1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
2 @c 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010
3 @c Free Software Foundation, Inc.
4 @c This is part of the GCC manual.
5 @c For copying conditions, see the file gcc.texi.
12 @c man begin COPYRIGHT
13 Copyright @copyright{} 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
14 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010
15 Free Software Foundation, Inc.
17 Permission is granted to copy, distribute and/or modify this document
18 under the terms of the GNU Free Documentation License, Version 1.3 or
19 any later version published by the Free Software Foundation; with the
20 Invariant Sections being ``GNU General Public License'' and ``Funding
21 Free Software'', the Front-Cover texts being (a) (see below), and with
22 the Back-Cover Texts being (b) (see below). A copy of the license is
23 included in the gfdl(7) man page.
25 (a) The FSF's Front-Cover Text is:
29 (b) The FSF's Back-Cover Text is:
31 You have freedom to copy and modify this GNU Manual, like GNU
32 software. Copies published by the Free Software Foundation raise
33 funds for GNU development.
35 @c Set file name and title for the man page.
37 @settitle GNU project C and C++ compiler
39 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
40 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
41 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
42 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
43 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
44 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
45 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
47 Only the most useful options are listed here; see below for the
48 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
51 gpl(7), gfdl(7), fsf-funding(7),
52 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
53 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
54 @file{ld}, @file{binutils} and @file{gdb}.
57 For instructions on reporting bugs, see
61 See the Info entry for @command{gcc}, or
62 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
63 for contributors to GCC@.
68 @chapter GCC Command Options
69 @cindex GCC command options
70 @cindex command options
71 @cindex options, GCC command
73 @c man begin DESCRIPTION
74 When you invoke GCC, it normally does preprocessing, compilation,
75 assembly and linking. The ``overall options'' allow you to stop this
76 process at an intermediate stage. For example, the @option{-c} option
77 says not to run the linker. Then the output consists of object files
78 output by the assembler.
80 Other options are passed on to one stage of processing. Some options
81 control the preprocessor and others the compiler itself. Yet other
82 options control the assembler and linker; most of these are not
83 documented here, since you rarely need to use any of them.
85 @cindex C compilation options
86 Most of the command line options that you can use with GCC are useful
87 for C programs; when an option is only useful with another language
88 (usually C++), the explanation says so explicitly. If the description
89 for a particular option does not mention a source language, you can use
90 that option with all supported languages.
92 @cindex C++ compilation options
93 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
94 options for compiling C++ programs.
96 @cindex grouping options
97 @cindex options, grouping
98 The @command{gcc} program accepts options and file names as operands. Many
99 options have multi-letter names; therefore multiple single-letter options
100 may @emph{not} be grouped: @option{-dv} is very different from @w{@samp{-d
103 @cindex order of options
104 @cindex options, order
105 You can mix options and other arguments. For the most part, the order
106 you use doesn't matter. Order does matter when you use several
107 options of the same kind; for example, if you specify @option{-L} more
108 than once, the directories are searched in the order specified. Also,
109 the placement of the @option{-l} option is significant.
111 Many options have long names starting with @samp{-f} or with
112 @samp{-W}---for example,
113 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
114 these have both positive and negative forms; the negative form of
115 @option{-ffoo} would be @option{-fno-foo}. This manual documents
116 only one of these two forms, whichever one is not the default.
120 @xref{Option Index}, for an index to GCC's options.
123 * Option Summary:: Brief list of all options, without explanations.
124 * Overall Options:: Controlling the kind of output:
125 an executable, object files, assembler files,
126 or preprocessed source.
127 * Invoking G++:: Compiling C++ programs.
128 * C Dialect Options:: Controlling the variant of C language compiled.
129 * C++ Dialect Options:: Variations on C++.
130 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
132 * Language Independent Options:: Controlling how diagnostics should be
134 * Warning Options:: How picky should the compiler be?
135 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
136 * Optimize Options:: How much optimization?
137 * Preprocessor Options:: Controlling header files and macro definitions.
138 Also, getting dependency information for Make.
139 * Assembler Options:: Passing options to the assembler.
140 * Link Options:: Specifying libraries and so on.
141 * Directory Options:: Where to find header files and libraries.
142 Where to find the compiler executable files.
143 * Spec Files:: How to pass switches to sub-processes.
144 * Target Options:: Running a cross-compiler, or an old version of GCC.
145 * Submodel Options:: Specifying minor hardware or convention variations,
146 such as 68010 vs 68020.
147 * Code Gen Options:: Specifying conventions for function calls, data layout
149 * Environment Variables:: Env vars that affect GCC.
150 * Precompiled Headers:: Compiling a header once, and using it many times.
156 @section Option Summary
158 Here is a summary of all the options, grouped by type. Explanations are
159 in the following sections.
162 @item Overall Options
163 @xref{Overall Options,,Options Controlling the Kind of Output}.
164 @gccoptlist{-c -S -E -o @var{file} -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} @gol
168 -fdump-ada-spec@r{[}-slim@r{]}}
170 @item C Language Options
171 @xref{C Dialect Options,,Options Controlling C Dialect}.
172 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
173 -aux-info @var{filename} @gol
174 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
175 -fhosted -ffreestanding -fopenmp -fms-extensions @gol
176 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
177 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
178 -fsigned-bitfields -fsigned-char @gol
179 -funsigned-bitfields -funsigned-char}
181 @item C++ Language Options
182 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
183 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
184 -fconserve-space -ffriend-injection @gol
185 -fno-elide-constructors @gol
186 -fno-enforce-eh-specs @gol
187 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
188 -fno-implicit-templates @gol
189 -fno-implicit-inline-templates @gol
190 -fno-implement-inlines -fms-extensions @gol
191 -fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol
192 -fno-optional-diags -fpermissive @gol
193 -fno-pretty-templates @gol
194 -frepo -fno-rtti -fstats -ftemplate-depth=@var{n} @gol
195 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
196 -fno-default-inline -fvisibility-inlines-hidden @gol
197 -fvisibility-ms-compat @gol
198 -Wabi -Wconversion-null -Wctor-dtor-privacy @gol
199 -Wnoexcept -Wnon-virtual-dtor -Wreorder @gol
200 -Weffc++ -Wstrict-null-sentinel @gol
201 -Wno-non-template-friend -Wold-style-cast @gol
202 -Woverloaded-virtual -Wno-pmf-conversions @gol
205 @item Objective-C and Objective-C++ Language Options
206 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
207 Objective-C and Objective-C++ Dialects}.
208 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
209 -fgnu-runtime -fnext-runtime @gol
210 -fno-nil-receivers @gol
211 -fobjc-call-cxx-cdtors @gol
212 -fobjc-direct-dispatch @gol
213 -fobjc-exceptions @gol
215 -freplace-objc-classes @gol
218 -Wassign-intercept @gol
219 -Wno-protocol -Wselector @gol
220 -Wstrict-selector-match @gol
221 -Wundeclared-selector}
223 @item Language Independent Options
224 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
225 @gccoptlist{-fmessage-length=@var{n} @gol
226 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
227 -fdiagnostics-show-option}
229 @item Warning Options
230 @xref{Warning Options,,Options to Request or Suppress Warnings}.
231 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
232 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
233 -Wno-attributes -Wno-builtin-macro-redefined @gol
234 -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual @gol
235 -Wchar-subscripts -Wclobbered -Wcomment @gol
236 -Wconversion -Wcoverage-mismatch -Wcpp -Wno-deprecated @gol
237 -Wno-deprecated-declarations -Wdisabled-optimization @gol
238 -Wno-div-by-zero -Wempty-body -Wenum-compare -Wno-endif-labels @gol
239 -Werror -Werror=* @gol
240 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
241 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
242 -Wformat-security -Wformat-y2k @gol
243 -Wframe-larger-than=@var{len} -Wjump-misses-init -Wignored-qualifiers @gol
244 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
245 -Winit-self -Winline @gol
246 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
247 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
248 -Wlogical-op -Wlong-long @gol
249 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
250 -Wmissing-format-attribute -Wmissing-include-dirs @gol
252 -Wno-multichar -Wnonnull -Wno-overflow @gol
253 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
254 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
255 -Wpointer-arith -Wno-pointer-to-int-cast @gol
256 -Wredundant-decls @gol
257 -Wreturn-type -Wsequence-point -Wshadow @gol
258 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
259 -Wstrict-aliasing -Wstrict-aliasing=n @gol
260 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
261 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]} @gol
262 -Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand @gol
263 -Wsystem-headers -Wtrampolines -Wtrigraphs -Wtype-limits -Wundef @gol
264 -Wuninitialized -Wunknown-pragmas -Wno-pragmas @gol
265 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
266 -Wunused-label -Wunused-parameter -Wno-unused-result -Wunused-value -Wunused-variable @gol
267 -Wunused-but-set-parameter -Wunused-but-set-variable -Wvariadic-macros -Wvla @gol
268 -Wvolatile-register-var -Wwrite-strings}
270 @item C and Objective-C-only Warning Options
271 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
272 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
273 -Wold-style-declaration -Wold-style-definition @gol
274 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
275 -Wdeclaration-after-statement -Wpointer-sign}
277 @item Debugging Options
278 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
279 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
280 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
281 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
282 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
283 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
284 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
285 -fdump-statistics @gol
287 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
288 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
289 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
291 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
292 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
293 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
294 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
295 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
296 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
297 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
298 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
299 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
300 -fdump-tree-nrv -fdump-tree-vect @gol
301 -fdump-tree-sink @gol
302 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
303 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
304 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
305 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
306 -ftree-vectorizer-verbose=@var{n} @gol
307 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
308 -fdump-final-insns=@var{file} @gol
309 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
310 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
311 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
312 -fenable-icf-debug @gol
313 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
314 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
315 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
316 -ftest-coverage -ftime-report -fvar-tracking @gol
317 -fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol
318 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
319 -ggdb -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
320 -gvms -gxcoff -gxcoff+ @gol
321 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
322 -fdebug-prefix-map=@var{old}=@var{new} @gol
323 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
324 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
325 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
326 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
327 -print-prog-name=@var{program} -print-search-dirs -Q @gol
328 -print-sysroot -print-sysroot-headers-suffix @gol
329 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
331 @item Optimization Options
332 @xref{Optimize Options,,Options that Control Optimization}.
334 -falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
335 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
336 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
337 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
338 -fcheck-data-deps -fconserve-stack -fcprop-registers -fcrossjumping @gol
339 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules -fcx-limited-range @gol
340 -fdata-sections -fdce -fdce @gol
341 -fdelayed-branch -fdelete-null-pointer-checks -fdse -fdse @gol
342 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffast-math @gol
343 -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
344 -fforward-propagate -ffunction-sections @gol
345 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
346 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
347 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
348 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg -fipa-pta @gol
349 -fipa-profile -fipa-pure-const -fipa-reference -fipa-struct-reorg @gol
350 -fira-algorithm=@var{algorithm} @gol
351 -fira-region=@var{region} -fira-coalesce @gol
352 -fira-loop-pressure -fno-ira-share-save-slots @gol
353 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
354 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
355 -floop-block -floop-interchange -floop-strip-mine @gol
356 -floop-parallelize-all -flto -flto-compression-level -flto-report -fltrans @gol
357 -fltrans-output-list -fmerge-all-constants -fmerge-constants -fmodulo-sched @gol
358 -fmodulo-sched-allow-regmoves -fmove-loop-invariants -fmudflap @gol
359 -fmudflapir -fmudflapth -fno-branch-count-reg -fno-default-inline @gol
360 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
361 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
362 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
363 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
364 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
365 -fpartial-inlining -fpeel-loops -fpredictive-commoning -fprefetch-loop-arrays @gol
366 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
367 -fprofile-generate=@var{path} @gol
368 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
369 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
370 -freorder-blocks-and-partition -freorder-functions @gol
371 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
372 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
373 -fsched-spec-load -fsched-spec-load-dangerous @gol
374 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
375 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
376 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
377 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
378 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
379 -fselective-scheduling -fselective-scheduling2 @gol
380 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
381 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
382 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
383 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer @gol
384 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
385 -ftree-copyrename -ftree-dce -ftree-dominator-opts -ftree-dse @gol
386 -ftree-forwprop -ftree-fre -ftree-loop-if-convert -ftree-loop-im @gol
387 -ftree-phiprop -ftree-loop-distribution @gol
388 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
389 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-pta -ftree-reassoc @gol
390 -ftree-sink -ftree-sra -ftree-switch-conversion @gol
391 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
392 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
393 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
394 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
395 -fwhole-program -fwhopr[=@var{n}] -fwpa -fuse-linker-plugin @gol
396 --param @var{name}=@var{value}
397 -O -O0 -O1 -O2 -O3 -Os -Ofast}
399 @item Preprocessor Options
400 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
401 @gccoptlist{-A@var{question}=@var{answer} @gol
402 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
403 -C -dD -dI -dM -dN @gol
404 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
405 -idirafter @var{dir} @gol
406 -include @var{file} -imacros @var{file} @gol
407 -iprefix @var{file} -iwithprefix @var{dir} @gol
408 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
409 -imultilib @var{dir} -isysroot @var{dir} @gol
410 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
411 -P -fworking-directory -remap @gol
412 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
413 -Xpreprocessor @var{option}}
415 @item Assembler Option
416 @xref{Assembler Options,,Passing Options to the Assembler}.
417 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
420 @xref{Link Options,,Options for Linking}.
421 @gccoptlist{@var{object-file-name} -l@var{library} @gol
422 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
423 -s -static -static-libgcc -static-libstdc++ -shared @gol
424 -shared-libgcc -symbolic @gol
425 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
428 @item Directory Options
429 @xref{Directory Options,,Options for Directory Search}.
430 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir}}
431 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I-
434 @item Machine Dependent Options
435 @xref{Submodel Options,,Hardware Models and Configurations}.
436 @c This list is ordered alphanumerically by subsection name.
437 @c Try and put the significant identifier (CPU or system) first,
438 @c so users have a clue at guessing where the ones they want will be.
441 @gccoptlist{-EB -EL @gol
442 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
443 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
446 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
447 -mabi=@var{name} @gol
448 -mapcs-stack-check -mno-apcs-stack-check @gol
449 -mapcs-float -mno-apcs-float @gol
450 -mapcs-reentrant -mno-apcs-reentrant @gol
451 -msched-prolog -mno-sched-prolog @gol
452 -mlittle-endian -mbig-endian -mwords-little-endian @gol
453 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
454 -mfp16-format=@var{name}
455 -mthumb-interwork -mno-thumb-interwork @gol
456 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
457 -mstructure-size-boundary=@var{n} @gol
458 -mabort-on-noreturn @gol
459 -mlong-calls -mno-long-calls @gol
460 -msingle-pic-base -mno-single-pic-base @gol
461 -mpic-register=@var{reg} @gol
462 -mnop-fun-dllimport @gol
463 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
464 -mpoke-function-name @gol
466 -mtpcs-frame -mtpcs-leaf-frame @gol
467 -mcaller-super-interworking -mcallee-super-interworking @gol
469 -mword-relocations @gol
470 -mfix-cortex-m3-ldrd}
473 @gccoptlist{-mmcu=@var{mcu} -mno-interrupts @gol
474 -mcall-prologues -mtiny-stack -mint8}
476 @emph{Blackfin Options}
477 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
478 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
479 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
480 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
481 -mno-id-shared-library -mshared-library-id=@var{n} @gol
482 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
483 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
484 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
488 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
489 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
490 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
491 -mstack-align -mdata-align -mconst-align @gol
492 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
493 -melf -maout -melinux -mlinux -sim -sim2 @gol
494 -mmul-bug-workaround -mno-mul-bug-workaround}
497 @gccoptlist{-mmac -mpush-args}
499 @emph{Darwin Options}
500 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
501 -arch_only -bind_at_load -bundle -bundle_loader @gol
502 -client_name -compatibility_version -current_version @gol
504 -dependency-file -dylib_file -dylinker_install_name @gol
505 -dynamic -dynamiclib -exported_symbols_list @gol
506 -filelist -flat_namespace -force_cpusubtype_ALL @gol
507 -force_flat_namespace -headerpad_max_install_names @gol
509 -image_base -init -install_name -keep_private_externs @gol
510 -multi_module -multiply_defined -multiply_defined_unused @gol
511 -noall_load -no_dead_strip_inits_and_terms @gol
512 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
513 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
514 -private_bundle -read_only_relocs -sectalign @gol
515 -sectobjectsymbols -whyload -seg1addr @gol
516 -sectcreate -sectobjectsymbols -sectorder @gol
517 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
518 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
519 -segprot -segs_read_only_addr -segs_read_write_addr @gol
520 -single_module -static -sub_library -sub_umbrella @gol
521 -twolevel_namespace -umbrella -undefined @gol
522 -unexported_symbols_list -weak_reference_mismatches @gol
523 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
524 -mkernel -mone-byte-bool}
526 @emph{DEC Alpha Options}
527 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
528 -mieee -mieee-with-inexact -mieee-conformant @gol
529 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
530 -mtrap-precision=@var{mode} -mbuild-constants @gol
531 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
532 -mbwx -mmax -mfix -mcix @gol
533 -mfloat-vax -mfloat-ieee @gol
534 -mexplicit-relocs -msmall-data -mlarge-data @gol
535 -msmall-text -mlarge-text @gol
536 -mmemory-latency=@var{time}}
538 @emph{DEC Alpha/VMS Options}
539 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
542 @gccoptlist{-msmall-model -mno-lsim}
545 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
546 -mhard-float -msoft-float @gol
547 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
548 -mdouble -mno-double @gol
549 -mmedia -mno-media -mmuladd -mno-muladd @gol
550 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
551 -mlinked-fp -mlong-calls -malign-labels @gol
552 -mlibrary-pic -macc-4 -macc-8 @gol
553 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
554 -moptimize-membar -mno-optimize-membar @gol
555 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
556 -mvliw-branch -mno-vliw-branch @gol
557 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
558 -mno-nested-cond-exec -mtomcat-stats @gol
562 @emph{GNU/Linux Options}
563 @gccoptlist{-mglibc -muclibc -mbionic -mandroid @gol
564 -tno-android-cc -tno-android-ld}
566 @emph{H8/300 Options}
567 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
570 @gccoptlist{-march=@var{architecture-type} @gol
571 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
572 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
573 -mfixed-range=@var{register-range} @gol
574 -mjump-in-delay -mlinker-opt -mlong-calls @gol
575 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
576 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
577 -mno-jump-in-delay -mno-long-load-store @gol
578 -mno-portable-runtime -mno-soft-float @gol
579 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
580 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
581 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
582 -munix=@var{unix-std} -nolibdld -static -threads}
584 @emph{i386 and x86-64 Options}
585 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
586 -mfpmath=@var{unit} @gol
587 -masm=@var{dialect} -mno-fancy-math-387 @gol
588 -mno-fp-ret-in-387 -msoft-float @gol
589 -mno-wide-multiply -mrtd -malign-double @gol
590 -mpreferred-stack-boundary=@var{num}
591 -mincoming-stack-boundary=@var{num}
592 -mcld -mcx16 -msahf -mmovbe -mcrc32 -mrecip @gol
593 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
594 -maes -mpclmul -mfsgsbase -mrdrnd -mf16c -mfused-madd @gol
595 -msse4a -m3dnow -mpopcnt -mabm -mfma4 -mxop -mlwp @gol
596 -mthreads -mno-align-stringops -minline-all-stringops @gol
597 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
598 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
599 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
600 -mveclibabi=@var{type} -mpc32 -mpc64 -mpc80 -mstackrealign @gol
601 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
602 -mcmodel=@var{code-model} -mabi=@var{name} @gol
603 -m32 -m64 -mlarge-data-threshold=@var{num} @gol
607 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
608 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
609 -mconstant-gp -mauto-pic -mfused-madd @gol
610 -minline-float-divide-min-latency @gol
611 -minline-float-divide-max-throughput @gol
612 -mno-inline-float-divide @gol
613 -minline-int-divide-min-latency @gol
614 -minline-int-divide-max-throughput @gol
615 -mno-inline-int-divide @gol
616 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
617 -mno-inline-sqrt @gol
618 -mdwarf2-asm -mearly-stop-bits @gol
619 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
620 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
621 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
622 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
623 -msched-spec-ldc -msched-spec-control-ldc @gol
624 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
625 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
626 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
627 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
629 @emph{IA-64/VMS Options}
630 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
633 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
634 -msign-extend-enabled -muser-enabled}
636 @emph{M32R/D Options}
637 @gccoptlist{-m32r2 -m32rx -m32r @gol
639 -malign-loops -mno-align-loops @gol
640 -missue-rate=@var{number} @gol
641 -mbranch-cost=@var{number} @gol
642 -mmodel=@var{code-size-model-type} @gol
643 -msdata=@var{sdata-type} @gol
644 -mno-flush-func -mflush-func=@var{name} @gol
645 -mno-flush-trap -mflush-trap=@var{number} @gol
649 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
651 @emph{M680x0 Options}
652 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
653 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
654 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
655 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
656 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
657 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
658 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
659 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
662 @emph{M68hc1x Options}
663 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
664 -mauto-incdec -minmax -mlong-calls -mshort @gol
665 -msoft-reg-count=@var{count}}
668 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
669 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
670 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
671 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
672 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
675 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
676 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
677 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
678 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
682 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
683 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
684 -mips64 -mips64r2 @gol
685 -mips16 -mno-mips16 -mflip-mips16 @gol
686 -minterlink-mips16 -mno-interlink-mips16 @gol
687 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
688 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
689 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
690 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
691 -mfpu=@var{fpu-type} @gol
692 -msmartmips -mno-smartmips @gol
693 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
694 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
695 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
696 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
697 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
698 -membedded-data -mno-embedded-data @gol
699 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
700 -mcode-readable=@var{setting} @gol
701 -msplit-addresses -mno-split-addresses @gol
702 -mexplicit-relocs -mno-explicit-relocs @gol
703 -mcheck-zero-division -mno-check-zero-division @gol
704 -mdivide-traps -mdivide-breaks @gol
705 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
706 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
707 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
708 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
709 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
710 -mflush-func=@var{func} -mno-flush-func @gol
711 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
712 -mfp-exceptions -mno-fp-exceptions @gol
713 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
714 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
717 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
718 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
719 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
720 -mno-base-addresses -msingle-exit -mno-single-exit}
722 @emph{MN10300 Options}
723 @gccoptlist{-mmult-bug -mno-mult-bug @gol
724 -mam33 -mno-am33 @gol
725 -mam33-2 -mno-am33-2 @gol
726 -mreturn-pointer-on-d0 @gol
729 @emph{PDP-11 Options}
730 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
731 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
732 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
733 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
734 -mbranch-expensive -mbranch-cheap @gol
735 -msplit -mno-split -munix-asm -mdec-asm}
737 @emph{picoChip Options}
738 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N}
739 -msymbol-as-address -mno-inefficient-warnings}
741 @emph{PowerPC Options}
742 See RS/6000 and PowerPC Options.
744 @emph{RS/6000 and PowerPC Options}
745 @gccoptlist{-mcpu=@var{cpu-type} @gol
746 -mtune=@var{cpu-type} @gol
747 -mcmodel=@var{code-model} @gol
748 -mpower -mno-power -mpower2 -mno-power2 @gol
749 -mpowerpc -mpowerpc64 -mno-powerpc @gol
750 -maltivec -mno-altivec @gol
751 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
752 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
753 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
754 -mfprnd -mno-fprnd @gol
755 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
756 -mnew-mnemonics -mold-mnemonics @gol
757 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
758 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
759 -malign-power -malign-natural @gol
760 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
761 -msingle-float -mdouble-float -msimple-fpu @gol
762 -mstring -mno-string -mupdate -mno-update @gol
763 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
764 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
765 -mstrict-align -mno-strict-align -mrelocatable @gol
766 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
767 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
768 -mdynamic-no-pic -maltivec -mswdiv @gol
769 -mprioritize-restricted-insns=@var{priority} @gol
770 -msched-costly-dep=@var{dependence_type} @gol
771 -minsert-sched-nops=@var{scheme} @gol
772 -mcall-sysv -mcall-netbsd @gol
773 -maix-struct-return -msvr4-struct-return @gol
774 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
775 -mblock-move-inline-limit=@var{num} @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 @gol
789 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision -mno-recip-precision}
792 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
794 -mbig-endian-data -mlittle-endian-data @gol
797 -mas100-syntax -mno-as100-syntax@gol
799 -mmax-constant-size=@gol
801 -msave-acc-in-interrupts}
803 @emph{S/390 and zSeries Options}
804 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
805 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
806 -mlong-double-64 -mlong-double-128 @gol
807 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
808 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
809 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
810 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
811 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
814 @gccoptlist{-meb -mel @gol
818 -mscore5 -mscore5u -mscore7 -mscore7d}
821 @gccoptlist{-m1 -m2 -m2e @gol
822 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
824 -m4-nofpu -m4-single-only -m4-single -m4 @gol
825 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
826 -m5-64media -m5-64media-nofpu @gol
827 -m5-32media -m5-32media-nofpu @gol
828 -m5-compact -m5-compact-nofpu @gol
829 -mb -ml -mdalign -mrelax @gol
830 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
831 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
832 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
833 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
834 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
835 -maccumulate-outgoing-args -minvalid-symbols}
838 @gccoptlist{-mcpu=@var{cpu-type} @gol
839 -mtune=@var{cpu-type} @gol
840 -mcmodel=@var{code-model} @gol
841 -m32 -m64 -mapp-regs -mno-app-regs @gol
842 -mfaster-structs -mno-faster-structs @gol
843 -mfpu -mno-fpu -mhard-float -msoft-float @gol
844 -mhard-quad-float -msoft-quad-float @gol
845 -mimpure-text -mno-impure-text -mlittle-endian @gol
846 -mstack-bias -mno-stack-bias @gol
847 -munaligned-doubles -mno-unaligned-doubles @gol
848 -mv8plus -mno-v8plus -mvis -mno-vis
849 -threads -pthreads -pthread}
852 @gccoptlist{-mwarn-reloc -merror-reloc @gol
853 -msafe-dma -munsafe-dma @gol
855 -msmall-mem -mlarge-mem -mstdmain @gol
856 -mfixed-range=@var{register-range} @gol
858 -maddress-space-conversion -mno-address-space-conversion @gol
859 -mcache-size=@var{cache-size} @gol
860 -matomic-updates -mno-atomic-updates}
862 @emph{System V Options}
863 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
866 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
867 -mprolog-function -mno-prolog-function -mspace @gol
868 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
869 -mapp-regs -mno-app-regs @gol
870 -mdisable-callt -mno-disable-callt @gol
878 @gccoptlist{-mg -mgnu -munix}
880 @emph{VxWorks Options}
881 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
882 -Xbind-lazy -Xbind-now}
884 @emph{x86-64 Options}
885 See i386 and x86-64 Options.
887 @emph{i386 and x86-64 Windows Options}
888 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll
889 -mnop-fun-dllimport -mthread -municode -mwin32 -mwindows
890 -fno-set-stack-executable}
892 @emph{Xstormy16 Options}
895 @emph{Xtensa Options}
896 @gccoptlist{-mconst16 -mno-const16 @gol
897 -mfused-madd -mno-fused-madd @gol
899 -mserialize-volatile -mno-serialize-volatile @gol
900 -mtext-section-literals -mno-text-section-literals @gol
901 -mtarget-align -mno-target-align @gol
902 -mlongcalls -mno-longcalls}
904 @emph{zSeries Options}
905 See S/390 and zSeries Options.
907 @item Code Generation Options
908 @xref{Code Gen Options,,Options for Code Generation Conventions}.
909 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
910 -ffixed-@var{reg} -fexceptions @gol
911 -fnon-call-exceptions -funwind-tables @gol
912 -fasynchronous-unwind-tables @gol
913 -finhibit-size-directive -finstrument-functions @gol
914 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
915 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
916 -fno-common -fno-ident @gol
917 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
918 -fno-jump-tables @gol
919 -frecord-gcc-switches @gol
920 -freg-struct-return -fshort-enums @gol
921 -fshort-double -fshort-wchar @gol
922 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
923 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
924 -fno-stack-limit @gol
925 -fleading-underscore -ftls-model=@var{model} @gol
926 -ftrapv -fwrapv -fbounds-check @gol
931 * Overall Options:: Controlling the kind of output:
932 an executable, object files, assembler files,
933 or preprocessed source.
934 * C Dialect Options:: Controlling the variant of C language compiled.
935 * C++ Dialect Options:: Variations on C++.
936 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
938 * Language Independent Options:: Controlling how diagnostics should be
940 * Warning Options:: How picky should the compiler be?
941 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
942 * Optimize Options:: How much optimization?
943 * Preprocessor Options:: Controlling header files and macro definitions.
944 Also, getting dependency information for Make.
945 * Assembler Options:: Passing options to the assembler.
946 * Link Options:: Specifying libraries and so on.
947 * Directory Options:: Where to find header files and libraries.
948 Where to find the compiler executable files.
949 * Spec Files:: How to pass switches to sub-processes.
950 * Target Options:: Running a cross-compiler, or an old version of GCC.
953 @node Overall Options
954 @section Options Controlling the Kind of Output
956 Compilation can involve up to four stages: preprocessing, compilation
957 proper, assembly and linking, always in that order. GCC is capable of
958 preprocessing and compiling several files either into several
959 assembler input files, or into one assembler input file; then each
960 assembler input file produces an object file, and linking combines all
961 the object files (those newly compiled, and those specified as input)
962 into an executable file.
964 @cindex file name suffix
965 For any given input file, the file name suffix determines what kind of
970 C source code which must be preprocessed.
973 C source code which should not be preprocessed.
976 C++ source code which should not be preprocessed.
979 Objective-C source code. Note that you must link with the @file{libobjc}
980 library to make an Objective-C program work.
983 Objective-C source code which should not be preprocessed.
987 Objective-C++ source code. Note that you must link with the @file{libobjc}
988 library to make an Objective-C++ program work. Note that @samp{.M} refers
989 to a literal capital M@.
992 Objective-C++ source code which should not be preprocessed.
995 C, C++, Objective-C or Objective-C++ header file to be turned into a
996 precompiled header (default), or C, C++ header file to be turned into an
997 Ada spec (via the @option{-fdump-ada-spec} switch).
1000 @itemx @var{file}.cp
1001 @itemx @var{file}.cxx
1002 @itemx @var{file}.cpp
1003 @itemx @var{file}.CPP
1004 @itemx @var{file}.c++
1006 C++ source code which must be preprocessed. Note that in @samp{.cxx},
1007 the last two letters must both be literally @samp{x}. Likewise,
1008 @samp{.C} refers to a literal capital C@.
1012 Objective-C++ source code which must be preprocessed.
1014 @item @var{file}.mii
1015 Objective-C++ source code which should not be preprocessed.
1019 @itemx @var{file}.hp
1020 @itemx @var{file}.hxx
1021 @itemx @var{file}.hpp
1022 @itemx @var{file}.HPP
1023 @itemx @var{file}.h++
1024 @itemx @var{file}.tcc
1025 C++ header file to be turned into a precompiled header or Ada spec.
1028 @itemx @var{file}.for
1029 @itemx @var{file}.ftn
1030 Fixed form Fortran source code which should not be preprocessed.
1033 @itemx @var{file}.FOR
1034 @itemx @var{file}.fpp
1035 @itemx @var{file}.FPP
1036 @itemx @var{file}.FTN
1037 Fixed form Fortran source code which must be preprocessed (with the traditional
1040 @item @var{file}.f90
1041 @itemx @var{file}.f95
1042 @itemx @var{file}.f03
1043 @itemx @var{file}.f08
1044 Free form Fortran source code which should not be preprocessed.
1046 @item @var{file}.F90
1047 @itemx @var{file}.F95
1048 @itemx @var{file}.F03
1049 @itemx @var{file}.F08
1050 Free form Fortran source code which must be preprocessed (with the
1051 traditional preprocessor).
1053 @c FIXME: Descriptions of Java file types.
1059 @item @var{file}.ads
1060 Ada source code file which contains a library unit declaration (a
1061 declaration of a package, subprogram, or generic, or a generic
1062 instantiation), or a library unit renaming declaration (a package,
1063 generic, or subprogram renaming declaration). Such files are also
1066 @item @var{file}.adb
1067 Ada source code file containing a library unit body (a subprogram or
1068 package body). Such files are also called @dfn{bodies}.
1070 @c GCC also knows about some suffixes for languages not yet included:
1081 @itemx @var{file}.sx
1082 Assembler code which must be preprocessed.
1085 An object file to be fed straight into linking.
1086 Any file name with no recognized suffix is treated this way.
1090 You can specify the input language explicitly with the @option{-x} option:
1093 @item -x @var{language}
1094 Specify explicitly the @var{language} for the following input files
1095 (rather than letting the compiler choose a default based on the file
1096 name suffix). This option applies to all following input files until
1097 the next @option{-x} option. Possible values for @var{language} are:
1099 c c-header c-cpp-output
1100 c++ c++-header c++-cpp-output
1101 objective-c objective-c-header objective-c-cpp-output
1102 objective-c++ objective-c++-header objective-c++-cpp-output
1103 assembler assembler-with-cpp
1105 f77 f77-cpp-input f95 f95-cpp-input
1110 Turn off any specification of a language, so that subsequent files are
1111 handled according to their file name suffixes (as they are if @option{-x}
1112 has not been used at all).
1114 @item -pass-exit-codes
1115 @opindex pass-exit-codes
1116 Normally the @command{gcc} program will exit with the code of 1 if any
1117 phase of the compiler returns a non-success return code. If you specify
1118 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1119 numerically highest error produced by any phase that returned an error
1120 indication. The C, C++, and Fortran frontends return 4, if an internal
1121 compiler error is encountered.
1124 If you only want some of the stages of compilation, you can use
1125 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1126 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1127 @command{gcc} is to stop. Note that some combinations (for example,
1128 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1133 Compile or assemble the source files, but do not link. The linking
1134 stage simply is not done. The ultimate output is in the form of an
1135 object file for each source file.
1137 By default, the object file name for a source file is made by replacing
1138 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1140 Unrecognized input files, not requiring compilation or assembly, are
1145 Stop after the stage of compilation proper; do not assemble. The output
1146 is in the form of an assembler code file for each non-assembler input
1149 By default, the assembler file name for a source file is made by
1150 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1152 Input files that don't require compilation are ignored.
1156 Stop after the preprocessing stage; do not run the compiler proper. The
1157 output is in the form of preprocessed source code, which is sent to the
1160 Input files which don't require preprocessing are ignored.
1162 @cindex output file option
1165 Place output in file @var{file}. This applies regardless to whatever
1166 sort of output is being produced, whether it be an executable file,
1167 an object file, an assembler file or preprocessed C code.
1169 If @option{-o} is not specified, the default is to put an executable
1170 file in @file{a.out}, the object file for
1171 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1172 assembler file in @file{@var{source}.s}, a precompiled header file in
1173 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1178 Print (on standard error output) the commands executed to run the stages
1179 of compilation. Also print the version number of the compiler driver
1180 program and of the preprocessor and the compiler proper.
1184 Like @option{-v} except the commands are not executed and arguments
1185 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1186 This is useful for shell scripts to capture the driver-generated command lines.
1190 Use pipes rather than temporary files for communication between the
1191 various stages of compilation. This fails to work on some systems where
1192 the assembler is unable to read from a pipe; but the GNU assembler has
1197 If you are compiling multiple source files, this option tells the driver
1198 to pass all the source files to the compiler at once (for those
1199 languages for which the compiler can handle this). This will allow
1200 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1201 language for which this is supported is C@. If you pass source files for
1202 multiple languages to the driver, using this option, the driver will invoke
1203 the compiler(s) that support IMA once each, passing each compiler all the
1204 source files appropriate for it. For those languages that do not support
1205 IMA this option will be ignored, and the compiler will be invoked once for
1206 each source file in that language. If you use this option in conjunction
1207 with @option{-save-temps}, the compiler will generate multiple
1209 (one for each source file), but only one (combined) @file{.o} or
1214 Print (on the standard output) a description of the command line options
1215 understood by @command{gcc}. If the @option{-v} option is also specified
1216 then @option{--help} will also be passed on to the various processes
1217 invoked by @command{gcc}, so that they can display the command line options
1218 they accept. If the @option{-Wextra} option has also been specified
1219 (prior to the @option{--help} option), then command line options which
1220 have no documentation associated with them will also be displayed.
1223 @opindex target-help
1224 Print (on the standard output) a description of target-specific command
1225 line options for each tool. For some targets extra target-specific
1226 information may also be printed.
1228 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1229 Print (on the standard output) a description of the command line
1230 options understood by the compiler that fit into all specified classes
1231 and qualifiers. These are the supported classes:
1234 @item @samp{optimizers}
1235 This will display all of the optimization options supported by the
1238 @item @samp{warnings}
1239 This will display all of the options controlling warning messages
1240 produced by the compiler.
1243 This will display target-specific options. Unlike the
1244 @option{--target-help} option however, target-specific options of the
1245 linker and assembler will not be displayed. This is because those
1246 tools do not currently support the extended @option{--help=} syntax.
1249 This will display the values recognized by the @option{--param}
1252 @item @var{language}
1253 This will display the options supported for @var{language}, where
1254 @var{language} is the name of one of the languages supported in this
1258 This will display the options that are common to all languages.
1261 These are the supported qualifiers:
1264 @item @samp{undocumented}
1265 Display only those options which are undocumented.
1268 Display options which take an argument that appears after an equal
1269 sign in the same continuous piece of text, such as:
1270 @samp{--help=target}.
1272 @item @samp{separate}
1273 Display options which take an argument that appears as a separate word
1274 following the original option, such as: @samp{-o output-file}.
1277 Thus for example to display all the undocumented target-specific
1278 switches supported by the compiler the following can be used:
1281 --help=target,undocumented
1284 The sense of a qualifier can be inverted by prefixing it with the
1285 @samp{^} character, so for example to display all binary warning
1286 options (i.e., ones that are either on or off and that do not take an
1287 argument), which have a description the following can be used:
1290 --help=warnings,^joined,^undocumented
1293 The argument to @option{--help=} should not consist solely of inverted
1296 Combining several classes is possible, although this usually
1297 restricts the output by so much that there is nothing to display. One
1298 case where it does work however is when one of the classes is
1299 @var{target}. So for example to display all the target-specific
1300 optimization options the following can be used:
1303 --help=target,optimizers
1306 The @option{--help=} option can be repeated on the command line. Each
1307 successive use will display its requested class of options, skipping
1308 those that have already been displayed.
1310 If the @option{-Q} option appears on the command line before the
1311 @option{--help=} option, then the descriptive text displayed by
1312 @option{--help=} is changed. Instead of describing the displayed
1313 options, an indication is given as to whether the option is enabled,
1314 disabled or set to a specific value (assuming that the compiler
1315 knows this at the point where the @option{--help=} option is used).
1317 Here is a truncated example from the ARM port of @command{gcc}:
1320 % gcc -Q -mabi=2 --help=target -c
1321 The following options are target specific:
1323 -mabort-on-noreturn [disabled]
1327 The output is sensitive to the effects of previous command line
1328 options, so for example it is possible to find out which optimizations
1329 are enabled at @option{-O2} by using:
1332 -Q -O2 --help=optimizers
1335 Alternatively you can discover which binary optimizations are enabled
1336 by @option{-O3} by using:
1339 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1340 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1341 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1344 @item -no-canonical-prefixes
1345 @opindex no-canonical-prefixes
1346 Do not expand any symbolic links, resolve references to @samp{/../}
1347 or @samp{/./}, or make the path absolute when generating a relative
1352 Display the version number and copyrights of the invoked GCC@.
1356 Invoke all subcommands under a wrapper program. It takes a single
1357 comma separated list as an argument, which will be used to invoke
1361 gcc -c t.c -wrapper gdb,--args
1364 This will invoke all subprograms of gcc under "gdb --args",
1365 thus cc1 invocation will be "gdb --args cc1 ...".
1367 @item -fplugin=@var{name}.so
1368 Load the plugin code in file @var{name}.so, assumed to be a
1369 shared object to be dlopen'd by the compiler. The base name of
1370 the shared object file is used to identify the plugin for the
1371 purposes of argument parsing (See
1372 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1373 Each plugin should define the callback functions specified in the
1376 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1377 Define an argument called @var{key} with a value of @var{value}
1378 for the plugin called @var{name}.
1380 @item -fdump-ada-spec@r{[}-slim@r{]}
1381 For C and C++ source and include files, generate corresponding Ada
1382 specs. @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1383 GNAT User's Guide}, which provides detailed documentation on this feature.
1385 @include @value{srcdir}/../libiberty/at-file.texi
1389 @section Compiling C++ Programs
1391 @cindex suffixes for C++ source
1392 @cindex C++ source file suffixes
1393 C++ source files conventionally use one of the suffixes @samp{.C},
1394 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1395 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1396 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1397 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1398 files with these names and compiles them as C++ programs even if you
1399 call the compiler the same way as for compiling C programs (usually
1400 with the name @command{gcc}).
1404 However, the use of @command{gcc} does not add the C++ library.
1405 @command{g++} is a program that calls GCC and treats @samp{.c},
1406 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1407 files unless @option{-x} is used, and automatically specifies linking
1408 against the C++ library. This program is also useful when
1409 precompiling a C header file with a @samp{.h} extension for use in C++
1410 compilations. On many systems, @command{g++} is also installed with
1411 the name @command{c++}.
1413 @cindex invoking @command{g++}
1414 When you compile C++ programs, you may specify many of the same
1415 command-line options that you use for compiling programs in any
1416 language; or command-line options meaningful for C and related
1417 languages; or options that are meaningful only for C++ programs.
1418 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1419 explanations of options for languages related to C@.
1420 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1421 explanations of options that are meaningful only for C++ programs.
1423 @node C Dialect Options
1424 @section Options Controlling C Dialect
1425 @cindex dialect options
1426 @cindex language dialect options
1427 @cindex options, dialect
1429 The following options control the dialect of C (or languages derived
1430 from C, such as C++, Objective-C and Objective-C++) that the compiler
1434 @cindex ANSI support
1438 In C mode, this is equivalent to @samp{-std=c90}. In C++ mode, it is
1439 equivalent to @samp{-std=c++98}.
1441 This turns off certain features of GCC that are incompatible with ISO
1442 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1443 such as the @code{asm} and @code{typeof} keywords, and
1444 predefined macros such as @code{unix} and @code{vax} that identify the
1445 type of system you are using. It also enables the undesirable and
1446 rarely used ISO trigraph feature. For the C compiler,
1447 it disables recognition of C++ style @samp{//} comments as well as
1448 the @code{inline} keyword.
1450 The alternate keywords @code{__asm__}, @code{__extension__},
1451 @code{__inline__} and @code{__typeof__} continue to work despite
1452 @option{-ansi}. You would not want to use them in an ISO C program, of
1453 course, but it is useful to put them in header files that might be included
1454 in compilations done with @option{-ansi}. Alternate predefined macros
1455 such as @code{__unix__} and @code{__vax__} are also available, with or
1456 without @option{-ansi}.
1458 The @option{-ansi} option does not cause non-ISO programs to be
1459 rejected gratuitously. For that, @option{-pedantic} is required in
1460 addition to @option{-ansi}. @xref{Warning Options}.
1462 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1463 option is used. Some header files may notice this macro and refrain
1464 from declaring certain functions or defining certain macros that the
1465 ISO standard doesn't call for; this is to avoid interfering with any
1466 programs that might use these names for other things.
1468 Functions that would normally be built in but do not have semantics
1469 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1470 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1471 built-in functions provided by GCC}, for details of the functions
1476 Determine the language standard. @xref{Standards,,Language Standards
1477 Supported by GCC}, for details of these standard versions. This option
1478 is currently only supported when compiling C or C++.
1480 The compiler can accept several base standards, such as @samp{c90} or
1481 @samp{c++98}, and GNU dialects of those standards, such as
1482 @samp{gnu90} or @samp{gnu++98}. By specifying a base standard, the
1483 compiler will accept all programs following that standard and those
1484 using GNU extensions that do not contradict it. For example,
1485 @samp{-std=c90} turns off certain features of GCC that are
1486 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1487 keywords, but not other GNU extensions that do not have a meaning in
1488 ISO C90, such as omitting the middle term of a @code{?:}
1489 expression. On the other hand, by specifying a GNU dialect of a
1490 standard, all features the compiler support are enabled, even when
1491 those features change the meaning of the base standard and some
1492 strict-conforming programs may be rejected. The particular standard
1493 is used by @option{-pedantic} to identify which features are GNU
1494 extensions given that version of the standard. For example
1495 @samp{-std=gnu90 -pedantic} would warn about C++ style @samp{//}
1496 comments, while @samp{-std=gnu99 -pedantic} would not.
1498 A value for this option must be provided; possible values are
1504 Support all ISO C90 programs (certain GNU extensions that conflict
1505 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1507 @item iso9899:199409
1508 ISO C90 as modified in amendment 1.
1514 ISO C99. Note that this standard is not yet fully supported; see
1515 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1516 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1519 ISO C1X, the draft of the next revision of the ISO C standard.
1520 Support is limited and experimental and features enabled by this
1521 option may be changed or removed if changed in or removed from the
1526 GNU dialect of ISO C90 (including some C99 features). This
1527 is the default for C code.
1531 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1532 this will become the default. The name @samp{gnu9x} is deprecated.
1535 GNU dialect of ISO C1X. Support is limited and experimental and
1536 features enabled by this option may be changed or removed if changed
1537 in or removed from the standard draft.
1540 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1544 GNU dialect of @option{-std=c++98}. This is the default for
1548 The working draft of the upcoming ISO C++0x standard. This option
1549 enables experimental features that are likely to be included in
1550 C++0x. The working draft is constantly changing, and any feature that is
1551 enabled by this flag may be removed from future versions of GCC if it is
1552 not part of the C++0x standard.
1555 GNU dialect of @option{-std=c++0x}. This option enables
1556 experimental features that may be removed in future versions of GCC.
1559 @item -fgnu89-inline
1560 @opindex fgnu89-inline
1561 The option @option{-fgnu89-inline} tells GCC to use the traditional
1562 GNU semantics for @code{inline} functions when in C99 mode.
1563 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1564 is accepted and ignored by GCC versions 4.1.3 up to but not including
1565 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1566 C99 mode. Using this option is roughly equivalent to adding the
1567 @code{gnu_inline} function attribute to all inline functions
1568 (@pxref{Function Attributes}).
1570 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1571 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1572 specifies the default behavior). This option was first supported in
1573 GCC 4.3. This option is not supported in @option{-std=c90} or
1574 @option{-std=gnu90} mode.
1576 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1577 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1578 in effect for @code{inline} functions. @xref{Common Predefined
1579 Macros,,,cpp,The C Preprocessor}.
1581 @item -aux-info @var{filename}
1583 Output to the given filename prototyped declarations for all functions
1584 declared and/or defined in a translation unit, including those in header
1585 files. This option is silently ignored in any language other than C@.
1587 Besides declarations, the file indicates, in comments, the origin of
1588 each declaration (source file and line), whether the declaration was
1589 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1590 @samp{O} for old, respectively, in the first character after the line
1591 number and the colon), and whether it came from a declaration or a
1592 definition (@samp{C} or @samp{F}, respectively, in the following
1593 character). In the case of function definitions, a K&R-style list of
1594 arguments followed by their declarations is also provided, inside
1595 comments, after the declaration.
1599 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1600 keyword, so that code can use these words as identifiers. You can use
1601 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1602 instead. @option{-ansi} implies @option{-fno-asm}.
1604 In C++, this switch only affects the @code{typeof} keyword, since
1605 @code{asm} and @code{inline} are standard keywords. You may want to
1606 use the @option{-fno-gnu-keywords} flag instead, which has the same
1607 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1608 switch only affects the @code{asm} and @code{typeof} keywords, since
1609 @code{inline} is a standard keyword in ISO C99.
1612 @itemx -fno-builtin-@var{function}
1613 @opindex fno-builtin
1614 @cindex built-in functions
1615 Don't recognize built-in functions that do not begin with
1616 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1617 functions provided by GCC}, for details of the functions affected,
1618 including those which are not built-in functions when @option{-ansi} or
1619 @option{-std} options for strict ISO C conformance are used because they
1620 do not have an ISO standard meaning.
1622 GCC normally generates special code to handle certain built-in functions
1623 more efficiently; for instance, calls to @code{alloca} may become single
1624 instructions that adjust the stack directly, and calls to @code{memcpy}
1625 may become inline copy loops. The resulting code is often both smaller
1626 and faster, but since the function calls no longer appear as such, you
1627 cannot set a breakpoint on those calls, nor can you change the behavior
1628 of the functions by linking with a different library. In addition,
1629 when a function is recognized as a built-in function, GCC may use
1630 information about that function to warn about problems with calls to
1631 that function, or to generate more efficient code, even if the
1632 resulting code still contains calls to that function. For example,
1633 warnings are given with @option{-Wformat} for bad calls to
1634 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1635 known not to modify global memory.
1637 With the @option{-fno-builtin-@var{function}} option
1638 only the built-in function @var{function} is
1639 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1640 function is named that is not built-in in this version of GCC, this
1641 option is ignored. There is no corresponding
1642 @option{-fbuiltin-@var{function}} option; if you wish to enable
1643 built-in functions selectively when using @option{-fno-builtin} or
1644 @option{-ffreestanding}, you may define macros such as:
1647 #define abs(n) __builtin_abs ((n))
1648 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1653 @cindex hosted environment
1655 Assert that compilation takes place in a hosted environment. This implies
1656 @option{-fbuiltin}. A hosted environment is one in which the
1657 entire standard library is available, and in which @code{main} has a return
1658 type of @code{int}. Examples are nearly everything except a kernel.
1659 This is equivalent to @option{-fno-freestanding}.
1661 @item -ffreestanding
1662 @opindex ffreestanding
1663 @cindex hosted environment
1665 Assert that compilation takes place in a freestanding environment. This
1666 implies @option{-fno-builtin}. A freestanding environment
1667 is one in which the standard library may not exist, and program startup may
1668 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1669 This is equivalent to @option{-fno-hosted}.
1671 @xref{Standards,,Language Standards Supported by GCC}, for details of
1672 freestanding and hosted environments.
1676 @cindex openmp parallel
1677 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1678 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1679 compiler generates parallel code according to the OpenMP Application
1680 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1681 implies @option{-pthread}, and thus is only supported on targets that
1682 have support for @option{-pthread}.
1684 @item -fms-extensions
1685 @opindex fms-extensions
1686 Accept some non-standard constructs used in Microsoft header files.
1688 Some cases of unnamed fields in structures and unions are only
1689 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1690 fields within structs/unions}, for details.
1694 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1695 options for strict ISO C conformance) implies @option{-trigraphs}.
1697 @item -no-integrated-cpp
1698 @opindex no-integrated-cpp
1699 Performs a compilation in two passes: preprocessing and compiling. This
1700 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1701 @option{-B} option. The user supplied compilation step can then add in
1702 an additional preprocessing step after normal preprocessing but before
1703 compiling. The default is to use the integrated cpp (internal cpp)
1705 The semantics of this option will change if "cc1", "cc1plus", and
1706 "cc1obj" are merged.
