokey argument, as most of the duplicates will be eliminated later in
8185 cross jumping, so it may be set to much higher values than is the desired code
8188 @item tracer-min-branch-ratio
8190 Stop reverse growth when the reverse probability of best edge is less than this
8191 threshold (in percent).
8193 @item tracer-min-branch-ratio
8194 @itemx tracer-min-branch-ratio-feedback
8196 Stop forward growth if the best edge do have probability lower than this
8199 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
8200 compilation for profile feedback and one for compilation without. The value
8201 for compilation with profile feedback needs to be more conservative (higher) in
8202 order to make tracer effective.
8204 @item max-cse-path-length
8206 Maximum number of basic blocks on path that cse considers. The default is 10.
8209 The maximum instructions CSE process before flushing. The default is 1000.
8211 @item ggc-min-expand
8213 GCC uses a garbage collector to manage its own memory allocation. This
8214 parameter specifies the minimum percentage by which the garbage
8215 collector's heap should be allowed to expand between collections.
8216 Tuning this may improve compilation speed; it has no effect on code
8219 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
8220 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
8221 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
8222 GCC is not able to calculate RAM on a particular platform, the lower
8223 bound of 30% is used. Setting this parameter and
8224 @option{ggc-min-heapsize} to zero causes a full collection to occur at
8225 every opportunity. This is extremely slow, but can be useful for
8228 @item ggc-min-heapsize
8230 Minimum size of the garbage collector's heap before it begins bothering
8231 to collect garbage. The first collection occurs after the heap expands
8232 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
8233 tuning this may improve compilation speed, and has no effect on code
8236 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
8237 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
8238 with a lower bound of 4096 (four megabytes) and an upper bound of
8239 131072 (128 megabytes). If GCC is not able to calculate RAM on a
8240 particular platform, the lower bound is used. Setting this parameter
8241 very large effectively disables garbage collection. Setting this
8242 parameter and @option{ggc-min-expand} to zero causes a full collection
8243 to occur at every opportunity.
8245 @item max-reload-search-insns
8246 The maximum number of instruction reload should look backward for equivalent
8247 register. Increasing values mean more aggressive optimization, making the
8248 compile time increase with probably slightly better performance. The default
8251 @item max-cselib-memory-locations
8252 The maximum number of memory locations cselib should take into account.
8253 Increasing values mean more aggressive optimization, making the compile time
8254 increase with probably slightly better performance. The default value is 500.
8256 @item reorder-blocks-duplicate
8257 @itemx reorder-blocks-duplicate-feedback
8259 Used by basic block reordering pass to decide whether to use unconditional
8260 branch or duplicate the code on its destination. Code is duplicated when its
8261 estimated size is smaller than this value multiplied by the estimated size of
8262 unconditional jump in the hot spots of the program.
8264 The @option{reorder-block-duplicate-feedback} is used only when profile
8265 feedback is available and may be set to higher values than
8266 @option{reorder-block-duplicate} since information about the hot spots is more
8269 @item max-sched-ready-insns
8270 The maximum number of instructions ready to be issued the scheduler should
8271 consider at any given time during the first scheduling pass. Increasing
8272 values mean more thorough searches, making the compilation time increase
8273 with probably little benefit. The default value is 100.
8275 @item max-sched-region-blocks
8276 The maximum number of blocks in a region to be considered for
8277 interblock scheduling. The default value is 10.
8279 @item max-pipeline-region-blocks
8280 The maximum number of blocks in a region to be considered for
8281 pipelining in the selective scheduler. The default value is 15.
8283 @item max-sched-region-insns
8284 The maximum number of insns in a region to be considered for
8285 interblock scheduling. The default value is 100.
8287 @item max-pipeline-region-insns
8288 The maximum number of insns in a region to be considered for
8289 pipelining in the selective scheduler. The default value is 200.
8292 The minimum probability (in percents) of reaching a source block
8293 for interblock speculative scheduling. The default value is 40.
8295 @item max-sched-extend-regions-iters
8296 The maximum number of iterations through CFG to extend regions.
8297 0 - disable region extension,
8298 N - do at most N iterations.
8299 The default value is 0.
8301 @item max-sched-insn-conflict-delay
8302 The maximum conflict delay for an insn to be considered for speculative motion.
8303 The default value is 3.
8305 @item sched-spec-prob-cutoff
8306 The minimal probability of speculation success (in percents), so that
8307 speculative insn will be scheduled.
8308 The default value is 40.
8310 @item sched-mem-true-dep-cost
8311 Minimal distance (in CPU cycles) between store and load targeting same
8312 memory locations. The default value is 1.
8314 @item selsched-max-lookahead
8315 The maximum size of the lookahead window of selective scheduling. It is a
8316 depth of search for available instructions.
8317 The default value is 50.
8319 @item selsched-max-sched-times
8320 The maximum number of times that an instruction will be scheduled during
8321 selective scheduling. This is the limit on the number of iterations
8322 through which the instruction may be pipelined. The default value is 2.
8324 @item selsched-max-insns-to-rename
8325 The maximum number of best instructions in the ready list that are considered
8326 for renaming in the selective scheduler. The default value is 2.
8328 @item max-last-value-rtl
8329 The maximum size measured as number of RTLs that can be recorded in an expression
8330 in combiner for a pseudo register as last known value of that register. The default
8333 @item integer-share-limit
8334 Small integer constants can use a shared data structure, reducing the
8335 compiler's memory usage and increasing its speed. This sets the maximum
8336 value of a shared integer constant. The default value is 256.
8338 @item min-virtual-mappings
8339 Specifies the minimum number of virtual mappings in the incremental
8340 SSA updater that should be registered to trigger the virtual mappings
8341 heuristic defined by virtual-mappings-ratio. The default value is
8344 @item virtual-mappings-ratio
8345 If the number of virtual mappings is virtual-mappings-ratio bigger
8346 than the number of virtual symbols to be updated, then the incremental
8347 SSA updater switches to a full update for those symbols. The default
8350 @item ssp-buffer-size
8351 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
8352 protection when @option{-fstack-protection} is used.
8354 @item max-jump-thread-duplication-stmts
8355 Maximum number of statements allowed in a block that needs to be
8356 duplicated when threading jumps.
8358 @item max-fields-for-field-sensitive
8359 Maximum number of fields in a structure we will treat in
8360 a field sensitive manner during pointer analysis. The default is zero
8361 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
8363 @item prefetch-latency
8364 Estimate on average number of instructions that are executed before
8365 prefetch finishes. The distance we prefetch ahead is proportional
8366 to this constant. Increasing this number may also lead to less
8367 streams being prefetched (see @option{simultaneous-prefetches}).
8369 @item simultaneous-prefetches
8370 Maximum number of prefetches that can run at the same time.
8372 @item l1-cache-line-size
8373 The size of cache line in L1 cache, in bytes.
8376 The size of L1 cache, in kilobytes.
8379 The size of L2 cache, in kilobytes.
8381 @item min-insn-to-prefetch-ratio
8382 The minimum ratio between the number of instructions and the
8383 number of prefetches to enable prefetching in a loop with an
8386 @item prefetch-min-insn-to-mem-ratio
8387 The minimum ratio between the number of instructions and the
8388 number of memory references to enable prefetching in a loop.
8390 @item use-canonical-types
8391 Whether the compiler should use the ``canonical'' type system. By
8392 default, this should always be 1, which uses a more efficient internal
8393 mechanism for comparing types in C++ and Objective-C++. However, if
8394 bugs in the canonical type system are causing compilation failures,
8395 set this value to 0 to disable canonical types.
8397 @item switch-conversion-max-branch-ratio
8398 Switch initialization conversion will refuse to create arrays that are
8399 bigger than @option{switch-conversion-max-branch-ratio} times the number of
8400 branches in the switch.
8402 @item max-partial-antic-length
8403 Maximum length of the partial antic set computed during the tree
8404 partial redundancy elimination optimization (@option{-ftree-pre}) when
8405 optimizing at @option{-O3} and above. For some sorts of source code
8406 the enhanced partial redundancy elimination optimization can run away,
8407 consuming all of the memory available on the host machine. This
8408 parameter sets a limit on the length of the sets that are computed,
8409 which prevents the runaway behavior. Setting a value of 0 for
8410 this parameter will allow an unlimited set length.
8412 @item sccvn-max-scc-size
8413 Maximum size of a strongly connected component (SCC) during SCCVN
8414 processing. If this limit is hit, SCCVN processing for the whole
8415 function will not be done and optimizations depending on it will
8416 be disabled. The default maximum SCC size is 10000.
8418 @item ira-max-loops-num
8419 IRA uses a regional register allocation by default. If a function
8420 contains loops more than number given by the parameter, only at most
8421 given number of the most frequently executed loops will form regions
8422 for the regional register allocation. The default value of the
8425 @item ira-max-conflict-table-size
8426 Although IRA uses a sophisticated algorithm of compression conflict
8427 table, the table can be still big for huge functions. If the conflict
8428 table for a function could be more than size in MB given by the
8429 parameter, the conflict table is not built and faster, simpler, and
8430 lower quality register allocation algorithm will be used. The
8431 algorithm do not use pseudo-register conflicts. The default value of
8432 the parameter is 2000.
8434 @item ira-loop-reserved-regs
8435 IRA can be used to evaluate more accurate register pressure in loops
8436 for decision to move loop invariants (see @option{-O3}). The number
8437 of available registers reserved for some other purposes is described
8438 by this parameter. The default value of the parameter is 2 which is
8439 minimal number of registers needed for execution of typical
8440 instruction. This value is the best found from numerous experiments.
8442 @item loop-invariant-max-bbs-in-loop
8443 Loop invariant motion can be very expensive, both in compile time and
8444 in amount of needed compile time memory, with very large loops. Loops
8445 with more basic blocks than this parameter won't have loop invariant
8446 motion optimization performed on them. The default value of the
8447 parameter is 1000 for -O1 and 10000 for -O2 and above.
8449 @item min-nondebug-insn-uid
8450 Use uids starting at this parameter for nondebug insns. The range below
8451 the parameter is reserved exclusively for debug insns created by
8452 @option{-fvar-tracking-assignments}, but debug insns may get
8453 (non-overlapping) uids above it if the reserved range is exhausted.
8455 @item ipa-sra-ptr-growth-factor
8456 IPA-SRA will replace a pointer to an aggregate with one or more new
8457 parameters only when their cumulative size is less or equal to
8458 @option{ipa-sra-ptr-growth-factor} times the size of the original
8464 @node Preprocessor Options
8465 @section Options Controlling the Preprocessor
8466 @cindex preprocessor options
8467 @cindex options, preprocessor
8469 These options control the C preprocessor, which is run on each C source
8470 file before actual compilation.
8472 If you use the @option{-E} option, nothing is done except preprocessing.
8473 Some of these options make sense only together with @option{-E} because
8474 they cause the preprocessor output to be unsuitable for actual
8478 @item -Wp,@var{option}
8480 You can use @option{-Wp,@var{option}} to bypass the compiler driver
8481 and pass @var{option} directly through to the preprocessor. If
8482 @var{option} contains commas, it is split into multiple options at the
8483 commas. However, many options are modified, translated or interpreted
8484 by the compiler driver before being passed to the preprocessor, and
8485 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
8486 interface is undocumented and subject to change, so whenever possible
8487 you should avoid using @option{-Wp} and let the driver handle the
8490 @item -Xpreprocessor @var{option}
8491 @opindex Xpreprocessor
8492 Pass @var{option} as an option to the preprocessor. You can use this to
8493 supply system-specific preprocessor options which GCC does not know how to
8496 If you want to pass an option that takes an argument, you must use
8497 @option{-Xpreprocessor} twice, once for the option and once for the argument.
8500 @include cppopts.texi
8502 @node Assembler Options
8503 @section Passing Options to the Assembler
8505 @c prevent bad page break with this line
8506 You can pass options to the assembler.
8509 @item -Wa,@var{option}
8511 Pass @var{option} as an option to the assembler. If @var{option}
8512 contains commas, it is split into multiple options at the commas.
8514 @item -Xassembler @var{option}
8516 Pass @var{option} as an option to the assembler. You can use this to
8517 supply system-specific assembler options which GCC does not know how to
8520 If you want to pass an option that takes an argument, you must use
8521 @option{-Xassembler} twice, once for the option and once for the argument.
8526 @section Options for Linking
8527 @cindex link options
8528 @cindex options, linking
8530 These options come into play when the compiler links object files into
8531 an executable output file. They are meaningless if the compiler is
8532 not doing a link step.
8536 @item @var{object-file-name}
8537 A file name that does not end in a special recognized suffix is
8538 considered to name an object file or library. (Object files are
8539 distinguished from libraries by the linker according to the file
8540 contents.) If linking is done, these object files are used as input
8549 If any of these options is used, then the linker is not run, and
8550 object file names should not be used as arguments. @xref{Overall
8554 @item -l@var{library}
8555 @itemx -l @var{library}
8557 Search the library named @var{library} when linking. (The second
8558 alternative with the library as a separate argument is only for
8559 POSIX compliance and is not recommended.)
8561 It makes a difference where in the command you write this option; the
8562 linker searches and processes libraries and object files in the order they
8563 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
8564 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
8565 to functions in @samp{z}, those functions may not be loaded.
8567 The linker searches a standard list of directories for the library,
8568 which is actually a file named @file{lib@var{library}.a}. The linker
8569 then uses this file as if it had been specified precisely by name.
8571 The directories searched include several standard system directories
8572 plus any that you specify with @option{-L}.
8574 Normally the files found this way are library files---archive files
8575 whose members are object files. The linker handles an archive file by
8576 scanning through it for members which define symbols that have so far
8577 been referenced but not defined. But if the file that is found is an
8578 ordinary object file, it is linked in the usual fashion. The only
8579 difference between using an @option{-l} option and specifying a file name
8580 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
8581 and searches several directories.
8585 You need this special case of the @option{-l} option in order to
8586 link an Objective-C or Objective-C++ program.
8589 @opindex nostartfiles
8590 Do not use the standard system startup files when linking.
8591 The standard system libraries are used normally, unless @option{-nostdlib}
8592 or @option{-nodefaultlibs} is used.
8594 @item -nodefaultlibs
8595 @opindex nodefaultlibs
8596 Do not use the standard system libraries when linking.
8597 Only the libraries you specify will be passed to the linker, options
8598 specifying linkage of the system libraries, such as @code{-static-libgcc}
8599 or @code{-shared-libgcc}, will be ignored.
8600 The standard startup files are used normally, unless @option{-nostartfiles}
8601 is used. The compiler may generate calls to @code{memcmp},
8602 @code{memset}, @code{memcpy} and @code{memmove}.
8603 These entries are usually resolved by entries in
8604 libc. These entry points should be supplied through some other
8605 mechanism when this option is specified.
8609 Do not use the standard system startup files or libraries when linking.
8610 No startup files and only the libraries you specify will be passed to
8611 the linker, options specifying linkage of the system libraries, such as
8612 @code{-static-libgcc} or @code{-shared-libgcc}, will be ignored.
8613 The compiler may generate calls to @code{memcmp}, @code{memset},
8614 @code{memcpy} and @code{memmove}.
8615 These entries are usually resolved by entries in
8616 libc. These entry points should be supplied through some other
8617 mechanism when this option is specified.
8619 @cindex @option{-lgcc}, use with @option{-nostdlib}
8620 @cindex @option{-nostdlib} and unresolved references
8621 @cindex unresolved references and @option{-nostdlib}
8622 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
8623 @cindex @option{-nodefaultlibs} and unresolved references
8624 @cindex unresolved references and @option{-nodefaultlibs}
8625 One of the standard libraries bypassed by @option{-nostdlib} and
8626 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
8627 that GCC uses to overcome shortcomings of particular machines, or special
8628 needs for some languages.
8629 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
8630 Collection (GCC) Internals},
8631 for more discussion of @file{libgcc.a}.)
8632 In most cases, you need @file{libgcc.a} even when you want to avoid
8633 other standard libraries. In other words, when you specify @option{-nostdlib}
8634 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
8635 This ensures that you have no unresolved references to internal GCC
8636 library subroutines. (For example, @samp{__main}, used to ensure C++
8637 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
8638 GNU Compiler Collection (GCC) Internals}.)
8642 Produce a position independent executable on targets which support it.
8643 For predictable results, you must also specify the same set of options
8644 that were used to generate code (@option{-fpie}, @option{-fPIE},
8645 or model suboptions) when you specify this option.
8649 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
8650 that support it. This instructs the linker to add all symbols, not
8651 only used ones, to the dynamic symbol table. This option is needed
8652 for some uses of @code{dlopen} or to allow obtaining backtraces
8653 from within a program.
8657 Remove all symbol table and relocation information from the executable.
8661 On systems that support dynamic linking, this prevents linking with the shared
8662 libraries. On other systems, this option has no effect.
8666 Produce a shared object which can then be linked with other objects to
8667 form an executable. Not all systems support this option. For predictable
8668 results, you must also specify the same set of options that were used to
8669 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
8670 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
8671 needs to build supplementary stub code for constructors to work. On
8672 multi-libbed systems, @samp{gcc -shared} must select the correct support
8673 libraries to link against. Failing to supply the correct flags may lead
8674 to subtle defects. Supplying them in cases where they are not necessary
8677 @item -shared-libgcc
8678 @itemx -static-libgcc
8679 @opindex shared-libgcc
8680 @opindex static-libgcc
8681 On systems that provide @file{libgcc} as a shared library, these options
8682 force the use of either the shared or static version respectively.
8683 If no shared version of @file{libgcc} was built when the compiler was
8684 configured, these options have no effect.
8686 There are several situations in which an application should use the
8687 shared @file{libgcc} instead of the static version. The most common
8688 of these is when the application wishes to throw and catch exceptions
8689 across different shared libraries. In that case, each of the libraries
8690 as well as the application itself should use the shared @file{libgcc}.
8692 Therefore, the G++ and GCJ drivers automatically add
8693 @option{-shared-libgcc} whenever you build a shared library or a main
8694 executable, because C++ and Java programs typically use exceptions, so
8695 this is the right thing to do.
8697 If, instead, you use the GCC driver to create shared libraries, you may
8698 find that they will not always be linked with the shared @file{libgcc}.
8699 If GCC finds, at its configuration time, that you have a non-GNU linker
8700 or a GNU linker that does not support option @option{--eh-frame-hdr},
8701 it will link the shared version of @file{libgcc} into shared libraries
8702 by default. Otherwise, it will take advantage of the linker and optimize
8703 away the linking with the shared version of @file{libgcc}, linking with
8704 the static version of libgcc by default. This allows exceptions to
8705 propagate through such shared libraries, without incurring relocation
8706 costs at library load time.
8708 However, if a library or main executable is supposed to throw or catch
8709 exceptions, you must link it using the G++ or GCJ driver, as appropriate
8710 for the languages used in the program, or using the option
8711 @option{-shared-libgcc}, such that it is linked with the shared
8714 @item -static-libstdc++
8715 When the @command{g++} program is used to link a C++ program, it will
8716 normally automatically link against @option{libstdc++}. If
8717 @file{libstdc++} is available as a shared library, and the
8718 @option{-static} option is not used, then this will link against the
8719 shared version of @file{libstdc++}. That is normally fine. However, it
8720 is sometimes useful to freeze the version of @file{libstdc++} used by
8721 the program without going all the way to a fully static link. The
8722 @option{-static-libstdc++} option directs the @command{g++} driver to
8723 link @file{libstdc++} statically, without necessarily linking other
8724 libraries statically.
8728 Bind references to global symbols when building a shared object. Warn
8729 about any unresolved references (unless overridden by the link editor
8730 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
8733 @item -T @var{script}
8735 @cindex linker script
8736 Use @var{script} as the linker script. This option is supported by most
8737 systems using the GNU linker. On some targets, such as bare-board
8738 targets without an operating system, the @option{-T} option may be required
8739 when linking to avoid references to undefined symbols.
8741 @item -Xlinker @var{option}
8743 Pass @var{option} as an option to the linker. You can use this to
8744 supply system-specific linker options which GCC does not know how to
8747 If you want to pass an option that takes a separate argument, you must use
8748 @option{-Xlinker} twice, once for the option and once for the argument.
8749 For example, to pass @option{-assert definitions}, you must write
8750 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
8751 @option{-Xlinker "-assert definitions"}, because this passes the entire
8752 string as a single argument, which is not what the linker expects.
8754 When using the GNU linker, it is usually more convenient to pass
8755 arguments to linker options using the @option{@var{option}=@var{value}}
8756 syntax than as separate arguments. For example, you can specify
8757 @samp{-Xlinker -Map=output.map} rather than
8758 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
8759 this syntax for command-line options.
8761 @item -Wl,@var{option}
8763 Pass @var{option} as an option to the linker. If @var{option} contains
8764 commas, it is split into multiple options at the commas. You can use this
8765 syntax to pass an argument to the option.
8766 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
8767 linker. When using the GNU linker, you can also get the same effect with
8768 @samp{-Wl,-Map=output.map}.
8770 @item -u @var{symbol}
8772 Pretend the symbol @var{symbol} is undefined, to force linking of
8773 library modules to define it. You can use @option{-u} multiple times with
8774 different symbols to force loading of additional library modules.
8777 @node Directory Options
8778 @section Options for Directory Search
8779 @cindex directory options
8780 @cindex options, directory search
8783 These options specify directories to search for header files, for
8784 libraries and for parts of the compiler:
8789 Add the directory @var{dir} to the head of the list of directories to be
8790 searched for header files. This can be used to override a system header
8791 file, substituting your own version, since these directories are
8792 searched before the system header file directories. However, you should
8793 not use this option to add directories that contain vendor-supplied
8794 system header files (use @option{-isystem} for that). If you use more than
8795 one @option{-I} option, the directories are scanned in left-to-right
8796 order; the standard system directories come after.
8798 If a standard system include directory, or a directory specified with
8799 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
8800 option will be ignored. The directory will still be searched but as a
8801 system directory at its normal position in the system include chain.
8802 This is to ensure that GCC's procedure to fix buggy system headers and
8803 the ordering for the include_next directive are not inadvertently changed.
8804 If you really need to change the search order for system directories,
8805 use the @option{-nostdinc} and/or @option{-isystem} options.
8807 @item -iquote@var{dir}
8809 Add the directory @var{dir} to the head of the list of directories to
8810 be searched for header files only for the case of @samp{#include
8811 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
8812 otherwise just like @option{-I}.
8816 Add directory @var{dir} to the list of directories to be searched
8819 @item -B@var{prefix}
8821 This option specifies where to find the executables, libraries,
8822 include files, and data files of the compiler itself.
8824 The compiler driver program runs one or more of the subprograms
8825 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
8826 @var{prefix} as a prefix for each program it tries to run, both with and
8827 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
8829 For each subprogram to be run, the compiler driver first tries the
8830 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
8831 was not specified, the driver tries two standard prefixes, which are
8832 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
8833 those results in a file name that is found, the unmodified program
8834 name is searched for using the directories specified in your
8835 @env{PATH} environment variable.
8837 The compiler will check to see if the path provided by the @option{-B}
8838 refers to a directory, and if necessary it will add a directory
8839 separator character at the end of the path.
8841 @option{-B} prefixes that effectively specify directory names also apply
8842 to libraries in the linker, because the compiler translates these
8843 options into @option{-L} options for the linker. They also apply to
8844 includes files in the preprocessor, because the compiler translates these
8845 options into @option{-isystem} options for the preprocessor. In this case,
8846 the compiler appends @samp{include} to the prefix.
8848 The run-time support file @file{libgcc.a} can also be searched for using
8849 the @option{-B} prefix, if needed. If it is not found there, the two
8850 standard prefixes above are tried, and that is all. The file is left
8851 out of the link if it is not found by those means.
8853 Another way to specify a prefix much like the @option{-B} prefix is to use
8854 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
8857 As a special kludge, if the path provided by @option{-B} is
8858 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
8859 9, then it will be replaced by @file{[dir/]include}. This is to help
8860 with boot-strapping the compiler.
8862 @item -specs=@var{file}
8864 Process @var{file} after the compiler reads in the standard @file{specs}
8865 file, in order to override the defaults that the @file{gcc} driver
8866 program uses when determining what switches to pass to @file{cc1},
8867 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
8868 @option{-specs=@var{file}} can be specified on the command line, and they
8869 are processed in order, from left to right.
8871 @item --sysroot=@var{dir}
8873 Use @var{dir} as the logical root directory for headers and libraries.
8874 For example, if the compiler would normally search for headers in
8875 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
8876 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
8878 If you use both this option and the @option{-isysroot} option, then
8879 the @option{--sysroot} option will apply to libraries, but the
8880 @option{-isysroot} option will apply to header files.
8882 The GNU linker (beginning with version 2.16) has the necessary support
8883 for this option. If your linker does not support this option, the
8884 header file aspect of @option{--sysroot} will still work, but the
8885 library aspect will not.
8889 This option has been deprecated. Please use @option{-iquote} instead for
8890 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
8891 Any directories you specify with @option{-I} options before the @option{-I-}
8892 option are searched only for the case of @samp{#include "@var{file}"};
8893 they are not searched for @samp{#include <@var{file}>}.
8895 If additional directories are specified with @option{-I} options after
8896 the @option{-I-}, these directories are searched for all @samp{#include}
8897 directives. (Ordinarily @emph{all} @option{-I} directories are used
8900 In addition, the @option{-I-} option inhibits the use of the current
8901 directory (where the current input file came from) as the first search
8902 directory for @samp{#include "@var{file}"}. There is no way to
8903 override this effect of @option{-I-}. With @option{-I.} you can specify
8904 searching the directory which was current when the compiler was
8905 invoked. That is not exactly the same as what the preprocessor does
8906 by default, but it is often satisfactory.
8908 @option{-I-} does not inhibit the use of the standard system directories
8909 for header files. Thus, @option{-I-} and @option{-nostdinc} are
8916 @section Specifying subprocesses and the switches to pass to them
8919 @command{gcc} is a driver program. It performs its job by invoking a
8920 sequence of other programs to do the work of compiling, assembling and
8921 linking. GCC interprets its command-line parameters and uses these to
8922 deduce which programs it should invoke, and which command-line options
8923 it ought to place on their command lines. This behavior is controlled
8924 by @dfn{spec strings}. In most cases there is one spec string for each
8925 program that GCC can invoke, but a few programs have multiple spec
8926 strings to control their behavior. The spec strings built into GCC can
8927 be overridden by using the @option{-specs=} command-line switch to specify
8930 @dfn{Spec files} are plaintext files that are used to construct spec
8931 strings. They consist of a sequence of directives separated by blank
8932 lines. The type of directive is determined by the first non-whitespace
8933 character on the line and it can be one of the following:
8936 @item %@var{command}
8937 Issues a @var{command} to the spec file processor. The commands that can
8941 @item %include <@var{file}>
8943 Search for @var{file} and insert its text at the current point in the
8946 @item %include_noerr <@var{file}>
8947 @cindex %include_noerr
8948 Just like @samp{%include}, but do not generate an error message if the include
8949 file cannot be found.
8951 @item %rename @var{old_name} @var{new_name}
8953 Rename the spec string @var{old_name} to @var{new_name}.
8957 @item *[@var{spec_name}]:
8958 This tells the compiler to create, override or delete the named spec
8959 string. All lines after this directive up to the next directive or
8960 blank line are considered to be the text for the spec string. If this
8961 results in an empty string then the spec will be deleted. (Or, if the
8962 spec did not exist, then nothing will happened.) Otherwise, if the spec
8963 does not currently exist a new spec will be created. If the spec does
8964 exist then its contents will be overridden by the text of this
8965 directive, unless the first character of that text is the @samp{+}
8966 character, in which case the text will be appended to the spec.
8968 @item [@var{suffix}]:
8969 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
8970 and up to the next directive or blank line are considered to make up the
8971 spec string for the indicated suffix. When the compiler encounters an
8972 input file with the named suffix, it will processes the spec string in
8973 order to work out how to compile that file. For example:
8980 This says that any input file whose name ends in @samp{.ZZ} should be
8981 passed to the program @samp{z-compile}, which should be invoked with the
8982 command-line switch @option{-input} and with the result of performing the
8983 @samp{%i} substitution. (See below.)
8985 As an alternative to providing a spec string, the text that follows a
8986 suffix directive can be one of the following:
8989 @item @@@var{language}
8990 This says that the suffix is an alias for a known @var{language}. This is
8991 similar to using the @option{-x} command-line switch to GCC to specify a
8992 language explicitly. For example:
8999 Says that .ZZ files are, in fact, C++ source files.
9002 This causes an error messages saying:
9005 @var{name} compiler not installed on this system.
9009 GCC already has an extensive list of suffixes built into it.
9010 This directive will add an entry to the end of the list of suffixes, but
9011 since the list is searched from the end backwards, it is effectively
9012 possible to override earlier entries using this technique.
9016 GCC has the following spec strings built into it. Spec files can
9017 override these strings or create their own. Note that individual
9018 targets can also add their own spec strings to this list.
9021 asm Options to pass to the assembler
9022 asm_final Options to pass to the assembler post-processor
9023 cpp Options to pass to the C preprocessor
9024 cc1 Options to pass to the C compiler
9025 cc1plus Options to pass to the C++ compiler
9026 endfile Object files to include at the end of the link
9027 link Options to pass to the linker
9028 lib Libraries to include on the command line to the linker
9029 libgcc Decides which GCC support library to pass to the linker
9030 linker Sets the name of the linker
9031 predefines Defines to be passed to the C preprocessor
9032 signed_char Defines to pass to CPP to say whether @code{char} is signed
9034 startfile Object files to include at the start of the link
9037 Here is a small example of a spec file:
9043 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
9046 This example renames the spec called @samp{lib} to @samp{old_lib} and
9047 then overrides the previous definition of @samp{lib} with a new one.
9048 The new definition adds in some extra command-line options before
9049 including the text of the old definition.
9051 @dfn{Spec strings} are a list of command-line options to be passed to their
9052 corresponding program. In addition, the spec strings can contain
9053 @samp{%}-prefixed sequences to substitute variable text or to
9054 conditionally insert text into the command line. Using these constructs
9055 it is possible to generate quite complex command lines.
9057 Here is a table of all defined @samp{%}-sequences for spec
9058 strings. Note that spaces are not generated automatically around the
9059 results of expanding these sequences. Therefore you can concatenate them
9060 together or combine them with constant text in a single argument.
9064 Substitute one @samp{%} into the program name or argument.
9067 Substitute the name of the input file being processed.
9070 Substitute the basename of the input file being processed.
9071 This is the substring up to (and not including) the last period
9072 and not including the directory.
9075 This is the same as @samp{%b}, but include the file suffix (text after
9079 Marks the argument containing or following the @samp{%d} as a
9080 temporary file name, so that that file will be deleted if GCC exits
9081 successfully. Unlike @samp{%g}, this contributes no text to the
9084 @item %g@var{suffix}
9085 Substitute a file name that has suffix @var{suffix} and is chosen
9086 once per compilation, and mark the argument in the same way as
9087 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
9088 name is now chosen in a way that is hard to predict even when previously
9089 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
9090 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
9091 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
9092 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
9093 was simply substituted with a file name chosen once per compilation,
9094 without regard to any appended suffix (which was therefore treated
9095 just like ordinary text), making such attacks more likely to succeed.
9097 @item %u@var{suffix}
9098 Like @samp{%g}, but generates a new temporary file name even if
9099 @samp{%u@var{suffix}} was already seen.
9101 @item %U@var{suffix}
9102 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
9103 new one if there is no such last file name. In the absence of any
9104 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
9105 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
9106 would involve the generation of two distinct file names, one
9107 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
9108 simply substituted with a file name chosen for the previous @samp{%u},
9109 without regard to any appended suffix.
9111 @item %j@var{suffix}
9112 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
9113 writable, and if save-temps is off; otherwise, substitute the name
9114 of a temporary file, just like @samp{%u}. This temporary file is not
9115 meant for communication between processes, but rather as a junk
9118 @item %|@var{suffix}
9119 @itemx %m@var{suffix}
9120 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
9121 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
9122 all. These are the two most common ways to instruct a program that it
9123 should read from standard input or write to standard output. If you
9124 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
9125 construct: see for example @file{f/lang-specs.h}.
9127 @item %.@var{SUFFIX}
9128 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
9129 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
9130 terminated by the next space or %.
9133 Marks the argument containing or following the @samp{%w} as the
9134 designated output file of this compilation. This puts the argument
9135 into the sequence of arguments that @samp{%o} will substitute later.
9138 Substitutes the names of all the output files, with spaces
9139 automatically placed around them. You should write spaces
9140 around the @samp{%o} as well or the results are undefined.
9141 @samp{%o} is for use in the specs for running the linker.
9142 Input files whose names have no recognized suffix are not compiled
9143 at all, but they are included among the output files, so they will
9147 Substitutes the suffix for object files. Note that this is
9148 handled specially when it immediately follows @samp{%g, %u, or %U},
9149 because of the need for those to form complete file names. The
9150 handling is such that @samp{%O} is treated exactly as if it had already
9151 been substituted, except that @samp{%g, %u, and %U} do not currently
9152 support additional @var{suffix} characters following @samp{%O} as they would
9153 following, for example, @samp{.o}.
9156 Substitutes the standard macro predefinitions for the
9157 current target machine. Use this when running @code{cpp}.
9160 Like @samp{%p}, but puts @samp{__} before and after the name of each
9161 predefined macro, except for macros that start with @samp{__} or with
9162 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
9166 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
9167 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
9168 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
9169 and @option{-imultilib} as necessary.
9172 Current argument is the name of a library or startup file of some sort.
9173 Search for that file in a standard list of directories and substitute
9174 the full name found. The current working directory is included in the
9175 list of directories scanned.
9178 Current argument is the name of a linker script. Search for that file
9179 in the current list of directories to scan for libraries. If the file
9180 is located insert a @option{--script} option into the command line
9181 followed by the full path name found. If the file is not found then
9182 generate an error message. Note: the current working directory is not
9186 Print @var{str} as an error message. @var{str} is terminated by a newline.
9187 Use this when inconsistent options are detected.
9190 Substitute the contents of spec string @var{name} at this point.
9193 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
9195 @item %x@{@var{option}@}
9196 Accumulate an option for @samp{%X}.
9199 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
9203 Output the accumulated assembler options specified by @option{-Wa}.
9206 Output the accumulated preprocessor options specified by @option{-Wp}.
9209 Process the @code{asm} spec. This is used to compute the
9210 switches to be passed to the assembler.
9213 Process the @code{asm_final} spec. This is a spec string for
9214 passing switches to an assembler post-processor, if such a program is
9218 Process the @code{link} spec. This is the spec for computing the
9219 command line passed to the linker. Typically it will make use of the
9220 @samp{%L %G %S %D and %E} sequences.
9223 Dump out a @option{-L} option for each directory that GCC believes might
9224 contain startup files. If the target supports multilibs then the
9225 current multilib directory will be prepended to each of these paths.
9228 Process the @code{lib} spec. This is a spec string for deciding which
9229 libraries should be included on the command line to the linker.
9232 Process the @code{libgcc} spec. This is a spec string for deciding
9233 which GCC support library should be included on the command line to the linker.
9236 Process the @code{startfile} spec. This is a spec for deciding which
9237 object files should be the first ones passed to the linker. Typically
9238 this might be a file named @file{crt0.o}.
9241 Process the @code{endfile} spec. This is a spec string that specifies
9242 the last object files that will be passed to the linker.
9245 Process the @code{cpp} spec. This is used to construct the arguments
9246 to be passed to the C preprocessor.
9249 Process the @code{cc1} spec. This is used to construct the options to be
9250 passed to the actual C compiler (@samp{cc1}).
9253 Process the @code{cc1plus} spec. This is used to construct the options to be
9254 passed to the actual C++ compiler (@samp{cc1plus}).
9257 Substitute the variable part of a matched option. See below.
9258 Note that each comma in the substituted string is replaced by
9262 Remove all occurrences of @code{-S} from the command line. Note---this
9263 command is position dependent. @samp{%} commands in the spec string
9264 before this one will see @code{-S}, @samp{%} commands in the spec string
9265 after this one will not.
9267 @item %:@var{function}(@var{args})
9268 Call the named function @var{function}, passing it @var{args}.
9269 @var{args} is first processed as a nested spec string, then split
9270 into an argument vector in the usual fashion. The function returns
9271 a string which is processed as if it had appeared literally as part
9272 of the current spec.
9274 The following built-in spec functions are provided:
9278 The @code{getenv} spec function takes two arguments: an environment
9279 variable name and a string. If the environment variable is not
9280 defined, a fatal error is issued. Otherwise, the return value is the
9281 value of the environment variable concatenated with the string. For
9282 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
9285 %:getenv(TOPDIR /include)
9288 expands to @file{/path/to/top/include}.
9290 @item @code{if-exists}
9291 The @code{if-exists} spec function takes one argument, an absolute
9292 pathname to a file. If the file exists, @code{if-exists} returns the
9293 pathname. Here is a small example of its usage:
9297 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
9300 @item @code{if-exists-else}
9301 The @code{if-exists-else} spec function is similar to the @code{if-exists}
9302 spec function, except that it takes two arguments. The first argument is
9303 an absolute pathname to a file. If the file exists, @code{if-exists-else}
9304 returns the pathname. If it does not exist, it returns the second argument.