1708 @cindex traditional C language
1709 @cindex C language, traditional
1711 @itemx -traditional-cpp
1712 @opindex traditional-cpp
1713 @opindex traditional
1714 Formerly, these options caused GCC to attempt to emulate a pre-standard
1715 C compiler. They are now only supported with the @option{-E} switch.
1716 The preprocessor continues to support a pre-standard mode. See the GNU
1717 CPP manual for details.
1719 @item -fcond-mismatch
1720 @opindex fcond-mismatch
1721 Allow conditional expressions with mismatched types in the second and
1722 third arguments. The value of such an expression is void. This option
1723 is not supported for C++.
1725 @item -flax-vector-conversions
1726 @opindex flax-vector-conversions
1727 Allow implicit conversions between vectors with differing numbers of
1728 elements and/or incompatible element types. This option should not be
1731 @item -funsigned-char
1732 @opindex funsigned-char
1733 Let the type @code{char} be unsigned, like @code{unsigned char}.
1735 Each kind of machine has a default for what @code{char} should
1736 be. It is either like @code{unsigned char} by default or like
1737 @code{signed char} by default.
1739 Ideally, a portable program should always use @code{signed char} or
1740 @code{unsigned char} when it depends on the signedness of an object.
1741 But many programs have been written to use plain @code{char} and
1742 expect it to be signed, or expect it to be unsigned, depending on the
1743 machines they were written for. This option, and its inverse, let you
1744 make such a program work with the opposite default.
1746 The type @code{char} is always a distinct type from each of
1747 @code{signed char} or @code{unsigned char}, even though its behavior
1748 is always just like one of those two.
1751 @opindex fsigned-char
1752 Let the type @code{char} be signed, like @code{signed char}.
1754 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1755 the negative form of @option{-funsigned-char}. Likewise, the option
1756 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1758 @item -fsigned-bitfields
1759 @itemx -funsigned-bitfields
1760 @itemx -fno-signed-bitfields
1761 @itemx -fno-unsigned-bitfields
1762 @opindex fsigned-bitfields
1763 @opindex funsigned-bitfields
1764 @opindex fno-signed-bitfields
1765 @opindex fno-unsigned-bitfields
1766 These options control whether a bit-field is signed or unsigned, when the
1767 declaration does not use either @code{signed} or @code{unsigned}. By
1768 default, such a bit-field is signed, because this is consistent: the
1769 basic integer types such as @code{int} are signed types.
1772 @node C++ Dialect Options
1773 @section Options Controlling C++ Dialect
1775 @cindex compiler options, C++
1776 @cindex C++ options, command line
1777 @cindex options, C++
1778 This section describes the command-line options that are only meaningful
1779 for C++ programs; but you can also use most of the GNU compiler options
1780 regardless of what language your program is in. For example, you
1781 might compile a file @code{firstClass.C} like this:
1784 g++ -g -frepo -O -c firstClass.C
1788 In this example, only @option{-frepo} is an option meant
1789 only for C++ programs; you can use the other options with any
1790 language supported by GCC@.
1792 Here is a list of options that are @emph{only} for compiling C++ programs:
1796 @item -fabi-version=@var{n}
1797 @opindex fabi-version
1798 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1799 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1800 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1801 the version that conforms most closely to the C++ ABI specification.
1802 Therefore, the ABI obtained using version 0 will change as ABI bugs
1805 The default is version 2.
1807 Version 3 corrects an error in mangling a constant address as a
1810 Version 4 implements a standard mangling for vector types.
1812 See also @option{-Wabi}.
1814 @item -fno-access-control
1815 @opindex fno-access-control
1816 Turn off all access checking. This switch is mainly useful for working
1817 around bugs in the access control code.
1821 Check that the pointer returned by @code{operator new} is non-null
1822 before attempting to modify the storage allocated. This check is
1823 normally unnecessary because the C++ standard specifies that
1824 @code{operator new} will only return @code{0} if it is declared
1825 @samp{throw()}, in which case the compiler will always check the
1826 return value even without this option. In all other cases, when
1827 @code{operator new} has a non-empty exception specification, memory
1828 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1829 @samp{new (nothrow)}.
1831 @item -fconserve-space
1832 @opindex fconserve-space
1833 Put uninitialized or runtime-initialized global variables into the
1834 common segment, as C does. This saves space in the executable at the
1835 cost of not diagnosing duplicate definitions. If you compile with this
1836 flag and your program mysteriously crashes after @code{main()} has
1837 completed, you may have an object that is being destroyed twice because
1838 two definitions were merged.
1840 This option is no longer useful on most targets, now that support has
1841 been added for putting variables into BSS without making them common.
1843 @item -fno-deduce-init-list
1844 @opindex fno-deduce-init-list
1845 Disable deduction of a template type parameter as
1846 std::initializer_list from a brace-enclosed initializer list, i.e.
1849 template <class T> auto forward(T t) -> decltype (realfn (t))
1856 forward(@{1,2@}); // call forward<std::initializer_list<int>>
1860 This option is present because this deduction is an extension to the
1861 current specification in the C++0x working draft, and there was
1862 some concern about potential overload resolution problems.
1864 @item -ffriend-injection
1865 @opindex ffriend-injection
1866 Inject friend functions into the enclosing namespace, so that they are
1867 visible outside the scope of the class in which they are declared.
1868 Friend functions were documented to work this way in the old Annotated
1869 C++ Reference Manual, and versions of G++ before 4.1 always worked
1870 that way. However, in ISO C++ a friend function which is not declared
1871 in an enclosing scope can only be found using argument dependent
1872 lookup. This option causes friends to be injected as they were in
1875 This option is for compatibility, and may be removed in a future
1878 @item -fno-elide-constructors
1879 @opindex fno-elide-constructors
1880 The C++ standard allows an implementation to omit creating a temporary
1881 which is only used to initialize another object of the same type.
1882 Specifying this option disables that optimization, and forces G++ to
1883 call the copy constructor in all cases.
1885 @item -fno-enforce-eh-specs
1886 @opindex fno-enforce-eh-specs
1887 Don't generate code to check for violation of exception specifications
1888 at runtime. This option violates the C++ standard, but may be useful
1889 for reducing code size in production builds, much like defining
1890 @samp{NDEBUG}. This does not give user code permission to throw
1891 exceptions in violation of the exception specifications; the compiler
1892 will still optimize based on the specifications, so throwing an
1893 unexpected exception will result in undefined behavior.
1896 @itemx -fno-for-scope
1898 @opindex fno-for-scope
1899 If @option{-ffor-scope} is specified, the scope of variables declared in
1900 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1901 as specified by the C++ standard.
1902 If @option{-fno-for-scope} is specified, the scope of variables declared in
1903 a @i{for-init-statement} extends to the end of the enclosing scope,
1904 as was the case in old versions of G++, and other (traditional)
1905 implementations of C++.
1907 The default if neither flag is given to follow the standard,
1908 but to allow and give a warning for old-style code that would
1909 otherwise be invalid, or have different behavior.
1911 @item -fno-gnu-keywords
1912 @opindex fno-gnu-keywords
1913 Do not recognize @code{typeof} as a keyword, so that code can use this
1914 word as an identifier. You can use the keyword @code{__typeof__} instead.
1915 @option{-ansi} implies @option{-fno-gnu-keywords}.
1917 @item -fno-implicit-templates
1918 @opindex fno-implicit-templates
1919 Never emit code for non-inline templates which are instantiated
1920 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1921 @xref{Template Instantiation}, for more information.
1923 @item -fno-implicit-inline-templates
1924 @opindex fno-implicit-inline-templates
1925 Don't emit code for implicit instantiations of inline templates, either.
1926 The default is to handle inlines differently so that compiles with and
1927 without optimization will need the same set of explicit instantiations.
1929 @item -fno-implement-inlines
1930 @opindex fno-implement-inlines
1931 To save space, do not emit out-of-line copies of inline functions
1932 controlled by @samp{#pragma implementation}. This will cause linker
1933 errors if these functions are not inlined everywhere they are called.
1935 @item -fms-extensions
1936 @opindex fms-extensions
1937 Disable pedantic warnings about constructs used in MFC, such as implicit
1938 int and getting a pointer to member function via non-standard syntax.
1940 @item -fno-nonansi-builtins
1941 @opindex fno-nonansi-builtins
1942 Disable built-in declarations of functions that are not mandated by
1943 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1944 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1947 @opindex fnothrow-opt
1948 Treat a @code{throw()} exception specification as though it were a
1949 @code{noexcept} specification to reduce or eliminate the text size
1950 overhead relative to a function with no exception specification. If
1951 the function has local variables of types with non-trivial
1952 destructors, the exception specification will actually make the
1953 function smaller because the EH cleanups for those variables can be
1954 optimized away. The semantic effect is that an exception thrown out of
1955 a function with such an exception specification will result in a call
1956 to @code{terminate} rather than @code{unexpected}.
1958 @item -fno-operator-names
1959 @opindex fno-operator-names
1960 Do not treat the operator name keywords @code{and}, @code{bitand},
1961 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1962 synonyms as keywords.
1964 @item -fno-optional-diags
1965 @opindex fno-optional-diags
1966 Disable diagnostics that the standard says a compiler does not need to
1967 issue. Currently, the only such diagnostic issued by G++ is the one for
1968 a name having multiple meanings within a class.
1971 @opindex fpermissive
1972 Downgrade some diagnostics about nonconformant code from errors to
1973 warnings. Thus, using @option{-fpermissive} will allow some
1974 nonconforming code to compile.
1976 @item -fno-pretty-templates
1977 @opindex fno-pretty-templates
1978 When an error message refers to a specialization of a function
1979 template, the compiler will normally print the signature of the
1980 template followed by the template arguments and any typedefs or
1981 typenames in the signature (e.g. @code{void f(T) [with T = int]}
1982 rather than @code{void f(int)}) so that it's clear which template is
1983 involved. When an error message refers to a specialization of a class
1984 template, the compiler will omit any template arguments which match
1985 the default template arguments for that template. If either of these
1986 behaviors make it harder to understand the error message rather than
1987 easier, using @option{-fno-pretty-templates} will disable them.
1991 Enable automatic template instantiation at link time. This option also
1992 implies @option{-fno-implicit-templates}. @xref{Template
1993 Instantiation}, for more information.
1997 Disable generation of information about every class with virtual
1998 functions for use by the C++ runtime type identification features
1999 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
2000 of the language, you can save some space by using this flag. Note that
2001 exception handling uses the same information, but it will generate it as
2002 needed. The @samp{dynamic_cast} operator can still be used for casts that
2003 do not require runtime type information, i.e.@: casts to @code{void *} or to
2004 unambiguous base classes.
2008 Emit statistics about front-end processing at the end of the compilation.
2009 This information is generally only useful to the G++ development team.
2011 @item -fstrict-enums
2012 @opindex fstrict-enums
2013 Allow the compiler to optimize using the assumption that a value of
2014 enumeration type can only be one of the values of the enumeration (as
2015 defined in the C++ standard; basically, a value which can be
2016 represented in the minimum number of bits needed to represent all the
2017 enumerators). This assumption may not be valid if the program uses a
2018 cast to convert an arbitrary integer value to the enumeration type.
2020 @item -ftemplate-depth=@var{n}
2021 @opindex ftemplate-depth
2022 Set the maximum instantiation depth for template classes to @var{n}.
2023 A limit on the template instantiation depth is needed to detect
2024 endless recursions during template class instantiation. ANSI/ISO C++
2025 conforming programs must not rely on a maximum depth greater than 17
2026 (changed to 1024 in C++0x).
2028 @item -fno-threadsafe-statics
2029 @opindex fno-threadsafe-statics
2030 Do not emit the extra code to use the routines specified in the C++
2031 ABI for thread-safe initialization of local statics. You can use this
2032 option to reduce code size slightly in code that doesn't need to be
2035 @item -fuse-cxa-atexit
2036 @opindex fuse-cxa-atexit
2037 Register destructors for objects with static storage duration with the
2038 @code{__cxa_atexit} function rather than the @code{atexit} function.
2039 This option is required for fully standards-compliant handling of static
2040 destructors, but will only work if your C library supports
2041 @code{__cxa_atexit}.
2043 @item -fno-use-cxa-get-exception-ptr
2044 @opindex fno-use-cxa-get-exception-ptr
2045 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2046 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
2047 if the runtime routine is not available.
2049 @item -fvisibility-inlines-hidden
2050 @opindex fvisibility-inlines-hidden
2051 This switch declares that the user does not attempt to compare
2052 pointers to inline methods where the addresses of the two functions
2053 were taken in different shared objects.
2055 The effect of this is that GCC may, effectively, mark inline methods with
2056 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2057 appear in the export table of a DSO and do not require a PLT indirection
2058 when used within the DSO@. Enabling this option can have a dramatic effect
2059 on load and link times of a DSO as it massively reduces the size of the
2060 dynamic export table when the library makes heavy use of templates.
2062 The behavior of this switch is not quite the same as marking the
2063 methods as hidden directly, because it does not affect static variables
2064 local to the function or cause the compiler to deduce that
2065 the function is defined in only one shared object.
2067 You may mark a method as having a visibility explicitly to negate the
2068 effect of the switch for that method. For example, if you do want to
2069 compare pointers to a particular inline method, you might mark it as
2070 having default visibility. Marking the enclosing class with explicit
2071 visibility will have no effect.
2073 Explicitly instantiated inline methods are unaffected by this option
2074 as their linkage might otherwise cross a shared library boundary.
2075 @xref{Template Instantiation}.
2077 @item -fvisibility-ms-compat
2078 @opindex fvisibility-ms-compat
2079 This flag attempts to use visibility settings to make GCC's C++
2080 linkage model compatible with that of Microsoft Visual Studio.
2082 The flag makes these changes to GCC's linkage model:
2086 It sets the default visibility to @code{hidden}, like
2087 @option{-fvisibility=hidden}.
2090 Types, but not their members, are not hidden by default.
2093 The One Definition Rule is relaxed for types without explicit
2094 visibility specifications which are defined in more than one different
2095 shared object: those declarations are permitted if they would have
2096 been permitted when this option was not used.
2099 In new code it is better to use @option{-fvisibility=hidden} and
2100 export those classes which are intended to be externally visible.
2101 Unfortunately it is possible for code to rely, perhaps accidentally,
2102 on the Visual Studio behavior.
2104 Among the consequences of these changes are that static data members
2105 of the same type with the same name but defined in different shared
2106 objects will be different, so changing one will not change the other;
2107 and that pointers to function members defined in different shared
2108 objects may not compare equal. When this flag is given, it is a
2109 violation of the ODR to define types with the same name differently.
2113 Do not use weak symbol support, even if it is provided by the linker.
2114 By default, G++ will use weak symbols if they are available. This
2115 option exists only for testing, and should not be used by end-users;
2116 it will result in inferior code and has no benefits. This option may
2117 be removed in a future release of G++.
2121 Do not search for header files in the standard directories specific to
2122 C++, but do still search the other standard directories. (This option
2123 is used when building the C++ library.)
2126 In addition, these optimization, warning, and code generation options
2127 have meanings only for C++ programs:
2130 @item -fno-default-inline
2131 @opindex fno-default-inline
2132 Do not assume @samp{inline} for functions defined inside a class scope.
2133 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2134 functions will have linkage like inline functions; they just won't be
2137 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2140 Warn when G++ generates code that is probably not compatible with the
2141 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2142 all such cases, there are probably some cases that are not warned about,
2143 even though G++ is generating incompatible code. There may also be
2144 cases where warnings are emitted even though the code that is generated
2147 You should rewrite your code to avoid these warnings if you are
2148 concerned about the fact that code generated by G++ may not be binary
2149 compatible with code generated by other compilers.
2151 The known incompatibilities in @option{-fabi-version=2} (the default) include:
2156 A template with a non-type template parameter of reference type is
2157 mangled incorrectly:
2160 template <int &> struct S @{@};
2164 This is fixed in @option{-fabi-version=3}.
2167 SIMD vector types declared using @code{__attribute ((vector_size))} are
2168 mangled in a non-standard way that does not allow for overloading of
2169 functions taking vectors of different sizes.
2171 The mangling is changed in @option{-fabi-version=4}.
2174 The known incompatibilities in @option{-fabi-version=1} include:
2179 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2180 pack data into the same byte as a base class. For example:
2183 struct A @{ virtual void f(); int f1 : 1; @};
2184 struct B : public A @{ int f2 : 1; @};
2188 In this case, G++ will place @code{B::f2} into the same byte
2189 as@code{A::f1}; other compilers will not. You can avoid this problem
2190 by explicitly padding @code{A} so that its size is a multiple of the
2191 byte size on your platform; that will cause G++ and other compilers to
2192 layout @code{B} identically.
2195 Incorrect handling of tail-padding for virtual bases. G++ does not use
2196 tail padding when laying out virtual bases. For example:
2199 struct A @{ virtual void f(); char c1; @};
2200 struct B @{ B(); char c2; @};
2201 struct C : public A, public virtual B @{@};
2205 In this case, G++ will not place @code{B} into the tail-padding for
2206 @code{A}; other compilers will. You can avoid this problem by
2207 explicitly padding @code{A} so that its size is a multiple of its
2208 alignment (ignoring virtual base classes); that will cause G++ and other
2209 compilers to layout @code{C} identically.
2212 Incorrect handling of bit-fields with declared widths greater than that
2213 of their underlying types, when the bit-fields appear in a union. For
2217 union U @{ int i : 4096; @};
2221 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2222 union too small by the number of bits in an @code{int}.
2225 Empty classes can be placed at incorrect offsets. For example:
2235 struct C : public B, public A @{@};
2239 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2240 it should be placed at offset zero. G++ mistakenly believes that the
2241 @code{A} data member of @code{B} is already at offset zero.
2244 Names of template functions whose types involve @code{typename} or
2245 template template parameters can be mangled incorrectly.
2248 template <typename Q>
2249 void f(typename Q::X) @{@}
2251 template <template <typename> class Q>
2252 void f(typename Q<int>::X) @{@}
2256 Instantiations of these templates may be mangled incorrectly.
2260 It also warns psABI related changes. The known psABI changes at this
2266 For SYSV/x86-64, when passing union with long double, it is changed to
2267 pass in memory as specified in psABI. For example:
2277 @code{union U} will always be passed in memory.
2281 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2282 @opindex Wctor-dtor-privacy
2283 @opindex Wno-ctor-dtor-privacy
2284 Warn when a class seems unusable because all the constructors or
2285 destructors in that class are private, and it has neither friends nor
2286 public static member functions.
2288 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2290 @opindex Wno-noexcept
2291 Warn when a noexcept-expression evaluates to false because of a call
2292 to a function that does not have a non-throwing exception
2293 specification (i.e. @samp{throw()} or @samp{noexcept}) but is known by
2294 the compiler to never throw an exception.
2296 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2297 @opindex Wnon-virtual-dtor
2298 @opindex Wno-non-virtual-dtor
2299 Warn when a class has virtual functions and accessible non-virtual
2300 destructor, in which case it would be possible but unsafe to delete
2301 an instance of a derived class through a pointer to the base class.
2302 This warning is also enabled if -Weffc++ is specified.
2304 @item -Wreorder @r{(C++ and Objective-C++ only)}
2306 @opindex Wno-reorder
2307 @cindex reordering, warning
2308 @cindex warning for reordering of member initializers
2309 Warn when the order of member initializers given in the code does not
2310 match the order in which they must be executed. For instance:
2316 A(): j (0), i (1) @{ @}
2320 The compiler will rearrange the member initializers for @samp{i}
2321 and @samp{j} to match the declaration order of the members, emitting
2322 a warning to that effect. This warning is enabled by @option{-Wall}.
2325 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2328 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2331 Warn about violations of the following style guidelines from Scott Meyers'
2332 @cite{Effective C++} book:
2336 Item 11: Define a copy constructor and an assignment operator for classes
2337 with dynamically allocated memory.
2340 Item 12: Prefer initialization to assignment in constructors.
2343 Item 14: Make destructors virtual in base classes.
2346 Item 15: Have @code{operator=} return a reference to @code{*this}.
2349 Item 23: Don't try to return a reference when you must return an object.
2353 Also warn about violations of the following style guidelines from
2354 Scott Meyers' @cite{More Effective C++} book:
2358 Item 6: Distinguish between prefix and postfix forms of increment and
2359 decrement operators.
2362 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2366 When selecting this option, be aware that the standard library
2367 headers do not obey all of these guidelines; use @samp{grep -v}
2368 to filter out those warnings.
2370 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2371 @opindex Wstrict-null-sentinel
2372 @opindex Wno-strict-null-sentinel
2373 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2374 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2375 to @code{__null}. Although it is a null pointer constant not a null pointer,
2376 it is guaranteed to be of the same size as a pointer. But this use is
2377 not portable across different compilers.
2379 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2380 @opindex Wno-non-template-friend
2381 @opindex Wnon-template-friend
2382 Disable warnings when non-templatized friend functions are declared
2383 within a template. Since the advent of explicit template specification
2384 support in G++, if the name of the friend is an unqualified-id (i.e.,
2385 @samp{friend foo(int)}), the C++ language specification demands that the
2386 friend declare or define an ordinary, nontemplate function. (Section
2387 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2388 could be interpreted as a particular specialization of a templatized
2389 function. Because this non-conforming behavior is no longer the default
2390 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2391 check existing code for potential trouble spots and is on by default.
2392 This new compiler behavior can be turned off with
2393 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2394 but disables the helpful warning.
2396 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2397 @opindex Wold-style-cast
2398 @opindex Wno-old-style-cast
2399 Warn if an old-style (C-style) cast to a non-void type is used within
2400 a C++ program. The new-style casts (@samp{dynamic_cast},
2401 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2402 less vulnerable to unintended effects and much easier to search for.
2404 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2405 @opindex Woverloaded-virtual
2406 @opindex Wno-overloaded-virtual
2407 @cindex overloaded virtual fn, warning
2408 @cindex warning for overloaded virtual fn
2409 Warn when a function declaration hides virtual functions from a
2410 base class. For example, in:
2417 struct B: public A @{
2422 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2430 will fail to compile.
2432 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2433 @opindex Wno-pmf-conversions
2434 @opindex Wpmf-conversions
2435 Disable the diagnostic for converting a bound pointer to member function
2438 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2439 @opindex Wsign-promo
2440 @opindex Wno-sign-promo
2441 Warn when overload resolution chooses a promotion from unsigned or
2442 enumerated type to a signed type, over a conversion to an unsigned type of
2443 the same size. Previous versions of G++ would try to preserve
2444 unsignedness, but the standard mandates the current behavior.
2449 A& operator = (int);
2459 In this example, G++ will synthesize a default @samp{A& operator =
2460 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2463 @node Objective-C and Objective-C++ Dialect Options
2464 @section Options Controlling Objective-C and Objective-C++ Dialects
2466 @cindex compiler options, Objective-C and Objective-C++
2467 @cindex Objective-C and Objective-C++ options, command line
2468 @cindex options, Objective-C and Objective-C++
2469 (NOTE: This manual does not describe the Objective-C and Objective-C++
2470 languages themselves. See @xref{Standards,,Language Standards
2471 Supported by GCC}, for references.)
2473 This section describes the command-line options that are only meaningful
2474 for Objective-C and Objective-C++ programs, but you can also use most of
2475 the language-independent GNU compiler options.
2476 For example, you might compile a file @code{some_class.m} like this:
2479 gcc -g -fgnu-runtime -O -c some_class.m
2483 In this example, @option{-fgnu-runtime} is an option meant only for
2484 Objective-C and Objective-C++ programs; you can use the other options with
2485 any language supported by GCC@.
2487 Note that since Objective-C is an extension of the C language, Objective-C
2488 compilations may also use options specific to the C front-end (e.g.,
2489 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2490 C++-specific options (e.g., @option{-Wabi}).
2492 Here is a list of options that are @emph{only} for compiling Objective-C
2493 and Objective-C++ programs:
2496 @item -fconstant-string-class=@var{class-name}
2497 @opindex fconstant-string-class
2498 Use @var{class-name} as the name of the class to instantiate for each
2499 literal string specified with the syntax @code{@@"@dots{}"}. The default
2500 class name is @code{NXConstantString} if the GNU runtime is being used, and
2501 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2502 @option{-fconstant-cfstrings} option, if also present, will override the
2503 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2504 to be laid out as constant CoreFoundation strings.
2507 @opindex fgnu-runtime
2508 Generate object code compatible with the standard GNU Objective-C
2509 runtime. This is the default for most types of systems.
2511 @item -fnext-runtime
2512 @opindex fnext-runtime
2513 Generate output compatible with the NeXT runtime. This is the default
2514 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2515 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2518 @item -fno-nil-receivers
2519 @opindex fno-nil-receivers
2520 Assume that all Objective-C message dispatches (e.g.,
2521 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
2522 is not @code{nil}. This allows for more efficient entry points in the runtime
2523 to be used. Currently, this option is only available in conjunction with
2524 the NeXT runtime on Mac OS X 10.3 and later.
2526 @item -fobjc-call-cxx-cdtors
2527 @opindex fobjc-call-cxx-cdtors
2528 For each Objective-C class, check if any of its instance variables is a
2529 C++ object with a non-trivial default constructor. If so, synthesize a
2530 special @code{- (id) .cxx_construct} instance method that will run
2531 non-trivial default constructors on any such instance variables, in order,
2532 and then return @code{self}. Similarly, check if any instance variable
2533 is a C++ object with a non-trivial destructor, and if so, synthesize a
2534 special @code{- (void) .cxx_destruct} method that will run
2535 all such default destructors, in reverse order.
2537 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
2538 thusly generated will only operate on instance variables declared in the
2539 current Objective-C class, and not those inherited from superclasses. It
2540 is the responsibility of the Objective-C runtime to invoke all such methods
2541 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
2542 will be invoked by the runtime immediately after a new object
2543 instance is allocated; the @code{- (void) .cxx_destruct} methods will
2544 be invoked immediately before the runtime deallocates an object instance.
2546 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2547 support for invoking the @code{- (id) .cxx_construct} and
2548 @code{- (void) .cxx_destruct} methods.
2550 @item -fobjc-direct-dispatch
2551 @opindex fobjc-direct-dispatch
2552 Allow fast jumps to the message dispatcher. On Darwin this is
2553 accomplished via the comm page.
2555 @item -fobjc-exceptions
2556 @opindex fobjc-exceptions
2557 Enable syntactic support for structured exception handling in Objective-C,
2558 similar to what is offered by C++ and Java. This option is
2559 unavailable in conjunction with the NeXT runtime on Mac OS X 10.2 and
2568 @@catch (AnObjCClass *exc) @{
2575 @@catch (AnotherClass *exc) @{
2578 @@catch (id allOthers) @{
2588 The @code{@@throw} statement may appear anywhere in an Objective-C or
2589 Objective-C++ program; when used inside of a @code{@@catch} block, the
2590 @code{@@throw} may appear without an argument (as shown above), in which case
2591 the object caught by the @code{@@catch} will be rethrown.
2593 Note that only (pointers to) Objective-C objects may be thrown and
2594 caught using this scheme. When an object is thrown, it will be caught
2595 by the nearest @code{@@catch} clause capable of handling objects of that type,
2596 analogously to how @code{catch} blocks work in C++ and Java. A
2597 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2598 any and all Objective-C exceptions not caught by previous @code{@@catch}
2601 The @code{@@finally} clause, if present, will be executed upon exit from the
2602 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2603 regardless of whether any exceptions are thrown, caught or rethrown
2604 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2605 of the @code{finally} clause in Java.
2607 There are several caveats to using the new exception mechanism:
2611 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2612 idioms provided by the @code{NSException} class, the new
2613 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2614 systems, due to additional functionality needed in the (NeXT) Objective-C
2618 As mentioned above, the new exceptions do not support handling
2619 types other than Objective-C objects. Furthermore, when used from
2620 Objective-C++, the Objective-C exception model does not interoperate with C++
2621 exceptions at this time. This means you cannot @code{@@throw} an exception
2622 from Objective-C and @code{catch} it in C++, or vice versa
2623 (i.e., @code{throw @dots{} @@catch}).
2626 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2627 blocks for thread-safe execution:
2630 @@synchronized (ObjCClass *guard) @{
2635 Upon entering the @code{@@synchronized} block, a thread of execution shall
2636 first check whether a lock has been placed on the corresponding @code{guard}
2637 object by another thread. If it has, the current thread shall wait until
2638 the other thread relinquishes its lock. Once @code{guard} becomes available,
2639 the current thread will place its own lock on it, execute the code contained in
2640 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2641 making @code{guard} available to other threads).
2643 Unlike Java, Objective-C does not allow for entire methods to be marked
2644 @code{@@synchronized}. Note that throwing exceptions out of
2645 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2646 to be unlocked properly.
2650 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2652 @item -freplace-objc-classes
2653 @opindex freplace-objc-classes
2654 Emit a special marker instructing @command{ld(1)} not to statically link in
2655 the resulting object file, and allow @command{dyld(1)} to load it in at
2656 run time instead. This is used in conjunction with the Fix-and-Continue
2657 debugging mode, where the object file in question may be recompiled and
2658 dynamically reloaded in the course of program execution, without the need
2659 to restart the program itself. Currently, Fix-and-Continue functionality
2660 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2665 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2666 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2667 compile time) with static class references that get initialized at load time,
2668 which improves run-time performance. Specifying the @option{-fzero-link} flag
2669 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2670 to be retained. This is useful in Zero-Link debugging mode, since it allows
2671 for individual class implementations to be modified during program execution.
2675 Dump interface declarations for all classes seen in the source file to a
2676 file named @file{@var{sourcename}.decl}.
2678 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2679 @opindex Wassign-intercept
2680 @opindex Wno-assign-intercept
2681 Warn whenever an Objective-C assignment is being intercepted by the
2684 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2685 @opindex Wno-protocol
2687 If a class is declared to implement a protocol, a warning is issued for
2688 every method in the protocol that is not implemented by the class. The
2689 default behavior is to issue a warning for every method not explicitly
2690 implemented in the class, even if a method implementation is inherited
2691 from the superclass. If you use the @option{-Wno-protocol} option, then
2692 methods inherited from the superclass are considered to be implemented,
2693 and no warning is issued for them.
2695 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2697 @opindex Wno-selector
2698 Warn if multiple methods of different types for the same selector are
2699 found during compilation. The check is performed on the list of methods
2700 in the final stage of compilation. Additionally, a check is performed
2701 for each selector appearing in a @code{@@selector(@dots{})}
2702 expression, and a corresponding method for that selector has been found
2703 during compilation. Because these checks scan the method table only at
2704 the end of compilation, these warnings are not produced if the final
2705 stage of compilation is not reached, for example because an error is
2706 found during compilation, or because the @option{-fsyntax-only} option is
2709 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2710 @opindex Wstrict-selector-match
2711 @opindex Wno-strict-selector-match
2712 Warn if multiple methods with differing argument and/or return types are
2713 found for a given selector when attempting to send a message using this
2714 selector to a receiver of type @code{id} or @code{Class}. When this flag
2715 is off (which is the default behavior), the compiler will omit such warnings
2716 if any differences found are confined to types which share the same size
2719 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2720 @opindex Wundeclared-selector
2721 @opindex Wno-undeclared-selector
2722 Warn if a @code{@@selector(@dots{})} expression referring to an
2723 undeclared selector is found. A selector is considered undeclared if no
2724 method with that name has been declared before the
2725 @code{@@selector(@dots{})} expression, either explicitly in an
2726 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2727 an @code{@@implementation} section. This option always performs its
2728 checks as soon as a @code{@@selector(@dots{})} expression is found,
2729 while @option{-Wselector} only performs its checks in the final stage of
2730 compilation. This also enforces the coding style convention
2731 that methods and selectors must be declared before being used.
2733 @item -print-objc-runtime-info
2734 @opindex print-objc-runtime-info
2735 Generate C header describing the largest structure that is passed by
2740 @node Language Independent Options
2741 @section Options to Control Diagnostic Messages Formatting
2742 @cindex options to control diagnostics formatting
2743 @cindex diagnostic messages
2744 @cindex message formatting
2746 Traditionally, diagnostic messages have been formatted irrespective of
2747 the output device's aspect (e.g.@: its width, @dots{}). The options described
2748 below can be used to control the diagnostic messages formatting
2749 algorithm, e.g.@: how many characters per line, how often source location
2750 information should be reported. Right now, only the C++ front end can
2751 honor these options. However it is expected, in the near future, that
2752 the remaining front ends would be able to digest them correctly.
2755 @item -fmessage-length=@var{n}
2756 @opindex fmessage-length
2757 Try to format error messages so that they fit on lines of about @var{n}
2758 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2759 the front ends supported by GCC@. If @var{n} is zero, then no
2760 line-wrapping will be done; each error message will appear on a single
2763 @opindex fdiagnostics-show-location
2764 @item -fdiagnostics-show-location=once
2765 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2766 reporter to emit @emph{once} source location information; that is, in
2767 case the message is too long to fit on a single physical line and has to
2768 be wrapped, the source location won't be emitted (as prefix) again,
2769 over and over, in subsequent continuation lines. This is the default
2772 @item -fdiagnostics-show-location=every-line
2773 Only meaningful in line-wrapping mode. Instructs the diagnostic
2774 messages reporter to emit the same source location information (as
2775 prefix) for physical lines that result from the process of breaking
2776 a message which is too long to fit on a single line.
2778 @item -fdiagnostics-show-option
2779 @opindex fdiagnostics-show-option
2780 This option instructs the diagnostic machinery to add text to each
2781 diagnostic emitted, which indicates which command line option directly
2782 controls that diagnostic, when such an option is known to the
2783 diagnostic machinery.
2785 @item -Wcoverage-mismatch
2786 @opindex Wcoverage-mismatch
2787 Warn if feedback profiles do not match when using the
2788 @option{-fprofile-use} option.
2789 If a source file was changed between @option{-fprofile-gen} and
2790 @option{-fprofile-use}, the files with the profile feedback can fail
2791 to match the source file and GCC can not use the profile feedback
2792 information. By default, this warning is enabled and is treated as an
2793 error. @option{-Wno-coverage-mismatch} can be used to disable the
2794 warning or @option{-Wno-error=coverage-mismatch} can be used to
2795 disable the error. Disable the error for this warning can result in
2796 poorly optimized code, so disabling the error is useful only in the
2797 case of very minor changes such as bug fixes to an existing code-base.
2798 Completely disabling the warning is not recommended.
2802 @node Warning Options
2803 @section Options to Request or Suppress Warnings
2804 @cindex options to control warnings
2805 @cindex warning messages
2806 @cindex messages, warning
2807 @cindex suppressing warnings
2809 Warnings are diagnostic messages that report constructions which
2810 are not inherently erroneous but which are risky or suggest there
2811 may have been an error.
2813 The following language-independent options do not enable specific
2814 warnings but control the kinds of diagnostics produced by GCC.
2817 @cindex syntax checking
2819 @opindex fsyntax-only
2820 Check the code for syntax errors, but don't do anything beyond that.
2824 Inhibit all warning messages.
2829 Make all warnings into errors.
2834 Make the specified warning into an error. The specifier for a warning
2835 is appended, for example @option{-Werror=switch} turns the warnings
2836 controlled by @option{-Wswitch} into errors. This switch takes a
2837 negative form, to be used to negate @option{-Werror} for specific
2838 warnings, for example @option{-Wno-error=switch} makes
2839 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2840 is in effect. You can use the @option{-fdiagnostics-show-option}
2841 option to have each controllable warning amended with the option which
2842 controls it, to determine what to use with this option.
2844 Note that specifying @option{-Werror=}@var{foo} automatically implies
2845 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2848 @item -Wfatal-errors
2849 @opindex Wfatal-errors
2850 @opindex Wno-fatal-errors
2851 This option causes the compiler to abort compilation on the first error
2852 occurred rather than trying to keep going and printing further error
2857 You can request many specific warnings with options beginning
2858 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2859 implicit declarations. Each of these specific warning options also
2860 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2861 example, @option{-Wno-implicit}. This manual lists only one of the
2862 two forms, whichever is not the default. For further,
2863 language-specific options also refer to @ref{C++ Dialect Options} and
2864 @ref{Objective-C and Objective-C++ Dialect Options}.
2866 When an unrecognized warning label is requested (e.g.,
2867 @option{-Wunknown-warning}), GCC will emit a diagnostic stating
2868 that the option is not recognized. However, if the @samp{-Wno-} form
2869 is used, the behavior is slightly different: No diagnostic will be
2870 produced for @option{-Wno-unknown-warning} unless other diagnostics
2871 are being produced. This allows the use of new @option{-Wno-} options
2872 with old compilers, but if something goes wrong, the compiler will
2873 warn that an unrecognized option was used.
2878 Issue all the warnings demanded by strict ISO C and ISO C++;
2879 reject all programs that use forbidden extensions, and some other
2880 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2881 version of the ISO C standard specified by any @option{-std} option used.
2883 Valid ISO C and ISO C++ programs should compile properly with or without
2884 this option (though a rare few will require @option{-ansi} or a
2885 @option{-std} option specifying the required version of ISO C)@. However,
2886 without this option, certain GNU extensions and traditional C and C++
2887 features are supported as well. With this option, they are rejected.
2889 @option{-pedantic} does not cause warning messages for use of the
2890 alternate keywords whose names begin and end with @samp{__}. Pedantic
2891 warnings are also disabled in the expression that follows
2892 @code{__extension__}. However, only system header files should use
2893 these escape routes; application programs should avoid them.
2894 @xref{Alternate Keywords}.
2896 Some users try to use @option{-pedantic} to check programs for strict ISO
2897 C conformance. They soon find that it does not do quite what they want:
2898 it finds some non-ISO practices, but not all---only those for which
2899 ISO C @emph{requires} a diagnostic, and some others for which
2900 diagnostics have been added.
2902 A feature to report any failure to conform to ISO C might be useful in
2903 some instances, but would require considerable additional work and would
2904 be quite different from @option{-pedantic}. We don't have plans to
2905 support such a feature in the near future.
2907 Where the standard specified with @option{-std} represents a GNU
2908 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
2909 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2910 extended dialect is based. Warnings from @option{-pedantic} are given
2911 where they are required by the base standard. (It would not make sense
2912 for such warnings to be given only for features not in the specified GNU
2913 C dialect, since by definition the GNU dialects of C include all
2914 features the compiler supports with the given option, and there would be
2915 nothing to warn about.)
2917 @item -pedantic-errors
2918 @opindex pedantic-errors
2919 Like @option{-pedantic}, except that errors are produced rather than
2925 This enables all the warnings about constructions that some users
2926 consider questionable, and that are easy to avoid (or modify to
2927 prevent the warning), even in conjunction with macros. This also
2928 enables some language-specific warnings described in @ref{C++ Dialect
2929 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2931 @option{-Wall} turns on the following warning flags:
2933 @gccoptlist{-Waddress @gol
2934 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2936 -Wchar-subscripts @gol
2937 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
2938 -Wimplicit-int @r{(C and Objective-C only)} @gol
2939 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
2942 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2943 -Wmissing-braces @gol
2949 -Wsequence-point @gol
2950 -Wsign-compare @r{(only in C++)} @gol
2951 -Wstrict-aliasing @gol
2952 -Wstrict-overflow=1 @gol
2955 -Wuninitialized @gol
2956 -Wunknown-pragmas @gol
2957 -Wunused-function @gol
2960 -Wunused-variable @gol
2961 -Wvolatile-register-var @gol
2964 Note that some warning flags are not implied by @option{-Wall}. Some of
2965 them warn about constructions that users generally do not consider
2966 questionable, but which occasionally you might wish to check for;
2967 others warn about constructions that are necessary or hard to avoid in
2968 some cases, and there is no simple way to modify the code to suppress
2969 the warning. Some of them are enabled by @option{-Wextra} but many of
2970 them must be enabled individually.
2976 This enables some extra warning flags that are not enabled by
2977 @option{-Wall}. (This option used to be called @option{-W}. The older
2978 name is still supported, but the newer name is more descriptive.)
2980 @gccoptlist{-Wclobbered @gol
2982 -Wignored-qualifiers @gol
2983 -Wmissing-field-initializers @gol
2984 -Wmissing-parameter-type @r{(C only)} @gol
2985 -Wold-style-declaration @r{(C only)} @gol
2986 -Woverride-init @gol
2989 -Wuninitialized @gol
2990 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2991 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
2994 The option @option{-Wextra} also prints warning messages for the
3000 A pointer is compared against integer zero with @samp{<}, @samp{<=},
3001 @samp{>}, or @samp{>=}.
3004 (C++ only) An enumerator and a non-enumerator both appear in a
3005 conditional expression.
3008 (C++ only) Ambiguous virtual bases.
3011 (C++ only) Subscripting an array which has been declared @samp{register}.
3014 (C++ only) Taking the address of a variable which has been declared
3018 (C++ only) A base class is not initialized in a derived class' copy
3023 @item -Wchar-subscripts
3024 @opindex Wchar-subscripts
3025 @opindex Wno-char-subscripts
3026 Warn if an array subscript has type @code{char}. This is a common cause
3027 of error, as programmers often forget that this type is signed on some
3029 This warning is enabled by @option{-Wall}.
3033 @opindex Wno-comment
3034 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3035 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3036 This warning is enabled by @option{-Wall}.
3039 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3041 Suppress warning messages emitted by @code{#warning} directives.
3046 @opindex ffreestanding
3047 @opindex fno-builtin
3048 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3049 the arguments supplied have types appropriate to the format string
3050 specified, and that the conversions specified in the format string make
3051 sense. This includes standard functions, and others specified by format
3052 attributes (@pxref{Function Attributes}), in the @code{printf},
3053 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3054 not in the C standard) families (or other target-specific families).
3055 Which functions are checked without format attributes having been
3056 specified depends on the standard version selected, and such checks of
3057 functions without the attribute specified are disabled by
3058 @option{-ffreestanding} or @option{-fno-builtin}.
3060 The formats are checked against the format features supported by GNU
3061 libc version 2.2. These include all ISO C90 and C99 features, as well
3062 as features from the Single Unix Specification and some BSD and GNU
3063 extensions. Other library implementations may not support all these
3064 features; GCC does not support warning about features that go beyond a
3065 particular library's limitations. However, if @option{-pedantic} is used
3066 with @option{-Wformat}, warnings will be given about format features not
3067 in the selected standard version (but not for @code{strfmon} formats,
3068 since those are not in any version of the C standard). @xref{C Dialect
3069 Options,,Options Controlling C Dialect}.
3071 Since @option{-Wformat} also checks for null format arguments for
3072 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
3074 @option{-Wformat} is included in @option{-Wall}. For more control over some
3075 aspects of format checking, the options @option{-Wformat-y2k},
3076 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
3077 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
3078 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
3081 @opindex Wformat-y2k
3082 @opindex Wno-format-y2k
3083 If @option{-Wformat} is specified, also warn about @code{strftime}
3084 formats which may yield only a two-digit year.
3086 @item -Wno-format-contains-nul
3087 @opindex Wno-format-contains-nul
3088 @opindex Wformat-contains-nul
3089 If @option{-Wformat} is specified, do not warn about format strings that
3092 @item -Wno-format-extra-args
3093 @opindex Wno-format-extra-args
3094 @opindex Wformat-extra-args
3095 If @option{-Wformat} is specified, do not warn about excess arguments to a
3096 @code{printf} or @code{scanf} format function. The C standard specifies
3097 that such arguments are ignored.
3099 Where the unused arguments lie between used arguments that are
3100 specified with @samp{$} operand number specifications, normally
3101 warnings are still given, since the implementation could not know what
3102 type to pass to @code{va_arg} to skip the unused arguments. However,
3103 in the case of @code{scanf} formats, this option will suppress the
3104 warning if the unused arguments are all pointers, since the Single
3105 Unix Specification says that such unused arguments are allowed.
3107 @item -Wno-format-zero-length @r{(C and Objective-C only)}
3108 @opindex Wno-format-zero-length
3109 @opindex Wformat-zero-length
3110 If @option{-Wformat} is specified, do not warn about zero-length formats.
3111 The C standard specifies that zero-length formats are allowed.
3113 @item -Wformat-nonliteral
3114 @opindex Wformat-nonliteral
3115 @opindex Wno-format-nonliteral
3116 If @option{-Wformat} is specified, also warn if the format string is not a
3117 string literal and so cannot be checked, unless the format function
3118 takes its format arguments as a @code{va_list}.
3120 @item -Wformat-security
3121 @opindex Wformat-security
3122 @opindex Wno-format-security
3123 If @option{-Wformat} is specified, also warn about uses of format
3124 functions that represent possible security problems. At present, this
3125 warns about calls to @code{printf} and @code{scanf} functions where the
3126 format string is not a string literal and there are no format arguments,
3127 as in @code{printf (foo);}. This may be a security hole if the format
3128 string came from untrusted input and contains @samp{%n}. (This is
3129 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3130 in future warnings may be added to @option{-Wformat-security} that are not
3131 included in @option{-Wformat-nonliteral}.)
3135 @opindex Wno-format=2
3136 Enable @option{-Wformat} plus format checks not included in
3137 @option{-Wformat}. Currently equivalent to @samp{-Wformat
3138 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
3140 @item -Wnonnull @r{(C and Objective-C only)}
3142 @opindex Wno-nonnull
3143 Warn about passing a null pointer for arguments marked as
3144 requiring a non-null value by the @code{nonnull} function attribute.
3146 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3147 can be disabled with the @option{-Wno-nonnull} option.
3149 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3151 @opindex Wno-init-self
3152 Warn about uninitialized variables which are initialized with themselves.
3153 Note this option can only be used with the @option{-Wuninitialized} option.
3155 For example, GCC will warn about @code{i} being uninitialized in the
3156 following snippet only when @option{-Winit-self} has been specified:
3167 @item -Wimplicit-int @r{(C and Objective-C only)}
3168 @opindex Wimplicit-int
3169 @opindex Wno-implicit-int
3170 Warn when a declaration does not specify a type.
3171 This warning is enabled by @option{-Wall}.
3173 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3174 @opindex Wimplicit-function-declaration
3175 @opindex Wno-implicit-function-declaration
3176 Give a warning whenever a function is used before being declared. In
3177 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3178 enabled by default and it is made into an error by
3179 @option{-pedantic-errors}. This warning is also enabled by
3182 @item -Wimplicit @r{(C and Objective-C only)}
3184 @opindex Wno-implicit
3185 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3186 This warning is enabled by @option{-Wall}.
3188 @item -Wignored-qualifiers @r{(C and C++ only)}
3189 @opindex Wignored-qualifiers
3190 @opindex Wno-ignored-qualifiers
3191 Warn if the return type of a function has a type qualifier
3192 such as @code{const}. For ISO C such a type qualifier has no effect,
3193 since the value returned by a function is not an lvalue.
3194 For C++, the warning is only emitted for scalar types or @code{void}.
3195 ISO C prohibits qualified @code{void} return types on function
3196 definitions, so such return types always receive a warning
3197 even without this option.
3199 This warning is also enabled by @option{-Wextra}.
3204 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3205 a function with external linkage, returning int, taking either zero
3206 arguments, two, or three arguments of appropriate types. This warning
3207 is enabled by default in C++ and is enabled by either @option{-Wall}
3208 or @option{-pedantic}.
3210 @item -Wmissing-braces
3211 @opindex Wmissing-braces
3212 @opindex Wno-missing-braces
3213 Warn if an aggregate or union initializer is not fully bracketed. In
3214 the following example, the initializer for @samp{a} is not fully
3215 bracketed, but that for @samp{b} is fully bracketed.
3218 int a[2][2] = @{ 0, 1, 2, 3 @};
3219 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3222 This warning is enabled by @option{-Wall}.
3224 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3225 @opindex Wmissing-include-dirs
3226 @opindex Wno-missing-include-dirs
3227 Warn if a user-supplied include directory does not exist.
3230 @opindex Wparentheses
3231 @opindex Wno-parentheses
3232 Warn if parentheses are omitted in certain contexts, such
3233 as when there is an assignment in a context where a truth value
3234 is expected, or when operators are nested whose precedence people
3235 often get confused about.
3237 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3238 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3239 interpretation from that of ordinary mathematical notation.
3241 Also warn about constructions where there may be confusion to which
3242 @code{if} statement an @code{else} branch belongs. Here is an example of
3257 In C/C++, every @code{else} branch belongs to the innermost possible
3258 @code{if} statement, which in this example is @code{if (b)}. This is
3259 often not what the programmer expected, as illustrated in the above
3260 example by indentation the programmer chose. When there is the
3261 potential for this confusion, GCC will issue a warning when this flag
3262 is specified. To eliminate the warning, add explicit braces around
3263 the innermost @code{if} statement so there is no way the @code{else}
3264 could belong to the enclosing @code{if}. The resulting code would
3281 Also warn for dangerous uses of the
3282 ?: with omitted middle operand GNU extension. When the condition
3283 in the ?: operator is a boolean expression the omitted value will
3284 be always 1. Often the user expects it to be a value computed
3285 inside the conditional expression instead.
3287 This warning is enabled by @option{-Wall}.
3289 @item -Wsequence-point
3290 @opindex Wsequence-point
3291 @opindex Wno-sequence-point
3292 Warn about code that may have undefined semantics because of violations
3293 of sequence point rules in the C and C++ standards.
3295 The C and C++ standards defines the order in which expressions in a C/C++
3296 program are evaluated in terms of @dfn{sequence points}, which represent
3297 a partial ordering between the execution of parts of the program: those
3298 executed before the sequence point, and those executed after it. These
3299 occur after the evaluation of a full expression (one which is not part
3300 of a larger expression), after the evaluation of the first operand of a
3301 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3302 function is called (but after the evaluation of its arguments and the
3303 expression denoting the called function), and in certain other places.
3304 Other than as expressed by the sequence point rules, the order of
3305 evaluation of subexpressions of an expression is not specified. All
3306 these rules describe only a partial order rather than a total order,
3307 since, for example, if two functions are called within one expression
3308 with no sequence point between them, the order in which the functions
3309 are called is not specified. However, the standards committee have
3310 ruled that function calls do not overlap.
3312 It is not specified when between sequence points modifications to the
3313 values of objects take effect. Programs whose behavior depends on this
3314 have undefined behavior; the C and C++ standards specify that ``Between
3315 the previous and next sequence point an object shall have its stored
3316 value modified at most once by the evaluation of an expression.
3317 Furthermore, the prior value shall be read only to determine the value
3318 to be stored.''. If a program breaks these rules, the results on any
3319 particular implementation are entirely unpredictable.
3321 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3322 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3323 diagnosed by this option, and it may give an occasional false positive
3324 result, but in general it has been found fairly effective at detecting
3325 this sort of problem in programs.
3327 The standard is worded confusingly, therefore there is some debate
3328 over the precise meaning of the sequence point rules in subtle cases.
3329 Links to discussions of the problem, including proposed formal
3330 definitions, may be found on the GCC readings page, at
3331 @w{@uref{http://gcc.gnu.org/readings.html}}.
3333 This warning is enabled by @option{-Wall} for C and C++.
3336 @opindex Wreturn-type
3337 @opindex Wno-return-type
3338 Warn whenever a function is defined with a return-type that defaults
3339 to @code{int}. Also warn about any @code{return} statement with no
3340 return-value in a function whose return-type is not @code{void}
3341 (falling off the end of the function body is considered returning
3342 without a value), and about a @code{return} statement with an
3343 expression in a function whose return-type is @code{void}.
3345 For C++, a function without return type always produces a diagnostic
3346 message, even when @option{-Wno-return-type} is specified. The only
3347 exceptions are @samp{main} and functions defined in system headers.
3349 This warning is enabled by @option{-Wall}.
3354 Warn whenever a @code{switch} statement has an index of enumerated type
3355 and lacks a @code{case} for one or more of the named codes of that
3356 enumeration. (The presence of a @code{default} label prevents this
3357 warning.) @code{case} labels outside the enumeration range also
3358 provoke warnings when this option is used (even if there is a
3359 @code{default} label).
3360 This warning is enabled by @option{-Wall}.
3362 @item -Wswitch-default
3363 @opindex Wswitch-default
3364 @opindex Wno-switch-default
3365 Warn whenever a @code{switch} statement does not have a @code{default}
3369 @opindex Wswitch-enum
3370 @opindex Wno-switch-enum
3371 Warn whenever a @code{switch} statement has an index of enumerated type
3372 and lacks a @code{case} for one or more of the named codes of that
3373 enumeration. @code{case} labels outside the enumeration range also
3374 provoke warnings when this option is used. The only difference
3375 between @option{-Wswitch} and this option is that this option gives a
3376 warning about an omitted enumeration code even if there is a
3377 @code{default} label.
3379 @item -Wsync-nand @r{(C and C++ only)}
3381 @opindex Wno-sync-nand
3382 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3383 built-in functions are used. These functions changed semantics in GCC 4.4.
3387 @opindex Wno-trigraphs
3388 Warn if any trigraphs are encountered that might change the meaning of
3389 the program (trigraphs within comments are not warned about).
3390 This warning is enabled by @option{-Wall}.
3392 @item -Wunused-but-set-parameter
3393 @opindex Wunused-but-set-parameter
3394 @opindex Wno-unused-but-set-parameter
3395 Warn whenever a function parameter is assigned to, but otherwise unused
3396 (aside from its declaration).
3398 To suppress this warning use the @samp{unused} attribute
3399 (@pxref{Variable Attributes}).
3401 This warning is also enabled by @option{-Wunused} together with
3404 @item -Wunused-but-set-variable
3405 @opindex Wunused-but-set-variable
3406 @opindex Wno-unused-but-set-variable
3407 Warn whenever a local variable is assigned to, but otherwise unused
3408 (aside from its declaration).
3409 This warning is enabled by @option{-Wall}.
3411 To suppress this warning use the @samp{unused} attribute
3412 (@pxref{Variable Attributes}).
3414 This warning is also enabled by @option{-Wunused}, which is enabled
3417 @item -Wunused-function
3418 @opindex Wunused-function
3419 @opindex Wno-unused-function
3420 Warn whenever a static function is declared but not defined or a
3421 non-inline static function is unused.
3422 This warning is enabled by @option{-Wall}.
3424 @item -Wunused-label
3425 @opindex Wunused-label
3426 @opindex Wno-unused-label
3427 Warn whenever a label is declared but not used.
3428 This warning is enabled by @option{-Wall}.
3430 To suppress this warning use the @samp{unused} attribute
3431 (@pxref{Variable Attributes}).
3433 @item -Wunused-parameter
3434 @opindex Wunused-parameter
3435 @opindex Wno-unused-parameter
3436 Warn whenever a function parameter is unused aside from its declaration.
3438 To suppress this warning use the @samp{unused} attribute
3439 (@pxref{Variable Attributes}).
3441 @item -Wno-unused-result
3442 @opindex Wunused-result
3443 @opindex Wno-unused-result
3444 Do not warn if a caller of a function marked with attribute
3445 @code{warn_unused_result} (@pxref{Variable Attributes}) does not use
3446 its return value. The default is @option{-Wunused-result}.
3448 @item -Wunused-variable
3449 @opindex Wunused-variable
3450 @opindex Wno-unused-variable
3451 Warn whenever a local variable or non-constant static variable is unused
3452 aside from its declaration.
3453 This warning is enabled by @option{-Wall}.
3455 To suppress this warning use the @samp{unused} attribute
3456 (@pxref{Variable Attributes}).
3458 @item -Wunused-value
3459 @opindex Wunused-value
3460 @opindex Wno-unused-value
3461 Warn whenever a statement computes a result that is explicitly not
3462 used. To suppress this warning cast the unused expression to
3463 @samp{void}. This includes an expression-statement or the left-hand
3464 side of a comma expression that contains no side effects. For example,
3465 an expression such as @samp{x[i,j]} will cause a warning, while
3466 @samp{x[(void)i,j]} will not.
3468 This warning is enabled by @option{-Wall}.
3473 All the above @option{-Wunused} options combined.
3475 In order to get a warning about an unused function parameter, you must
3476 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3477 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3479 @item -Wuninitialized
3480 @opindex Wuninitialized
3481 @opindex Wno-uninitialized
3482 Warn if an automatic variable is used without first being initialized
3483 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3484 warn if a non-static reference or non-static @samp{const} member
3485 appears in a class without constructors.
3487 If you want to warn about code which uses the uninitialized value of the
3488 variable in its own initializer, use the @option{-Winit-self} option.
3490 These warnings occur for individual uninitialized or clobbered
3491 elements of structure, union or array variables as well as for
3492 variables which are uninitialized or clobbered as a whole. They do
3493 not occur for variables or elements declared @code{volatile}. Because
3494 these warnings depend on optimization, the exact variables or elements
3495 for which there are warnings will depend on the precise optimization
3496 options and version of GCC used.
3498 Note that there may be no warning about a variable that is used only
3499 to compute a value that itself is never used, because such
3500 computations may be deleted by data flow analysis before the warnings
3503 These warnings are made optional because GCC is not smart
3504 enough to see all the reasons why the code might be correct
3505 despite appearing to have an error. Here is one example of how
3526 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3527 always initialized, but GCC doesn't know this. Here is
3528 another common case:
3533 if (change_y) save_y = y, y = new_y;
3535 if (change_y) y = save_y;
3540 This has no bug because @code{save_y} is used only if it is set.
3542 @cindex @code{longjmp} warnings
3543 This option also warns when a non-volatile automatic variable might be
3544 changed by a call to @code{longjmp}. These warnings as well are possible
3545 only in optimizing compilation.
3547 The compiler sees only the calls to @code{setjmp}. It cannot know
3548 where @code{longjmp} will be called; in fact, a signal handler could
3549 call it at any point in the code. As a result, you may get a warning
3550 even when there is in fact no problem because @code{longjmp} cannot
3551 in fact be called at the place which would cause a problem.
3553 Some spurious warnings can be avoided if you declare all the functions
3554 you use that never return as @code{noreturn}. @xref{Function
3557 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3559 @item -Wunknown-pragmas
3560 @opindex Wunknown-pragmas
3561 @opindex Wno-unknown-pragmas
3562 @cindex warning for unknown pragmas
3563 @cindex unknown pragmas, warning
3564 @cindex pragmas, warning of unknown
3565 Warn when a #pragma directive is encountered which is not understood by
3566 GCC@. If this command line option is used, warnings will even be issued
3567 for unknown pragmas in system header files. This is not the case if
3568 the warnings were only enabled by the @option{-Wall} command line option.
3571 @opindex Wno-pragmas
3573 Do not warn about misuses of pragmas, such as incorrect parameters,
3574 invalid syntax, or conflicts between pragmas. See also
3575 @samp{-Wunknown-pragmas}.
3577 @item -Wstrict-aliasing
3578 @opindex Wstrict-aliasing
3579 @opindex Wno-strict-aliasing
3580 This option is only active when @option{-fstrict-aliasing} is active.
3581 It warns about code which might break the strict aliasing rules that the
3582 compiler is using for optimization. The warning does not catch all
3583 cases, but does attempt to catch the more common pitfalls. It is
3584 included in @option{-Wall}.
3585 It is equivalent to @option{-Wstrict-aliasing=3}
3587 @item -Wstrict-aliasing=n
3588 @opindex Wstrict-aliasing=n
3589 @opindex Wno-strict-aliasing=n
3590 This option is only active when @option{-fstrict-aliasing} is active.
3591 It warns about code which might break the strict aliasing rules that the
3592 compiler is using for optimization.
3593 Higher levels correspond to higher accuracy (fewer false positives).
3594 Higher levels also correspond to more effort, similar to the way -O works.
3595 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3598 Level 1: Most aggressive, quick, least accurate.
3599 Possibly useful when higher levels
3600 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3601 false negatives. However, it has many false positives.
3602 Warns for all pointer conversions between possibly incompatible types,
3603 even if never dereferenced. Runs in the frontend only.
3605 Level 2: Aggressive, quick, not too precise.
3606 May still have many false positives (not as many as level 1 though),
3607 and few false negatives (but possibly more than level 1).
3608 Unlike level 1, it only warns when an address is taken. Warns about
3609 incomplete types. Runs in the frontend only.
3611 Level 3 (default for @option{-Wstrict-aliasing}):
3612 Should have very few false positives and few false
3613 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3614 Takes care of the common pun+dereference pattern in the frontend:
3615 @code{*(int*)&some_float}.
3616 If optimization is enabled, it also runs in the backend, where it deals
3617 with multiple statement cases using flow-sensitive points-to information.
3618 Only warns when the converted pointer is dereferenced.
3619 Does not warn about incomplete types.
3621 @item -Wstrict-overflow
3622 @itemx -Wstrict-overflow=@var{n}
3623 @opindex Wstrict-overflow
3624 @opindex Wno-strict-overflow
3625 This option is only active when @option{-fstrict-overflow} is active.
3626 It warns about cases where the compiler optimizes based on the
3627 assumption that signed overflow does not occur. Note that it does not
3628 warn about all cases where the code might overflow: it only warns
3629 about cases where the compiler implements some optimization. Thus
3630 this warning depends on the optimization level.
3632 An optimization which assumes that signed overflow does not occur is
3633 perfectly safe if the values of the variables involved are such that
3634 overflow never does, in fact, occur. Therefore this warning can
3635 easily give a false positive: a warning about code which is not
3636 actually a problem. To help focus on important issues, several
3637 warning levels are defined. No warnings are issued for the use of
3638 undefined signed overflow when estimating how many iterations a loop
3639 will require, in particular when determining whether a loop will be
3643 @item -Wstrict-overflow=1
3644 Warn about cases which are both questionable and easy to avoid. For
3645 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3646 compiler will simplify this to @code{1}. This level of
3647 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3648 are not, and must be explicitly requested.
3650 @item -Wstrict-overflow=2
3651 Also warn about other cases where a comparison is simplified to a
3652 constant. For example: @code{abs (x) >= 0}. This can only be
3653 simplified when @option{-fstrict-overflow} is in effect, because
3654 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3655 zero. @option{-Wstrict-overflow} (with no level) is the same as
3656 @option{-Wstrict-overflow=2}.
3658 @item -Wstrict-overflow=3
3659 Also warn about other cases where a comparison is simplified. For
3660 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3662 @item -Wstrict-overflow=4
3663 Also warn about other simplifications not covered by the above cases.
3664 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3666 @item -Wstrict-overflow=5
3667 Also warn about cases where the compiler reduces the magnitude of a
3668 constant involved in a comparison. For example: @code{x + 2 > y} will
3669 be simplified to @code{x + 1 >= y}. This is reported only at the
3670 highest warning level because this simplification applies to many
3671 comparisons, so this warning level will give a very large number of
3675 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]}
3676 @opindex Wsuggest-attribute=
3677 @opindex Wno-suggest-attribute=
3678 Warn for cases where adding an attribute may be beneficial. The
3679 attributes currently supported are listed below.
3682 @item -Wsuggest-attribute=pure
3683 @itemx -Wsuggest-attribute=const
3684 @itemx -Wsuggest-attribute=noreturn
3685 @opindex Wsuggest-attribute=pure
3686 @opindex Wno-suggest-attribute=pure
3687 @opindex Wsuggest-attribute=const
3688 @opindex Wno-suggest-attribute=const
3689 @opindex Wsuggest-attribute=noreturn
3690 @opindex Wno-suggest-attribute=noreturn
3692 Warn about functions which might be candidates for attributes
3693 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
3694 functions visible in other compilation units or (in the case of @code{pure} and
3695 @code{const}) if it cannot prove that the function returns normally. A function
3696 returns normally if it doesn't contain an infinite loop nor returns abnormally
3697 by throwing, calling @code{abort()} or trapping. This analysis requires option
3698 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
3699 higher. Higher optimization levels improve the accuracy of the analysis.
3702 @item -Warray-bounds
3703 @opindex Wno-array-bounds
3704 @opindex Warray-bounds
3705 This option is only active when @option{-ftree-vrp} is active
3706 (default for @option{-O2} and above). It warns about subscripts to arrays
3707 that are always out of bounds. This warning is enabled by @option{-Wall}.
3709 @item -Wno-div-by-zero
3710 @opindex Wno-div-by-zero
3711 @opindex Wdiv-by-zero
3712 Do not warn about compile-time integer division by zero. Floating point
3713 division by zero is not warned about, as it can be a legitimate way of
3714 obtaining infinities and NaNs.
3716 @item -Wsystem-headers
3717 @opindex Wsystem-headers
3718 @opindex Wno-system-headers
3719 @cindex warnings from system headers
3720 @cindex system headers, warnings from
3721 Print warning messages for constructs found in system header files.
3722 Warnings from system headers are normally suppressed, on the assumption
3723 that they usually do not indicate real problems and would only make the
3724 compiler output harder to read. Using this command line option tells
3725 GCC to emit warnings from system headers as if they occurred in user
3726 code. However, note that using @option{-Wall} in conjunction with this
3727 option will @emph{not} warn about unknown pragmas in system
3728 headers---for that, @option{-Wunknown-pragmas} must also be used.
3731 @opindex Wtrampolines
3732 @opindex Wno-trampolines
3733 Warn about trampolines generated for pointers to nested functions.
3735 A trampoline is a small piece of data or code that is created at run
3736 time on the stack when the address of a nested function is taken, and
3737 is used to call the nested function indirectly. For some targets, it
3738 is made up of data only and thus requires no special treatment. But,
3739 for most targets, it is made up of code and thus requires the stack
3740 to be made executable in order for the program to work properly.
3743 @opindex Wfloat-equal
3744 @opindex Wno-float-equal
3745 Warn if floating point values are used in equality comparisons.
3747 The idea behind this is that sometimes it is convenient (for the
3748 programmer) to consider floating-point values as approximations to
3749 infinitely precise real numbers. If you are doing this, then you need
3750 to compute (by analyzing the code, or in some other way) the maximum or
3751 likely maximum error that the computation introduces, and allow for it
3752 when performing comparisons (and when producing output, but that's a
3753 different problem). In particular, instead of testing for equality, you
3754 would check to see whether the two values have ranges that overlap; and
3755 this is done with the relational operators, so equality comparisons are
3758 @item -Wtraditional @r{(C and Objective-C only)}
3759 @opindex Wtraditional
3760 @opindex Wno-traditional
3761 Warn about certain constructs that behave differently in traditional and
3762 ISO C@. Also warn about ISO C constructs that have no traditional C
3763 equivalent, and/or problematic constructs which should be avoided.
3767 Macro parameters that appear within string literals in the macro body.
3768 In traditional C macro replacement takes place within string literals,
3769 but does not in ISO C@.
3772 In traditional C, some preprocessor directives did not exist.
3773 Traditional preprocessors would only consider a line to be a directive
3774 if the @samp{#} appeared in column 1 on the line. Therefore
3775 @option{-Wtraditional} warns about directives that traditional C
3776 understands but would ignore because the @samp{#} does not appear as the
3777 first character on the line. It also suggests you hide directives like
3778 @samp{#pragma} not understood by traditional C by indenting them. Some
3779 traditional implementations would not recognize @samp{#elif}, so it
3780 suggests avoiding it altogether.
3783 A function-like macro that appears without arguments.
3786 The unary plus operator.
3789 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3790 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3791 constants.) Note, these suffixes appear in macros defined in the system
3792 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3793 Use of these macros in user code might normally lead to spurious
3794 warnings, however GCC's integrated preprocessor has enough context to
3795 avoid warning in these cases.
3798 A function declared external in one block and then used after the end of
3802 A @code{switch} statement has an operand of type @code{long}.
3805 A non-@code{static} function declaration follows a @code{static} one.
3806 This construct is not accepted by some traditional C compilers.
3809 The ISO type of an integer constant has a different width or
3810 signedness from its traditional type. This warning is only issued if
3811 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3812 typically represent bit patterns, are not warned about.
3815 Usage of ISO string concatenation is detected.
3818 Initialization of automatic aggregates.
3821 Identifier conflicts with labels. Traditional C lacks a separate
3822 namespace for labels.
3825 Initialization of unions. If the initializer is zero, the warning is
3826 omitted. This is done under the assumption that the zero initializer in
3827 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3828 initializer warnings and relies on default initialization to zero in the
3832 Conversions by prototypes between fixed/floating point values and vice
3833 versa. The absence of these prototypes when compiling with traditional
3834 C would cause serious problems. This is a subset of the possible
3835 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3838 Use of ISO C style function definitions. This warning intentionally is
3839 @emph{not} issued for prototype declarations or variadic functions
3840 because these ISO C features will appear in your code when using
3841 libiberty's traditional C compatibility macros, @code{PARAMS} and
3842 @code{VPARAMS}. This warning is also bypassed for nested functions
3843 because that feature is already a GCC extension and thus not relevant to
3844 traditional C compatibility.
3847 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3848 @opindex Wtraditional-conversion
3849 @opindex Wno-traditional-conversion
3850 Warn if a prototype causes a type conversion that is different from what
3851 would happen to the same argument in the absence of a prototype. This
3852 includes conversions of fixed point to floating and vice versa, and
3853 conversions changing the width or signedness of a fixed point argument
3854 except when the same as the default promotion.
3856 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3857 @opindex Wdeclaration-after-statement
3858 @opindex Wno-declaration-after-statement
3859 Warn when a declaration is found after a statement in a block. This
3860 construct, known from C++, was introduced with ISO C99 and is by default
3861 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3862 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3867 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3869 @item -Wno-endif-labels
3870 @opindex Wno-endif-labels
3871 @opindex Wendif-labels
3872 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3877 Warn whenever a local variable or type declaration shadows another variable,
3878 parameter, type, or class member (in C++), or whenever a built-in function
3879 is shadowed. Note that in C++, the compiler will not warn if a local variable
3880 shadows a struct/class/enum, but will warn if it shadows an explicit typedef.
3882 @item -Wlarger-than=@var{len}
3883 @opindex Wlarger-than=@var{len}
3884 @opindex Wlarger-than-@var{len}
3885 Warn whenever an object of larger than @var{len} bytes is defined.
3887 @item -Wframe-larger-than=@var{len}
3888 @opindex Wframe-larger-than
3889 Warn if the size of a function frame is larger than @var{len} bytes.
3890 The computation done to determine the stack frame size is approximate
3891 and not conservative.
3892 The actual requirements may be somewhat greater than @var{len}
3893 even if you do not get a warning. In addition, any space allocated
3894 via @code{alloca}, variable-length arrays, or related constructs
3895 is not included by the compiler when determining
3896 whether or not to issue a warning.
3898 @item -Wunsafe-loop-optimizations
3899 @opindex Wunsafe-loop-optimizations
3900 @opindex Wno-unsafe-loop-optimizations
3901 Warn if the loop cannot be optimized because the compiler could not
3902 assume anything on the bounds of the loop indices. With
3903 @option{-funsafe-loop-optimizations} warn if the compiler made
3906 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
3907 @opindex Wno-pedantic-ms-format
3908 @opindex Wpedantic-ms-format
3909 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
3910 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
3911 depending on the MS runtime, when you are using the options @option{-Wformat}
3912 and @option{-pedantic} without gnu-extensions.
3914 @item -Wpointer-arith
3915 @opindex Wpointer-arith
3916 @opindex Wno-pointer-arith
3917 Warn about anything that depends on the ``size of'' a function type or
3918 of @code{void}. GNU C assigns these types a size of 1, for
3919 convenience in calculations with @code{void *} pointers and pointers
3920 to functions. In C++, warn also when an arithmetic operation involves
3921 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3924 @opindex Wtype-limits
3925 @opindex Wno-type-limits
3926 Warn if a comparison is always true or always false due to the limited
3927 range of the data type, but do not warn for constant expressions. For
3928 example, warn if an unsigned variable is compared against zero with
3929 @samp{<} or @samp{>=}. This warning is also enabled by
3932 @item -Wbad-function-cast @r{(C and Objective-C only)}
3933 @opindex Wbad-function-cast
3934 @opindex Wno-bad-function-cast
3935 Warn whenever a function call is cast to a non-matching type.
3936 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3938 @item -Wc++-compat @r{(C and Objective-C only)}
3939 Warn about ISO C constructs that are outside of the common subset of
3940 ISO C and ISO C++, e.g.@: request for implicit conversion from
3941 @code{void *} to a pointer to non-@code{void} type.
3943 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3944 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3945 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3946 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3950 @opindex Wno-cast-qual
3951 Warn whenever a pointer is cast so as to remove a type qualifier from
3952 the target type. For example, warn if a @code{const char *} is cast
3953 to an ordinary @code{char *}.
3955 Also warn when making a cast which introduces a type qualifier in an
3956 unsafe way. For example, casting @code{char **} to @code{const char **}
3957 is unsafe, as in this example:
3960 /* p is char ** value. */
3961 const char **q = (const char **) p;
3962 /* Assignment of readonly string to const char * is OK. */
3964 /* Now char** pointer points to read-only memory. */
3969 @opindex Wcast-align
3970 @opindex Wno-cast-align
3971 Warn whenever a pointer is cast such that the required alignment of the
3972 target is increased. For example, warn if a @code{char *} is cast to
3973 an @code{int *} on machines where integers can only be accessed at
3974 two- or four-byte boundaries.
3976 @item -Wwrite-strings
3977 @opindex Wwrite-strings
3978 @opindex Wno-write-strings
3979 When compiling C, give string constants the type @code{const
3980 char[@var{length}]} so that copying the address of one into a
3981 non-@code{const} @code{char *} pointer will get a warning. These
3982 warnings will help you find at compile time code that can try to write
3983 into a string constant, but only if you have been very careful about
3984 using @code{const} in declarations and prototypes. Otherwise, it will
3985 just be a nuisance. This is why we did not make @option{-Wall} request
3988 When compiling C++, warn about the deprecated conversion from string
3989 literals to @code{char *}. This warning is enabled by default for C++
3994 @opindex Wno-clobbered
3995 Warn for variables that might be changed by @samp{longjmp} or
3996 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
3999 @opindex Wconversion
4000 @opindex Wno-conversion
4001 Warn for implicit conversions that may alter a value. This includes
4002 conversions between real and integer, like @code{abs (x)} when
4003 @code{x} is @code{double}; conversions between signed and unsigned,
4004 like @code{unsigned ui = -1}; and conversions to smaller types, like
4005 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
4006 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
4007 changed by the conversion like in @code{abs (2.0)}. Warnings about
4008 conversions between signed and unsigned integers can be disabled by
4009 using @option{-Wno-sign-conversion}.
4011 For C++, also warn for confusing overload resolution for user-defined
4012 conversions; and conversions that will never use a type conversion
4013 operator: conversions to @code{void}, the same type, a base class or a
4014 reference to them. Warnings about conversions between signed and
4015 unsigned integers are disabled by default in C++ unless
4016 @option{-Wsign-conversion} is explicitly enabled.
4018 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
4019 @opindex Wconversion-null
4020 @opindex Wno-conversion-null
4021 Do not warn for conversions between @code{NULL} and non-pointer
4022 types. @option{-Wconversion-null} is enabled by default.
4025 @opindex Wempty-body
4026 @opindex Wno-empty-body
4027 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
4028 while} statement. This warning is also enabled by @option{-Wextra}.
4030 @item -Wenum-compare
4031 @opindex Wenum-compare
4032 @opindex Wno-enum-compare
4033 Warn about a comparison between values of different enum types. In C++
4034 this warning is enabled by default. In C this warning is enabled by
4037 @item -Wjump-misses-init @r{(C, Objective-C only)}
4038 @opindex Wjump-misses-init
4039 @opindex Wno-jump-misses-init
4040 Warn if a @code{goto} statement or a @code{switch} statement jumps
4041 forward across the initialization of a variable, or jumps backward to a
4042 label after the variable has been initialized. This only warns about
4043 variables which are initialized when they are declared. This warning is
4044 only supported for C and Objective C; in C++ this sort of branch is an
4047 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
4048 can be disabled with the @option{-Wno-jump-misses-init} option.
4050 @item -Wsign-compare
4051 @opindex Wsign-compare
4052 @opindex Wno-sign-compare
4053 @cindex warning for comparison of signed and unsigned values
4054 @cindex comparison of signed and unsigned values, warning
4055 @cindex signed and unsigned values, comparison warning
4056 Warn when a comparison between signed and unsigned values could produce
4057 an incorrect result when the signed value is converted to unsigned.
4058 This warning is also enabled by @option{-Wextra}; to get the other warnings
4059 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
4061 @item -Wsign-conversion
4062 @opindex Wsign-conversion
4063 @opindex Wno-sign-conversion
4064 Warn for implicit conversions that may change the sign of an integer
4065 value, like assigning a signed integer expression to an unsigned
4066 integer variable. An explicit cast silences the warning. In C, this
4067 option is enabled also by @option{-Wconversion}.
4071 @opindex Wno-address
4072 Warn about suspicious uses of memory addresses. These include using
4073 the address of a function in a conditional expression, such as
4074 @code{void func(void); if (func)}, and comparisons against the memory
4075 address of a string literal, such as @code{if (x == "abc")}. Such
4076 uses typically indicate a programmer error: the address of a function
4077 always evaluates to true, so their use in a conditional usually
4078 indicate that the programmer forgot the parentheses in a function
4079 call; and comparisons against string literals result in unspecified
4080 behavior and are not portable in C, so they usually indicate that the
4081 programmer intended to use @code{strcmp}. This warning is enabled by
4085 @opindex Wlogical-op
4086 @opindex Wno-logical-op
4087 Warn about suspicious uses of logical operators in expressions.
4088 This includes using logical operators in contexts where a
4089 bit-wise operator is likely to be expected.
4091 @item -Waggregate-return
4092 @opindex Waggregate-return
4093 @opindex Wno-aggregate-return
4094 Warn if any functions that return structures or unions are defined or
4095 called. (In languages where you can return an array, this also elicits
4098 @item -Wno-attributes
4099 @opindex Wno-attributes
4100 @opindex Wattributes
4101 Do not warn if an unexpected @code{__attribute__} is used, such as
4102 unrecognized attributes, function attributes applied to variables,
4103 etc. This will not stop errors for incorrect use of supported
4106 @item -Wno-builtin-macro-redefined
4107 @opindex Wno-builtin-macro-redefined
4108 @opindex Wbuiltin-macro-redefined
4109 Do not warn if certain built-in macros are redefined. This suppresses
4110 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
4111 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
4113 @item -Wstrict-prototypes @r{(C and Objective-C only)}
4114 @opindex Wstrict-prototypes
4115 @opindex Wno-strict-prototypes
4116 Warn if a function is declared or defined without specifying the
4117 argument types. (An old-style function definition is permitted without
4118 a warning if preceded by a declaration which specifies the argument
4121 @item -Wold-style-declaration @r{(C and Objective-C only)}
4122 @opindex Wold-style-declaration
4123 @opindex Wno-old-style-declaration
4124 Warn for obsolescent usages, according to the C Standard, in a
4125 declaration. For example, warn if storage-class specifiers like
4126 @code{static} are not the first things in a declaration. This warning
4127 is also enabled by @option{-Wextra}.
4129 @item -Wold-style-definition @r{(C and Objective-C only)}
4130 @opindex Wold-style-definition
4131 @opindex Wno-old-style-definition
4132 Warn if an old-style function definition is used. A warning is given
4133 even if there is a previous prototype.
4135 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
4136 @opindex Wmissing-parameter-type
4137 @opindex Wno-missing-parameter-type
4138 A function parameter is declared without a type specifier in K&R-style
4145 This warning is also enabled by @option{-Wextra}.
4147 @item -Wmissing-prototypes @r{(C and Objective-C only)}
4148 @opindex Wmissing-prototypes
4149 @opindex Wno-missing-prototypes
4150 Warn if a global function is defined without a previous prototype
4151 declaration. This warning is issued even if the definition itself
4152 provides a prototype. The aim is to detect global functions that fail
4153 to be declared in header files.
4155 @item -Wmissing-declarations
4156 @opindex Wmissing-declarations
4157 @opindex Wno-missing-declarations
4158 Warn if a global function is defined without a previous declaration.
4159 Do so even if the definition itself provides a prototype.
4160 Use this option to detect global functions that are not declared in
4161 header files. In C++, no warnings are issued for function templates,
4162 or for inline functions, or for functions in anonymous namespaces.
4164 @item -Wmissing-field-initializers
4165 @opindex Wmissing-field-initializers
4166 @opindex Wno-missing-field-initializers
4170 Warn if a structure's initializer has some fields missing. For
4171 example, the following code would cause such a warning, because
4172 @code{x.h} is implicitly zero:
4175 struct s @{ int f, g, h; @};
4176 struct s x = @{ 3, 4 @};
4179 This option does not warn about designated initializers, so the following
4180 modification would not trigger a warning:
4183 struct s @{ int f, g, h; @};
4184 struct s x = @{ .f = 3, .g = 4 @};
4187 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
4188 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
4190 @item -Wmissing-format-attribute
4191 @opindex Wmissing-format-attribute
4192 @opindex Wno-missing-format-attribute
4195 Warn about function pointers which might be candidates for @code{format}
4196 attributes. Note these are only possible candidates, not absolute ones.
4197 GCC will guess that function pointers with @code{format} attributes that
4198 are used in assignment, initialization, parameter passing or return
4199 statements should have a corresponding @code{format} attribute in the
4200 resulting type. I.e.@: the left-hand side of the assignment or
4201 initialization, the type of the parameter variable, or the return type
4202 of the containing function respectively should also have a @code{format}
4203 attribute to avoid the warning.
4205 GCC will also warn about function definitions which might be
4206 candidates for @code{format} attributes. Again, these are only
4207 possible candidates. GCC will guess that @code{format} attributes
4208 might be appropriate for any function that calls a function like
4209 @code{vprintf} or @code{vscanf}, but this might not always be the
4210 case, and some functions for which @code{format} attributes are
4211 appropriate may not be detected.
4213 @item -Wno-multichar
4214 @opindex Wno-multichar
4216 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4217 Usually they indicate a typo in the user's code, as they have
4218 implementation-defined values, and should not be used in portable code.
4220 @item -Wnormalized=<none|id|nfc|nfkc>
4221 @opindex Wnormalized=
4224 @cindex character set, input normalization
4225 In ISO C and ISO C++, two identifiers are different if they are
4226 different sequences of characters. However, sometimes when characters
4227 outside the basic ASCII character set are used, you can have two
4228 different character sequences that look the same. To avoid confusion,
4229 the ISO 10646 standard sets out some @dfn{normalization rules} which
4230 when applied ensure that two sequences that look the same are turned into
4231 the same sequence. GCC can warn you if you are using identifiers which
4232 have not been normalized; this option controls that warning.
4234 There are four levels of warning that GCC supports. The default is
4235 @option{-Wnormalized=nfc}, which warns about any identifier which is
4236 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4237 recommended form for most uses.
4239 Unfortunately, there are some characters which ISO C and ISO C++ allow
4240 in identifiers that when turned into NFC aren't allowable as
4241 identifiers. That is, there's no way to use these symbols in portable
4242 ISO C or C++ and have all your identifiers in NFC@.
4243 @option{-Wnormalized=id} suppresses the warning for these characters.
4244 It is hoped that future versions of the standards involved will correct
4245 this, which is why this option is not the default.
4247 You can switch the warning off for all characters by writing
4248 @option{-Wnormalized=none}. You would only want to do this if you
4249 were using some other normalization scheme (like ``D''), because
4250 otherwise you can easily create bugs that are literally impossible to see.
4252 Some characters in ISO 10646 have distinct meanings but look identical
4253 in some fonts or display methodologies, especially once formatting has
4254 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4255 LETTER N'', will display just like a regular @code{n} which has been
4256 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4257 normalization scheme to convert all these into a standard form as
4258 well, and GCC will warn if your code is not in NFKC if you use
4259 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4260 about every identifier that contains the letter O because it might be
4261 confused with the digit 0, and so is not the default, but may be
4262 useful as a local coding convention if the programming environment is
4263 unable to be fixed to display these characters distinctly.
4265 @item -Wno-deprecated
4266 @opindex Wno-deprecated
4267 @opindex Wdeprecated
4268 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4270 @item -Wno-deprecated-declarations
4271 @opindex Wno-deprecated-declarations
4272 @opindex Wdeprecated-declarations
4273 Do not warn about uses of functions (@pxref{Function Attributes}),
4274 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4275 Attributes}) marked as deprecated by using the @code{deprecated}
4279 @opindex Wno-overflow
4281 Do not warn about compile-time overflow in constant expressions.
4283 @item -Woverride-init @r{(C and Objective-C only)}
4284 @opindex Woverride-init
4285 @opindex Wno-override-init
4289 Warn if an initialized field without side effects is overridden when
4290 using designated initializers (@pxref{Designated Inits, , Designated
4293 This warning is included in @option{-Wextra}. To get other
4294 @option{-Wextra} warnings without this one, use @samp{-Wextra
4295 -Wno-override-init}.
4300 Warn if a structure is given the packed attribute, but the packed
4301 attribute has no effect on the layout or size of the structure.
4302 Such structures may be mis-aligned for little benefit. For
4303 instance, in this code, the variable @code{f.x} in @code{struct bar}
4304 will be misaligned even though @code{struct bar} does not itself
4305 have the packed attribute:
4312 @} __attribute__((packed));
4320 @item -Wpacked-bitfield-compat
4321 @opindex Wpacked-bitfield-compat
4322 @opindex Wno-packed-bitfield-compat
4323 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4324 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4325 the change can lead to differences in the structure layout. GCC
4326 informs you when the offset of such a field has changed in GCC 4.4.
4327 For example there is no longer a 4-bit padding between field @code{a}
4328 and @code{b} in this structure:
4335 @} __attribute__ ((packed));
4338 This warning is enabled by default. Use
4339 @option{-Wno-packed-bitfield-compat} to disable this warning.
4344 Warn if padding is included in a structure, either to align an element
4345 of the structure or to align the whole structure. Sometimes when this
4346 happens it is possible to rearrange the fields of the structure to
4347 reduce the padding and so make the structure smaller.
4349 @item -Wredundant-decls
4350 @opindex Wredundant-decls
4351 @opindex Wno-redundant-decls
4352 Warn if anything is declared more than once in the same scope, even in
4353 cases where multiple declaration is valid and changes nothing.
4355 @item -Wnested-externs @r{(C and Objective-C only)}
4356 @opindex Wnested-externs
4357 @opindex Wno-nested-externs
4358 Warn if an @code{extern} declaration is encountered within a function.
4363 Warn if a function can not be inlined and it was declared as inline.
4364 Even with this option, the compiler will not warn about failures to
4365 inline functions declared in system headers.
4367 The compiler uses a variety of heuristics to determine whether or not
4368 to inline a function. For example, the compiler takes into account
4369 the size of the function being inlined and the amount of inlining
4370 that has already been done in the current function. Therefore,
4371 seemingly insignificant changes in the source program can cause the
4372 warnings produced by @option{-Winline} to appear or disappear.
4374 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4375 @opindex Wno-invalid-offsetof
4376 @opindex Winvalid-offsetof
4377 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4378 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4379 to a non-POD type is undefined. In existing C++ implementations,
4380 however, @samp{offsetof} typically gives meaningful results even when
4381 applied to certain kinds of non-POD types. (Such as a simple
4382 @samp{struct} that fails to be a POD type only by virtue of having a
4383 constructor.) This flag is for users who are aware that they are
4384 writing nonportable code and who have deliberately chosen to ignore the
4387 The restrictions on @samp{offsetof} may be relaxed in a future version
4388 of the C++ standard.
4390 @item -Wno-int-to-pointer-cast
4391 @opindex Wno-int-to-pointer-cast
4392 @opindex Wint-to-pointer-cast
4393 Suppress warnings from casts to pointer type of an integer of a
4394 different size. In C++, casting to a pointer type of smaller size is
4395 an error. @option{Wint-to-pointer-cast} is enabled by default.
4398 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4399 @opindex Wno-pointer-to-int-cast
4400 @opindex Wpointer-to-int-cast
4401 Suppress warnings from casts from a pointer to an integer type of a
4405 @opindex Winvalid-pch
4406 @opindex Wno-invalid-pch
4407 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4408 the search path but can't be used.
4412 @opindex Wno-long-long
4413 Warn if @samp{long long} type is used. This is enabled by either
4414 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4415 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4417 @item -Wvariadic-macros
4418 @opindex Wvariadic-macros
4419 @opindex Wno-variadic-macros
4420 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4421 alternate syntax when in pedantic ISO C99 mode. This is default.
4422 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4427 Warn if variable length array is used in the code.
4428 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4429 the variable length array.
4431 @item -Wvolatile-register-var
4432 @opindex Wvolatile-register-var
4433 @opindex Wno-volatile-register-var
4434 Warn if a register variable is declared volatile. The volatile
4435 modifier does not inhibit all optimizations that may eliminate reads
4436 and/or writes to register variables. This warning is enabled by
4439 @item -Wdisabled-optimization
4440 @opindex Wdisabled-optimization
4441 @opindex Wno-disabled-optimization
4442 Warn if a requested optimization pass is disabled. This warning does
4443 not generally indicate that there is anything wrong with your code; it
4444 merely indicates that GCC's optimizers were unable to handle the code
4445 effectively. Often, the problem is that your code is too big or too
4446 complex; GCC will refuse to optimize programs when the optimization
4447 itself is likely to take inordinate amounts of time.
4449 @item -Wpointer-sign @r{(C and Objective-C only)}
4450 @opindex Wpointer-sign
4451 @opindex Wno-pointer-sign
4452 Warn for pointer argument passing or assignment with different signedness.
4453 This option is only supported for C and Objective-C@. It is implied by
4454 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4455 @option{-Wno-pointer-sign}.
4457 @item -Wstack-protector
4458 @opindex Wstack-protector
4459 @opindex Wno-stack-protector
4460 This option is only active when @option{-fstack-protector} is active. It
4461 warns about functions that will not be protected against stack smashing.
4464 @opindex Wno-mudflap
4465 Suppress warnings about constructs that cannot be instrumented by
4468 @item -Woverlength-strings
4469 @opindex Woverlength-strings
4470 @opindex Wno-overlength-strings
4471 Warn about string constants which are longer than the ``minimum
4472 maximum'' length specified in the C standard. Modern compilers
4473 generally allow string constants which are much longer than the
4474 standard's minimum limit, but very portable programs should avoid
4475 using longer strings.
4477 The limit applies @emph{after} string constant concatenation, and does
4478 not count the trailing NUL@. In C90, the limit was 509 characters; in
4479 C99, it was raised to 4095. C++98 does not specify a normative
4480 minimum maximum, so we do not diagnose overlength strings in C++@.
4482 This option is implied by @option{-pedantic}, and can be disabled with
4483 @option{-Wno-overlength-strings}.
4485 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
4486 @opindex Wunsuffixed-float-constants
4488 GCC will issue a warning for any floating constant that does not have
4489 a suffix. When used together with @option{-Wsystem-headers} it will
4490 warn about such constants in system header files. This can be useful
4491 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4492 from the decimal floating-point extension to C99.
4495 @node Debugging Options
4496 @section Options for Debugging Your Program or GCC
4497 @cindex options, debugging
4498 @cindex debugging information options
4500 GCC has various special options that are used for debugging
4501 either your program or GCC:
4506 Produce debugging information in the operating system's native format
4507 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4510 On most systems that use stabs format, @option{-g} enables use of extra
4511 debugging information that only GDB can use; this extra information
4512 makes debugging work better in GDB but will probably make other debuggers
4514 refuse to read the program. If you want to control for certain whether
4515 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4516 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4518 GCC allows you to use @option{-g} with
4519 @option{-O}. The shortcuts taken by optimized code may occasionally
4520 produce surprising results: some variables you declared may not exist
4521 at all; flow of control may briefly move where you did not expect it;
4522 some statements may not be executed because they compute constant
4523 results or their values were already at hand; some statements may
4524 execute in different places because they were moved out of loops.
4526 Nevertheless it proves possible to debug optimized output. This makes
4527 it reasonable to use the optimizer for programs that might have bugs.
4529 The following options are useful when GCC is generated with the
4530 capability for more than one debugging format.
4534 Produce debugging information for use by GDB@. This means to use the
4535 most expressive format available (DWARF 2, stabs, or the native format
4536 if neither of those are supported), including GDB extensions if at all
4541 Produce debugging information in stabs format (if that is supported),
4542 without GDB extensions. This is the format used by DBX on most BSD
4543 systems. On MIPS, Alpha and System V Release 4 systems this option
4544 produces stabs debugging output which is not understood by DBX or SDB@.
4545 On System V Release 4 systems this option requires the GNU assembler.
4547 @item -feliminate-unused-debug-symbols
4548 @opindex feliminate-unused-debug-symbols
4549 Produce debugging information in stabs format (if that is supported),
4550 for only symbols that are actually used.
4552 @item -femit-class-debug-always
4553 Instead of emitting debugging information for a C++ class in only one
4554 object file, emit it in all object files using the class. This option
4555 should be used only with debuggers that are unable to handle the way GCC
4556 normally emits debugging information for classes because using this
4557 option will increase the size of debugging information by as much as a
4562 Produce debugging information in stabs format (if that is supported),
4563 using GNU extensions understood only by the GNU debugger (GDB)@. The
4564 use of these extensions is likely to make other debuggers crash or
4565 refuse to read the program.
4569 Produce debugging information in COFF format (if that is supported).
4570 This is the format used by SDB on most System V systems prior to
4575 Produce debugging information in XCOFF format (if that is supported).
4576 This is the format used by the DBX debugger on IBM RS/6000 systems.
4580 Produce debugging information in XCOFF format (if that is supported),
4581 using GNU extensions understood only by the GNU debugger (GDB)@. The
4582 use of these extensions is likely to make other debuggers crash or
4583 refuse to read the program, and may cause assemblers other than the GNU
4584 assembler (GAS) to fail with an error.
4586 @item -gdwarf-@var{version}
4587 @opindex gdwarf-@var{version}
4588 Produce debugging information in DWARF format (if that is
4589 supported). This is the format used by DBX on IRIX 6. The value
4590 of @var{version} may be either 2, 3 or 4; the default version is 2.
4592 Note that with DWARF version 2 some ports require, and will always
4593 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4595 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
4596 for maximum benefit.
4598 @item -gstrict-dwarf
4599 @opindex gstrict-dwarf
4600 Disallow using extensions of later DWARF standard version than selected
4601 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
4602 DWARF extensions from later standard versions is allowed.
4604 @item -gno-strict-dwarf
4605 @opindex gno-strict-dwarf
4606 Allow using extensions of later DWARF standard version than selected with
4607 @option{-gdwarf-@var{version}}.
4611 Produce debugging information in VMS debug format (if that is
4612 supported). This is the format used by DEBUG on VMS systems.
4615 @itemx -ggdb@var{level}
4616 @itemx -gstabs@var{level}
4617 @itemx -gcoff@var{level}
4618 @itemx -gxcoff@var{level}
4619 @itemx -gvms@var{level}
4620 Request debugging information and also use @var{level} to specify how
4621 much information. The default level is 2.
4623 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4626 Level 1 produces minimal information, enough for making backtraces in
4627 parts of the program that you don't plan to debug. This includes
4628 descriptions of functions and external variables, but no information
4629 about local variables and no line numbers.
4631 Level 3 includes extra information, such as all the macro definitions
4632 present in the program. Some debuggers support macro expansion when
4633 you use @option{-g3}.
4635 @option{-gdwarf-2} does not accept a concatenated debug level, because
4636 GCC used to support an option @option{-gdwarf} that meant to generate
4637 debug information in version 1 of the DWARF format (which is very
4638 different from version 2), and it would have been too confusing. That
4639 debug format is long obsolete, but the option cannot be changed now.
4640 Instead use an additional @option{-g@var{level}} option to change the
4641 debug level for DWARF.
4645 Turn off generation of debug info, if leaving out this option would have
4646 generated it, or turn it on at level 2 otherwise. The position of this
4647 argument in the command line does not matter, it takes effect after all
4648 other options are processed, and it does so only once, no matter how
4649 many times it is given. This is mainly intended to be used with
4650 @option{-fcompare-debug}.
4652 @item -fdump-final-insns@r{[}=@var{file}@r{]}
4653 @opindex fdump-final-insns
4654 Dump the final internal representation (RTL) to @var{file}. If the
4655 optional argument is omitted (or if @var{file} is @code{.}), the name
4656 of the dump file will be determined by appending @code{.gkd} to the
4657 compilation output file name.
4659 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4660 @opindex fcompare-debug
4661 @opindex fno-compare-debug
4662 If no error occurs during compilation, run the compiler a second time,
4663 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4664 passed to the second compilation. Dump the final internal
4665 representation in both compilations, and print an error if they differ.
4667 If the equal sign is omitted, the default @option{-gtoggle} is used.
4669 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4670 and nonzero, implicitly enables @option{-fcompare-debug}. If
4671 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4672 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4675 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4676 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4677 of the final representation and the second compilation, preventing even
4678 @env{GCC_COMPARE_DEBUG} from taking effect.
4680 To verify full coverage during @option{-fcompare-debug} testing, set
4681 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4682 which GCC will reject as an invalid option in any actual compilation
4683 (rather than preprocessing, assembly or linking). To get just a
4684 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4685 not overridden} will do.
4687 @item -fcompare-debug-second
4688 @opindex fcompare-debug-second
4689 This option is implicitly passed to the compiler for the second
4690 compilation requested by @option{-fcompare-debug}, along with options to
4691 silence warnings, and omitting other options that would cause
4692 side-effect compiler outputs to files or to the standard output. Dump
4693 files and preserved temporary files are renamed so as to contain the
4694 @code{.gk} additional extension during the second compilation, to avoid
4695 overwriting those generated by the first.
4697 When this option is passed to the compiler driver, it causes the
4698 @emph{first} compilation to be skipped, which makes it useful for little
4699 other than debugging the compiler proper.
4701 @item -feliminate-dwarf2-dups
4702 @opindex feliminate-dwarf2-dups
4703 Compress DWARF2 debugging information by eliminating duplicated
4704 information about each symbol. This option only makes sense when
4705 generating DWARF2 debugging information with @option{-gdwarf-2}.
4707 @item -femit-struct-debug-baseonly
4708 Emit debug information for struct-like types
4709 only when the base name of the compilation source file
4710 matches the base name of file in which the struct was defined.
4712 This option substantially reduces the size of debugging information,
4713 but at significant potential loss in type information to the debugger.
4714 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4715 See @option{-femit-struct-debug-detailed} for more detailed control.
4717 This option works only with DWARF 2.
4719 @item -femit-struct-debug-reduced
4720 Emit debug information for struct-like types
4721 only when the base name of the compilation source file
4722 matches the base name of file in which the type was defined,
4723 unless the struct is a template or defined in a system header.
4725 This option significantly reduces the size of debugging information,
4726 with some potential loss in type information to the debugger.
4727 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4728 See @option{-femit-struct-debug-detailed} for more detailed control.
4730 This option works only with DWARF 2.
4732 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4733 Specify the struct-like types
4734 for which the compiler will generate debug information.
4735 The intent is to reduce duplicate struct debug information
4736 between different object files within the same program.
4738 This option is a detailed version of
4739 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4740 which will serve for most needs.
4742 A specification has the syntax
4743 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4745 The optional first word limits the specification to
4746 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4747 A struct type is used directly when it is the type of a variable, member.
4748 Indirect uses arise through pointers to structs.
4749 That is, when use of an incomplete struct would be legal, the use is indirect.
4751 @samp{struct one direct; struct two * indirect;}.
4753 The optional second word limits the specification to
4754 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4755 Generic structs are a bit complicated to explain.
4756 For C++, these are non-explicit specializations of template classes,
4757 or non-template classes within the above.
4758 Other programming languages have generics,
4759 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4761 The third word specifies the source files for those
4762 structs for which the compiler will emit debug information.
4763 The values @samp{none} and @samp{any} have the normal meaning.
4764 The value @samp{base} means that
4765 the base of name of the file in which the type declaration appears
4766 must match the base of the name of the main compilation file.
4767 In practice, this means that
4768 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4769 but types declared in other header will not.
4770 The value @samp{sys} means those types satisfying @samp{base}
4771 or declared in system or compiler headers.
4773 You may need to experiment to determine the best settings for your application.
4775 The default is @samp{-femit-struct-debug-detailed=all}.
4777 This option works only with DWARF 2.
4779 @item -fenable-icf-debug
4780 @opindex fenable-icf-debug
4781 Generate additional debug information to support identical code folding (ICF).
4782 This option only works with DWARF version 2 or higher.
4784 @item -fno-merge-debug-strings
4785 @opindex fmerge-debug-strings
4786 @opindex fno-merge-debug-strings
4787 Direct the linker to not merge together strings in the debugging
4788 information which are identical in different object files. Merging is
4789 not supported by all assemblers or linkers. Merging decreases the size
4790 of the debug information in the output file at the cost of increasing
4791 link processing time. Merging is enabled by default.
4793 @item -fdebug-prefix-map=@var{old}=@var{new}
4794 @opindex fdebug-prefix-map
4795 When compiling files in directory @file{@var{old}}, record debugging
4796 information describing them as in @file{@var{new}} instead.
4798 @item -fno-dwarf2-cfi-asm
4799 @opindex fdwarf2-cfi-asm
4800 @opindex fno-dwarf2-cfi-asm
4801 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4802 instead of using GAS @code{.cfi_*} directives.
4804 @cindex @command{prof}
4807 Generate extra code to write profile information suitable for the
4808 analysis program @command{prof}. You must use this option when compiling
4809 the source files you want data about, and you must also use it when
4812 @cindex @command{gprof}
4815 Generate extra code to write profile information suitable for the
4816 analysis program @command{gprof}. You must use this option when compiling
4817 the source files you want data about, and you must also use it when
4822 Makes the compiler print out each function name as it is compiled, and
4823 print some statistics about each pass when it finishes.
4826 @opindex ftime-report
4827 Makes the compiler print some statistics about the time consumed by each
4828 pass when it finishes.
4831 @opindex fmem-report
4832 Makes the compiler print some statistics about permanent memory
4833 allocation when it finishes.
4835 @item -fpre-ipa-mem-report
4836 @opindex fpre-ipa-mem-report
4837 @item -fpost-ipa-mem-report
4838 @opindex fpost-ipa-mem-report
4839 Makes the compiler print some statistics about permanent memory
4840 allocation before or after interprocedural optimization.
4842 @item -fprofile-arcs
4843 @opindex fprofile-arcs
4844 Add code so that program flow @dfn{arcs} are instrumented. During
4845 execution the program records how many times each branch and call is
4846 executed and how many times it is taken or returns. When the compiled
4847 program exits it saves this data to a file called
4848 @file{@var{auxname}.gcda} for each source file. The data may be used for
4849 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4850 test coverage analysis (@option{-ftest-coverage}). Each object file's
4851 @var{auxname} is generated from the name of the output file, if
4852 explicitly specified and it is not the final executable, otherwise it is
4853 the basename of the source file. In both cases any suffix is removed
4854 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4855 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4856 @xref{Cross-profiling}.