9305 This way, @code{if-exists-else} can be used to select one file or another,
9306 based on the existence of the first. Here is a small example of its usage:
9310 crt0%O%s %:if-exists(crti%O%s) \
9311 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
9314 @item @code{replace-outfile}
9315 The @code{replace-outfile} spec function takes two arguments. It looks for the
9316 first argument in the outfiles array and replaces it with the second argument. Here
9317 is a small example of its usage:
9320 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
9323 @item @code{print-asm-header}
9324 The @code{print-asm-header} function takes no arguments and simply
9325 prints a banner like:
9331 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
9334 It is used to separate compiler options from assembler options
9335 in the @option{--target-help} output.
9339 Substitutes the @code{-S} switch, if that switch was given to GCC@.
9340 If that switch was not specified, this substitutes nothing. Note that
9341 the leading dash is omitted when specifying this option, and it is
9342 automatically inserted if the substitution is performed. Thus the spec
9343 string @samp{%@{foo@}} would match the command-line option @option{-foo}
9344 and would output the command line option @option{-foo}.
9346 @item %W@{@code{S}@}
9347 Like %@{@code{S}@} but mark last argument supplied within as a file to be
9350 @item %@{@code{S}*@}
9351 Substitutes all the switches specified to GCC whose names start
9352 with @code{-S}, but which also take an argument. This is used for
9353 switches like @option{-o}, @option{-D}, @option{-I}, etc.
9354 GCC considers @option{-o foo} as being
9355 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
9356 text, including the space. Thus two arguments would be generated.
9358 @item %@{@code{S}*&@code{T}*@}
9359 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
9360 (the order of @code{S} and @code{T} in the spec is not significant).
9361 There can be any number of ampersand-separated variables; for each the
9362 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
9364 @item %@{@code{S}:@code{X}@}
9365 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
9367 @item %@{!@code{S}:@code{X}@}
9368 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
9370 @item %@{@code{S}*:@code{X}@}
9371 Substitutes @code{X} if one or more switches whose names start with
9372 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
9373 once, no matter how many such switches appeared. However, if @code{%*}
9374 appears somewhere in @code{X}, then @code{X} will be substituted once
9375 for each matching switch, with the @code{%*} replaced by the part of
9376 that switch that matched the @code{*}.
9378 @item %@{.@code{S}:@code{X}@}
9379 Substitutes @code{X}, if processing a file with suffix @code{S}.
9381 @item %@{!.@code{S}:@code{X}@}
9382 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
9384 @item %@{,@code{S}:@code{X}@}
9385 Substitutes @code{X}, if processing a file for language @code{S}.
9387 @item %@{!,@code{S}:@code{X}@}
9388 Substitutes @code{X}, if not processing a file for language @code{S}.
9390 @item %@{@code{S}|@code{P}:@code{X}@}
9391 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
9392 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
9393 @code{*} sequences as well, although they have a stronger binding than
9394 the @samp{|}. If @code{%*} appears in @code{X}, all of the
9395 alternatives must be starred, and only the first matching alternative
9398 For example, a spec string like this:
9401 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
9404 will output the following command-line options from the following input
9405 command-line options:
9410 -d fred.c -foo -baz -boggle
9411 -d jim.d -bar -baz -boggle
9414 @item %@{S:X; T:Y; :D@}
9416 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
9417 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
9418 be as many clauses as you need. This may be combined with @code{.},
9419 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
9424 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
9425 construct may contain other nested @samp{%} constructs or spaces, or
9426 even newlines. They are processed as usual, as described above.
9427 Trailing white space in @code{X} is ignored. White space may also
9428 appear anywhere on the left side of the colon in these constructs,
9429 except between @code{.} or @code{*} and the corresponding word.
9431 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
9432 handled specifically in these constructs. If another value of
9433 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
9434 @option{-W} switch is found later in the command line, the earlier
9435 switch value is ignored, except with @{@code{S}*@} where @code{S} is
9436 just one letter, which passes all matching options.
9438 The character @samp{|} at the beginning of the predicate text is used to
9439 indicate that a command should be piped to the following command, but
9440 only if @option{-pipe} is specified.
9442 It is built into GCC which switches take arguments and which do not.
9443 (You might think it would be useful to generalize this to allow each
9444 compiler's spec to say which switches take arguments. But this cannot
9445 be done in a consistent fashion. GCC cannot even decide which input
9446 files have been specified without knowing which switches take arguments,
9447 and it must know which input files to compile in order to tell which
9450 GCC also knows implicitly that arguments starting in @option{-l} are to be
9451 treated as compiler output files, and passed to the linker in their
9452 proper position among the other output files.
9454 @c man begin OPTIONS
9456 @node Target Options
9457 @section Specifying Target Machine and Compiler Version
9458 @cindex target options
9459 @cindex cross compiling
9460 @cindex specifying machine version
9461 @cindex specifying compiler version and target machine
9462 @cindex compiler version, specifying
9463 @cindex target machine, specifying
9465 The usual way to run GCC is to run the executable called @file{gcc}, or
9466 @file{<machine>-gcc} when cross-compiling, or
9467 @file{<machine>-gcc-<version>} to run a version other than the one that
9468 was installed last. Sometimes this is inconvenient, so GCC provides
9469 options that will switch to another cross-compiler or version.
9472 @item -b @var{machine}
9474 The argument @var{machine} specifies the target machine for compilation.
9476 The value to use for @var{machine} is the same as was specified as the
9477 machine type when configuring GCC as a cross-compiler. For
9478 example, if a cross-compiler was configured with @samp{configure
9479 arm-elf}, meaning to compile for an arm processor with elf binaries,
9480 then you would specify @option{-b arm-elf} to run that cross compiler.
9481 Because there are other options beginning with @option{-b}, the
9482 configuration must contain a hyphen, or @option{-b} alone should be one
9483 argument followed by the configuration in the next argument.
9485 @item -V @var{version}
9487 The argument @var{version} specifies which version of GCC to run.
9488 This is useful when multiple versions are installed. For example,
9489 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
9492 The @option{-V} and @option{-b} options work by running the
9493 @file{<machine>-gcc-<version>} executable, so there's no real reason to
9494 use them if you can just run that directly.
9496 @node Submodel Options
9497 @section Hardware Models and Configurations
9498 @cindex submodel options
9499 @cindex specifying hardware config
9500 @cindex hardware models and configurations, specifying
9501 @cindex machine dependent options
9503 Earlier we discussed the standard option @option{-b} which chooses among
9504 different installed compilers for completely different target
9505 machines, such as VAX vs.@: 68000 vs.@: 80386.
9507 In addition, each of these target machine types can have its own
9508 special options, starting with @samp{-m}, to choose among various
9509 hardware models or configurations---for example, 68010 vs 68020,
9510 floating coprocessor or none. A single installed version of the
9511 compiler can compile for any model or configuration, according to the
9514 Some configurations of the compiler also support additional special
9515 options, usually for compatibility with other compilers on the same
9518 @c This list is ordered alphanumerically by subsection name.
9519 @c It should be the same order and spelling as these options are listed
9520 @c in Machine Dependent Options
9526 * Blackfin Options::
9530 * DEC Alpha Options::
9531 * DEC Alpha/VMS Options::
9534 * GNU/Linux Options::
9537 * i386 and x86-64 Options::
9538 * i386 and x86-64 Windows Options::
9540 * IA-64/VMS Options::
9552 * picoChip Options::
9554 * RS/6000 and PowerPC Options::
9556 * S/390 and zSeries Options::
9561 * System V Options::
9566 * Xstormy16 Options::
9572 @subsection ARC Options
9575 These options are defined for ARC implementations:
9580 Compile code for little endian mode. This is the default.
9584 Compile code for big endian mode.
9587 @opindex mmangle-cpu
9588 Prepend the name of the cpu to all public symbol names.
9589 In multiple-processor systems, there are many ARC variants with different
9590 instruction and register set characteristics. This flag prevents code
9591 compiled for one cpu to be linked with code compiled for another.
9592 No facility exists for handling variants that are ``almost identical''.
9593 This is an all or nothing option.
9595 @item -mcpu=@var{cpu}
9597 Compile code for ARC variant @var{cpu}.
9598 Which variants are supported depend on the configuration.
9599 All variants support @option{-mcpu=base}, this is the default.
9601 @item -mtext=@var{text-section}
9602 @itemx -mdata=@var{data-section}
9603 @itemx -mrodata=@var{readonly-data-section}
9607 Put functions, data, and readonly data in @var{text-section},
9608 @var{data-section}, and @var{readonly-data-section} respectively
9609 by default. This can be overridden with the @code{section} attribute.
9610 @xref{Variable Attributes}.
9612 @item -mfix-cortex-m3-ldrd
9613 @opindex mfix-cortex-m3-ldrd
9614 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
9615 with overlapping destination and base registers are used. This option avoids
9616 generating these instructions. This option is enabled by default when
9617 @option{-mcpu=cortex-m3} is specified.
9622 @subsection ARM Options
9625 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
9629 @item -mabi=@var{name}
9631 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
9632 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
9635 @opindex mapcs-frame
9636 Generate a stack frame that is compliant with the ARM Procedure Call
9637 Standard for all functions, even if this is not strictly necessary for
9638 correct execution of the code. Specifying @option{-fomit-frame-pointer}
9639 with this option will cause the stack frames not to be generated for
9640 leaf functions. The default is @option{-mno-apcs-frame}.
9644 This is a synonym for @option{-mapcs-frame}.
9647 @c not currently implemented
9648 @item -mapcs-stack-check
9649 @opindex mapcs-stack-check
9650 Generate code to check the amount of stack space available upon entry to
9651 every function (that actually uses some stack space). If there is
9652 insufficient space available then either the function
9653 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
9654 called, depending upon the amount of stack space required. The run time
9655 system is required to provide these functions. The default is
9656 @option{-mno-apcs-stack-check}, since this produces smaller code.
9658 @c not currently implemented
9660 @opindex mapcs-float
9661 Pass floating point arguments using the float point registers. This is
9662 one of the variants of the APCS@. This option is recommended if the
9663 target hardware has a floating point unit or if a lot of floating point
9664 arithmetic is going to be performed by the code. The default is
9665 @option{-mno-apcs-float}, since integer only code is slightly increased in
9666 size if @option{-mapcs-float} is used.
9668 @c not currently implemented
9669 @item -mapcs-reentrant
9670 @opindex mapcs-reentrant
9671 Generate reentrant, position independent code. The default is
9672 @option{-mno-apcs-reentrant}.
9675 @item -mthumb-interwork
9676 @opindex mthumb-interwork
9677 Generate code which supports calling between the ARM and Thumb
9678 instruction sets. Without this option the two instruction sets cannot
9679 be reliably used inside one program. The default is
9680 @option{-mno-thumb-interwork}, since slightly larger code is generated
9681 when @option{-mthumb-interwork} is specified.
9683 @item -mno-sched-prolog
9684 @opindex mno-sched-prolog
9685 Prevent the reordering of instructions in the function prolog, or the
9686 merging of those instruction with the instructions in the function's
9687 body. This means that all functions will start with a recognizable set
9688 of instructions (or in fact one of a choice from a small set of
9689 different function prologues), and this information can be used to
9690 locate the start if functions inside an executable piece of code. The
9691 default is @option{-msched-prolog}.
9693 @item -mfloat-abi=@var{name}
9695 Specifies which floating-point ABI to use. Permissible values
9696 are: @samp{soft}, @samp{softfp} and @samp{hard}.
9698 Specifying @samp{soft} causes GCC to generate output containing
9699 library calls for floating-point operations.
9700 @samp{softfp} allows the generation of code using hardware floating-point
9701 instructions, but still uses the soft-float calling conventions.
9702 @samp{hard} allows generation of floating-point instructions
9703 and uses FPU-specific calling conventions.
9705 The default depends on the specific target configuration. Note that
9706 the hard-float and soft-float ABIs are not link-compatible; you must
9707 compile your entire program with the same ABI, and link with a
9708 compatible set of libraries.
9711 @opindex mhard-float
9712 Equivalent to @option{-mfloat-abi=hard}.
9715 @opindex msoft-float
9716 Equivalent to @option{-mfloat-abi=soft}.
9718 @item -mlittle-endian
9719 @opindex mlittle-endian
9720 Generate code for a processor running in little-endian mode. This is
9721 the default for all standard configurations.
9724 @opindex mbig-endian
9725 Generate code for a processor running in big-endian mode; the default is
9726 to compile code for a little-endian processor.
9728 @item -mwords-little-endian
9729 @opindex mwords-little-endian
9730 This option only applies when generating code for big-endian processors.
9731 Generate code for a little-endian word order but a big-endian byte
9732 order. That is, a byte order of the form @samp{32107654}. Note: this
9733 option should only be used if you require compatibility with code for
9734 big-endian ARM processors generated by versions of the compiler prior to
9737 @item -mcpu=@var{name}
9739 This specifies the name of the target ARM processor. GCC uses this name
9740 to determine what kind of instructions it can emit when generating
9741 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
9742 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
9743 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
9744 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
9745 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
9747 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
9748 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
9749 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
9750 @samp{strongarm1110},
9751 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
9752 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
9753 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
9754 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
9755 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
9756 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
9757 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
9758 @samp{cortex-a5}, @samp{cortex-a8}, @samp{cortex-a9},
9759 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-m3},
9762 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9764 @item -mtune=@var{name}
9766 This option is very similar to the @option{-mcpu=} option, except that
9767 instead of specifying the actual target processor type, and hence
9768 restricting which instructions can be used, it specifies that GCC should
9769 tune the performance of the code as if the target were of the type
9770 specified in this option, but still choosing the instructions that it
9771 will generate based on the cpu specified by a @option{-mcpu=} option.
9772 For some ARM implementations better performance can be obtained by using
9775 @item -march=@var{name}
9777 This specifies the name of the target ARM architecture. GCC uses this
9778 name to determine what kind of instructions it can emit when generating
9779 assembly code. This option can be used in conjunction with or instead
9780 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
9781 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
9782 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
9783 @samp{armv6}, @samp{armv6j},
9784 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
9785 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
9786 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9788 @item -mfpu=@var{name}
9789 @itemx -mfpe=@var{number}
9790 @itemx -mfp=@var{number}
9794 This specifies what floating point hardware (or hardware emulation) is
9795 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
9796 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-fp16},
9797 @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd}, @samp{vfpv3xd-fp16},
9798 @samp{neon}, @samp{neon-fp16}, @samp{vfpv4}, @samp{vfpv4-d16},
9799 @samp{fpv4-sp-d16} and @samp{neon-vfpv4}.
9800 @option{-mfp} and @option{-mfpe} are synonyms for
9801 @option{-mfpu}=@samp{fpe}@var{number}, for compatibility with older versions
9804 If @option{-msoft-float} is specified this specifies the format of
9805 floating point values.
9807 @item -mfp16-format=@var{name}
9808 @opindex mfp16-format
9809 Specify the format of the @code{__fp16} half-precision floating-point type.
9810 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
9811 the default is @samp{none}, in which case the @code{__fp16} type is not
9812 defined. @xref{Half-Precision}, for more information.
9814 @item -mstructure-size-boundary=@var{n}
9815 @opindex mstructure-size-boundary
9816 The size of all structures and unions will be rounded up to a multiple
9817 of the number of bits set by this option. Permissible values are 8, 32
9818 and 64. The default value varies for different toolchains. For the COFF
9819 targeted toolchain the default value is 8. A value of 64 is only allowed
9820 if the underlying ABI supports it.
9822 Specifying the larger number can produce faster, more efficient code, but
9823 can also increase the size of the program. Different values are potentially
9824 incompatible. Code compiled with one value cannot necessarily expect to
9825 work with code or libraries compiled with another value, if they exchange
9826 information using structures or unions.
9828 @item -mabort-on-noreturn
9829 @opindex mabort-on-noreturn
9830 Generate a call to the function @code{abort} at the end of a
9831 @code{noreturn} function. It will be executed if the function tries to
9835 @itemx -mno-long-calls
9836 @opindex mlong-calls
9837 @opindex mno-long-calls
9838 Tells the compiler to perform function calls by first loading the
9839 address of the function into a register and then performing a subroutine
9840 call on this register. This switch is needed if the target function
9841 will lie outside of the 64 megabyte addressing range of the offset based
9842 version of subroutine call instruction.
9844 Even if this switch is enabled, not all function calls will be turned
9845 into long calls. The heuristic is that static functions, functions
9846 which have the @samp{short-call} attribute, functions that are inside
9847 the scope of a @samp{#pragma no_long_calls} directive and functions whose
9848 definitions have already been compiled within the current compilation
9849 unit, will not be turned into long calls. The exception to this rule is
9850 that weak function definitions, functions with the @samp{long-call}
9851 attribute or the @samp{section} attribute, and functions that are within
9852 the scope of a @samp{#pragma long_calls} directive, will always be
9853 turned into long calls.
9855 This feature is not enabled by default. Specifying
9856 @option{-mno-long-calls} will restore the default behavior, as will
9857 placing the function calls within the scope of a @samp{#pragma
9858 long_calls_off} directive. Note these switches have no effect on how
9859 the compiler generates code to handle function calls via function
9862 @item -msingle-pic-base
9863 @opindex msingle-pic-base
9864 Treat the register used for PIC addressing as read-only, rather than
9865 loading it in the prologue for each function. The run-time system is
9866 responsible for initializing this register with an appropriate value
9867 before execution begins.
9869 @item -mpic-register=@var{reg}
9870 @opindex mpic-register
9871 Specify the register to be used for PIC addressing. The default is R10
9872 unless stack-checking is enabled, when R9 is used.
9874 @item -mcirrus-fix-invalid-insns
9875 @opindex mcirrus-fix-invalid-insns
9876 @opindex mno-cirrus-fix-invalid-insns
9877 Insert NOPs into the instruction stream to in order to work around
9878 problems with invalid Maverick instruction combinations. This option
9879 is only valid if the @option{-mcpu=ep9312} option has been used to
9880 enable generation of instructions for the Cirrus Maverick floating
9881 point co-processor. This option is not enabled by default, since the
9882 problem is only present in older Maverick implementations. The default
9883 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
9886 @item -mpoke-function-name
9887 @opindex mpoke-function-name
9888 Write the name of each function into the text section, directly
9889 preceding the function prologue. The generated code is similar to this:
9893 .ascii "arm_poke_function_name", 0
9896 .word 0xff000000 + (t1 - t0)
9897 arm_poke_function_name
9899 stmfd sp!, @{fp, ip, lr, pc@}
9903 When performing a stack backtrace, code can inspect the value of
9904 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
9905 location @code{pc - 12} and the top 8 bits are set, then we know that
9906 there is a function name embedded immediately preceding this location
9907 and has length @code{((pc[-3]) & 0xff000000)}.
9911 Generate code for the Thumb instruction set. The default is to
9912 use the 32-bit ARM instruction set.
9913 This option automatically enables either 16-bit Thumb-1 or
9914 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
9915 and @option{-march=@var{name}} options. This option is not passed to the
9916 assembler. If you want to force assembler files to be interpreted as Thumb code,
9917 either add a @samp{.thumb} directive to the source or pass the @option{-mthumb}
9918 option directly to the assembler by prefixing it with @option{-Wa}.
9921 @opindex mtpcs-frame
9922 Generate a stack frame that is compliant with the Thumb Procedure Call
9923 Standard for all non-leaf functions. (A leaf function is one that does
9924 not call any other functions.) The default is @option{-mno-tpcs-frame}.
9926 @item -mtpcs-leaf-frame
9927 @opindex mtpcs-leaf-frame
9928 Generate a stack frame that is compliant with the Thumb Procedure Call
9929 Standard for all leaf functions. (A leaf function is one that does
9930 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
9932 @item -mcallee-super-interworking
9933 @opindex mcallee-super-interworking
9934 Gives all externally visible functions in the file being compiled an ARM
9935 instruction set header which switches to Thumb mode before executing the
9936 rest of the function. This allows these functions to be called from
9937 non-interworking code. This option is not valid in AAPCS configurations
9938 because interworking is enabled by default.
9940 @item -mcaller-super-interworking
9941 @opindex mcaller-super-interworking
9942 Allows calls via function pointers (including virtual functions) to
9943 execute correctly regardless of whether the target code has been
9944 compiled for interworking or not. There is a small overhead in the cost
9945 of executing a function pointer if this option is enabled. This option
9946 is not valid in AAPCS configurations because interworking is enabled
9949 @item -mtp=@var{name}
9951 Specify the access model for the thread local storage pointer. The valid
9952 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
9953 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
9954 (supported in the arm6k architecture), and @option{auto}, which uses the
9955 best available method for the selected processor. The default setting is
9958 @item -mword-relocations
9959 @opindex mword-relocations
9960 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
9961 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
9962 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
9968 @subsection AVR Options
9971 These options are defined for AVR implementations:
9974 @item -mmcu=@var{mcu}
9976 Specify ATMEL AVR instruction set or MCU type.
9978 Instruction set avr1 is for the minimal AVR core, not supported by the C
9979 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
9980 attiny11, attiny12, attiny15, attiny28).
9982 Instruction set avr2 (default) is for the classic AVR core with up to
9983 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
9984 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
9985 at90c8534, at90s8535).
9987 Instruction set avr3 is for the classic AVR core with up to 128K program
9988 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
9990 Instruction set avr4 is for the enhanced AVR core with up to 8K program
9991 memory space (MCU types: atmega8, atmega83, atmega85).
9993 Instruction set avr5 is for the enhanced AVR core with up to 128K program
9994 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
9995 atmega64, atmega128, at43usb355, at94k).
9997 @item -mno-interrupts
9998 @opindex mno-interrupts
9999 Generated code is not compatible with hardware interrupts.
10000 Code size will be smaller.
10002 @item -mcall-prologues
10003 @opindex mcall-prologues
10004 Functions prologues/epilogues expanded as call to appropriate
10005 subroutines. Code size will be smaller.
10008 @opindex mtiny-stack
10009 Change only the low 8 bits of the stack pointer.
10013 Assume int to be 8 bit integer. This affects the sizes of all types: A
10014 char will be 1 byte, an int will be 1 byte, a long will be 2 bytes
10015 and long long will be 4 bytes. Please note that this option does not
10016 comply to the C standards, but it will provide you with smaller code
10020 @node Blackfin Options
10021 @subsection Blackfin Options
10022 @cindex Blackfin Options
10025 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
10027 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
10028 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
10029 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
10030 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
10031 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
10032 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
10033 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
10035 The optional @var{sirevision} specifies the silicon revision of the target
10036 Blackfin processor. Any workarounds available for the targeted silicon revision
10037 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
10038 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
10039 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
10040 hexadecimal digits representing the major and minor numbers in the silicon
10041 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
10042 is not defined. If @var{sirevision} is @samp{any}, the
10043 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
10044 If this optional @var{sirevision} is not used, GCC assumes the latest known
10045 silicon revision of the targeted Blackfin processor.
10047 Support for @samp{bf561} is incomplete. For @samp{bf561},
10048 Only the processor macro is defined.
10049 Without this option, @samp{bf532} is used as the processor by default.
10050 The corresponding predefined processor macros for @var{cpu} is to
10051 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
10052 provided by libgloss to be linked in if @option{-msim} is not given.
10056 Specifies that the program will be run on the simulator. This causes
10057 the simulator BSP provided by libgloss to be linked in. This option
10058 has effect only for @samp{bfin-elf} toolchain.
10059 Certain other options, such as @option{-mid-shared-library} and
10060 @option{-mfdpic}, imply @option{-msim}.
10062 @item -momit-leaf-frame-pointer
10063 @opindex momit-leaf-frame-pointer
10064 Don't keep the frame pointer in a register for leaf functions. This
10065 avoids the instructions to save, set up and restore frame pointers and
10066 makes an extra register available in leaf functions. The option
10067 @option{-fomit-frame-pointer} removes the frame pointer for all functions
10068 which might make debugging harder.
10070 @item -mspecld-anomaly
10071 @opindex mspecld-anomaly
10072 When enabled, the compiler will ensure that the generated code does not
10073 contain speculative loads after jump instructions. If this option is used,
10074 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
10076 @item -mno-specld-anomaly
10077 @opindex mno-specld-anomaly
10078 Don't generate extra code to prevent speculative loads from occurring.
10080 @item -mcsync-anomaly
10081 @opindex mcsync-anomaly
10082 When enabled, the compiler will ensure that the generated code does not
10083 contain CSYNC or SSYNC instructions too soon after conditional branches.
10084 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
10086 @item -mno-csync-anomaly
10087 @opindex mno-csync-anomaly
10088 Don't generate extra code to prevent CSYNC or SSYNC instructions from
10089 occurring too soon after a conditional branch.
10093 When enabled, the compiler is free to take advantage of the knowledge that
10094 the entire program fits into the low 64k of memory.
10097 @opindex mno-low-64k
10098 Assume that the program is arbitrarily large. This is the default.
10100 @item -mstack-check-l1
10101 @opindex mstack-check-l1
10102 Do stack checking using information placed into L1 scratchpad memory by the
10105 @item -mid-shared-library
10106 @opindex mid-shared-library
10107 Generate code that supports shared libraries via the library ID method.
10108 This allows for execute in place and shared libraries in an environment
10109 without virtual memory management. This option implies @option{-fPIC}.
10110 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10112 @item -mno-id-shared-library
10113 @opindex mno-id-shared-library
10114 Generate code that doesn't assume ID based shared libraries are being used.
10115 This is the default.
10117 @item -mleaf-id-shared-library
10118 @opindex mleaf-id-shared-library
10119 Generate code that supports shared libraries via the library ID method,
10120 but assumes that this library or executable won't link against any other
10121 ID shared libraries. That allows the compiler to use faster code for jumps
10124 @item -mno-leaf-id-shared-library
10125 @opindex mno-leaf-id-shared-library
10126 Do not assume that the code being compiled won't link against any ID shared
10127 libraries. Slower code will be generated for jump and call insns.
10129 @item -mshared-library-id=n
10130 @opindex mshared-library-id
10131 Specified the identification number of the ID based shared library being
10132 compiled. Specifying a value of 0 will generate more compact code, specifying
10133 other values will force the allocation of that number to the current
10134 library but is no more space or time efficient than omitting this option.
10138 Generate code that allows the data segment to be located in a different
10139 area of memory from the text segment. This allows for execute in place in
10140 an environment without virtual memory management by eliminating relocations
10141 against the text section.
10143 @item -mno-sep-data
10144 @opindex mno-sep-data
10145 Generate code that assumes that the data segment follows the text segment.
10146 This is the default.
10149 @itemx -mno-long-calls
10150 @opindex mlong-calls
10151 @opindex mno-long-calls
10152 Tells the compiler to perform function calls by first loading the
10153 address of the function into a register and then performing a subroutine
10154 call on this register. This switch is needed if the target function
10155 will lie outside of the 24 bit addressing range of the offset based
10156 version of subroutine call instruction.
10158 This feature is not enabled by default. Specifying
10159 @option{-mno-long-calls} will restore the default behavior. Note these
10160 switches have no effect on how the compiler generates code to handle
10161 function calls via function pointers.
10165 Link with the fast floating-point library. This library relaxes some of
10166 the IEEE floating-point standard's rules for checking inputs against
10167 Not-a-Number (NAN), in the interest of performance.
10170 @opindex minline-plt
10171 Enable inlining of PLT entries in function calls to functions that are
10172 not known to bind locally. It has no effect without @option{-mfdpic}.
10175 @opindex mmulticore
10176 Build standalone application for multicore Blackfin processor. Proper
10177 start files and link scripts will be used to support multicore.
10178 This option defines @code{__BFIN_MULTICORE}. It can only be used with
10179 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
10180 @option{-mcorea} or @option{-mcoreb}. If it's used without
10181 @option{-mcorea} or @option{-mcoreb}, single application/dual core
10182 programming model is used. In this model, the main function of Core B
10183 should be named as coreb_main. If it's used with @option{-mcorea} or
10184 @option{-mcoreb}, one application per core programming model is used.
10185 If this option is not used, single core application programming
10190 Build standalone application for Core A of BF561 when using
10191 one application per core programming model. Proper start files
10192 and link scripts will be used to support Core A. This option
10193 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
10197 Build standalone application for Core B of BF561 when using
10198 one application per core programming model. Proper start files
10199 and link scripts will be used to support Core B. This option
10200 defines @code{__BFIN_COREB}. When this option is used, coreb_main
10201 should be used instead of main. It must be used with
10202 @option{-mmulticore}.
10206 Build standalone application for SDRAM. Proper start files and
10207 link scripts will be used to put the application into SDRAM.
10208 Loader should initialize SDRAM before loading the application
10209 into SDRAM. This option defines @code{__BFIN_SDRAM}.
10213 Assume that ICPLBs are enabled at runtime. This has an effect on certain
10214 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
10215 are enabled; for standalone applications the default is off.
10219 @subsection CRIS Options
10220 @cindex CRIS Options
10222 These options are defined specifically for the CRIS ports.
10225 @item -march=@var{architecture-type}
10226 @itemx -mcpu=@var{architecture-type}
10229 Generate code for the specified architecture. The choices for
10230 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
10231 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
10232 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
10235 @item -mtune=@var{architecture-type}
10237 Tune to @var{architecture-type} everything applicable about the generated
10238 code, except for the ABI and the set of available instructions. The
10239 choices for @var{architecture-type} are the same as for
10240 @option{-march=@var{architecture-type}}.
10242 @item -mmax-stack-frame=@var{n}
10243 @opindex mmax-stack-frame
10244 Warn when the stack frame of a function exceeds @var{n} bytes.
10250 The options @option{-metrax4} and @option{-metrax100} are synonyms for
10251 @option{-march=v3} and @option{-march=v8} respectively.
10253 @item -mmul-bug-workaround
10254 @itemx -mno-mul-bug-workaround
10255 @opindex mmul-bug-workaround
10256 @opindex mno-mul-bug-workaround
10257 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
10258 models where it applies. This option is active by default.
10262 Enable CRIS-specific verbose debug-related information in the assembly
10263 code. This option also has the effect to turn off the @samp{#NO_APP}
10264 formatted-code indicator to the assembler at the beginning of the
10269 Do not use condition-code results from previous instruction; always emit
10270 compare and test instructions before use of condition codes.
10272 @item -mno-side-effects
10273 @opindex mno-side-effects
10274 Do not emit instructions with side-effects in addressing modes other than
10277 @item -mstack-align
10278 @itemx -mno-stack-align
10279 @itemx -mdata-align
10280 @itemx -mno-data-align
10281 @itemx -mconst-align
10282 @itemx -mno-const-align
10283 @opindex mstack-align
10284 @opindex mno-stack-align
10285 @opindex mdata-align
10286 @opindex mno-data-align
10287 @opindex mconst-align
10288 @opindex mno-const-align
10289 These options (no-options) arranges (eliminate arrangements) for the
10290 stack-frame, individual data and constants to be aligned for the maximum
10291 single data access size for the chosen CPU model. The default is to
10292 arrange for 32-bit alignment. ABI details such as structure layout are
10293 not affected by these options.
10301 Similar to the stack- data- and const-align options above, these options
10302 arrange for stack-frame, writable data and constants to all be 32-bit,
10303 16-bit or 8-bit aligned. The default is 32-bit alignment.
10305 @item -mno-prologue-epilogue
10306 @itemx -mprologue-epilogue
10307 @opindex mno-prologue-epilogue
10308 @opindex mprologue-epilogue
10309 With @option{-mno-prologue-epilogue}, the normal function prologue and
10310 epilogue that sets up the stack-frame are omitted and no return
10311 instructions or return sequences are generated in the code. Use this
10312 option only together with visual inspection of the compiled code: no
10313 warnings or errors are generated when call-saved registers must be saved,
10314 or storage for local variable needs to be allocated.
10318 @opindex mno-gotplt
10320 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
10321 instruction sequences that load addresses for functions from the PLT part
10322 of the GOT rather than (traditional on other architectures) calls to the
10323 PLT@. The default is @option{-mgotplt}.
10327 Legacy no-op option only recognized with the cris-axis-elf and
10328 cris-axis-linux-gnu targets.
10332 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
10336 This option, recognized for the cris-axis-elf arranges
10337 to link with input-output functions from a simulator library. Code,
10338 initialized data and zero-initialized data are allocated consecutively.
10342 Like @option{-sim}, but pass linker options to locate initialized data at
10343 0x40000000 and zero-initialized data at 0x80000000.
10347 @subsection CRX Options
10348 @cindex CRX Options
10350 These options are defined specifically for the CRX ports.
10356 Enable the use of multiply-accumulate instructions. Disabled by default.
10359 @opindex mpush-args
10360 Push instructions will be used to pass outgoing arguments when functions
10361 are called. Enabled by default.
10364 @node Darwin Options
10365 @subsection Darwin Options
10366 @cindex Darwin options
10368 These options are defined for all architectures running the Darwin operating
10371 FSF GCC on Darwin does not create ``fat'' object files; it will create
10372 an object file for the single architecture that it was built to
10373 target. Apple's GCC on Darwin does create ``fat'' files if multiple
10374 @option{-arch} options are used; it does so by running the compiler or
10375 linker multiple times and joining the results together with
10378 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
10379 @samp{i686}) is determined by the flags that specify the ISA
10380 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
10381 @option{-force_cpusubtype_ALL} option can be used to override this.
10383 The Darwin tools vary in their behavior when presented with an ISA
10384 mismatch. The assembler, @file{as}, will only permit instructions to
10385 be used that are valid for the subtype of the file it is generating,
10386 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
10387 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
10388 and print an error if asked to create a shared library with a less
10389 restrictive subtype than its input files (for instance, trying to put
10390 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
10391 for executables, @file{ld}, will quietly give the executable the most
10392 restrictive subtype of any of its input files.
10397 Add the framework directory @var{dir} to the head of the list of
10398 directories to be searched for header files. These directories are
10399 interleaved with those specified by @option{-I} options and are
10400 scanned in a left-to-right order.
10402 A framework directory is a directory with frameworks in it. A
10403 framework is a directory with a @samp{"Headers"} and/or
10404 @samp{"PrivateHeaders"} directory contained directly in it that ends
10405 in @samp{".framework"}. The name of a framework is the name of this
10406 directory excluding the @samp{".framework"}. Headers associated with
10407 the framework are found in one of those two directories, with
10408 @samp{"Headers"} being searched first. A subframework is a framework
10409 directory that is in a framework's @samp{"Frameworks"} directory.
10410 Includes of subframework headers can only appear in a header of a
10411 framework that contains the subframework, or in a sibling subframework
10412 header. Two subframeworks are siblings if they occur in the same
10413 framework. A subframework should not have the same name as a
10414 framework, a warning will be issued if this is violated. Currently a
10415 subframework cannot have subframeworks, in the future, the mechanism
10416 may be extended to support this. The standard frameworks can be found
10417 in @samp{"/System/Library/Frameworks"} and
10418 @samp{"/Library/Frameworks"}. An example include looks like
10419 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
10420 the name of the framework and header.h is found in the
10421 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
10423 @item -iframework@var{dir}
10424 @opindex iframework
10425 Like @option{-F} except the directory is a treated as a system
10426 directory. The main difference between this @option{-iframework} and
10427 @option{-F} is that with @option{-iframework} the compiler does not
10428 warn about constructs contained within header files found via
10429 @var{dir}. This option is valid only for the C family of languages.
10433 Emit debugging information for symbols that are used. For STABS
10434 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
10435 This is by default ON@.
10439 Emit debugging information for all symbols and types.
10441 @item -mmacosx-version-min=@var{version}
10442 The earliest version of MacOS X that this executable will run on
10443 is @var{version}. Typical values of @var{version} include @code{10.1},
10444 @code{10.2}, and @code{10.3.9}.
10446 If the compiler was built to use the system's headers by default,
10447 then the default for this option is the system version on which the
10448 compiler is running, otherwise the default is to make choices which
10449 are compatible with as many systems and code bases as possible.
10453 Enable kernel development mode. The @option{-mkernel} option sets
10454 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
10455 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
10456 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
10457 applicable. This mode also sets @option{-mno-altivec},
10458 @option{-msoft-float}, @option{-fno-builtin} and
10459 @option{-mlong-branch} for PowerPC targets.
10461 @item -mone-byte-bool
10462 @opindex mone-byte-bool
10463 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
10464 By default @samp{sizeof(bool)} is @samp{4} when compiling for
10465 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
10466 option has no effect on x86.
10468 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
10469 to generate code that is not binary compatible with code generated
10470 without that switch. Using this switch may require recompiling all
10471 other modules in a program, including system libraries. Use this
10472 switch to conform to a non-default data model.
10474 @item -mfix-and-continue
10475 @itemx -ffix-and-continue
10476 @itemx -findirect-data
10477 @opindex mfix-and-continue
10478 @opindex ffix-and-continue
10479 @opindex findirect-data
10480 Generate code suitable for fast turn around development. Needed to
10481 enable gdb to dynamically load @code{.o} files into already running
10482 programs. @option{-findirect-data} and @option{-ffix-and-continue}
10483 are provided for backwards compatibility.