4858 @cindex @command{gcov}
4862 This option is used to compile and link code instrumented for coverage
4863 analysis. The option is a synonym for @option{-fprofile-arcs}
4864 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4865 linking). See the documentation for those options for more details.
4870 Compile the source files with @option{-fprofile-arcs} plus optimization
4871 and code generation options. For test coverage analysis, use the
4872 additional @option{-ftest-coverage} option. You do not need to profile
4873 every source file in a program.
4876 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4877 (the latter implies the former).
4880 Run the program on a representative workload to generate the arc profile
4881 information. This may be repeated any number of times. You can run
4882 concurrent instances of your program, and provided that the file system
4883 supports locking, the data files will be correctly updated. Also
4884 @code{fork} calls are detected and correctly handled (double counting
4888 For profile-directed optimizations, compile the source files again with
4889 the same optimization and code generation options plus
4890 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4891 Control Optimization}).
4894 For test coverage analysis, use @command{gcov} to produce human readable
4895 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4896 @command{gcov} documentation for further information.
4900 With @option{-fprofile-arcs}, for each function of your program GCC
4901 creates a program flow graph, then finds a spanning tree for the graph.
4902 Only arcs that are not on the spanning tree have to be instrumented: the
4903 compiler adds code to count the number of times that these arcs are
4904 executed. When an arc is the only exit or only entrance to a block, the
4905 instrumentation code can be added to the block; otherwise, a new basic
4906 block must be created to hold the instrumentation code.
4909 @item -ftest-coverage
4910 @opindex ftest-coverage
4911 Produce a notes file that the @command{gcov} code-coverage utility
4912 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4913 show program coverage. Each source file's note file is called
4914 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4915 above for a description of @var{auxname} and instructions on how to
4916 generate test coverage data. Coverage data will match the source files
4917 more closely, if you do not optimize.
4919 @item -fdbg-cnt-list
4920 @opindex fdbg-cnt-list
4921 Print the name and the counter upperbound for all debug counters.
4923 @item -fdbg-cnt=@var{counter-value-list}
4925 Set the internal debug counter upperbound. @var{counter-value-list}
4926 is a comma-separated list of @var{name}:@var{value} pairs
4927 which sets the upperbound of each debug counter @var{name} to @var{value}.
4928 All debug counters have the initial upperbound of @var{UINT_MAX},
4929 thus dbg_cnt() returns true always unless the upperbound is set by this option.
4930 e.g. With -fdbg-cnt=dce:10,tail_call:0
4931 dbg_cnt(dce) will return true only for first 10 invocations
4932 and dbg_cnt(tail_call) will return false always.
4934 @item -d@var{letters}
4935 @itemx -fdump-rtl-@var{pass}
4937 Says to make debugging dumps during compilation at times specified by
4938 @var{letters}. This is used for debugging the RTL-based passes of the
4939 compiler. The file names for most of the dumps are made by appending
4940 a pass number and a word to the @var{dumpname}, and the files are
4941 created in the directory of the output file. @var{dumpname} is
4942 generated from the name of the output file, if explicitly specified
4943 and it is not an executable, otherwise it is the basename of the
4944 source file. These switches may have different effects when
4945 @option{-E} is used for preprocessing.
4947 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
4948 @option{-d} option @var{letters}. Here are the possible
4949 letters for use in @var{pass} and @var{letters}, and their meanings:
4953 @item -fdump-rtl-alignments
4954 @opindex fdump-rtl-alignments
4955 Dump after branch alignments have been computed.
4957 @item -fdump-rtl-asmcons
4958 @opindex fdump-rtl-asmcons
4959 Dump after fixing rtl statements that have unsatisfied in/out constraints.
4961 @item -fdump-rtl-auto_inc_dec
4962 @opindex fdump-rtl-auto_inc_dec
4963 Dump after auto-inc-dec discovery. This pass is only run on
4964 architectures that have auto inc or auto dec instructions.
4966 @item -fdump-rtl-barriers
4967 @opindex fdump-rtl-barriers
4968 Dump after cleaning up the barrier instructions.
4970 @item -fdump-rtl-bbpart
4971 @opindex fdump-rtl-bbpart
4972 Dump after partitioning hot and cold basic blocks.
4974 @item -fdump-rtl-bbro
4975 @opindex fdump-rtl-bbro
4976 Dump after block reordering.
4978 @item -fdump-rtl-btl1
4979 @itemx -fdump-rtl-btl2
4980 @opindex fdump-rtl-btl2
4981 @opindex fdump-rtl-btl2
4982 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
4983 after the two branch
4984 target load optimization passes.
4986 @item -fdump-rtl-bypass
4987 @opindex fdump-rtl-bypass
4988 Dump after jump bypassing and control flow optimizations.
4990 @item -fdump-rtl-combine
4991 @opindex fdump-rtl-combine
4992 Dump after the RTL instruction combination pass.
4994 @item -fdump-rtl-compgotos
4995 @opindex fdump-rtl-compgotos
4996 Dump after duplicating the computed gotos.
4998 @item -fdump-rtl-ce1
4999 @itemx -fdump-rtl-ce2
5000 @itemx -fdump-rtl-ce3
5001 @opindex fdump-rtl-ce1
5002 @opindex fdump-rtl-ce2
5003 @opindex fdump-rtl-ce3
5004 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
5005 @option{-fdump-rtl-ce3} enable dumping after the three
5006 if conversion passes.
5008 @itemx -fdump-rtl-cprop_hardreg
5009 @opindex fdump-rtl-cprop_hardreg
5010 Dump after hard register copy propagation.
5012 @itemx -fdump-rtl-csa
5013 @opindex fdump-rtl-csa
5014 Dump after combining stack adjustments.
5016 @item -fdump-rtl-cse1
5017 @itemx -fdump-rtl-cse2
5018 @opindex fdump-rtl-cse1
5019 @opindex fdump-rtl-cse2
5020 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
5021 the two common sub-expression elimination passes.
5023 @itemx -fdump-rtl-dce
5024 @opindex fdump-rtl-dce
5025 Dump after the standalone dead code elimination passes.
5027 @itemx -fdump-rtl-dbr
5028 @opindex fdump-rtl-dbr
5029 Dump after delayed branch scheduling.
5031 @item -fdump-rtl-dce1
5032 @itemx -fdump-rtl-dce2
5033 @opindex fdump-rtl-dce1
5034 @opindex fdump-rtl-dce2
5035 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
5036 the two dead store elimination passes.
5039 @opindex fdump-rtl-eh
5040 Dump after finalization of EH handling code.
5042 @item -fdump-rtl-eh_ranges
5043 @opindex fdump-rtl-eh_ranges
5044 Dump after conversion of EH handling range regions.
5046 @item -fdump-rtl-expand
5047 @opindex fdump-rtl-expand
5048 Dump after RTL generation.
5050 @item -fdump-rtl-fwprop1
5051 @itemx -fdump-rtl-fwprop2
5052 @opindex fdump-rtl-fwprop1
5053 @opindex fdump-rtl-fwprop2
5054 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
5055 dumping after the two forward propagation passes.
5057 @item -fdump-rtl-gcse1
5058 @itemx -fdump-rtl-gcse2
5059 @opindex fdump-rtl-gcse1
5060 @opindex fdump-rtl-gcse2
5061 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
5062 after global common subexpression elimination.
5064 @item -fdump-rtl-init-regs
5065 @opindex fdump-rtl-init-regs
5066 Dump after the initialization of the registers.
5068 @item -fdump-rtl-initvals
5069 @opindex fdump-rtl-initvals
5070 Dump after the computation of the initial value sets.
5072 @itemx -fdump-rtl-into_cfglayout
5073 @opindex fdump-rtl-into_cfglayout
5074 Dump after converting to cfglayout mode.
5076 @item -fdump-rtl-ira
5077 @opindex fdump-rtl-ira
5078 Dump after iterated register allocation.
5080 @item -fdump-rtl-jump
5081 @opindex fdump-rtl-jump
5082 Dump after the second jump optimization.
5084 @item -fdump-rtl-loop2
5085 @opindex fdump-rtl-loop2
5086 @option{-fdump-rtl-loop2} enables dumping after the rtl
5087 loop optimization passes.
5089 @item -fdump-rtl-mach
5090 @opindex fdump-rtl-mach
5091 Dump after performing the machine dependent reorganization pass, if that
5094 @item -fdump-rtl-mode_sw
5095 @opindex fdump-rtl-mode_sw
5096 Dump after removing redundant mode switches.
5098 @item -fdump-rtl-rnreg
5099 @opindex fdump-rtl-rnreg
5100 Dump after register renumbering.
5102 @itemx -fdump-rtl-outof_cfglayout
5103 @opindex fdump-rtl-outof_cfglayout
5104 Dump after converting from cfglayout mode.
5106 @item -fdump-rtl-peephole2
5107 @opindex fdump-rtl-peephole2
5108 Dump after the peephole pass.
5110 @item -fdump-rtl-postreload
5111 @opindex fdump-rtl-postreload
5112 Dump after post-reload optimizations.
5114 @itemx -fdump-rtl-pro_and_epilogue
5115 @opindex fdump-rtl-pro_and_epilogue
5116 Dump after generating the function pro and epilogues.
5118 @item -fdump-rtl-regmove
5119 @opindex fdump-rtl-regmove
5120 Dump after the register move pass.
5122 @item -fdump-rtl-sched1
5123 @itemx -fdump-rtl-sched2
5124 @opindex fdump-rtl-sched1
5125 @opindex fdump-rtl-sched2
5126 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
5127 after the basic block scheduling passes.
5129 @item -fdump-rtl-see
5130 @opindex fdump-rtl-see
5131 Dump after sign extension elimination.
5133 @item -fdump-rtl-seqabstr
5134 @opindex fdump-rtl-seqabstr
5135 Dump after common sequence discovery.
5137 @item -fdump-rtl-shorten
5138 @opindex fdump-rtl-shorten
5139 Dump after shortening branches.
5141 @item -fdump-rtl-sibling
5142 @opindex fdump-rtl-sibling
5143 Dump after sibling call optimizations.
5145 @item -fdump-rtl-split1
5146 @itemx -fdump-rtl-split2
5147 @itemx -fdump-rtl-split3
5148 @itemx -fdump-rtl-split4
5149 @itemx -fdump-rtl-split5
5150 @opindex fdump-rtl-split1
5151 @opindex fdump-rtl-split2
5152 @opindex fdump-rtl-split3
5153 @opindex fdump-rtl-split4
5154 @opindex fdump-rtl-split5
5155 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
5156 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
5157 @option{-fdump-rtl-split5} enable dumping after five rounds of
5158 instruction splitting.
5160 @item -fdump-rtl-sms
5161 @opindex fdump-rtl-sms
5162 Dump after modulo scheduling. This pass is only run on some
5165 @item -fdump-rtl-stack
5166 @opindex fdump-rtl-stack
5167 Dump after conversion from GCC's "flat register file" registers to the
5168 x87's stack-like registers. This pass is only run on x86 variants.
5170 @item -fdump-rtl-subreg1
5171 @itemx -fdump-rtl-subreg2
5172 @opindex fdump-rtl-subreg1
5173 @opindex fdump-rtl-subreg2
5174 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
5175 the two subreg expansion passes.
5177 @item -fdump-rtl-unshare
5178 @opindex fdump-rtl-unshare
5179 Dump after all rtl has been unshared.
5181 @item -fdump-rtl-vartrack
5182 @opindex fdump-rtl-vartrack
5183 Dump after variable tracking.
5185 @item -fdump-rtl-vregs
5186 @opindex fdump-rtl-vregs
5187 Dump after converting virtual registers to hard registers.
5189 @item -fdump-rtl-web
5190 @opindex fdump-rtl-web
5191 Dump after live range splitting.
5193 @item -fdump-rtl-regclass
5194 @itemx -fdump-rtl-subregs_of_mode_init
5195 @itemx -fdump-rtl-subregs_of_mode_finish
5196 @itemx -fdump-rtl-dfinit
5197 @itemx -fdump-rtl-dfinish
5198 @opindex fdump-rtl-regclass
5199 @opindex fdump-rtl-subregs_of_mode_init
5200 @opindex fdump-rtl-subregs_of_mode_finish
5201 @opindex fdump-rtl-dfinit
5202 @opindex fdump-rtl-dfinish
5203 These dumps are defined but always produce empty files.
5205 @item -fdump-rtl-all
5206 @opindex fdump-rtl-all
5207 Produce all the dumps listed above.
5211 Annotate the assembler output with miscellaneous debugging information.
5215 Dump all macro definitions, at the end of preprocessing, in addition to
5220 Produce a core dump whenever an error occurs.
5224 Print statistics on memory usage, at the end of the run, to
5229 Annotate the assembler output with a comment indicating which
5230 pattern and alternative was used. The length of each instruction is
5235 Dump the RTL in the assembler output as a comment before each instruction.
5236 Also turns on @option{-dp} annotation.
5240 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
5241 dump a representation of the control flow graph suitable for viewing with VCG
5242 to @file{@var{file}.@var{pass}.vcg}.
5246 Just generate RTL for a function instead of compiling it. Usually used
5247 with @option{-fdump-rtl-expand}.
5251 Dump debugging information during parsing, to standard error.
5255 @opindex fdump-noaddr
5256 When doing debugging dumps, suppress address output. This makes it more
5257 feasible to use diff on debugging dumps for compiler invocations with
5258 different compiler binaries and/or different
5259 text / bss / data / heap / stack / dso start locations.
5261 @item -fdump-unnumbered
5262 @opindex fdump-unnumbered
5263 When doing debugging dumps, suppress instruction numbers and address output.
5264 This makes it more feasible to use diff on debugging dumps for compiler
5265 invocations with different options, in particular with and without
5268 @item -fdump-unnumbered-links
5269 @opindex fdump-unnumbered-links
5270 When doing debugging dumps (see @option{-d} option above), suppress
5271 instruction numbers for the links to the previous and next instructions
5274 @item -fdump-translation-unit @r{(C++ only)}
5275 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5276 @opindex fdump-translation-unit
5277 Dump a representation of the tree structure for the entire translation
5278 unit to a file. The file name is made by appending @file{.tu} to the
5279 source file name, and the file is created in the same directory as the
5280 output file. If the @samp{-@var{options}} form is used, @var{options}
5281 controls the details of the dump as described for the
5282 @option{-fdump-tree} options.
5284 @item -fdump-class-hierarchy @r{(C++ only)}
5285 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5286 @opindex fdump-class-hierarchy
5287 Dump a representation of each class's hierarchy and virtual function
5288 table layout to a file. The file name is made by appending
5289 @file{.class} to the source file name, and the file is created in the
5290 same directory as the output file. If the @samp{-@var{options}} form
5291 is used, @var{options} controls the details of the dump as described
5292 for the @option{-fdump-tree} options.
5294 @item -fdump-ipa-@var{switch}
5296 Control the dumping at various stages of inter-procedural analysis
5297 language tree to a file. The file name is generated by appending a
5298 switch specific suffix to the source file name, and the file is created
5299 in the same directory as the output file. The following dumps are
5304 Enables all inter-procedural analysis dumps.
5307 Dumps information about call-graph optimization, unused function removal,
5308 and inlining decisions.
5311 Dump after function inlining.
5315 @item -fdump-statistics-@var{option}
5316 @opindex fdump-statistics
5317 Enable and control dumping of pass statistics in a separate file. The
5318 file name is generated by appending a suffix ending in
5319 @samp{.statistics} to the source file name, and the file is created in
5320 the same directory as the output file. If the @samp{-@var{option}}
5321 form is used, @samp{-stats} will cause counters to be summed over the
5322 whole compilation unit while @samp{-details} will dump every event as
5323 the passes generate them. The default with no option is to sum
5324 counters for each function compiled.
5326 @item -fdump-tree-@var{switch}
5327 @itemx -fdump-tree-@var{switch}-@var{options}
5329 Control the dumping at various stages of processing the intermediate
5330 language tree to a file. The file name is generated by appending a
5331 switch specific suffix to the source file name, and the file is
5332 created in the same directory as the output file. If the
5333 @samp{-@var{options}} form is used, @var{options} is a list of
5334 @samp{-} separated options that control the details of the dump. Not
5335 all options are applicable to all dumps, those which are not
5336 meaningful will be ignored. The following options are available
5340 Print the address of each node. Usually this is not meaningful as it
5341 changes according to the environment and source file. Its primary use
5342 is for tying up a dump file with a debug environment.
5344 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
5345 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
5346 use working backward from mangled names in the assembly file.
5348 Inhibit dumping of members of a scope or body of a function merely
5349 because that scope has been reached. Only dump such items when they
5350 are directly reachable by some other path. When dumping pretty-printed
5351 trees, this option inhibits dumping the bodies of control structures.
5353 Print a raw representation of the tree. By default, trees are
5354 pretty-printed into a C-like representation.
5356 Enable more detailed dumps (not honored by every dump option).
5358 Enable dumping various statistics about the pass (not honored by every dump
5361 Enable showing basic block boundaries (disabled in raw dumps).
5363 Enable showing virtual operands for every statement.
5365 Enable showing line numbers for statements.
5367 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5369 Enable showing the tree dump for each statement.
5371 Enable showing the EH region number holding each statement.
5373 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5374 and @option{lineno}.
5377 The following tree dumps are possible:
5381 @opindex fdump-tree-original
5382 Dump before any tree based optimization, to @file{@var{file}.original}.
5385 @opindex fdump-tree-optimized
5386 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5389 @opindex fdump-tree-gimple
5390 Dump each function before and after the gimplification pass to a file. The
5391 file name is made by appending @file{.gimple} to the source file name.
5394 @opindex fdump-tree-cfg
5395 Dump the control flow graph of each function to a file. The file name is
5396 made by appending @file{.cfg} to the source file name.
5399 @opindex fdump-tree-vcg
5400 Dump the control flow graph of each function to a file in VCG format. The
5401 file name is made by appending @file{.vcg} to the source file name. Note
5402 that if the file contains more than one function, the generated file cannot
5403 be used directly by VCG@. You will need to cut and paste each function's
5404 graph into its own separate file first.
5407 @opindex fdump-tree-ch
5408 Dump each function after copying loop headers. The file name is made by
5409 appending @file{.ch} to the source file name.
5412 @opindex fdump-tree-ssa
5413 Dump SSA related information to a file. The file name is made by appending
5414 @file{.ssa} to the source file name.
5417 @opindex fdump-tree-alias
5418 Dump aliasing information for each function. The file name is made by
5419 appending @file{.alias} to the source file name.
5422 @opindex fdump-tree-ccp
5423 Dump each function after CCP@. The file name is made by appending
5424 @file{.ccp} to the source file name.
5427 @opindex fdump-tree-storeccp
5428 Dump each function after STORE-CCP@. The file name is made by appending
5429 @file{.storeccp} to the source file name.
5432 @opindex fdump-tree-pre
5433 Dump trees after partial redundancy elimination. The file name is made
5434 by appending @file{.pre} to the source file name.
5437 @opindex fdump-tree-fre
5438 Dump trees after full redundancy elimination. The file name is made
5439 by appending @file{.fre} to the source file name.
5442 @opindex fdump-tree-copyprop
5443 Dump trees after copy propagation. The file name is made
5444 by appending @file{.copyprop} to the source file name.
5446 @item store_copyprop
5447 @opindex fdump-tree-store_copyprop
5448 Dump trees after store copy-propagation. The file name is made
5449 by appending @file{.store_copyprop} to the source file name.
5452 @opindex fdump-tree-dce
5453 Dump each function after dead code elimination. The file name is made by
5454 appending @file{.dce} to the source file name.
5457 @opindex fdump-tree-mudflap
5458 Dump each function after adding mudflap instrumentation. The file name is
5459 made by appending @file{.mudflap} to the source file name.
5462 @opindex fdump-tree-sra
5463 Dump each function after performing scalar replacement of aggregates. The
5464 file name is made by appending @file{.sra} to the source file name.
5467 @opindex fdump-tree-sink
5468 Dump each function after performing code sinking. The file name is made
5469 by appending @file{.sink} to the source file name.
5472 @opindex fdump-tree-dom
5473 Dump each function after applying dominator tree optimizations. The file
5474 name is made by appending @file{.dom} to the source file name.
5477 @opindex fdump-tree-dse
5478 Dump each function after applying dead store elimination. The file
5479 name is made by appending @file{.dse} to the source file name.
5482 @opindex fdump-tree-phiopt
5483 Dump each function after optimizing PHI nodes into straightline code. The file
5484 name is made by appending @file{.phiopt} to the source file name.
5487 @opindex fdump-tree-forwprop
5488 Dump each function after forward propagating single use variables. The file
5489 name is made by appending @file{.forwprop} to the source file name.
5492 @opindex fdump-tree-copyrename
5493 Dump each function after applying the copy rename optimization. The file
5494 name is made by appending @file{.copyrename} to the source file name.
5497 @opindex fdump-tree-nrv
5498 Dump each function after applying the named return value optimization on
5499 generic trees. The file name is made by appending @file{.nrv} to the source
5503 @opindex fdump-tree-vect
5504 Dump each function after applying vectorization of loops. The file name is
5505 made by appending @file{.vect} to the source file name.
5508 @opindex fdump-tree-slp
5509 Dump each function after applying vectorization of basic blocks. The file name
5510 is made by appending @file{.slp} to the source file name.
5513 @opindex fdump-tree-vrp
5514 Dump each function after Value Range Propagation (VRP). The file name
5515 is made by appending @file{.vrp} to the source file name.
5518 @opindex fdump-tree-all
5519 Enable all the available tree dumps with the flags provided in this option.
5522 @item -ftree-vectorizer-verbose=@var{n}
5523 @opindex ftree-vectorizer-verbose
5524 This option controls the amount of debugging output the vectorizer prints.
5525 This information is written to standard error, unless
5526 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5527 in which case it is output to the usual dump listing file, @file{.vect}.
5528 For @var{n}=0 no diagnostic information is reported.
5529 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5530 and the total number of loops that got vectorized.
5531 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5532 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5533 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5534 level that @option{-fdump-tree-vect-stats} uses.
5535 Higher verbosity levels mean either more information dumped for each
5536 reported loop, or same amount of information reported for more loops:
5537 if @var{n}=3, vectorizer cost model information is reported.
5538 If @var{n}=4, alignment related information is added to the reports.
5539 If @var{n}=5, data-references related information (e.g.@: memory dependences,
5540 memory access-patterns) is added to the reports.
5541 If @var{n}=6, the vectorizer reports also non-vectorized inner-most loops
5542 that did not pass the first analysis phase (i.e., may not be countable, or
5543 may have complicated control-flow).
5544 If @var{n}=7, the vectorizer reports also non-vectorized nested loops.
5545 If @var{n}=8, SLP related information is added to the reports.
5546 For @var{n}=9, all the information the vectorizer generates during its
5547 analysis and transformation is reported. This is the same verbosity level
5548 that @option{-fdump-tree-vect-details} uses.
5550 @item -frandom-seed=@var{string}
5551 @opindex frandom-seed
5552 This option provides a seed that GCC uses when it would otherwise use
5553 random numbers. It is used to generate certain symbol names
5554 that have to be different in every compiled file. It is also used to
5555 place unique stamps in coverage data files and the object files that
5556 produce them. You can use the @option{-frandom-seed} option to produce
5557 reproducibly identical object files.
5559 The @var{string} should be different for every file you compile.
5561 @item -fsched-verbose=@var{n}
5562 @opindex fsched-verbose
5563 On targets that use instruction scheduling, this option controls the
5564 amount of debugging output the scheduler prints. This information is
5565 written to standard error, unless @option{-fdump-rtl-sched1} or
5566 @option{-fdump-rtl-sched2} is specified, in which case it is output
5567 to the usual dump listing file, @file{.sched1} or @file{.sched2}
5568 respectively. However for @var{n} greater than nine, the output is
5569 always printed to standard error.
5571 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5572 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5573 For @var{n} greater than one, it also output basic block probabilities,
5574 detailed ready list information and unit/insn info. For @var{n} greater
5575 than two, it includes RTL at abort point, control-flow and regions info.
5576 And for @var{n} over four, @option{-fsched-verbose} also includes
5580 @itemx -save-temps=cwd
5582 Store the usual ``temporary'' intermediate files permanently; place them
5583 in the current directory and name them based on the source file. Thus,
5584 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5585 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5586 preprocessed @file{foo.i} output file even though the compiler now
5587 normally uses an integrated preprocessor.
5589 When used in combination with the @option{-x} command line option,
5590 @option{-save-temps} is sensible enough to avoid over writing an
5591 input source file with the same extension as an intermediate file.
5592 The corresponding intermediate file may be obtained by renaming the
5593 source file before using @option{-save-temps}.
5595 If you invoke GCC in parallel, compiling several different source
5596 files that share a common base name in different subdirectories or the
5597 same source file compiled for multiple output destinations, it is
5598 likely that the different parallel compilers will interfere with each
5599 other, and overwrite the temporary files. For instance:
5602 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5603 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5606 may result in @file{foo.i} and @file{foo.o} being written to
5607 simultaneously by both compilers.
5609 @item -save-temps=obj
5610 @opindex save-temps=obj
5611 Store the usual ``temporary'' intermediate files permanently. If the
5612 @option{-o} option is used, the temporary files are based on the
5613 object file. If the @option{-o} option is not used, the
5614 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5619 gcc -save-temps=obj -c foo.c
5620 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5621 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5624 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5625 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5626 @file{dir2/yfoobar.o}.
5628 @item -time@r{[}=@var{file}@r{]}
5630 Report the CPU time taken by each subprocess in the compilation
5631 sequence. For C source files, this is the compiler proper and assembler
5632 (plus the linker if linking is done).
5634 Without the specification of an output file, the output looks like this:
5641 The first number on each line is the ``user time'', that is time spent
5642 executing the program itself. The second number is ``system time'',
5643 time spent executing operating system routines on behalf of the program.
5644 Both numbers are in seconds.
5646 With the specification of an output file, the output is appended to the
5647 named file, and it looks like this:
5650 0.12 0.01 cc1 @var{options}
5651 0.00 0.01 as @var{options}
5654 The ``user time'' and the ``system time'' are moved before the program
5655 name, and the options passed to the program are displayed, so that one
5656 can later tell what file was being compiled, and with which options.
5658 @item -fvar-tracking
5659 @opindex fvar-tracking
5660 Run variable tracking pass. It computes where variables are stored at each
5661 position in code. Better debugging information is then generated
5662 (if the debugging information format supports this information).
5664 It is enabled by default when compiling with optimization (@option{-Os},
5665 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5666 the debug info format supports it.
5668 @item -fvar-tracking-assignments
5669 @opindex fvar-tracking-assignments
5670 @opindex fno-var-tracking-assignments
5671 Annotate assignments to user variables early in the compilation and
5672 attempt to carry the annotations over throughout the compilation all the
5673 way to the end, in an attempt to improve debug information while
5674 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
5676 It can be enabled even if var-tracking is disabled, in which case
5677 annotations will be created and maintained, but discarded at the end.
5679 @item -fvar-tracking-assignments-toggle
5680 @opindex fvar-tracking-assignments-toggle
5681 @opindex fno-var-tracking-assignments-toggle
5682 Toggle @option{-fvar-tracking-assignments}, in the same way that
5683 @option{-gtoggle} toggles @option{-g}.
5685 @item -print-file-name=@var{library}
5686 @opindex print-file-name
5687 Print the full absolute name of the library file @var{library} that
5688 would be used when linking---and don't do anything else. With this
5689 option, GCC does not compile or link anything; it just prints the
5692 @item -print-multi-directory
5693 @opindex print-multi-directory
5694 Print the directory name corresponding to the multilib selected by any
5695 other switches present in the command line. This directory is supposed
5696 to exist in @env{GCC_EXEC_PREFIX}.
5698 @item -print-multi-lib
5699 @opindex print-multi-lib
5700 Print the mapping from multilib directory names to compiler switches
5701 that enable them. The directory name is separated from the switches by
5702 @samp{;}, and each switch starts with an @samp{@@} instead of the
5703 @samp{-}, without spaces between multiple switches. This is supposed to
5704 ease shell-processing.
5706 @item -print-multi-os-directory
5707 @opindex print-multi-os-directory
5708 Print the path to OS libraries for the selected
5709 multilib, relative to some @file{lib} subdirectory. If OS libraries are
5710 present in the @file{lib} subdirectory and no multilibs are used, this is
5711 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
5712 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
5713 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
5714 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
5716 @item -print-prog-name=@var{program}
5717 @opindex print-prog-name
5718 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5720 @item -print-libgcc-file-name
5721 @opindex print-libgcc-file-name
5722 Same as @option{-print-file-name=libgcc.a}.
5724 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5725 but you do want to link with @file{libgcc.a}. You can do
5728 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5731 @item -print-search-dirs
5732 @opindex print-search-dirs
5733 Print the name of the configured installation directory and a list of
5734 program and library directories @command{gcc} will search---and don't do anything else.
5736 This is useful when @command{gcc} prints the error message
5737 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5738 To resolve this you either need to put @file{cpp0} and the other compiler
5739 components where @command{gcc} expects to find them, or you can set the environment
5740 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5741 Don't forget the trailing @samp{/}.
5742 @xref{Environment Variables}.
5744 @item -print-sysroot
5745 @opindex print-sysroot
5746 Print the target sysroot directory that will be used during
5747 compilation. This is the target sysroot specified either at configure
5748 time or using the @option{--sysroot} option, possibly with an extra
5749 suffix that depends on compilation options. If no target sysroot is
5750 specified, the option prints nothing.
5752 @item -print-sysroot-headers-suffix
5753 @opindex print-sysroot-headers-suffix
5754 Print the suffix added to the target sysroot when searching for
5755 headers, or give an error if the compiler is not configured with such
5756 a suffix---and don't do anything else.
5759 @opindex dumpmachine
5760 Print the compiler's target machine (for example,
5761 @samp{i686-pc-linux-gnu})---and don't do anything else.
5764 @opindex dumpversion
5765 Print the compiler version (for example, @samp{3.0})---and don't do
5770 Print the compiler's built-in specs---and don't do anything else. (This
5771 is used when GCC itself is being built.) @xref{Spec Files}.
5773 @item -feliminate-unused-debug-types
5774 @opindex feliminate-unused-debug-types
5775 Normally, when producing DWARF2 output, GCC will emit debugging
5776 information for all types declared in a compilation
5777 unit, regardless of whether or not they are actually used
5778 in that compilation unit. Sometimes this is useful, such as
5779 if, in the debugger, you want to cast a value to a type that is
5780 not actually used in your program (but is declared). More often,
5781 however, this results in a significant amount of wasted space.
5782 With this option, GCC will avoid producing debug symbol output
5783 for types that are nowhere used in the source file being compiled.
5786 @node Optimize Options
5787 @section Options That Control Optimization
5788 @cindex optimize options
5789 @cindex options, optimization
5791 These options control various sorts of optimizations.
5793 Without any optimization option, the compiler's goal is to reduce the
5794 cost of compilation and to make debugging produce the expected
5795 results. Statements are independent: if you stop the program with a
5796 breakpoint between statements, you can then assign a new value to any
5797 variable or change the program counter to any other statement in the
5798 function and get exactly the results you would expect from the source
5801 Turning on optimization flags makes the compiler attempt to improve
5802 the performance and/or code size at the expense of compilation time
5803 and possibly the ability to debug the program.
5805 The compiler performs optimization based on the knowledge it has of the
5806 program. Compiling multiple files at once to a single output file mode allows
5807 the compiler to use information gained from all of the files when compiling
5810 Not all optimizations are controlled directly by a flag. Only
5811 optimizations that have a flag are listed in this section.
5813 Most optimizations are only enabled if an @option{-O} level is set on
5814 the command line. Otherwise they are disabled, even if individual
5815 optimization flags are specified.
5817 Depending on the target and how GCC was configured, a slightly different
5818 set of optimizations may be enabled at each @option{-O} level than
5819 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
5820 to find out the exact set of optimizations that are enabled at each level.
5821 @xref{Overall Options}, for examples.
5828 Optimize. Optimizing compilation takes somewhat more time, and a lot
5829 more memory for a large function.
5831 With @option{-O}, the compiler tries to reduce code size and execution
5832 time, without performing any optimizations that take a great deal of
5835 @option{-O} turns on the following optimization flags:
5838 -fcprop-registers @gol
5841 -fdelayed-branch @gol
5843 -fguess-branch-probability @gol
5844 -fif-conversion2 @gol
5845 -fif-conversion @gol
5846 -fipa-pure-const @gol
5848 -fipa-reference @gol
5850 -fsplit-wide-types @gol
5851 -ftree-builtin-call-dce @gol
5854 -ftree-copyrename @gol
5856 -ftree-dominator-opts @gol
5858 -ftree-forwprop @gol
5866 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5867 where doing so does not interfere with debugging.
5871 Optimize even more. GCC performs nearly all supported optimizations
5872 that do not involve a space-speed tradeoff.
5873 As compared to @option{-O}, this option increases both compilation time
5874 and the performance of the generated code.
5876 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5877 also turns on the following optimization flags:
5878 @gccoptlist{-fthread-jumps @gol
5879 -falign-functions -falign-jumps @gol
5880 -falign-loops -falign-labels @gol
5883 -fcse-follow-jumps -fcse-skip-blocks @gol
5884 -fdelete-null-pointer-checks @gol
5885 -fexpensive-optimizations @gol
5886 -fgcse -fgcse-lm @gol
5887 -finline-small-functions @gol
5888 -findirect-inlining @gol
5890 -foptimize-sibling-calls @gol
5891 -fpartial-inlining @gol
5894 -freorder-blocks -freorder-functions @gol
5895 -frerun-cse-after-loop @gol
5896 -fsched-interblock -fsched-spec @gol
5897 -fschedule-insns -fschedule-insns2 @gol
5898 -fstrict-aliasing -fstrict-overflow @gol
5899 -ftree-switch-conversion @gol
5903 Please note the warning under @option{-fgcse} about
5904 invoking @option{-O2} on programs that use computed gotos.
5908 Optimize yet more. @option{-O3} turns on all optimizations specified
5909 by @option{-O2} and also turns on the @option{-finline-functions},
5910 @option{-funswitch-loops}, @option{-fpredictive-commoning},
5911 @option{-fgcse-after-reload} and @option{-ftree-vectorize} options.
5915 Reduce compilation time and make debugging produce the expected
5916 results. This is the default.
5920 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5921 do not typically increase code size. It also performs further
5922 optimizations designed to reduce code size.
5924 @option{-Os} disables the following optimization flags:
5925 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5926 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5927 -fprefetch-loop-arrays -ftree-vect-loop-version}
5931 Disregard strict standards compliance. @option{-Ofast} enables all
5932 @option{-O3} optimizations. It also enables optimizations that are not
5933 valid for all standard compliant programs.
5934 It turns on @option{-ffast-math}.
5936 If you use multiple @option{-O} options, with or without level numbers,
5937 the last such option is the one that is effective.
5940 Options of the form @option{-f@var{flag}} specify machine-independent
5941 flags. Most flags have both positive and negative forms; the negative
5942 form of @option{-ffoo} would be @option{-fno-foo}. In the table
5943 below, only one of the forms is listed---the one you typically will
5944 use. You can figure out the other form by either removing @samp{no-}
5947 The following options control specific optimizations. They are either
5948 activated by @option{-O} options or are related to ones that are. You
5949 can use the following flags in the rare cases when ``fine-tuning'' of
5950 optimizations to be performed is desired.
5953 @item -fno-default-inline
5954 @opindex fno-default-inline
5955 Do not make member functions inline by default merely because they are
5956 defined inside the class scope (C++ only). Otherwise, when you specify
5957 @w{@option{-O}}, member functions defined inside class scope are compiled
5958 inline by default; i.e., you don't need to add @samp{inline} in front of
5959 the member function name.
5961 @item -fno-defer-pop
5962 @opindex fno-defer-pop
5963 Always pop the arguments to each function call as soon as that function
5964 returns. For machines which must pop arguments after a function call,
5965 the compiler normally lets arguments accumulate on the stack for several
5966 function calls and pops them all at once.
5968 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5970 @item -fforward-propagate
5971 @opindex fforward-propagate
5972 Perform a forward propagation pass on RTL@. The pass tries to combine two
5973 instructions and checks if the result can be simplified. If loop unrolling
5974 is active, two passes are performed and the second is scheduled after
5977 This option is enabled by default at optimization levels @option{-O},
5978 @option{-O2}, @option{-O3}, @option{-Os}.
5980 @item -fomit-frame-pointer
5981 @opindex fomit-frame-pointer
5982 Don't keep the frame pointer in a register for functions that
5983 don't need one. This avoids the instructions to save, set up and
5984 restore frame pointers; it also makes an extra register available
5985 in many functions. @strong{It also makes debugging impossible on
5988 On some machines, such as the VAX, this flag has no effect, because
5989 the standard calling sequence automatically handles the frame pointer
5990 and nothing is saved by pretending it doesn't exist. The
5991 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
5992 whether a target machine supports this flag. @xref{Registers,,Register
5993 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
5995 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5997 @item -foptimize-sibling-calls
5998 @opindex foptimize-sibling-calls
5999 Optimize sibling and tail recursive calls.
6001 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6005 Don't pay attention to the @code{inline} keyword. Normally this option
6006 is used to keep the compiler from expanding any functions inline.
6007 Note that if you are not optimizing, no functions can be expanded inline.
6009 @item -finline-small-functions
6010 @opindex finline-small-functions
6011 Integrate functions into their callers when their body is smaller than expected
6012 function call code (so overall size of program gets smaller). The compiler
6013 heuristically decides which functions are simple enough to be worth integrating
6016 Enabled at level @option{-O2}.
6018 @item -findirect-inlining
6019 @opindex findirect-inlining
6020 Inline also indirect calls that are discovered to be known at compile
6021 time thanks to previous inlining. This option has any effect only
6022 when inlining itself is turned on by the @option{-finline-functions}
6023 or @option{-finline-small-functions} options.
6025 Enabled at level @option{-O2}.
6027 @item -finline-functions
6028 @opindex finline-functions
6029 Integrate all simple functions into their callers. The compiler
6030 heuristically decides which functions are simple enough to be worth
6031 integrating in this way.
6033 If all calls to a given function are integrated, and the function is
6034 declared @code{static}, then the function is normally not output as
6035 assembler code in its own right.
6037 Enabled at level @option{-O3}.
6039 @item -finline-functions-called-once
6040 @opindex finline-functions-called-once
6041 Consider all @code{static} functions called once for inlining into their
6042 caller even if they are not marked @code{inline}. If a call to a given
6043 function is integrated, then the function is not output as assembler code
6046 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
6048 @item -fearly-inlining
6049 @opindex fearly-inlining
6050 Inline functions marked by @code{always_inline} and functions whose body seems
6051 smaller than the function call overhead early before doing
6052 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
6053 makes profiling significantly cheaper and usually inlining faster on programs
6054 having large chains of nested wrapper functions.
6060 Perform interprocedural scalar replacement of aggregates, removal of
6061 unused parameters and replacement of parameters passed by reference
6062 by parameters passed by value.
6064 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
6066 @item -finline-limit=@var{n}
6067 @opindex finline-limit
6068 By default, GCC limits the size of functions that can be inlined. This flag
6069 allows coarse control of this limit. @var{n} is the size of functions that
6070 can be inlined in number of pseudo instructions.
6072 Inlining is actually controlled by a number of parameters, which may be
6073 specified individually by using @option{--param @var{name}=@var{value}}.
6074 The @option{-finline-limit=@var{n}} option sets some of these parameters
6078 @item max-inline-insns-single
6079 is set to @var{n}/2.
6080 @item max-inline-insns-auto
6081 is set to @var{n}/2.
6084 See below for a documentation of the individual
6085 parameters controlling inlining and for the defaults of these parameters.
6087 @emph{Note:} there may be no value to @option{-finline-limit} that results
6088 in default behavior.
6090 @emph{Note:} pseudo instruction represents, in this particular context, an
6091 abstract measurement of function's size. In no way does it represent a count
6092 of assembly instructions and as such its exact meaning might change from one
6093 release to an another.
6095 @item -fkeep-inline-functions
6096 @opindex fkeep-inline-functions
6097 In C, emit @code{static} functions that are declared @code{inline}
6098 into the object file, even if the function has been inlined into all
6099 of its callers. This switch does not affect functions using the
6100 @code{extern inline} extension in GNU C90@. In C++, emit any and all
6101 inline functions into the object file.
6103 @item -fkeep-static-consts
6104 @opindex fkeep-static-consts
6105 Emit variables declared @code{static const} when optimization isn't turned
6106 on, even if the variables aren't referenced.
6108 GCC enables this option by default. If you want to force the compiler to
6109 check if the variable was referenced, regardless of whether or not
6110 optimization is turned on, use the @option{-fno-keep-static-consts} option.
6112 @item -fmerge-constants
6113 @opindex fmerge-constants
6114 Attempt to merge identical constants (string constants and floating point
6115 constants) across compilation units.
6117 This option is the default for optimized compilation if the assembler and
6118 linker support it. Use @option{-fno-merge-constants} to inhibit this
6121 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6123 @item -fmerge-all-constants
6124 @opindex fmerge-all-constants
6125 Attempt to merge identical constants and identical variables.
6127 This option implies @option{-fmerge-constants}. In addition to
6128 @option{-fmerge-constants} this considers e.g.@: even constant initialized
6129 arrays or initialized constant variables with integral or floating point
6130 types. Languages like C or C++ require each variable, including multiple
6131 instances of the same variable in recursive calls, to have distinct locations,
6132 so using this option will result in non-conforming
6135 @item -fmodulo-sched
6136 @opindex fmodulo-sched
6137 Perform swing modulo scheduling immediately before the first scheduling
6138 pass. This pass looks at innermost loops and reorders their
6139 instructions by overlapping different iterations.
6141 @item -fmodulo-sched-allow-regmoves
6142 @opindex fmodulo-sched-allow-regmoves
6143 Perform more aggressive SMS based modulo scheduling with register moves
6144 allowed. By setting this flag certain anti-dependences edges will be
6145 deleted which will trigger the generation of reg-moves based on the
6146 life-range analysis. This option is effective only with
6147 @option{-fmodulo-sched} enabled.
6149 @item -fno-branch-count-reg
6150 @opindex fno-branch-count-reg
6151 Do not use ``decrement and branch'' instructions on a count register,
6152 but instead generate a sequence of instructions that decrement a
6153 register, compare it against zero, then branch based upon the result.
6154 This option is only meaningful on architectures that support such
6155 instructions, which include x86, PowerPC, IA-64 and S/390.
6157 The default is @option{-fbranch-count-reg}.
6159 @item -fno-function-cse
6160 @opindex fno-function-cse
6161 Do not put function addresses in registers; make each instruction that
6162 calls a constant function contain the function's address explicitly.
6164 This option results in less efficient code, but some strange hacks
6165 that alter the assembler output may be confused by the optimizations
6166 performed when this option is not used.
6168 The default is @option{-ffunction-cse}
6170 @item -fno-zero-initialized-in-bss
6171 @opindex fno-zero-initialized-in-bss
6172 If the target supports a BSS section, GCC by default puts variables that
6173 are initialized to zero into BSS@. This can save space in the resulting
6176 This option turns off this behavior because some programs explicitly
6177 rely on variables going to the data section. E.g., so that the
6178 resulting executable can find the beginning of that section and/or make
6179 assumptions based on that.
6181 The default is @option{-fzero-initialized-in-bss}.
6183 @item -fmudflap -fmudflapth -fmudflapir
6187 @cindex bounds checking
6189 For front-ends that support it (C and C++), instrument all risky
6190 pointer/array dereferencing operations, some standard library
6191 string/heap functions, and some other associated constructs with
6192 range/validity tests. Modules so instrumented should be immune to
6193 buffer overflows, invalid heap use, and some other classes of C/C++
6194 programming errors. The instrumentation relies on a separate runtime
6195 library (@file{libmudflap}), which will be linked into a program if
6196 @option{-fmudflap} is given at link time. Run-time behavior of the
6197 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
6198 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
6201 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
6202 link if your program is multi-threaded. Use @option{-fmudflapir}, in
6203 addition to @option{-fmudflap} or @option{-fmudflapth}, if
6204 instrumentation should ignore pointer reads. This produces less
6205 instrumentation (and therefore faster execution) and still provides
6206 some protection against outright memory corrupting writes, but allows
6207 erroneously read data to propagate within a program.
6209 @item -fthread-jumps
6210 @opindex fthread-jumps
6211 Perform optimizations where we check to see if a jump branches to a
6212 location where another comparison subsumed by the first is found. If
6213 so, the first branch is redirected to either the destination of the
6214 second branch or a point immediately following it, depending on whether
6215 the condition is known to be true or false.
6217 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6219 @item -fsplit-wide-types
6220 @opindex fsplit-wide-types
6221 When using a type that occupies multiple registers, such as @code{long
6222 long} on a 32-bit system, split the registers apart and allocate them
6223 independently. This normally generates better code for those types,
6224 but may make debugging more difficult.
6226 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
6229 @item -fcse-follow-jumps
6230 @opindex fcse-follow-jumps
6231 In common subexpression elimination (CSE), scan through jump instructions
6232 when the target of the jump is not reached by any other path. For
6233 example, when CSE encounters an @code{if} statement with an
6234 @code{else} clause, CSE will follow the jump when the condition
6237 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6239 @item -fcse-skip-blocks
6240 @opindex fcse-skip-blocks
6241 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
6242 follow jumps which conditionally skip over blocks. When CSE
6243 encounters a simple @code{if} statement with no else clause,
6244 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6245 body of the @code{if}.
6247 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6249 @item -frerun-cse-after-loop
6250 @opindex frerun-cse-after-loop
6251 Re-run common subexpression elimination after loop optimizations has been
6254 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6258 Perform a global common subexpression elimination pass.
6259 This pass also performs global constant and copy propagation.
6261 @emph{Note:} When compiling a program using computed gotos, a GCC
6262 extension, you may get better runtime performance if you disable
6263 the global common subexpression elimination pass by adding
6264 @option{-fno-gcse} to the command line.
6266 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6270 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
6271 attempt to move loads which are only killed by stores into themselves. This
6272 allows a loop containing a load/store sequence to be changed to a load outside
6273 the loop, and a copy/store within the loop.
6275 Enabled by default when gcse is enabled.
6279 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6280 global common subexpression elimination. This pass will attempt to move
6281 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6282 loops containing a load/store sequence can be changed to a load before
6283 the loop and a store after the loop.
6285 Not enabled at any optimization level.
6289 When @option{-fgcse-las} is enabled, the global common subexpression
6290 elimination pass eliminates redundant loads that come after stores to the
6291 same memory location (both partial and full redundancies).
6293 Not enabled at any optimization level.
6295 @item -fgcse-after-reload
6296 @opindex fgcse-after-reload
6297 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6298 pass is performed after reload. The purpose of this pass is to cleanup
6301 @item -funsafe-loop-optimizations
6302 @opindex funsafe-loop-optimizations
6303 If given, the loop optimizer will assume that loop indices do not
6304 overflow, and that the loops with nontrivial exit condition are not
6305 infinite. This enables a wider range of loop optimizations even if
6306 the loop optimizer itself cannot prove that these assumptions are valid.
6307 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
6308 if it finds this kind of loop.
6310 @item -fcrossjumping
6311 @opindex fcrossjumping
6312 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6313 resulting code may or may not perform better than without cross-jumping.
6315 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6317 @item -fauto-inc-dec
6318 @opindex fauto-inc-dec
6319 Combine increments or decrements of addresses with memory accesses.
6320 This pass is always skipped on architectures that do not have
6321 instructions to support this. Enabled by default at @option{-O} and
6322 higher on architectures that support this.
6326 Perform dead code elimination (DCE) on RTL@.
6327 Enabled by default at @option{-O} and higher.
6331 Perform dead store elimination (DSE) on RTL@.
6332 Enabled by default at @option{-O} and higher.
6334 @item -fif-conversion
6335 @opindex fif-conversion
6336 Attempt to transform conditional jumps into branch-less equivalents. This
6337 include use of conditional moves, min, max, set flags and abs instructions, and
6338 some tricks doable by standard arithmetics. The use of conditional execution
6339 on chips where it is available is controlled by @code{if-conversion2}.
6341 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6343 @item -fif-conversion2
6344 @opindex fif-conversion2
6345 Use conditional execution (where available) to transform conditional jumps into
6346 branch-less equivalents.
6348 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6350 @item -fdelete-null-pointer-checks
6351 @opindex fdelete-null-pointer-checks
6352 Assume that programs cannot safely dereference null pointers, and that
6353 no code or data element resides there. This enables simple constant
6354 folding optimizations at all optimization levels. In addition, other
6355 optimization passes in GCC use this flag to control global dataflow
6356 analyses that eliminate useless checks for null pointers; these assume
6357 that if a pointer is checked after it has already been dereferenced,
6360 Note however that in some environments this assumption is not true.
6361 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6362 for programs which depend on that behavior.
6364 Some targets, especially embedded ones, disable this option at all levels.
6365 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6366 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6367 are enabled independently at different optimization levels.
6369 @item -fexpensive-optimizations
6370 @opindex fexpensive-optimizations
6371 Perform a number of minor optimizations that are relatively expensive.
6373 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6375 @item -foptimize-register-move
6377 @opindex foptimize-register-move
6379 Attempt to reassign register numbers in move instructions and as
6380 operands of other simple instructions in order to maximize the amount of
6381 register tying. This is especially helpful on machines with two-operand
6384 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6387 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6389 @item -fira-algorithm=@var{algorithm}
6390 Use specified coloring algorithm for the integrated register
6391 allocator. The @var{algorithm} argument should be @code{priority} or
6392 @code{CB}. The first algorithm specifies Chow's priority coloring,
6393 the second one specifies Chaitin-Briggs coloring. The second
6394 algorithm can be unimplemented for some architectures. If it is
6395 implemented, it is the default because Chaitin-Briggs coloring as a
6396 rule generates a better code.
6398 @item -fira-region=@var{region}
6399 Use specified regions for the integrated register allocator. The
6400 @var{region} argument should be one of @code{all}, @code{mixed}, or
6401 @code{one}. The first value means using all loops as register
6402 allocation regions, the second value which is the default means using
6403 all loops except for loops with small register pressure as the
6404 regions, and third one means using all function as a single region.
6405 The first value can give best result for machines with small size and
6406 irregular register set, the third one results in faster and generates
6407 decent code and the smallest size code, and the default value usually
6408 give the best results in most cases and for most architectures.
6410 @item -fira-coalesce
6411 @opindex fira-coalesce
6412 Do optimistic register coalescing. This option might be profitable for
6413 architectures with big regular register files.
6415 @item -fira-loop-pressure
6416 @opindex fira-loop-pressure
6417 Use IRA to evaluate register pressure in loops for decision to move
6418 loop invariants. Usage of this option usually results in generation
6419 of faster and smaller code on machines with big register files (>= 32
6420 registers) but it can slow compiler down.
6422 This option is enabled at level @option{-O3} for some targets.
6424 @item -fno-ira-share-save-slots
6425 @opindex fno-ira-share-save-slots
6426 Switch off sharing stack slots used for saving call used hard
6427 registers living through a call. Each hard register will get a
6428 separate stack slot and as a result function stack frame will be
6431 @item -fno-ira-share-spill-slots
6432 @opindex fno-ira-share-spill-slots
6433 Switch off sharing stack slots allocated for pseudo-registers. Each
6434 pseudo-register which did not get a hard register will get a separate
6435 stack slot and as a result function stack frame will be bigger.
6437 @item -fira-verbose=@var{n}
6438 @opindex fira-verbose
6439 Set up how verbose dump file for the integrated register allocator
6440 will be. Default value is 5. If the value is greater or equal to 10,
6441 the dump file will be stderr as if the value were @var{n} minus 10.
6443 @item -fdelayed-branch
6444 @opindex fdelayed-branch
6445 If supported for the target machine, attempt to reorder instructions
6446 to exploit instruction slots available after delayed branch
6449 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6451 @item -fschedule-insns
6452 @opindex fschedule-insns
6453 If supported for the target machine, attempt to reorder instructions to
6454 eliminate execution stalls due to required data being unavailable. This
6455 helps machines that have slow floating point or memory load instructions
6456 by allowing other instructions to be issued until the result of the load
6457 or floating point instruction is required.
6459 Enabled at levels @option{-O2}, @option{-O3}.
6461 @item -fschedule-insns2
6462 @opindex fschedule-insns2
6463 Similar to @option{-fschedule-insns}, but requests an additional pass of
6464 instruction scheduling after register allocation has been done. This is
6465 especially useful on machines with a relatively small number of
6466 registers and where memory load instructions take more than one cycle.
6468 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6470 @item -fno-sched-interblock
6471 @opindex fno-sched-interblock
6472 Don't schedule instructions across basic blocks. This is normally
6473 enabled by default when scheduling before register allocation, i.e.@:
6474 with @option{-fschedule-insns} or at @option{-O2} or higher.
6476 @item -fno-sched-spec
6477 @opindex fno-sched-spec
6478 Don't allow speculative motion of non-load instructions. This is normally
6479 enabled by default when scheduling before register allocation, i.e.@:
6480 with @option{-fschedule-insns} or at @option{-O2} or higher.
6482 @item -fsched-pressure
6483 @opindex fsched-pressure
6484 Enable register pressure sensitive insn scheduling before the register
6485 allocation. This only makes sense when scheduling before register
6486 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
6487 @option{-O2} or higher. Usage of this option can improve the
6488 generated code and decrease its size by preventing register pressure
6489 increase above the number of available hard registers and as a
6490 consequence register spills in the register allocation.
6492 @item -fsched-spec-load
6493 @opindex fsched-spec-load
6494 Allow speculative motion of some load instructions. This only makes
6495 sense when scheduling before register allocation, i.e.@: with
6496 @option{-fschedule-insns} or at @option{-O2} or higher.
6498 @item -fsched-spec-load-dangerous
6499 @opindex fsched-spec-load-dangerous
6500 Allow speculative motion of more load instructions. This only makes
6501 sense when scheduling before register allocation, i.e.@: with
6502 @option{-fschedule-insns} or at @option{-O2} or higher.
6504 @item -fsched-stalled-insns
6505 @itemx -fsched-stalled-insns=@var{n}
6506 @opindex fsched-stalled-insns
6507 Define how many insns (if any) can be moved prematurely from the queue
6508 of stalled insns into the ready list, during the second scheduling pass.
6509 @option{-fno-sched-stalled-insns} means that no insns will be moved
6510 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6511 on how many queued insns can be moved prematurely.
6512 @option{-fsched-stalled-insns} without a value is equivalent to
6513 @option{-fsched-stalled-insns=1}.
6515 @item -fsched-stalled-insns-dep
6516 @itemx -fsched-stalled-insns-dep=@var{n}
6517 @opindex fsched-stalled-insns-dep
6518 Define how many insn groups (cycles) will be examined for a dependency
6519 on a stalled insn that is candidate for premature removal from the queue
6520 of stalled insns. This has an effect only during the second scheduling pass,
6521 and only if @option{-fsched-stalled-insns} is used.
6522 @option{-fno-sched-stalled-insns-dep} is equivalent to
6523 @option{-fsched-stalled-insns-dep=0}.
6524 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6525 @option{-fsched-stalled-insns-dep=1}.
6527 @item -fsched2-use-superblocks
6528 @opindex fsched2-use-superblocks
6529 When scheduling after register allocation, do use superblock scheduling
6530 algorithm. Superblock scheduling allows motion across basic block boundaries
6531 resulting on faster schedules. This option is experimental, as not all machine
6532 descriptions used by GCC model the CPU closely enough to avoid unreliable
6533 results from the algorithm.
6535 This only makes sense when scheduling after register allocation, i.e.@: with
6536 @option{-fschedule-insns2} or at @option{-O2} or higher.
6538 @item -fsched-group-heuristic
6539 @opindex fsched-group-heuristic
6540 Enable the group heuristic in the scheduler. This heuristic favors
6541 the instruction that belongs to a schedule group. This is enabled
6542 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6543 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6545 @item -fsched-critical-path-heuristic
6546 @opindex fsched-critical-path-heuristic
6547 Enable the critical-path heuristic in the scheduler. This heuristic favors
6548 instructions on the critical path. This is enabled by default when
6549 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6550 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6552 @item -fsched-spec-insn-heuristic
6553 @opindex fsched-spec-insn-heuristic
6554 Enable the speculative instruction heuristic in the scheduler. This
6555 heuristic favors speculative instructions with greater dependency weakness.
6556 This is enabled by default when scheduling is enabled, i.e.@:
6557 with @option{-fschedule-insns} or @option{-fschedule-insns2}
6558 or at @option{-O2} or higher.
6560 @item -fsched-rank-heuristic
6561 @opindex fsched-rank-heuristic
6562 Enable the rank heuristic in the scheduler. This heuristic favors
6563 the instruction belonging to a basic block with greater size or frequency.
6564 This is enabled by default when scheduling is enabled, i.e.@:
6565 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6566 at @option{-O2} or higher.
6568 @item -fsched-last-insn-heuristic
6569 @opindex fsched-last-insn-heuristic
6570 Enable the last-instruction heuristic in the scheduler. This heuristic
6571 favors the instruction that is less dependent on the last instruction
6572 scheduled. This is enabled by default when scheduling is enabled,
6573 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6574 at @option{-O2} or higher.
6576 @item -fsched-dep-count-heuristic
6577 @opindex fsched-dep-count-heuristic
6578 Enable the dependent-count heuristic in the scheduler. This heuristic
6579 favors the instruction that has more instructions depending on it.
6580 This is enabled by default when scheduling is enabled, i.e.@:
6581 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6582 at @option{-O2} or higher.
6584 @item -freschedule-modulo-scheduled-loops
6585 @opindex freschedule-modulo-scheduled-loops
6586 The modulo scheduling comes before the traditional scheduling, if a loop
6587 was modulo scheduled we may want to prevent the later scheduling passes
6588 from changing its schedule, we use this option to control that.
6590 @item -fselective-scheduling
6591 @opindex fselective-scheduling
6592 Schedule instructions using selective scheduling algorithm. Selective
6593 scheduling runs instead of the first scheduler pass.
6595 @item -fselective-scheduling2
6596 @opindex fselective-scheduling2
6597 Schedule instructions using selective scheduling algorithm. Selective
6598 scheduling runs instead of the second scheduler pass.
6600 @item -fsel-sched-pipelining
6601 @opindex fsel-sched-pipelining
6602 Enable software pipelining of innermost loops during selective scheduling.
6603 This option has no effect until one of @option{-fselective-scheduling} or
6604 @option{-fselective-scheduling2} is turned on.
6606 @item -fsel-sched-pipelining-outer-loops
6607 @opindex fsel-sched-pipelining-outer-loops
6608 When pipelining loops during selective scheduling, also pipeline outer loops.
6609 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6611 @item -fcaller-saves
6612 @opindex fcaller-saves
6613 Enable values to be allocated in registers that will be clobbered by
6614 function calls, by emitting extra instructions to save and restore the
6615 registers around such calls. Such allocation is done only when it
6616 seems to result in better code than would otherwise be produced.
6618 This option is always enabled by default on certain machines, usually
6619 those which have no call-preserved registers to use instead.
6621 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6623 @item -fconserve-stack
6624 @opindex fconserve-stack
6625 Attempt to minimize stack usage. The compiler will attempt to use less
6626 stack space, even if that makes the program slower. This option
6627 implies setting the @option{large-stack-frame} parameter to 100
6628 and the @option{large-stack-frame-growth} parameter to 400.
6630 @item -ftree-reassoc
6631 @opindex ftree-reassoc
6632 Perform reassociation on trees. This flag is enabled by default
6633 at @option{-O} and higher.
6637 Perform partial redundancy elimination (PRE) on trees. This flag is
6638 enabled by default at @option{-O2} and @option{-O3}.
6640 @item -ftree-forwprop
6641 @opindex ftree-forwprop
6642 Perform forward propagation on trees. This flag is enabled by default
6643 at @option{-O} and higher.
6647 Perform full redundancy elimination (FRE) on trees. The difference
6648 between FRE and PRE is that FRE only considers expressions
6649 that are computed on all paths leading to the redundant computation.
6650 This analysis is faster than PRE, though it exposes fewer redundancies.
6651 This flag is enabled by default at @option{-O} and higher.
6653 @item -ftree-phiprop
6654 @opindex ftree-phiprop
6655 Perform hoisting of loads from conditional pointers on trees. This
6656 pass is enabled by default at @option{-O} and higher.
6658 @item -ftree-copy-prop
6659 @opindex ftree-copy-prop
6660 Perform copy propagation on trees. This pass eliminates unnecessary
6661 copy operations. This flag is enabled by default at @option{-O} and
6664 @item -fipa-pure-const
6665 @opindex fipa-pure-const
6666 Discover which functions are pure or constant.
6667 Enabled by default at @option{-O} and higher.
6669 @item -fipa-reference
6670 @opindex fipa-reference
6671 Discover which static variables do not escape cannot escape the
6673 Enabled by default at @option{-O} and higher.
6675 @item -fipa-struct-reorg
6676 @opindex fipa-struct-reorg
6677 Perform structure reorganization optimization, that change C-like structures
6678 layout in order to better utilize spatial locality. This transformation is
6679 affective for programs containing arrays of structures. Available in two
6680 compilation modes: profile-based (enabled with @option{-fprofile-generate})
6681 or static (which uses built-in heuristics). It works only in whole program
6682 mode, so it requires @option{-fwhole-program} and @option{-combine} to be
6683 enabled. Structures considered @samp{cold} by this transformation are not
6684 affected (see @option{--param struct-reorg-cold-struct-ratio=@var{value}}).
6686 With this flag, the program debug info reflects a new structure layout.
6690 Perform interprocedural pointer analysis and interprocedural modification
6691 and reference analysis. This option can cause excessive memory and
6692 compile-time usage on large compilation units. It is not enabled by
6693 default at any optimization level.
6696 @opindex fipa-profile
6697 Perform interprocedural profile propagation. The functions called only from
6698 cold functions are marked as cold. Also functions executed once (such as
6699 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
6700 functions and loop less parts of functions executed once are then optimized for
6702 Enabled by default at @option{-O} and higher.
6706 Perform interprocedural constant propagation.
6707 This optimization analyzes the program to determine when values passed
6708 to functions are constants and then optimizes accordingly.
6709 This optimization can substantially increase performance
6710 if the application has constants passed to functions.
6711 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
6713 @item -fipa-cp-clone
6714 @opindex fipa-cp-clone
6715 Perform function cloning to make interprocedural constant propagation stronger.
6716 When enabled, interprocedural constant propagation will perform function cloning
6717 when externally visible function can be called with constant arguments.
6718 Because this optimization can create multiple copies of functions,
6719 it may significantly increase code size
6720 (see @option{--param ipcp-unit-growth=@var{value}}).
6721 This flag is enabled by default at @option{-O3}.
6723 @item -fipa-matrix-reorg
6724 @opindex fipa-matrix-reorg
6725 Perform matrix flattening and transposing.
6726 Matrix flattening tries to replace an @math{m}-dimensional matrix
6727 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
6728 This reduces the level of indirection needed for accessing the elements
6729 of the matrix. The second optimization is matrix transposing that
6730 attempts to change the order of the matrix's dimensions in order to
6731 improve cache locality.
6732 Both optimizations need the @option{-fwhole-program} flag.
6733 Transposing is enabled only if profiling information is available.
6737 Perform forward store motion on trees. This flag is
6738 enabled by default at @option{-O} and higher.
6742 Perform sparse conditional constant propagation (CCP) on trees. This
6743 pass only operates on local scalar variables and is enabled by default
6744 at @option{-O} and higher.
6746 @item -ftree-switch-conversion
6747 Perform conversion of simple initializations in a switch to
6748 initializations from a scalar array. This flag is enabled by default
6749 at @option{-O2} and higher.
6753 Perform dead code elimination (DCE) on trees. This flag is enabled by
6754 default at @option{-O} and higher.
6756 @item -ftree-builtin-call-dce
6757 @opindex ftree-builtin-call-dce
6758 Perform conditional dead code elimination (DCE) for calls to builtin functions
6759 that may set @code{errno} but are otherwise side-effect free. This flag is
6760 enabled by default at @option{-O2} and higher if @option{-Os} is not also
6763 @item -ftree-dominator-opts
6764 @opindex ftree-dominator-opts
6765 Perform a variety of simple scalar cleanups (constant/copy
6766 propagation, redundancy elimination, range propagation and expression
6767 simplification) based on a dominator tree traversal. This also
6768 performs jump threading (to reduce jumps to jumps). This flag is
6769 enabled by default at @option{-O} and higher.
6773 Perform dead store elimination (DSE) on trees. A dead store is a store into
6774 a memory location which will later be overwritten by another store without
6775 any intervening loads. In this case the earlier store can be deleted. This
6776 flag is enabled by default at @option{-O} and higher.
6780 Perform loop header copying on trees. This is beneficial since it increases
6781 effectiveness of code motion optimizations. It also saves one jump. This flag
6782 is enabled by default at @option{-O} and higher. It is not enabled
6783 for @option{-Os}, since it usually increases code size.
6785 @item -ftree-loop-optimize
6786 @opindex ftree-loop-optimize
6787 Perform loop optimizations on trees. This flag is enabled by default
6788 at @option{-O} and higher.
6790 @item -ftree-loop-linear
6791 @opindex ftree-loop-linear
6792 Perform linear loop transformations on tree. This flag can improve cache
6793 performance and allow further loop optimizations to take place.
6795 @item -floop-interchange
6796 Perform loop interchange transformations on loops. Interchanging two
6797 nested loops switches the inner and outer loops. For example, given a
6802 A(J, I) = A(J, I) * C
6806 loop interchange will transform the loop as if the user had written:
6810 A(J, I) = A(J, I) * C
6814 which can be beneficial when @code{N} is larger than the caches,
6815 because in Fortran, the elements of an array are stored in memory
6816 contiguously by column, and the original loop iterates over rows,
6817 potentially creating at each access a cache miss. This optimization
6818 applies to all the languages supported by GCC and is not limited to
6819 Fortran. To use this code transformation, GCC has to be configured
6820 with @option{--with-ppl} and @option{--with-cloog} to enable the
6821 Graphite loop transformation infrastructure.
6823 @item -floop-strip-mine
6824 Perform loop strip mining transformations on loops. Strip mining
6825 splits a loop into two nested loops. The outer loop has strides
6826 equal to the strip size and the inner loop has strides of the
6827 original loop within a strip. The strip length can be changed
6828 using the @option{loop-block-tile-size} parameter. For example,
6835 loop strip mining will transform the loop as if the user had written:
6838 DO I = II, min (II + 50, N)
6843 This optimization applies to all the languages supported by GCC and is
6844 not limited to Fortran. To use this code transformation, GCC has to
6845 be configured with @option{--with-ppl} and @option{--with-cloog} to
6846 enable the Graphite loop transformation infrastructure.
6849 Perform loop blocking transformations on loops. Blocking strip mines
6850 each loop in the loop nest such that the memory accesses of the
6851 element loops fit inside caches. The strip length can be changed
6852 using the @option{loop-block-tile-size} parameter. For example, given
6857 A(J, I) = B(I) + C(J)
6861 loop blocking will transform the loop as if the user had written:
6865 DO I = II, min (II + 50, N)
6866 DO J = JJ, min (JJ + 50, M)
6867 A(J, I) = B(I) + C(J)
6873 which can be beneficial when @code{M} is larger than the caches,
6874 because the innermost loop will iterate over a smaller amount of data
6875 that can be kept in the caches. This optimization applies to all the
6876 languages supported by GCC and is not limited to Fortran. To use this
6877 code transformation, GCC has to be configured with @option{--with-ppl}
6878 and @option{--with-cloog} to enable the Graphite loop transformation
6881 @item -fgraphite-identity
6882 @opindex fgraphite-identity
6883 Enable the identity transformation for graphite. For every SCoP we generate
6884 the polyhedral representation and transform it back to gimple. Using
6885 @option{-fgraphite-identity} we can check the costs or benefits of the
6886 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
6887 are also performed by the code generator CLooG, like index splitting and
6888 dead code elimination in loops.
6890 @item -floop-parallelize-all
6891 Use the Graphite data dependence analysis to identify loops that can
6892 be parallelized. Parallelize all the loops that can be analyzed to
6893 not contain loop carried dependences without checking that it is
6894 profitable to parallelize the loops.
6896 @item -fcheck-data-deps
6897 @opindex fcheck-data-deps
6898 Compare the results of several data dependence analyzers. This option
6899 is used for debugging the data dependence analyzers.
6901 @item -ftree-loop-if-convert
6902 Attempt to transform conditional jumps in the innermost loops to
6903 branch-less equivalents. The intent is to remove control-flow from
6904 the innermost loops in order to improve the ability of the
6905 vectorization pass to handle these loops. This is enabled by default
6906 if vectorization is enabled.
6908 @item -ftree-loop-distribution
6909 Perform loop distribution. This flag can improve cache performance on
6910 big loop bodies and allow further loop optimizations, like
6911 parallelization or vectorization, to take place. For example, the loop
6928 @item -ftree-loop-im
6929 @opindex ftree-loop-im
6930 Perform loop invariant motion on trees. This pass moves only invariants that
6931 would be hard to handle at RTL level (function calls, operations that expand to
6932 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
6933 operands of conditions that are invariant out of the loop, so that we can use
6934 just trivial invariantness analysis in loop unswitching. The pass also includes
6937 @item -ftree-loop-ivcanon
6938 @opindex ftree-loop-ivcanon
6939 Create a canonical counter for number of iterations in the loop for that
6940 determining number of iterations requires complicated analysis. Later
6941 optimizations then may determine the number easily. Useful especially
6942 in connection with unrolling.
6946 Perform induction variable optimizations (strength reduction, induction
6947 variable merging and induction variable elimination) on trees.
6949 @item -ftree-parallelize-loops=n
6950 @opindex ftree-parallelize-loops
6951 Parallelize loops, i.e., split their iteration space to run in n threads.
6952 This is only possible for loops whose iterations are independent
6953 and can be arbitrarily reordered. The optimization is only
6954 profitable on multiprocessor machines, for loops that are CPU-intensive,
6955 rather than constrained e.g.@: by memory bandwidth. This option
6956 implies @option{-pthread}, and thus is only supported on targets
6957 that have support for @option{-pthread}.
6961 Perform function-local points-to analysis on trees. This flag is
6962 enabled by default at @option{-O} and higher.
6966 Perform scalar replacement of aggregates. This pass replaces structure
6967 references with scalars to prevent committing structures to memory too
6968 early. This flag is enabled by default at @option{-O} and higher.
6970 @item -ftree-copyrename
6971 @opindex ftree-copyrename
6972 Perform copy renaming on trees. This pass attempts to rename compiler
6973 temporaries to other variables at copy locations, usually resulting in
6974 variable names which more closely resemble the original variables. This flag
6975 is enabled by default at @option{-O} and higher.
6979 Perform temporary expression replacement during the SSA->normal phase. Single
6980 use/single def temporaries are replaced at their use location with their
6981 defining expression. This results in non-GIMPLE code, but gives the expanders
6982 much more complex trees to work on resulting in better RTL generation. This is
6983 enabled by default at @option{-O} and higher.
6985 @item -ftree-vectorize
6986 @opindex ftree-vectorize
6987 Perform loop vectorization on trees. This flag is enabled by default at
6990 @item -ftree-slp-vectorize
6991 @opindex ftree-slp-vectorize
6992 Perform basic block vectorization on trees. This flag is enabled by default at
6993 @option{-O3} and when @option{-ftree-vectorize} is enabled.
6995 @item -ftree-vect-loop-version
6996 @opindex ftree-vect-loop-version
6997 Perform loop versioning when doing loop vectorization on trees. When a loop
6998 appears to be vectorizable except that data alignment or data dependence cannot
6999 be determined at compile time then vectorized and non-vectorized versions of
7000 the loop are generated along with runtime checks for alignment or dependence
7001 to control which version is executed. This option is enabled by default
7002 except at level @option{-Os} where it is disabled.
7004 @item -fvect-cost-model
7005 @opindex fvect-cost-model
7006 Enable cost model for vectorization.
7010 Perform Value Range Propagation on trees. This is similar to the
7011 constant propagation pass, but instead of values, ranges of values are
7012 propagated. This allows the optimizers to remove unnecessary range
7013 checks like array bound checks and null pointer checks. This is
7014 enabled by default at @option{-O2} and higher. Null pointer check
7015 elimination is only done if @option{-fdelete-null-pointer-checks} is
7020 Perform tail duplication to enlarge superblock size. This transformation
7021 simplifies the control flow of the function allowing other optimizations to do
7024 @item -funroll-loops
7025 @opindex funroll-loops
7026 Unroll loops whose number of iterations can be determined at compile
7027 time or upon entry to the loop. @option{-funroll-loops} implies
7028 @option{-frerun-cse-after-loop}. This option makes code larger,
7029 and may or may not make it run faster.
7031 @item -funroll-all-loops
7032 @opindex funroll-all-loops
7033 Unroll all loops, even if their number of iterations is uncertain when
7034 the loop is entered. This usually makes programs run more slowly.
7035 @option{-funroll-all-loops} implies the same options as
7036 @option{-funroll-loops},
7038 @item -fsplit-ivs-in-unroller
7039 @opindex fsplit-ivs-in-unroller
7040 Enables expressing of values of induction variables in later iterations
7041 of the unrolled loop using the value in the first iteration. This breaks
7042 long dependency chains, thus improving efficiency of the scheduling passes.
7044 Combination of @option{-fweb} and CSE is often sufficient to obtain the
7045 same effect. However in cases the loop body is more complicated than
7046 a single basic block, this is not reliable. It also does not work at all
7047 on some of the architectures due to restrictions in the CSE pass.
7049 This optimization is enabled by default.
7051 @item -fvariable-expansion-in-unroller
7052 @opindex fvariable-expansion-in-unroller
7053 With this option, the compiler will create multiple copies of some
7054 local variables when unrolling a loop which can result in superior code.
7056 @item -fpartial-inlining
7057 @opindex fpartial-inlining
7058 Inline parts of functions. This option has any effect only
7059 when inlining itself is turned on by the @option{-finline-functions}
7060 or @option{-finline-small-functions} options.
7062 Enabled at level @option{-O2}.
7064 @item -fpredictive-commoning
7065 @opindex fpredictive-commoning
7066 Perform predictive commoning optimization, i.e., reusing computations
7067 (especially memory loads and stores) performed in previous
7068 iterations of loops.
7070 This option is enabled at level @option{-O3}.
7072 @item -fprefetch-loop-arrays
7073 @opindex fprefetch-loop-arrays
7074 If supported by the target machine, generate instructions to prefetch
7075 memory to improve the performance of loops that access large arrays.
7077 This option may generate better or worse code; results are highly
7078 dependent on the structure of loops within the source code.
7080 Disabled at level @option{-Os}.
7083 @itemx -fno-peephole2
7084 @opindex fno-peephole
7085 @opindex fno-peephole2
7086 Disable any machine-specific peephole optimizations. The difference
7087 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
7088 are implemented in the compiler; some targets use one, some use the
7089 other, a few use both.
7091 @option{-fpeephole} is enabled by default.
7092 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7094 @item -fno-guess-branch-probability
7095 @opindex fno-guess-branch-probability
7096 Do not guess branch probabilities using heuristics.
7098 GCC will use heuristics to guess branch probabilities if they are
7099 not provided by profiling feedback (@option{-fprofile-arcs}). These
7100 heuristics are based on the control flow graph. If some branch probabilities
7101 are specified by @samp{__builtin_expect}, then the heuristics will be
7102 used to guess branch probabilities for the rest of the control flow graph,
7103 taking the @samp{__builtin_expect} info into account. The interactions
7104 between the heuristics and @samp{__builtin_expect} can be complex, and in
7105 some cases, it may be useful to disable the heuristics so that the effects
7106 of @samp{__builtin_expect} are easier to understand.
7108 The default is @option{-fguess-branch-probability} at levels
7109 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7111 @item -freorder-blocks
7112 @opindex freorder-blocks
7113 Reorder basic blocks in the compiled function in order to reduce number of
7114 taken branches and improve code locality.
7116 Enabled at levels @option{-O2}, @option{-O3}.
7118 @item -freorder-blocks-and-partition
7119 @opindex freorder-blocks-and-partition
7120 In addition to reordering basic blocks in the compiled function, in order
7121 to reduce number of taken branches, partitions hot and cold basic blocks
7122 into separate sections of the assembly and .o files, to improve
7123 paging and cache locality performance.
7125 This optimization is automatically turned off in the presence of
7126 exception handling, for linkonce sections, for functions with a user-defined
7127 section attribute and on any architecture that does not support named
7130 @item -freorder-functions
7131 @opindex freorder-functions
7132 Reorder functions in the object file in order to
7133 improve code locality. This is implemented by using special
7134 subsections @code{.text.hot} for most frequently executed functions and
7135 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
7136 the linker so object file format must support named sections and linker must
7137 place them in a reasonable way.
7139 Also profile feedback must be available in to make this option effective. See
7140 @option{-fprofile-arcs} for details.
7142 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7144 @item -fstrict-aliasing
7145 @opindex fstrict-aliasing
7146 Allow the compiler to assume the strictest aliasing rules applicable to
7147 the language being compiled. For C (and C++), this activates
7148 optimizations based on the type of expressions. In particular, an
7149 object of one type is assumed never to reside at the same address as an
7150 object of a different type, unless the types are almost the same. For
7151 example, an @code{unsigned int} can alias an @code{int}, but not a
7152 @code{void*} or a @code{double}. A character type may alias any other
7155 @anchor{Type-punning}Pay special attention to code like this:
7168 The practice of reading from a different union member than the one most
7169 recently written to (called ``type-punning'') is common. Even with
7170 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
7171 is accessed through the union type. So, the code above will work as
7172 expected. @xref{Structures unions enumerations and bit-fields
7173 implementation}. However, this code might not:
7184 Similarly, access by taking the address, casting the resulting pointer
7185 and dereferencing the result has undefined behavior, even if the cast
7186 uses a union type, e.g.:
7190 return ((union a_union *) &d)->i;
7194 The @option{-fstrict-aliasing} option is enabled at levels
7195 @option{-O2}, @option{-O3}, @option{-Os}.
7197 @item -fstrict-overflow
7198 @opindex fstrict-overflow
7199 Allow the compiler to assume strict signed overflow rules, depending
7200 on the language being compiled. For C (and C++) this means that
7201 overflow when doing arithmetic with signed numbers is undefined, which
7202 means that the compiler may assume that it will not happen. This
7203 permits various optimizations. For example, the compiler will assume
7204 that an expression like @code{i + 10 > i} will always be true for
7205 signed @code{i}. This assumption is only valid if signed overflow is
7206 undefined, as the expression is false if @code{i + 10} overflows when
7207 using twos complement arithmetic. When this option is in effect any
7208 attempt to determine whether an operation on signed numbers will
7209 overflow must be written carefully to not actually involve overflow.
7211 This option also allows the compiler to assume strict pointer
7212 semantics: given a pointer to an object, if adding an offset to that
7213 pointer does not produce a pointer to the same object, the addition is
7214 undefined. This permits the compiler to conclude that @code{p + u >
7215 p} is always true for a pointer @code{p} and unsigned integer
7216 @code{u}. This assumption is only valid because pointer wraparound is
7217 undefined, as the expression is false if @code{p + u} overflows using
7218 twos complement arithmetic.
7220 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
7221 that integer signed overflow is fully defined: it wraps. When
7222 @option{-fwrapv} is used, there is no difference between
7223 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
7224 integers. With @option{-fwrapv} certain types of overflow are
7225 permitted. For example, if the compiler gets an overflow when doing
7226 arithmetic on constants, the overflowed value can still be used with
7227 @option{-fwrapv}, but not otherwise.
7229 The @option{-fstrict-overflow} option is enabled at levels
7230 @option{-O2}, @option{-O3}, @option{-Os}.
7232 @item -falign-functions
7233 @itemx -falign-functions=@var{n}
7234 @opindex falign-functions
7235 Align the start of functions to the next power-of-two greater than
7236 @var{n}, skipping up to @var{n} bytes. For instance,
7237 @option{-falign-functions=32} aligns functions to the next 32-byte
7238 boundary, but @option{-falign-functions=24} would align to the next
7239 32-byte boundary only if this can be done by skipping 23 bytes or less.
7241 @option{-fno-align-functions} and @option{-falign-functions=1} are
7242 equivalent and mean that functions will not be aligned.
7244 Some assemblers only support this flag when @var{n} is a power of two;
7245 in that case, it is rounded up.
7247 If @var{n} is not specified or is zero, use a machine-dependent default.
7249 Enabled at levels @option{-O2}, @option{-O3}.
7251 @item -falign-labels
7252 @itemx -falign-labels=@var{n}
7253 @opindex falign-labels
7254 Align all branch targets to a power-of-two boundary, skipping up to
7255 @var{n} bytes like @option{-falign-functions}. This option can easily
7256 make code slower, because it must insert dummy operations for when the
7257 branch target is reached in the usual flow of the code.
7259 @option{-fno-align-labels} and @option{-falign-labels=1} are
7260 equivalent and mean that labels will not be aligned.
7262 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
7263 are greater than this value, then their values are used instead.
7265 If @var{n} is not specified or is zero, use a machine-dependent default
7266 which is very likely to be @samp{1}, meaning no alignment.
7268 Enabled at levels @option{-O2}, @option{-O3}.
7271 @itemx -falign-loops=@var{n}
7272 @opindex falign-loops
7273 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7274 like @option{-falign-functions}. The hope is that the loop will be
7275 executed many times, which will make up for any execution of the dummy
7278 @option{-fno-align-loops} and @option{-falign-loops=1} are
7279 equivalent and mean that loops will not be aligned.
7281 If @var{n} is not specified or is zero, use a machine-dependent default.
7283 Enabled at levels @option{-O2}, @option{-O3}.
7286 @itemx -falign-jumps=@var{n}
7287 @opindex falign-jumps
7288 Align branch targets to a power-of-two boundary, for branch targets
7289 where the targets can only be reached by jumping, skipping up to @var{n}
7290 bytes like @option{-falign-functions}. In this case, no dummy operations
7293 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7294 equivalent and mean that loops will not be aligned.
7296 If @var{n} is not specified or is zero, use a machine-dependent default.
7298 Enabled at levels @option{-O2}, @option{-O3}.
7300 @item -funit-at-a-time
7301 @opindex funit-at-a-time
7302 This option is left for compatibility reasons. @option{-funit-at-a-time}
7303 has no effect, while @option{-fno-unit-at-a-time} implies
7304 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7308 @item -fno-toplevel-reorder
7309 @opindex fno-toplevel-reorder
7310 Do not reorder top-level functions, variables, and @code{asm}
7311 statements. Output them in the same order that they appear in the
7312 input file. When this option is used, unreferenced static variables
7313 will not be removed. This option is intended to support existing code
7314 which relies on a particular ordering. For new code, it is better to
7317 Enabled at level @option{-O0}. When disabled explicitly, it also imply
7318 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
7323 Constructs webs as commonly used for register allocation purposes and assign
7324 each web individual pseudo register. This allows the register allocation pass
7325 to operate on pseudos directly, but also strengthens several other optimization
7326 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7327 however, make debugging impossible, since variables will no longer stay in a
7330 Enabled by default with @option{-funroll-loops}.
7332 @item -fwhole-program
7333 @opindex fwhole-program
7334 Assume that the current compilation unit represents the whole program being
7335 compiled. All public functions and variables with the exception of @code{main}
7336 and those merged by attribute @code{externally_visible} become static functions
7337 and in effect are optimized more aggressively by interprocedural optimizers. If @command{gold} is used as the linker plugin, @code{externally_visible} attributes are automatically added to functions (not variable yet due to a current @command{gold} issue) that are accessed outside of LTO objects according to resolution file produced by @command{gold}. For other linkers that cannot generate resolution file, explicit @code{externally_visible} attributes are still necessary.
7338 While this option is equivalent to proper use of the @code{static} keyword for
7339 programs consisting of a single file, in combination with option
7340 @option{-combine}, @option{-flto} or @option{-fwhopr} this flag can be used to
7341 compile many smaller scale programs since the functions and variables become
7342 local for the whole combined compilation unit, not for the single source file
7345 This option implies @option{-fwhole-file} for Fortran programs.
7349 This option runs the standard link-time optimizer. When invoked
7350 with source code, it generates GIMPLE (one of GCC's internal
7351 representations) and writes it to special ELF sections in the object
7352 file. When the object files are linked together, all the function
7353 bodies are read from these ELF sections and instantiated as if they
7354 had been part of the same translation unit.
7356 To use the link-timer optimizer, @option{-flto} needs to be specified at
7357 compile time and during the final link. For example,
7360 gcc -c -O2 -flto foo.c
7361 gcc -c -O2 -flto bar.c
7362 gcc -o myprog -flto -O2 foo.o bar.o
7365 The first two invocations to GCC will save a bytecode representation
7366 of GIMPLE into special ELF sections inside @file{foo.o} and
7367 @file{bar.o}. The final invocation will read the GIMPLE bytecode from
7368 @file{foo.o} and @file{bar.o}, merge the two files into a single
7369 internal image, and compile the result as usual. Since both
7370 @file{foo.o} and @file{bar.o} are merged into a single image, this
7371 causes all the inter-procedural analyses and optimizations in GCC to
7372 work across the two files as if they were a single one. This means,
7373 for example, that the inliner will be able to inline functions in
7374 @file{bar.o} into functions in @file{foo.o} and vice-versa.
7376 Another (simpler) way to enable link-time optimization is,
7379 gcc -o myprog -flto -O2 foo.c bar.c
7382 The above will generate bytecode for @file{foo.c} and @file{bar.c},
7383 merge them together into a single GIMPLE representation and optimize
7384 them as usual to produce @file{myprog}.
7386 The only important thing to keep in mind is that to enable link-time
7387 optimizations the @option{-flto} flag needs to be passed to both the
7388 compile and the link commands.
7390 Note that when a file is compiled with @option{-flto}, the generated
7391 object file will be larger than a regular object file because it will
7392 contain GIMPLE bytecodes and the usual final code. This means that
7393 object files with LTO information can be linked as a normal object
7394 file. So, in the previous example, if the final link is done with
7397 gcc -o myprog foo.o bar.o
7400 The only difference will be that no inter-procedural optimizations
7401 will be applied to produce @file{myprog}. The two object files
7402 @file{foo.o} and @file{bar.o} will be simply sent to the regular
7405 Additionally, the optimization flags used to compile individual files
7406 are not necessarily related to those used at link-time. For instance,
7409 gcc -c -O0 -flto foo.c
7410 gcc -c -O0 -flto bar.c
7411 gcc -o myprog -flto -O3 foo.o bar.o
7414 This will produce individual object files with unoptimized assembler
7415 code, but the resulting binary @file{myprog} will be optimized at
7416 @option{-O3}. Now, if the final binary is generated without
7417 @option{-flto}, then @file{myprog} will not be optimized.
7419 When producing the final binary with @option{-flto}, GCC will only
7420 apply link-time optimizations to those files that contain bytecode.
7421 Therefore, you can mix and match object files and libraries with
7422 GIMPLE bytecodes and final object code. GCC will automatically select
7423 which files to optimize in LTO mode and which files to link without
7426 There are some code generation flags that GCC will preserve when
7427 generating bytecodes, as they need to be used during the final link
7428 stage. Currently, the following options are saved into the GIMPLE
7429 bytecode files: @option{-fPIC}, @option{-fcommon} and all the
7430 @option{-m} target flags.
7432 At link time, these options are read-in and reapplied. Note that the
7433 current implementation makes no attempt at recognizing conflicting
7434 values for these options. If two or more files have a conflicting
7435 value (e.g., one file is compiled with @option{-fPIC} and another
7436 isn't), the compiler will simply use the last value read from the
7437 bytecode files. It is recommended, then, that all the files
7438 participating in the same link be compiled with the same options.
7440 Another feature of LTO is that it is possible to apply interprocedural
7441 optimizations on files written in different languages. This requires
7442 some support in the language front end. Currently, the C, C++ and
7443 Fortran front ends are capable of emitting GIMPLE bytecodes, so
7444 something like this should work
7449 gfortran -c -flto baz.f90
7450 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
7453 Notice that the final link is done with @command{g++} to get the C++
7454 runtime libraries and @option{-lgfortran} is added to get the Fortran
7455 runtime libraries. In general, when mixing languages in LTO mode, you
7456 should use the same link command used when mixing languages in a
7457 regular (non-LTO) compilation. This means that if your build process
7458 was mixing languages before, all you need to add is @option{-flto} to
7459 all the compile and link commands.
7461 If LTO encounters objects with C linkage declared with incompatible
7462 types in separate translation units to be linked together (undefined
7463 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
7464 issued. The behavior is still undefined at runtime.
7466 If object files containing GIMPLE bytecode are stored in a library
7467 archive, say @file{libfoo.a}, it is possible to extract and use them
7468 in an LTO link if you are using @command{gold} as the linker (which,
7469 in turn requires GCC to be configured with @option{--enable-gold}).
7470 To enable this feature, use the flag @option{-fuse-linker-plugin} at
7474 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
7477 With the linker plugin enabled, @command{gold} will extract the needed
7478 GIMPLE files from @file{libfoo.a} and pass them on to the running GCC
7479 to make them part of the aggregated GIMPLE image to be optimized.
7481 If you are not using @command{gold} and/or do not specify
7482 @option{-fuse-linker-plugin} then the objects inside @file{libfoo.a}
7483 will be extracted and linked as usual, but they will not participate
7484 in the LTO optimization process.
7486 Link time optimizations do not require the presence of the whole
7487 program to operate. If the program does not require any symbols to
7488 be exported, it is possible to combine @option{-flto} and
7489 @option{-fwhopr} with @option{-fwhole-program} to allow the
7490 interprocedural optimizers to use more aggressive assumptions which
7491 may lead to improved optimization opportunities.
7493 Regarding portability: the current implementation of LTO makes no
7494 attempt at generating bytecode that can be ported between different
7495 types of hosts. The bytecode files are versioned and there is a
7496 strict version check, so bytecode files generated in one version of
7497 GCC will not work with an older/newer version of GCC.
7499 Link time optimization does not play well with generating debugging
7500 information. Combining @option{-flto} or @option{-fwhopr} with
7501 @option{-g} is experimental.
7503 This option is disabled by default.
7505 @item -fwhopr[=@var{n}]
7507 This option is identical in functionality to @option{-flto} but it
7508 differs in how the final link stage is executed. Instead of loading
7509 all the function bodies in memory, the callgraph is analyzed and
7510 optimization decisions are made (whole program analysis or WPA). Once
7511 optimization decisions are made, the callgraph is partitioned and the
7512 different sections are compiled separately (local transformations or
7513 LTRANS)@. This process allows optimizations on very large programs
7514 that otherwise would not fit in memory. This option enables
7515 @option{-fwpa} and @option{-fltrans} automatically.
7517 If you specify the optional @var{n} the link stage is executed in
7518 parallel using @var{n} parallel jobs by utilizing an installed
7519 @command{make} program. The environment variable @env{MAKE} may be
7520 used to override the program used.
7522 Disabled by default.
7526 This is an internal option used by GCC when compiling with
7527 @option{-fwhopr}. You should never need to use it.
7529 This option runs the link-time optimizer in the whole-program-analysis
7530 (WPA) mode, which reads in summary information from all inputs and
7531 performs a whole-program analysis based on summary information only.
7532 It generates object files for subsequent runs of the link-time
7533 optimizer where individual object files are optimized using both
7534 summary information from the WPA mode and the actual function bodies.
7535 It then drives the LTRANS phase.
7537 Disabled by default.
7541 This is an internal option used by GCC when compiling with
7542 @option{-fwhopr}. You should never need to use it.
7544 This option runs the link-time optimizer in the local-transformation (LTRANS)
7545 mode, which reads in output from a previous run of the LTO in WPA mode.
7546 In the LTRANS mode, LTO optimizes an object and produces the final assembly.
7548 Disabled by default.
7550 @item -fltrans-output-list=@var{file}
7551 @opindex fltrans-output-list
7552 This is an internal option used by GCC when compiling with
7553 @option{-fwhopr}. You should never need to use it.
7555 This option specifies a file to which the names of LTRANS output files are
7556 written. This option is only meaningful in conjunction with @option{-fwpa}.
7558 Disabled by default.
7560 @item -flto-compression-level=@var{n}
7561 This option specifies the level of compression used for intermediate
7562 language written to LTO object files, and is only meaningful in
7563 conjunction with LTO mode (@option{-fwhopr}, @option{-flto}). Valid
7564 values are 0 (no compression) to 9 (maximum compression). Values
7565 outside this range are clamped to either 0 or 9. If the option is not
7566 given, a default balanced compression setting is used.
7569 Prints a report with internal details on the workings of the link-time
7570 optimizer. The contents of this report vary from version to version,
7571 it is meant to be useful to GCC developers when processing object
7572 files in LTO mode (via @option{-fwhopr} or @option{-flto}).
7574 Disabled by default.
7576 @item -fuse-linker-plugin
7577 Enables the extraction of objects with GIMPLE bytecode information
7578 from library archives. This option relies on features available only
7579 in @command{gold}, so to use this you must configure GCC with
7580 @option{--enable-gold}. See @option{-flto} for a description on the
7581 effect of this flag and how to use it.
7583 Disabled by default.
7585 @item -fcprop-registers
7586 @opindex fcprop-registers
7587 After register allocation and post-register allocation instruction splitting,
7588 we perform a copy-propagation pass to try to reduce scheduling dependencies
7589 and occasionally eliminate the copy.
7591 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7593 @item -fprofile-correction
7594 @opindex fprofile-correction
7595 Profiles collected using an instrumented binary for multi-threaded programs may
7596 be inconsistent due to missed counter updates. When this option is specified,
7597 GCC will use heuristics to correct or smooth out such inconsistencies. By
7598 default, GCC will emit an error message when an inconsistent profile is detected.
7600 @item -fprofile-dir=@var{path}
7601 @opindex fprofile-dir
7603 Set the directory to search the profile data files in to @var{path}.
7604 This option affects only the profile data generated by
7605 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
7606 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
7607 and its related options.
7608 By default, GCC will use the current directory as @var{path}
7609 thus the profile data file will appear in the same directory as the object file.
7611 @item -fprofile-generate
7612 @itemx -fprofile-generate=@var{path}
7613 @opindex fprofile-generate
7615 Enable options usually used for instrumenting application to produce
7616 profile useful for later recompilation with profile feedback based
7617 optimization. You must use @option{-fprofile-generate} both when
7618 compiling and when linking your program.
7620 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
7622 If @var{path} is specified, GCC will look at the @var{path} to find
7623 the profile feedback data files. See @option{-fprofile-dir}.
7626 @itemx -fprofile-use=@var{path}
7627 @opindex fprofile-use
7628 Enable profile feedback directed optimizations, and optimizations
7629 generally profitable only with profile feedback available.
7631 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
7632 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
7634 By default, GCC emits an error message if the feedback profiles do not
7635 match the source code. This error can be turned into a warning by using
7636 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
7639 If @var{path} is specified, GCC will look at the @var{path} to find
7640 the profile feedback data files. See @option{-fprofile-dir}.
7643 The following options control compiler behavior regarding floating
7644 point arithmetic. These options trade off between speed and
7645 correctness. All must be specifically enabled.
7649 @opindex ffloat-store
7650 Do not store floating point variables in registers, and inhibit other
7651 options that might change whether a floating point value is taken from a
7654 @cindex floating point precision
7655 This option prevents undesirable excess precision on machines such as
7656 the 68000 where the floating registers (of the 68881) keep more
7657 precision than a @code{double} is supposed to have. Similarly for the
7658 x86 architecture. For most programs, the excess precision does only
7659 good, but a few programs rely on the precise definition of IEEE floating
7660 point. Use @option{-ffloat-store} for such programs, after modifying
7661 them to store all pertinent intermediate computations into variables.
7663 @item -fexcess-precision=@var{style}
7664 @opindex fexcess-precision
7665 This option allows further control over excess precision on machines
7666 where floating-point registers have more precision than the IEEE
7667 @code{float} and @code{double} types and the processor does not
7668 support operations rounding to those types. By default,
7669 @option{-fexcess-precision=fast} is in effect; this means that
7670 operations are carried out in the precision of the registers and that
7671 it is unpredictable when rounding to the types specified in the source
7672 code takes place. When compiling C, if
7673 @option{-fexcess-precision=standard} is specified then excess
7674 precision will follow the rules specified in ISO C99; in particular,
7675 both casts and assignments cause values to be rounded to their
7676 semantic types (whereas @option{-ffloat-store} only affects
7677 assignments). This option is enabled by default for C if a strict
7678 conformance option such as @option{-std=c99} is used.
7681 @option{-fexcess-precision=standard} is not implemented for languages
7682 other than C, and has no effect if
7683 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
7684 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
7685 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
7686 semantics apply without excess precision, and in the latter, rounding
7691 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
7692 @option{-ffinite-math-only}, @option{-fno-rounding-math},
7693 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
7695 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
7697 This option is not turned on by any @option{-O} option since
7698 it can result in incorrect output for programs which depend on
7699 an exact implementation of IEEE or ISO rules/specifications for
7700 math functions. It may, however, yield faster code for programs
7701 that do not require the guarantees of these specifications.
7703 @item -fno-math-errno
7704 @opindex fno-math-errno
7705 Do not set ERRNO after calling math functions that are executed
7706 with a single instruction, e.g., sqrt. A program that relies on
7707 IEEE exceptions for math error handling may want to use this flag
7708 for speed while maintaining IEEE arithmetic compatibility.
7710 This option is not turned on by any @option{-O} option since
7711 it can result in incorrect output for programs which depend on
7712 an exact implementation of IEEE or ISO rules/specifications for
7713 math functions. It may, however, yield faster code for programs
7714 that do not require the guarantees of these specifications.
7716 The default is @option{-fmath-errno}.
7718 On Darwin systems, the math library never sets @code{errno}. There is
7719 therefore no reason for the compiler to consider the possibility that
7720 it might, and @option{-fno-math-errno} is the default.
7722 @item -funsafe-math-optimizations
7723 @opindex funsafe-math-optimizations
7725 Allow optimizations for floating-point arithmetic that (a) assume
7726 that arguments and results are valid and (b) may violate IEEE or
7727 ANSI standards. When used at link-time, it may include libraries
7728 or startup files that change the default FPU control word or other
7729 similar optimizations.
7731 This option is not turned on by any @option{-O} option since
7732 it can result in incorrect output for programs which depend on
7733 an exact implementation of IEEE or ISO rules/specifications for
7734 math functions. It may, however, yield faster code for programs
7735 that do not require the guarantees of these specifications.
7736 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
7737 @option{-fassociative-math} and @option{-freciprocal-math}.
7739 The default is @option{-fno-unsafe-math-optimizations}.
7741 @item -fassociative-math
7742 @opindex fassociative-math
7744 Allow re-association of operands in series of floating-point operations.
7745 This violates the ISO C and C++ language standard by possibly changing
7746 computation result. NOTE: re-ordering may change the sign of zero as
7747 well as ignore NaNs and inhibit or create underflow or overflow (and
7748 thus cannot be used on a code which relies on rounding behavior like
7749 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
7750 and thus may not be used when ordered comparisons are required.
7751 This option requires that both @option{-fno-signed-zeros} and
7752 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
7753 much sense with @option{-frounding-math}. For Fortran the option
7754 is automatically enabled when both @option{-fno-signed-zeros} and
7755 @option{-fno-trapping-math} are in effect.
7757 The default is @option{-fno-associative-math}.
7759 @item -freciprocal-math
7760 @opindex freciprocal-math
7762 Allow the reciprocal of a value to be used instead of dividing by
7763 the value if this enables optimizations. For example @code{x / y}
7764 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
7765 is subject to common subexpression elimination. Note that this loses
7766 precision and increases the number of flops operating on the value.
7768 The default is @option{-fno-reciprocal-math}.
7770 @item -ffinite-math-only
7771 @opindex ffinite-math-only
7772 Allow optimizations for floating-point arithmetic that assume
7773 that arguments and results are not NaNs or +-Infs.
7775 This option is not turned on by any @option{-O} option since
7776 it can result in incorrect output for programs which depend on
7777 an exact implementation of IEEE or ISO rules/specifications for
7778 math functions. It may, however, yield faster code for programs
7779 that do not require the guarantees of these specifications.
7781 The default is @option{-fno-finite-math-only}.
7783 @item -fno-signed-zeros
7784 @opindex fno-signed-zeros
7785 Allow optimizations for floating point arithmetic that ignore the
7786 signedness of zero. IEEE arithmetic specifies the behavior of
7787 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
7788 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
7789 This option implies that the sign of a zero result isn't significant.
7791 The default is @option{-fsigned-zeros}.
7793 @item -fno-trapping-math
7794 @opindex fno-trapping-math
7795 Compile code assuming that floating-point operations cannot generate
7796 user-visible traps. These traps include division by zero, overflow,
7797 underflow, inexact result and invalid operation. This option requires
7798 that @option{-fno-signaling-nans} be in effect. Setting this option may
7799 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
7801 This option should never be turned on by any @option{-O} option since
7802 it can result in incorrect output for programs which depend on
7803 an exact implementation of IEEE or ISO rules/specifications for
7806 The default is @option{-ftrapping-math}.
7808 @item -frounding-math
7809 @opindex frounding-math
7810 Disable transformations and optimizations that assume default floating
7811 point rounding behavior. This is round-to-zero for all floating point
7812 to integer conversions, and round-to-nearest for all other arithmetic
7813 truncations. This option should be specified for programs that change
7814 the FP rounding mode dynamically, or that may be executed with a
7815 non-default rounding mode. This option disables constant folding of
7816 floating point expressions at compile-time (which may be affected by
7817 rounding mode) and arithmetic transformations that are unsafe in the
7818 presence of sign-dependent rounding modes.
7820 The default is @option{-fno-rounding-math}.
7822 This option is experimental and does not currently guarantee to
7823 disable all GCC optimizations that are affected by rounding mode.
7824 Future versions of GCC may provide finer control of this setting
7825 using C99's @code{FENV_ACCESS} pragma. This command line option
7826 will be used to specify the default state for @code{FENV_ACCESS}.
7828 @item -fsignaling-nans
7829 @opindex fsignaling-nans
7830 Compile code assuming that IEEE signaling NaNs may generate user-visible
7831 traps during floating-point operations. Setting this option disables
7832 optimizations that may change the number of exceptions visible with
7833 signaling NaNs. This option implies @option{-ftrapping-math}.