10487 Loads all members of static archive libraries.
10488 See man ld(1) for more information.
10490 @item -arch_errors_fatal
10491 @opindex arch_errors_fatal
10492 Cause the errors having to do with files that have the wrong architecture
10495 @item -bind_at_load
10496 @opindex bind_at_load
10497 Causes the output file to be marked such that the dynamic linker will
10498 bind all undefined references when the file is loaded or launched.
10502 Produce a Mach-o bundle format file.
10503 See man ld(1) for more information.
10505 @item -bundle_loader @var{executable}
10506 @opindex bundle_loader
10507 This option specifies the @var{executable} that will be loading the build
10508 output file being linked. See man ld(1) for more information.
10511 @opindex dynamiclib
10512 When passed this option, GCC will produce a dynamic library instead of
10513 an executable when linking, using the Darwin @file{libtool} command.
10515 @item -force_cpusubtype_ALL
10516 @opindex force_cpusubtype_ALL
10517 This causes GCC's output file to have the @var{ALL} subtype, instead of
10518 one controlled by the @option{-mcpu} or @option{-march} option.
10520 @item -allowable_client @var{client_name}
10521 @itemx -client_name
10522 @itemx -compatibility_version
10523 @itemx -current_version
10525 @itemx -dependency-file
10527 @itemx -dylinker_install_name
10529 @itemx -exported_symbols_list
10531 @itemx -flat_namespace
10532 @itemx -force_flat_namespace
10533 @itemx -headerpad_max_install_names
10536 @itemx -install_name
10537 @itemx -keep_private_externs
10538 @itemx -multi_module
10539 @itemx -multiply_defined
10540 @itemx -multiply_defined_unused
10542 @itemx -no_dead_strip_inits_and_terms
10543 @itemx -nofixprebinding
10544 @itemx -nomultidefs
10546 @itemx -noseglinkedit
10547 @itemx -pagezero_size
10549 @itemx -prebind_all_twolevel_modules
10550 @itemx -private_bundle
10551 @itemx -read_only_relocs
10553 @itemx -sectobjectsymbols
10557 @itemx -sectobjectsymbols
10560 @itemx -segs_read_only_addr
10561 @itemx -segs_read_write_addr
10562 @itemx -seg_addr_table
10563 @itemx -seg_addr_table_filename
10564 @itemx -seglinkedit
10566 @itemx -segs_read_only_addr
10567 @itemx -segs_read_write_addr
10568 @itemx -single_module
10570 @itemx -sub_library
10571 @itemx -sub_umbrella
10572 @itemx -twolevel_namespace
10575 @itemx -unexported_symbols_list
10576 @itemx -weak_reference_mismatches
10577 @itemx -whatsloaded
10578 @opindex allowable_client
10579 @opindex client_name
10580 @opindex compatibility_version
10581 @opindex current_version
10582 @opindex dead_strip
10583 @opindex dependency-file
10584 @opindex dylib_file
10585 @opindex dylinker_install_name
10587 @opindex exported_symbols_list
10589 @opindex flat_namespace
10590 @opindex force_flat_namespace
10591 @opindex headerpad_max_install_names
10592 @opindex image_base
10594 @opindex install_name
10595 @opindex keep_private_externs
10596 @opindex multi_module
10597 @opindex multiply_defined
10598 @opindex multiply_defined_unused
10599 @opindex noall_load
10600 @opindex no_dead_strip_inits_and_terms
10601 @opindex nofixprebinding
10602 @opindex nomultidefs
10604 @opindex noseglinkedit
10605 @opindex pagezero_size
10607 @opindex prebind_all_twolevel_modules
10608 @opindex private_bundle
10609 @opindex read_only_relocs
10611 @opindex sectobjectsymbols
10614 @opindex sectcreate
10615 @opindex sectobjectsymbols
10618 @opindex segs_read_only_addr
10619 @opindex segs_read_write_addr
10620 @opindex seg_addr_table
10621 @opindex seg_addr_table_filename
10622 @opindex seglinkedit
10624 @opindex segs_read_only_addr
10625 @opindex segs_read_write_addr
10626 @opindex single_module
10628 @opindex sub_library
10629 @opindex sub_umbrella
10630 @opindex twolevel_namespace
10633 @opindex unexported_symbols_list
10634 @opindex weak_reference_mismatches
10635 @opindex whatsloaded
10636 These options are passed to the Darwin linker. The Darwin linker man page
10637 describes them in detail.
10640 @node DEC Alpha Options
10641 @subsection DEC Alpha Options
10643 These @samp{-m} options are defined for the DEC Alpha implementations:
10646 @item -mno-soft-float
10647 @itemx -msoft-float
10648 @opindex mno-soft-float
10649 @opindex msoft-float
10650 Use (do not use) the hardware floating-point instructions for
10651 floating-point operations. When @option{-msoft-float} is specified,
10652 functions in @file{libgcc.a} will be used to perform floating-point
10653 operations. Unless they are replaced by routines that emulate the
10654 floating-point operations, or compiled in such a way as to call such
10655 emulations routines, these routines will issue floating-point
10656 operations. If you are compiling for an Alpha without floating-point
10657 operations, you must ensure that the library is built so as not to call
10660 Note that Alpha implementations without floating-point operations are
10661 required to have floating-point registers.
10664 @itemx -mno-fp-regs
10666 @opindex mno-fp-regs
10667 Generate code that uses (does not use) the floating-point register set.
10668 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
10669 register set is not used, floating point operands are passed in integer
10670 registers as if they were integers and floating-point results are passed
10671 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
10672 so any function with a floating-point argument or return value called by code
10673 compiled with @option{-mno-fp-regs} must also be compiled with that
10676 A typical use of this option is building a kernel that does not use,
10677 and hence need not save and restore, any floating-point registers.
10681 The Alpha architecture implements floating-point hardware optimized for
10682 maximum performance. It is mostly compliant with the IEEE floating
10683 point standard. However, for full compliance, software assistance is
10684 required. This option generates code fully IEEE compliant code
10685 @emph{except} that the @var{inexact-flag} is not maintained (see below).
10686 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
10687 defined during compilation. The resulting code is less efficient but is
10688 able to correctly support denormalized numbers and exceptional IEEE
10689 values such as not-a-number and plus/minus infinity. Other Alpha
10690 compilers call this option @option{-ieee_with_no_inexact}.
10692 @item -mieee-with-inexact
10693 @opindex mieee-with-inexact
10694 This is like @option{-mieee} except the generated code also maintains
10695 the IEEE @var{inexact-flag}. Turning on this option causes the
10696 generated code to implement fully-compliant IEEE math. In addition to
10697 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
10698 macro. On some Alpha implementations the resulting code may execute
10699 significantly slower than the code generated by default. Since there is
10700 very little code that depends on the @var{inexact-flag}, you should
10701 normally not specify this option. Other Alpha compilers call this
10702 option @option{-ieee_with_inexact}.
10704 @item -mfp-trap-mode=@var{trap-mode}
10705 @opindex mfp-trap-mode
10706 This option controls what floating-point related traps are enabled.
10707 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
10708 The trap mode can be set to one of four values:
10712 This is the default (normal) setting. The only traps that are enabled
10713 are the ones that cannot be disabled in software (e.g., division by zero
10717 In addition to the traps enabled by @samp{n}, underflow traps are enabled
10721 Like @samp{u}, but the instructions are marked to be safe for software
10722 completion (see Alpha architecture manual for details).
10725 Like @samp{su}, but inexact traps are enabled as well.
10728 @item -mfp-rounding-mode=@var{rounding-mode}
10729 @opindex mfp-rounding-mode
10730 Selects the IEEE rounding mode. Other Alpha compilers call this option
10731 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
10736 Normal IEEE rounding mode. Floating point numbers are rounded towards
10737 the nearest machine number or towards the even machine number in case
10741 Round towards minus infinity.
10744 Chopped rounding mode. Floating point numbers are rounded towards zero.
10747 Dynamic rounding mode. A field in the floating point control register
10748 (@var{fpcr}, see Alpha architecture reference manual) controls the
10749 rounding mode in effect. The C library initializes this register for
10750 rounding towards plus infinity. Thus, unless your program modifies the
10751 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
10754 @item -mtrap-precision=@var{trap-precision}
10755 @opindex mtrap-precision
10756 In the Alpha architecture, floating point traps are imprecise. This
10757 means without software assistance it is impossible to recover from a
10758 floating trap and program execution normally needs to be terminated.
10759 GCC can generate code that can assist operating system trap handlers
10760 in determining the exact location that caused a floating point trap.
10761 Depending on the requirements of an application, different levels of
10762 precisions can be selected:
10766 Program precision. This option is the default and means a trap handler
10767 can only identify which program caused a floating point exception.
10770 Function precision. The trap handler can determine the function that
10771 caused a floating point exception.
10774 Instruction precision. The trap handler can determine the exact
10775 instruction that caused a floating point exception.
10778 Other Alpha compilers provide the equivalent options called
10779 @option{-scope_safe} and @option{-resumption_safe}.
10781 @item -mieee-conformant
10782 @opindex mieee-conformant
10783 This option marks the generated code as IEEE conformant. You must not
10784 use this option unless you also specify @option{-mtrap-precision=i} and either
10785 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
10786 is to emit the line @samp{.eflag 48} in the function prologue of the
10787 generated assembly file. Under DEC Unix, this has the effect that
10788 IEEE-conformant math library routines will be linked in.
10790 @item -mbuild-constants
10791 @opindex mbuild-constants
10792 Normally GCC examines a 32- or 64-bit integer constant to
10793 see if it can construct it from smaller constants in two or three
10794 instructions. If it cannot, it will output the constant as a literal and
10795 generate code to load it from the data segment at runtime.
10797 Use this option to require GCC to construct @emph{all} integer constants
10798 using code, even if it takes more instructions (the maximum is six).
10800 You would typically use this option to build a shared library dynamic
10801 loader. Itself a shared library, it must relocate itself in memory
10802 before it can find the variables and constants in its own data segment.
10808 Select whether to generate code to be assembled by the vendor-supplied
10809 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
10827 Indicate whether GCC should generate code to use the optional BWX,
10828 CIX, FIX and MAX instruction sets. The default is to use the instruction
10829 sets supported by the CPU type specified via @option{-mcpu=} option or that
10830 of the CPU on which GCC was built if none was specified.
10833 @itemx -mfloat-ieee
10834 @opindex mfloat-vax
10835 @opindex mfloat-ieee
10836 Generate code that uses (does not use) VAX F and G floating point
10837 arithmetic instead of IEEE single and double precision.
10839 @item -mexplicit-relocs
10840 @itemx -mno-explicit-relocs
10841 @opindex mexplicit-relocs
10842 @opindex mno-explicit-relocs
10843 Older Alpha assemblers provided no way to generate symbol relocations
10844 except via assembler macros. Use of these macros does not allow
10845 optimal instruction scheduling. GNU binutils as of version 2.12
10846 supports a new syntax that allows the compiler to explicitly mark
10847 which relocations should apply to which instructions. This option
10848 is mostly useful for debugging, as GCC detects the capabilities of
10849 the assembler when it is built and sets the default accordingly.
10852 @itemx -mlarge-data
10853 @opindex msmall-data
10854 @opindex mlarge-data
10855 When @option{-mexplicit-relocs} is in effect, static data is
10856 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
10857 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
10858 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
10859 16-bit relocations off of the @code{$gp} register. This limits the
10860 size of the small data area to 64KB, but allows the variables to be
10861 directly accessed via a single instruction.
10863 The default is @option{-mlarge-data}. With this option the data area
10864 is limited to just below 2GB@. Programs that require more than 2GB of
10865 data must use @code{malloc} or @code{mmap} to allocate the data in the
10866 heap instead of in the program's data segment.
10868 When generating code for shared libraries, @option{-fpic} implies
10869 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
10872 @itemx -mlarge-text
10873 @opindex msmall-text
10874 @opindex mlarge-text
10875 When @option{-msmall-text} is used, the compiler assumes that the
10876 code of the entire program (or shared library) fits in 4MB, and is
10877 thus reachable with a branch instruction. When @option{-msmall-data}
10878 is used, the compiler can assume that all local symbols share the
10879 same @code{$gp} value, and thus reduce the number of instructions
10880 required for a function call from 4 to 1.
10882 The default is @option{-mlarge-text}.
10884 @item -mcpu=@var{cpu_type}
10886 Set the instruction set and instruction scheduling parameters for
10887 machine type @var{cpu_type}. You can specify either the @samp{EV}
10888 style name or the corresponding chip number. GCC supports scheduling
10889 parameters for the EV4, EV5 and EV6 family of processors and will
10890 choose the default values for the instruction set from the processor
10891 you specify. If you do not specify a processor type, GCC will default
10892 to the processor on which the compiler was built.
10894 Supported values for @var{cpu_type} are
10900 Schedules as an EV4 and has no instruction set extensions.
10904 Schedules as an EV5 and has no instruction set extensions.
10908 Schedules as an EV5 and supports the BWX extension.
10913 Schedules as an EV5 and supports the BWX and MAX extensions.
10917 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
10921 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
10924 Native Linux/GNU toolchains also support the value @samp{native},
10925 which selects the best architecture option for the host processor.
10926 @option{-mcpu=native} has no effect if GCC does not recognize
10929 @item -mtune=@var{cpu_type}
10931 Set only the instruction scheduling parameters for machine type
10932 @var{cpu_type}. The instruction set is not changed.
10934 Native Linux/GNU toolchains also support the value @samp{native},
10935 which selects the best architecture option for the host processor.
10936 @option{-mtune=native} has no effect if GCC does not recognize
10939 @item -mmemory-latency=@var{time}
10940 @opindex mmemory-latency
10941 Sets the latency the scheduler should assume for typical memory
10942 references as seen by the application. This number is highly
10943 dependent on the memory access patterns used by the application
10944 and the size of the external cache on the machine.
10946 Valid options for @var{time} are
10950 A decimal number representing clock cycles.
10956 The compiler contains estimates of the number of clock cycles for
10957 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
10958 (also called Dcache, Scache, and Bcache), as well as to main memory.
10959 Note that L3 is only valid for EV5.
10964 @node DEC Alpha/VMS Options
10965 @subsection DEC Alpha/VMS Options
10967 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
10970 @item -mvms-return-codes
10971 @opindex mvms-return-codes
10972 Return VMS condition codes from main. The default is to return POSIX
10973 style condition (e.g.@: error) codes.
10975 @item -mdebug-main=@var{prefix}
10976 @opindex mdebug-main=@var{prefix}
10977 Flag the first routine whose name starts with @var{prefix} as the main
10978 routine for the debugger.
10982 Default to 64bit memory allocation routines.
10986 @subsection FR30 Options
10987 @cindex FR30 Options
10989 These options are defined specifically for the FR30 port.
10993 @item -msmall-model
10994 @opindex msmall-model
10995 Use the small address space model. This can produce smaller code, but
10996 it does assume that all symbolic values and addresses will fit into a
11001 Assume that run-time support has been provided and so there is no need
11002 to include the simulator library (@file{libsim.a}) on the linker
11008 @subsection FRV Options
11009 @cindex FRV Options
11015 Only use the first 32 general purpose registers.
11020 Use all 64 general purpose registers.
11025 Use only the first 32 floating point registers.
11030 Use all 64 floating point registers
11033 @opindex mhard-float
11035 Use hardware instructions for floating point operations.
11038 @opindex msoft-float
11040 Use library routines for floating point operations.
11045 Dynamically allocate condition code registers.
11050 Do not try to dynamically allocate condition code registers, only
11051 use @code{icc0} and @code{fcc0}.
11056 Change ABI to use double word insns.
11061 Do not use double word instructions.
11066 Use floating point double instructions.
11069 @opindex mno-double
11071 Do not use floating point double instructions.
11076 Use media instructions.
11081 Do not use media instructions.
11086 Use multiply and add/subtract instructions.
11089 @opindex mno-muladd
11091 Do not use multiply and add/subtract instructions.
11096 Select the FDPIC ABI, that uses function descriptors to represent
11097 pointers to functions. Without any PIC/PIE-related options, it
11098 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
11099 assumes GOT entries and small data are within a 12-bit range from the
11100 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
11101 are computed with 32 bits.
11102 With a @samp{bfin-elf} target, this option implies @option{-msim}.
11105 @opindex minline-plt
11107 Enable inlining of PLT entries in function calls to functions that are
11108 not known to bind locally. It has no effect without @option{-mfdpic}.
11109 It's enabled by default if optimizing for speed and compiling for
11110 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
11111 optimization option such as @option{-O3} or above is present in the
11117 Assume a large TLS segment when generating thread-local code.
11122 Do not assume a large TLS segment when generating thread-local code.
11127 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
11128 that is known to be in read-only sections. It's enabled by default,
11129 except for @option{-fpic} or @option{-fpie}: even though it may help
11130 make the global offset table smaller, it trades 1 instruction for 4.
11131 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
11132 one of which may be shared by multiple symbols, and it avoids the need
11133 for a GOT entry for the referenced symbol, so it's more likely to be a
11134 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
11136 @item -multilib-library-pic
11137 @opindex multilib-library-pic
11139 Link with the (library, not FD) pic libraries. It's implied by
11140 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
11141 @option{-fpic} without @option{-mfdpic}. You should never have to use
11145 @opindex mlinked-fp
11147 Follow the EABI requirement of always creating a frame pointer whenever
11148 a stack frame is allocated. This option is enabled by default and can
11149 be disabled with @option{-mno-linked-fp}.
11152 @opindex mlong-calls
11154 Use indirect addressing to call functions outside the current
11155 compilation unit. This allows the functions to be placed anywhere
11156 within the 32-bit address space.
11158 @item -malign-labels
11159 @opindex malign-labels
11161 Try to align labels to an 8-byte boundary by inserting nops into the
11162 previous packet. This option only has an effect when VLIW packing
11163 is enabled. It doesn't create new packets; it merely adds nops to
11166 @item -mlibrary-pic
11167 @opindex mlibrary-pic
11169 Generate position-independent EABI code.
11174 Use only the first four media accumulator registers.
11179 Use all eight media accumulator registers.
11184 Pack VLIW instructions.
11189 Do not pack VLIW instructions.
11192 @opindex mno-eflags
11194 Do not mark ABI switches in e_flags.
11197 @opindex mcond-move
11199 Enable the use of conditional-move instructions (default).
11201 This switch is mainly for debugging the compiler and will likely be removed
11202 in a future version.
11204 @item -mno-cond-move
11205 @opindex mno-cond-move
11207 Disable the use of conditional-move instructions.
11209 This switch is mainly for debugging the compiler and will likely be removed
11210 in a future version.
11215 Enable the use of conditional set instructions (default).
11217 This switch is mainly for debugging the compiler and will likely be removed
11218 in a future version.
11223 Disable the use of conditional set instructions.
11225 This switch is mainly for debugging the compiler and will likely be removed
11226 in a future version.
11229 @opindex mcond-exec
11231 Enable the use of conditional execution (default).
11233 This switch is mainly for debugging the compiler and will likely be removed
11234 in a future version.
11236 @item -mno-cond-exec
11237 @opindex mno-cond-exec
11239 Disable the use of conditional execution.
11241 This switch is mainly for debugging the compiler and will likely be removed
11242 in a future version.
11244 @item -mvliw-branch
11245 @opindex mvliw-branch
11247 Run a pass to pack branches into VLIW instructions (default).
11249 This switch is mainly for debugging the compiler and will likely be removed
11250 in a future version.
11252 @item -mno-vliw-branch
11253 @opindex mno-vliw-branch
11255 Do not run a pass to pack branches into VLIW instructions.
11257 This switch is mainly for debugging the compiler and will likely be removed
11258 in a future version.
11260 @item -mmulti-cond-exec
11261 @opindex mmulti-cond-exec
11263 Enable optimization of @code{&&} and @code{||} in conditional execution
11266 This switch is mainly for debugging the compiler and will likely be removed
11267 in a future version.
11269 @item -mno-multi-cond-exec
11270 @opindex mno-multi-cond-exec
11272 Disable optimization of @code{&&} and @code{||} in conditional execution.
11274 This switch is mainly for debugging the compiler and will likely be removed
11275 in a future version.
11277 @item -mnested-cond-exec
11278 @opindex mnested-cond-exec
11280 Enable nested conditional execution optimizations (default).
11282 This switch is mainly for debugging the compiler and will likely be removed
11283 in a future version.
11285 @item -mno-nested-cond-exec
11286 @opindex mno-nested-cond-exec
11288 Disable nested conditional execution optimizations.
11290 This switch is mainly for debugging the compiler and will likely be removed
11291 in a future version.
11293 @item -moptimize-membar
11294 @opindex moptimize-membar
11296 This switch removes redundant @code{membar} instructions from the
11297 compiler generated code. It is enabled by default.
11299 @item -mno-optimize-membar
11300 @opindex mno-optimize-membar
11302 This switch disables the automatic removal of redundant @code{membar}
11303 instructions from the generated code.
11305 @item -mtomcat-stats
11306 @opindex mtomcat-stats
11308 Cause gas to print out tomcat statistics.
11310 @item -mcpu=@var{cpu}
11313 Select the processor type for which to generate code. Possible values are
11314 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
11315 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
11319 @node GNU/Linux Options
11320 @subsection GNU/Linux Options
11322 These @samp{-m} options are defined for GNU/Linux targets:
11327 Use the GNU C library instead of uClibc. This is the default except
11328 on @samp{*-*-linux-*uclibc*} targets.
11332 Use uClibc instead of the GNU C library. This is the default on
11333 @samp{*-*-linux-*uclibc*} targets.
11336 @node H8/300 Options
11337 @subsection H8/300 Options
11339 These @samp{-m} options are defined for the H8/300 implementations:
11344 Shorten some address references at link time, when possible; uses the
11345 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
11346 ld, Using ld}, for a fuller description.
11350 Generate code for the H8/300H@.
11354 Generate code for the H8S@.
11358 Generate code for the H8S and H8/300H in the normal mode. This switch
11359 must be used either with @option{-mh} or @option{-ms}.
11363 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
11367 Make @code{int} data 32 bits by default.
11370 @opindex malign-300
11371 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
11372 The default for the H8/300H and H8S is to align longs and floats on 4
11374 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
11375 This option has no effect on the H8/300.
11379 @subsection HPPA Options
11380 @cindex HPPA Options
11382 These @samp{-m} options are defined for the HPPA family of computers:
11385 @item -march=@var{architecture-type}
11387 Generate code for the specified architecture. The choices for
11388 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
11389 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
11390 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
11391 architecture option for your machine. Code compiled for lower numbered
11392 architectures will run on higher numbered architectures, but not the
11395 @item -mpa-risc-1-0
11396 @itemx -mpa-risc-1-1
11397 @itemx -mpa-risc-2-0
11398 @opindex mpa-risc-1-0
11399 @opindex mpa-risc-1-1
11400 @opindex mpa-risc-2-0
11401 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
11404 @opindex mbig-switch
11405 Generate code suitable for big switch tables. Use this option only if
11406 the assembler/linker complain about out of range branches within a switch
11409 @item -mjump-in-delay
11410 @opindex mjump-in-delay
11411 Fill delay slots of function calls with unconditional jump instructions
11412 by modifying the return pointer for the function call to be the target
11413 of the conditional jump.
11415 @item -mdisable-fpregs
11416 @opindex mdisable-fpregs
11417 Prevent floating point registers from being used in any manner. This is
11418 necessary for compiling kernels which perform lazy context switching of
11419 floating point registers. If you use this option and attempt to perform
11420 floating point operations, the compiler will abort.
11422 @item -mdisable-indexing
11423 @opindex mdisable-indexing
11424 Prevent the compiler from using indexing address modes. This avoids some
11425 rather obscure problems when compiling MIG generated code under MACH@.
11427 @item -mno-space-regs
11428 @opindex mno-space-regs
11429 Generate code that assumes the target has no space registers. This allows
11430 GCC to generate faster indirect calls and use unscaled index address modes.
11432 Such code is suitable for level 0 PA systems and kernels.
11434 @item -mfast-indirect-calls
11435 @opindex mfast-indirect-calls
11436 Generate code that assumes calls never cross space boundaries. This
11437 allows GCC to emit code which performs faster indirect calls.
11439 This option will not work in the presence of shared libraries or nested
11442 @item -mfixed-range=@var{register-range}
11443 @opindex mfixed-range
11444 Generate code treating the given register range as fixed registers.
11445 A fixed register is one that the register allocator can not use. This is
11446 useful when compiling kernel code. A register range is specified as
11447 two registers separated by a dash. Multiple register ranges can be
11448 specified separated by a comma.
11450 @item -mlong-load-store
11451 @opindex mlong-load-store
11452 Generate 3-instruction load and store sequences as sometimes required by
11453 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
11456 @item -mportable-runtime
11457 @opindex mportable-runtime
11458 Use the portable calling conventions proposed by HP for ELF systems.
11462 Enable the use of assembler directives only GAS understands.
11464 @item -mschedule=@var{cpu-type}
11466 Schedule code according to the constraints for the machine type
11467 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
11468 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
11469 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
11470 proper scheduling option for your machine. The default scheduling is
11474 @opindex mlinker-opt
11475 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
11476 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
11477 linkers in which they give bogus error messages when linking some programs.
11480 @opindex msoft-float
11481 Generate output containing library calls for floating point.
11482 @strong{Warning:} the requisite libraries are not available for all HPPA
11483 targets. Normally the facilities of the machine's usual C compiler are
11484 used, but this cannot be done directly in cross-compilation. You must make
11485 your own arrangements to provide suitable library functions for
11488 @option{-msoft-float} changes the calling convention in the output file;
11489 therefore, it is only useful if you compile @emph{all} of a program with
11490 this option. In particular, you need to compile @file{libgcc.a}, the
11491 library that comes with GCC, with @option{-msoft-float} in order for
11496 Generate the predefine, @code{_SIO}, for server IO@. The default is
11497 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
11498 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
11499 options are available under HP-UX and HI-UX@.
11503 Use GNU ld specific options. This passes @option{-shared} to ld when
11504 building a shared library. It is the default when GCC is configured,
11505 explicitly or implicitly, with the GNU linker. This option does not
11506 have any affect on which ld is called, it only changes what parameters
11507 are passed to that ld. The ld that is called is determined by the
11508 @option{--with-ld} configure option, GCC's program search path, and
11509 finally by the user's @env{PATH}. The linker used by GCC can be printed
11510 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
11511 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11515 Use HP ld specific options. This passes @option{-b} to ld when building
11516 a shared library and passes @option{+Accept TypeMismatch} to ld on all
11517 links. It is the default when GCC is configured, explicitly or
11518 implicitly, with the HP linker. This option does not have any affect on
11519 which ld is called, it only changes what parameters are passed to that
11520 ld. The ld that is called is determined by the @option{--with-ld}
11521 configure option, GCC's program search path, and finally by the user's
11522 @env{PATH}. The linker used by GCC can be printed using @samp{which
11523 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
11524 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11527 @opindex mno-long-calls
11528 Generate code that uses long call sequences. This ensures that a call
11529 is always able to reach linker generated stubs. The default is to generate
11530 long calls only when the distance from the call site to the beginning
11531 of the function or translation unit, as the case may be, exceeds a
11532 predefined limit set by the branch type being used. The limits for
11533 normal calls are 7,600,000 and 240,000 bytes, respectively for the
11534 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
11537 Distances are measured from the beginning of functions when using the
11538 @option{-ffunction-sections} option, or when using the @option{-mgas}
11539 and @option{-mno-portable-runtime} options together under HP-UX with
11542 It is normally not desirable to use this option as it will degrade
11543 performance. However, it may be useful in large applications,
11544 particularly when partial linking is used to build the application.
11546 The types of long calls used depends on the capabilities of the
11547 assembler and linker, and the type of code being generated. The
11548 impact on systems that support long absolute calls, and long pic
11549 symbol-difference or pc-relative calls should be relatively small.
11550 However, an indirect call is used on 32-bit ELF systems in pic code
11551 and it is quite long.
11553 @item -munix=@var{unix-std}
11555 Generate compiler predefines and select a startfile for the specified
11556 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
11557 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
11558 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
11559 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
11560 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
11563 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
11564 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
11565 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
11566 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
11567 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
11568 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
11570 It is @emph{important} to note that this option changes the interfaces
11571 for various library routines. It also affects the operational behavior
11572 of the C library. Thus, @emph{extreme} care is needed in using this
11575 Library code that is intended to operate with more than one UNIX
11576 standard must test, set and restore the variable @var{__xpg4_extended_mask}
11577 as appropriate. Most GNU software doesn't provide this capability.
11581 Suppress the generation of link options to search libdld.sl when the
11582 @option{-static} option is specified on HP-UX 10 and later.
11586 The HP-UX implementation of setlocale in libc has a dependency on
11587 libdld.sl. There isn't an archive version of libdld.sl. Thus,
11588 when the @option{-static} option is specified, special link options
11589 are needed to resolve this dependency.
11591 On HP-UX 10 and later, the GCC driver adds the necessary options to
11592 link with libdld.sl when the @option{-static} option is specified.
11593 This causes the resulting binary to be dynamic. On the 64-bit port,
11594 the linkers generate dynamic binaries by default in any case. The
11595 @option{-nolibdld} option can be used to prevent the GCC driver from
11596 adding these link options.
11600 Add support for multithreading with the @dfn{dce thread} library
11601 under HP-UX@. This option sets flags for both the preprocessor and
11605 @node i386 and x86-64 Options
11606 @subsection Intel 386 and AMD x86-64 Options
11607 @cindex i386 Options
11608 @cindex x86-64 Options
11609 @cindex Intel 386 Options
11610 @cindex AMD x86-64 Options
11612 These @samp{-m} options are defined for the i386 and x86-64 family of
11616 @item -mtune=@var{cpu-type}
11618 Tune to @var{cpu-type} everything applicable about the generated code, except
11619 for the ABI and the set of available instructions. The choices for
11620 @var{cpu-type} are:
11623 Produce code optimized for the most common IA32/AMD64/EM64T processors.
11624 If you know the CPU on which your code will run, then you should use
11625 the corresponding @option{-mtune} option instead of
11626 @option{-mtune=generic}. But, if you do not know exactly what CPU users
11627 of your application will have, then you should use this option.
11629 As new processors are deployed in the marketplace, the behavior of this
11630 option will change. Therefore, if you upgrade to a newer version of
11631 GCC, the code generated option will change to reflect the processors
11632 that were most common when that version of GCC was released.
11634 There is no @option{-march=generic} option because @option{-march}
11635 indicates the instruction set the compiler can use, and there is no
11636 generic instruction set applicable to all processors. In contrast,
11637 @option{-mtune} indicates the processor (or, in this case, collection of
11638 processors) for which the code is optimized.
11640 This selects the CPU to tune for at compilation time by determining
11641 the processor type of the compiling machine. Using @option{-mtune=native}
11642 will produce code optimized for the local machine under the constraints
11643 of the selected instruction set. Using @option{-march=native} will
11644 enable all instruction subsets supported by the local machine (hence
11645 the result might not run on different machines).
11647 Original Intel's i386 CPU@.
11649 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
11650 @item i586, pentium
11651 Intel Pentium CPU with no MMX support.
11653 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
11655 Intel PentiumPro CPU@.
11657 Same as @code{generic}, but when used as @code{march} option, PentiumPro
11658 instruction set will be used, so the code will run on all i686 family chips.
11660 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
11661 @item pentium3, pentium3m
11662 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
11665 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
11666 support. Used by Centrino notebooks.
11667 @item pentium4, pentium4m
11668 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
11670 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
11673 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
11674 SSE2 and SSE3 instruction set support.
11676 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
11677 instruction set support.
11679 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
11680 instruction set support.
11682 AMD K6 CPU with MMX instruction set support.
11684 Improved versions of AMD K6 CPU with MMX and 3dNOW!@: instruction set support.
11685 @item athlon, athlon-tbird
11686 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and SSE prefetch instructions
11688 @item athlon-4, athlon-xp, athlon-mp
11689 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW!@: and full SSE
11690 instruction set support.
11691 @item k8, opteron, athlon64, athlon-fx
11692 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
11693 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW!@: and 64-bit instruction set extensions.)
11694 @item k8-sse3, opteron-sse3, athlon64-sse3
11695 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
11696 @item amdfam10, barcelona
11697 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
11698 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3dNOW!, enhanced 3dNOW!, ABM and 64-bit
11699 instruction set extensions.)
11701 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
11704 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!@:
11705 instruction set support.
11707 Via C3 CPU with MMX and 3dNOW!@: instruction set support. (No scheduling is
11708 implemented for this chip.)
11710 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
11711 implemented for this chip.)
11713 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
11716 While picking a specific @var{cpu-type} will schedule things appropriately
11717 for that particular chip, the compiler will not generate any code that
11718 does not run on the i386 without the @option{-march=@var{cpu-type}} option
11721 @item -march=@var{cpu-type}
11723 Generate instructions for the machine type @var{cpu-type}. The choices
11724 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
11725 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
11727 @item -mcpu=@var{cpu-type}
11729 A deprecated synonym for @option{-mtune}.
11731 @item -mfpmath=@var{unit}
11733 Generate floating point arithmetics for selected unit @var{unit}. The choices
11734 for @var{unit} are:
11738 Use the standard 387 floating point coprocessor present majority of chips and
11739 emulated otherwise. Code compiled with this option will run almost everywhere.
11740 The temporary results are computed in 80bit precision instead of precision
11741 specified by the type resulting in slightly different results compared to most
11742 of other chips. See @option{-ffloat-store} for more detailed description.
11744 This is the default choice for i386 compiler.
11747 Use scalar floating point instructions present in the SSE instruction set.
11748 This instruction set is supported by Pentium3 and newer chips, in the AMD line
11749 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
11750 instruction set supports only single precision arithmetics, thus the double and
11751 extended precision arithmetics is still done using 387. Later version, present
11752 only in Pentium4 and the future AMD x86-64 chips supports double precision
11755 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
11756 or @option{-msse2} switches to enable SSE extensions and make this option
11757 effective. For the x86-64 compiler, these extensions are enabled by default.
11759 The resulting code should be considerably faster in the majority of cases and avoid
11760 the numerical instability problems of 387 code, but may break some existing
11761 code that expects temporaries to be 80bit.
11763 This is the default choice for the x86-64 compiler.
11768 Attempt to utilize both instruction sets at once. This effectively double the
11769 amount of available registers and on chips with separate execution units for
11770 387 and SSE the execution resources too. Use this option with care, as it is
11771 still experimental, because the GCC register allocator does not model separate
11772 functional units well resulting in instable performance.
11775 @item -masm=@var{dialect}
11776 @opindex masm=@var{dialect}
11777 Output asm instructions using selected @var{dialect}. Supported
11778 choices are @samp{intel} or @samp{att} (the default one). Darwin does
11779 not support @samp{intel}.
11782 @itemx -mno-ieee-fp
11784 @opindex mno-ieee-fp
11785 Control whether or not the compiler uses IEEE floating point
11786 comparisons. These handle correctly the case where the result of a
11787 comparison is unordered.
11790 @opindex msoft-float
11791 Generate output containing library calls for floating point.
11792 @strong{Warning:} the requisite libraries are not part of GCC@.
11793 Normally the facilities of the machine's usual C compiler are used, but
11794 this can't be done directly in cross-compilation. You must make your
11795 own arrangements to provide suitable library functions for
11798 On machines where a function returns floating point results in the 80387
11799 register stack, some floating point opcodes may be emitted even if
11800 @option{-msoft-float} is used.
11802 @item -mno-fp-ret-in-387
11803 @opindex mno-fp-ret-in-387
11804 Do not use the FPU registers for return values of functions.
11806 The usual calling convention has functions return values of types
11807 @code{float} and @code{double} in an FPU register, even if there
11808 is no FPU@. The idea is that the operating system should emulate
11811 The option @option{-mno-fp-ret-in-387} causes such values to be returned
11812 in ordinary CPU registers instead.
11814 @item -mno-fancy-math-387
11815 @opindex mno-fancy-math-387
11816 Some 387 emulators do not support the @code{sin}, @code{cos} and
11817 @code{sqrt} instructions for the 387. Specify this option to avoid
11818 generating those instructions. This option is the default on FreeBSD,
11819 OpenBSD and NetBSD@. This option is overridden when @option{-march}
11820 indicates that the target cpu will always have an FPU and so the
11821 instruction will not need emulation. As of revision 2.6.1, these
11822 instructions are not generated unless you also use the
11823 @option{-funsafe-math-optimizations} switch.
11825 @item -malign-double
11826 @itemx -mno-align-double
11827 @opindex malign-double
11828 @opindex mno-align-double
11829 Control whether GCC aligns @code{double}, @code{long double}, and
11830 @code{long long} variables on a two word boundary or a one word
11831 boundary. Aligning @code{double} variables on a two word boundary will
11832 produce code that runs somewhat faster on a @samp{Pentium} at the
11833 expense of more memory.
11835 On x86-64, @option{-malign-double} is enabled by default.
11837 @strong{Warning:} if you use the @option{-malign-double} switch,
11838 structures containing the above types will be aligned differently than
11839 the published application binary interface specifications for the 386
11840 and will not be binary compatible with structures in code compiled
11841 without that switch.
11843 @item -m96bit-long-double
11844 @itemx -m128bit-long-double
11845 @opindex m96bit-long-double
11846 @opindex m128bit-long-double
11847 These switches control the size of @code{long double} type. The i386
11848 application binary interface specifies the size to be 96 bits,
11849 so @option{-m96bit-long-double} is the default in 32 bit mode.
11851 Modern architectures (Pentium and newer) would prefer @code{long double}
11852 to be aligned to an 8 or 16 byte boundary. In arrays or structures
11853 conforming to the ABI, this would not be possible. So specifying a
11854 @option{-m128bit-long-double} will align @code{long double}
11855 to a 16 byte boundary by padding the @code{long double} with an additional
11858 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
11859 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
11861 Notice that neither of these options enable any extra precision over the x87
11862 standard of 80 bits for a @code{long double}.
11864 @strong{Warning:} if you override the default value for your target ABI, the
11865 structures and arrays containing @code{long double} variables will change
11866 their size as well as function calling convention for function taking
11867 @code{long double} will be modified. Hence they will not be binary
11868 compatible with arrays or structures in code compiled without that switch.
11870 @item -mlarge-data-threshold=@var{number}
11871 @opindex mlarge-data-threshold=@var{number}
11872 When @option{-mcmodel=medium} is specified, the data greater than
11873 @var{threshold} are placed in large data section. This value must be the
11874 same across all object linked into the binary and defaults to 65535.
11878 Use a different function-calling convention, in which functions that
11879 take a fixed number of arguments return with the @code{ret} @var{num}
11880 instruction, which pops their arguments while returning. This saves one
11881 instruction in the caller since there is no need to pop the arguments
11884 You can specify that an individual function is called with this calling
11885 sequence with the function attribute @samp{stdcall}. You can also
11886 override the @option{-mrtd} option by using the function attribute
11887 @samp{cdecl}. @xref{Function Attributes}.
11889 @strong{Warning:} this calling convention is incompatible with the one
11890 normally used on Unix, so you cannot use it if you need to call
11891 libraries compiled with the Unix compiler.
11893 Also, you must provide function prototypes for all functions that
11894 take variable numbers of arguments (including @code{printf});
11895 otherwise incorrect code will be generated for calls to those
11898 In addition, seriously incorrect code will result if you call a
11899 function with too many arguments. (Normally, extra arguments are
11900 harmlessly ignored.)
11902 @item -mregparm=@var{num}
11904 Control how many registers are used to pass integer arguments. By
11905 default, no registers are used to pass arguments, and at most 3
11906 registers can be used. You can control this behavior for a specific
11907 function by using the function attribute @samp{regparm}.
11908 @xref{Function Attributes}.
11910 @strong{Warning:} if you use this switch, and
11911 @var{num} is nonzero, then you must build all modules with the same
11912 value, including any libraries. This includes the system libraries and
11916 @opindex msseregparm
11917 Use SSE register passing conventions for float and double arguments
11918 and return values. You can control this behavior for a specific
11919 function by using the function attribute @samp{sseregparm}.
11920 @xref{Function Attributes}.
11922 @strong{Warning:} if you use this switch then you must build all
11923 modules with the same value, including any libraries. This includes
11924 the system libraries and startup modules.
11933 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
11934 is specified, the significands of results of floating-point operations are
11935 rounded to 24 bits (single precision); @option{-mpc64} rounds the
11936 significands of results of floating-point operations to 53 bits (double
11937 precision) and @option{-mpc80} rounds the significands of results of
11938 floating-point operations to 64 bits (extended double precision), which is
11939 the default. When this option is used, floating-point operations in higher
11940 precisions are not available to the programmer without setting the FPU
11941 control word explicitly.
11943 Setting the rounding of floating-point operations to less than the default
11944 80 bits can speed some programs by 2% or more. Note that some mathematical
11945 libraries assume that extended precision (80 bit) floating-point operations
11946 are enabled by default; routines in such libraries could suffer significant
11947 loss of accuracy, typically through so-called "catastrophic cancellation",
11948 when this option is used to set the precision to less than extended precision.
11950 @item -mstackrealign
11951 @opindex mstackrealign
11952 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
11953 option will generate an alternate prologue and epilogue that realigns the
11954 runtime stack if necessary. This supports mixing legacy codes that keep
11955 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
11956 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
11957 applicable to individual functions.
11959 @item -mpreferred-stack-boundary=@var{num}
11960 @opindex mpreferred-stack-boundary
11961 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
11962 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
11963 the default is 4 (16 bytes or 128 bits).
11965 @item -mincoming-stack-boundary=@var{num}
11966 @opindex mincoming-stack-boundary
11967 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
11968 boundary. If @option{-mincoming-stack-boundary} is not specified,
11969 the one specified by @option{-mpreferred-stack-boundary} will be used.
11971 On Pentium and PentiumPro, @code{double} and @code{long double} values
11972 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
11973 suffer significant run time performance penalties. On Pentium III, the
11974 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
11975 properly if it is not 16 byte aligned.
11977 To ensure proper alignment of this values on the stack, the stack boundary
11978 must be as aligned as that required by any value stored on the stack.
11979 Further, every function must be generated such that it keeps the stack
11980 aligned. Thus calling a function compiled with a higher preferred
11981 stack boundary from a function compiled with a lower preferred stack
11982 boundary will most likely misalign the stack. It is recommended that
11983 libraries that use callbacks always use the default setting.
11985 This extra alignment does consume extra stack space, and generally
11986 increases code size. Code that is sensitive to stack space usage, such
11987 as embedded systems and operating system kernels, may want to reduce the
11988 preferred alignment to @option{-mpreferred-stack-boundary=2}.
12032 These switches enable or disable the use of instructions in the MMX,
12033 SSE, SSE2, SSE3, SSSE3, SSE4.1, AVX, AES, PCLMUL, SSE4A, FMA4, XOP,
12034 LWP, ABM or 3DNow!@: extended instruction sets.
12035 These extensions are also available as built-in functions: see
12036 @ref{X86 Built-in Functions}, for details of the functions enabled and
12037 disabled by these switches.
12039 To have SSE/SSE2 instructions generated automatically from floating-point
12040 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
12042 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
12043 generates new AVX instructions or AVX equivalence for all SSEx instructions
12046 These options will enable GCC to use these extended instructions in
12047 generated code, even without @option{-mfpmath=sse}. Applications which
12048 perform runtime CPU detection must compile separate files for each
12049 supported architecture, using the appropriate flags. In particular,
12050 the file containing the CPU detection code should be compiled without
12055 This option instructs GCC to emit a @code{cld} instruction in the prologue
12056 of functions that use string instructions. String instructions depend on
12057 the DF flag to select between autoincrement or autodecrement mode. While the
12058 ABI specifies the DF flag to be cleared on function entry, some operating
12059 systems violate this specification by not clearing the DF flag in their
12060 exception dispatchers. The exception handler can be invoked with the DF flag
12061 set which leads to wrong direction mode, when string instructions are used.
12062 This option can be enabled by default on 32-bit x86 targets by configuring
12063 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
12064 instructions can be suppressed with the @option{-mno-cld} compiler option
12069 This option will enable GCC to use CMPXCHG16B instruction in generated code.
12070 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
12071 data types. This is useful for high resolution counters that could be updated
12072 by multiple processors (or cores). This instruction is generated as part of
12073 atomic built-in functions: see @ref{Atomic Builtins} for details.
12077 This option will enable GCC to use SAHF instruction in generated 64-bit code.
12078 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
12079 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
12080 SAHF are load and store instructions, respectively, for certain status flags.
12081 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
12082 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
12086 This option will enable GCC to use movbe instruction to implement
12087 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
12091 This option will enable built-in functions, @code{__builtin_ia32_crc32qi},
12092 @code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and
12093 @code{__builtin_ia32_crc32di} to generate the crc32 machine instruction.
12097 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
12098 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
12099 to increase precision instead of DIVSS and SQRTSS (and their vectorized
12100 variants) for single precision floating point arguments. These instructions
12101 are generated only when @option{-funsafe-math-optimizations} is enabled
12102 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
12103 Note that while the throughput of the sequence is higher than the throughput
12104 of the non-reciprocal instruction, the precision of the sequence can be
12105 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
12107 Note that GCC implements 1.0f/sqrtf(x) in terms of RSQRTSS (or RSQRTPS)
12108 already with @option{-ffast-math} (or the above option combination), and
12109 doesn't need @option{-mrecip}.
12111 @item -mveclibabi=@var{type}
12112 @opindex mveclibabi
12113 Specifies the ABI type to use for vectorizing intrinsics using an
12114 external library. Supported types are @code{svml} for the Intel short
12115 vector math library and @code{acml} for the AMD math core library style
12116 of interfacing. GCC will currently emit calls to @code{vmldExp2},
12117 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
12118 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
12119 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
12120 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
12121 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
12122 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
12123 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
12124 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
12125 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
12126 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
12127 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
12128 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
12129 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
12130 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
12131 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
12132 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
12133 compatible library will have to be specified at link time.
12135 @item -mabi=@var{name}
12137 Generate code for the specified calling convention. Permissible values
12138 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
12139 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
12140 ABI when targeting Windows. On all other systems, the default is the
12141 SYSV ABI. You can control this behavior for a specific function by
12142 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
12143 @xref{Function Attributes}.
12146 @itemx -mno-push-args
12147 @opindex mpush-args
12148 @opindex mno-push-args
12149 Use PUSH operations to store outgoing parameters. This method is shorter
12150 and usually equally fast as method using SUB/MOV operations and is enabled
12151 by default. In some cases disabling it may improve performance because of
12152 improved scheduling and reduced dependencies.
12154 @item -maccumulate-outgoing-args
12155 @opindex maccumulate-outgoing-args
12156 If enabled, the maximum amount of space required for outgoing arguments will be
12157 computed in the function prologue. This is faster on most modern CPUs
12158 because of reduced dependencies, improved scheduling and reduced stack usage
12159 when preferred stack boundary is not equal to 2. The drawback is a notable
12160 increase in code size. This switch implies @option{-mno-push-args}.
12164 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
12165 on thread-safe exception handling must compile and link all code with the
12166 @option{-mthreads} option. When compiling, @option{-mthreads} defines
12167 @option{-D_MT}; when linking, it links in a special thread helper library
12168 @option{-lmingwthrd} which cleans up per thread exception handling data.
12170 @item -mno-align-stringops
12171 @opindex mno-align-stringops
12172 Do not align destination of inlined string operations. This switch reduces
12173 code size and improves performance in case the destination is already aligned,
12174 but GCC doesn't know about it.
12176 @item -minline-all-stringops
12177 @opindex minline-all-stringops
12178 By default GCC inlines string operations only when destination is known to be
12179 aligned at least to 4 byte boundary. This enables more inlining, increase code
12180 size, but may improve performance of code that depends on fast memcpy, strlen
12181 and memset for short lengths.
12183 @item -minline-stringops-dynamically
12184 @opindex minline-stringops-dynamically
12185 For string operation of unknown size, inline runtime checks so for small
12186 blocks inline code is used, while for large blocks library call is used.
12188 @item -mstringop-strategy=@var{alg}
12189 @opindex mstringop-strategy=@var{alg}
12190 Overwrite internal decision heuristic about particular algorithm to inline
12191 string operation with. The allowed values are @code{rep_byte},
12192 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
12193 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
12194 expanding inline loop, @code{libcall} for always expanding library call.
12196 @item -momit-leaf-frame-pointer
12197 @opindex momit-leaf-frame-pointer
12198 Don't keep the frame pointer in a register for leaf functions. This
12199 avoids the instructions to save, set up and restore frame pointers and
12200 makes an extra register available in leaf functions. The option
12201 @option{-fomit-frame-pointer} removes the frame pointer for all functions
12202 which might make debugging harder.
12204 @item -mtls-direct-seg-refs
12205 @itemx -mno-tls-direct-seg-refs
12206 @opindex mtls-direct-seg-refs
12207 Controls whether TLS variables may be accessed with offsets from the
12208 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
12209 or whether the thread base pointer must be added. Whether or not this
12210 is legal depends on the operating system, and whether it maps the
12211 segment to cover the entire TLS area.
12213 For systems that use GNU libc, the default is on.
12216 @itemx -mno-sse2avx
12218 Specify that the assembler should encode SSE instructions with VEX
12219 prefix. The option @option{-mavx} turns this on by default.
12222 These @samp{-m} switches are supported in addition to the above
12223 on AMD x86-64 processors in 64-bit environments.
12230 Generate code for a 32-bit or 64-bit environment.
12231 The 32-bit environment sets int, long and pointer to 32 bits and
12232 generates code that runs on any i386 system.
12233 The 64-bit environment sets int to 32 bits and long and pointer
12234 to 64 bits and generates code for AMD's x86-64 architecture. For
12235 darwin only the -m64 option turns off the @option{-fno-pic} and
12236 @option{-mdynamic-no-pic} options.
12238 @item -mno-red-zone
12239 @opindex mno-red-zone
12240 Do not use a so called red zone for x86-64 code. The red zone is mandated
12241 by the x86-64 ABI, it is a 128-byte area beyond the location of the
12242 stack pointer that will not be modified by signal or interrupt handlers
12243 and therefore can be used for temporary data without adjusting the stack
12244 pointer. The flag @option{-mno-red-zone} disables this red zone.
12246 @item -mcmodel=small
12247 @opindex mcmodel=small
12248 Generate code for the small code model: the program and its symbols must
12249 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
12250 Programs can be statically or dynamically linked. This is the default
12253 @item -mcmodel=kernel
12254 @opindex mcmodel=kernel
12255 Generate code for the kernel code model. The kernel runs in the
12256 negative 2 GB of the address space.
12257 This model has to be used for Linux kernel code.
12259 @item -mcmodel=medium
12260 @opindex mcmodel=medium
12261 Generate code for the medium model: The program is linked in the lower 2
12262 GB of the address space. Small symbols are also placed there. Symbols
12263 with sizes larger than @option{-mlarge-data-threshold} are put into
12264 large data or bss sections and can be located above 2GB. Programs can
12265 be statically or dynamically linked.
12267 @item -mcmodel=large
12268 @opindex mcmodel=large
12269 Generate code for the large model: This model makes no assumptions
12270 about addresses and sizes of sections.
12273 @node IA-64 Options
12274 @subsection IA-64 Options
12275 @cindex IA-64 Options
12277 These are the @samp{-m} options defined for the Intel IA-64 architecture.
12281 @opindex mbig-endian
12282 Generate code for a big endian target. This is the default for HP-UX@.
12284 @item -mlittle-endian
12285 @opindex mlittle-endian
12286 Generate code for a little endian target. This is the default for AIX5
12292 @opindex mno-gnu-as
12293 Generate (or don't) code for the GNU assembler. This is the default.
12294 @c Also, this is the default if the configure option @option{--with-gnu-as}
12300 @opindex mno-gnu-ld
12301 Generate (or don't) code for the GNU linker. This is the default.
12302 @c Also, this is the default if the configure option @option{--with-gnu-ld}
12307 Generate code that does not use a global pointer register. The result
12308 is not position independent code, and violates the IA-64 ABI@.
12310 @item -mvolatile-asm-stop
12311 @itemx -mno-volatile-asm-stop
12312 @opindex mvolatile-asm-stop
12313 @opindex mno-volatile-asm-stop
12314 Generate (or don't) a stop bit immediately before and after volatile asm
12317 @item -mregister-names
12318 @itemx -mno-register-names
12319 @opindex mregister-names
12320 @opindex mno-register-names
12321 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
12322 the stacked registers. This may make assembler output more readable.
12328 Disable (or enable) optimizations that use the small data section. This may
12329 be useful for working around optimizer bugs.
12331 @item -mconstant-gp
12332 @opindex mconstant-gp
12333 Generate code that uses a single constant global pointer value. This is
12334 useful when compiling kernel code.
12338 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
12339 This is useful when compiling firmware code.
12341 @item -minline-float-divide-min-latency
12342 @opindex minline-float-divide-min-latency
12343 Generate code for inline divides of floating point values
12344 using the minimum latency algorithm.
12346 @item -minline-float-divide-max-throughput
12347 @opindex minline-float-divide-max-throughput
12348 Generate code for inline divides of floating point values
12349 using the maximum throughput algorithm.
12351 @item -mno-inline-float-divide
12352 @opindex mno-inline-float-divide
12353 Do not generate inline code for divides of floating point values.
12355 @item -minline-int-divide-min-latency
12356 @opindex minline-int-divide-min-latency
12357 Generate code for inline divides of integer values
12358 using the minimum latency algorithm.
12360 @item -minline-int-divide-max-throughput
12361 @opindex minline-int-divide-max-throughput
12362 Generate code for inline divides of integer values
12363 using the maximum throughput algorithm.
12365 @item -mno-inline-int-divide
12366 @opindex mno-inline-int-divide
12367 Do not generate inline code for divides of integer values.
12369 @item -minline-sqrt-min-latency
12370 @opindex minline-sqrt-min-latency
12371 Generate code for inline square roots
12372 using the minimum latency algorithm.
12374 @item -minline-sqrt-max-throughput
12375 @opindex minline-sqrt-max-throughput
12376 Generate code for inline square roots
12377 using the maximum throughput algorithm.
12379 @item -mno-inline-sqrt
12380 @opindex mno-inline-sqrt
12381 Do not generate inline code for sqrt.
12384 @itemx -mno-fused-madd
12385 @opindex mfused-madd
12386 @opindex mno-fused-madd
12387 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
12388 instructions. The default is to use these instructions.
12390 @item -mno-dwarf2-asm
12391 @itemx -mdwarf2-asm
12392 @opindex mno-dwarf2-asm
12393 @opindex mdwarf2-asm
12394 Don't (or do) generate assembler code for the DWARF2 line number debugging
12395 info. This may be useful when not using the GNU assembler.
12397 @item -mearly-stop-bits
12398 @itemx -mno-early-stop-bits
12399 @opindex mearly-stop-bits
12400 @opindex mno-early-stop-bits
12401 Allow stop bits to be placed earlier than immediately preceding the
12402 instruction that triggered the stop bit. This can improve instruction
12403 scheduling, but does not always do so.
12405 @item -mfixed-range=@var{register-range}
12406 @opindex mfixed-range
12407 Generate code treating the given register range as fixed registers.
12408 A fixed register is one that the register allocator can not use. This is
12409 useful when compiling kernel code. A register range is specified as
12410 two registers separated by a dash. Multiple register ranges can be
12411 specified separated by a comma.
12413 @item -mtls-size=@var{tls-size}
12415 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
12418 @item -mtune=@var{cpu-type}
12420 Tune the instruction scheduling for a particular CPU, Valid values are
12421 itanium, itanium1, merced, itanium2, and mckinley.
12427 Generate code for a 32-bit or 64-bit environment.
12428 The 32-bit environment sets int, long and pointer to 32 bits.
12429 The 64-bit environment sets int to 32 bits and long and pointer
12430 to 64 bits. These are HP-UX specific flags.
12432 @item -mno-sched-br-data-spec
12433 @itemx -msched-br-data-spec
12434 @opindex mno-sched-br-data-spec
12435 @opindex msched-br-data-spec
12436 (Dis/En)able data speculative scheduling before reload.
12437 This will result in generation of the ld.a instructions and
12438 the corresponding check instructions (ld.c / chk.a).
12439 The default is 'disable'.
12441 @item -msched-ar-data-spec
12442 @itemx -mno-sched-ar-data-spec
12443 @opindex msched-ar-data-spec
12444 @opindex mno-sched-ar-data-spec
12445 (En/Dis)able data speculative scheduling after reload.
12446 This will result in generation of the ld.a instructions and
12447 the corresponding check instructions (ld.c / chk.a).
12448 The default is 'enable'.
12450 @item -mno-sched-control-spec
12451 @itemx -msched-control-spec
12452 @opindex mno-sched-control-spec
12453 @opindex msched-control-spec
12454 (Dis/En)able control speculative scheduling. This feature is
12455 available only during region scheduling (i.e.@: before reload).
12456 This will result in generation of the ld.s instructions and
12457 the corresponding check instructions chk.s .
12458 The default is 'disable'.
12460 @item -msched-br-in-data-spec
12461 @itemx -mno-sched-br-in-data-spec
12462 @opindex msched-br-in-data-spec
12463 @opindex mno-sched-br-in-data-spec
12464 (En/Dis)able speculative scheduling of the instructions that
12465 are dependent on the data speculative loads before reload.
12466 This is effective only with @option{-msched-br-data-spec} enabled.
12467 The default is 'enable'.
12469 @item -msched-ar-in-data-spec
12470 @itemx -mno-sched-ar-in-data-spec
12471 @opindex msched-ar-in-data-spec
12472 @opindex mno-sched-ar-in-data-spec
12473 (En/Dis)able speculative scheduling of the instructions that
12474 are dependent on the data speculative loads after reload.
12475 This is effective only with @option{-msched-ar-data-spec} enabled.
12476 The default is 'enable'.
12478 @item -msched-in-control-spec
12479 @itemx -mno-sched-in-control-spec
12480 @opindex msched-in-control-spec
12481 @opindex mno-sched-in-control-spec
12482 (En/Dis)able speculative scheduling of the instructions that
12483 are dependent on the control speculative loads.
12484 This is effective only with @option{-msched-control-spec} enabled.
12485 The default is 'enable'.
12487 @item -mno-sched-prefer-non-data-spec-insns
12488 @itemx -msched-prefer-non-data-spec-insns
12489 @opindex mno-sched-prefer-non-data-spec-insns
12490 @opindex msched-prefer-non-data-spec-insns
12491 If enabled, data speculative instructions will be chosen for schedule
12492 only if there are no other choices at the moment. This will make
12493 the use of the data speculation much more conservative.
12494 The default is 'disable'.
12496 @item -mno-sched-prefer-non-control-spec-insns
12497 @itemx -msched-prefer-non-control-spec-insns
12498 @opindex mno-sched-prefer-non-control-spec-insns
12499 @opindex msched-prefer-non-control-spec-insns
12500 If enabled, control speculative instructions will be chosen for schedule
12501 only if there are no other choices at the moment. This will make
12502 the use of the control speculation much more conservative.
12503 The default is 'disable'.
12505 @item -mno-sched-count-spec-in-critical-path
12506 @itemx -msched-count-spec-in-critical-path
12507 @opindex mno-sched-count-spec-in-critical-path
12508 @opindex msched-count-spec-in-critical-path
12509 If enabled, speculative dependencies will be considered during
12510 computation of the instructions priorities. This will make the use of the
12511 speculation a bit more conservative.
12512 The default is 'disable'.
12514 @item -msched-spec-ldc
12515 @opindex msched-spec-ldc
12516 Use a simple data speculation check. This option is on by default.
12518 @item -msched-control-spec-ldc
12519 @opindex msched-spec-ldc
12520 Use a simple check for control speculation. This option is on by default.
12522 @item -msched-stop-bits-after-every-cycle
12523 @opindex msched-stop-bits-after-every-cycle
12524 Place a stop bit after every cycle when scheduling. This option is on
12527 @item -msched-fp-mem-deps-zero-cost
12528 @opindex msched-fp-mem-deps-zero-cost
12529 Assume that floating-point stores and loads are not likely to cause a conflict
12530 when placed into the same instruction group. This option is disabled by
12533 @item -msel-sched-dont-check-control-spec
12534 @opindex msel-sched-dont-check-control-spec
12535 Generate checks for control speculation in selective scheduling.
12536 This flag is disabled by default.
12538 @item -msched-max-memory-insns=@var{max-insns}
12539 @opindex msched-max-memory-insns
12540 Limit on the number of memory insns per instruction group, giving lower
12541 priority to subsequent memory insns attempting to schedule in the same
12542 instruction group. Frequently useful to prevent cache bank conflicts.
12543 The default value is 1.
12545 @item -msched-max-memory-insns-hard-limit
12546 @opindex msched-max-memory-insns-hard-limit
12547 Disallow more than `msched-max-memory-insns' in instruction group.
12548 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
12549 when limit is reached but may still schedule memory operations.
12553 @node IA-64/VMS Options
12554 @subsection IA-64/VMS Options
12556 These @samp{-m} options are defined for the IA-64/VMS implementations:
12559 @item -mvms-return-codes
12560 @opindex mvms-return-codes
12561 Return VMS condition codes from main. The default is to return POSIX
12562 style condition (e.g.@ error) codes.
12564 @item -mdebug-main=@var{prefix}
12565 @opindex mdebug-main=@var{prefix}
12566 Flag the first routine whose name starts with @var{prefix} as the main
12567 routine for the debugger.
12571 Default to 64bit memory allocation routines.
12575 @subsection LM32 Options
12576 @cindex LM32 options
12578 These @option{-m} options are defined for the Lattice Mico32 architecture:
12581 @item -mbarrel-shift-enabled
12582 @opindex mbarrel-shift-enabled
12583 Enable barrel-shift instructions.
12585 @item -mdivide-enabled
12586 @opindex mdivide-enabled
12587 Enable divide and modulus instructions.
12589 @item -mmultiply-enabled
12590 @opindex multiply-enabled
12591 Enable multiply instructions.
12593 @item -msign-extend-enabled
12594 @opindex msign-extend-enabled
12595 Enable sign extend instructions.
12597 @item -muser-enabled
12598 @opindex muser-enabled
12599 Enable user-defined instructions.
12604 @subsection M32C Options
12605 @cindex M32C options
12608 @item -mcpu=@var{name}
12610 Select the CPU for which code is generated. @var{name} may be one of
12611 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
12612 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
12613 the M32C/80 series.
12617 Specifies that the program will be run on the simulator. This causes
12618 an alternate runtime library to be linked in which supports, for
12619 example, file I/O@. You must not use this option when generating
12620 programs that will run on real hardware; you must provide your own
12621 runtime library for whatever I/O functions are needed.
12623 @item -memregs=@var{number}
12625 Specifies the number of memory-based pseudo-registers GCC will use
12626 during code generation. These pseudo-registers will be used like real
12627 registers, so there is a tradeoff between GCC's ability to fit the
12628 code into available registers, and the performance penalty of using
12629 memory instead of registers. Note that all modules in a program must
12630 be compiled with the same value for this option. Because of that, you
12631 must not use this option with the default runtime libraries gcc
12636 @node M32R/D Options
12637 @subsection M32R/D Options
12638 @cindex M32R/D options
12640 These @option{-m} options are defined for Renesas M32R/D architectures:
12645 Generate code for the M32R/2@.
12649 Generate code for the M32R/X@.
12653 Generate code for the M32R@. This is the default.
12655 @item -mmodel=small
12656 @opindex mmodel=small
12657 Assume all objects live in the lower 16MB of memory (so that their addresses
12658 can be loaded with the @code{ld24} instruction), and assume all subroutines
12659 are reachable with the @code{bl} instruction.
12660 This is the default.
12662 The addressability of a particular object can be set with the
12663 @code{model} attribute.
12665 @item -mmodel=medium
12666 @opindex mmodel=medium
12667 Assume objects may be anywhere in the 32-bit address space (the compiler
12668 will generate @code{seth/add3} instructions to load their addresses), and
12669 assume all subroutines are reachable with the @code{bl} instruction.
12671 @item -mmodel=large
12672 @opindex mmodel=large
12673 Assume objects may be anywhere in the 32-bit address space (the compiler
12674 will generate @code{seth/add3} instructions to load their addresses), and
12675 assume subroutines may not be reachable with the @code{bl} instruction
12676 (the compiler will generate the much slower @code{seth/add3/jl}
12677 instruction sequence).
12680 @opindex msdata=none
12681 Disable use of the small data area. Variables will be put into
12682 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
12683 @code{section} attribute has been specified).
12684 This is the default.
12686 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
12687 Objects may be explicitly put in the small data area with the
12688 @code{section} attribute using one of these sections.
12690 @item -msdata=sdata
12691 @opindex msdata=sdata
12692 Put small global and static data in the small data area, but do not
12693 generate special code to reference them.
12696 @opindex msdata=use
12697 Put small global and static data in the small data area, and generate
12698 special instructions to reference them.
12702 @cindex smaller data references
12703 Put global and static objects less than or equal to @var{num} bytes
12704 into the small data or bss sections instead of the normal data or bss
12705 sections. The default value of @var{num} is 8.
12706 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
12707 for this option to have any effect.
12709 All modules should be compiled with the same @option{-G @var{num}} value.
12710 Compiling with different values of @var{num} may or may not work; if it
12711 doesn't the linker will give an error message---incorrect code will not be
12716 Makes the M32R specific code in the compiler display some statistics
12717 that might help in debugging programs.
12719 @item -malign-loops
12720 @opindex malign-loops
12721 Align all loops to a 32-byte boundary.
12723 @item -mno-align-loops
12724 @opindex mno-align-loops
12725 Do not enforce a 32-byte alignment for loops. This is the default.
12727 @item -missue-rate=@var{number}
12728 @opindex missue-rate=@var{number}
12729 Issue @var{number} instructions per cycle. @var{number} can only be 1
12732 @item -mbranch-cost=@var{number}
12733 @opindex mbranch-cost=@var{number}
12734 @var{number} can only be 1 or 2. If it is 1 then branches will be
12735 preferred over conditional code, if it is 2, then the opposite will
12738 @item -mflush-trap=@var{number}
12739 @opindex mflush-trap=@var{number}
12740 Specifies the trap number to use to flush the cache. The default is
12741 12. Valid numbers are between 0 and 15 inclusive.
12743 @item -mno-flush-trap
12744 @opindex mno-flush-trap
12745 Specifies that the cache cannot be flushed by using a trap.
12747 @item -mflush-func=@var{name}
12748 @opindex mflush-func=@var{name}
12749 Specifies the name of the operating system function to call to flush
12750 the cache. The default is @emph{_flush_cache}, but a function call
12751 will only be used if a trap is not available.
12753 @item -mno-flush-func
12754 @opindex mno-flush-func
12755 Indicates that there is no OS function for flushing the cache.
12759 @node M680x0 Options
12760 @subsection M680x0 Options
12761 @cindex M680x0 options
12763 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
12764 The default settings depend on which architecture was selected when
12765 the compiler was configured; the defaults for the most common choices
12769 @item -march=@var{arch}
12771 Generate code for a specific M680x0 or ColdFire instruction set
12772 architecture. Permissible values of @var{arch} for M680x0
12773 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
12774 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
12775 architectures are selected according to Freescale's ISA classification
12776 and the permissible values are: @samp{isaa}, @samp{isaaplus},
12777 @samp{isab} and @samp{isac}.
12779 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
12780 code for a ColdFire target. The @var{arch} in this macro is one of the
12781 @option{-march} arguments given above.
12783 When used together, @option{-march} and @option{-mtune} select code
12784 that runs on a family of similar processors but that is optimized
12785 for a particular microarchitecture.
12787 @item -mcpu=@var{cpu}
12789 Generate code for a specific M680x0 or ColdFire processor.
12790 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
12791 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
12792 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
12793 below, which also classifies the CPUs into families:
12795 @multitable @columnfractions 0.20 0.80
12796 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
12797 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
12798 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
12799 @item @samp{5206e} @tab @samp{5206e}
12800 @item @samp{5208} @tab @samp{5207} @samp{5208}
12801 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
12802 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
12803 @item @samp{5216} @tab @samp{5214} @samp{5216}
12804 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
12805 @item @samp{5225} @tab @samp{5224} @samp{5225}
12806 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
12807 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
12808 @item @samp{5249} @tab @samp{5249}
12809 @item @samp{5250} @tab @samp{5250}
12810 @item @samp{5271} @tab @samp{5270} @samp{5271}
12811 @item @samp{5272} @tab @samp{5272}
12812 @item @samp{5275} @tab @samp{5274} @samp{5275}
12813 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
12814 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
12815 @item @samp{5307} @tab @samp{5307}
12816 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
12817 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
12818 @item @samp{5407} @tab @samp{5407}
12819 @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}
12822 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
12823 @var{arch} is compatible with @var{cpu}. Other combinations of
12824 @option{-mcpu} and @option{-march} are rejected.
12826 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
12827 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
12828 where the value of @var{family} is given by the table above.
12830 @item -mtune=@var{tune}
12832 Tune the code for a particular microarchitecture, within the
12833 constraints set by @option{-march} and @option{-mcpu}.
12834 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
12835 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
12836 and @samp{cpu32}. The ColdFire microarchitectures
12837 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
12839 You can also use @option{-mtune=68020-40} for code that needs
12840 to run relatively well on 68020, 68030 and 68040 targets.
12841 @option{-mtune=68020-60} is similar but includes 68060 targets
12842 as well. These two options select the same tuning decisions as
12843 @option{-m68020-40} and @option{-m68020-60} respectively.
12845 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
12846 when tuning for 680x0 architecture @var{arch}. It also defines
12847 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
12848 option is used. If gcc is tuning for a range of architectures,
12849 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
12850 it defines the macros for every architecture in the range.
12852 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
12853 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
12854 of the arguments given above.
12860 Generate output for a 68000. This is the default
12861 when the compiler is configured for 68000-based systems.
12862 It is equivalent to @option{-march=68000}.
12864 Use this option for microcontrollers with a 68000 or EC000 core,
12865 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
12869 Generate output for a 68010. This is the default
12870 when the compiler is configured for 68010-based systems.
12871 It is equivalent to @option{-march=68010}.
12877 Generate output for a 68020. This is the default
12878 when the compiler is configured for 68020-based systems.
12879 It is equivalent to @option{-march=68020}.
12883 Generate output for a 68030. This is the default when the compiler is
12884 configured for 68030-based systems. It is equivalent to
12885 @option{-march=68030}.
12889 Generate output for a 68040. This is the default when the compiler is
12890 configured for 68040-based systems. It is equivalent to
12891 @option{-march=68040}.
12893 This option inhibits the use of 68881/68882 instructions that have to be
12894 emulated by software on the 68040. Use this option if your 68040 does not
12895 have code to emulate those instructions.
12899 Generate output for a 68060. This is the default when the compiler is
12900 configured for 68060-based systems. It is equivalent to
12901 @option{-march=68060}.
12903 This option inhibits the use of 68020 and 68881/68882 instructions that
12904 have to be emulated by software on the 68060. Use this option if your 68060
12905 does not have code to emulate those instructions.
12909 Generate output for a CPU32. This is the default
12910 when the compiler is configured for CPU32-based systems.
12911 It is equivalent to @option{-march=cpu32}.
12913 Use this option for microcontrollers with a
12914 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
12915 68336, 68340, 68341, 68349 and 68360.
12919 Generate output for a 520X ColdFire CPU@. This is the default
12920 when the compiler is configured for 520X-based systems.
12921 It is equivalent to @option{-mcpu=5206}, and is now deprecated
12922 in favor of that option.
12924 Use this option for microcontroller with a 5200 core, including
12925 the MCF5202, MCF5203, MCF5204 and MCF5206.
12929 Generate output for a 5206e ColdFire CPU@. The option is now
12930 deprecated in favor of the equivalent @option{-mcpu=5206e}.
12934 Generate output for a member of the ColdFire 528X family.
12935 The option is now deprecated in favor of the equivalent
12936 @option{-mcpu=528x}.
12940 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
12941 in favor of the equivalent @option{-mcpu=5307}.
12945 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
12946 in favor of the equivalent @option{-mcpu=5407}.
12950 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
12951 This includes use of hardware floating point instructions.
12952 The option is equivalent to @option{-mcpu=547x}, and is now
12953 deprecated in favor of that option.
12957 Generate output for a 68040, without using any of the new instructions.
12958 This results in code which can run relatively efficiently on either a
12959 68020/68881 or a 68030 or a 68040. The generated code does use the
12960 68881 instructions that are emulated on the 68040.
12962 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
12966 Generate output for a 68060, without using any of the new instructions.
12967 This results in code which can run relatively efficiently on either a
12968 68020/68881 or a 68030 or a 68040. The generated code does use the
12969 68881 instructions that are emulated on the 68060.
12971 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
12975 @opindex mhard-float
12977 Generate floating-point instructions. This is the default for 68020
12978 and above, and for ColdFire devices that have an FPU@. It defines the
12979 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
12980 on ColdFire targets.
12983 @opindex msoft-float
12984 Do not generate floating-point instructions; use library calls instead.
12985 This is the default for 68000, 68010, and 68832 targets. It is also
12986 the default for ColdFire devices that have no FPU.
12992 Generate (do not generate) ColdFire hardware divide and remainder
12993 instructions. If @option{-march} is used without @option{-mcpu},
12994 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
12995 architectures. Otherwise, the default is taken from the target CPU
12996 (either the default CPU, or the one specified by @option{-mcpu}). For
12997 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
12998 @option{-mcpu=5206e}.
13000 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
13004 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13005 Additionally, parameters passed on the stack are also aligned to a
13006 16-bit boundary even on targets whose API mandates promotion to 32-bit.