7835 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
7838 The default is @option{-fno-signaling-nans}.
7840 This option is experimental and does not currently guarantee to
7841 disable all GCC optimizations that affect signaling NaN behavior.
7843 @item -fsingle-precision-constant
7844 @opindex fsingle-precision-constant
7845 Treat floating point constant as single precision constant instead of
7846 implicitly converting it to double precision constant.
7848 @item -fcx-limited-range
7849 @opindex fcx-limited-range
7850 When enabled, this option states that a range reduction step is not
7851 needed when performing complex division. Also, there is no checking
7852 whether the result of a complex multiplication or division is @code{NaN
7853 + I*NaN}, with an attempt to rescue the situation in that case. The
7854 default is @option{-fno-cx-limited-range}, but is enabled by
7855 @option{-ffast-math}.
7857 This option controls the default setting of the ISO C99
7858 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
7861 @item -fcx-fortran-rules
7862 @opindex fcx-fortran-rules
7863 Complex multiplication and division follow Fortran rules. Range
7864 reduction is done as part of complex division, but there is no checking
7865 whether the result of a complex multiplication or division is @code{NaN
7866 + I*NaN}, with an attempt to rescue the situation in that case.
7868 The default is @option{-fno-cx-fortran-rules}.
7872 The following options control optimizations that may improve
7873 performance, but are not enabled by any @option{-O} options. This
7874 section includes experimental options that may produce broken code.
7877 @item -fbranch-probabilities
7878 @opindex fbranch-probabilities
7879 After running a program compiled with @option{-fprofile-arcs}
7880 (@pxref{Debugging Options,, Options for Debugging Your Program or
7881 @command{gcc}}), you can compile it a second time using
7882 @option{-fbranch-probabilities}, to improve optimizations based on
7883 the number of times each branch was taken. When the program
7884 compiled with @option{-fprofile-arcs} exits it saves arc execution
7885 counts to a file called @file{@var{sourcename}.gcda} for each source
7886 file. The information in this data file is very dependent on the
7887 structure of the generated code, so you must use the same source code
7888 and the same optimization options for both compilations.
7890 With @option{-fbranch-probabilities}, GCC puts a
7891 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
7892 These can be used to improve optimization. Currently, they are only
7893 used in one place: in @file{reorg.c}, instead of guessing which path a
7894 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
7895 exactly determine which path is taken more often.
7897 @item -fprofile-values
7898 @opindex fprofile-values
7899 If combined with @option{-fprofile-arcs}, it adds code so that some
7900 data about values of expressions in the program is gathered.
7902 With @option{-fbranch-probabilities}, it reads back the data gathered
7903 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
7904 notes to instructions for their later usage in optimizations.
7906 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
7910 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
7911 a code to gather information about values of expressions.
7913 With @option{-fbranch-probabilities}, it reads back the data gathered
7914 and actually performs the optimizations based on them.
7915 Currently the optimizations include specialization of division operation
7916 using the knowledge about the value of the denominator.
7918 @item -frename-registers
7919 @opindex frename-registers
7920 Attempt to avoid false dependencies in scheduled code by making use
7921 of registers left over after register allocation. This optimization
7922 will most benefit processors with lots of registers. Depending on the
7923 debug information format adopted by the target, however, it can
7924 make debugging impossible, since variables will no longer stay in
7925 a ``home register''.
7927 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
7931 Perform tail duplication to enlarge superblock size. This transformation
7932 simplifies the control flow of the function allowing other optimizations to do
7935 Enabled with @option{-fprofile-use}.
7937 @item -funroll-loops
7938 @opindex funroll-loops
7939 Unroll loops whose number of iterations can be determined at compile time or
7940 upon entry to the loop. @option{-funroll-loops} implies
7941 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
7942 It also turns on complete loop peeling (i.e.@: complete removal of loops with
7943 small constant number of iterations). This option makes code larger, and may
7944 or may not make it run faster.
7946 Enabled with @option{-fprofile-use}.
7948 @item -funroll-all-loops
7949 @opindex funroll-all-loops
7950 Unroll all loops, even if their number of iterations is uncertain when
7951 the loop is entered. This usually makes programs run more slowly.
7952 @option{-funroll-all-loops} implies the same options as
7953 @option{-funroll-loops}.
7956 @opindex fpeel-loops
7957 Peels the loops for that there is enough information that they do not
7958 roll much (from profile feedback). It also turns on complete loop peeling
7959 (i.e.@: complete removal of loops with small constant number of iterations).
7961 Enabled with @option{-fprofile-use}.
7963 @item -fmove-loop-invariants
7964 @opindex fmove-loop-invariants
7965 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
7966 at level @option{-O1}
7968 @item -funswitch-loops
7969 @opindex funswitch-loops
7970 Move branches with loop invariant conditions out of the loop, with duplicates
7971 of the loop on both branches (modified according to result of the condition).
7973 @item -ffunction-sections
7974 @itemx -fdata-sections
7975 @opindex ffunction-sections
7976 @opindex fdata-sections
7977 Place each function or data item into its own section in the output
7978 file if the target supports arbitrary sections. The name of the
7979 function or the name of the data item determines the section's name
7982 Use these options on systems where the linker can perform optimizations
7983 to improve locality of reference in the instruction space. Most systems
7984 using the ELF object format and SPARC processors running Solaris 2 have
7985 linkers with such optimizations. AIX may have these optimizations in
7988 Only use these options when there are significant benefits from doing
7989 so. When you specify these options, the assembler and linker will
7990 create larger object and executable files and will also be slower.
7991 You will not be able to use @code{gprof} on all systems if you
7992 specify this option and you may have problems with debugging if
7993 you specify both this option and @option{-g}.
7995 @item -fbranch-target-load-optimize
7996 @opindex fbranch-target-load-optimize
7997 Perform branch target register load optimization before prologue / epilogue
7999 The use of target registers can typically be exposed only during reload,
8000 thus hoisting loads out of loops and doing inter-block scheduling needs
8001 a separate optimization pass.
8003 @item -fbranch-target-load-optimize2
8004 @opindex fbranch-target-load-optimize2
8005 Perform branch target register load optimization after prologue / epilogue
8008 @item -fbtr-bb-exclusive
8009 @opindex fbtr-bb-exclusive
8010 When performing branch target register load optimization, don't reuse
8011 branch target registers in within any basic block.
8013 @item -fstack-protector
8014 @opindex fstack-protector
8015 Emit extra code to check for buffer overflows, such as stack smashing
8016 attacks. This is done by adding a guard variable to functions with
8017 vulnerable objects. This includes functions that call alloca, and
8018 functions with buffers larger than 8 bytes. The guards are initialized
8019 when a function is entered and then checked when the function exits.
8020 If a guard check fails, an error message is printed and the program exits.
8022 @item -fstack-protector-all
8023 @opindex fstack-protector-all
8024 Like @option{-fstack-protector} except that all functions are protected.
8026 @item -fsection-anchors
8027 @opindex fsection-anchors
8028 Try to reduce the number of symbolic address calculations by using
8029 shared ``anchor'' symbols to address nearby objects. This transformation
8030 can help to reduce the number of GOT entries and GOT accesses on some
8033 For example, the implementation of the following function @code{foo}:
8037 int foo (void) @{ return a + b + c; @}
8040 would usually calculate the addresses of all three variables, but if you
8041 compile it with @option{-fsection-anchors}, it will access the variables
8042 from a common anchor point instead. The effect is similar to the
8043 following pseudocode (which isn't valid C):
8048 register int *xr = &x;
8049 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
8053 Not all targets support this option.
8055 @item --param @var{name}=@var{value}
8057 In some places, GCC uses various constants to control the amount of
8058 optimization that is done. For example, GCC will not inline functions
8059 that contain more that a certain number of instructions. You can
8060 control some of these constants on the command-line using the
8061 @option{--param} option.
8063 The names of specific parameters, and the meaning of the values, are
8064 tied to the internals of the compiler, and are subject to change
8065 without notice in future releases.
8067 In each case, the @var{value} is an integer. The allowable choices for
8068 @var{name} are given in the following table:
8071 @item struct-reorg-cold-struct-ratio
8072 The threshold ratio (as a percentage) between a structure frequency
8073 and the frequency of the hottest structure in the program. This parameter
8074 is used by struct-reorg optimization enabled by @option{-fipa-struct-reorg}.
8075 We say that if the ratio of a structure frequency, calculated by profiling,
8076 to the hottest structure frequency in the program is less than this
8077 parameter, then structure reorganization is not applied to this structure.
8080 @item predictable-branch-outcome
8081 When branch is predicted to be taken with probability lower than this threshold
8082 (in percent), then it is considered well predictable. The default is 10.
8084 @item max-crossjump-edges
8085 The maximum number of incoming edges to consider for crossjumping.
8086 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
8087 the number of edges incoming to each block. Increasing values mean
8088 more aggressive optimization, making the compile time increase with
8089 probably small improvement in executable size.
8091 @item min-crossjump-insns
8092 The minimum number of instructions which must be matched at the end
8093 of two blocks before crossjumping will be performed on them. This
8094 value is ignored in the case where all instructions in the block being
8095 crossjumped from are matched. The default value is 5.
8097 @item max-grow-copy-bb-insns
8098 The maximum code size expansion factor when copying basic blocks
8099 instead of jumping. The expansion is relative to a jump instruction.
8100 The default value is 8.
8102 @item max-goto-duplication-insns
8103 The maximum number of instructions to duplicate to a block that jumps
8104 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
8105 passes, GCC factors computed gotos early in the compilation process,
8106 and unfactors them as late as possible. Only computed jumps at the
8107 end of a basic blocks with no more than max-goto-duplication-insns are
8108 unfactored. The default value is 8.
8110 @item max-delay-slot-insn-search
8111 The maximum number of instructions to consider when looking for an
8112 instruction to fill a delay slot. If more than this arbitrary number of
8113 instructions is searched, the time savings from filling the delay slot
8114 will be minimal so stop searching. Increasing values mean more
8115 aggressive optimization, making the compile time increase with probably
8116 small improvement in executable run time.
8118 @item max-delay-slot-live-search
8119 When trying to fill delay slots, the maximum number of instructions to
8120 consider when searching for a block with valid live register
8121 information. Increasing this arbitrarily chosen value means more
8122 aggressive optimization, increasing the compile time. This parameter
8123 should be removed when the delay slot code is rewritten to maintain the
8126 @item max-gcse-memory
8127 The approximate maximum amount of memory that will be allocated in
8128 order to perform the global common subexpression elimination
8129 optimization. If more memory than specified is required, the
8130 optimization will not be done.
8132 @item max-pending-list-length
8133 The maximum number of pending dependencies scheduling will allow
8134 before flushing the current state and starting over. Large functions
8135 with few branches or calls can create excessively large lists which
8136 needlessly consume memory and resources.
8138 @item max-inline-insns-single
8139 Several parameters control the tree inliner used in gcc.
8140 This number sets the maximum number of instructions (counted in GCC's
8141 internal representation) in a single function that the tree inliner
8142 will consider for inlining. This only affects functions declared
8143 inline and methods implemented in a class declaration (C++).
8144 The default value is 300.
8146 @item max-inline-insns-auto
8147 When you use @option{-finline-functions} (included in @option{-O3}),
8148 a lot of functions that would otherwise not be considered for inlining
8149 by the compiler will be investigated. To those functions, a different
8150 (more restrictive) limit compared to functions declared inline can
8152 The default value is 40.
8154 @item large-function-insns
8155 The limit specifying really large functions. For functions larger than this
8156 limit after inlining, inlining is constrained by
8157 @option{--param large-function-growth}. This parameter is useful primarily
8158 to avoid extreme compilation time caused by non-linear algorithms used by the
8160 The default value is 2700.
8162 @item large-function-growth
8163 Specifies maximal growth of large function caused by inlining in percents.
8164 The default value is 100 which limits large function growth to 2.0 times
8167 @item large-unit-insns
8168 The limit specifying large translation unit. Growth caused by inlining of
8169 units larger than this limit is limited by @option{--param inline-unit-growth}.
8170 For small units this might be too tight (consider unit consisting of function A
8171 that is inline and B that just calls A three time. If B is small relative to
8172 A, the growth of unit is 300\% and yet such inlining is very sane. For very
8173 large units consisting of small inlineable functions however the overall unit
8174 growth limit is needed to avoid exponential explosion of code size. Thus for
8175 smaller units, the size is increased to @option{--param large-unit-insns}
8176 before applying @option{--param inline-unit-growth}. The default is 10000
8178 @item inline-unit-growth
8179 Specifies maximal overall growth of the compilation unit caused by inlining.
8180 The default value is 30 which limits unit growth to 1.3 times the original
8183 @item ipcp-unit-growth
8184 Specifies maximal overall growth of the compilation unit caused by
8185 interprocedural constant propagation. The default value is 10 which limits
8186 unit growth to 1.1 times the original size.
8188 @item large-stack-frame
8189 The limit specifying large stack frames. While inlining the algorithm is trying
8190 to not grow past this limit too much. Default value is 256 bytes.
8192 @item large-stack-frame-growth
8193 Specifies maximal growth of large stack frames caused by inlining in percents.
8194 The default value is 1000 which limits large stack frame growth to 11 times
8197 @item max-inline-insns-recursive
8198 @itemx max-inline-insns-recursive-auto
8199 Specifies maximum number of instructions out-of-line copy of self recursive inline
8200 function can grow into by performing recursive inlining.
8202 For functions declared inline @option{--param max-inline-insns-recursive} is
8203 taken into account. For function not declared inline, recursive inlining
8204 happens only when @option{-finline-functions} (included in @option{-O3}) is
8205 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
8206 default value is 450.
8208 @item max-inline-recursive-depth
8209 @itemx max-inline-recursive-depth-auto
8210 Specifies maximum recursion depth used by the recursive inlining.
8212 For functions declared inline @option{--param max-inline-recursive-depth} is
8213 taken into account. For function not declared inline, recursive inlining
8214 happens only when @option{-finline-functions} (included in @option{-O3}) is
8215 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
8218 @item min-inline-recursive-probability
8219 Recursive inlining is profitable only for function having deep recursion
8220 in average and can hurt for function having little recursion depth by
8221 increasing the prologue size or complexity of function body to other
8224 When profile feedback is available (see @option{-fprofile-generate}) the actual
8225 recursion depth can be guessed from probability that function will recurse via
8226 given call expression. This parameter limits inlining only to call expression
8227 whose probability exceeds given threshold (in percents). The default value is
8230 @item early-inlining-insns
8231 Specify growth that early inliner can make. In effect it increases amount of
8232 inlining for code having large abstraction penalty. The default value is 8.
8234 @item max-early-inliner-iterations
8235 @itemx max-early-inliner-iterations
8236 Limit of iterations of early inliner. This basically bounds number of nested
8237 indirect calls early inliner can resolve. Deeper chains are still handled by
8240 @item min-vect-loop-bound
8241 The minimum number of iterations under which a loop will not get vectorized
8242 when @option{-ftree-vectorize} is used. The number of iterations after
8243 vectorization needs to be greater than the value specified by this option
8244 to allow vectorization. The default value is 0.
8246 @item gcse-cost-distance-ratio
8247 Scaling factor in calculation of maximum distance an expression
8248 can be moved by GCSE optimizations. This is currently supported only in
8249 code hoisting pass. The bigger the ratio, the more agressive code hoisting
8250 will be with simple expressions, i.e., the expressions which have cost
8251 less than @option{gcse-unrestricted-cost}. Specifying 0 will disable
8252 hoisting of simple expressions. The default value is 10.
8254 @item gcse-unrestricted-cost
8255 Cost, roughly measured as the cost of a single typical machine
8256 instruction, at which GCSE optimizations will not constrain
8257 the distance an expression can travel. This is currently
8258 supported only in code hoisting pass. The lesser the cost,
8259 the more aggressive code hoisting will be. Specifying 0 will
8260 allow all expressions to travel unrestricted distances.
8261 The default value is 3.
8263 @item max-unrolled-insns
8264 The maximum number of instructions that a loop should have if that loop
8265 is unrolled, and if the loop is unrolled, it determines how many times
8266 the loop code is unrolled.
8268 @item max-average-unrolled-insns
8269 The maximum number of instructions biased by probabilities of their execution
8270 that a loop should have if that loop is unrolled, and if the loop is unrolled,
8271 it determines how many times the loop code is unrolled.
8273 @item max-unroll-times
8274 The maximum number of unrollings of a single loop.
8276 @item max-peeled-insns
8277 The maximum number of instructions that a loop should have if that loop
8278 is peeled, and if the loop is peeled, it determines how many times
8279 the loop code is peeled.
8281 @item max-peel-times
8282 The maximum number of peelings of a single loop.
8284 @item max-completely-peeled-insns
8285 The maximum number of insns of a completely peeled loop.
8287 @item max-completely-peel-times
8288 The maximum number of iterations of a loop to be suitable for complete peeling.
8290 @item max-completely-peel-loop-nest-depth
8291 The maximum depth of a loop nest suitable for complete peeling.
8293 @item max-unswitch-insns
8294 The maximum number of insns of an unswitched loop.
8296 @item max-unswitch-level
8297 The maximum number of branches unswitched in a single loop.
8300 The minimum cost of an expensive expression in the loop invariant motion.
8302 @item iv-consider-all-candidates-bound
8303 Bound on number of candidates for induction variables below that
8304 all candidates are considered for each use in induction variable
8305 optimizations. Only the most relevant candidates are considered
8306 if there are more candidates, to avoid quadratic time complexity.
8308 @item iv-max-considered-uses
8309 The induction variable optimizations give up on loops that contain more
8310 induction variable uses.
8312 @item iv-always-prune-cand-set-bound
8313 If number of candidates in the set is smaller than this value,
8314 we always try to remove unnecessary ivs from the set during its
8315 optimization when a new iv is added to the set.
8317 @item scev-max-expr-size
8318 Bound on size of expressions used in the scalar evolutions analyzer.
8319 Large expressions slow the analyzer.
8321 @item omega-max-vars
8322 The maximum number of variables in an Omega constraint system.
8323 The default value is 128.
8325 @item omega-max-geqs
8326 The maximum number of inequalities in an Omega constraint system.
8327 The default value is 256.
8330 The maximum number of equalities in an Omega constraint system.
8331 The default value is 128.
8333 @item omega-max-wild-cards
8334 The maximum number of wildcard variables that the Omega solver will
8335 be able to insert. The default value is 18.
8337 @item omega-hash-table-size
8338 The size of the hash table in the Omega solver. The default value is
8341 @item omega-max-keys
8342 The maximal number of keys used by the Omega solver. The default
8345 @item omega-eliminate-redundant-constraints
8346 When set to 1, use expensive methods to eliminate all redundant
8347 constraints. The default value is 0.
8349 @item vect-max-version-for-alignment-checks
8350 The maximum number of runtime checks that can be performed when
8351 doing loop versioning for alignment in the vectorizer. See option
8352 ftree-vect-loop-version for more information.
8354 @item vect-max-version-for-alias-checks
8355 The maximum number of runtime checks that can be performed when
8356 doing loop versioning for alias in the vectorizer. See option
8357 ftree-vect-loop-version for more information.
8359 @item max-iterations-to-track
8361 The maximum number of iterations of a loop the brute force algorithm
8362 for analysis of # of iterations of the loop tries to evaluate.
8364 @item hot-bb-count-fraction
8365 Select fraction of the maximal count of repetitions of basic block in program
8366 given basic block needs to have to be considered hot.
8368 @item hot-bb-frequency-fraction
8369 Select fraction of the maximal frequency of executions of basic block in
8370 function given basic block needs to have to be considered hot
8372 @item max-predicted-iterations
8373 The maximum number of loop iterations we predict statically. This is useful
8374 in cases where function contain single loop with known bound and other loop
8375 with unknown. We predict the known number of iterations correctly, while
8376 the unknown number of iterations average to roughly 10. This means that the
8377 loop without bounds would appear artificially cold relative to the other one.
8379 @item align-threshold
8381 Select fraction of the maximal frequency of executions of basic block in
8382 function given basic block will get aligned.
8384 @item align-loop-iterations
8386 A loop expected to iterate at lest the selected number of iterations will get
8389 @item tracer-dynamic-coverage
8390 @itemx tracer-dynamic-coverage-feedback
8392 This value is used to limit superblock formation once the given percentage of
8393 executed instructions is covered. This limits unnecessary code size
8396 The @option{tracer-dynamic-coverage-feedback} is used only when profile
8397 feedback is available. The real profiles (as opposed to statically estimated
8398 ones) are much less balanced allowing the threshold to be larger value.
8400 @item tracer-max-code-growth
8401 Stop tail duplication once code growth has reached given percentage. This is
8402 rather hokey argument, as most of the duplicates will be eliminated later in
8403 cross jumping, so it may be set to much higher values than is the desired code
8406 @item tracer-min-branch-ratio
8408 Stop reverse growth when the reverse probability of best edge is less than this
8409 threshold (in percent).
8411 @item tracer-min-branch-ratio
8412 @itemx tracer-min-branch-ratio-feedback
8414 Stop forward growth if the best edge do have probability lower than this
8417 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
8418 compilation for profile feedback and one for compilation without. The value
8419 for compilation with profile feedback needs to be more conservative (higher) in
8420 order to make tracer effective.
8422 @item max-cse-path-length
8424 Maximum number of basic blocks on path that cse considers. The default is 10.
8427 The maximum instructions CSE process before flushing. The default is 1000.
8429 @item ggc-min-expand
8431 GCC uses a garbage collector to manage its own memory allocation. This
8432 parameter specifies the minimum percentage by which the garbage
8433 collector's heap should be allowed to expand between collections.
8434 Tuning this may improve compilation speed; it has no effect on code
8437 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
8438 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
8439 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
8440 GCC is not able to calculate RAM on a particular platform, the lower
8441 bound of 30% is used. Setting this parameter and
8442 @option{ggc-min-heapsize} to zero causes a full collection to occur at
8443 every opportunity. This is extremely slow, but can be useful for
8446 @item ggc-min-heapsize
8448 Minimum size of the garbage collector's heap before it begins bothering
8449 to collect garbage. The first collection occurs after the heap expands
8450 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
8451 tuning this may improve compilation speed, and has no effect on code
8454 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
8455 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
8456 with a lower bound of 4096 (four megabytes) and an upper bound of
8457 131072 (128 megabytes). If GCC is not able to calculate RAM on a
8458 particular platform, the lower bound is used. Setting this parameter
8459 very large effectively disables garbage collection. Setting this
8460 parameter and @option{ggc-min-expand} to zero causes a full collection
8461 to occur at every opportunity.
8463 @item max-reload-search-insns
8464 The maximum number of instruction reload should look backward for equivalent
8465 register. Increasing values mean more aggressive optimization, making the
8466 compile time increase with probably slightly better performance. The default
8469 @item max-cselib-memory-locations
8470 The maximum number of memory locations cselib should take into account.
8471 Increasing values mean more aggressive optimization, making the compile time
8472 increase with probably slightly better performance. The default value is 500.
8474 @item reorder-blocks-duplicate
8475 @itemx reorder-blocks-duplicate-feedback
8477 Used by basic block reordering pass to decide whether to use unconditional
8478 branch or duplicate the code on its destination. Code is duplicated when its
8479 estimated size is smaller than this value multiplied by the estimated size of
8480 unconditional jump in the hot spots of the program.
8482 The @option{reorder-block-duplicate-feedback} is used only when profile
8483 feedback is available and may be set to higher values than
8484 @option{reorder-block-duplicate} since information about the hot spots is more
8487 @item max-sched-ready-insns
8488 The maximum number of instructions ready to be issued the scheduler should
8489 consider at any given time during the first scheduling pass. Increasing
8490 values mean more thorough searches, making the compilation time increase
8491 with probably little benefit. The default value is 100.
8493 @item max-sched-region-blocks
8494 The maximum number of blocks in a region to be considered for
8495 interblock scheduling. The default value is 10.
8497 @item max-pipeline-region-blocks
8498 The maximum number of blocks in a region to be considered for
8499 pipelining in the selective scheduler. The default value is 15.
8501 @item max-sched-region-insns
8502 The maximum number of insns in a region to be considered for
8503 interblock scheduling. The default value is 100.
8505 @item max-pipeline-region-insns
8506 The maximum number of insns in a region to be considered for
8507 pipelining in the selective scheduler. The default value is 200.
8510 The minimum probability (in percents) of reaching a source block
8511 for interblock speculative scheduling. The default value is 40.
8513 @item max-sched-extend-regions-iters
8514 The maximum number of iterations through CFG to extend regions.
8515 0 - disable region extension,
8516 N - do at most N iterations.
8517 The default value is 0.
8519 @item max-sched-insn-conflict-delay
8520 The maximum conflict delay for an insn to be considered for speculative motion.
8521 The default value is 3.
8523 @item sched-spec-prob-cutoff
8524 The minimal probability of speculation success (in percents), so that
8525 speculative insn will be scheduled.
8526 The default value is 40.
8528 @item sched-mem-true-dep-cost
8529 Minimal distance (in CPU cycles) between store and load targeting same
8530 memory locations. The default value is 1.
8532 @item selsched-max-lookahead
8533 The maximum size of the lookahead window of selective scheduling. It is a
8534 depth of search for available instructions.
8535 The default value is 50.
8537 @item selsched-max-sched-times
8538 The maximum number of times that an instruction will be scheduled during
8539 selective scheduling. This is the limit on the number of iterations
8540 through which the instruction may be pipelined. The default value is 2.
8542 @item selsched-max-insns-to-rename
8543 The maximum number of best instructions in the ready list that are considered
8544 for renaming in the selective scheduler. The default value is 2.
8546 @item max-last-value-rtl
8547 The maximum size measured as number of RTLs that can be recorded in an expression
8548 in combiner for a pseudo register as last known value of that register. The default
8551 @item integer-share-limit
8552 Small integer constants can use a shared data structure, reducing the
8553 compiler's memory usage and increasing its speed. This sets the maximum
8554 value of a shared integer constant. The default value is 256.
8556 @item min-virtual-mappings
8557 Specifies the minimum number of virtual mappings in the incremental
8558 SSA updater that should be registered to trigger the virtual mappings
8559 heuristic defined by virtual-mappings-ratio. The default value is
8562 @item virtual-mappings-ratio
8563 If the number of virtual mappings is virtual-mappings-ratio bigger
8564 than the number of virtual symbols to be updated, then the incremental
8565 SSA updater switches to a full update for those symbols. The default
8568 @item ssp-buffer-size
8569 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
8570 protection when @option{-fstack-protection} is used.
8572 @item max-jump-thread-duplication-stmts
8573 Maximum number of statements allowed in a block that needs to be
8574 duplicated when threading jumps.
8576 @item max-fields-for-field-sensitive
8577 Maximum number of fields in a structure we will treat in
8578 a field sensitive manner during pointer analysis. The default is zero
8579 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
8581 @item prefetch-latency
8582 Estimate on average number of instructions that are executed before
8583 prefetch finishes. The distance we prefetch ahead is proportional
8584 to this constant. Increasing this number may also lead to less
8585 streams being prefetched (see @option{simultaneous-prefetches}).
8587 @item simultaneous-prefetches
8588 Maximum number of prefetches that can run at the same time.
8590 @item l1-cache-line-size
8591 The size of cache line in L1 cache, in bytes.
8594 The size of L1 cache, in kilobytes.
8597 The size of L2 cache, in kilobytes.
8599 @item min-insn-to-prefetch-ratio
8600 The minimum ratio between the number of instructions and the
8601 number of prefetches to enable prefetching in a loop.
8603 @item prefetch-min-insn-to-mem-ratio
8604 The minimum ratio between the number of instructions and the
8605 number of memory references to enable prefetching in a loop.
8607 @item use-canonical-types
8608 Whether the compiler should use the ``canonical'' type system. By
8609 default, this should always be 1, which uses a more efficient internal
8610 mechanism for comparing types in C++ and Objective-C++. However, if
8611 bugs in the canonical type system are causing compilation failures,
8612 set this value to 0 to disable canonical types.
8614 @item switch-conversion-max-branch-ratio
8615 Switch initialization conversion will refuse to create arrays that are
8616 bigger than @option{switch-conversion-max-branch-ratio} times the number of
8617 branches in the switch.
8619 @item max-partial-antic-length
8620 Maximum length of the partial antic set computed during the tree
8621 partial redundancy elimination optimization (@option{-ftree-pre}) when
8622 optimizing at @option{-O3} and above. For some sorts of source code
8623 the enhanced partial redundancy elimination optimization can run away,
8624 consuming all of the memory available on the host machine. This
8625 parameter sets a limit on the length of the sets that are computed,
8626 which prevents the runaway behavior. Setting a value of 0 for
8627 this parameter will allow an unlimited set length.
8629 @item sccvn-max-scc-size
8630 Maximum size of a strongly connected component (SCC) during SCCVN
8631 processing. If this limit is hit, SCCVN processing for the whole
8632 function will not be done and optimizations depending on it will
8633 be disabled. The default maximum SCC size is 10000.
8635 @item ira-max-loops-num
8636 IRA uses a regional register allocation by default. If a function
8637 contains loops more than number given by the parameter, only at most
8638 given number of the most frequently executed loops will form regions
8639 for the regional register allocation. The default value of the
8642 @item ira-max-conflict-table-size
8643 Although IRA uses a sophisticated algorithm of compression conflict
8644 table, the table can be still big for huge functions. If the conflict
8645 table for a function could be more than size in MB given by the
8646 parameter, the conflict table is not built and faster, simpler, and
8647 lower quality register allocation algorithm will be used. The
8648 algorithm do not use pseudo-register conflicts. The default value of
8649 the parameter is 2000.
8651 @item ira-loop-reserved-regs
8652 IRA can be used to evaluate more accurate register pressure in loops
8653 for decision to move loop invariants (see @option{-O3}). The number
8654 of available registers reserved for some other purposes is described
8655 by this parameter. The default value of the parameter is 2 which is
8656 minimal number of registers needed for execution of typical
8657 instruction. This value is the best found from numerous experiments.
8659 @item loop-invariant-max-bbs-in-loop
8660 Loop invariant motion can be very expensive, both in compile time and
8661 in amount of needed compile time memory, with very large loops. Loops
8662 with more basic blocks than this parameter won't have loop invariant
8663 motion optimization performed on them. The default value of the
8664 parameter is 1000 for -O1 and 10000 for -O2 and above.
8666 @item max-vartrack-size
8667 Sets a maximum number of hash table slots to use during variable
8668 tracking dataflow analysis of any function. If this limit is exceeded
8669 with variable tracking at assignments enabled, analysis for that
8670 function is retried without it, after removing all debug insns from
8671 the function. If the limit is exceeded even without debug insns, var
8672 tracking analysis is completely disabled for the function. Setting
8673 the parameter to zero makes it unlimited.
8675 @item min-nondebug-insn-uid
8676 Use uids starting at this parameter for nondebug insns. The range below
8677 the parameter is reserved exclusively for debug insns created by
8678 @option{-fvar-tracking-assignments}, but debug insns may get
8679 (non-overlapping) uids above it if the reserved range is exhausted.
8681 @item ipa-sra-ptr-growth-factor
8682 IPA-SRA will replace a pointer to an aggregate with one or more new
8683 parameters only when their cumulative size is less or equal to
8684 @option{ipa-sra-ptr-growth-factor} times the size of the original
8687 @item graphite-max-nb-scop-params
8688 To avoid exponential effects in the Graphite loop transforms, the
8689 number of parameters in a Static Control Part (SCoP) is bounded. The
8690 default value is 10 parameters. A variable whose value is unknown at
8691 compile time and defined outside a SCoP is a parameter of the SCoP.
8693 @item graphite-max-bbs-per-function
8694 To avoid exponential effects in the detection of SCoPs, the size of
8695 the functions analyzed by Graphite is bounded. The default value is
8698 @item loop-block-tile-size
8699 Loop blocking or strip mining transforms, enabled with
8700 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
8701 loop in the loop nest by a given number of iterations. The strip
8702 length can be changed using the @option{loop-block-tile-size}
8703 parameter. The default value is 51 iterations.
8708 @node Preprocessor Options
8709 @section Options Controlling the Preprocessor
8710 @cindex preprocessor options
8711 @cindex options, preprocessor
8713 These options control the C preprocessor, which is run on each C source
8714 file before actual compilation.
8716 If you use the @option{-E} option, nothing is done except preprocessing.
8717 Some of these options make sense only together with @option{-E} because
8718 they cause the preprocessor output to be unsuitable for actual
8722 @item -Wp,@var{option}
8724 You can use @option{-Wp,@var{option}} to bypass the compiler driver
8725 and pass @var{option} directly through to the preprocessor. If
8726 @var{option} contains commas, it is split into multiple options at the
8727 commas. However, many options are modified, translated or interpreted
8728 by the compiler driver before being passed to the preprocessor, and
8729 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
8730 interface is undocumented and subject to change, so whenever possible
8731 you should avoid using @option{-Wp} and let the driver handle the
8734 @item -Xpreprocessor @var{option}
8735 @opindex Xpreprocessor
8736 Pass @var{option} as an option to the preprocessor. You can use this to
8737 supply system-specific preprocessor options which GCC does not know how to
8740 If you want to pass an option that takes an argument, you must use
8741 @option{-Xpreprocessor} twice, once for the option and once for the argument.
8744 @include cppopts.texi
8746 @node Assembler Options
8747 @section Passing Options to the Assembler
8749 @c prevent bad page break with this line
8750 You can pass options to the assembler.
8753 @item -Wa,@var{option}
8755 Pass @var{option} as an option to the assembler. If @var{option}
8756 contains commas, it is split into multiple options at the commas.
8758 @item -Xassembler @var{option}
8760 Pass @var{option} as an option to the assembler. You can use this to
8761 supply system-specific assembler options which GCC does not know how to
8764 If you want to pass an option that takes an argument, you must use
8765 @option{-Xassembler} twice, once for the option and once for the argument.
8770 @section Options for Linking
8771 @cindex link options
8772 @cindex options, linking
8774 These options come into play when the compiler links object files into
8775 an executable output file. They are meaningless if the compiler is
8776 not doing a link step.
8780 @item @var{object-file-name}
8781 A file name that does not end in a special recognized suffix is
8782 considered to name an object file or library. (Object files are
8783 distinguished from libraries by the linker according to the file
8784 contents.) If linking is done, these object files are used as input
8793 If any of these options is used, then the linker is not run, and
8794 object file names should not be used as arguments. @xref{Overall
8798 @item -l@var{library}
8799 @itemx -l @var{library}
8801 Search the library named @var{library} when linking. (The second
8802 alternative with the library as a separate argument is only for
8803 POSIX compliance and is not recommended.)
8805 It makes a difference where in the command you write this option; the
8806 linker searches and processes libraries and object files in the order they
8807 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
8808 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
8809 to functions in @samp{z}, those functions may not be loaded.
8811 The linker searches a standard list of directories for the library,
8812 which is actually a file named @file{lib@var{library}.a}. The linker
8813 then uses this file as if it had been specified precisely by name.
8815 The directories searched include several standard system directories
8816 plus any that you specify with @option{-L}.
8818 Normally the files found this way are library files---archive files
8819 whose members are object files. The linker handles an archive file by
8820 scanning through it for members which define symbols that have so far
8821 been referenced but not defined. But if the file that is found is an
8822 ordinary object file, it is linked in the usual fashion. The only
8823 difference between using an @option{-l} option and specifying a file name
8824 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
8825 and searches several directories.
8829 You need this special case of the @option{-l} option in order to
8830 link an Objective-C or Objective-C++ program.
8833 @opindex nostartfiles
8834 Do not use the standard system startup files when linking.
8835 The standard system libraries are used normally, unless @option{-nostdlib}
8836 or @option{-nodefaultlibs} is used.
8838 @item -nodefaultlibs
8839 @opindex nodefaultlibs
8840 Do not use the standard system libraries when linking.
8841 Only the libraries you specify will be passed to the linker, options
8842 specifying linkage of the system libraries, such as @code{-static-libgcc}
8843 or @code{-shared-libgcc}, will be ignored.
8844 The standard startup files are used normally, unless @option{-nostartfiles}
8845 is used. The compiler may generate calls to @code{memcmp},
8846 @code{memset}, @code{memcpy} and @code{memmove}.
8847 These entries are usually resolved by entries in
8848 libc. These entry points should be supplied through some other
8849 mechanism when this option is specified.
8853 Do not use the standard system startup files or libraries when linking.
8854 No startup files and only the libraries you specify will be passed to
8855 the linker, options specifying linkage of the system libraries, such as
8856 @code{-static-libgcc} or @code{-shared-libgcc}, will be ignored.
8857 The compiler may generate calls to @code{memcmp}, @code{memset},
8858 @code{memcpy} and @code{memmove}.
8859 These entries are usually resolved by entries in
8860 libc. These entry points should be supplied through some other
8861 mechanism when this option is specified.
8863 @cindex @option{-lgcc}, use with @option{-nostdlib}
8864 @cindex @option{-nostdlib} and unresolved references
8865 @cindex unresolved references and @option{-nostdlib}
8866 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
8867 @cindex @option{-nodefaultlibs} and unresolved references
8868 @cindex unresolved references and @option{-nodefaultlibs}
8869 One of the standard libraries bypassed by @option{-nostdlib} and
8870 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
8871 that GCC uses to overcome shortcomings of particular machines, or special
8872 needs for some languages.
8873 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
8874 Collection (GCC) Internals},
8875 for more discussion of @file{libgcc.a}.)
8876 In most cases, you need @file{libgcc.a} even when you want to avoid
8877 other standard libraries. In other words, when you specify @option{-nostdlib}
8878 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
8879 This ensures that you have no unresolved references to internal GCC
8880 library subroutines. (For example, @samp{__main}, used to ensure C++
8881 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
8882 GNU Compiler Collection (GCC) Internals}.)
8886 Produce a position independent executable on targets which support it.
8887 For predictable results, you must also specify the same set of options
8888 that were used to generate code (@option{-fpie}, @option{-fPIE},
8889 or model suboptions) when you specify this option.
8893 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
8894 that support it. This instructs the linker to add all symbols, not
8895 only used ones, to the dynamic symbol table. This option is needed
8896 for some uses of @code{dlopen} or to allow obtaining backtraces
8897 from within a program.
8901 Remove all symbol table and relocation information from the executable.
8905 On systems that support dynamic linking, this prevents linking with the shared
8906 libraries. On other systems, this option has no effect.
8910 Produce a shared object which can then be linked with other objects to
8911 form an executable. Not all systems support this option. For predictable
8912 results, you must also specify the same set of options that were used to
8913 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
8914 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
8915 needs to build supplementary stub code for constructors to work. On
8916 multi-libbed systems, @samp{gcc -shared} must select the correct support
8917 libraries to link against. Failing to supply the correct flags may lead
8918 to subtle defects. Supplying them in cases where they are not necessary
8921 @item -shared-libgcc
8922 @itemx -static-libgcc
8923 @opindex shared-libgcc
8924 @opindex static-libgcc
8925 On systems that provide @file{libgcc} as a shared library, these options
8926 force the use of either the shared or static version respectively.
8927 If no shared version of @file{libgcc} was built when the compiler was
8928 configured, these options have no effect.
8930 There are several situations in which an application should use the
8931 shared @file{libgcc} instead of the static version. The most common
8932 of these is when the application wishes to throw and catch exceptions
8933 across different shared libraries. In that case, each of the libraries
8934 as well as the application itself should use the shared @file{libgcc}.
8936 Therefore, the G++ and GCJ drivers automatically add
8937 @option{-shared-libgcc} whenever you build a shared library or a main
8938 executable, because C++ and Java programs typically use exceptions, so
8939 this is the right thing to do.
8941 If, instead, you use the GCC driver to create shared libraries, you may
8942 find that they will not always be linked with the shared @file{libgcc}.
8943 If GCC finds, at its configuration time, that you have a non-GNU linker
8944 or a GNU linker that does not support option @option{--eh-frame-hdr},
8945 it will link the shared version of @file{libgcc} into shared libraries
8946 by default. Otherwise, it will take advantage of the linker and optimize
8947 away the linking with the shared version of @file{libgcc}, linking with
8948 the static version of libgcc by default. This allows exceptions to
8949 propagate through such shared libraries, without incurring relocation
8950 costs at library load time.
8952 However, if a library or main executable is supposed to throw or catch
8953 exceptions, you must link it using the G++ or GCJ driver, as appropriate
8954 for the languages used in the program, or using the option
8955 @option{-shared-libgcc}, such that it is linked with the shared
8958 @item -static-libstdc++
8959 When the @command{g++} program is used to link a C++ program, it will
8960 normally automatically link against @option{libstdc++}. If
8961 @file{libstdc++} is available as a shared library, and the
8962 @option{-static} option is not used, then this will link against the
8963 shared version of @file{libstdc++}. That is normally fine. However, it
8964 is sometimes useful to freeze the version of @file{libstdc++} used by
8965 the program without going all the way to a fully static link. The
8966 @option{-static-libstdc++} option directs the @command{g++} driver to
8967 link @file{libstdc++} statically, without necessarily linking other
8968 libraries statically.
8972 Bind references to global symbols when building a shared object. Warn
8973 about any unresolved references (unless overridden by the link editor
8974 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
8977 @item -T @var{script}
8979 @cindex linker script
8980 Use @var{script} as the linker script. This option is supported by most
8981 systems using the GNU linker. On some targets, such as bare-board
8982 targets without an operating system, the @option{-T} option may be required
8983 when linking to avoid references to undefined symbols.
8985 @item -Xlinker @var{option}
8987 Pass @var{option} as an option to the linker. You can use this to
8988 supply system-specific linker options which GCC does not know how to
8991 If you want to pass an option that takes a separate argument, you must use
8992 @option{-Xlinker} twice, once for the option and once for the argument.
8993 For example, to pass @option{-assert definitions}, you must write
8994 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
8995 @option{-Xlinker "-assert definitions"}, because this passes the entire
8996 string as a single argument, which is not what the linker expects.
8998 When using the GNU linker, it is usually more convenient to pass
8999 arguments to linker options using the @option{@var{option}=@var{value}}
9000 syntax than as separate arguments. For example, you can specify
9001 @samp{-Xlinker -Map=output.map} rather than
9002 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
9003 this syntax for command-line options.
9005 @item -Wl,@var{option}
9007 Pass @var{option} as an option to the linker. If @var{option} contains
9008 commas, it is split into multiple options at the commas. You can use this
9009 syntax to pass an argument to the option.
9010 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
9011 linker. When using the GNU linker, you can also get the same effect with
9012 @samp{-Wl,-Map=output.map}.
9014 @item -u @var{symbol}
9016 Pretend the symbol @var{symbol} is undefined, to force linking of
9017 library modules to define it. You can use @option{-u} multiple times with
9018 different symbols to force loading of additional library modules.
9021 @node Directory Options
9022 @section Options for Directory Search
9023 @cindex directory options
9024 @cindex options, directory search
9027 These options specify directories to search for header files, for
9028 libraries and for parts of the compiler:
9033 Add the directory @var{dir} to the head of the list of directories to be
9034 searched for header files. This can be used to override a system header
9035 file, substituting your own version, since these directories are
9036 searched before the system header file directories. However, you should
9037 not use this option to add directories that contain vendor-supplied
9038 system header files (use @option{-isystem} for that). If you use more than
9039 one @option{-I} option, the directories are scanned in left-to-right
9040 order; the standard system directories come after.
9042 If a standard system include directory, or a directory specified with
9043 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
9044 option will be ignored. The directory will still be searched but as a
9045 system directory at its normal position in the system include chain.
9046 This is to ensure that GCC's procedure to fix buggy system headers and
9047 the ordering for the include_next directive are not inadvertently changed.
9048 If you really need to change the search order for system directories,
9049 use the @option{-nostdinc} and/or @option{-isystem} options.
9051 @item -iplugindir=@var{dir}
9052 Set the directory to search for plugins which are passed
9053 by @option{-fplugin=@var{name}} instead of
9054 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
9055 to be used by the user, but only passed by the driver.
9057 @item -iquote@var{dir}
9059 Add the directory @var{dir} to the head of the list of directories to
9060 be searched for header files only for the case of @samp{#include
9061 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
9062 otherwise just like @option{-I}.
9066 Add directory @var{dir} to the list of directories to be searched
9069 @item -B@var{prefix}
9071 This option specifies where to find the executables, libraries,
9072 include files, and data files of the compiler itself.
9074 The compiler driver program runs one or more of the subprograms
9075 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
9076 @var{prefix} as a prefix for each program it tries to run, both with and
9077 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
9079 For each subprogram to be run, the compiler driver first tries the
9080 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
9081 was not specified, the driver tries two standard prefixes, which are
9082 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
9083 those results in a file name that is found, the unmodified program
9084 name is searched for using the directories specified in your
9085 @env{PATH} environment variable.
9087 The compiler will check to see if the path provided by the @option{-B}
9088 refers to a directory, and if necessary it will add a directory
9089 separator character at the end of the path.
9091 @option{-B} prefixes that effectively specify directory names also apply
9092 to libraries in the linker, because the compiler translates these
9093 options into @option{-L} options for the linker. They also apply to
9094 includes files in the preprocessor, because the compiler translates these
9095 options into @option{-isystem} options for the preprocessor. In this case,
9096 the compiler appends @samp{include} to the prefix.
9098 The run-time support file @file{libgcc.a} can also be searched for using
9099 the @option{-B} prefix, if needed. If it is not found there, the two
9100 standard prefixes above are tried, and that is all. The file is left
9101 out of the link if it is not found by those means.
9103 Another way to specify a prefix much like the @option{-B} prefix is to use
9104 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
9107 As a special kludge, if the path provided by @option{-B} is
9108 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
9109 9, then it will be replaced by @file{[dir/]include}. This is to help
9110 with boot-strapping the compiler.
9112 @item -specs=@var{file}
9114 Process @var{file} after the compiler reads in the standard @file{specs}
9115 file, in order to override the defaults that the @file{gcc} driver
9116 program uses when determining what switches to pass to @file{cc1},
9117 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
9118 @option{-specs=@var{file}} can be specified on the command line, and they
9119 are processed in order, from left to right.
9121 @item --sysroot=@var{dir}
9123 Use @var{dir} as the logical root directory for headers and libraries.
9124 For example, if the compiler would normally search for headers in
9125 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
9126 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
9128 If you use both this option and the @option{-isysroot} option, then
9129 the @option{--sysroot} option will apply to libraries, but the
9130 @option{-isysroot} option will apply to header files.
9132 The GNU linker (beginning with version 2.16) has the necessary support
9133 for this option. If your linker does not support this option, the
9134 header file aspect of @option{--sysroot} will still work, but the
9135 library aspect will not.
9139 This option has been deprecated. Please use @option{-iquote} instead for
9140 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
9141 Any directories you specify with @option{-I} options before the @option{-I-}
9142 option are searched only for the case of @samp{#include "@var{file}"};
9143 they are not searched for @samp{#include <@var{file}>}.
9145 If additional directories are specified with @option{-I} options after
9146 the @option{-I-}, these directories are searched for all @samp{#include}
9147 directives. (Ordinarily @emph{all} @option{-I} directories are used
9150 In addition, the @option{-I-} option inhibits the use of the current
9151 directory (where the current input file came from) as the first search
9152 directory for @samp{#include "@var{file}"}. There is no way to
9153 override this effect of @option{-I-}. With @option{-I.} you can specify
9154 searching the directory which was current when the compiler was
9155 invoked. That is not exactly the same as what the preprocessor does
9156 by default, but it is often satisfactory.
9158 @option{-I-} does not inhibit the use of the standard system directories
9159 for header files. Thus, @option{-I-} and @option{-nostdinc} are
9166 @section Specifying subprocesses and the switches to pass to them
9169 @command{gcc} is a driver program. It performs its job by invoking a
9170 sequence of other programs to do the work of compiling, assembling and
9171 linking. GCC interprets its command-line parameters and uses these to
9172 deduce which programs it should invoke, and which command-line options
9173 it ought to place on their command lines. This behavior is controlled
9174 by @dfn{spec strings}. In most cases there is one spec string for each
9175 program that GCC can invoke, but a few programs have multiple spec
9176 strings to control their behavior. The spec strings built into GCC can
9177 be overridden by using the @option{-specs=} command-line switch to specify
9180 @dfn{Spec files} are plaintext files that are used to construct spec
9181 strings. They consist of a sequence of directives separated by blank
9182 lines. The type of directive is determined by the first non-whitespace
9183 character on the line and it can be one of the following:
9186 @item %@var{command}
9187 Issues a @var{command} to the spec file processor. The commands that can
9191 @item %include <@var{file}>
9193 Search for @var{file} and insert its text at the current point in the
9196 @item %include_noerr <@var{file}>
9197 @cindex %include_noerr
9198 Just like @samp{%include}, but do not generate an error message if the include
9199 file cannot be found.
9201 @item %rename @var{old_name} @var{new_name}
9203 Rename the spec string @var{old_name} to @var{new_name}.
9207 @item *[@var{spec_name}]:
9208 This tells the compiler to create, override or delete the named spec
9209 string. All lines after this directive up to the next directive or
9210 blank line are considered to be the text for the spec string. If this
9211 results in an empty string then the spec will be deleted. (Or, if the
9212 spec did not exist, then nothing will happened.) Otherwise, if the spec
9213 does not currently exist a new spec will be created. If the spec does
9214 exist then its contents will be overridden by the text of this
9215 directive, unless the first character of that text is the @samp{+}
9216 character, in which case the text will be appended to the spec.
9218 @item [@var{suffix}]:
9219 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
9220 and up to the next directive or blank line are considered to make up the
9221 spec string for the indicated suffix. When the compiler encounters an
9222 input file with the named suffix, it will processes the spec string in
9223 order to work out how to compile that file. For example:
9230 This says that any input file whose name ends in @samp{.ZZ} should be
9231 passed to the program @samp{z-compile}, which should be invoked with the
9232 command-line switch @option{-input} and with the result of performing the
9233 @samp{%i} substitution. (See below.)
9235 As an alternative to providing a spec string, the text that follows a
9236 suffix directive can be one of the following:
9239 @item @@@var{language}
9240 This says that the suffix is an alias for a known @var{language}. This is
9241 similar to using the @option{-x} command-line switch to GCC to specify a
9242 language explicitly. For example:
9249 Says that .ZZ files are, in fact, C++ source files.
9252 This causes an error messages saying:
9255 @var{name} compiler not installed on this system.
9259 GCC already has an extensive list of suffixes built into it.
9260 This directive will add an entry to the end of the list of suffixes, but
9261 since the list is searched from the end backwards, it is effectively
9262 possible to override earlier entries using this technique.
9266 GCC has the following spec strings built into it. Spec files can
9267 override these strings or create their own. Note that individual
9268 targets can also add their own spec strings to this list.
9271 asm Options to pass to the assembler
9272 asm_final Options to pass to the assembler post-processor
9273 cpp Options to pass to the C preprocessor
9274 cc1 Options to pass to the C compiler
9275 cc1plus Options to pass to the C++ compiler
9276 endfile Object files to include at the end of the link
9277 link Options to pass to the linker
9278 lib Libraries to include on the command line to the linker
9279 libgcc Decides which GCC support library to pass to the linker
9280 linker Sets the name of the linker
9281 predefines Defines to be passed to the C preprocessor
9282 signed_char Defines to pass to CPP to say whether @code{char} is signed
9284 startfile Object files to include at the start of the link
9287 Here is a small example of a spec file:
9293 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
9296 This example renames the spec called @samp{lib} to @samp{old_lib} and
9297 then overrides the previous definition of @samp{lib} with a new one.