13010 Do not consider type @code{int} to be 16 bits wide. This is the default.
13013 @itemx -mno-bitfield
13014 @opindex mnobitfield
13015 @opindex mno-bitfield
13016 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
13017 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
13021 Do use the bit-field instructions. The @option{-m68020} option implies
13022 @option{-mbitfield}. This is the default if you use a configuration
13023 designed for a 68020.
13027 Use a different function-calling convention, in which functions
13028 that take a fixed number of arguments return with the @code{rtd}
13029 instruction, which pops their arguments while returning. This
13030 saves one instruction in the caller since there is no need to pop
13031 the arguments there.
13033 This calling convention is incompatible with the one normally
13034 used on Unix, so you cannot use it if you need to call libraries
13035 compiled with the Unix compiler.
13037 Also, you must provide function prototypes for all functions that
13038 take variable numbers of arguments (including @code{printf});
13039 otherwise incorrect code will be generated for calls to those
13042 In addition, seriously incorrect code will result if you call a
13043 function with too many arguments. (Normally, extra arguments are
13044 harmlessly ignored.)
13046 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
13047 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
13051 Do not use the calling conventions selected by @option{-mrtd}.
13052 This is the default.
13055 @itemx -mno-align-int
13056 @opindex malign-int
13057 @opindex mno-align-int
13058 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
13059 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
13060 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
13061 Aligning variables on 32-bit boundaries produces code that runs somewhat
13062 faster on processors with 32-bit busses at the expense of more memory.
13064 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
13065 align structures containing the above types differently than
13066 most published application binary interface specifications for the m68k.
13070 Use the pc-relative addressing mode of the 68000 directly, instead of
13071 using a global offset table. At present, this option implies @option{-fpic},
13072 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
13073 not presently supported with @option{-mpcrel}, though this could be supported for
13074 68020 and higher processors.
13076 @item -mno-strict-align
13077 @itemx -mstrict-align
13078 @opindex mno-strict-align
13079 @opindex mstrict-align
13080 Do not (do) assume that unaligned memory references will be handled by
13084 Generate code that allows the data segment to be located in a different
13085 area of memory from the text segment. This allows for execute in place in
13086 an environment without virtual memory management. This option implies
13089 @item -mno-sep-data
13090 Generate code that assumes that the data segment follows the text segment.
13091 This is the default.
13093 @item -mid-shared-library
13094 Generate code that supports shared libraries via the library ID method.
13095 This allows for execute in place and shared libraries in an environment
13096 without virtual memory management. This option implies @option{-fPIC}.
13098 @item -mno-id-shared-library
13099 Generate code that doesn't assume ID based shared libraries are being used.
13100 This is the default.
13102 @item -mshared-library-id=n
13103 Specified the identification number of the ID based shared library being
13104 compiled. Specifying a value of 0 will generate more compact code, specifying
13105 other values will force the allocation of that number to the current
13106 library but is no more space or time efficient than omitting this option.
13112 When generating position-independent code for ColdFire, generate code
13113 that works if the GOT has more than 8192 entries. This code is
13114 larger and slower than code generated without this option. On M680x0
13115 processors, this option is not needed; @option{-fPIC} suffices.
13117 GCC normally uses a single instruction to load values from the GOT@.
13118 While this is relatively efficient, it only works if the GOT
13119 is smaller than about 64k. Anything larger causes the linker
13120 to report an error such as:
13122 @cindex relocation truncated to fit (ColdFire)
13124 relocation truncated to fit: R_68K_GOT16O foobar
13127 If this happens, you should recompile your code with @option{-mxgot}.
13128 It should then work with very large GOTs. However, code generated with
13129 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
13130 the value of a global symbol.
13132 Note that some linkers, including newer versions of the GNU linker,
13133 can create multiple GOTs and sort GOT entries. If you have such a linker,
13134 you should only need to use @option{-mxgot} when compiling a single
13135 object file that accesses more than 8192 GOT entries. Very few do.
13137 These options have no effect unless GCC is generating
13138 position-independent code.
13142 @node M68hc1x Options
13143 @subsection M68hc1x Options
13144 @cindex M68hc1x options
13146 These are the @samp{-m} options defined for the 68hc11 and 68hc12
13147 microcontrollers. The default values for these options depends on
13148 which style of microcontroller was selected when the compiler was configured;
13149 the defaults for the most common choices are given below.
13156 Generate output for a 68HC11. This is the default
13157 when the compiler is configured for 68HC11-based systems.
13163 Generate output for a 68HC12. This is the default
13164 when the compiler is configured for 68HC12-based systems.
13170 Generate output for a 68HCS12.
13172 @item -mauto-incdec
13173 @opindex mauto-incdec
13174 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
13181 Enable the use of 68HC12 min and max instructions.
13184 @itemx -mno-long-calls
13185 @opindex mlong-calls
13186 @opindex mno-long-calls
13187 Treat all calls as being far away (near). If calls are assumed to be
13188 far away, the compiler will use the @code{call} instruction to
13189 call a function and the @code{rtc} instruction for returning.
13193 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13195 @item -msoft-reg-count=@var{count}
13196 @opindex msoft-reg-count
13197 Specify the number of pseudo-soft registers which are used for the
13198 code generation. The maximum number is 32. Using more pseudo-soft
13199 register may or may not result in better code depending on the program.
13200 The default is 4 for 68HC11 and 2 for 68HC12.
13204 @node MCore Options
13205 @subsection MCore Options
13206 @cindex MCore options
13208 These are the @samp{-m} options defined for the Motorola M*Core
13214 @itemx -mno-hardlit
13216 @opindex mno-hardlit
13217 Inline constants into the code stream if it can be done in two
13218 instructions or less.
13224 Use the divide instruction. (Enabled by default).
13226 @item -mrelax-immediate
13227 @itemx -mno-relax-immediate
13228 @opindex mrelax-immediate
13229 @opindex mno-relax-immediate
13230 Allow arbitrary sized immediates in bit operations.
13232 @item -mwide-bitfields
13233 @itemx -mno-wide-bitfields
13234 @opindex mwide-bitfields
13235 @opindex mno-wide-bitfields
13236 Always treat bit-fields as int-sized.
13238 @item -m4byte-functions
13239 @itemx -mno-4byte-functions
13240 @opindex m4byte-functions
13241 @opindex mno-4byte-functions
13242 Force all functions to be aligned to a four byte boundary.
13244 @item -mcallgraph-data
13245 @itemx -mno-callgraph-data
13246 @opindex mcallgraph-data
13247 @opindex mno-callgraph-data
13248 Emit callgraph information.
13251 @itemx -mno-slow-bytes
13252 @opindex mslow-bytes
13253 @opindex mno-slow-bytes
13254 Prefer word access when reading byte quantities.
13256 @item -mlittle-endian
13257 @itemx -mbig-endian
13258 @opindex mlittle-endian
13259 @opindex mbig-endian
13260 Generate code for a little endian target.
13266 Generate code for the 210 processor.
13270 Assume that run-time support has been provided and so omit the
13271 simulator library (@file{libsim.a)} from the linker command line.
13273 @item -mstack-increment=@var{size}
13274 @opindex mstack-increment
13275 Set the maximum amount for a single stack increment operation. Large
13276 values can increase the speed of programs which contain functions
13277 that need a large amount of stack space, but they can also trigger a
13278 segmentation fault if the stack is extended too much. The default
13284 @subsection MeP Options
13285 @cindex MeP options
13291 Enables the @code{abs} instruction, which is the absolute difference
13292 between two registers.
13296 Enables all the optional instructions - average, multiply, divide, bit
13297 operations, leading zero, absolute difference, min/max, clip, and
13303 Enables the @code{ave} instruction, which computes the average of two
13306 @item -mbased=@var{n}
13308 Variables of size @var{n} bytes or smaller will be placed in the
13309 @code{.based} section by default. Based variables use the @code{$tp}
13310 register as a base register, and there is a 128 byte limit to the
13311 @code{.based} section.
13315 Enables the bit operation instructions - bit test (@code{btstm}), set
13316 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
13317 test-and-set (@code{tas}).
13319 @item -mc=@var{name}
13321 Selects which section constant data will be placed in. @var{name} may
13322 be @code{tiny}, @code{near}, or @code{far}.
13326 Enables the @code{clip} instruction. Note that @code{-mclip} is not
13327 useful unless you also provide @code{-mminmax}.
13329 @item -mconfig=@var{name}
13331 Selects one of the build-in core configurations. Each MeP chip has
13332 one or more modules in it; each module has a core CPU and a variety of
13333 coprocessors, optional instructions, and peripherals. The
13334 @code{MeP-Integrator} tool, not part of GCC, provides these
13335 configurations through this option; using this option is the same as
13336 using all the corresponding command line options. The default
13337 configuration is @code{default}.
13341 Enables the coprocessor instructions. By default, this is a 32-bit
13342 coprocessor. Note that the coprocessor is normally enabled via the
13343 @code{-mconfig=} option.
13347 Enables the 32-bit coprocessor's instructions.
13351 Enables the 64-bit coprocessor's instructions.
13355 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
13359 Causes constant variables to be placed in the @code{.near} section.
13363 Enables the @code{div} and @code{divu} instructions.
13367 Generate big-endian code.
13371 Generate little-endian code.
13373 @item -mio-volatile
13374 @opindex mio-volatile
13375 Tells the compiler that any variable marked with the @code{io}
13376 attribute is to be considered volatile.
13380 Causes variables to be assigned to the @code{.far} section by default.
13384 Enables the @code{leadz} (leading zero) instruction.
13388 Causes variables to be assigned to the @code{.near} section by default.
13392 Enables the @code{min} and @code{max} instructions.
13396 Enables the multiplication and multiply-accumulate instructions.
13400 Disables all the optional instructions enabled by @code{-mall-opts}.
13404 Enables the @code{repeat} and @code{erepeat} instructions, used for
13405 low-overhead looping.
13409 Causes all variables to default to the @code{.tiny} section. Note
13410 that there is a 65536 byte limit to this section. Accesses to these
13411 variables use the @code{%gp} base register.
13415 Enables the saturation instructions. Note that the compiler does not
13416 currently generate these itself, but this option is included for
13417 compatibility with other tools, like @code{as}.
13421 Link the SDRAM-based runtime instead of the default ROM-based runtime.
13425 Link the simulator runtime libraries.
13429 Link the simulator runtime libraries, excluding built-in support
13430 for reset and exception vectors and tables.
13434 Causes all functions to default to the @code{.far} section. Without
13435 this option, functions default to the @code{.near} section.
13437 @item -mtiny=@var{n}
13439 Variables that are @var{n} bytes or smaller will be allocated to the
13440 @code{.tiny} section. These variables use the @code{$gp} base
13441 register. The default for this option is 4, but note that there's a
13442 65536 byte limit to the @code{.tiny} section.
13447 @subsection MIPS Options
13448 @cindex MIPS options
13454 Generate big-endian code.
13458 Generate little-endian code. This is the default for @samp{mips*el-*-*}
13461 @item -march=@var{arch}
13463 Generate code that will run on @var{arch}, which can be the name of a
13464 generic MIPS ISA, or the name of a particular processor.
13466 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
13467 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
13468 The processor names are:
13469 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
13470 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
13471 @samp{5kc}, @samp{5kf},
13473 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
13474 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
13475 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
13476 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
13477 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
13478 @samp{loongson2e}, @samp{loongson2f},
13482 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
13483 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
13484 @samp{rm7000}, @samp{rm9000},
13485 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
13488 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
13489 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
13491 The special value @samp{from-abi} selects the
13492 most compatible architecture for the selected ABI (that is,
13493 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
13495 Native Linux/GNU toolchains also support the value @samp{native},
13496 which selects the best architecture option for the host processor.
13497 @option{-march=native} has no effect if GCC does not recognize
13500 In processor names, a final @samp{000} can be abbreviated as @samp{k}
13501 (for example, @samp{-march=r2k}). Prefixes are optional, and
13502 @samp{vr} may be written @samp{r}.
13504 Names of the form @samp{@var{n}f2_1} refer to processors with
13505 FPUs clocked at half the rate of the core, names of the form
13506 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
13507 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
13508 processors with FPUs clocked a ratio of 3:2 with respect to the core.
13509 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
13510 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
13511 accepted as synonyms for @samp{@var{n}f1_1}.
13513 GCC defines two macros based on the value of this option. The first
13514 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
13515 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
13516 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
13517 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
13518 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
13520 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
13521 above. In other words, it will have the full prefix and will not
13522 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
13523 the macro names the resolved architecture (either @samp{"mips1"} or
13524 @samp{"mips3"}). It names the default architecture when no
13525 @option{-march} option is given.
13527 @item -mtune=@var{arch}
13529 Optimize for @var{arch}. Among other things, this option controls
13530 the way instructions are scheduled, and the perceived cost of arithmetic
13531 operations. The list of @var{arch} values is the same as for
13534 When this option is not used, GCC will optimize for the processor
13535 specified by @option{-march}. By using @option{-march} and
13536 @option{-mtune} together, it is possible to generate code that will
13537 run on a family of processors, but optimize the code for one
13538 particular member of that family.
13540 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
13541 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
13542 @samp{-march} ones described above.
13546 Equivalent to @samp{-march=mips1}.
13550 Equivalent to @samp{-march=mips2}.
13554 Equivalent to @samp{-march=mips3}.
13558 Equivalent to @samp{-march=mips4}.
13562 Equivalent to @samp{-march=mips32}.
13566 Equivalent to @samp{-march=mips32r2}.
13570 Equivalent to @samp{-march=mips64}.
13574 Equivalent to @samp{-march=mips64r2}.
13579 @opindex mno-mips16
13580 Generate (do not generate) MIPS16 code. If GCC is targetting a
13581 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
13583 MIPS16 code generation can also be controlled on a per-function basis
13584 by means of @code{mips16} and @code{nomips16} attributes.
13585 @xref{Function Attributes}, for more information.
13587 @item -mflip-mips16
13588 @opindex mflip-mips16
13589 Generate MIPS16 code on alternating functions. This option is provided
13590 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
13591 not intended for ordinary use in compiling user code.
13593 @item -minterlink-mips16
13594 @itemx -mno-interlink-mips16
13595 @opindex minterlink-mips16
13596 @opindex mno-interlink-mips16
13597 Require (do not require) that non-MIPS16 code be link-compatible with
13600 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
13601 it must either use a call or an indirect jump. @option{-minterlink-mips16}
13602 therefore disables direct jumps unless GCC knows that the target of the
13603 jump is not MIPS16.
13615 Generate code for the given ABI@.
13617 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
13618 generates 64-bit code when you select a 64-bit architecture, but you
13619 can use @option{-mgp32} to get 32-bit code instead.
13621 For information about the O64 ABI, see
13622 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
13624 GCC supports a variant of the o32 ABI in which floating-point registers
13625 are 64 rather than 32 bits wide. You can select this combination with
13626 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
13627 and @samp{mfhc1} instructions and is therefore only supported for
13628 MIPS32R2 processors.
13630 The register assignments for arguments and return values remain the
13631 same, but each scalar value is passed in a single 64-bit register
13632 rather than a pair of 32-bit registers. For example, scalar
13633 floating-point values are returned in @samp{$f0} only, not a
13634 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
13635 remains the same, but all 64 bits are saved.
13638 @itemx -mno-abicalls
13640 @opindex mno-abicalls
13641 Generate (do not generate) code that is suitable for SVR4-style
13642 dynamic objects. @option{-mabicalls} is the default for SVR4-based
13647 Generate (do not generate) code that is fully position-independent,
13648 and that can therefore be linked into shared libraries. This option
13649 only affects @option{-mabicalls}.
13651 All @option{-mabicalls} code has traditionally been position-independent,
13652 regardless of options like @option{-fPIC} and @option{-fpic}. However,
13653 as an extension, the GNU toolchain allows executables to use absolute
13654 accesses for locally-binding symbols. It can also use shorter GP
13655 initialization sequences and generate direct calls to locally-defined
13656 functions. This mode is selected by @option{-mno-shared}.
13658 @option{-mno-shared} depends on binutils 2.16 or higher and generates
13659 objects that can only be linked by the GNU linker. However, the option
13660 does not affect the ABI of the final executable; it only affects the ABI
13661 of relocatable objects. Using @option{-mno-shared} will generally make
13662 executables both smaller and quicker.
13664 @option{-mshared} is the default.
13670 Assume (do not assume) that the static and dynamic linkers
13671 support PLTs and copy relocations. This option only affects
13672 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
13673 has no effect without @samp{-msym32}.
13675 You can make @option{-mplt} the default by configuring
13676 GCC with @option{--with-mips-plt}. The default is
13677 @option{-mno-plt} otherwise.
13683 Lift (do not lift) the usual restrictions on the size of the global
13686 GCC normally uses a single instruction to load values from the GOT@.
13687 While this is relatively efficient, it will only work if the GOT
13688 is smaller than about 64k. Anything larger will cause the linker
13689 to report an error such as:
13691 @cindex relocation truncated to fit (MIPS)
13693 relocation truncated to fit: R_MIPS_GOT16 foobar
13696 If this happens, you should recompile your code with @option{-mxgot}.
13697 It should then work with very large GOTs, although it will also be
13698 less efficient, since it will take three instructions to fetch the
13699 value of a global symbol.
13701 Note that some linkers can create multiple GOTs. If you have such a
13702 linker, you should only need to use @option{-mxgot} when a single object
13703 file accesses more than 64k's worth of GOT entries. Very few do.
13705 These options have no effect unless GCC is generating position
13710 Assume that general-purpose registers are 32 bits wide.
13714 Assume that general-purpose registers are 64 bits wide.
13718 Assume that floating-point registers are 32 bits wide.
13722 Assume that floating-point registers are 64 bits wide.
13725 @opindex mhard-float
13726 Use floating-point coprocessor instructions.
13729 @opindex msoft-float
13730 Do not use floating-point coprocessor instructions. Implement
13731 floating-point calculations using library calls instead.
13733 @item -msingle-float
13734 @opindex msingle-float
13735 Assume that the floating-point coprocessor only supports single-precision
13738 @item -mdouble-float
13739 @opindex mdouble-float
13740 Assume that the floating-point coprocessor supports double-precision
13741 operations. This is the default.
13747 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
13748 implement atomic memory built-in functions. When neither option is
13749 specified, GCC will use the instructions if the target architecture
13752 @option{-mllsc} is useful if the runtime environment can emulate the
13753 instructions and @option{-mno-llsc} can be useful when compiling for
13754 nonstandard ISAs. You can make either option the default by
13755 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
13756 respectively. @option{--with-llsc} is the default for some
13757 configurations; see the installation documentation for details.
13763 Use (do not use) revision 1 of the MIPS DSP ASE@.
13764 @xref{MIPS DSP Built-in Functions}. This option defines the
13765 preprocessor macro @samp{__mips_dsp}. It also defines
13766 @samp{__mips_dsp_rev} to 1.
13772 Use (do not use) revision 2 of the MIPS DSP ASE@.
13773 @xref{MIPS DSP Built-in Functions}. This option defines the
13774 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
13775 It also defines @samp{__mips_dsp_rev} to 2.
13778 @itemx -mno-smartmips
13779 @opindex msmartmips
13780 @opindex mno-smartmips
13781 Use (do not use) the MIPS SmartMIPS ASE.
13783 @item -mpaired-single
13784 @itemx -mno-paired-single
13785 @opindex mpaired-single
13786 @opindex mno-paired-single
13787 Use (do not use) paired-single floating-point instructions.
13788 @xref{MIPS Paired-Single Support}. This option requires
13789 hardware floating-point support to be enabled.
13795 Use (do not use) MIPS Digital Media Extension instructions.
13796 This option can only be used when generating 64-bit code and requires
13797 hardware floating-point support to be enabled.
13802 @opindex mno-mips3d
13803 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
13804 The option @option{-mips3d} implies @option{-mpaired-single}.
13810 Use (do not use) MT Multithreading instructions.
13814 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
13815 an explanation of the default and the way that the pointer size is
13820 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
13822 The default size of @code{int}s, @code{long}s and pointers depends on
13823 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
13824 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
13825 32-bit @code{long}s. Pointers are the same size as @code{long}s,
13826 or the same size as integer registers, whichever is smaller.
13832 Assume (do not assume) that all symbols have 32-bit values, regardless
13833 of the selected ABI@. This option is useful in combination with
13834 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
13835 to generate shorter and faster references to symbolic addresses.
13839 Put definitions of externally-visible data in a small data section
13840 if that data is no bigger than @var{num} bytes. GCC can then access
13841 the data more efficiently; see @option{-mgpopt} for details.
13843 The default @option{-G} option depends on the configuration.
13845 @item -mlocal-sdata
13846 @itemx -mno-local-sdata
13847 @opindex mlocal-sdata
13848 @opindex mno-local-sdata
13849 Extend (do not extend) the @option{-G} behavior to local data too,
13850 such as to static variables in C@. @option{-mlocal-sdata} is the
13851 default for all configurations.
13853 If the linker complains that an application is using too much small data,
13854 you might want to try rebuilding the less performance-critical parts with
13855 @option{-mno-local-sdata}. You might also want to build large
13856 libraries with @option{-mno-local-sdata}, so that the libraries leave
13857 more room for the main program.
13859 @item -mextern-sdata
13860 @itemx -mno-extern-sdata
13861 @opindex mextern-sdata
13862 @opindex mno-extern-sdata
13863 Assume (do not assume) that externally-defined data will be in
13864 a small data section if that data is within the @option{-G} limit.
13865 @option{-mextern-sdata} is the default for all configurations.
13867 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
13868 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
13869 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
13870 is placed in a small data section. If @var{Var} is defined by another
13871 module, you must either compile that module with a high-enough
13872 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
13873 definition. If @var{Var} is common, you must link the application
13874 with a high-enough @option{-G} setting.
13876 The easiest way of satisfying these restrictions is to compile
13877 and link every module with the same @option{-G} option. However,
13878 you may wish to build a library that supports several different
13879 small data limits. You can do this by compiling the library with
13880 the highest supported @option{-G} setting and additionally using
13881 @option{-mno-extern-sdata} to stop the library from making assumptions
13882 about externally-defined data.
13888 Use (do not use) GP-relative accesses for symbols that are known to be
13889 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
13890 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
13893 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
13894 might not hold the value of @code{_gp}. For example, if the code is
13895 part of a library that might be used in a boot monitor, programs that
13896 call boot monitor routines will pass an unknown value in @code{$gp}.
13897 (In such situations, the boot monitor itself would usually be compiled
13898 with @option{-G0}.)
13900 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
13901 @option{-mno-extern-sdata}.
13903 @item -membedded-data
13904 @itemx -mno-embedded-data
13905 @opindex membedded-data
13906 @opindex mno-embedded-data
13907 Allocate variables to the read-only data section first if possible, then
13908 next in the small data section if possible, otherwise in data. This gives
13909 slightly slower code than the default, but reduces the amount of RAM required
13910 when executing, and thus may be preferred for some embedded systems.
13912 @item -muninit-const-in-rodata
13913 @itemx -mno-uninit-const-in-rodata
13914 @opindex muninit-const-in-rodata
13915 @opindex mno-uninit-const-in-rodata
13916 Put uninitialized @code{const} variables in the read-only data section.
13917 This option is only meaningful in conjunction with @option{-membedded-data}.
13919 @item -mcode-readable=@var{setting}
13920 @opindex mcode-readable
13921 Specify whether GCC may generate code that reads from executable sections.
13922 There are three possible settings:
13925 @item -mcode-readable=yes
13926 Instructions may freely access executable sections. This is the
13929 @item -mcode-readable=pcrel
13930 MIPS16 PC-relative load instructions can access executable sections,
13931 but other instructions must not do so. This option is useful on 4KSc
13932 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
13933 It is also useful on processors that can be configured to have a dual
13934 instruction/data SRAM interface and that, like the M4K, automatically
13935 redirect PC-relative loads to the instruction RAM.
13937 @item -mcode-readable=no
13938 Instructions must not access executable sections. This option can be
13939 useful on targets that are configured to have a dual instruction/data
13940 SRAM interface but that (unlike the M4K) do not automatically redirect
13941 PC-relative loads to the instruction RAM.
13944 @item -msplit-addresses
13945 @itemx -mno-split-addresses
13946 @opindex msplit-addresses
13947 @opindex mno-split-addresses
13948 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
13949 relocation operators. This option has been superseded by
13950 @option{-mexplicit-relocs} but is retained for backwards compatibility.
13952 @item -mexplicit-relocs
13953 @itemx -mno-explicit-relocs
13954 @opindex mexplicit-relocs
13955 @opindex mno-explicit-relocs
13956 Use (do not use) assembler relocation operators when dealing with symbolic
13957 addresses. The alternative, selected by @option{-mno-explicit-relocs},
13958 is to use assembler macros instead.
13960 @option{-mexplicit-relocs} is the default if GCC was configured
13961 to use an assembler that supports relocation operators.
13963 @item -mcheck-zero-division
13964 @itemx -mno-check-zero-division
13965 @opindex mcheck-zero-division
13966 @opindex mno-check-zero-division
13967 Trap (do not trap) on integer division by zero.
13969 The default is @option{-mcheck-zero-division}.
13971 @item -mdivide-traps
13972 @itemx -mdivide-breaks
13973 @opindex mdivide-traps
13974 @opindex mdivide-breaks
13975 MIPS systems check for division by zero by generating either a
13976 conditional trap or a break instruction. Using traps results in
13977 smaller code, but is only supported on MIPS II and later. Also, some
13978 versions of the Linux kernel have a bug that prevents trap from
13979 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
13980 allow conditional traps on architectures that support them and
13981 @option{-mdivide-breaks} to force the use of breaks.
13983 The default is usually @option{-mdivide-traps}, but this can be
13984 overridden at configure time using @option{--with-divide=breaks}.
13985 Divide-by-zero checks can be completely disabled using
13986 @option{-mno-check-zero-division}.
13991 @opindex mno-memcpy
13992 Force (do not force) the use of @code{memcpy()} for non-trivial block
13993 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
13994 most constant-sized copies.
13997 @itemx -mno-long-calls
13998 @opindex mlong-calls
13999 @opindex mno-long-calls
14000 Disable (do not disable) use of the @code{jal} instruction. Calling
14001 functions using @code{jal} is more efficient but requires the caller
14002 and callee to be in the same 256 megabyte segment.
14004 This option has no effect on abicalls code. The default is
14005 @option{-mno-long-calls}.
14011 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
14012 instructions, as provided by the R4650 ISA@.
14015 @itemx -mno-fused-madd
14016 @opindex mfused-madd
14017 @opindex mno-fused-madd
14018 Enable (disable) use of the floating point multiply-accumulate
14019 instructions, when they are available. The default is
14020 @option{-mfused-madd}.
14022 When multiply-accumulate instructions are used, the intermediate
14023 product is calculated to infinite precision and is not subject to
14024 the FCSR Flush to Zero bit. This may be undesirable in some
14029 Tell the MIPS assembler to not run its preprocessor over user
14030 assembler files (with a @samp{.s} suffix) when assembling them.
14033 @itemx -mno-fix-r4000
14034 @opindex mfix-r4000
14035 @opindex mno-fix-r4000
14036 Work around certain R4000 CPU errata:
14039 A double-word or a variable shift may give an incorrect result if executed
14040 immediately after starting an integer division.
14042 A double-word or a variable shift may give an incorrect result if executed
14043 while an integer multiplication is in progress.
14045 An integer division may give an incorrect result if started in a delay slot
14046 of a taken branch or a jump.
14050 @itemx -mno-fix-r4400
14051 @opindex mfix-r4400
14052 @opindex mno-fix-r4400
14053 Work around certain R4400 CPU errata:
14056 A double-word or a variable shift may give an incorrect result if executed
14057 immediately after starting an integer division.
14061 @itemx -mno-fix-r10000
14062 @opindex mfix-r10000
14063 @opindex mno-fix-r10000
14064 Work around certain R10000 errata:
14067 @code{ll}/@code{sc} sequences may not behave atomically on revisions
14068 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
14071 This option can only be used if the target architecture supports
14072 branch-likely instructions. @option{-mfix-r10000} is the default when
14073 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
14077 @itemx -mno-fix-vr4120
14078 @opindex mfix-vr4120
14079 Work around certain VR4120 errata:
14082 @code{dmultu} does not always produce the correct result.
14084 @code{div} and @code{ddiv} do not always produce the correct result if one
14085 of the operands is negative.
14087 The workarounds for the division errata rely on special functions in
14088 @file{libgcc.a}. At present, these functions are only provided by
14089 the @code{mips64vr*-elf} configurations.
14091 Other VR4120 errata require a nop to be inserted between certain pairs of
14092 instructions. These errata are handled by the assembler, not by GCC itself.
14095 @opindex mfix-vr4130
14096 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
14097 workarounds are implemented by the assembler rather than by GCC,
14098 although GCC will avoid using @code{mflo} and @code{mfhi} if the
14099 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
14100 instructions are available instead.
14103 @itemx -mno-fix-sb1
14105 Work around certain SB-1 CPU core errata.
14106 (This flag currently works around the SB-1 revision 2
14107 ``F1'' and ``F2'' floating point errata.)
14109 @item -mr10k-cache-barrier=@var{setting}
14110 @opindex mr10k-cache-barrier
14111 Specify whether GCC should insert cache barriers to avoid the
14112 side-effects of speculation on R10K processors.
14114 In common with many processors, the R10K tries to predict the outcome
14115 of a conditional branch and speculatively executes instructions from
14116 the ``taken'' branch. It later aborts these instructions if the
14117 predicted outcome was wrong. However, on the R10K, even aborted
14118 instructions can have side effects.
14120 This problem only affects kernel stores and, depending on the system,
14121 kernel loads. As an example, a speculatively-executed store may load
14122 the target memory into cache and mark the cache line as dirty, even if
14123 the store itself is later aborted. If a DMA operation writes to the
14124 same area of memory before the ``dirty'' line is flushed, the cached
14125 data will overwrite the DMA-ed data. See the R10K processor manual
14126 for a full description, including other potential problems.
14128 One workaround is to insert cache barrier instructions before every memory
14129 access that might be speculatively executed and that might have side
14130 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
14131 controls GCC's implementation of this workaround. It assumes that
14132 aborted accesses to any byte in the following regions will not have
14137 the memory occupied by the current function's stack frame;
14140 the memory occupied by an incoming stack argument;
14143 the memory occupied by an object with a link-time-constant address.
14146 It is the kernel's responsibility to ensure that speculative
14147 accesses to these regions are indeed safe.
14149 If the input program contains a function declaration such as:
14155 then the implementation of @code{foo} must allow @code{j foo} and
14156 @code{jal foo} to be executed speculatively. GCC honors this
14157 restriction for functions it compiles itself. It expects non-GCC
14158 functions (such as hand-written assembly code) to do the same.
14160 The option has three forms:
14163 @item -mr10k-cache-barrier=load-store
14164 Insert a cache barrier before a load or store that might be
14165 speculatively executed and that might have side effects even
14168 @item -mr10k-cache-barrier=store
14169 Insert a cache barrier before a store that might be speculatively
14170 executed and that might have side effects even if aborted.
14172 @item -mr10k-cache-barrier=none
14173 Disable the insertion of cache barriers. This is the default setting.
14176 @item -mflush-func=@var{func}
14177 @itemx -mno-flush-func
14178 @opindex mflush-func
14179 Specifies the function to call to flush the I and D caches, or to not
14180 call any such function. If called, the function must take the same
14181 arguments as the common @code{_flush_func()}, that is, the address of the
14182 memory range for which the cache is being flushed, the size of the
14183 memory range, and the number 3 (to flush both caches). The default
14184 depends on the target GCC was configured for, but commonly is either
14185 @samp{_flush_func} or @samp{__cpu_flush}.
14187 @item mbranch-cost=@var{num}
14188 @opindex mbranch-cost
14189 Set the cost of branches to roughly @var{num} ``simple'' instructions.
14190 This cost is only a heuristic and is not guaranteed to produce
14191 consistent results across releases. A zero cost redundantly selects
14192 the default, which is based on the @option{-mtune} setting.
14194 @item -mbranch-likely
14195 @itemx -mno-branch-likely
14196 @opindex mbranch-likely
14197 @opindex mno-branch-likely
14198 Enable or disable use of Branch Likely instructions, regardless of the
14199 default for the selected architecture. By default, Branch Likely
14200 instructions may be generated if they are supported by the selected
14201 architecture. An exception is for the MIPS32 and MIPS64 architectures
14202 and processors which implement those architectures; for those, Branch
14203 Likely instructions will not be generated by default because the MIPS32
14204 and MIPS64 architectures specifically deprecate their use.
14206 @item -mfp-exceptions
14207 @itemx -mno-fp-exceptions
14208 @opindex mfp-exceptions
14209 Specifies whether FP exceptions are enabled. This affects how we schedule
14210 FP instructions for some processors. The default is that FP exceptions are
14213 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
14214 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
14217 @item -mvr4130-align
14218 @itemx -mno-vr4130-align
14219 @opindex mvr4130-align
14220 The VR4130 pipeline is two-way superscalar, but can only issue two
14221 instructions together if the first one is 8-byte aligned. When this
14222 option is enabled, GCC will align pairs of instructions that it
14223 thinks should execute in parallel.
14225 This option only has an effect when optimizing for the VR4130.
14226 It normally makes code faster, but at the expense of making it bigger.
14227 It is enabled by default at optimization level @option{-O3}.
14232 Enable (disable) generation of @code{synci} instructions on
14233 architectures that support it. The @code{synci} instructions (if
14234 enabled) will be generated when @code{__builtin___clear_cache()} is
14237 This option defaults to @code{-mno-synci}, but the default can be
14238 overridden by configuring with @code{--with-synci}.
14240 When compiling code for single processor systems, it is generally safe
14241 to use @code{synci}. However, on many multi-core (SMP) systems, it
14242 will not invalidate the instruction caches on all cores and may lead
14243 to undefined behavior.
14245 @item -mrelax-pic-calls
14246 @itemx -mno-relax-pic-calls
14247 @opindex mrelax-pic-calls
14248 Try to turn PIC calls that are normally dispatched via register
14249 @code{$25} into direct calls. This is only possible if the linker can
14250 resolve the destination at link-time and if the destination is within
14251 range for a direct call.
14253 @option{-mrelax-pic-calls} is the default if GCC was configured to use
14254 an assembler and a linker that supports the @code{.reloc} assembly
14255 directive and @code{-mexplicit-relocs} is in effect. With
14256 @code{-mno-explicit-relocs}, this optimization can be performed by the
14257 assembler and the linker alone without help from the compiler.
14259 @item -mmcount-ra-address
14260 @itemx -mno-mcount-ra-address
14261 @opindex mmcount-ra-address
14262 @opindex mno-mcount-ra-address
14263 Emit (do not emit) code that allows @code{_mcount} to modify the
14264 calling function's return address. When enabled, this option extends
14265 the usual @code{_mcount} interface with a new @var{ra-address}
14266 parameter, which has type @code{intptr_t *} and is passed in register
14267 @code{$12}. @code{_mcount} can then modify the return address by
14268 doing both of the following:
14271 Returning the new address in register @code{$31}.
14273 Storing the new address in @code{*@var{ra-address}},
14274 if @var{ra-address} is nonnull.
14277 The default is @option{-mno-mcount-ra-address}.
14282 @subsection MMIX Options
14283 @cindex MMIX Options
14285 These options are defined for the MMIX:
14289 @itemx -mno-libfuncs
14291 @opindex mno-libfuncs
14292 Specify that intrinsic library functions are being compiled, passing all
14293 values in registers, no matter the size.
14296 @itemx -mno-epsilon
14298 @opindex mno-epsilon
14299 Generate floating-point comparison instructions that compare with respect
14300 to the @code{rE} epsilon register.
14302 @item -mabi=mmixware
14304 @opindex mabi=mmixware
14306 Generate code that passes function parameters and return values that (in
14307 the called function) are seen as registers @code{$0} and up, as opposed to
14308 the GNU ABI which uses global registers @code{$231} and up.
14310 @item -mzero-extend
14311 @itemx -mno-zero-extend
14312 @opindex mzero-extend
14313 @opindex mno-zero-extend
14314 When reading data from memory in sizes shorter than 64 bits, use (do not
14315 use) zero-extending load instructions by default, rather than
14316 sign-extending ones.
14319 @itemx -mno-knuthdiv
14321 @opindex mno-knuthdiv
14322 Make the result of a division yielding a remainder have the same sign as
14323 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
14324 remainder follows the sign of the dividend. Both methods are
14325 arithmetically valid, the latter being almost exclusively used.
14327 @item -mtoplevel-symbols
14328 @itemx -mno-toplevel-symbols
14329 @opindex mtoplevel-symbols
14330 @opindex mno-toplevel-symbols
14331 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
14332 code can be used with the @code{PREFIX} assembly directive.
14336 Generate an executable in the ELF format, rather than the default
14337 @samp{mmo} format used by the @command{mmix} simulator.
14339 @item -mbranch-predict
14340 @itemx -mno-branch-predict
14341 @opindex mbranch-predict
14342 @opindex mno-branch-predict
14343 Use (do not use) the probable-branch instructions, when static branch
14344 prediction indicates a probable branch.
14346 @item -mbase-addresses
14347 @itemx -mno-base-addresses
14348 @opindex mbase-addresses
14349 @opindex mno-base-addresses
14350 Generate (do not generate) code that uses @emph{base addresses}. Using a
14351 base address automatically generates a request (handled by the assembler
14352 and the linker) for a constant to be set up in a global register. The
14353 register is used for one or more base address requests within the range 0
14354 to 255 from the value held in the register. The generally leads to short
14355 and fast code, but the number of different data items that can be
14356 addressed is limited. This means that a program that uses lots of static
14357 data may require @option{-mno-base-addresses}.
14359 @item -msingle-exit
14360 @itemx -mno-single-exit
14361 @opindex msingle-exit
14362 @opindex mno-single-exit
14363 Force (do not force) generated code to have a single exit point in each
14367 @node MN10300 Options
14368 @subsection MN10300 Options
14369 @cindex MN10300 options
14371 These @option{-m} options are defined for Matsushita MN10300 architectures:
14376 Generate code to avoid bugs in the multiply instructions for the MN10300
14377 processors. This is the default.
14379 @item -mno-mult-bug
14380 @opindex mno-mult-bug
14381 Do not generate code to avoid bugs in the multiply instructions for the
14382 MN10300 processors.
14386 Generate code which uses features specific to the AM33 processor.
14390 Do not generate code which uses features specific to the AM33 processor. This
14393 @item -mreturn-pointer-on-d0
14394 @opindex mreturn-pointer-on-d0
14395 When generating a function which returns a pointer, return the pointer
14396 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
14397 only in a0, and attempts to call such functions without a prototype
14398 would result in errors. Note that this option is on by default; use
14399 @option{-mno-return-pointer-on-d0} to disable it.
14403 Do not link in the C run-time initialization object file.
14407 Indicate to the linker that it should perform a relaxation optimization pass
14408 to shorten branches, calls and absolute memory addresses. This option only
14409 has an effect when used on the command line for the final link step.
14411 This option makes symbolic debugging impossible.
14414 @node PDP-11 Options
14415 @subsection PDP-11 Options
14416 @cindex PDP-11 Options
14418 These options are defined for the PDP-11:
14423 Use hardware FPP floating point. This is the default. (FIS floating
14424 point on the PDP-11/40 is not supported.)
14427 @opindex msoft-float
14428 Do not use hardware floating point.
14432 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
14436 Return floating-point results in memory. This is the default.
14440 Generate code for a PDP-11/40.
14444 Generate code for a PDP-11/45. This is the default.
14448 Generate code for a PDP-11/10.
14450 @item -mbcopy-builtin
14451 @opindex mbcopy-builtin
14452 Use inline @code{movmemhi} patterns for copying memory. This is the
14457 Do not use inline @code{movmemhi} patterns for copying memory.
14463 Use 16-bit @code{int}. This is the default.
14469 Use 32-bit @code{int}.
14472 @itemx -mno-float32
14474 @opindex mno-float32
14475 Use 64-bit @code{float}. This is the default.
14478 @itemx -mno-float64
14480 @opindex mno-float64
14481 Use 32-bit @code{float}.
14485 Use @code{abshi2} pattern. This is the default.
14489 Do not use @code{abshi2} pattern.
14491 @item -mbranch-expensive
14492 @opindex mbranch-expensive
14493 Pretend that branches are expensive. This is for experimenting with
14494 code generation only.
14496 @item -mbranch-cheap
14497 @opindex mbranch-cheap
14498 Do not pretend that branches are expensive. This is the default.
14502 Generate code for a system with split I&D@.
14506 Generate code for a system without split I&D@. This is the default.
14510 Use Unix assembler syntax. This is the default when configured for
14511 @samp{pdp11-*-bsd}.
14515 Use DEC assembler syntax. This is the default when configured for any
14516 PDP-11 target other than @samp{pdp11-*-bsd}.
14519 @node picoChip Options
14520 @subsection picoChip Options
14521 @cindex picoChip options
14523 These @samp{-m} options are defined for picoChip implementations:
14527 @item -mae=@var{ae_type}
14529 Set the instruction set, register set, and instruction scheduling
14530 parameters for array element type @var{ae_type}. Supported values
14531 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
14533 @option{-mae=ANY} selects a completely generic AE type. Code
14534 generated with this option will run on any of the other AE types. The
14535 code will not be as efficient as it would be if compiled for a specific
14536 AE type, and some types of operation (e.g., multiplication) will not
14537 work properly on all types of AE.
14539 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
14540 for compiled code, and is the default.
14542 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
14543 option may suffer from poor performance of byte (char) manipulation,
14544 since the DSP AE does not provide hardware support for byte load/stores.
14546 @item -msymbol-as-address
14547 Enable the compiler to directly use a symbol name as an address in a
14548 load/store instruction, without first loading it into a
14549 register. Typically, the use of this option will generate larger
14550 programs, which run faster than when the option isn't used. However, the
14551 results vary from program to program, so it is left as a user option,
14552 rather than being permanently enabled.
14554 @item -mno-inefficient-warnings
14555 Disables warnings about the generation of inefficient code. These
14556 warnings can be generated, for example, when compiling code which
14557 performs byte-level memory operations on the MAC AE type. The MAC AE has
14558 no hardware support for byte-level memory operations, so all byte
14559 load/stores must be synthesized from word load/store operations. This is
14560 inefficient and a warning will be generated indicating to the programmer
14561 that they should rewrite the code to avoid byte operations, or to target
14562 an AE type which has the necessary hardware support. This option enables
14563 the warning to be turned off.
14567 @node PowerPC Options
14568 @subsection PowerPC Options
14569 @cindex PowerPC options
14571 These are listed under @xref{RS/6000 and PowerPC Options}.
14573 @node RS/6000 and PowerPC Options
14574 @subsection IBM RS/6000 and PowerPC Options
14575 @cindex RS/6000 and PowerPC Options
14576 @cindex IBM RS/6000 and PowerPC Options
14578 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
14585 @itemx -mno-powerpc
14586 @itemx -mpowerpc-gpopt
14587 @itemx -mno-powerpc-gpopt
14588 @itemx -mpowerpc-gfxopt
14589 @itemx -mno-powerpc-gfxopt
14591 @itemx -mno-powerpc64
14595 @itemx -mno-popcntb
14597 @itemx -mno-popcntd
14605 @itemx -mno-hard-dfp
14609 @opindex mno-power2
14611 @opindex mno-powerpc
14612 @opindex mpowerpc-gpopt
14613 @opindex mno-powerpc-gpopt
14614 @opindex mpowerpc-gfxopt
14615 @opindex mno-powerpc-gfxopt
14616 @opindex mpowerpc64
14617 @opindex mno-powerpc64
14621 @opindex mno-popcntb
14623 @opindex mno-popcntd
14629 @opindex mno-mfpgpr
14631 @opindex mno-hard-dfp
14632 GCC supports two related instruction set architectures for the
14633 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
14634 instructions supported by the @samp{rios} chip set used in the original
14635 RS/6000 systems and the @dfn{PowerPC} instruction set is the
14636 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
14637 the IBM 4xx, 6xx, and follow-on microprocessors.
14639 Neither architecture is a subset of the other. However there is a
14640 large common subset of instructions supported by both. An MQ
14641 register is included in processors supporting the POWER architecture.
14643 You use these options to specify which instructions are available on the
14644 processor you are using. The default value of these options is
14645 determined when configuring GCC@. Specifying the
14646 @option{-mcpu=@var{cpu_type}} overrides the specification of these
14647 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
14648 rather than the options listed above.
14650 The @option{-mpower} option allows GCC to generate instructions that
14651 are found only in the POWER architecture and to use the MQ register.
14652 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
14653 to generate instructions that are present in the POWER2 architecture but
14654 not the original POWER architecture.
14656 The @option{-mpowerpc} option allows GCC to generate instructions that
14657 are found only in the 32-bit subset of the PowerPC architecture.
14658 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
14659 GCC to use the optional PowerPC architecture instructions in the
14660 General Purpose group, including floating-point square root. Specifying
14661 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
14662 use the optional PowerPC architecture instructions in the Graphics
14663 group, including floating-point select.
14665 The @option{-mmfcrf} option allows GCC to generate the move from
14666 condition register field instruction implemented on the POWER4
14667 processor and other processors that support the PowerPC V2.01
14669 The @option{-mpopcntb} option allows GCC to generate the popcount and
14670 double precision FP reciprocal estimate instruction implemented on the
14671 POWER5 processor and other processors that support the PowerPC V2.02
14673 The @option{-mpopcntd} option allows GCC to generate the popcount
14674 instruction implemented on the POWER7 processor and other processors
14675 that support the PowerPC V2.06 architecture.
14676 The @option{-mfprnd} option allows GCC to generate the FP round to
14677 integer instructions implemented on the POWER5+ processor and other
14678 processors that support the PowerPC V2.03 architecture.
14679 The @option{-mcmpb} option allows GCC to generate the compare bytes
14680 instruction implemented on the POWER6 processor and other processors
14681 that support the PowerPC V2.05 architecture.
14682 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
14683 general purpose register instructions implemented on the POWER6X
14684 processor and other processors that support the extended PowerPC V2.05
14686 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
14687 point instructions implemented on some POWER processors.
14689 The @option{-mpowerpc64} option allows GCC to generate the additional
14690 64-bit instructions that are found in the full PowerPC64 architecture
14691 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
14692 @option{-mno-powerpc64}.
14694 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
14695 will use only the instructions in the common subset of both
14696 architectures plus some special AIX common-mode calls, and will not use
14697 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
14698 permits GCC to use any instruction from either architecture and to
14699 allow use of the MQ register; specify this for the Motorola MPC601.
14701 @item -mnew-mnemonics
14702 @itemx -mold-mnemonics
14703 @opindex mnew-mnemonics
14704 @opindex mold-mnemonics
14705 Select which mnemonics to use in the generated assembler code. With
14706 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
14707 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
14708 assembler mnemonics defined for the POWER architecture. Instructions
14709 defined in only one architecture have only one mnemonic; GCC uses that
14710 mnemonic irrespective of which of these options is specified.
14712 GCC defaults to the mnemonics appropriate for the architecture in
14713 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
14714 value of these option. Unless you are building a cross-compiler, you
14715 should normally not specify either @option{-mnew-mnemonics} or
14716 @option{-mold-mnemonics}, but should instead accept the default.
14718 @item -mcpu=@var{cpu_type}
14720 Set architecture type, register usage, choice of mnemonics, and
14721 instruction scheduling parameters for machine type @var{cpu_type}.
14722 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
14723 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
14724 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
14725 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
14726 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
14727 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
14728 @samp{e300c3}, @samp{e500mc}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
14729 @samp{power}, @samp{power2}, @samp{power3}, @samp{power4},
14730 @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x}, @samp{power7},
14731 @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
14732 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
14734 @option{-mcpu=common} selects a completely generic processor. Code
14735 generated under this option will run on any POWER or PowerPC processor.
14736 GCC will use only the instructions in the common subset of both
14737 architectures, and will not use the MQ register. GCC assumes a generic
14738 processor model for scheduling purposes.
14740 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
14741 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
14742 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
14743 types, with an appropriate, generic processor model assumed for
14744 scheduling purposes.
14746 The other options specify a specific processor. Code generated under
14747 those options will run best on that processor, and may not run at all on
14750 The @option{-mcpu} options automatically enable or disable the
14753 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
14754 -mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol
14755 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
14756 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx}
14758 The particular options set for any particular CPU will vary between
14759 compiler versions, depending on what setting seems to produce optimal
14760 code for that CPU; it doesn't necessarily reflect the actual hardware's
14761 capabilities. If you wish to set an individual option to a particular
14762 value, you may specify it after the @option{-mcpu} option, like
14763 @samp{-mcpu=970 -mno-altivec}.
14765 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
14766 not enabled or disabled by the @option{-mcpu} option at present because
14767 AIX does not have full support for these options. You may still
14768 enable or disable them individually if you're sure it'll work in your
14771 @item -mtune=@var{cpu_type}
14773 Set the instruction scheduling parameters for machine type
14774 @var{cpu_type}, but do not set the architecture type, register usage, or
14775 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
14776 values for @var{cpu_type} are used for @option{-mtune} as for
14777 @option{-mcpu}. If both are specified, the code generated will use the
14778 architecture, registers, and mnemonics set by @option{-mcpu}, but the
14779 scheduling parameters set by @option{-mtune}.
14785 Generate code to compute division as reciprocal estimate and iterative
14786 refinement, creating opportunities for increased throughput. This
14787 feature requires: optional PowerPC Graphics instruction set for single
14788 precision and FRE instruction for double precision, assuming divides
14789 cannot generate user-visible traps, and the domain values not include
14790 Infinities, denormals or zero denominator.
14793 @itemx -mno-altivec
14795 @opindex mno-altivec
14796 Generate code that uses (does not use) AltiVec instructions, and also
14797 enable the use of built-in functions that allow more direct access to
14798 the AltiVec instruction set. You may also need to set
14799 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
14805 @opindex mno-vrsave
14806 Generate VRSAVE instructions when generating AltiVec code.
14808 @item -mgen-cell-microcode
14809 @opindex mgen-cell-microcode
14810 Generate Cell microcode instructions
14812 @item -mwarn-cell-microcode
14813 @opindex mwarn-cell-microcode
14814 Warning when a Cell microcode instruction is going to emitted. An example
14815 of a Cell microcode instruction is a variable shift.
14818 @opindex msecure-plt
14819 Generate code that allows ld and ld.so to build executables and shared
14820 libraries with non-exec .plt and .got sections. This is a PowerPC
14821 32-bit SYSV ABI option.
14825 Generate code that uses a BSS .plt section that ld.so fills in, and
14826 requires .plt and .got sections that are both writable and executable.
14827 This is a PowerPC 32-bit SYSV ABI option.
14833 This switch enables or disables the generation of ISEL instructions.
14835 @item -misel=@var{yes/no}
14836 This switch has been deprecated. Use @option{-misel} and
14837 @option{-mno-isel} instead.
14843 This switch enables or disables the generation of SPE simd
14849 @opindex mno-paired
14850 This switch enables or disables the generation of PAIRED simd
14853 @item -mspe=@var{yes/no}
14854 This option has been deprecated. Use @option{-mspe} and
14855 @option{-mno-spe} instead.
14861 Generate code that uses (does not use) vector/scalar (VSX)
14862 instructions, and also enable the use of built-in functions that allow
14863 more direct access to the VSX instruction set.
14865 @item -mfloat-gprs=@var{yes/single/double/no}
14866 @itemx -mfloat-gprs
14867 @opindex mfloat-gprs
14868 This switch enables or disables the generation of floating point
14869 operations on the general purpose registers for architectures that
14872 The argument @var{yes} or @var{single} enables the use of
14873 single-precision floating point operations.
14875 The argument @var{double} enables the use of single and
14876 double-precision floating point operations.
14878 The argument @var{no} disables floating point operations on the
14879 general purpose registers.
14881 This option is currently only available on the MPC854x.
14887 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
14888 targets (including GNU/Linux). The 32-bit environment sets int, long
14889 and pointer to 32 bits and generates code that runs on any PowerPC
14890 variant. The 64-bit environment sets int to 32 bits and long and
14891 pointer to 64 bits, and generates code for PowerPC64, as for
14892 @option{-mpowerpc64}.
14895 @itemx -mno-fp-in-toc
14896 @itemx -mno-sum-in-toc
14897 @itemx -mminimal-toc
14899 @opindex mno-fp-in-toc
14900 @opindex mno-sum-in-toc
14901 @opindex mminimal-toc
14902 Modify generation of the TOC (Table Of Contents), which is created for
14903 every executable file. The @option{-mfull-toc} option is selected by
14904 default. In that case, GCC will allocate at least one TOC entry for
14905 each unique non-automatic variable reference in your program. GCC
14906 will also place floating-point constants in the TOC@. However, only
14907 16,384 entries are available in the TOC@.
14909 If you receive a linker error message that saying you have overflowed
14910 the available TOC space, you can reduce the amount of TOC space used
14911 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
14912 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
14913 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
14914 generate code to calculate the sum of an address and a constant at
14915 run-time instead of putting that sum into the TOC@. You may specify one
14916 or both of these options. Each causes GCC to produce very slightly
14917 slower and larger code at the expense of conserving TOC space.
14919 If you still run out of space in the TOC even when you specify both of
14920 these options, specify @option{-mminimal-toc} instead. This option causes
14921 GCC to make only one TOC entry for every file. When you specify this
14922 option, GCC will produce code that is slower and larger but which
14923 uses extremely little TOC space. You may wish to use this option
14924 only on files that contain less frequently executed code.
14930 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
14931 @code{long} type, and the infrastructure needed to support them.
14932 Specifying @option{-maix64} implies @option{-mpowerpc64} and
14933 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
14934 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
14937 @itemx -mno-xl-compat
14938 @opindex mxl-compat
14939 @opindex mno-xl-compat
14940 Produce code that conforms more closely to IBM XL compiler semantics
14941 when using AIX-compatible ABI@. Pass floating-point arguments to
14942 prototyped functions beyond the register save area (RSA) on the stack
14943 in addition to argument FPRs. Do not assume that most significant
14944 double in 128-bit long double value is properly rounded when comparing
14945 values and converting to double. Use XL symbol names for long double
14948 The AIX calling convention was extended but not initially documented to
14949 handle an obscure K&R C case of calling a function that takes the
14950 address of its arguments with fewer arguments than declared. IBM XL
14951 compilers access floating point arguments which do not fit in the
14952 RSA from the stack when a subroutine is compiled without
14953 optimization. Because always storing floating-point arguments on the
14954 stack is inefficient and rarely needed, this option is not enabled by
14955 default and only is necessary when calling subroutines compiled by IBM
14956 XL compilers without optimization.
14960 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
14961 application written to use message passing with special startup code to
14962 enable the application to run. The system must have PE installed in the
14963 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
14964 must be overridden with the @option{-specs=} option to specify the
14965 appropriate directory location. The Parallel Environment does not
14966 support threads, so the @option{-mpe} option and the @option{-pthread}
14967 option are incompatible.
14969 @item -malign-natural
14970 @itemx -malign-power
14971 @opindex malign-natural
14972 @opindex malign-power
14973 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
14974 @option{-malign-natural} overrides the ABI-defined alignment of larger
14975 types, such as floating-point doubles, on their natural size-based boundary.
14976 The option @option{-malign-power} instructs GCC to follow the ABI-specified
14977 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
14979 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
14983 @itemx -mhard-float
14984 @opindex msoft-float
14985 @opindex mhard-float
14986 Generate code that does not use (uses) the floating-point register set.
14987 Software floating point emulation is provided if you use the
14988 @option{-msoft-float} option, and pass the option to GCC when linking.
14990 @item -msingle-float
14991 @itemx -mdouble-float
14992 @opindex msingle-float
14993 @opindex mdouble-float
14994 Generate code for single or double-precision floating point operations.
14995 @option{-mdouble-float} implies @option{-msingle-float}.
14998 @opindex msimple-fpu
14999 Do not generate sqrt and div instructions for hardware floating point unit.
15003 Specify type of floating point unit. Valid values are @var{sp_lite}
15004 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
15005 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
15006 and @var{dp_full} (equivalent to -mdouble-float).
15009 @opindex mxilinx-fpu
15010 Perform optimizations for floating point unit on Xilinx PPC 405/440.
15013 @itemx -mno-multiple
15015 @opindex mno-multiple
15016 Generate code that uses (does not use) the load multiple word
15017 instructions and the store multiple word instructions. These
15018 instructions are generated by default on POWER systems, and not
15019 generated on PowerPC systems. Do not use @option{-mmultiple} on little
15020 endian PowerPC systems, since those instructions do not work when the
15021 processor is in little endian mode. The exceptions are PPC740 and
15022 PPC750 which permit the instructions usage in little endian mode.
15027 @opindex mno-string
15028 Generate code that uses (does not use) the load string instructions
15029 and the store string word instructions to save multiple registers and
15030 do small block moves. These instructions are generated by default on
15031 POWER systems, and not generated on PowerPC systems. Do not use
15032 @option{-mstring} on little endian PowerPC systems, since those
15033 instructions do not work when the processor is in little endian mode.
15034 The exceptions are PPC740 and PPC750 which permit the instructions
15035 usage in little endian mode.
15040 @opindex mno-update
15041 Generate code that uses (does not use) the load or store instructions
15042 that update the base register to the address of the calculated memory
15043 location. These instructions are generated by default. If you use
15044 @option{-mno-update}, there is a small window between the time that the
15045 stack pointer is updated and the address of the previous frame is
15046 stored, which means code that walks the stack frame across interrupts or
15047 signals may get corrupted data.
15049 @item -mavoid-indexed-addresses
15050 @item -mno-avoid-indexed-addresses
15051 @opindex mavoid-indexed-addresses
15052 @opindex mno-avoid-indexed-addresses
15053 Generate code that tries to avoid (not avoid) the use of indexed load
15054 or store instructions. These instructions can incur a performance
15055 penalty on Power6 processors in certain situations, such as when
15056 stepping through large arrays that cross a 16M boundary. This option
15057 is enabled by default when targetting Power6 and disabled otherwise.
15060 @itemx -mno-fused-madd
15061 @opindex mfused-madd
15062 @opindex mno-fused-madd
15063 Generate code that uses (does not use) the floating point multiply and
15064 accumulate instructions. These instructions are generated by default if
15065 hardware floating is used.
15071 Generate code that uses (does not use) the half-word multiply and
15072 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
15073 These instructions are generated by default when targetting those
15080 Generate code that uses (does not use) the string-search @samp{dlmzb}
15081 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
15082 generated by default when targetting those processors.
15084 @item -mno-bit-align
15086 @opindex mno-bit-align
15087 @opindex mbit-align
15088 On System V.4 and embedded PowerPC systems do not (do) force structures
15089 and unions that contain bit-fields to be aligned to the base type of the
15092 For example, by default a structure containing nothing but 8
15093 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
15094 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
15095 the structure would be aligned to a 1 byte boundary and be one byte in
15098 @item -mno-strict-align
15099 @itemx -mstrict-align
15100 @opindex mno-strict-align
15101 @opindex mstrict-align
15102 On System V.4 and embedded PowerPC systems do not (do) assume that
15103 unaligned memory references will be handled by the system.
15105 @item -mrelocatable
15106 @itemx -mno-relocatable
15107 @opindex mrelocatable
15108 @opindex mno-relocatable
15109 On embedded PowerPC systems generate code that allows (does not allow)
15110 the program to be relocated to a different address at runtime. If you
15111 use @option{-mrelocatable} on any module, all objects linked together must
15112 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
15114 @item -mrelocatable-lib
15115 @itemx -mno-relocatable-lib
15116 @opindex mrelocatable-lib
15117 @opindex mno-relocatable-lib
15118 On embedded PowerPC systems generate code that allows (does not allow)
15119 the program to be relocated to a different address at runtime. Modules
15120 compiled with @option{-mrelocatable-lib} can be linked with either modules
15121 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
15122 with modules compiled with the @option{-mrelocatable} options.
15128 On System V.4 and embedded PowerPC systems do not (do) assume that
15129 register 2 contains a pointer to a global area pointing to the addresses
15130 used in the program.
15133 @itemx -mlittle-endian
15135 @opindex mlittle-endian
15136 On System V.4 and embedded PowerPC systems compile code for the
15137 processor in little endian mode. The @option{-mlittle-endian} option is
15138 the same as @option{-mlittle}.
15141 @itemx -mbig-endian
15143 @opindex mbig-endian
15144 On System V.4 and embedded PowerPC systems compile code for the
15145 processor in big endian mode. The @option{-mbig-endian} option is
15146 the same as @option{-mbig}.
15148 @item -mdynamic-no-pic
15149 @opindex mdynamic-no-pic
15150 On Darwin and Mac OS X systems, compile code so that it is not
15151 relocatable, but that its external references are relocatable. The
15152 resulting code is suitable for applications, but not shared
15155 @item -mprioritize-restricted-insns=@var{priority}
15156 @opindex mprioritize-restricted-insns
15157 This option controls the priority that is assigned to
15158 dispatch-slot restricted instructions during the second scheduling
15159 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
15160 @var{no/highest/second-highest} priority to dispatch slot restricted
15163 @item -msched-costly-dep=@var{dependence_type}
15164 @opindex msched-costly-dep
15165 This option controls which dependences are considered costly
15166 by the target during instruction scheduling. The argument
15167 @var{dependence_type} takes one of the following values:
15168 @var{no}: no dependence is costly,
15169 @var{all}: all dependences are costly,
15170 @var{true_store_to_load}: a true dependence from store to load is costly,
15171 @var{store_to_load}: any dependence from store to load is costly,
15172 @var{number}: any dependence which latency >= @var{number} is costly.
15174 @item -minsert-sched-nops=@var{scheme}
15175 @opindex minsert-sched-nops
15176 This option controls which nop insertion scheme will be used during
15177 the second scheduling pass. The argument @var{scheme} takes one of the
15179 @var{no}: Don't insert nops.
15180 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
15181 according to the scheduler's grouping.
15182 @var{regroup_exact}: Insert nops to force costly dependent insns into
15183 separate groups. Insert exactly as many nops as needed to force an insn
15184 to a new group, according to the estimated processor grouping.
15185 @var{number}: Insert nops to force costly dependent insns into
15186 separate groups. Insert @var{number} nops to force an insn to a new group.
15189 @opindex mcall-sysv
15190 On System V.4 and embedded PowerPC systems compile code using calling
15191 conventions that adheres to the March 1995 draft of the System V
15192 Application Binary Interface, PowerPC processor supplement. This is the
15193 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
15195 @item -mcall-sysv-eabi
15197 @opindex mcall-sysv-eabi
15198 @opindex mcall-eabi
15199 Specify both @option{-mcall-sysv} and @option{-meabi} options.
15201 @item -mcall-sysv-noeabi
15202 @opindex mcall-sysv-noeabi
15203 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
15205 @item -mcall-aixdesc
15207 On System V.4 and embedded PowerPC systems compile code for the AIX
15211 @opindex mcall-linux
15212 On System V.4 and embedded PowerPC systems compile code for the
15213 Linux-based GNU system.
15217 On System V.4 and embedded PowerPC systems compile code for the
15218 Hurd-based GNU system.
15220 @item -mcall-freebsd
15221 @opindex mcall-freebsd
15222 On System V.4 and embedded PowerPC systems compile code for the
15223 FreeBSD operating system.
15225 @item -mcall-netbsd
15226 @opindex mcall-netbsd
15227 On System V.4 and embedded PowerPC systems compile code for the
15228 NetBSD operating system.
15230 @item -mcall-openbsd
15231 @opindex mcall-netbsd
15232 On System V.4 and embedded PowerPC systems compile code for the
15233 OpenBSD operating system.
15235 @item -maix-struct-return
15236 @opindex maix-struct-return
15237 Return all structures in memory (as specified by the AIX ABI)@.
15239 @item -msvr4-struct-return
15240 @opindex msvr4-struct-return
15241 Return structures smaller than 8 bytes in registers (as specified by the
15244 @item -mabi=@var{abi-type}
15246 Extend the current ABI with a particular extension, or remove such extension.
15247 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
15248 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
15252 Extend the current ABI with SPE ABI extensions. This does not change
15253 the default ABI, instead it adds the SPE ABI extensions to the current
15257 @opindex mabi=no-spe
15258 Disable Booke SPE ABI extensions for the current ABI@.
15260 @item -mabi=ibmlongdouble
15261 @opindex mabi=ibmlongdouble
15262 Change the current ABI to use IBM extended precision long double.
15263 This is a PowerPC 32-bit SYSV ABI option.
15265 @item -mabi=ieeelongdouble
15266 @opindex mabi=ieeelongdouble
15267 Change the current ABI to use IEEE extended precision long double.
15268 This is a PowerPC 32-bit Linux ABI option.
15271 @itemx -mno-prototype
15272 @opindex mprototype
15273 @opindex mno-prototype
15274 On System V.4 and embedded PowerPC systems assume that all calls to
15275 variable argument functions are properly prototyped. Otherwise, the
15276 compiler must insert an instruction before every non prototyped call to
15277 set or clear bit 6 of the condition code register (@var{CR}) to
15278 indicate whether floating point values were passed in the floating point
15279 registers in case the function takes a variable arguments. With
15280 @option{-mprototype}, only calls to prototyped variable argument functions
15281 will set or clear the bit.
15285 On embedded PowerPC systems, assume that the startup module is called
15286 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
15287 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
15292 On embedded PowerPC systems, assume that the startup module is called
15293 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
15298 On embedded PowerPC systems, assume that the startup module is called
15299 @file{crt0.o} and the standard C libraries are @file{libads.a} and
15302 @item -myellowknife
15303 @opindex myellowknife
15304 On embedded PowerPC systems, assume that the startup module is called
15305 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
15310 On System V.4 and embedded PowerPC systems, specify that you are
15311 compiling for a VxWorks system.
15315 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
15316 header to indicate that @samp{eabi} extended relocations are used.
15322 On System V.4 and embedded PowerPC systems do (do not) adhere to the
15323 Embedded Applications Binary Interface (eabi) which is a set of
15324 modifications to the System V.4 specifications. Selecting @option{-meabi}
15325 means that the stack is aligned to an 8 byte boundary, a function
15326 @code{__eabi} is called to from @code{main} to set up the eabi
15327 environment, and the @option{-msdata} option can use both @code{r2} and
15328 @code{r13} to point to two separate small data areas. Selecting
15329 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
15330 do not call an initialization function from @code{main}, and the
15331 @option{-msdata} option will only use @code{r13} to point to a single
15332 small data area. The @option{-meabi} option is on by default if you
15333 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
15336 @opindex msdata=eabi
15337 On System V.4 and embedded PowerPC systems, put small initialized
15338 @code{const} global and static data in the @samp{.sdata2} section, which
15339 is pointed to by register @code{r2}. Put small initialized
15340 non-@code{const} global and static data in the @samp{.sdata} section,
15341 which is pointed to by register @code{r13}. Put small uninitialized
15342 global and static data in the @samp{.sbss} section, which is adjacent to
15343 the @samp{.sdata} section. The @option{-msdata=eabi} option is
15344 incompatible with the @option{-mrelocatable} option. The
15345 @option{-msdata=eabi} option also sets the @option{-memb} option.
15348 @opindex msdata=sysv
15349 On System V.4 and embedded PowerPC systems, put small global and static
15350 data in the @samp{.sdata} section, which is pointed to by register
15351 @code{r13}. Put small uninitialized global and static data in the
15352 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
15353 The @option{-msdata=sysv} option is incompatible with the
15354 @option{-mrelocatable} option.
15356 @item -msdata=default
15358 @opindex msdata=default
15360 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
15361 compile code the same as @option{-msdata=eabi}, otherwise compile code the
15362 same as @option{-msdata=sysv}.
15365 @opindex msdata=data
15366 On System V.4 and embedded PowerPC systems, put small global
15367 data in the @samp{.sdata} section. Put small uninitialized global
15368 data in the @samp{.sbss} section. Do not use register @code{r13}
15369 to address small data however. This is the default behavior unless
15370 other @option{-msdata} options are used.
15374 @opindex msdata=none
15376 On embedded PowerPC systems, put all initialized global and static data
15377 in the @samp{.data} section, and all uninitialized data in the
15378 @samp{.bss} section.
15382 @cindex smaller data references (PowerPC)
15383 @cindex .sdata/.sdata2 references (PowerPC)
15384 On embedded PowerPC systems, put global and static items less than or
15385 equal to @var{num} bytes into the small data or bss sections instead of
15386 the normal data or bss section. By default, @var{num} is 8. The
15387 @option{-G @var{num}} switch is also passed to the linker.
15388 All modules should be compiled with the same @option{-G @var{num}} value.
15391 @itemx -mno-regnames
15393 @opindex mno-regnames
15394 On System V.4 and embedded PowerPC systems do (do not) emit register
15395 names in the assembly language output using symbolic forms.
15398 @itemx -mno-longcall
15400 @opindex mno-longcall
15401 By default assume that all calls are far away so that a longer more
15402 expensive calling sequence is required. This is required for calls
15403 further than 32 megabytes (33,554,432 bytes) from the current location.
15404 A short call will be generated if the compiler knows
15405 the call cannot be that far away. This setting can be overridden by
15406 the @code{shortcall} function attribute, or by @code{#pragma
15409 Some linkers are capable of detecting out-of-range calls and generating
15410 glue code on the fly. On these systems, long calls are unnecessary and
15411 generate slower code. As of this writing, the AIX linker can do this,
15412 as can the GNU linker for PowerPC/64. It is planned to add this feature
15413 to the GNU linker for 32-bit PowerPC systems as well.
15415 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
15416 callee, L42'', plus a ``branch island'' (glue code). The two target
15417 addresses represent the callee and the ``branch island''. The
15418 Darwin/PPC linker will prefer the first address and generate a ``bl
15419 callee'' if the PPC ``bl'' instruction will reach the callee directly;
15420 otherwise, the linker will generate ``bl L42'' to call the ``branch
15421 island''. The ``branch island'' is appended to the body of the
15422 calling function; it computes the full 32-bit address of the callee
15425 On Mach-O (Darwin) systems, this option directs the compiler emit to
15426 the glue for every direct call, and the Darwin linker decides whether
15427 to use or discard it.
15429 In the future, we may cause GCC to ignore all longcall specifications
15430 when the linker is known to generate glue.
15432 @item -mtls-markers
15433 @itemx -mno-tls-markers
15434 @opindex mtls-markers
15435 @opindex mno-tls-markers
15436 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
15437 specifying the function argument. The relocation allows ld to
15438 reliably associate function call with argument setup instructions for
15439 TLS optimization, which in turn allows gcc to better schedule the
15444 Adds support for multithreading with the @dfn{pthreads} library.
15445 This option sets flags for both the preprocessor and linker.
15450 @subsection RX Options
15453 These @option{-m} options are defined for RX implementations:
15456 @item -m64bit-doubles
15457 @itemx -m32bit-doubles
15460 @opindex m64bit-doubles
15461 @opindex m32bit-doubles
15464 Make the @code{double} data type be 64-bits (@option{-m64bit-doubles})
15465 or 32-bits (@option{-m32bit-doubles}) in size. The default is
15466 @option{-m64bit-doubles}. @emph{Note} the RX's hardware floating
15467 point instructions are only used for 32-bit floating point values, and
15468 then only if @option{-ffast-math} has been specified on the command
15469 line. This is because the RX FPU instructions do not properly support
15470 denormal (or sub-normal) values.
15472 The options @option{-fpu} and @option{-nofpu} have been provided at
15473 the request of Rensas for compatibility with their toolchain. The
15474 @option{-mfpu} option enables the use of RX FPU instructions by
15475 selecting 32-bit doubles and enabling unsafe math optimizations. The
15476 @option{-mnofpu} option disables the use of RX FPU instructions, even
15477 if @option{-m32bit-doubles} is active and unsafe math optimizations
15480 @item -mcpu=@var{name}
15481 @itemx -patch=@var{name}
15484 Selects the type of RX CPU to be targeted. Currently on two types are
15485 supported, the generic @var{RX600} and the specific @var{RX610}. The
15486 only difference between them is that the @var{RX610} does not support
15487 the @code{MVTIPL} instruction.
15489 @item -mbig-endian-data
15490 @itemx -mlittle-endian-data
15491 @opindex mbig-endian-data
15492 @opindex mlittle-endian-data
15493 Store data (but not code) in the big-endian format. The default is
15494 @option{-mlittle-endian-data}, ie to store data in the little endian
15497 @item -msmall-data-limit=@var{N}
15498 @opindex msmall-data-limit
15499 Specifies the maximum size in bytes of global and static variables
15500 which can be placed into the small data area. Using the small data
15501 area can lead to smaller and faster code, but the size of area is
15502 limited and it is up to the programmer to ensure that the area does
15503 not overflow. Also when the small data area is used one of the RX's
15504 registers (@code{r13}) is reserved for use pointing to this area, so
15505 it is no longer available for use by the compiler. This could result
15506 in slower and/or larger code if variables which once could have been
15507 held in @code{r13} are now pushed onto the stack.
15509 Note, common variables (variables which have not been initialised) and
15510 constants are not placed into the small data area as they are assigned
15511 to other sections in the output executeable.