9298 The new definition adds in some extra command-line options before
9299 including the text of the old definition.
9301 @dfn{Spec strings} are a list of command-line options to be passed to their
9302 corresponding program. In addition, the spec strings can contain
9303 @samp{%}-prefixed sequences to substitute variable text or to
9304 conditionally insert text into the command line. Using these constructs
9305 it is possible to generate quite complex command lines.
9307 Here is a table of all defined @samp{%}-sequences for spec
9308 strings. Note that spaces are not generated automatically around the
9309 results of expanding these sequences. Therefore you can concatenate them
9310 together or combine them with constant text in a single argument.
9314 Substitute one @samp{%} into the program name or argument.
9317 Substitute the name of the input file being processed.
9320 Substitute the basename of the input file being processed.
9321 This is the substring up to (and not including) the last period
9322 and not including the directory.
9325 This is the same as @samp{%b}, but include the file suffix (text after
9329 Marks the argument containing or following the @samp{%d} as a
9330 temporary file name, so that that file will be deleted if GCC exits
9331 successfully. Unlike @samp{%g}, this contributes no text to the
9334 @item %g@var{suffix}
9335 Substitute a file name that has suffix @var{suffix} and is chosen
9336 once per compilation, and mark the argument in the same way as
9337 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
9338 name is now chosen in a way that is hard to predict even when previously
9339 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
9340 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
9341 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
9342 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
9343 was simply substituted with a file name chosen once per compilation,
9344 without regard to any appended suffix (which was therefore treated
9345 just like ordinary text), making such attacks more likely to succeed.
9347 @item %u@var{suffix}
9348 Like @samp{%g}, but generates a new temporary file name even if
9349 @samp{%u@var{suffix}} was already seen.
9351 @item %U@var{suffix}
9352 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
9353 new one if there is no such last file name. In the absence of any
9354 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
9355 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
9356 would involve the generation of two distinct file names, one
9357 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
9358 simply substituted with a file name chosen for the previous @samp{%u},
9359 without regard to any appended suffix.
9361 @item %j@var{suffix}
9362 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
9363 writable, and if save-temps is off; otherwise, substitute the name
9364 of a temporary file, just like @samp{%u}. This temporary file is not
9365 meant for communication between processes, but rather as a junk
9368 @item %|@var{suffix}
9369 @itemx %m@var{suffix}
9370 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
9371 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
9372 all. These are the two most common ways to instruct a program that it
9373 should read from standard input or write to standard output. If you
9374 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
9375 construct: see for example @file{f/lang-specs.h}.
9377 @item %.@var{SUFFIX}
9378 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
9379 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
9380 terminated by the next space or %.
9383 Marks the argument containing or following the @samp{%w} as the
9384 designated output file of this compilation. This puts the argument
9385 into the sequence of arguments that @samp{%o} will substitute later.
9388 Substitutes the names of all the output files, with spaces
9389 automatically placed around them. You should write spaces
9390 around the @samp{%o} as well or the results are undefined.
9391 @samp{%o} is for use in the specs for running the linker.
9392 Input files whose names have no recognized suffix are not compiled
9393 at all, but they are included among the output files, so they will
9397 Substitutes the suffix for object files. Note that this is
9398 handled specially when it immediately follows @samp{%g, %u, or %U},
9399 because of the need for those to form complete file names. The
9400 handling is such that @samp{%O} is treated exactly as if it had already
9401 been substituted, except that @samp{%g, %u, and %U} do not currently
9402 support additional @var{suffix} characters following @samp{%O} as they would
9403 following, for example, @samp{.o}.
9406 Substitutes the standard macro predefinitions for the
9407 current target machine. Use this when running @code{cpp}.
9410 Like @samp{%p}, but puts @samp{__} before and after the name of each
9411 predefined macro, except for macros that start with @samp{__} or with
9412 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
9416 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
9417 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
9418 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
9419 and @option{-imultilib} as necessary.
9422 Current argument is the name of a library or startup file of some sort.
9423 Search for that file in a standard list of directories and substitute
9424 the full name found. The current working directory is included in the
9425 list of directories scanned.
9428 Current argument is the name of a linker script. Search for that file
9429 in the current list of directories to scan for libraries. If the file
9430 is located insert a @option{--script} option into the command line
9431 followed by the full path name found. If the file is not found then
9432 generate an error message. Note: the current working directory is not
9436 Print @var{str} as an error message. @var{str} is terminated by a newline.
9437 Use this when inconsistent options are detected.
9440 Substitute the contents of spec string @var{name} at this point.
9443 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
9445 @item %x@{@var{option}@}
9446 Accumulate an option for @samp{%X}.
9449 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
9453 Output the accumulated assembler options specified by @option{-Wa}.
9456 Output the accumulated preprocessor options specified by @option{-Wp}.
9459 Process the @code{asm} spec. This is used to compute the
9460 switches to be passed to the assembler.
9463 Process the @code{asm_final} spec. This is a spec string for
9464 passing switches to an assembler post-processor, if such a program is
9468 Process the @code{link} spec. This is the spec for computing the
9469 command line passed to the linker. Typically it will make use of the
9470 @samp{%L %G %S %D and %E} sequences.
9473 Dump out a @option{-L} option for each directory that GCC believes might
9474 contain startup files. If the target supports multilibs then the
9475 current multilib directory will be prepended to each of these paths.
9478 Process the @code{lib} spec. This is a spec string for deciding which
9479 libraries should be included on the command line to the linker.
9482 Process the @code{libgcc} spec. This is a spec string for deciding
9483 which GCC support library should be included on the command line to the linker.
9486 Process the @code{startfile} spec. This is a spec for deciding which
9487 object files should be the first ones passed to the linker. Typically
9488 this might be a file named @file{crt0.o}.
9491 Process the @code{endfile} spec. This is a spec string that specifies
9492 the last object files that will be passed to the linker.
9495 Process the @code{cpp} spec. This is used to construct the arguments
9496 to be passed to the C preprocessor.
9499 Process the @code{cc1} spec. This is used to construct the options to be
9500 passed to the actual C compiler (@samp{cc1}).
9503 Process the @code{cc1plus} spec. This is used to construct the options to be
9504 passed to the actual C++ compiler (@samp{cc1plus}).
9507 Substitute the variable part of a matched option. See below.
9508 Note that each comma in the substituted string is replaced by
9512 Remove all occurrences of @code{-S} from the command line. Note---this
9513 command is position dependent. @samp{%} commands in the spec string
9514 before this one will see @code{-S}, @samp{%} commands in the spec string
9515 after this one will not.
9517 @item %:@var{function}(@var{args})
9518 Call the named function @var{function}, passing it @var{args}.
9519 @var{args} is first processed as a nested spec string, then split
9520 into an argument vector in the usual fashion. The function returns
9521 a string which is processed as if it had appeared literally as part
9522 of the current spec.
9524 The following built-in spec functions are provided:
9528 The @code{getenv} spec function takes two arguments: an environment
9529 variable name and a string. If the environment variable is not
9530 defined, a fatal error is issued. Otherwise, the return value is the
9531 value of the environment variable concatenated with the string. For
9532 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
9535 %:getenv(TOPDIR /include)
9538 expands to @file{/path/to/top/include}.
9540 @item @code{if-exists}
9541 The @code{if-exists} spec function takes one argument, an absolute
9542 pathname to a file. If the file exists, @code{if-exists} returns the
9543 pathname. Here is a small example of its usage:
9547 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
9550 @item @code{if-exists-else}
9551 The @code{if-exists-else} spec function is similar to the @code{if-exists}
9552 spec function, except that it takes two arguments. The first argument is
9553 an absolute pathname to a file. If the file exists, @code{if-exists-else}
9554 returns the pathname. If it does not exist, it returns the second argument.
9555 This way, @code{if-exists-else} can be used to select one file or another,
9556 based on the existence of the first. Here is a small example of its usage:
9560 crt0%O%s %:if-exists(crti%O%s) \
9561 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
9564 @item @code{replace-outfile}
9565 The @code{replace-outfile} spec function takes two arguments. It looks for the
9566 first argument in the outfiles array and replaces it with the second argument. Here
9567 is a small example of its usage:
9570 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
9573 @item @code{print-asm-header}
9574 The @code{print-asm-header} function takes no arguments and simply
9575 prints a banner like:
9581 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
9584 It is used to separate compiler options from assembler options
9585 in the @option{--target-help} output.
9589 Substitutes the @code{-S} switch, if that switch was given to GCC@.
9590 If that switch was not specified, this substitutes nothing. Note that
9591 the leading dash is omitted when specifying this option, and it is
9592 automatically inserted if the substitution is performed. Thus the spec
9593 string @samp{%@{foo@}} would match the command-line option @option{-foo}
9594 and would output the command line option @option{-foo}.
9596 @item %W@{@code{S}@}
9597 Like %@{@code{S}@} but mark last argument supplied within as a file to be
9600 @item %@{@code{S}*@}
9601 Substitutes all the switches specified to GCC whose names start
9602 with @code{-S}, but which also take an argument. This is used for
9603 switches like @option{-o}, @option{-D}, @option{-I}, etc.
9604 GCC considers @option{-o foo} as being
9605 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
9606 text, including the space. Thus two arguments would be generated.
9608 @item %@{@code{S}*&@code{T}*@}
9609 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
9610 (the order of @code{S} and @code{T} in the spec is not significant).
9611 There can be any number of ampersand-separated variables; for each the
9612 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
9614 @item %@{@code{S}:@code{X}@}
9615 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
9617 @item %@{!@code{S}:@code{X}@}
9618 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
9620 @item %@{@code{S}*:@code{X}@}
9621 Substitutes @code{X} if one or more switches whose names start with
9622 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
9623 once, no matter how many such switches appeared. However, if @code{%*}
9624 appears somewhere in @code{X}, then @code{X} will be substituted once
9625 for each matching switch, with the @code{%*} replaced by the part of
9626 that switch that matched the @code{*}.
9628 @item %@{.@code{S}:@code{X}@}
9629 Substitutes @code{X}, if processing a file with suffix @code{S}.
9631 @item %@{!.@code{S}:@code{X}@}
9632 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
9634 @item %@{,@code{S}:@code{X}@}
9635 Substitutes @code{X}, if processing a file for language @code{S}.
9637 @item %@{!,@code{S}:@code{X}@}
9638 Substitutes @code{X}, if not processing a file for language @code{S}.
9640 @item %@{@code{S}|@code{P}:@code{X}@}
9641 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
9642 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
9643 @code{*} sequences as well, although they have a stronger binding than
9644 the @samp{|}. If @code{%*} appears in @code{X}, all of the
9645 alternatives must be starred, and only the first matching alternative
9648 For example, a spec string like this:
9651 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
9654 will output the following command-line options from the following input
9655 command-line options:
9660 -d fred.c -foo -baz -boggle
9661 -d jim.d -bar -baz -boggle
9664 @item %@{S:X; T:Y; :D@}
9666 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
9667 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
9668 be as many clauses as you need. This may be combined with @code{.},
9669 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
9674 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
9675 construct may contain other nested @samp{%} constructs or spaces, or
9676 even newlines. They are processed as usual, as described above.
9677 Trailing white space in @code{X} is ignored. White space may also
9678 appear anywhere on the left side of the colon in these constructs,
9679 except between @code{.} or @code{*} and the corresponding word.
9681 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
9682 handled specifically in these constructs. If another value of
9683 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
9684 @option{-W} switch is found later in the command line, the earlier
9685 switch value is ignored, except with @{@code{S}*@} where @code{S} is
9686 just one letter, which passes all matching options.
9688 The character @samp{|} at the beginning of the predicate text is used to
9689 indicate that a command should be piped to the following command, but
9690 only if @option{-pipe} is specified.
9692 It is built into GCC which switches take arguments and which do not.
9693 (You might think it would be useful to generalize this to allow each
9694 compiler's spec to say which switches take arguments. But this cannot
9695 be done in a consistent fashion. GCC cannot even decide which input
9696 files have been specified without knowing which switches take arguments,
9697 and it must know which input files to compile in order to tell which
9700 GCC also knows implicitly that arguments starting in @option{-l} are to be
9701 treated as compiler output files, and passed to the linker in their
9702 proper position among the other output files.
9704 @c man begin OPTIONS
9706 @node Target Options
9707 @section Specifying Target Machine and Compiler Version
9708 @cindex target options
9709 @cindex cross compiling
9710 @cindex specifying machine version
9711 @cindex specifying compiler version and target machine
9712 @cindex compiler version, specifying
9713 @cindex target machine, specifying
9715 The usual way to run GCC is to run the executable called @file{gcc}, or
9716 @file{<machine>-gcc} when cross-compiling, or
9717 @file{<machine>-gcc-<version>} to run a version other than the one that
9720 @node Submodel Options
9721 @section Hardware Models and Configurations
9722 @cindex submodel options
9723 @cindex specifying hardware config
9724 @cindex hardware models and configurations, specifying
9725 @cindex machine dependent options
9727 Each target machine types can have its own
9728 special options, starting with @samp{-m}, to choose among various
9729 hardware models or configurations---for example, 68010 vs 68020,
9730 floating coprocessor or none. A single installed version of the
9731 compiler can compile for any model or configuration, according to the
9734 Some configurations of the compiler also support additional special
9735 options, usually for compatibility with other compilers on the same
9738 @c This list is ordered alphanumerically by subsection name.
9739 @c It should be the same order and spelling as these options are listed
9740 @c in Machine Dependent Options
9746 * Blackfin Options::
9750 * DEC Alpha Options::
9751 * DEC Alpha/VMS Options::
9754 * GNU/Linux Options::
9757 * i386 and x86-64 Options::
9758 * i386 and x86-64 Windows Options::
9760 * IA-64/VMS Options::
9772 * picoChip Options::
9774 * RS/6000 and PowerPC Options::
9776 * S/390 and zSeries Options::
9781 * System V Options::
9786 * Xstormy16 Options::
9792 @subsection ARC Options
9795 These options are defined for ARC implementations:
9800 Compile code for little endian mode. This is the default.
9804 Compile code for big endian mode.
9807 @opindex mmangle-cpu
9808 Prepend the name of the cpu to all public symbol names.
9809 In multiple-processor systems, there are many ARC variants with different
9810 instruction and register set characteristics. This flag prevents code
9811 compiled for one cpu to be linked with code compiled for another.
9812 No facility exists for handling variants that are ``almost identical''.
9813 This is an all or nothing option.
9815 @item -mcpu=@var{cpu}
9817 Compile code for ARC variant @var{cpu}.
9818 Which variants are supported depend on the configuration.
9819 All variants support @option{-mcpu=base}, this is the default.
9821 @item -mtext=@var{text-section}
9822 @itemx -mdata=@var{data-section}
9823 @itemx -mrodata=@var{readonly-data-section}
9827 Put functions, data, and readonly data in @var{text-section},
9828 @var{data-section}, and @var{readonly-data-section} respectively
9829 by default. This can be overridden with the @code{section} attribute.
9830 @xref{Variable Attributes}.
9835 @subsection ARM Options
9838 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
9842 @item -mabi=@var{name}
9844 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
9845 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
9848 @opindex mapcs-frame
9849 Generate a stack frame that is compliant with the ARM Procedure Call
9850 Standard for all functions, even if this is not strictly necessary for
9851 correct execution of the code. Specifying @option{-fomit-frame-pointer}
9852 with this option will cause the stack frames not to be generated for
9853 leaf functions. The default is @option{-mno-apcs-frame}.
9857 This is a synonym for @option{-mapcs-frame}.
9860 @c not currently implemented
9861 @item -mapcs-stack-check
9862 @opindex mapcs-stack-check
9863 Generate code to check the amount of stack space available upon entry to
9864 every function (that actually uses some stack space). If there is
9865 insufficient space available then either the function
9866 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
9867 called, depending upon the amount of stack space required. The run time
9868 system is required to provide these functions. The default is
9869 @option{-mno-apcs-stack-check}, since this produces smaller code.
9871 @c not currently implemented
9873 @opindex mapcs-float
9874 Pass floating point arguments using the float point registers. This is
9875 one of the variants of the APCS@. This option is recommended if the
9876 target hardware has a floating point unit or if a lot of floating point
9877 arithmetic is going to be performed by the code. The default is
9878 @option{-mno-apcs-float}, since integer only code is slightly increased in
9879 size if @option{-mapcs-float} is used.
9881 @c not currently implemented
9882 @item -mapcs-reentrant
9883 @opindex mapcs-reentrant
9884 Generate reentrant, position independent code. The default is
9885 @option{-mno-apcs-reentrant}.
9888 @item -mthumb-interwork
9889 @opindex mthumb-interwork
9890 Generate code which supports calling between the ARM and Thumb
9891 instruction sets. Without this option the two instruction sets cannot
9892 be reliably used inside one program. The default is
9893 @option{-mno-thumb-interwork}, since slightly larger code is generated
9894 when @option{-mthumb-interwork} is specified.
9896 @item -mno-sched-prolog
9897 @opindex mno-sched-prolog
9898 Prevent the reordering of instructions in the function prolog, or the
9899 merging of those instruction with the instructions in the function's
9900 body. This means that all functions will start with a recognizable set
9901 of instructions (or in fact one of a choice from a small set of
9902 different function prologues), and this information can be used to
9903 locate the start if functions inside an executable piece of code. The
9904 default is @option{-msched-prolog}.
9906 @item -mfloat-abi=@var{name}
9908 Specifies which floating-point ABI to use. Permissible values
9909 are: @samp{soft}, @samp{softfp} and @samp{hard}.
9911 Specifying @samp{soft} causes GCC to generate output containing
9912 library calls for floating-point operations.
9913 @samp{softfp} allows the generation of code using hardware floating-point
9914 instructions, but still uses the soft-float calling conventions.
9915 @samp{hard} allows generation of floating-point instructions
9916 and uses FPU-specific calling conventions.
9918 The default depends on the specific target configuration. Note that
9919 the hard-float and soft-float ABIs are not link-compatible; you must
9920 compile your entire program with the same ABI, and link with a
9921 compatible set of libraries.
9924 @opindex mhard-float
9925 Equivalent to @option{-mfloat-abi=hard}.
9928 @opindex msoft-float
9929 Equivalent to @option{-mfloat-abi=soft}.
9931 @item -mlittle-endian
9932 @opindex mlittle-endian
9933 Generate code for a processor running in little-endian mode. This is
9934 the default for all standard configurations.
9937 @opindex mbig-endian
9938 Generate code for a processor running in big-endian mode; the default is
9939 to compile code for a little-endian processor.
9941 @item -mwords-little-endian
9942 @opindex mwords-little-endian
9943 This option only applies when generating code for big-endian processors.
9944 Generate code for a little-endian word order but a big-endian byte
9945 order. That is, a byte order of the form @samp{32107654}. Note: this
9946 option should only be used if you require compatibility with code for
9947 big-endian ARM processors generated by versions of the compiler prior to
9950 @item -mcpu=@var{name}
9952 This specifies the name of the target ARM processor. GCC uses this name
9953 to determine what kind of instructions it can emit when generating
9954 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
9955 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
9956 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
9957 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
9958 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
9960 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
9961 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
9962 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
9963 @samp{strongarm1110},
9964 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
9965 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
9966 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
9967 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
9968 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
9969 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
9970 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
9971 @samp{cortex-a5}, @samp{cortex-a8}, @samp{cortex-a9},
9972 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-m4}, @samp{cortex-m3},
9975 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
9977 @item -mtune=@var{name}
9979 This option is very similar to the @option{-mcpu=} option, except that
9980 instead of specifying the actual target processor type, and hence
9981 restricting which instructions can be used, it specifies that GCC should
9982 tune the performance of the code as if the target were of the type
9983 specified in this option, but still choosing the instructions that it
9984 will generate based on the cpu specified by a @option{-mcpu=} option.
9985 For some ARM implementations better performance can be obtained by using
9988 @item -march=@var{name}
9990 This specifies the name of the target ARM architecture. GCC uses this
9991 name to determine what kind of instructions it can emit when generating
9992 assembly code. This option can be used in conjunction with or instead
9993 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
9994 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
9995 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
9996 @samp{armv6}, @samp{armv6j},
9997 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
9998 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
9999 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
10001 @item -mfpu=@var{name}
10002 @itemx -mfpe=@var{number}
10003 @itemx -mfp=@var{number}
10007 This specifies what floating point hardware (or hardware emulation) is
10008 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
10009 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-fp16},
10010 @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd}, @samp{vfpv3xd-fp16},
10011 @samp{neon}, @samp{neon-fp16}, @samp{vfpv4}, @samp{vfpv4-d16},
10012 @samp{fpv4-sp-d16} and @samp{neon-vfpv4}.
10013 @option{-mfp} and @option{-mfpe} are synonyms for
10014 @option{-mfpu}=@samp{fpe}@var{number}, for compatibility with older versions
10017 If @option{-msoft-float} is specified this specifies the format of
10018 floating point values.
10020 If the selected floating-point hardware includes the NEON extension
10021 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
10022 operations will not be used by GCC's auto-vectorization pass unless
10023 @option{-funsafe-math-optimizations} is also specified. This is
10024 because NEON hardware does not fully implement the IEEE 754 standard for
10025 floating-point arithmetic (in particular denormal values are treated as
10026 zero), so the use of NEON instructions may lead to a loss of precision.
10028 @item -mfp16-format=@var{name}
10029 @opindex mfp16-format
10030 Specify the format of the @code{__fp16} half-precision floating-point type.
10031 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
10032 the default is @samp{none}, in which case the @code{__fp16} type is not
10033 defined. @xref{Half-Precision}, for more information.
10035 @item -mstructure-size-boundary=@var{n}
10036 @opindex mstructure-size-boundary
10037 The size of all structures and unions will be rounded up to a multiple
10038 of the number of bits set by this option. Permissible values are 8, 32
10039 and 64. The default value varies for different toolchains. For the COFF
10040 targeted toolchain the default value is 8. A value of 64 is only allowed
10041 if the underlying ABI supports it.
10043 Specifying the larger number can produce faster, more efficient code, but
10044 can also increase the size of the program. Different values are potentially
10045 incompatible. Code compiled with one value cannot necessarily expect to
10046 work with code or libraries compiled with another value, if they exchange
10047 information using structures or unions.
10049 @item -mabort-on-noreturn
10050 @opindex mabort-on-noreturn
10051 Generate a call to the function @code{abort} at the end of a
10052 @code{noreturn} function. It will be executed if the function tries to
10056 @itemx -mno-long-calls
10057 @opindex mlong-calls
10058 @opindex mno-long-calls
10059 Tells the compiler to perform function calls by first loading the
10060 address of the function into a register and then performing a subroutine
10061 call on this register. This switch is needed if the target function
10062 will lie outside of the 64 megabyte addressing range of the offset based
10063 version of subroutine call instruction.
10065 Even if this switch is enabled, not all function calls will be turned
10066 into long calls. The heuristic is that static functions, functions
10067 which have the @samp{short-call} attribute, functions that are inside
10068 the scope of a @samp{#pragma no_long_calls} directive and functions whose
10069 definitions have already been compiled within the current compilation
10070 unit, will not be turned into long calls. The exception to this rule is
10071 that weak function definitions, functions with the @samp{long-call}
10072 attribute or the @samp{section} attribute, and functions that are within
10073 the scope of a @samp{#pragma long_calls} directive, will always be
10074 turned into long calls.
10076 This feature is not enabled by default. Specifying
10077 @option{-mno-long-calls} will restore the default behavior, as will
10078 placing the function calls within the scope of a @samp{#pragma
10079 long_calls_off} directive. Note these switches have no effect on how
10080 the compiler generates code to handle function calls via function
10083 @item -msingle-pic-base
10084 @opindex msingle-pic-base
10085 Treat the register used for PIC addressing as read-only, rather than
10086 loading it in the prologue for each function. The run-time system is
10087 responsible for initializing this register with an appropriate value
10088 before execution begins.
10090 @item -mpic-register=@var{reg}
10091 @opindex mpic-register
10092 Specify the register to be used for PIC addressing. The default is R10
10093 unless stack-checking is enabled, when R9 is used.
10095 @item -mcirrus-fix-invalid-insns
10096 @opindex mcirrus-fix-invalid-insns
10097 @opindex mno-cirrus-fix-invalid-insns
10098 Insert NOPs into the instruction stream to in order to work around
10099 problems with invalid Maverick instruction combinations. This option
10100 is only valid if the @option{-mcpu=ep9312} option has been used to
10101 enable generation of instructions for the Cirrus Maverick floating
10102 point co-processor. This option is not enabled by default, since the
10103 problem is only present in older Maverick implementations. The default
10104 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
10107 @item -mpoke-function-name
10108 @opindex mpoke-function-name
10109 Write the name of each function into the text section, directly
10110 preceding the function prologue. The generated code is similar to this:
10114 .ascii "arm_poke_function_name", 0
10117 .word 0xff000000 + (t1 - t0)
10118 arm_poke_function_name
10120 stmfd sp!, @{fp, ip, lr, pc@}
10124 When performing a stack backtrace, code can inspect the value of
10125 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
10126 location @code{pc - 12} and the top 8 bits are set, then we know that
10127 there is a function name embedded immediately preceding this location
10128 and has length @code{((pc[-3]) & 0xff000000)}.
10132 Generate code for the Thumb instruction set. The default is to
10133 use the 32-bit ARM instruction set.
10134 This option automatically enables either 16-bit Thumb-1 or
10135 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
10136 and @option{-march=@var{name}} options. This option is not passed to the
10137 assembler. If you want to force assembler files to be interpreted as Thumb code,
10138 either add a @samp{.thumb} directive to the source or pass the @option{-mthumb}
10139 option directly to the assembler by prefixing it with @option{-Wa}.
10142 @opindex mtpcs-frame
10143 Generate a stack frame that is compliant with the Thumb Procedure Call
10144 Standard for all non-leaf functions. (A leaf function is one that does
10145 not call any other functions.) The default is @option{-mno-tpcs-frame}.
10147 @item -mtpcs-leaf-frame
10148 @opindex mtpcs-leaf-frame
10149 Generate a stack frame that is compliant with the Thumb Procedure Call
10150 Standard for all leaf functions. (A leaf function is one that does
10151 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
10153 @item -mcallee-super-interworking
10154 @opindex mcallee-super-interworking
10155 Gives all externally visible functions in the file being compiled an ARM
10156 instruction set header which switches to Thumb mode before executing the
10157 rest of the function. This allows these functions to be called from
10158 non-interworking code. This option is not valid in AAPCS configurations
10159 because interworking is enabled by default.
10161 @item -mcaller-super-interworking
10162 @opindex mcaller-super-interworking
10163 Allows calls via function pointers (including virtual functions) to
10164 execute correctly regardless of whether the target code has been
10165 compiled for interworking or not. There is a small overhead in the cost
10166 of executing a function pointer if this option is enabled. This option
10167 is not valid in AAPCS configurations because interworking is enabled
10170 @item -mtp=@var{name}
10172 Specify the access model for the thread local storage pointer. The valid
10173 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
10174 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
10175 (supported in the arm6k architecture), and @option{auto}, which uses the
10176 best available method for the selected processor. The default setting is
10179 @item -mword-relocations
10180 @opindex mword-relocations
10181 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
10182 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
10183 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
10186 @item -mfix-cortex-m3-ldrd
10187 @opindex mfix-cortex-m3-ldrd
10188 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
10189 with overlapping destination and base registers are used. This option avoids
10190 generating these instructions. This option is enabled by default when
10191 @option{-mcpu=cortex-m3} is specified.
10196 @subsection AVR Options
10197 @cindex AVR Options
10199 These options are defined for AVR implementations:
10202 @item -mmcu=@var{mcu}
10204 Specify ATMEL AVR instruction set or MCU type.
10206 Instruction set avr1 is for the minimal AVR core, not supported by the C
10207 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
10208 attiny11, attiny12, attiny15, attiny28).
10210 Instruction set avr2 (default) is for the classic AVR core with up to
10211 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
10212 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
10213 at90c8534, at90s8535).
10215 Instruction set avr3 is for the classic AVR core with up to 128K program
10216 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
10218 Instruction set avr4 is for the enhanced AVR core with up to 8K program
10219 memory space (MCU types: atmega8, atmega83, atmega85).
10221 Instruction set avr5 is for the enhanced AVR core with up to 128K program
10222 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
10223 atmega64, atmega128, at43usb355, at94k).
10225 @item -mno-interrupts
10226 @opindex mno-interrupts
10227 Generated code is not compatible with hardware interrupts.
10228 Code size will be smaller.
10230 @item -mcall-prologues
10231 @opindex mcall-prologues
10232 Functions prologues/epilogues expanded as call to appropriate
10233 subroutines. Code size will be smaller.
10236 @opindex mtiny-stack
10237 Change only the low 8 bits of the stack pointer.
10241 Assume int to be 8 bit integer. This affects the sizes of all types: A
10242 char will be 1 byte, an int will be 1 byte, a long will be 2 bytes
10243 and long long will be 4 bytes. Please note that this option does not
10244 comply to the C standards, but it will provide you with smaller code
10248 @node Blackfin Options
10249 @subsection Blackfin Options
10250 @cindex Blackfin Options
10253 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
10255 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
10256 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
10257 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
10258 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
10259 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
10260 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
10261 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
10263 The optional @var{sirevision} specifies the silicon revision of the target
10264 Blackfin processor. Any workarounds available for the targeted silicon revision
10265 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
10266 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
10267 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
10268 hexadecimal digits representing the major and minor numbers in the silicon
10269 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
10270 is not defined. If @var{sirevision} is @samp{any}, the
10271 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
10272 If this optional @var{sirevision} is not used, GCC assumes the latest known
10273 silicon revision of the targeted Blackfin processor.
10275 Support for @samp{bf561} is incomplete. For @samp{bf561},
10276 Only the processor macro is defined.
10277 Without this option, @samp{bf532} is used as the processor by default.
10278 The corresponding predefined processor macros for @var{cpu} is to
10279 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
10280 provided by libgloss to be linked in if @option{-msim} is not given.
10284 Specifies that the program will be run on the simulator. This causes
10285 the simulator BSP provided by libgloss to be linked in. This option
10286 has effect only for @samp{bfin-elf} toolchain.
10287 Certain other options, such as @option{-mid-shared-library} and
10288 @option{-mfdpic}, imply @option{-msim}.
10290 @item -momit-leaf-frame-pointer
10291 @opindex momit-leaf-frame-pointer
10292 Don't keep the frame pointer in a register for leaf functions. This
10293 avoids the instructions to save, set up and restore frame pointers and
10294 makes an extra register available in leaf functions. The option
10295 @option{-fomit-frame-pointer} removes the frame pointer for all functions
10296 which might make debugging harder.
10298 @item -mspecld-anomaly
10299 @opindex mspecld-anomaly
10300 When enabled, the compiler will ensure that the generated code does not
10301 contain speculative loads after jump instructions. If this option is used,
10302 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
10304 @item -mno-specld-anomaly
10305 @opindex mno-specld-anomaly
10306 Don't generate extra code to prevent speculative loads from occurring.
10308 @item -mcsync-anomaly
10309 @opindex mcsync-anomaly
10310 When enabled, the compiler will ensure that the generated code does not
10311 contain CSYNC or SSYNC instructions too soon after conditional branches.
10312 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
10314 @item -mno-csync-anomaly
10315 @opindex mno-csync-anomaly
10316 Don't generate extra code to prevent CSYNC or SSYNC instructions from
10317 occurring too soon after a conditional branch.
10321 When enabled, the compiler is free to take advantage of the knowledge that
10322 the entire program fits into the low 64k of memory.
10325 @opindex mno-low-64k
10326 Assume that the program is arbitrarily large. This is the default.
10328 @item -mstack-check-l1
10329 @opindex mstack-check-l1
10330 Do stack checking using information placed into L1 scratchpad memory by the
10333 @item -mid-shared-library
10334 @opindex mid-shared-library
10335 Generate code that supports shared libraries via the library ID method.
10336 This allows for execute in place and shared libraries in an environment
10337 without virtual memory management. This option implies @option{-fPIC}.
10338 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10340 @item -mno-id-shared-library
10341 @opindex mno-id-shared-library
10342 Generate code that doesn't assume ID based shared libraries are being used.
10343 This is the default.
10345 @item -mleaf-id-shared-library
10346 @opindex mleaf-id-shared-library
10347 Generate code that supports shared libraries via the library ID method,
10348 but assumes that this library or executable won't link against any other
10349 ID shared libraries. That allows the compiler to use faster code for jumps
10352 @item -mno-leaf-id-shared-library
10353 @opindex mno-leaf-id-shared-library
10354 Do not assume that the code being compiled won't link against any ID shared
10355 libraries. Slower code will be generated for jump and call insns.
10357 @item -mshared-library-id=n
10358 @opindex mshared-library-id
10359 Specified the identification number of the ID based shared library being
10360 compiled. Specifying a value of 0 will generate more compact code, specifying
10361 other values will force the allocation of that number to the current
10362 library but is no more space or time efficient than omitting this option.
10366 Generate code that allows the data segment to be located in a different
10367 area of memory from the text segment. This allows for execute in place in
10368 an environment without virtual memory management by eliminating relocations
10369 against the text section.
10371 @item -mno-sep-data
10372 @opindex mno-sep-data
10373 Generate code that assumes that the data segment follows the text segment.
10374 This is the default.
10377 @itemx -mno-long-calls
10378 @opindex mlong-calls
10379 @opindex mno-long-calls
10380 Tells the compiler to perform function calls by first loading the
10381 address of the function into a register and then performing a subroutine
10382 call on this register. This switch is needed if the target function
10383 will lie outside of the 24 bit addressing range of the offset based
10384 version of subroutine call instruction.
10386 This feature is not enabled by default. Specifying
10387 @option{-mno-long-calls} will restore the default behavior. Note these
10388 switches have no effect on how the compiler generates code to handle
10389 function calls via function pointers.
10393 Link with the fast floating-point library. This library relaxes some of
10394 the IEEE floating-point standard's rules for checking inputs against
10395 Not-a-Number (NAN), in the interest of performance.
10398 @opindex minline-plt
10399 Enable inlining of PLT entries in function calls to functions that are
10400 not known to bind locally. It has no effect without @option{-mfdpic}.
10403 @opindex mmulticore
10404 Build standalone application for multicore Blackfin processor. Proper
10405 start files and link scripts will be used to support multicore.
10406 This option defines @code{__BFIN_MULTICORE}. It can only be used with
10407 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
10408 @option{-mcorea} or @option{-mcoreb}. If it's used without
10409 @option{-mcorea} or @option{-mcoreb}, single application/dual core
10410 programming model is used. In this model, the main function of Core B
10411 should be named as coreb_main. If it's used with @option{-mcorea} or
10412 @option{-mcoreb}, one application per core programming model is used.
10413 If this option is not used, single core application programming
10418 Build standalone application for Core A of BF561 when using
10419 one application per core programming model. Proper start files
10420 and link scripts will be used to support Core A. This option
10421 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
10425 Build standalone application for Core B of BF561 when using
10426 one application per core programming model. Proper start files
10427 and link scripts will be used to support Core B. This option
10428 defines @code{__BFIN_COREB}. When this option is used, coreb_main
10429 should be used instead of main. It must be used with
10430 @option{-mmulticore}.
10434 Build standalone application for SDRAM. Proper start files and
10435 link scripts will be used to put the application into SDRAM.
10436 Loader should initialize SDRAM before loading the application
10437 into SDRAM. This option defines @code{__BFIN_SDRAM}.
10441 Assume that ICPLBs are enabled at runtime. This has an effect on certain
10442 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
10443 are enabled; for standalone applications the default is off.
10447 @subsection CRIS Options
10448 @cindex CRIS Options
10450 These options are defined specifically for the CRIS ports.
10453 @item -march=@var{architecture-type}
10454 @itemx -mcpu=@var{architecture-type}
10457 Generate code for the specified architecture. The choices for
10458 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
10459 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
10460 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
10463 @item -mtune=@var{architecture-type}
10465 Tune to @var{architecture-type} everything applicable about the generated
10466 code, except for the ABI and the set of available instructions. The
10467 choices for @var{architecture-type} are the same as for
10468 @option{-march=@var{architecture-type}}.
10470 @item -mmax-stack-frame=@var{n}
10471 @opindex mmax-stack-frame
10472 Warn when the stack frame of a function exceeds @var{n} bytes.
10478 The options @option{-metrax4} and @option{-metrax100} are synonyms for
10479 @option{-march=v3} and @option{-march=v8} respectively.
10481 @item -mmul-bug-workaround
10482 @itemx -mno-mul-bug-workaround
10483 @opindex mmul-bug-workaround
10484 @opindex mno-mul-bug-workaround
10485 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
10486 models where it applies. This option is active by default.
10490 Enable CRIS-specific verbose debug-related information in the assembly
10491 code. This option also has the effect to turn off the @samp{#NO_APP}
10492 formatted-code indicator to the assembler at the beginning of the
10497 Do not use condition-code results from previous instruction; always emit
10498 compare and test instructions before use of condition codes.
10500 @item -mno-side-effects
10501 @opindex mno-side-effects
10502 Do not emit instructions with side-effects in addressing modes other than
10505 @item -mstack-align
10506 @itemx -mno-stack-align
10507 @itemx -mdata-align
10508 @itemx -mno-data-align
10509 @itemx -mconst-align
10510 @itemx -mno-const-align
10511 @opindex mstack-align
10512 @opindex mno-stack-align
10513 @opindex mdata-align
10514 @opindex mno-data-align
10515 @opindex mconst-align
10516 @opindex mno-const-align
10517 These options (no-options) arranges (eliminate arrangements) for the
10518 stack-frame, individual data and constants to be aligned for the maximum
10519 single data access size for the chosen CPU model. The default is to
10520 arrange for 32-bit alignment. ABI details such as structure layout are
10521 not affected by these options.
10529 Similar to the stack- data- and const-align options above, these options
10530 arrange for stack-frame, writable data and constants to all be 32-bit,
10531 16-bit or 8-bit aligned. The default is 32-bit alignment.
10533 @item -mno-prologue-epilogue
10534 @itemx -mprologue-epilogue
10535 @opindex mno-prologue-epilogue
10536 @opindex mprologue-epilogue
10537 With @option{-mno-prologue-epilogue}, the normal function prologue and
10538 epilogue that sets up the stack-frame are omitted and no return
10539 instructions or return sequences are generated in the code. Use this
10540 option only together with visual inspection of the compiled code: no
10541 warnings or errors are generated when call-saved registers must be saved,
10542 or storage for local variable needs to be allocated.
10546 @opindex mno-gotplt
10548 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
10549 instruction sequences that load addresses for functions from the PLT part
10550 of the GOT rather than (traditional on other architectures) calls to the
10551 PLT@. The default is @option{-mgotplt}.
10555 Legacy no-op option only recognized with the cris-axis-elf and
10556 cris-axis-linux-gnu targets.
10560 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
10564 This option, recognized for the cris-axis-elf arranges
10565 to link with input-output functions from a simulator library. Code,
10566 initialized data and zero-initialized data are allocated consecutively.
10570 Like @option{-sim}, but pass linker options to locate initialized data at
10571 0x40000000 and zero-initialized data at 0x80000000.
10575 @subsection CRX Options
10576 @cindex CRX Options
10578 These options are defined specifically for the CRX ports.
10584 Enable the use of multiply-accumulate instructions. Disabled by default.
10587 @opindex mpush-args
10588 Push instructions will be used to pass outgoing arguments when functions
10589 are called. Enabled by default.
10592 @node Darwin Options
10593 @subsection Darwin Options
10594 @cindex Darwin options
10596 These options are defined for all architectures running the Darwin operating
10599 FSF GCC on Darwin does not create ``fat'' object files; it will create
10600 an object file for the single architecture that it was built to
10601 target. Apple's GCC on Darwin does create ``fat'' files if multiple
10602 @option{-arch} options are used; it does so by running the compiler or
10603 linker multiple times and joining the results together with
10606 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
10607 @samp{i686}) is determined by the flags that specify the ISA
10608 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
10609 @option{-force_cpusubtype_ALL} option can be used to override this.
10611 The Darwin tools vary in their behavior when presented with an ISA
10612 mismatch. The assembler, @file{as}, will only permit instructions to
10613 be used that are valid for the subtype of the file it is generating,
10614 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
10615 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
10616 and print an error if asked to create a shared library with a less
10617 restrictive subtype than its input files (for instance, trying to put
10618 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
10619 for executables, @file{ld}, will quietly give the executable the most
10620 restrictive subtype of any of its input files.
10625 Add the framework directory @var{dir} to the head of the list of
10626 directories to be searched for header files. These directories are
10627 interleaved with those specified by @option{-I} options and are
10628 scanned in a left-to-right order.
10630 A framework directory is a directory with frameworks in it. A
10631 framework is a directory with a @samp{"Headers"} and/or
10632 @samp{"PrivateHeaders"} directory contained directly in it that ends
10633 in @samp{".framework"}. The name of a framework is the name of this
10634 directory excluding the @samp{".framework"}. Headers associated with
10635 the framework are found in one of those two directories, with
10636 @samp{"Headers"} being searched first. A subframework is a framework
10637 directory that is in a framework's @samp{"Frameworks"} directory.
10638 Includes of subframework headers can only appear in a header of a
10639 framework that contains the subframework, or in a sibling subframework
10640 header. Two subframeworks are siblings if they occur in the same
10641 framework. A subframework should not have the same name as a
10642 framework, a warning will be issued if this is violated. Currently a
10643 subframework cannot have subframeworks, in the future, the mechanism
10644 may be extended to support this. The standard frameworks can be found
10645 in @samp{"/System/Library/Frameworks"} and
10646 @samp{"/Library/Frameworks"}. An example include looks like
10647 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
10648 the name of the framework and header.h is found in the
10649 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
10651 @item -iframework@var{dir}
10652 @opindex iframework
10653 Like @option{-F} except the directory is a treated as a system
10654 directory. The main difference between this @option{-iframework} and
10655 @option{-F} is that with @option{-iframework} the compiler does not
10656 warn about constructs contained within header files found via
10657 @var{dir}. This option is valid only for the C family of languages.
10661 Emit debugging information for symbols that are used. For STABS
10662 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
10663 This is by default ON@.
10667 Emit debugging information for all symbols and types.
10669 @item -mmacosx-version-min=@var{version}
10670 The earliest version of MacOS X that this executable will run on
10671 is @var{version}. Typical values of @var{version} include @code{10.1},
10672 @code{10.2}, and @code{10.3.9}.
10674 If the compiler was built to use the system's headers by default,
10675 then the default for this option is the system version on which the
10676 compiler is running, otherwise the default is to make choices which
10677 are compatible with as many systems and code bases as possible.
10681 Enable kernel development mode. The @option{-mkernel} option sets
10682 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
10683 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
10684 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
10685 applicable. This mode also sets @option{-mno-altivec},
10686 @option{-msoft-float}, @option{-fno-builtin} and
10687 @option{-mlong-branch} for PowerPC targets.
10689 @item -mone-byte-bool
10690 @opindex mone-byte-bool
10691 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
10692 By default @samp{sizeof(bool)} is @samp{4} when compiling for
10693 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
10694 option has no effect on x86.
10696 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
10697 to generate code that is not binary compatible with code generated
10698 without that switch. Using this switch may require recompiling all
10699 other modules in a program, including system libraries. Use this
10700 switch to conform to a non-default data model.
10702 @item -mfix-and-continue
10703 @itemx -ffix-and-continue
10704 @itemx -findirect-data
10705 @opindex mfix-and-continue
10706 @opindex ffix-and-continue
10707 @opindex findirect-data
10708 Generate code suitable for fast turn around development. Needed to
10709 enable gdb to dynamically load @code{.o} files into already running
10710 programs. @option{-findirect-data} and @option{-ffix-and-continue}
10711 are provided for backwards compatibility.
10715 Loads all members of static archive libraries.
10716 See man ld(1) for more information.
10718 @item -arch_errors_fatal
10719 @opindex arch_errors_fatal
10720 Cause the errors having to do with files that have the wrong architecture
10723 @item -bind_at_load
10724 @opindex bind_at_load
10725 Causes the output file to be marked such that the dynamic linker will
10726 bind all undefined references when the file is loaded or launched.
10730 Produce a Mach-o bundle format file.
10731 See man ld(1) for more information.
10733 @item -bundle_loader @var{executable}
10734 @opindex bundle_loader
10735 This option specifies the @var{executable} that will be loading the build
10736 output file being linked. See man ld(1) for more information.
10739 @opindex dynamiclib
10740 When passed this option, GCC will produce a dynamic library instead of
10741 an executable when linking, using the Darwin @file{libtool} command.
10743 @item -force_cpusubtype_ALL
10744 @opindex force_cpusubtype_ALL
10745 This causes GCC's output file to have the @var{ALL} subtype, instead of
10746 one controlled by the @option{-mcpu} or @option{-march} option.
10748 @item -allowable_client @var{client_name}
10749 @itemx -client_name
10750 @itemx -compatibility_version
10751 @itemx -current_version
10753 @itemx -dependency-file
10755 @itemx -dylinker_install_name
10757 @itemx -exported_symbols_list
10759 @itemx -flat_namespace
10760 @itemx -force_flat_namespace
10761 @itemx -headerpad_max_install_names
10764 @itemx -install_name
10765 @itemx -keep_private_externs
10766 @itemx -multi_module
10767 @itemx -multiply_defined
10768 @itemx -multiply_defined_unused
10770 @itemx -no_dead_strip_inits_and_terms
10771 @itemx -nofixprebinding
10772 @itemx -nomultidefs
10774 @itemx -noseglinkedit
10775 @itemx -pagezero_size
10777 @itemx -prebind_all_twolevel_modules
10778 @itemx -private_bundle
10779 @itemx -read_only_relocs
10781 @itemx -sectobjectsymbols
10785 @itemx -sectobjectsymbols
10788 @itemx -segs_read_only_addr
10789 @itemx -segs_read_write_addr
10790 @itemx -seg_addr_table
10791 @itemx -seg_addr_table_filename
10792 @itemx -seglinkedit
10794 @itemx -segs_read_only_addr
10795 @itemx -segs_read_write_addr
10796 @itemx -single_module
10798 @itemx -sub_library
10799 @itemx -sub_umbrella
10800 @itemx -twolevel_namespace
10803 @itemx -unexported_symbols_list
10804 @itemx -weak_reference_mismatches
10805 @itemx -whatsloaded
10806 @opindex allowable_client
10807 @opindex client_name
10808 @opindex compatibility_version
10809 @opindex current_version
10810 @opindex dead_strip
10811 @opindex dependency-file
10812 @opindex dylib_file
10813 @opindex dylinker_install_name
10815 @opindex exported_symbols_list
10817 @opindex flat_namespace
10818 @opindex force_flat_namespace
10819 @opindex headerpad_max_install_names
10820 @opindex image_base
10822 @opindex install_name
10823 @opindex keep_private_externs
10824 @opindex multi_module
10825 @opindex multiply_defined
10826 @opindex multiply_defined_unused
10827 @opindex noall_load
10828 @opindex no_dead_strip_inits_and_terms
10829 @opindex nofixprebinding
10830 @opindex nomultidefs
10832 @opindex noseglinkedit
10833 @opindex pagezero_size
10835 @opindex prebind_all_twolevel_modules
10836 @opindex private_bundle
10837 @opindex read_only_relocs
10839 @opindex sectobjectsymbols
10842 @opindex sectcreate
10843 @opindex sectobjectsymbols
10846 @opindex segs_read_only_addr
10847 @opindex segs_read_write_addr
10848 @opindex seg_addr_table
10849 @opindex seg_addr_table_filename
10850 @opindex seglinkedit
10852 @opindex segs_read_only_addr
10853 @opindex segs_read_write_addr
10854 @opindex single_module
10856 @opindex sub_library
10857 @opindex sub_umbrella
10858 @opindex twolevel_namespace
10861 @opindex unexported_symbols_list
10862 @opindex weak_reference_mismatches
10863 @opindex whatsloaded
10864 These options are passed to the Darwin linker. The Darwin linker man page
10865 describes them in detail.
10868 @node DEC Alpha Options
10869 @subsection DEC Alpha Options
10871 These @samp{-m} options are defined for the DEC Alpha implementations:
10874 @item -mno-soft-float
10875 @itemx -msoft-float
10876 @opindex mno-soft-float
10877 @opindex msoft-float
10878 Use (do not use) the hardware floating-point instructions for
10879 floating-point operations. When @option{-msoft-float} is specified,
10880 functions in @file{libgcc.a} will be used to perform floating-point
10881 operations. Unless they are replaced by routines that emulate the
10882 floating-point operations, or compiled in such a way as to call such
10883 emulations routines, these routines will issue floating-point
10884 operations. If you are compiling for an Alpha without floating-point
10885 operations, you must ensure that the library is built so as not to call
10888 Note that Alpha implementations without floating-point operations are
10889 required to have floating-point registers.
10892 @itemx -mno-fp-regs
10894 @opindex mno-fp-regs
10895 Generate code that uses (does not use) the floating-point register set.
10896 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
10897 register set is not used, floating point operands are passed in integer
10898 registers as if they were integers and floating-point results are passed
10899 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
10900 so any function with a floating-point argument or return value called by code
10901 compiled with @option{-mno-fp-regs} must also be compiled with that
10904 A typical use of this option is building a kernel that does not use,
10905 and hence need not save and restore, any floating-point registers.
10909 The Alpha architecture implements floating-point hardware optimized for
10910 maximum performance. It is mostly compliant with the IEEE floating
10911 point standard. However, for full compliance, software assistance is
10912 required. This option generates code fully IEEE compliant code
10913 @emph{except} that the @var{inexact-flag} is not maintained (see below).
10914 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
10915 defined during compilation. The resulting code is less efficient but is
10916 able to correctly support denormalized numbers and exceptional IEEE
10917 values such as not-a-number and plus/minus infinity. Other Alpha
10918 compilers call this option @option{-ieee_with_no_inexact}.
10920 @item -mieee-with-inexact
10921 @opindex mieee-with-inexact
10922 This is like @option{-mieee} except the generated code also maintains
10923 the IEEE @var{inexact-flag}. Turning on this option causes the
10924 generated code to implement fully-compliant IEEE math. In addition to
10925 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
10926 macro. On some Alpha implementations the resulting code may execute
10927 significantly slower than the code generated by default. Since there is
10928 very little code that depends on the @var{inexact-flag}, you should
10929 normally not specify this option. Other Alpha compilers call this
10930 option @option{-ieee_with_inexact}.
10932 @item -mfp-trap-mode=@var{trap-mode}
10933 @opindex mfp-trap-mode
10934 This option controls what floating-point related traps are enabled.
10935 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
10936 The trap mode can be set to one of four values:
10940 This is the default (normal) setting. The only traps that are enabled
10941 are the ones that cannot be disabled in software (e.g., division by zero
10945 In addition to the traps enabled by @samp{n}, underflow traps are enabled
10949 Like @samp{u}, but the instructions are marked to be safe for software
10950 completion (see Alpha architecture manual for details).