15513 The default value is zero, which disables this feature. Note, this
15514 feature is not enabled by default with higher optimization levels
15515 (@option{-O2} etc) because of the potentially deterimental effects of
15516 reserving register @code{r13}. It is up to the programmer to
15517 experiment and discover whether this feature is of benefit to their
15524 Use the simulator runtime. The default is to use the libgloss board
15527 @item -mas100-syntax
15528 @item -mno-as100-syntax
15529 @opindex mas100-syntax
15530 @opindex mno-as100-syntax
15531 When generating assembler output use a syntax that is compatible with
15532 Renesas's AS100 assembler. This syntax can also be handled by the GAS
15533 assembler but it has some restrictions so generating it is not the
15536 @item -mmax-constant-size=@var{N}
15537 @opindex mmax-constant-size
15538 Specifies the maxium size, in bytes, of a constant that can be used as
15539 an operand in a RX instruction. Although the RX instruction set does
15540 allow consants of up to 4 bytes in length to be used in instructions,
15541 a longer value equates to a longer instruction. Thus in some
15542 circumstances it can be beneficial to restrict the size of constants
15543 that are used in instructions. Constants that are too big are instead
15544 placed into a constant pool and referenced via register indirection.
15546 The value @var{N} can be between 0 and 3. A value of 0, the default,
15547 means that constants of any size are allowed.
15551 Enable linker relaxation. Linker relaxation is a process whereby the
15552 linker will attempt to reduce the size of a program by finding shorter
15553 versions of various instructions. Disabled by default.
15555 @item -mint-register=@var{N}
15556 @opindex mint-register
15557 Specify the number of registers to reserve for fast interrupt handler
15558 functions. The value @var{N} can be between 0 and 4. A value of 1
15559 means that register @code{r13} will be reserved for ther exclusive use
15560 of fast interrupt handlers. A value of 2 reserves @code{r13} and
15561 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
15562 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
15563 A value of 0, the default, does not reserve any registers.
15565 @item -msave-acc-in-interrupts
15566 @opindex msave-acc-in-interrupts
15567 Specifies that interrupt handler functions should preserve the
15568 accumulator register. This is only necessary if normal code might use
15569 the accumulator register, for example because it performs 64-bit
15570 multiplications. The default is to ignore the accumulator as this
15571 makes the interrupt handlers faster.
15575 @emph{Note:} The generic GCC command line @option{-ffixed-@var{reg}}
15576 has special significance to the RX port when used with the
15577 @code{interrupt} function attribute. This attribute indicates a
15578 function intended to process fast interrupts. GCC will will ensure
15579 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
15580 and/or @code{r13} and only provided that the normal use of the
15581 corresponding registers have been restricted via the
15582 @option{-ffixed-@var{reg}} or @option{-mint-register} command line
15585 @node S/390 and zSeries Options
15586 @subsection S/390 and zSeries Options
15587 @cindex S/390 and zSeries Options
15589 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
15593 @itemx -msoft-float
15594 @opindex mhard-float
15595 @opindex msoft-float
15596 Use (do not use) the hardware floating-point instructions and registers
15597 for floating-point operations. When @option{-msoft-float} is specified,
15598 functions in @file{libgcc.a} will be used to perform floating-point
15599 operations. When @option{-mhard-float} is specified, the compiler
15600 generates IEEE floating-point instructions. This is the default.
15603 @itemx -mno-hard-dfp
15605 @opindex mno-hard-dfp
15606 Use (do not use) the hardware decimal-floating-point instructions for
15607 decimal-floating-point operations. When @option{-mno-hard-dfp} is
15608 specified, functions in @file{libgcc.a} will be used to perform
15609 decimal-floating-point operations. When @option{-mhard-dfp} is
15610 specified, the compiler generates decimal-floating-point hardware
15611 instructions. This is the default for @option{-march=z9-ec} or higher.
15613 @item -mlong-double-64
15614 @itemx -mlong-double-128
15615 @opindex mlong-double-64
15616 @opindex mlong-double-128
15617 These switches control the size of @code{long double} type. A size
15618 of 64bit makes the @code{long double} type equivalent to the @code{double}
15619 type. This is the default.
15622 @itemx -mno-backchain
15623 @opindex mbackchain
15624 @opindex mno-backchain
15625 Store (do not store) the address of the caller's frame as backchain pointer
15626 into the callee's stack frame.
15627 A backchain may be needed to allow debugging using tools that do not understand
15628 DWARF-2 call frame information.
15629 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
15630 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
15631 the backchain is placed into the topmost word of the 96/160 byte register
15634 In general, code compiled with @option{-mbackchain} is call-compatible with
15635 code compiled with @option{-mmo-backchain}; however, use of the backchain
15636 for debugging purposes usually requires that the whole binary is built with
15637 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
15638 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
15639 to build a linux kernel use @option{-msoft-float}.
15641 The default is to not maintain the backchain.
15643 @item -mpacked-stack
15644 @itemx -mno-packed-stack
15645 @opindex mpacked-stack
15646 @opindex mno-packed-stack
15647 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
15648 specified, the compiler uses the all fields of the 96/160 byte register save
15649 area only for their default purpose; unused fields still take up stack space.
15650 When @option{-mpacked-stack} is specified, register save slots are densely
15651 packed at the top of the register save area; unused space is reused for other
15652 purposes, allowing for more efficient use of the available stack space.
15653 However, when @option{-mbackchain} is also in effect, the topmost word of
15654 the save area is always used to store the backchain, and the return address
15655 register is always saved two words below the backchain.
15657 As long as the stack frame backchain is not used, code generated with
15658 @option{-mpacked-stack} is call-compatible with code generated with
15659 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
15660 S/390 or zSeries generated code that uses the stack frame backchain at run
15661 time, not just for debugging purposes. Such code is not call-compatible
15662 with code compiled with @option{-mpacked-stack}. Also, note that the
15663 combination of @option{-mbackchain},
15664 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
15665 to build a linux kernel use @option{-msoft-float}.
15667 The default is to not use the packed stack layout.
15670 @itemx -mno-small-exec
15671 @opindex msmall-exec
15672 @opindex mno-small-exec
15673 Generate (or do not generate) code using the @code{bras} instruction
15674 to do subroutine calls.
15675 This only works reliably if the total executable size does not
15676 exceed 64k. The default is to use the @code{basr} instruction instead,
15677 which does not have this limitation.
15683 When @option{-m31} is specified, generate code compliant to the
15684 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
15685 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
15686 particular to generate 64-bit instructions. For the @samp{s390}
15687 targets, the default is @option{-m31}, while the @samp{s390x}
15688 targets default to @option{-m64}.
15694 When @option{-mzarch} is specified, generate code using the
15695 instructions available on z/Architecture.
15696 When @option{-mesa} is specified, generate code using the
15697 instructions available on ESA/390. Note that @option{-mesa} is
15698 not possible with @option{-m64}.
15699 When generating code compliant to the GNU/Linux for S/390 ABI,
15700 the default is @option{-mesa}. When generating code compliant
15701 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
15707 Generate (or do not generate) code using the @code{mvcle} instruction
15708 to perform block moves. When @option{-mno-mvcle} is specified,
15709 use a @code{mvc} loop instead. This is the default unless optimizing for
15716 Print (or do not print) additional debug information when compiling.
15717 The default is to not print debug information.
15719 @item -march=@var{cpu-type}
15721 Generate code that will run on @var{cpu-type}, which is the name of a system
15722 representing a certain processor type. Possible values for
15723 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
15724 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
15725 When generating code using the instructions available on z/Architecture,
15726 the default is @option{-march=z900}. Otherwise, the default is
15727 @option{-march=g5}.
15729 @item -mtune=@var{cpu-type}
15731 Tune to @var{cpu-type} everything applicable about the generated code,
15732 except for the ABI and the set of available instructions.
15733 The list of @var{cpu-type} values is the same as for @option{-march}.
15734 The default is the value used for @option{-march}.
15737 @itemx -mno-tpf-trace
15738 @opindex mtpf-trace
15739 @opindex mno-tpf-trace
15740 Generate code that adds (does not add) in TPF OS specific branches to trace
15741 routines in the operating system. This option is off by default, even
15742 when compiling for the TPF OS@.
15745 @itemx -mno-fused-madd
15746 @opindex mfused-madd
15747 @opindex mno-fused-madd
15748 Generate code that uses (does not use) the floating point multiply and
15749 accumulate instructions. These instructions are generated by default if
15750 hardware floating point is used.
15752 @item -mwarn-framesize=@var{framesize}
15753 @opindex mwarn-framesize
15754 Emit a warning if the current function exceeds the given frame size. Because
15755 this is a compile time check it doesn't need to be a real problem when the program
15756 runs. It is intended to identify functions which most probably cause
15757 a stack overflow. It is useful to be used in an environment with limited stack
15758 size e.g.@: the linux kernel.
15760 @item -mwarn-dynamicstack
15761 @opindex mwarn-dynamicstack
15762 Emit a warning if the function calls alloca or uses dynamically
15763 sized arrays. This is generally a bad idea with a limited stack size.
15765 @item -mstack-guard=@var{stack-guard}
15766 @itemx -mstack-size=@var{stack-size}
15767 @opindex mstack-guard
15768 @opindex mstack-size
15769 If these options are provided the s390 back end emits additional instructions in
15770 the function prologue which trigger a trap if the stack size is @var{stack-guard}
15771 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
15772 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
15773 the frame size of the compiled function is chosen.
15774 These options are intended to be used to help debugging stack overflow problems.
15775 The additionally emitted code causes only little overhead and hence can also be
15776 used in production like systems without greater performance degradation. The given
15777 values have to be exact powers of 2 and @var{stack-size} has to be greater than
15778 @var{stack-guard} without exceeding 64k.
15779 In order to be efficient the extra code makes the assumption that the stack starts
15780 at an address aligned to the value given by @var{stack-size}.
15781 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
15784 @node Score Options
15785 @subsection Score Options
15786 @cindex Score Options
15788 These options are defined for Score implementations:
15793 Compile code for big endian mode. This is the default.
15797 Compile code for little endian mode.
15801 Disable generate bcnz instruction.
15805 Enable generate unaligned load and store instruction.
15809 Enable the use of multiply-accumulate instructions. Disabled by default.
15813 Specify the SCORE5 as the target architecture.
15817 Specify the SCORE5U of the target architecture.
15821 Specify the SCORE7 as the target architecture. This is the default.
15825 Specify the SCORE7D as the target architecture.
15829 @subsection SH Options
15831 These @samp{-m} options are defined for the SH implementations:
15836 Generate code for the SH1.
15840 Generate code for the SH2.
15843 Generate code for the SH2e.
15847 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
15848 that the floating-point unit is not used.
15850 @item -m2a-single-only
15851 @opindex m2a-single-only
15852 Generate code for the SH2a-FPU, in such a way that no double-precision
15853 floating point operations are used.
15856 @opindex m2a-single
15857 Generate code for the SH2a-FPU assuming the floating-point unit is in
15858 single-precision mode by default.
15862 Generate code for the SH2a-FPU assuming the floating-point unit is in
15863 double-precision mode by default.
15867 Generate code for the SH3.
15871 Generate code for the SH3e.
15875 Generate code for the SH4 without a floating-point unit.
15877 @item -m4-single-only
15878 @opindex m4-single-only
15879 Generate code for the SH4 with a floating-point unit that only
15880 supports single-precision arithmetic.
15884 Generate code for the SH4 assuming the floating-point unit is in
15885 single-precision mode by default.
15889 Generate code for the SH4.
15893 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
15894 floating-point unit is not used.
15896 @item -m4a-single-only
15897 @opindex m4a-single-only
15898 Generate code for the SH4a, in such a way that no double-precision
15899 floating point operations are used.
15902 @opindex m4a-single
15903 Generate code for the SH4a assuming the floating-point unit is in
15904 single-precision mode by default.
15908 Generate code for the SH4a.
15912 Same as @option{-m4a-nofpu}, except that it implicitly passes
15913 @option{-dsp} to the assembler. GCC doesn't generate any DSP
15914 instructions at the moment.
15918 Compile code for the processor in big endian mode.
15922 Compile code for the processor in little endian mode.
15926 Align doubles at 64-bit boundaries. Note that this changes the calling
15927 conventions, and thus some functions from the standard C library will
15928 not work unless you recompile it first with @option{-mdalign}.
15932 Shorten some address references at link time, when possible; uses the
15933 linker option @option{-relax}.
15937 Use 32-bit offsets in @code{switch} tables. The default is to use
15942 Enable the use of bit manipulation instructions on SH2A.
15946 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
15947 alignment constraints.
15951 Comply with the calling conventions defined by Renesas.
15955 Comply with the calling conventions defined by Renesas.
15959 Comply with the calling conventions defined for GCC before the Renesas
15960 conventions were available. This option is the default for all
15961 targets of the SH toolchain except for @samp{sh-symbianelf}.
15964 @opindex mnomacsave
15965 Mark the @code{MAC} register as call-clobbered, even if
15966 @option{-mhitachi} is given.
15970 Increase IEEE-compliance of floating-point code.
15971 At the moment, this is equivalent to @option{-fno-finite-math-only}.
15972 When generating 16 bit SH opcodes, getting IEEE-conforming results for
15973 comparisons of NANs / infinities incurs extra overhead in every
15974 floating point comparison, therefore the default is set to
15975 @option{-ffinite-math-only}.
15977 @item -minline-ic_invalidate
15978 @opindex minline-ic_invalidate
15979 Inline code to invalidate instruction cache entries after setting up
15980 nested function trampolines.
15981 This option has no effect if -musermode is in effect and the selected
15982 code generation option (e.g. -m4) does not allow the use of the icbi
15984 If the selected code generation option does not allow the use of the icbi
15985 instruction, and -musermode is not in effect, the inlined code will
15986 manipulate the instruction cache address array directly with an associative
15987 write. This not only requires privileged mode, but it will also
15988 fail if the cache line had been mapped via the TLB and has become unmapped.
15992 Dump instruction size and location in the assembly code.
15995 @opindex mpadstruct
15996 This option is deprecated. It pads structures to multiple of 4 bytes,
15997 which is incompatible with the SH ABI@.
16001 Optimize for space instead of speed. Implied by @option{-Os}.
16004 @opindex mprefergot
16005 When generating position-independent code, emit function calls using
16006 the Global Offset Table instead of the Procedure Linkage Table.
16010 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
16011 if the inlined code would not work in user mode.
16012 This is the default when the target is @code{sh-*-linux*}.
16014 @item -multcost=@var{number}
16015 @opindex multcost=@var{number}
16016 Set the cost to assume for a multiply insn.
16018 @item -mdiv=@var{strategy}
16019 @opindex mdiv=@var{strategy}
16020 Set the division strategy to use for SHmedia code. @var{strategy} must be
16021 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
16022 inv:call2, inv:fp .
16023 "fp" performs the operation in floating point. This has a very high latency,
16024 but needs only a few instructions, so it might be a good choice if
16025 your code has enough easily exploitable ILP to allow the compiler to
16026 schedule the floating point instructions together with other instructions.
16027 Division by zero causes a floating point exception.
16028 "inv" uses integer operations to calculate the inverse of the divisor,
16029 and then multiplies the dividend with the inverse. This strategy allows
16030 cse and hoisting of the inverse calculation. Division by zero calculates
16031 an unspecified result, but does not trap.
16032 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
16033 have been found, or if the entire operation has been hoisted to the same
16034 place, the last stages of the inverse calculation are intertwined with the
16035 final multiply to reduce the overall latency, at the expense of using a few
16036 more instructions, and thus offering fewer scheduling opportunities with
16038 "call" calls a library function that usually implements the inv:minlat
16040 This gives high code density for m5-*media-nofpu compilations.
16041 "call2" uses a different entry point of the same library function, where it
16042 assumes that a pointer to a lookup table has already been set up, which
16043 exposes the pointer load to cse / code hoisting optimizations.
16044 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
16045 code generation, but if the code stays unoptimized, revert to the "call",
16046 "call2", or "fp" strategies, respectively. Note that the
16047 potentially-trapping side effect of division by zero is carried by a
16048 separate instruction, so it is possible that all the integer instructions
16049 are hoisted out, but the marker for the side effect stays where it is.
16050 A recombination to fp operations or a call is not possible in that case.
16051 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
16052 that the inverse calculation was nor separated from the multiply, they speed
16053 up division where the dividend fits into 20 bits (plus sign where applicable),
16054 by inserting a test to skip a number of operations in this case; this test
16055 slows down the case of larger dividends. inv20u assumes the case of a such
16056 a small dividend to be unlikely, and inv20l assumes it to be likely.
16058 @item -mdivsi3_libfunc=@var{name}
16059 @opindex mdivsi3_libfunc=@var{name}
16060 Set the name of the library function used for 32 bit signed division to
16061 @var{name}. This only affect the name used in the call and inv:call
16062 division strategies, and the compiler will still expect the same
16063 sets of input/output/clobbered registers as if this option was not present.
16065 @item -mfixed-range=@var{register-range}
16066 @opindex mfixed-range
16067 Generate code treating the given register range as fixed registers.
16068 A fixed register is one that the register allocator can not use. This is
16069 useful when compiling kernel code. A register range is specified as
16070 two registers separated by a dash. Multiple register ranges can be
16071 specified separated by a comma.
16073 @item -madjust-unroll
16074 @opindex madjust-unroll
16075 Throttle unrolling to avoid thrashing target registers.
16076 This option only has an effect if the gcc code base supports the
16077 TARGET_ADJUST_UNROLL_MAX target hook.
16079 @item -mindexed-addressing
16080 @opindex mindexed-addressing
16081 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
16082 This is only safe if the hardware and/or OS implement 32 bit wrap-around
16083 semantics for the indexed addressing mode. The architecture allows the
16084 implementation of processors with 64 bit MMU, which the OS could use to
16085 get 32 bit addressing, but since no current hardware implementation supports
16086 this or any other way to make the indexed addressing mode safe to use in
16087 the 32 bit ABI, the default is -mno-indexed-addressing.
16089 @item -mgettrcost=@var{number}
16090 @opindex mgettrcost=@var{number}
16091 Set the cost assumed for the gettr instruction to @var{number}.
16092 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
16096 Assume pt* instructions won't trap. This will generally generate better
16097 scheduled code, but is unsafe on current hardware. The current architecture
16098 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
16099 This has the unintentional effect of making it unsafe to schedule ptabs /
16100 ptrel before a branch, or hoist it out of a loop. For example,
16101 __do_global_ctors, a part of libgcc that runs constructors at program
16102 startup, calls functions in a list which is delimited by @minus{}1. With the
16103 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
16104 That means that all the constructors will be run a bit quicker, but when
16105 the loop comes to the end of the list, the program crashes because ptabs
16106 loads @minus{}1 into a target register. Since this option is unsafe for any
16107 hardware implementing the current architecture specification, the default
16108 is -mno-pt-fixed. Unless the user specifies a specific cost with
16109 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
16110 this deters register allocation using target registers for storing
16113 @item -minvalid-symbols
16114 @opindex minvalid-symbols
16115 Assume symbols might be invalid. Ordinary function symbols generated by
16116 the compiler will always be valid to load with movi/shori/ptabs or
16117 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
16118 to generate symbols that will cause ptabs / ptrel to trap.
16119 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
16120 It will then prevent cross-basic-block cse, hoisting and most scheduling
16121 of symbol loads. The default is @option{-mno-invalid-symbols}.
16124 @node SPARC Options
16125 @subsection SPARC Options
16126 @cindex SPARC options
16128 These @samp{-m} options are supported on the SPARC:
16131 @item -mno-app-regs
16133 @opindex mno-app-regs
16135 Specify @option{-mapp-regs} to generate output using the global registers
16136 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
16139 To be fully SVR4 ABI compliant at the cost of some performance loss,
16140 specify @option{-mno-app-regs}. You should compile libraries and system
16141 software with this option.
16144 @itemx -mhard-float
16146 @opindex mhard-float
16147 Generate output containing floating point instructions. This is the
16151 @itemx -msoft-float
16153 @opindex msoft-float
16154 Generate output containing library calls for floating point.
16155 @strong{Warning:} the requisite libraries are not available for all SPARC
16156 targets. Normally the facilities of the machine's usual C compiler are
16157 used, but this cannot be done directly in cross-compilation. You must make
16158 your own arrangements to provide suitable library functions for
16159 cross-compilation. The embedded targets @samp{sparc-*-aout} and
16160 @samp{sparclite-*-*} do provide software floating point support.
16162 @option{-msoft-float} changes the calling convention in the output file;
16163 therefore, it is only useful if you compile @emph{all} of a program with
16164 this option. In particular, you need to compile @file{libgcc.a}, the
16165 library that comes with GCC, with @option{-msoft-float} in order for
16168 @item -mhard-quad-float
16169 @opindex mhard-quad-float
16170 Generate output containing quad-word (long double) floating point
16173 @item -msoft-quad-float
16174 @opindex msoft-quad-float
16175 Generate output containing library calls for quad-word (long double)
16176 floating point instructions. The functions called are those specified
16177 in the SPARC ABI@. This is the default.
16179 As of this writing, there are no SPARC implementations that have hardware
16180 support for the quad-word floating point instructions. They all invoke
16181 a trap handler for one of these instructions, and then the trap handler
16182 emulates the effect of the instruction. Because of the trap handler overhead,
16183 this is much slower than calling the ABI library routines. Thus the
16184 @option{-msoft-quad-float} option is the default.
16186 @item -mno-unaligned-doubles
16187 @itemx -munaligned-doubles
16188 @opindex mno-unaligned-doubles
16189 @opindex munaligned-doubles
16190 Assume that doubles have 8 byte alignment. This is the default.
16192 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
16193 alignment only if they are contained in another type, or if they have an
16194 absolute address. Otherwise, it assumes they have 4 byte alignment.
16195 Specifying this option avoids some rare compatibility problems with code
16196 generated by other compilers. It is not the default because it results
16197 in a performance loss, especially for floating point code.
16199 @item -mno-faster-structs
16200 @itemx -mfaster-structs
16201 @opindex mno-faster-structs
16202 @opindex mfaster-structs
16203 With @option{-mfaster-structs}, the compiler assumes that structures
16204 should have 8 byte alignment. This enables the use of pairs of
16205 @code{ldd} and @code{std} instructions for copies in structure
16206 assignment, in place of twice as many @code{ld} and @code{st} pairs.
16207 However, the use of this changed alignment directly violates the SPARC
16208 ABI@. Thus, it's intended only for use on targets where the developer
16209 acknowledges that their resulting code will not be directly in line with
16210 the rules of the ABI@.
16212 @item -mimpure-text
16213 @opindex mimpure-text
16214 @option{-mimpure-text}, used in addition to @option{-shared}, tells
16215 the compiler to not pass @option{-z text} to the linker when linking a
16216 shared object. Using this option, you can link position-dependent
16217 code into a shared object.
16219 @option{-mimpure-text} suppresses the ``relocations remain against
16220 allocatable but non-writable sections'' linker error message.
16221 However, the necessary relocations will trigger copy-on-write, and the
16222 shared object is not actually shared across processes. Instead of
16223 using @option{-mimpure-text}, you should compile all source code with
16224 @option{-fpic} or @option{-fPIC}.
16226 This option is only available on SunOS and Solaris.
16228 @item -mcpu=@var{cpu_type}
16230 Set the instruction set, register set, and instruction scheduling parameters
16231 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
16232 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
16233 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
16234 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
16235 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
16237 Default instruction scheduling parameters are used for values that select
16238 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
16239 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
16241 Here is a list of each supported architecture and their supported
16246 v8: supersparc, hypersparc
16247 sparclite: f930, f934, sparclite86x
16249 v9: ultrasparc, ultrasparc3, niagara, niagara2
16252 By default (unless configured otherwise), GCC generates code for the V7
16253 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
16254 additionally optimizes it for the Cypress CY7C602 chip, as used in the
16255 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
16256 SPARCStation 1, 2, IPX etc.
16258 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
16259 architecture. The only difference from V7 code is that the compiler emits
16260 the integer multiply and integer divide instructions which exist in SPARC-V8
16261 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
16262 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
16265 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
16266 the SPARC architecture. This adds the integer multiply, integer divide step
16267 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
16268 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
16269 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
16270 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
16271 MB86934 chip, which is the more recent SPARClite with FPU@.
16273 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
16274 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
16275 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
16276 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
16277 optimizes it for the TEMIC SPARClet chip.
16279 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
16280 architecture. This adds 64-bit integer and floating-point move instructions,
16281 3 additional floating-point condition code registers and conditional move
16282 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
16283 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
16284 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
16285 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
16286 @option{-mcpu=niagara}, the compiler additionally optimizes it for
16287 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
16288 additionally optimizes it for Sun UltraSPARC T2 chips.
16290 @item -mtune=@var{cpu_type}
16292 Set the instruction scheduling parameters for machine type
16293 @var{cpu_type}, but do not set the instruction set or register set that the
16294 option @option{-mcpu=@var{cpu_type}} would.
16296 The same values for @option{-mcpu=@var{cpu_type}} can be used for
16297 @option{-mtune=@var{cpu_type}}, but the only useful values are those
16298 that select a particular cpu implementation. Those are @samp{cypress},
16299 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
16300 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
16301 @samp{ultrasparc3}, @samp{niagara}, and @samp{niagara2}.
16306 @opindex mno-v8plus
16307 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
16308 difference from the V8 ABI is that the global and out registers are
16309 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
16310 mode for all SPARC-V9 processors.
16316 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
16317 Visual Instruction Set extensions. The default is @option{-mno-vis}.
16320 These @samp{-m} options are supported in addition to the above
16321 on SPARC-V9 processors in 64-bit environments:
16324 @item -mlittle-endian
16325 @opindex mlittle-endian
16326 Generate code for a processor running in little-endian mode. It is only
16327 available for a few configurations and most notably not on Solaris and Linux.
16333 Generate code for a 32-bit or 64-bit environment.
16334 The 32-bit environment sets int, long and pointer to 32 bits.
16335 The 64-bit environment sets int to 32 bits and long and pointer
16338 @item -mcmodel=medlow
16339 @opindex mcmodel=medlow
16340 Generate code for the Medium/Low code model: 64-bit addresses, programs
16341 must be linked in the low 32 bits of memory. Programs can be statically
16342 or dynamically linked.
16344 @item -mcmodel=medmid
16345 @opindex mcmodel=medmid
16346 Generate code for the Medium/Middle code model: 64-bit addresses, programs
16347 must be linked in the low 44 bits of memory, the text and data segments must
16348 be less than 2GB in size and the data segment must be located within 2GB of
16351 @item -mcmodel=medany
16352 @opindex mcmodel=medany
16353 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
16354 may be linked anywhere in memory, the text and data segments must be less
16355 than 2GB in size and the data segment must be located within 2GB of the
16358 @item -mcmodel=embmedany
16359 @opindex mcmodel=embmedany
16360 Generate code for the Medium/Anywhere code model for embedded systems:
16361 64-bit addresses, the text and data segments must be less than 2GB in
16362 size, both starting anywhere in memory (determined at link time). The
16363 global register %g4 points to the base of the data segment. Programs
16364 are statically linked and PIC is not supported.
16367 @itemx -mno-stack-bias
16368 @opindex mstack-bias
16369 @opindex mno-stack-bias
16370 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
16371 frame pointer if present, are offset by @minus{}2047 which must be added back
16372 when making stack frame references. This is the default in 64-bit mode.
16373 Otherwise, assume no such offset is present.
16376 These switches are supported in addition to the above on Solaris:
16381 Add support for multithreading using the Solaris threads library. This
16382 option sets flags for both the preprocessor and linker. This option does
16383 not affect the thread safety of object code produced by the compiler or
16384 that of libraries supplied with it.
16388 Add support for multithreading using the POSIX threads library. This
16389 option sets flags for both the preprocessor and linker. This option does
16390 not affect the thread safety of object code produced by the compiler or
16391 that of libraries supplied with it.
16395 This is a synonym for @option{-pthreads}.
16399 @subsection SPU Options
16400 @cindex SPU options
16402 These @samp{-m} options are supported on the SPU:
16406 @itemx -merror-reloc
16407 @opindex mwarn-reloc
16408 @opindex merror-reloc
16410 The loader for SPU does not handle dynamic relocations. By default, GCC
16411 will give an error when it generates code that requires a dynamic
16412 relocation. @option{-mno-error-reloc} disables the error,
16413 @option{-mwarn-reloc} will generate a warning instead.
16416 @itemx -munsafe-dma
16418 @opindex munsafe-dma
16420 Instructions which initiate or test completion of DMA must not be
16421 reordered with respect to loads and stores of the memory which is being
16422 accessed. Users typically address this problem using the volatile
16423 keyword, but that can lead to inefficient code in places where the
16424 memory is known to not change. Rather than mark the memory as volatile
16425 we treat the DMA instructions as potentially effecting all memory. With
16426 @option{-munsafe-dma} users must use the volatile keyword to protect
16429 @item -mbranch-hints
16430 @opindex mbranch-hints
16432 By default, GCC will generate a branch hint instruction to avoid
16433 pipeline stalls for always taken or probably taken branches. A hint
16434 will not be generated closer than 8 instructions away from its branch.
16435 There is little reason to disable them, except for debugging purposes,
16436 or to make an object a little bit smaller.
16440 @opindex msmall-mem
16441 @opindex mlarge-mem
16443 By default, GCC generates code assuming that addresses are never larger
16444 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
16445 a full 32 bit address.
16450 By default, GCC links against startup code that assumes the SPU-style
16451 main function interface (which has an unconventional parameter list).
16452 With @option{-mstdmain}, GCC will link your program against startup
16453 code that assumes a C99-style interface to @code{main}, including a
16454 local copy of @code{argv} strings.
16456 @item -mfixed-range=@var{register-range}
16457 @opindex mfixed-range
16458 Generate code treating the given register range as fixed registers.
16459 A fixed register is one that the register allocator can not use. This is
16460 useful when compiling kernel code. A register range is specified as
16461 two registers separated by a dash. Multiple register ranges can be
16462 specified separated by a comma.
16468 Compile code assuming that pointers to the PPU address space accessed
16469 via the @code{__ea} named address space qualifier are either 32 or 64
16470 bits wide. The default is 32 bits. As this is an ABI changing option,
16471 all object code in an executable must be compiled with the same setting.
16473 @item -maddress-space-conversion
16474 @itemx -mno-address-space-conversion
16475 @opindex maddress-space-conversion
16476 @opindex mno-address-space-conversion
16477 Allow/disallow treating the @code{__ea} address space as superset
16478 of the generic address space. This enables explicit type casts
16479 between @code{__ea} and generic pointer as well as implicit
16480 conversions of generic pointers to @code{__ea} pointers. The
16481 default is to allow address space pointer conversions.
16483 @item -mcache-size=@var{cache-size}
16484 @opindex mcache-size
16485 This option controls the version of libgcc that the compiler links to an
16486 executable and selects a software-managed cache for accessing variables
16487 in the @code{__ea} address space with a particular cache size. Possible
16488 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
16489 and @samp{128}. The default cache size is 64KB.
16491 @item -matomic-updates
16492 @itemx -mno-atomic-updates
16493 @opindex matomic-updates
16494 @opindex mno-atomic-updates
16495 This option controls the version of libgcc that the compiler links to an
16496 executable and selects whether atomic updates to the software-managed
16497 cache of PPU-side variables are used. If you use atomic updates, changes
16498 to a PPU variable from SPU code using the @code{__ea} named address space
16499 qualifier will not interfere with changes to other PPU variables residing
16500 in the same cache line from PPU code. If you do not use atomic updates,
16501 such interference may occur; however, writing back cache lines will be
16502 more efficient. The default behavior is to use atomic updates.
16505 @itemx -mdual-nops=@var{n}
16506 @opindex mdual-nops
16507 By default, GCC will insert nops to increase dual issue when it expects
16508 it to increase performance. @var{n} can be a value from 0 to 10. A
16509 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
16510 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
16512 @item -mhint-max-nops=@var{n}
16513 @opindex mhint-max-nops
16514 Maximum number of nops to insert for a branch hint. A branch hint must
16515 be at least 8 instructions away from the branch it is effecting. GCC
16516 will insert up to @var{n} nops to enforce this, otherwise it will not
16517 generate the branch hint.
16519 @item -mhint-max-distance=@var{n}
16520 @opindex mhint-max-distance
16521 The encoding of the branch hint instruction limits the hint to be within
16522 256 instructions of the branch it is effecting. By default, GCC makes
16523 sure it is within 125.
16526 @opindex msafe-hints
16527 Work around a hardware bug which causes the SPU to stall indefinitely.
16528 By default, GCC will insert the @code{hbrp} instruction to make sure
16529 this stall won't happen.
16533 @node System V Options
16534 @subsection Options for System V
16536 These additional options are available on System V Release 4 for
16537 compatibility with other compilers on those systems:
16542 Create a shared object.
16543 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
16547 Identify the versions of each tool used by the compiler, in a
16548 @code{.ident} assembler directive in the output.
16552 Refrain from adding @code{.ident} directives to the output file (this is
16555 @item -YP,@var{dirs}
16557 Search the directories @var{dirs}, and no others, for libraries
16558 specified with @option{-l}.
16560 @item -Ym,@var{dir}
16562 Look in the directory @var{dir} to find the M4 preprocessor.
16563 The assembler uses this option.
16564 @c This is supposed to go with a -Yd for predefined M4 macro files, but
16565 @c the generic assembler that comes with Solaris takes just -Ym.
16569 @subsection V850 Options
16570 @cindex V850 Options
16572 These @samp{-m} options are defined for V850 implementations:
16576 @itemx -mno-long-calls
16577 @opindex mlong-calls
16578 @opindex mno-long-calls
16579 Treat all calls as being far away (near). If calls are assumed to be
16580 far away, the compiler will always load the functions address up into a
16581 register, and call indirect through the pointer.
16587 Do not optimize (do optimize) basic blocks that use the same index
16588 pointer 4 or more times to copy pointer into the @code{ep} register, and
16589 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
16590 option is on by default if you optimize.
16592 @item -mno-prolog-function
16593 @itemx -mprolog-function
16594 @opindex mno-prolog-function
16595 @opindex mprolog-function
16596 Do not use (do use) external functions to save and restore registers
16597 at the prologue and epilogue of a function. The external functions
16598 are slower, but use less code space if more than one function saves
16599 the same number of registers. The @option{-mprolog-function} option
16600 is on by default if you optimize.
16604 Try to make the code as small as possible. At present, this just turns
16605 on the @option{-mep} and @option{-mprolog-function} options.
16607 @item -mtda=@var{n}
16609 Put static or global variables whose size is @var{n} bytes or less into
16610 the tiny data area that register @code{ep} points to. The tiny data
16611 area can hold up to 256 bytes in total (128 bytes for byte references).
16613 @item -msda=@var{n}
16615 Put static or global variables whose size is @var{n} bytes or less into
16616 the small data area that register @code{gp} points to. The small data
16617 area can hold up to 64 kilobytes.
16619 @item -mzda=@var{n}
16621 Put static or global variables whose size is @var{n} bytes or less into
16622 the first 32 kilobytes of memory.
16626 Specify that the target processor is the V850.
16629 @opindex mbig-switch
16630 Generate code suitable for big switch tables. Use this option only if
16631 the assembler/linker complain about out of range branches within a switch
16636 This option will cause r2 and r5 to be used in the code generated by
16637 the compiler. This setting is the default.
16639 @item -mno-app-regs
16640 @opindex mno-app-regs
16641 This option will cause r2 and r5 to be treated as fixed registers.
16645 Specify that the target processor is the V850E1. The preprocessor
16646 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
16647 this option is used.
16651 Specify that the target processor is the V850E@. The preprocessor
16652 constant @samp{__v850e__} will be defined if this option is used.
16654 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
16655 are defined then a default target processor will be chosen and the
16656 relevant @samp{__v850*__} preprocessor constant will be defined.
16658 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
16659 defined, regardless of which processor variant is the target.
16661 @item -mdisable-callt
16662 @opindex mdisable-callt
16663 This option will suppress generation of the CALLT instruction for the
16664 v850e and v850e1 flavors of the v850 architecture. The default is
16665 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
16670 @subsection VAX Options
16671 @cindex VAX options
16673 These @samp{-m} options are defined for the VAX:
16678 Do not output certain jump instructions (@code{aobleq} and so on)
16679 that the Unix assembler for the VAX cannot handle across long
16684 Do output those jump instructions, on the assumption that you
16685 will assemble with the GNU assembler.
16689 Output code for g-format floating point numbers instead of d-format.
16692 @node VxWorks Options
16693 @subsection VxWorks Options
16694 @cindex VxWorks Options
16696 The options in this section are defined for all VxWorks targets.
16697 Options specific to the target hardware are listed with the other
16698 options for that target.
16703 GCC can generate code for both VxWorks kernels and real time processes
16704 (RTPs). This option switches from the former to the latter. It also
16705 defines the preprocessor macro @code{__RTP__}.
16708 @opindex non-static
16709 Link an RTP executable against shared libraries rather than static
16710 libraries. The options @option{-static} and @option{-shared} can
16711 also be used for RTPs (@pxref{Link Options}); @option{-static}
16718 These options are passed down to the linker. They are defined for
16719 compatibility with Diab.
16722 @opindex Xbind-lazy
16723 Enable lazy binding of function calls. This option is equivalent to
16724 @option{-Wl,-z,now} and is defined for compatibility with Diab.
16728 Disable lazy binding of function calls. This option is the default and
16729 is defined for compatibility with Diab.
16732 @node x86-64 Options
16733 @subsection x86-64 Options
16734 @cindex x86-64 options
16736 These are listed under @xref{i386 and x86-64 Options}.
16738 @node i386 and x86-64 Windows Options
16739 @subsection i386 and x86-64 Windows Options
16740 @cindex i386 and x86-64 Windows Options
16742 These additional options are available for Windows targets:
16747 This option is available for Cygwin and MinGW targets. It
16748 specifies that a console application is to be generated, by
16749 instructing the linker to set the PE header subsystem type
16750 required for console applications.
16751 This is the default behavior for Cygwin and MinGW targets.
16755 This option is available for Cygwin targets. It specifies that
16756 the Cygwin internal interface is to be used for predefined
16757 preprocessor macros, C runtime libraries and related linker
16758 paths and options. For Cygwin targets this is the default behavior.
16759 This option is deprecated and will be removed in a future release.
16762 @opindex mno-cygwin
16763 This option is available for Cygwin targets. It specifies that
16764 the MinGW internal interface is to be used instead of Cygwin's, by
16765 setting MinGW-related predefined macros and linker paths and default
16767 This option is deprecated and will be removed in a future release.
16771 This option is available for Cygwin and MinGW targets. It
16772 specifies that a DLL - a dynamic link library - is to be
16773 generated, enabling the selection of the required runtime
16774 startup object and entry point.
16776 @item -mnop-fun-dllimport
16777 @opindex mnop-fun-dllimport
16778 This option is available for Cygwin and MinGW targets. It
16779 specifies that the dllimport attribute should be ignored.
16783 This option is available for MinGW targets. It specifies
16784 that MinGW-specific thread support is to be used.
16788 This option is available for mingw-w64 targets. It specifies
16789 that the UNICODE macro is getting pre-defined and that the
16790 unicode capable runtime startup code is choosen.
16794 This option is available for Cygwin and MinGW targets. It
16795 specifies that the typical Windows pre-defined macros are to
16796 be set in the pre-processor, but does not influence the choice
16797 of runtime library/startup code.
16801 This option is available for Cygwin and MinGW targets. It
16802 specifies that a GUI application is to be generated by
16803 instructing the linker to set the PE header subsystem type
16806 @item -fno-set-stack-executable
16807 @opindex fno-set-stack-executable
16808 This option is available for MinGW targets. It specifies that
16809 the executable flag for stack used by nested functions isn't
16810 set. This is necessary for binaries running in kernel mode of
16811 Windows, as there the user32 API, which is used to set executable
16812 privileges, isn't available.
16814 @item -mpe-aligned-commons
16815 @opindex mpe-aligned-commons
16816 This option is available for Cygwin and MinGW targets. It
16817 specifies that the GNU extension to the PE file format that
16818 permits the correct alignment of COMMON variables should be
16819 used when generating code. It will be enabled by default if
16820 GCC detects that the target assembler found during configuration
16821 supports the feature.
16824 See also under @ref{i386 and x86-64 Options} for standard options.
16826 @node Xstormy16 Options
16827 @subsection Xstormy16 Options
16828 @cindex Xstormy16 Options
16830 These options are defined for Xstormy16:
16835 Choose startup files and linker script suitable for the simulator.
16838 @node Xtensa Options
16839 @subsection Xtensa Options
16840 @cindex Xtensa Options
16842 These options are supported for Xtensa targets:
16846 @itemx -mno-const16
16848 @opindex mno-const16
16849 Enable or disable use of @code{CONST16} instructions for loading
16850 constant values. The @code{CONST16} instruction is currently not a
16851 standard option from Tensilica. When enabled, @code{CONST16}
16852 instructions are always used in place of the standard @code{L32R}
16853 instructions. The use of @code{CONST16} is enabled by default only if
16854 the @code{L32R} instruction is not available.
16857 @itemx -mno-fused-madd
16858 @opindex mfused-madd
16859 @opindex mno-fused-madd
16860 Enable or disable use of fused multiply/add and multiply/subtract
16861 instructions in the floating-point option. This has no effect if the
16862 floating-point option is not also enabled. Disabling fused multiply/add
16863 and multiply/subtract instructions forces the compiler to use separate
16864 instructions for the multiply and add/subtract operations. This may be
16865 desirable in some cases where strict IEEE 754-compliant results are
16866 required: the fused multiply add/subtract instructions do not round the
16867 intermediate result, thereby producing results with @emph{more} bits of
16868 precision than specified by the IEEE standard. Disabling fused multiply
16869 add/subtract instructions also ensures that the program output is not
16870 sensitive to the compiler's ability to combine multiply and add/subtract
16873 @item -mserialize-volatile
16874 @itemx -mno-serialize-volatile
16875 @opindex mserialize-volatile
16876 @opindex mno-serialize-volatile
16877 When this option is enabled, GCC inserts @code{MEMW} instructions before
16878 @code{volatile} memory references to guarantee sequential consistency.
16879 The default is @option{-mserialize-volatile}. Use
16880 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
16882 @item -mtext-section-literals
16883 @itemx -mno-text-section-literals
16884 @opindex mtext-section-literals
16885 @opindex mno-text-section-literals
16886 Control the treatment of literal pools. The default is
16887 @option{-mno-text-section-literals}, which places literals in a separate
16888 section in the output file. This allows the literal pool to be placed
16889 in a data RAM/ROM, and it also allows the linker to combine literal
16890 pools from separate object files to remove redundant literals and
16891 improve code size. With @option{-mtext-section-literals}, the literals
16892 are interspersed in the text section in order to keep them as close as
16893 possible to their references. This may be necessary for large assembly
16896 @item -mtarget-align
16897 @itemx -mno-target-align
16898 @opindex mtarget-align
16899 @opindex mno-target-align
16900 When this option is enabled, GCC instructs the assembler to
16901 automatically align instructions to reduce branch penalties at the
16902 expense of some code density. The assembler attempts to widen density
16903 instructions to align branch targets and the instructions following call
16904 instructions. If there are not enough preceding safe density
16905 instructions to align a target, no widening will be performed. The
16906 default is @option{-mtarget-align}. These options do not affect the
16907 treatment of auto-aligned instructions like @code{LOOP}, which the
16908 assembler will always align, either by widening density instructions or
16909 by inserting no-op instructions.
16912 @itemx -mno-longcalls
16913 @opindex mlongcalls
16914 @opindex mno-longcalls
16915 When this option is enabled, GCC instructs the assembler to translate
16916 direct calls to indirect calls unless it can determine that the target
16917 of a direct call is in the range allowed by the call instruction. This
16918 translation typically occurs for calls to functions in other source
16919 files. Specifically, the assembler translates a direct @code{CALL}
16920 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
16921 The default is @option{-mno-longcalls}. This option should be used in
16922 programs where the call target can potentially be out of range. This
16923 option is implemented in the assembler, not the compiler, so the
16924 assembly code generated by GCC will still show direct call
16925 instructions---look at the disassembled object code to see the actual
16926 instructions. Note that the assembler will use an indirect call for
16927 every cross-file call, not just those that really will be out of range.
16930 @node zSeries Options
16931 @subsection zSeries Options
16932 @cindex zSeries options
16934 These are listed under @xref{S/390 and zSeries Options}.
16936 @node Code Gen Options
16937 @section Options for Code Generation Conventions
16938 @cindex code generation conventions
16939 @cindex options, code generation
16940 @cindex run-time options
16942 These machine-independent options control the interface conventions
16943 used in code generation.
16945 Most of them have both positive and negative forms; the negative form
16946 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
16947 one of the forms is listed---the one which is not the default. You
16948 can figure out the other form by either removing @samp{no-} or adding
16952 @item -fbounds-check
16953 @opindex fbounds-check
16954 For front-ends that support it, generate additional code to check that
16955 indices used to access arrays are within the declared range. This is
16956 currently only supported by the Java and Fortran front-ends, where
16957 this option defaults to true and false respectively.
16961 This option generates traps for signed overflow on addition, subtraction,
16962 multiplication operations.
16966 This option instructs the compiler to assume that signed arithmetic
16967 overflow of addition, subtraction and multiplication wraps around
16968 using twos-complement representation. This flag enables some optimizations
16969 and disables others. This option is enabled by default for the Java
16970 front-end, as required by the Java language specification.
16973 @opindex fexceptions
16974 Enable exception handling. Generates extra code needed to propagate
16975 exceptions. For some targets, this implies GCC will generate frame
16976 unwind information for all functions, which can produce significant data
16977 size overhead, although it does not affect execution. If you do not
16978 specify this option, GCC will enable it by default for languages like
16979 C++ which normally require exception handling, and disable it for
16980 languages like C that do not normally require it. However, you may need
16981 to enable this option when compiling C code that needs to interoperate
16982 properly with exception handlers written in C++. You may also wish to
16983 disable this option if you are compiling older C++ programs that don't
16984 use exception handling.
16986 @item -fnon-call-exceptions
16987 @opindex fnon-call-exceptions
16988 Generate code that allows trapping instructions to throw exceptions.
16989 Note that this requires platform-specific runtime support that does
16990 not exist everywhere. Moreover, it only allows @emph{trapping}
16991 instructions to throw exceptions, i.e.@: memory references or floating
16992 point instructions. It does not allow exceptions to be thrown from
16993 arbitrary signal handlers such as @code{SIGALRM}.
16995 @item -funwind-tables
16996 @opindex funwind-tables
16997 Similar to @option{-fexceptions}, except that it will just generate any needed
16998 static data, but will not affect the generated code in any other way.
16999 You will normally not enable this option; instead, a language processor
17000 that needs this handling would enable it on your behalf.
17002 @item -fasynchronous-unwind-tables
17003 @opindex fasynchronous-unwind-tables
17004 Generate unwind table in dwarf2 format, if supported by target machine. The
17005 table is exact at each instruction boundary, so it can be used for stack
17006 unwinding from asynchronous events (such as debugger or garbage collector).
17008 @item -fpcc-struct-return
17009 @opindex fpcc-struct-return
17010 Return ``short'' @code{struct} and @code{union} values in memory like
17011 longer ones, rather than in registers. This convention is less
17012 efficient, but it has the advantage of allowing intercallability between
17013 GCC-compiled files and files compiled with other compilers, particularly
17014 the Portable C Compiler (pcc).
17016 The precise convention for returning structures in memory depends
17017 on the target configuration macros.
17019 Short structures and unions are those whose size and alignment match
17020 that of some integer type.
17022 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
17023 switch is not binary compatible with code compiled with the
17024 @option{-freg-struct-return} switch.
17025 Use it to conform to a non-default application binary interface.
17027 @item -freg-struct-return
17028 @opindex freg-struct-return
17029 Return @code{struct} and @code{union} values in registers when possible.
17030 This is more efficient for small structures than
17031 @option{-fpcc-struct-return}.
17033 If you specify neither @option{-fpcc-struct-return} nor
17034 @option{-freg-struct-return}, GCC defaults to whichever convention is
17035 standard for the target. If there is no standard convention, GCC
17036 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
17037 the principal compiler. In those cases, we can choose the standard, and
17038 we chose the more efficient register return alternative.
17040 @strong{Warning:} code compiled with the @option{-freg-struct-return}
17041 switch is not binary compatible with code compiled with the
17042 @option{-fpcc-struct-return} switch.
17043 Use it to conform to a non-default application binary interface.
17045 @item -fshort-enums
17046 @opindex fshort-enums
17047 Allocate to an @code{enum} type only as many bytes as it needs for the
17048 declared range of possible values. Specifically, the @code{enum} type
17049 will be equivalent to the smallest integer type which has enough room.
17051 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
17052 code that is not binary compatible with code generated without that switch.
17053 Use it to conform to a non-default application binary interface.
17055 @item -fshort-double
17056 @opindex fshort-double
17057 Use the same size for @code{double} as for @code{float}.
17059 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
17060 code that is not binary compatible with code generated without that switch.
17061 Use it to conform to a non-default application binary interface.
17063 @item -fshort-wchar
17064 @opindex fshort-wchar
17065 Override the underlying type for @samp{wchar_t} to be @samp{short
17066 unsigned int} instead of the default for the target. This option is
17067 useful for building programs to run under WINE@.
17069 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
17070 code that is not binary compatible with code generated without that switch.
17071 Use it to conform to a non-default application binary interface.
17074 @opindex fno-common
17075 In C code, controls the placement of uninitialized global variables.
17076 Unix C compilers have traditionally permitted multiple definitions of
17077 such variables in different compilation units by placing the variables
17079 This is the behavior specified by @option{-fcommon}, and is the default
17080 for GCC on most targets.
17081 On the other hand, this behavior is not required by ISO C, and on some
17082 targets may carry a speed or code size penalty on variable references.
17083 The @option{-fno-common} option specifies that the compiler should place
17084 uninitialized global variables in the data section of the object file,
17085 rather than generating them as common blocks.
17086 This has the effect that if the same variable is declared
17087 (without @code{extern}) in two different compilations,
17088 you will get a multiple-definition error when you link them.
17089 In this case, you must compile with @option{-fcommon} instead.
17090 Compiling with @option{-fno-common} is useful on targets for which
17091 it provides better performance, or if you wish to verify that the
17092 program will work on other systems which always treat uninitialized
17093 variable declarations this way.
17097 Ignore the @samp{#ident} directive.
17099 @item -finhibit-size-directive
17100 @opindex finhibit-size-directive
17101 Don't output a @code{.size} assembler directive, or anything else that
17102 would cause trouble if the function is split in the middle, and the
17103 two halves are placed at locations far apart in memory. This option is
17104 used when compiling @file{crtstuff.c}; you should not need to use it
17107 @item -fverbose-asm
17108 @opindex fverbose-asm
17109 Put extra commentary information in the generated assembly code to
17110 make it more readable. This option is generally only of use to those
17111 who actually need to read the generated assembly code (perhaps while
17112 debugging the compiler itself).
17114 @option{-fno-verbose-asm}, the default, causes the
17115 extra information to be omitted and is useful when comparing two assembler
17118 @item -frecord-gcc-switches
17119 @opindex frecord-gcc-switches
17120 This switch causes the command line that was used to invoke the
17121 compiler to be recorded into the object file that is being created.
17122 This switch is only implemented on some targets and the exact format
17123 of the recording is target and binary file format dependent, but it
17124 usually takes the form of a section containing ASCII text. This
17125 switch is related to the @option{-fverbose-asm} switch, but that
17126 switch only records information in the assembler output file as
17127 comments, so it never reaches the object file.
17131 @cindex global offset table
17133 Generate position-independent code (PIC) suitable for use in a shared
17134 library, if supported for the target machine. Such code accesses all
17135 constant addresses through a global offset table (GOT)@. The dynamic
17136 loader resolves the GOT entries when the program starts (the dynamic
17137 loader is not part of GCC; it is part of the operating system). If
17138 the GOT size for the linked executable exceeds a machine-specific
17139 maximum size, you get an error message from the linker indicating that
17140 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
17141 instead. (These maximums are 8k on the SPARC and 32k
17142 on the m68k and RS/6000. The 386 has no such limit.)
17144 Position-independent code requires special support, and therefore works
17145 only on certain machines. For the 386, GCC supports PIC for System V
17146 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
17147 position-independent.
17149 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
17154 If supported for the target machine, emit position-independent code,
17155 suitable for dynamic linking and avoiding any limit on the size of the
17156 global offset table. This option makes a difference on the m68k,
17157 PowerPC and SPARC@.
17159 Position-independent code requires special support, and therefore works
17160 only on certain machines.
17162 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
17169 These options are similar to @option{-fpic} and @option{-fPIC}, but
17170 generated position independent code can be only linked into executables.
17171 Usually these options are used when @option{-pie} GCC option will be
17172 used during linking.
17174 @option{-fpie} and @option{-fPIE} both define the macros
17175 @code{__pie__} and @code{__PIE__}. The macros have the value 1
17176 for @option{-fpie} and 2 for @option{-fPIE}.
17178 @item -fno-jump-tables
17179 @opindex fno-jump-tables
17180 Do not use jump tables for switch statements even where it would be
17181 more efficient than other code generation strategies. This option is
17182 of use in conjunction with @option{-fpic} or @option{-fPIC} for
17183 building code which forms part of a dynamic linker and cannot
17184 reference the address of a jump table. On some targets, jump tables
17185 do not require a GOT and this option is not needed.
17187 @item -ffixed-@var{reg}
17189 Treat the register named @var{reg} as a fixed register; generated code
17190 should never refer to it (except perhaps as a stack pointer, frame
17191 pointer or in some other fixed role).
17193 @var{reg} must be the name of a register. The register names accepted
17194 are machine-specific and are defined in the @code{REGISTER_NAMES}
17195 macro in the machine description macro file.
17197 This flag does not have a negative form, because it specifies a
17200 @item -fcall-used-@var{reg}
17201 @opindex fcall-used
17202 Treat the register named @var{reg} as an allocable register that is
17203 clobbered by function calls. It may be allocated for temporaries or
17204 variables that do not live across a call. Functions compiled this way
17205 will not save and restore the register @var{reg}.
17207 It is an error to used this flag with the frame pointer or stack pointer.
17208 Use of this flag for other registers that have fixed pervasive roles in
17209 the machine's execution model will produce disastrous results.
17211 This flag does not have a negative form, because it specifies a
17214 @item -fcall-saved-@var{reg}
17215 @opindex fcall-saved
17216 Treat the register named @var{reg} as an allocable register saved by
17217 functions. It may be allocated even for temporaries or variables that
17218 live across a call. Functions compiled this way will save and restore
17219 the register @var{reg} if they use it.
17221 It is an error to used this flag with the frame pointer or stack pointer.
17222 Use of this flag for other registers that have fixed pervasive roles in
17223 the machine's execution model will produce disastrous results.
17225 A different sort of disaster will result from the use of this flag for
17226 a register in which function values may be returned.
17228 This flag does not have a negative form, because it specifies a
17231 @item -fpack-struct[=@var{n}]
17232 @opindex fpack-struct
17233 Without a value specified, pack all structure members together without
17234 holes. When a value is specified (which must be a small power of two), pack
17235 structure members according to this value, representing the maximum
17236 alignment (that is, objects with default alignment requirements larger than
17237 this will be output potentially unaligned at the next fitting location.
17239 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
17240 code that is not binary compatible with code generated without that switch.
17241 Additionally, it makes the code suboptimal.
17242 Use it to conform to a non-default application binary interface.
17244 @item -finstrument-functions
17245 @opindex finstrument-functions
17246 Generate instrumentation calls for entry and exit to functions. Just
17247 after function entry and just before function exit, the following
17248 profiling functions will be called with the address of the current
17249 function and its call site. (On some platforms,
17250 @code{__builtin_return_address} does not work beyond the current
17251 function, so the call site information may not be available to the
17252 profiling functions otherwise.)
17255 void __cyg_profile_func_enter (void *this_fn,
17257 void __cyg_profile_func_exit (void *this_fn,
17261 The first argument is the address of the start of the current function,
17262 which may be looked up exactly in the symbol table.
17264 This instrumentation is also done for functions expanded inline in other
17265 functions. The profiling calls will indicate where, conceptually, the
17266 inline function is entered and exited. This means that addressable
17267 versions of such functions must be available. If all your uses of a
17268 function are expanded inline, this may mean an additional expansion of
17269 code size. If you use @samp{extern inline} in your C code, an
17270 addressable version of such functions must be provided. (This is
17271 normally the case anyways, but if you get lucky and the optimizer always
17272 expands the functions inline, you might have gotten away without
17273 providing static copies.)
17275 A function may be given the attribute @code{no_instrument_function}, in
17276 which case this instrumentation will not be done. This can be used, for
17277 example, for the profiling functions listed above, high-priority
17278 interrupt routines, and any functions from which the profiling functions
17279 cannot safely be called (perhaps signal handlers, if the profiling
17280 routines generate output or allocate memory).
17282 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
17283 @opindex finstrument-functions-exclude-file-list
17285 Set the list of functions that are excluded from instrumentation (see
17286 the description of @code{-finstrument-functions}). If the file that
17287 contains a function definition matches with one of @var{file}, then
17288 that function is not instrumented. The match is done on substrings:
17289 if the @var{file} parameter is a substring of the file name, it is
17290 considered to be a match.
17293 @code{-finstrument-functions-exclude-file-list=/bits/stl,include/sys}
17294 will exclude any inline function defined in files whose pathnames
17295 contain @code{/bits/stl} or @code{include/sys}.
17297 If, for some reason, you want to include letter @code{','} in one of
17298 @var{sym}, write @code{'\,'}. For example,
17299 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
17300 (note the single quote surrounding the option).
17302 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
17303 @opindex finstrument-functions-exclude-function-list
17305 This is similar to @code{-finstrument-functions-exclude-file-list},
17306 but this option sets the list of function names to be excluded from
17307 instrumentation. The function name to be matched is its user-visible
17308 name, such as @code{vector<int> blah(const vector<int> &)}, not the
17309 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
17310 match is done on substrings: if the @var{sym} parameter is a substring
17311 of the function name, it is considered to be a match. For C99 and C++
17312 extended identifiers, the function name must be given in UTF-8, not
17313 using universal character names.
17315 @item -fstack-check
17316 @opindex fstack-check
17317 Generate code to verify that you do not go beyond the boundary of the
17318 stack. You should specify this flag if you are running in an
17319 environment with multiple threads, but only rarely need to specify it in
17320 a single-threaded environment since stack overflow is automatically
17321 detected on nearly all systems if there is only one stack.
17323 Note that this switch does not actually cause checking to be done; the
17324 operating system or the language runtime must do that. The switch causes
17325 generation of code to ensure that they see the stack being extended.
17327 You can additionally specify a string parameter: @code{no} means no
17328 checking, @code{generic} means force the use of old-style checking,
17329 @code{specific} means use the best checking method and is equivalent
17330 to bare @option{-fstack-check}.
17332 Old-style checking is a generic mechanism that requires no specific
17333 target support in the compiler but comes with the following drawbacks:
17337 Modified allocation strategy for large objects: they will always be
17338 allocated dynamically if their size exceeds a fixed threshold.
17341 Fixed limit on the size of the static frame of functions: when it is
17342 topped by a particular function, stack checking is not reliable and
17343 a warning is issued by the compiler.
17346 Inefficiency: because of both the modified allocation strategy and the
17347 generic implementation, the performances of the code are hampered.
17350 Note that old-style stack checking is also the fallback method for
17351 @code{specific} if no target support has been added in the compiler.
17353 @item -fstack-limit-register=@var{reg}
17354 @itemx -fstack-limit-symbol=@var{sym}
17355 @itemx -fno-stack-limit
17356 @opindex fstack-limit-register
17357 @opindex fstack-limit-symbol
17358 @opindex fno-stack-limit
17359 Generate code to ensure that the stack does not grow beyond a certain value,
17360 either the value of a register or the address of a symbol. If the stack
17361 would grow beyond the value, a signal is raised. For most targets,
17362 the signal is raised before the stack overruns the boundary, so
17363 it is possible to catch the signal without taking special precautions.
17365 For instance, if the stack starts at absolute address @samp{0x80000000}
17366 and grows downwards, you can use the flags
17367 @option{-fstack-limit-symbol=__stack_limit} and
17368 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
17369 of 128KB@. Note that this may only work with the GNU linker.
17371 @cindex aliasing of parameters
17372 @cindex parameters, aliased
17373 @item -fargument-alias
17374 @itemx -fargument-noalias
17375 @itemx -fargument-noalias-global
17376 @itemx -fargument-noalias-anything
17377 @opindex fargument-alias
17378 @opindex fargument-noalias
17379 @opindex fargument-noalias-global
17380 @opindex fargument-noalias-anything
17381 Specify the possible relationships among parameters and between
17382 parameters and global data.
17384 @option{-fargument-alias} specifies that arguments (parameters) may
17385 alias each other and may alias global storage.@*
17386 @option{-fargument-noalias} specifies that arguments do not alias
17387 each other, but may alias global storage.@*
17388 @option{-fargument-noalias-global} specifies that arguments do not
17389 alias each other and do not alias global storage.
17390 @option{-fargument-noalias-anything} specifies that arguments do not
17391 alias any other storage.
17393 Each language will automatically use whatever option is required by
17394 the language standard. You should not need to use these options yourself.
17396 @item -fleading-underscore
17397 @opindex fleading-underscore
17398 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
17399 change the way C symbols are represented in the object file. One use
17400 is to help link with legacy assembly code.
17402 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
17403 generate code that is not binary compatible with code generated without that
17404 switch. Use it to conform to a non-default application binary interface.
17405 Not all targets provide complete support for this switch.
17407 @item -ftls-model=@var{model}
17408 @opindex ftls-model
17409 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
17410 The @var{model} argument should be one of @code{global-dynamic},
17411 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
17413 The default without @option{-fpic} is @code{initial-exec}; with
17414 @option{-fpic} the default is @code{global-dynamic}.
17416 @item -fvisibility=@var{default|internal|hidden|protected}
17417 @opindex fvisibility
17418 Set the default ELF image symbol visibility to the specified option---all
17419 symbols will be marked with this unless overridden within the code.
17420 Using this feature can very substantially improve linking and
17421 load times of shared object libraries, produce more optimized
17422 code, provide near-perfect API export and prevent symbol clashes.
17423 It is @strong{strongly} recommended that you use this in any shared objects
17426 Despite the nomenclature, @code{default} always means public ie;
17427 available to be linked against from outside the shared object.
17428 @code{protected} and @code{internal} are pretty useless in real-world
17429 usage so the only other commonly used option will be @code{hidden}.
17430 The default if @option{-fvisibility} isn't specified is
17431 @code{default}, i.e., make every
17432 symbol public---this causes the same behavior as previous versions of
17435 A good explanation of the benefits offered by ensuring ELF
17436 symbols have the correct visibility is given by ``How To Write
17437 Shared Libraries'' by Ulrich Drepper (which can be found at
17438 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
17439 solution made possible by this option to marking things hidden when
17440 the default is public is to make the default hidden and mark things
17441 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
17442 and @code{__attribute__ ((visibility("default")))} instead of
17443 @code{__declspec(dllexport)} you get almost identical semantics with
17444 identical syntax. This is a great boon to those working with
17445 cross-platform projects.
17447 For those adding visibility support to existing code, you may find
17448 @samp{#pragma GCC visibility} of use. This works by you enclosing
17449 the declarations you wish to set visibility for with (for example)
17450 @samp{#pragma GCC visibility push(hidden)} and
17451 @samp{#pragma GCC visibility pop}.
17452 Bear in mind that symbol visibility should be viewed @strong{as
17453 part of the API interface contract} and thus all new code should
17454 always specify visibility when it is not the default ie; declarations
17455 only for use within the local DSO should @strong{always} be marked explicitly
17456 as hidden as so to avoid PLT indirection overheads---making this
17457 abundantly clear also aids readability and self-documentation of the code.
17458 Note that due to ISO C++ specification requirements, operator new and
17459 operator delete must always be of default visibility.
17461 Be aware that headers from outside your project, in particular system
17462 headers and headers from any other library you use, may not be
17463 expecting to be compiled with visibility other than the default. You
17464 may need to explicitly say @samp{#pragma GCC visibility push(default)}
17465 before including any such headers.
17467 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
17468 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
17469 no modifications. However, this means that calls to @samp{extern}
17470 functions with no explicit visibility will use the PLT, so it is more
17471 effective to use @samp{__attribute ((visibility))} and/or
17472 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
17473 declarations should be treated as hidden.
17475 Note that @samp{-fvisibility} does affect C++ vague linkage
17476 entities. This means that, for instance, an exception class that will
17477 be thrown between DSOs must be explicitly marked with default
17478 visibility so that the @samp{type_info} nodes will be unified between
17481 An overview of these techniques, their benefits and how to use them
17482 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
17488 @node Environment Variables
17489 @section Environment Variables Affecting GCC
17490 @cindex environment variables
17492 @c man begin ENVIRONMENT
17493 This section describes several environment variables that affect how GCC
17494 operates. Some of them work by specifying directories or prefixes to use
17495 when searching for various kinds of files. Some are used to specify other
17496 aspects of the compilation environment.
17498 Note that you can also specify places to search using options such as
17499 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
17500 take precedence over places specified using environment variables, which
17501 in turn take precedence over those specified by the configuration of GCC@.
17502 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
17503 GNU Compiler Collection (GCC) Internals}.
17508 @c @itemx LC_COLLATE
17510 @c @itemx LC_MONETARY
17511 @c @itemx LC_NUMERIC
17516 @c @findex LC_COLLATE
17517 @findex LC_MESSAGES
17518 @c @findex LC_MONETARY
17519 @c @findex LC_NUMERIC
17523 These environment variables control the way that GCC uses
17524 localization information that allow GCC to work with different
17525 national conventions. GCC inspects the locale categories
17526 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
17527 so. These locale categories can be set to any value supported by your
17528 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
17529 Kingdom encoded in UTF-8.
17531 The @env{LC_CTYPE} environment variable specifies character
17532 classification. GCC uses it to determine the character boundaries in
17533 a string; this is needed for some multibyte encodings that contain quote
17534 and escape characters that would otherwise be interpreted as a string
17537 The @env{LC_MESSAGES} environment variable specifies the language to
17538 use in diagnostic messages.
17540 If the @env{LC_ALL} environment variable is set, it overrides the value
17541 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
17542 and @env{LC_MESSAGES} default to the value of the @env{LANG}
17543 environment variable. If none of these variables are set, GCC
17544 defaults to traditional C English behavior.
17548 If @env{TMPDIR} is set, it specifies the directory to use for temporary
17549 files. GCC uses temporary files to hold the output of one stage of
17550 compilation which is to be used as input to the next stage: for example,
17551 the output of the preprocessor, which is the input to the compiler
17554 @item GCC_EXEC_PREFIX
17555 @findex GCC_EXEC_PREFIX
17556 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
17557 names of the subprograms executed by the compiler. No slash is added
17558 when this prefix is combined with the name of a subprogram, but you can
17559 specify a prefix that ends with a slash if you wish.
17561 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
17562 an appropriate prefix to use based on the pathname it was invoked with.
17564 If GCC cannot find the subprogram using the specified prefix, it
17565 tries looking in the usual places for the subprogram.
17567 The default value of @env{GCC_EXEC_PREFIX} is
17568 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
17569 the installed compiler. In many cases @var{prefix} is the value
17570 of @code{prefix} when you ran the @file{configure} script.
17572 Other prefixes specified with @option{-B} take precedence over this prefix.
17574 This prefix is also used for finding files such as @file{crt0.o} that are
17577 In addition, the prefix is used in an unusual way in finding the
17578 directories to search for header files. For each of the standard
17579 directories whose name normally begins with @samp{/usr/local/lib/gcc}
17580 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
17581 replacing that beginning with the specified prefix to produce an
17582 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
17583 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
17584 These alternate directories are searched first; the standard directories
17585 come next. If a standard directory begins with the configured
17586 @var{prefix} then the value of @var{prefix} is replaced by
17587 @env{GCC_EXEC_PREFIX} when looking for header files.
17589 @item COMPILER_PATH
17590 @findex COMPILER_PATH
17591 The value of @env{COMPILER_PATH} is a colon-separated list of
17592 directories, much like @env{PATH}. GCC tries the directories thus
17593 specified when searching for subprograms, if it can't find the
17594 subprograms using @env{GCC_EXEC_PREFIX}.
17597 @findex LIBRARY_PATH
17598 The value of @env{LIBRARY_PATH} is a colon-separated list of
17599 directories, much like @env{PATH}. When configured as a native compiler,
17600 GCC tries the directories thus specified when searching for special
17601 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
17602 using GCC also uses these directories when searching for ordinary
17603 libraries for the @option{-l} option (but directories specified with
17604 @option{-L} come first).
17608 @cindex locale definition
17609 This variable is used to pass locale information to the compiler. One way in
17610 which this information is used is to determine the character set to be used
17611 when character literals, string literals and comments are parsed in C and C++.
17612 When the compiler is configured to allow multibyte characters,
17613 the following values for @env{LANG} are recognized:
17617 Recognize JIS characters.
17619 Recognize SJIS characters.
17621 Recognize EUCJP characters.
17624 If @env{LANG} is not defined, or if it has some other value, then the
17625 compiler will use mblen and mbtowc as defined by the default locale to
17626 recognize and translate multibyte characters.
17630 Some additional environments variables affect the behavior of the
17633 @include cppenv.texi
17637 @node Precompiled Headers
17638 @section Using Precompiled Headers
17639 @cindex precompiled headers
17640 @cindex speed of compilation
17642 Often large projects have many header files that are included in every
17643 source file. The time the compiler takes to process these header files
17644 over and over again can account for nearly all of the time required to
17645 build the project. To make builds faster, GCC allows users to
17646 `precompile' a header file; then, if builds can use the precompiled
17647 header file they will be much faster.
17649 To create a precompiled header file, simply compile it as you would any
17650 other file, if necessary using the @option{-x} option to make the driver
17651 treat it as a C or C++ header file. You will probably want to use a
17652 tool like @command{make} to keep the precompiled header up-to-date when
17653 the headers it contains change.
17655 A precompiled header file will be searched for when @code{#include} is
17656 seen in the compilation. As it searches for the included file
17657 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
17658 compiler looks for a precompiled header in each directory just before it
17659 looks for the include file in that directory. The name searched for is
17660 the name specified in the @code{#include} with @samp{.gch} appended. If
17661 the precompiled header file can't be used, it is ignored.
17663 For instance, if you have @code{#include "all.h"}, and you have
17664 @file{all.h.gch} in the same directory as @file{all.h}, then the
17665 precompiled header file will be used if possible, and the original
17666 header will be used otherwise.
17668 Alternatively, you might decide to put the precompiled header file in a
17669 directory and use @option{-I} to ensure that directory is searched
17670 before (or instead of) the directory containing the original header.
17671 Then, if you want to check that the precompiled header file is always
17672 used, you can put a file of the same name as the original header in this
17673 directory containing an @code{#error} command.
17675 This also works with @option{-include}. So yet another way to use
17676 precompiled headers, good for projects not designed with precompiled
17677 header files in mind, is to simply take most of the header files used by
17678 a project, include them from another header file, precompile that header
17679 file, and @option{-include} the precompiled header. If the header files
17680 have guards against multiple inclusion, they will be skipped because
17681 they've already been included (in the precompiled header).
17683 If you need to precompile the same header file for different
17684 languages, targets, or compiler options, you can instead make a
17685 @emph{directory} named like @file{all.h.gch}, and put each precompiled
17686 header in the directory, perhaps using @option{-o}. It doesn't matter
17687 what you call the files in the directory, every precompiled header in
17688 the directory will be considered. The first precompiled header
17689 encountered in the directory that is valid for this compilation will
17690 be used; they're searched in no particular order.
17692 There are many other possibilities, limited only by your imagination,
17693 good sense, and the constraints of your build system.
17695 A precompiled header file can be used only when these conditions apply:
17699 Only one precompiled header can be used in a particular compilation.
17702 A precompiled header can't be used once the first C token is seen. You
17703 can have preprocessor directives before a precompiled header; you can
17704 even include a precompiled header from inside another header, so long as
17705 there are no C tokens before the @code{#include}.
17708 The precompiled header file must be produced for the same language as
17709 the current compilation. You can't use a C precompiled header for a C++
17713 The precompiled header file must have been produced by the same compiler
17714 binary as the current compilation is using.
17717 Any macros defined before the precompiled header is included must
17718 either be defined in the same way as when the precompiled header was
17719 generated, or must not affect the precompiled header, which usually
17720 means that they don't appear in the precompiled header at all.
17722 The @option{-D} option is one way to define a macro before a
17723 precompiled header is included; using a @code{#define} can also do it.
17724 There are also some options that define macros implicitly, like
17725 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
17728 @item If debugging information is output when using the precompiled
17729 header, using @option{-g} or similar, the same kind of debugging information
17730 must have been output when building the precompiled header. However,
17731 a precompiled header built using @option{-g} can be used in a compilation
17732 when no debugging information is being output.
17734 @item The same @option{-m} options must generally be used when building
17735 and using the precompiled header. @xref{Submodel Options},
17736 for any cases where this rule is relaxed.
17738 @item Each of the following options must be the same when building and using
17739 the precompiled header:
17741 @gccoptlist{-fexceptions}
17744 Some other command-line options starting with @option{-f},
17745 @option{-p}, or @option{-O} must be defined in the same way as when
17746 the precompiled header was generated. At present, it's not clear
17747 which options are safe to change and which are not; the safest choice
17748 is to use exactly the same options when generating and using the
17749 precompiled header. The following are known to be safe:
17751 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
17752 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
17753 -fsched-verbose=<number> -fschedule-insns -fvisibility= @gol
17758 For all of these except the last, the compiler will automatically
17759 ignore the precompiled header if the conditions aren't met. If you
17760 find an option combination that doesn't work and doesn't cause the
17761 precompiled header to be ignored, please consider filing a bug report,
17764 If you do use differing options when generating and using the
17765 precompiled header, the actual behavior will be a mixture of the
17766 behavior for the options. For instance, if you use @option{-g} to
17767 generate the precompiled header but not when using it, you may or may
17768 not get debugging information for routines in the precompiled header.