10953 Like @samp{su}, but inexact traps are enabled as well.
10956 @item -mfp-rounding-mode=@var{rounding-mode}
10957 @opindex mfp-rounding-mode
10958 Selects the IEEE rounding mode. Other Alpha compilers call this option
10959 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
10964 Normal IEEE rounding mode. Floating point numbers are rounded towards
10965 the nearest machine number or towards the even machine number in case
10969 Round towards minus infinity.
10972 Chopped rounding mode. Floating point numbers are rounded towards zero.
10975 Dynamic rounding mode. A field in the floating point control register
10976 (@var{fpcr}, see Alpha architecture reference manual) controls the
10977 rounding mode in effect. The C library initializes this register for
10978 rounding towards plus infinity. Thus, unless your program modifies the
10979 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
10982 @item -mtrap-precision=@var{trap-precision}
10983 @opindex mtrap-precision
10984 In the Alpha architecture, floating point traps are imprecise. This
10985 means without software assistance it is impossible to recover from a
10986 floating trap and program execution normally needs to be terminated.
10987 GCC can generate code that can assist operating system trap handlers
10988 in determining the exact location that caused a floating point trap.
10989 Depending on the requirements of an application, different levels of
10990 precisions can be selected:
10994 Program precision. This option is the default and means a trap handler
10995 can only identify which program caused a floating point exception.
10998 Function precision. The trap handler can determine the function that
10999 caused a floating point exception.
11002 Instruction precision. The trap handler can determine the exact
11003 instruction that caused a floating point exception.
11006 Other Alpha compilers provide the equivalent options called
11007 @option{-scope_safe} and @option{-resumption_safe}.
11009 @item -mieee-conformant
11010 @opindex mieee-conformant
11011 This option marks the generated code as IEEE conformant. You must not
11012 use this option unless you also specify @option{-mtrap-precision=i} and either
11013 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
11014 is to emit the line @samp{.eflag 48} in the function prologue of the
11015 generated assembly file. Under DEC Unix, this has the effect that
11016 IEEE-conformant math library routines will be linked in.
11018 @item -mbuild-constants
11019 @opindex mbuild-constants
11020 Normally GCC examines a 32- or 64-bit integer constant to
11021 see if it can construct it from smaller constants in two or three
11022 instructions. If it cannot, it will output the constant as a literal and
11023 generate code to load it from the data segment at runtime.
11025 Use this option to require GCC to construct @emph{all} integer constants
11026 using code, even if it takes more instructions (the maximum is six).
11028 You would typically use this option to build a shared library dynamic
11029 loader. Itself a shared library, it must relocate itself in memory
11030 before it can find the variables and constants in its own data segment.
11036 Select whether to generate code to be assembled by the vendor-supplied
11037 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
11055 Indicate whether GCC should generate code to use the optional BWX,
11056 CIX, FIX and MAX instruction sets. The default is to use the instruction
11057 sets supported by the CPU type specified via @option{-mcpu=} option or that
11058 of the CPU on which GCC was built if none was specified.
11061 @itemx -mfloat-ieee
11062 @opindex mfloat-vax
11063 @opindex mfloat-ieee
11064 Generate code that uses (does not use) VAX F and G floating point
11065 arithmetic instead of IEEE single and double precision.
11067 @item -mexplicit-relocs
11068 @itemx -mno-explicit-relocs
11069 @opindex mexplicit-relocs
11070 @opindex mno-explicit-relocs
11071 Older Alpha assemblers provided no way to generate symbol relocations
11072 except via assembler macros. Use of these macros does not allow
11073 optimal instruction scheduling. GNU binutils as of version 2.12
11074 supports a new syntax that allows the compiler to explicitly mark
11075 which relocations should apply to which instructions. This option
11076 is mostly useful for debugging, as GCC detects the capabilities of
11077 the assembler when it is built and sets the default accordingly.
11080 @itemx -mlarge-data
11081 @opindex msmall-data
11082 @opindex mlarge-data
11083 When @option{-mexplicit-relocs} is in effect, static data is
11084 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
11085 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
11086 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
11087 16-bit relocations off of the @code{$gp} register. This limits the
11088 size of the small data area to 64KB, but allows the variables to be
11089 directly accessed via a single instruction.
11091 The default is @option{-mlarge-data}. With this option the data area
11092 is limited to just below 2GB@. Programs that require more than 2GB of
11093 data must use @code{malloc} or @code{mmap} to allocate the data in the
11094 heap instead of in the program's data segment.
11096 When generating code for shared libraries, @option{-fpic} implies
11097 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
11100 @itemx -mlarge-text
11101 @opindex msmall-text
11102 @opindex mlarge-text
11103 When @option{-msmall-text} is used, the compiler assumes that the
11104 code of the entire program (or shared library) fits in 4MB, and is
11105 thus reachable with a branch instruction. When @option{-msmall-data}
11106 is used, the compiler can assume that all local symbols share the
11107 same @code{$gp} value, and thus reduce the number of instructions
11108 required for a function call from 4 to 1.
11110 The default is @option{-mlarge-text}.
11112 @item -mcpu=@var{cpu_type}
11114 Set the instruction set and instruction scheduling parameters for
11115 machine type @var{cpu_type}. You can specify either the @samp{EV}
11116 style name or the corresponding chip number. GCC supports scheduling
11117 parameters for the EV4, EV5 and EV6 family of processors and will
11118 choose the default values for the instruction set from the processor
11119 you specify. If you do not specify a processor type, GCC will default
11120 to the processor on which the compiler was built.
11122 Supported values for @var{cpu_type} are
11128 Schedules as an EV4 and has no instruction set extensions.
11132 Schedules as an EV5 and has no instruction set extensions.
11136 Schedules as an EV5 and supports the BWX extension.
11141 Schedules as an EV5 and supports the BWX and MAX extensions.
11145 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
11149 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
11152 Native Linux/GNU toolchains also support the value @samp{native},
11153 which selects the best architecture option for the host processor.
11154 @option{-mcpu=native} has no effect if GCC does not recognize
11157 @item -mtune=@var{cpu_type}
11159 Set only the instruction scheduling parameters for machine type
11160 @var{cpu_type}. The instruction set is not changed.
11162 Native Linux/GNU toolchains also support the value @samp{native},
11163 which selects the best architecture option for the host processor.
11164 @option{-mtune=native} has no effect if GCC does not recognize
11167 @item -mmemory-latency=@var{time}
11168 @opindex mmemory-latency
11169 Sets the latency the scheduler should assume for typical memory
11170 references as seen by the application. This number is highly
11171 dependent on the memory access patterns used by the application
11172 and the size of the external cache on the machine.
11174 Valid options for @var{time} are
11178 A decimal number representing clock cycles.
11184 The compiler contains estimates of the number of clock cycles for
11185 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
11186 (also called Dcache, Scache, and Bcache), as well as to main memory.
11187 Note that L3 is only valid for EV5.
11192 @node DEC Alpha/VMS Options
11193 @subsection DEC Alpha/VMS Options
11195 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
11198 @item -mvms-return-codes
11199 @opindex mvms-return-codes
11200 Return VMS condition codes from main. The default is to return POSIX
11201 style condition (e.g.@: error) codes.
11203 @item -mdebug-main=@var{prefix}
11204 @opindex mdebug-main=@var{prefix}
11205 Flag the first routine whose name starts with @var{prefix} as the main
11206 routine for the debugger.
11210 Default to 64bit memory allocation routines.
11214 @subsection FR30 Options
11215 @cindex FR30 Options
11217 These options are defined specifically for the FR30 port.
11221 @item -msmall-model
11222 @opindex msmall-model
11223 Use the small address space model. This can produce smaller code, but
11224 it does assume that all symbolic values and addresses will fit into a
11229 Assume that run-time support has been provided and so there is no need
11230 to include the simulator library (@file{libsim.a}) on the linker
11236 @subsection FRV Options
11237 @cindex FRV Options
11243 Only use the first 32 general purpose registers.
11248 Use all 64 general purpose registers.
11253 Use only the first 32 floating point registers.
11258 Use all 64 floating point registers
11261 @opindex mhard-float
11263 Use hardware instructions for floating point operations.
11266 @opindex msoft-float
11268 Use library routines for floating point operations.
11273 Dynamically allocate condition code registers.
11278 Do not try to dynamically allocate condition code registers, only
11279 use @code{icc0} and @code{fcc0}.
11284 Change ABI to use double word insns.
11289 Do not use double word instructions.
11294 Use floating point double instructions.
11297 @opindex mno-double
11299 Do not use floating point double instructions.
11304 Use media instructions.
11309 Do not use media instructions.
11314 Use multiply and add/subtract instructions.
11317 @opindex mno-muladd
11319 Do not use multiply and add/subtract instructions.
11324 Select the FDPIC ABI, that uses function descriptors to represent
11325 pointers to functions. Without any PIC/PIE-related options, it
11326 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
11327 assumes GOT entries and small data are within a 12-bit range from the
11328 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
11329 are computed with 32 bits.
11330 With a @samp{bfin-elf} target, this option implies @option{-msim}.
11333 @opindex minline-plt
11335 Enable inlining of PLT entries in function calls to functions that are
11336 not known to bind locally. It has no effect without @option{-mfdpic}.
11337 It's enabled by default if optimizing for speed and compiling for
11338 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
11339 optimization option such as @option{-O3} or above is present in the
11345 Assume a large TLS segment when generating thread-local code.
11350 Do not assume a large TLS segment when generating thread-local code.
11355 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
11356 that is known to be in read-only sections. It's enabled by default,
11357 except for @option{-fpic} or @option{-fpie}: even though it may help
11358 make the global offset table smaller, it trades 1 instruction for 4.
11359 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
11360 one of which may be shared by multiple symbols, and it avoids the need
11361 for a GOT entry for the referenced symbol, so it's more likely to be a
11362 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
11364 @item -multilib-library-pic
11365 @opindex multilib-library-pic
11367 Link with the (library, not FD) pic libraries. It's implied by
11368 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
11369 @option{-fpic} without @option{-mfdpic}. You should never have to use
11373 @opindex mlinked-fp
11375 Follow the EABI requirement of always creating a frame pointer whenever
11376 a stack frame is allocated. This option is enabled by default and can
11377 be disabled with @option{-mno-linked-fp}.
11380 @opindex mlong-calls
11382 Use indirect addressing to call functions outside the current
11383 compilation unit. This allows the functions to be placed anywhere
11384 within the 32-bit address space.
11386 @item -malign-labels
11387 @opindex malign-labels
11389 Try to align labels to an 8-byte boundary by inserting nops into the
11390 previous packet. This option only has an effect when VLIW packing
11391 is enabled. It doesn't create new packets; it merely adds nops to
11394 @item -mlibrary-pic
11395 @opindex mlibrary-pic
11397 Generate position-independent EABI code.
11402 Use only the first four media accumulator registers.
11407 Use all eight media accumulator registers.
11412 Pack VLIW instructions.
11417 Do not pack VLIW instructions.
11420 @opindex mno-eflags
11422 Do not mark ABI switches in e_flags.
11425 @opindex mcond-move
11427 Enable the use of conditional-move instructions (default).
11429 This switch is mainly for debugging the compiler and will likely be removed
11430 in a future version.
11432 @item -mno-cond-move
11433 @opindex mno-cond-move
11435 Disable the use of conditional-move instructions.
11437 This switch is mainly for debugging the compiler and will likely be removed
11438 in a future version.
11443 Enable the use of conditional set instructions (default).
11445 This switch is mainly for debugging the compiler and will likely be removed
11446 in a future version.
11451 Disable the use of conditional set instructions.
11453 This switch is mainly for debugging the compiler and will likely be removed
11454 in a future version.
11457 @opindex mcond-exec
11459 Enable the use of conditional execution (default).
11461 This switch is mainly for debugging the compiler and will likely be removed
11462 in a future version.
11464 @item -mno-cond-exec
11465 @opindex mno-cond-exec
11467 Disable the use of conditional execution.
11469 This switch is mainly for debugging the compiler and will likely be removed
11470 in a future version.
11472 @item -mvliw-branch
11473 @opindex mvliw-branch
11475 Run a pass to pack branches into VLIW instructions (default).
11477 This switch is mainly for debugging the compiler and will likely be removed
11478 in a future version.
11480 @item -mno-vliw-branch
11481 @opindex mno-vliw-branch
11483 Do not run a pass to pack branches into VLIW instructions.
11485 This switch is mainly for debugging the compiler and will likely be removed
11486 in a future version.
11488 @item -mmulti-cond-exec
11489 @opindex mmulti-cond-exec
11491 Enable optimization of @code{&&} and @code{||} in conditional execution
11494 This switch is mainly for debugging the compiler and will likely be removed
11495 in a future version.
11497 @item -mno-multi-cond-exec
11498 @opindex mno-multi-cond-exec
11500 Disable optimization of @code{&&} and @code{||} in conditional execution.
11502 This switch is mainly for debugging the compiler and will likely be removed
11503 in a future version.
11505 @item -mnested-cond-exec
11506 @opindex mnested-cond-exec
11508 Enable nested conditional execution optimizations (default).
11510 This switch is mainly for debugging the compiler and will likely be removed
11511 in a future version.
11513 @item -mno-nested-cond-exec
11514 @opindex mno-nested-cond-exec
11516 Disable nested conditional execution optimizations.
11518 This switch is mainly for debugging the compiler and will likely be removed
11519 in a future version.
11521 @item -moptimize-membar
11522 @opindex moptimize-membar
11524 This switch removes redundant @code{membar} instructions from the
11525 compiler generated code. It is enabled by default.
11527 @item -mno-optimize-membar
11528 @opindex mno-optimize-membar
11530 This switch disables the automatic removal of redundant @code{membar}
11531 instructions from the generated code.
11533 @item -mtomcat-stats
11534 @opindex mtomcat-stats
11536 Cause gas to print out tomcat statistics.
11538 @item -mcpu=@var{cpu}
11541 Select the processor type for which to generate code. Possible values are
11542 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
11543 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
11547 @node GNU/Linux Options
11548 @subsection GNU/Linux Options
11550 These @samp{-m} options are defined for GNU/Linux targets:
11555 Use the GNU C library. This is the default except
11556 on @samp{*-*-linux-*uclibc*} and @samp{*-*-linux-*android*} targets.
11560 Use uClibc C library. This is the default on
11561 @samp{*-*-linux-*uclibc*} targets.
11565 Use Bionic C library. This is the default on
11566 @samp{*-*-linux-*android*} targets.
11570 Compile code compatible with Android platform. This is the default on
11571 @samp{*-*-linux-*android*} targets.
11573 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
11574 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
11575 this option makes the GCC driver pass Android-specific options to the linker.
11576 Finally, this option causes the preprocessor macro @code{__ANDROID__}
11579 @item -tno-android-cc
11580 @opindex tno-android-cc
11581 Disable compilation effects of @option{-mandroid}, i.e., do not enable
11582 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
11583 @option{-fno-rtti} by default.
11585 @item -tno-android-ld
11586 @opindex tno-android-ld
11587 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
11588 linking options to the linker.
11592 @node H8/300 Options
11593 @subsection H8/300 Options
11595 These @samp{-m} options are defined for the H8/300 implementations:
11600 Shorten some address references at link time, when possible; uses the
11601 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
11602 ld, Using ld}, for a fuller description.
11606 Generate code for the H8/300H@.
11610 Generate code for the H8S@.
11614 Generate code for the H8S and H8/300H in the normal mode. This switch
11615 must be used either with @option{-mh} or @option{-ms}.
11619 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
11623 Make @code{int} data 32 bits by default.
11626 @opindex malign-300
11627 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
11628 The default for the H8/300H and H8S is to align longs and floats on 4
11630 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
11631 This option has no effect on the H8/300.
11635 @subsection HPPA Options
11636 @cindex HPPA Options
11638 These @samp{-m} options are defined for the HPPA family of computers:
11641 @item -march=@var{architecture-type}
11643 Generate code for the specified architecture. The choices for
11644 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
11645 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
11646 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
11647 architecture option for your machine. Code compiled for lower numbered
11648 architectures will run on higher numbered architectures, but not the
11651 @item -mpa-risc-1-0
11652 @itemx -mpa-risc-1-1
11653 @itemx -mpa-risc-2-0
11654 @opindex mpa-risc-1-0
11655 @opindex mpa-risc-1-1
11656 @opindex mpa-risc-2-0
11657 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
11660 @opindex mbig-switch
11661 Generate code suitable for big switch tables. Use this option only if
11662 the assembler/linker complain about out of range branches within a switch
11665 @item -mjump-in-delay
11666 @opindex mjump-in-delay
11667 Fill delay slots of function calls with unconditional jump instructions
11668 by modifying the return pointer for the function call to be the target
11669 of the conditional jump.
11671 @item -mdisable-fpregs
11672 @opindex mdisable-fpregs
11673 Prevent floating point registers from being used in any manner. This is
11674 necessary for compiling kernels which perform lazy context switching of
11675 floating point registers. If you use this option and attempt to perform
11676 floating point operations, the compiler will abort.
11678 @item -mdisable-indexing
11679 @opindex mdisable-indexing
11680 Prevent the compiler from using indexing address modes. This avoids some
11681 rather obscure problems when compiling MIG generated code under MACH@.
11683 @item -mno-space-regs
11684 @opindex mno-space-regs
11685 Generate code that assumes the target has no space registers. This allows
11686 GCC to generate faster indirect calls and use unscaled index address modes.
11688 Such code is suitable for level 0 PA systems and kernels.
11690 @item -mfast-indirect-calls
11691 @opindex mfast-indirect-calls
11692 Generate code that assumes calls never cross space boundaries. This
11693 allows GCC to emit code which performs faster indirect calls.
11695 This option will not work in the presence of shared libraries or nested
11698 @item -mfixed-range=@var{register-range}
11699 @opindex mfixed-range
11700 Generate code treating the given register range as fixed registers.
11701 A fixed register is one that the register allocator can not use. This is
11702 useful when compiling kernel code. A register range is specified as
11703 two registers separated by a dash. Multiple register ranges can be
11704 specified separated by a comma.
11706 @item -mlong-load-store
11707 @opindex mlong-load-store
11708 Generate 3-instruction load and store sequences as sometimes required by
11709 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
11712 @item -mportable-runtime
11713 @opindex mportable-runtime
11714 Use the portable calling conventions proposed by HP for ELF systems.
11718 Enable the use of assembler directives only GAS understands.
11720 @item -mschedule=@var{cpu-type}
11722 Schedule code according to the constraints for the machine type
11723 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
11724 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
11725 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
11726 proper scheduling option for your machine. The default scheduling is
11730 @opindex mlinker-opt
11731 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
11732 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
11733 linkers in which they give bogus error messages when linking some programs.
11736 @opindex msoft-float
11737 Generate output containing library calls for floating point.
11738 @strong{Warning:} the requisite libraries are not available for all HPPA
11739 targets. Normally the facilities of the machine's usual C compiler are
11740 used, but this cannot be done directly in cross-compilation. You must make
11741 your own arrangements to provide suitable library functions for
11744 @option{-msoft-float} changes the calling convention in the output file;
11745 therefore, it is only useful if you compile @emph{all} of a program with
11746 this option. In particular, you need to compile @file{libgcc.a}, the
11747 library that comes with GCC, with @option{-msoft-float} in order for
11752 Generate the predefine, @code{_SIO}, for server IO@. The default is
11753 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
11754 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
11755 options are available under HP-UX and HI-UX@.
11759 Use GNU ld specific options. This passes @option{-shared} to ld when
11760 building a shared library. It is the default when GCC is configured,
11761 explicitly or implicitly, with the GNU linker. This option does not
11762 have any affect on which ld is called, it only changes what parameters
11763 are passed to that ld. The ld that is called is determined by the
11764 @option{--with-ld} configure option, GCC's program search path, and
11765 finally by the user's @env{PATH}. The linker used by GCC can be printed
11766 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
11767 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11771 Use HP ld specific options. This passes @option{-b} to ld when building
11772 a shared library and passes @option{+Accept TypeMismatch} to ld on all
11773 links. It is the default when GCC is configured, explicitly or
11774 implicitly, with the HP linker. This option does not have any affect on
11775 which ld is called, it only changes what parameters are passed to that
11776 ld. The ld that is called is determined by the @option{--with-ld}
11777 configure option, GCC's program search path, and finally by the user's
11778 @env{PATH}. The linker used by GCC can be printed using @samp{which
11779 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
11780 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
11783 @opindex mno-long-calls
11784 Generate code that uses long call sequences. This ensures that a call
11785 is always able to reach linker generated stubs. The default is to generate
11786 long calls only when the distance from the call site to the beginning
11787 of the function or translation unit, as the case may be, exceeds a
11788 predefined limit set by the branch type being used. The limits for
11789 normal calls are 7,600,000 and 240,000 bytes, respectively for the
11790 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
11793 Distances are measured from the beginning of functions when using the
11794 @option{-ffunction-sections} option, or when using the @option{-mgas}
11795 and @option{-mno-portable-runtime} options together under HP-UX with
11798 It is normally not desirable to use this option as it will degrade
11799 performance. However, it may be useful in large applications,
11800 particularly when partial linking is used to build the application.
11802 The types of long calls used depends on the capabilities of the
11803 assembler and linker, and the type of code being generated. The
11804 impact on systems that support long absolute calls, and long pic
11805 symbol-difference or pc-relative calls should be relatively small.
11806 However, an indirect call is used on 32-bit ELF systems in pic code
11807 and it is quite long.
11809 @item -munix=@var{unix-std}
11811 Generate compiler predefines and select a startfile for the specified
11812 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
11813 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
11814 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
11815 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
11816 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
11819 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
11820 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
11821 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
11822 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
11823 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
11824 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
11826 It is @emph{important} to note that this option changes the interfaces
11827 for various library routines. It also affects the operational behavior
11828 of the C library. Thus, @emph{extreme} care is needed in using this
11831 Library code that is intended to operate with more than one UNIX
11832 standard must test, set and restore the variable @var{__xpg4_extended_mask}
11833 as appropriate. Most GNU software doesn't provide this capability.
11837 Suppress the generation of link options to search libdld.sl when the
11838 @option{-static} option is specified on HP-UX 10 and later.
11842 The HP-UX implementation of setlocale in libc has a dependency on
11843 libdld.sl. There isn't an archive version of libdld.sl. Thus,
11844 when the @option{-static} option is specified, special link options
11845 are needed to resolve this dependency.
11847 On HP-UX 10 and later, the GCC driver adds the necessary options to
11848 link with libdld.sl when the @option{-static} option is specified.
11849 This causes the resulting binary to be dynamic. On the 64-bit port,
11850 the linkers generate dynamic binaries by default in any case. The
11851 @option{-nolibdld} option can be used to prevent the GCC driver from
11852 adding these link options.
11856 Add support for multithreading with the @dfn{dce thread} library
11857 under HP-UX@. This option sets flags for both the preprocessor and
11861 @node i386 and x86-64 Options
11862 @subsection Intel 386 and AMD x86-64 Options
11863 @cindex i386 Options
11864 @cindex x86-64 Options
11865 @cindex Intel 386 Options
11866 @cindex AMD x86-64 Options
11868 These @samp{-m} options are defined for the i386 and x86-64 family of
11872 @item -mtune=@var{cpu-type}
11874 Tune to @var{cpu-type} everything applicable about the generated code, except
11875 for the ABI and the set of available instructions. The choices for
11876 @var{cpu-type} are:
11879 Produce code optimized for the most common IA32/AMD64/EM64T processors.
11880 If you know the CPU on which your code will run, then you should use
11881 the corresponding @option{-mtune} option instead of
11882 @option{-mtune=generic}. But, if you do not know exactly what CPU users
11883 of your application will have, then you should use this option.
11885 As new processors are deployed in the marketplace, the behavior of this
11886 option will change. Therefore, if you upgrade to a newer version of
11887 GCC, the code generated option will change to reflect the processors
11888 that were most common when that version of GCC was released.
11890 There is no @option{-march=generic} option because @option{-march}
11891 indicates the instruction set the compiler can use, and there is no
11892 generic instruction set applicable to all processors. In contrast,
11893 @option{-mtune} indicates the processor (or, in this case, collection of
11894 processors) for which the code is optimized.
11896 This selects the CPU to tune for at compilation time by determining
11897 the processor type of the compiling machine. Using @option{-mtune=native}
11898 will produce code optimized for the local machine under the constraints
11899 of the selected instruction set. Using @option{-march=native} will
11900 enable all instruction subsets supported by the local machine (hence
11901 the result might not run on different machines).
11903 Original Intel's i386 CPU@.
11905 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
11906 @item i586, pentium
11907 Intel Pentium CPU with no MMX support.
11909 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
11911 Intel PentiumPro CPU@.
11913 Same as @code{generic}, but when used as @code{march} option, PentiumPro
11914 instruction set will be used, so the code will run on all i686 family chips.
11916 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
11917 @item pentium3, pentium3m
11918 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
11921 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
11922 support. Used by Centrino notebooks.
11923 @item pentium4, pentium4m
11924 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
11926 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
11929 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
11930 SSE2 and SSE3 instruction set support.
11932 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
11933 instruction set support.
11935 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
11936 instruction set support.
11938 AMD K6 CPU with MMX instruction set support.
11940 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
11941 @item athlon, athlon-tbird
11942 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
11944 @item athlon-4, athlon-xp, athlon-mp
11945 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
11946 instruction set support.
11947 @item k8, opteron, athlon64, athlon-fx
11948 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
11949 MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit instruction set extensions.)
11950 @item k8-sse3, opteron-sse3, athlon64-sse3
11951 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
11952 @item amdfam10, barcelona
11953 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
11954 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
11955 instruction set extensions.)
11957 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
11960 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
11961 instruction set support.
11963 Via C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
11964 implemented for this chip.)
11966 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
11967 implemented for this chip.)
11969 Embedded AMD CPU with MMX and 3DNow!@: instruction set support.
11972 While picking a specific @var{cpu-type} will schedule things appropriately
11973 for that particular chip, the compiler will not generate any code that
11974 does not run on the i386 without the @option{-march=@var{cpu-type}} option
11977 @item -march=@var{cpu-type}
11979 Generate instructions for the machine type @var{cpu-type}. The choices
11980 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
11981 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
11983 @item -mcpu=@var{cpu-type}
11985 A deprecated synonym for @option{-mtune}.
11987 @item -mfpmath=@var{unit}
11989 Generate floating point arithmetics for selected unit @var{unit}. The choices
11990 for @var{unit} are:
11994 Use the standard 387 floating point coprocessor present majority of chips and
11995 emulated otherwise. Code compiled with this option will run almost everywhere.
11996 The temporary results are computed in 80bit precision instead of precision
11997 specified by the type resulting in slightly different results compared to most
11998 of other chips. See @option{-ffloat-store} for more detailed description.
12000 This is the default choice for i386 compiler.
12003 Use scalar floating point instructions present in the SSE instruction set.
12004 This instruction set is supported by Pentium3 and newer chips, in the AMD line
12005 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
12006 instruction set supports only single precision arithmetics, thus the double and
12007 extended precision arithmetics is still done using 387. Later version, present
12008 only in Pentium4 and the future AMD x86-64 chips supports double precision
12011 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
12012 or @option{-msse2} switches to enable SSE extensions and make this option
12013 effective. For the x86-64 compiler, these extensions are enabled by default.
12015 The resulting code should be considerably faster in the majority of cases and avoid
12016 the numerical instability problems of 387 code, but may break some existing
12017 code that expects temporaries to be 80bit.
12019 This is the default choice for the x86-64 compiler.
12024 Attempt to utilize both instruction sets at once. This effectively double the
12025 amount of available registers and on chips with separate execution units for
12026 387 and SSE the execution resources too. Use this option with care, as it is
12027 still experimental, because the GCC register allocator does not model separate
12028 functional units well resulting in instable performance.
12031 @item -masm=@var{dialect}
12032 @opindex masm=@var{dialect}
12033 Output asm instructions using selected @var{dialect}. Supported
12034 choices are @samp{intel} or @samp{att} (the default one). Darwin does
12035 not support @samp{intel}.
12038 @itemx -mno-ieee-fp
12040 @opindex mno-ieee-fp
12041 Control whether or not the compiler uses IEEE floating point
12042 comparisons. These handle correctly the case where the result of a
12043 comparison is unordered.
12046 @opindex msoft-float
12047 Generate output containing library calls for floating point.
12048 @strong{Warning:} the requisite libraries are not part of GCC@.
12049 Normally the facilities of the machine's usual C compiler are used, but
12050 this can't be done directly in cross-compilation. You must make your
12051 own arrangements to provide suitable library functions for
12054 On machines where a function returns floating point results in the 80387
12055 register stack, some floating point opcodes may be emitted even if
12056 @option{-msoft-float} is used.
12058 @item -mno-fp-ret-in-387
12059 @opindex mno-fp-ret-in-387
12060 Do not use the FPU registers for return values of functions.
12062 The usual calling convention has functions return values of types
12063 @code{float} and @code{double} in an FPU register, even if there
12064 is no FPU@. The idea is that the operating system should emulate
12067 The option @option{-mno-fp-ret-in-387} causes such values to be returned
12068 in ordinary CPU registers instead.
12070 @item -mno-fancy-math-387
12071 @opindex mno-fancy-math-387
12072 Some 387 emulators do not support the @code{sin}, @code{cos} and
12073 @code{sqrt} instructions for the 387. Specify this option to avoid
12074 generating those instructions. This option is the default on FreeBSD,
12075 OpenBSD and NetBSD@. This option is overridden when @option{-march}
12076 indicates that the target cpu will always have an FPU and so the
12077 instruction will not need emulation. As of revision 2.6.1, these
12078 instructions are not generated unless you also use the
12079 @option{-funsafe-math-optimizations} switch.
12081 @item -malign-double
12082 @itemx -mno-align-double
12083 @opindex malign-double
12084 @opindex mno-align-double
12085 Control whether GCC aligns @code{double}, @code{long double}, and
12086 @code{long long} variables on a two word boundary or a one word
12087 boundary. Aligning @code{double} variables on a two word boundary will
12088 produce code that runs somewhat faster on a @samp{Pentium} at the
12089 expense of more memory.
12091 On x86-64, @option{-malign-double} is enabled by default.
12093 @strong{Warning:} if you use the @option{-malign-double} switch,
12094 structures containing the above types will be aligned differently than
12095 the published application binary interface specifications for the 386
12096 and will not be binary compatible with structures in code compiled
12097 without that switch.
12099 @item -m96bit-long-double
12100 @itemx -m128bit-long-double
12101 @opindex m96bit-long-double
12102 @opindex m128bit-long-double
12103 These switches control the size of @code{long double} type. The i386
12104 application binary interface specifies the size to be 96 bits,
12105 so @option{-m96bit-long-double} is the default in 32 bit mode.
12107 Modern architectures (Pentium and newer) would prefer @code{long double}
12108 to be aligned to an 8 or 16 byte boundary. In arrays or structures
12109 conforming to the ABI, this would not be possible. So specifying a
12110 @option{-m128bit-long-double} will align @code{long double}
12111 to a 16 byte boundary by padding the @code{long double} with an additional
12114 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
12115 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
12117 Notice that neither of these options enable any extra precision over the x87
12118 standard of 80 bits for a @code{long double}.
12120 @strong{Warning:} if you override the default value for your target ABI, the
12121 structures and arrays containing @code{long double} variables will change
12122 their size as well as function calling convention for function taking
12123 @code{long double} will be modified. Hence they will not be binary
12124 compatible with arrays or structures in code compiled without that switch.
12126 @item -mlarge-data-threshold=@var{number}
12127 @opindex mlarge-data-threshold=@var{number}
12128 When @option{-mcmodel=medium} is specified, the data greater than
12129 @var{threshold} are placed in large data section. This value must be the
12130 same across all object linked into the binary and defaults to 65535.
12134 Use a different function-calling convention, in which functions that
12135 take a fixed number of arguments return with the @code{ret} @var{num}
12136 instruction, which pops their arguments while returning. This saves one
12137 instruction in the caller since there is no need to pop the arguments
12140 You can specify that an individual function is called with this calling
12141 sequence with the function attribute @samp{stdcall}. You can also
12142 override the @option{-mrtd} option by using the function attribute
12143 @samp{cdecl}. @xref{Function Attributes}.
12145 @strong{Warning:} this calling convention is incompatible with the one
12146 normally used on Unix, so you cannot use it if you need to call
12147 libraries compiled with the Unix compiler.
12149 Also, you must provide function prototypes for all functions that
12150 take variable numbers of arguments (including @code{printf});
12151 otherwise incorrect code will be generated for calls to those
12154 In addition, seriously incorrect code will result if you call a
12155 function with too many arguments. (Normally, extra arguments are
12156 harmlessly ignored.)
12158 @item -mregparm=@var{num}
12160 Control how many registers are used to pass integer arguments. By
12161 default, no registers are used to pass arguments, and at most 3
12162 registers can be used. You can control this behavior for a specific
12163 function by using the function attribute @samp{regparm}.
12164 @xref{Function Attributes}.
12166 @strong{Warning:} if you use this switch, and
12167 @var{num} is nonzero, then you must build all modules with the same
12168 value, including any libraries. This includes the system libraries and
12172 @opindex msseregparm
12173 Use SSE register passing conventions for float and double arguments
12174 and return values. You can control this behavior for a specific
12175 function by using the function attribute @samp{sseregparm}.
12176 @xref{Function Attributes}.
12178 @strong{Warning:} if you use this switch then you must build all
12179 modules with the same value, including any libraries. This includes
12180 the system libraries and startup modules.
12189 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
12190 is specified, the significands of results of floating-point operations are
12191 rounded to 24 bits (single precision); @option{-mpc64} rounds the
12192 significands of results of floating-point operations to 53 bits (double
12193 precision) and @option{-mpc80} rounds the significands of results of
12194 floating-point operations to 64 bits (extended double precision), which is
12195 the default. When this option is used, floating-point operations in higher
12196 precisions are not available to the programmer without setting the FPU
12197 control word explicitly.
12199 Setting the rounding of floating-point operations to less than the default
12200 80 bits can speed some programs by 2% or more. Note that some mathematical
12201 libraries assume that extended precision (80 bit) floating-point operations
12202 are enabled by default; routines in such libraries could suffer significant
12203 loss of accuracy, typically through so-called "catastrophic cancellation",
12204 when this option is used to set the precision to less than extended precision.
12206 @item -mstackrealign
12207 @opindex mstackrealign
12208 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
12209 option will generate an alternate prologue and epilogue that realigns the
12210 runtime stack if necessary. This supports mixing legacy codes that keep
12211 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
12212 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
12213 applicable to individual functions.
12215 @item -mpreferred-stack-boundary=@var{num}
12216 @opindex mpreferred-stack-boundary
12217 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
12218 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
12219 the default is 4 (16 bytes or 128 bits).
12221 @item -mincoming-stack-boundary=@var{num}
12222 @opindex mincoming-stack-boundary
12223 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
12224 boundary. If @option{-mincoming-stack-boundary} is not specified,
12225 the one specified by @option{-mpreferred-stack-boundary} will be used.
12227 On Pentium and PentiumPro, @code{double} and @code{long double} values
12228 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
12229 suffer significant run time performance penalties. On Pentium III, the
12230 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
12231 properly if it is not 16 byte aligned.
12233 To ensure proper alignment of this values on the stack, the stack boundary
12234 must be as aligned as that required by any value stored on the stack.
12235 Further, every function must be generated such that it keeps the stack
12236 aligned. Thus calling a function compiled with a higher preferred
12237 stack boundary from a function compiled with a lower preferred stack
12238 boundary will most likely misalign the stack. It is recommended that
12239 libraries that use callbacks always use the default setting.
12241 This extra alignment does consume extra stack space, and generally
12242 increases code size. Code that is sensitive to stack space usage, such
12243 as embedded systems and operating system kernels, may want to reduce the
12244 preferred alignment to @option{-mpreferred-stack-boundary=2}.
12269 @itemx -mno-fsgsbase
12294 These switches enable or disable the use of instructions in the MMX,
12295 SSE, SSE2, SSE3, SSSE3, SSE4.1, AVX, AES, PCLMUL, FSGSBASE, RDRND,
12296 F16C, SSE4A, FMA4, XOP, LWP, ABM or 3DNow!@: extended instruction sets.
12297 These extensions are also available as built-in functions: see
12298 @ref{X86 Built-in Functions}, for details of the functions enabled and
12299 disabled by these switches.
12301 To have SSE/SSE2 instructions generated automatically from floating-point
12302 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
12304 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
12305 generates new AVX instructions or AVX equivalence for all SSEx instructions
12308 These options will enable GCC to use these extended instructions in
12309 generated code, even without @option{-mfpmath=sse}. Applications which
12310 perform runtime CPU detection must compile separate files for each
12311 supported architecture, using the appropriate flags. In particular,
12312 the file containing the CPU detection code should be compiled without
12316 @itemx -mno-fused-madd
12317 @opindex mfused-madd
12318 @opindex mno-fused-madd
12319 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
12320 instructions. The default is to use these instructions.
12324 This option instructs GCC to emit a @code{cld} instruction in the prologue
12325 of functions that use string instructions. String instructions depend on
12326 the DF flag to select between autoincrement or autodecrement mode. While the
12327 ABI specifies the DF flag to be cleared on function entry, some operating
12328 systems violate this specification by not clearing the DF flag in their
12329 exception dispatchers. The exception handler can be invoked with the DF flag
12330 set which leads to wrong direction mode, when string instructions are used.
12331 This option can be enabled by default on 32-bit x86 targets by configuring
12332 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
12333 instructions can be suppressed with the @option{-mno-cld} compiler option
12338 This option will enable GCC to use CMPXCHG16B instruction in generated code.
12339 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
12340 data types. This is useful for high resolution counters that could be updated
12341 by multiple processors (or cores). This instruction is generated as part of
12342 atomic built-in functions: see @ref{Atomic Builtins} for details.
12346 This option will enable GCC to use SAHF instruction in generated 64-bit code.
12347 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
12348 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
12349 SAHF are load and store instructions, respectively, for certain status flags.
12350 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
12351 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
12355 This option will enable GCC to use movbe instruction to implement
12356 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
12360 This option will enable built-in functions, @code{__builtin_ia32_crc32qi},
12361 @code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and
12362 @code{__builtin_ia32_crc32di} to generate the crc32 machine instruction.
12366 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
12367 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
12368 to increase precision instead of DIVSS and SQRTSS (and their vectorized
12369 variants) for single precision floating point arguments. These instructions
12370 are generated only when @option{-funsafe-math-optimizations} is enabled
12371 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
12372 Note that while the throughput of the sequence is higher than the throughput
12373 of the non-reciprocal instruction, the precision of the sequence can be
12374 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
12376 Note that GCC implements 1.0f/sqrtf(x) in terms of RSQRTSS (or RSQRTPS)
12377 already with @option{-ffast-math} (or the above option combination), and
12378 doesn't need @option{-mrecip}.
12380 @item -mveclibabi=@var{type}
12381 @opindex mveclibabi
12382 Specifies the ABI type to use for vectorizing intrinsics using an
12383 external library. Supported types are @code{svml} for the Intel short
12384 vector math library and @code{acml} for the AMD math core library style
12385 of interfacing. GCC will currently emit calls to @code{vmldExp2},
12386 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
12387 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
12388 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
12389 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
12390 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
12391 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
12392 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
12393 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
12394 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
12395 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
12396 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
12397 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
12398 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
12399 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
12400 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
12401 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
12402 compatible library will have to be specified at link time.
12404 @item -mabi=@var{name}
12406 Generate code for the specified calling convention. Permissible values
12407 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
12408 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
12409 ABI when targeting Windows. On all other systems, the default is the
12410 SYSV ABI. You can control this behavior for a specific function by
12411 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
12412 @xref{Function Attributes}.
12415 @itemx -mno-push-args
12416 @opindex mpush-args
12417 @opindex mno-push-args
12418 Use PUSH operations to store outgoing parameters. This method is shorter
12419 and usually equally fast as method using SUB/MOV operations and is enabled
12420 by default. In some cases disabling it may improve performance because of
12421 improved scheduling and reduced dependencies.
12423 @item -maccumulate-outgoing-args
12424 @opindex maccumulate-outgoing-args
12425 If enabled, the maximum amount of space required for outgoing arguments will be
12426 computed in the function prologue. This is faster on most modern CPUs
12427 because of reduced dependencies, improved scheduling and reduced stack usage
12428 when preferred stack boundary is not equal to 2. The drawback is a notable
12429 increase in code size. This switch implies @option{-mno-push-args}.
12433 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
12434 on thread-safe exception handling must compile and link all code with the
12435 @option{-mthreads} option. When compiling, @option{-mthreads} defines
12436 @option{-D_MT}; when linking, it links in a special thread helper library
12437 @option{-lmingwthrd} which cleans up per thread exception handling data.
12439 @item -mno-align-stringops
12440 @opindex mno-align-stringops
12441 Do not align destination of inlined string operations. This switch reduces
12442 code size and improves performance in case the destination is already aligned,
12443 but GCC doesn't know about it.
12445 @item -minline-all-stringops
12446 @opindex minline-all-stringops
12447 By default GCC inlines string operations only when destination is known to be
12448 aligned at least to 4 byte boundary. This enables more inlining, increase code
12449 size, but may improve performance of code that depends on fast memcpy, strlen
12450 and memset for short lengths.
12452 @item -minline-stringops-dynamically
12453 @opindex minline-stringops-dynamically
12454 For string operation of unknown size, inline runtime checks so for small
12455 blocks inline code is used, while for large blocks library call is used.
12457 @item -mstringop-strategy=@var{alg}
12458 @opindex mstringop-strategy=@var{alg}
12459 Overwrite internal decision heuristic about particular algorithm to inline
12460 string operation with. The allowed values are @code{rep_byte},
12461 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
12462 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
12463 expanding inline loop, @code{libcall} for always expanding library call.
12465 @item -momit-leaf-frame-pointer
12466 @opindex momit-leaf-frame-pointer
12467 Don't keep the frame pointer in a register for leaf functions. This
12468 avoids the instructions to save, set up and restore frame pointers and
12469 makes an extra register available in leaf functions. The option
12470 @option{-fomit-frame-pointer} removes the frame pointer for all functions
12471 which might make debugging harder.
12473 @item -mtls-direct-seg-refs
12474 @itemx -mno-tls-direct-seg-refs
12475 @opindex mtls-direct-seg-refs
12476 Controls whether TLS variables may be accessed with offsets from the
12477 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
12478 or whether the thread base pointer must be added. Whether or not this
12479 is legal depends on the operating system, and whether it maps the
12480 segment to cover the entire TLS area.
12482 For systems that use GNU libc, the default is on.
12485 @itemx -mno-sse2avx
12487 Specify that the assembler should encode SSE instructions with VEX
12488 prefix. The option @option{-mavx} turns this on by default.
12491 These @samp{-m} switches are supported in addition to the above
12492 on AMD x86-64 processors in 64-bit environments.
12499 Generate code for a 32-bit or 64-bit environment.
12500 The 32-bit environment sets int, long and pointer to 32 bits and
12501 generates code that runs on any i386 system.
12502 The 64-bit environment sets int to 32 bits and long and pointer
12503 to 64 bits and generates code for AMD's x86-64 architecture. For
12504 darwin only the -m64 option turns off the @option{-fno-pic} and
12505 @option{-mdynamic-no-pic} options.
12507 @item -mno-red-zone
12508 @opindex mno-red-zone
12509 Do not use a so called red zone for x86-64 code. The red zone is mandated
12510 by the x86-64 ABI, it is a 128-byte area beyond the location of the
12511 stack pointer that will not be modified by signal or interrupt handlers
12512 and therefore can be used for temporary data without adjusting the stack
12513 pointer. The flag @option{-mno-red-zone} disables this red zone.
12515 @item -mcmodel=small
12516 @opindex mcmodel=small
12517 Generate code for the small code model: the program and its symbols must
12518 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
12519 Programs can be statically or dynamically linked. This is the default
12522 @item -mcmodel=kernel
12523 @opindex mcmodel=kernel
12524 Generate code for the kernel code model. The kernel runs in the
12525 negative 2 GB of the address space.
12526 This model has to be used for Linux kernel code.
12528 @item -mcmodel=medium
12529 @opindex mcmodel=medium
12530 Generate code for the medium model: The program is linked in the lower 2
12531 GB of the address space. Small symbols are also placed there. Symbols
12532 with sizes larger than @option{-mlarge-data-threshold} are put into
12533 large data or bss sections and can be located above 2GB. Programs can
12534 be statically or dynamically linked.
12536 @item -mcmodel=large
12537 @opindex mcmodel=large
12538 Generate code for the large model: This model makes no assumptions
12539 about addresses and sizes of sections.
12542 @node IA-64 Options
12543 @subsection IA-64 Options
12544 @cindex IA-64 Options
12546 These are the @samp{-m} options defined for the Intel IA-64 architecture.
12550 @opindex mbig-endian
12551 Generate code for a big endian target. This is the default for HP-UX@.
12553 @item -mlittle-endian
12554 @opindex mlittle-endian
12555 Generate code for a little endian target. This is the default for AIX5
12561 @opindex mno-gnu-as
12562 Generate (or don't) code for the GNU assembler. This is the default.
12563 @c Also, this is the default if the configure option @option{--with-gnu-as}
12569 @opindex mno-gnu-ld
12570 Generate (or don't) code for the GNU linker. This is the default.
12571 @c Also, this is the default if the configure option @option{--with-gnu-ld}
12576 Generate code that does not use a global pointer register. The result
12577 is not position independent code, and violates the IA-64 ABI@.
12579 @item -mvolatile-asm-stop
12580 @itemx -mno-volatile-asm-stop
12581 @opindex mvolatile-asm-stop
12582 @opindex mno-volatile-asm-stop
12583 Generate (or don't) a stop bit immediately before and after volatile asm
12586 @item -mregister-names
12587 @itemx -mno-register-names
12588 @opindex mregister-names
12589 @opindex mno-register-names
12590 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
12591 the stacked registers. This may make assembler output more readable.
12597 Disable (or enable) optimizations that use the small data section. This may
12598 be useful for working around optimizer bugs.
12600 @item -mconstant-gp
12601 @opindex mconstant-gp
12602 Generate code that uses a single constant global pointer value. This is
12603 useful when compiling kernel code.
12607 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
12608 This is useful when compiling firmware code.
12610 @item -minline-float-divide-min-latency
12611 @opindex minline-float-divide-min-latency
12612 Generate code for inline divides of floating point values
12613 using the minimum latency algorithm.
12615 @item -minline-float-divide-max-throughput
12616 @opindex minline-float-divide-max-throughput
12617 Generate code for inline divides of floating point values
12618 using the maximum throughput algorithm.
12620 @item -mno-inline-float-divide
12621 @opindex mno-inline-float-divide
12622 Do not generate inline code for divides of floating point values.
12624 @item -minline-int-divide-min-latency
12625 @opindex minline-int-divide-min-latency
12626 Generate code for inline divides of integer values
12627 using the minimum latency algorithm.
12629 @item -minline-int-divide-max-throughput
12630 @opindex minline-int-divide-max-throughput
12631 Generate code for inline divides of integer values
12632 using the maximum throughput algorithm.
12634 @item -mno-inline-int-divide
12635 @opindex mno-inline-int-divide
12636 Do not generate inline code for divides of integer values.
12638 @item -minline-sqrt-min-latency
12639 @opindex minline-sqrt-min-latency
12640 Generate code for inline square roots
12641 using the minimum latency algorithm.
12643 @item -minline-sqrt-max-throughput
12644 @opindex minline-sqrt-max-throughput
12645 Generate code for inline square roots
12646 using the maximum throughput algorithm.
12648 @item -mno-inline-sqrt
12649 @opindex mno-inline-sqrt
12650 Do not generate inline code for sqrt.
12653 @itemx -mno-fused-madd
12654 @opindex mfused-madd
12655 @opindex mno-fused-madd
12656 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
12657 instructions. The default is to use these instructions.
12659 @item -mno-dwarf2-asm
12660 @itemx -mdwarf2-asm
12661 @opindex mno-dwarf2-asm
12662 @opindex mdwarf2-asm
12663 Don't (or do) generate assembler code for the DWARF2 line number debugging
12664 info. This may be useful when not using the GNU assembler.
12666 @item -mearly-stop-bits
12667 @itemx -mno-early-stop-bits
12668 @opindex mearly-stop-bits
12669 @opindex mno-early-stop-bits
12670 Allow stop bits to be placed earlier than immediately preceding the
12671 instruction that triggered the stop bit. This can improve instruction
12672 scheduling, but does not always do so.
12674 @item -mfixed-range=@var{register-range}
12675 @opindex mfixed-range
12676 Generate code treating the given register range as fixed registers.
12677 A fixed register is one that the register allocator can not use. This is
12678 useful when compiling kernel code. A register range is specified as
12679 two registers separated by a dash. Multiple register ranges can be
12680 specified separated by a comma.
12682 @item -mtls-size=@var{tls-size}
12684 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
12687 @item -mtune=@var{cpu-type}
12689 Tune the instruction scheduling for a particular CPU, Valid values are
12690 itanium, itanium1, merced, itanium2, and mckinley.
12696 Generate code for a 32-bit or 64-bit environment.
12697 The 32-bit environment sets int, long and pointer to 32 bits.
12698 The 64-bit environment sets int to 32 bits and long and pointer
12699 to 64 bits. These are HP-UX specific flags.
12701 @item -mno-sched-br-data-spec
12702 @itemx -msched-br-data-spec
12703 @opindex mno-sched-br-data-spec
12704 @opindex msched-br-data-spec
12705 (Dis/En)able data speculative scheduling before reload.
12706 This will result in generation of the ld.a instructions and
12707 the corresponding check instructions (ld.c / chk.a).
12708 The default is 'disable'.
12710 @item -msched-ar-data-spec
12711 @itemx -mno-sched-ar-data-spec
12712 @opindex msched-ar-data-spec
12713 @opindex mno-sched-ar-data-spec
12714 (En/Dis)able data speculative scheduling after reload.
12715 This will result in generation of the ld.a instructions and
12716 the corresponding check instructions (ld.c / chk.a).
12717 The default is 'enable'.
12719 @item -mno-sched-control-spec
12720 @itemx -msched-control-spec
12721 @opindex mno-sched-control-spec
12722 @opindex msched-control-spec
12723 (Dis/En)able control speculative scheduling. This feature is
12724 available only during region scheduling (i.e.@: before reload).
12725 This will result in generation of the ld.s instructions and
12726 the corresponding check instructions chk.s .
12727 The default is 'disable'.
12729 @item -msched-br-in-data-spec
12730 @itemx -mno-sched-br-in-data-spec
12731 @opindex msched-br-in-data-spec
12732 @opindex mno-sched-br-in-data-spec
12733 (En/Dis)able speculative scheduling of the instructions that
12734 are dependent on the data speculative loads before reload.
12735 This is effective only with @option{-msched-br-data-spec} enabled.
12736 The default is 'enable'.
12738 @item -msched-ar-in-data-spec
12739 @itemx -mno-sched-ar-in-data-spec
12740 @opindex msched-ar-in-data-spec
12741 @opindex mno-sched-ar-in-data-spec
12742 (En/Dis)able speculative scheduling of the instructions that
12743 are dependent on the data speculative loads after reload.
12744 This is effective only with @option{-msched-ar-data-spec} enabled.
12745 The default is 'enable'.
12747 @item -msched-in-control-spec
12748 @itemx -mno-sched-in-control-spec
12749 @opindex msched-in-control-spec
12750 @opindex mno-sched-in-control-spec
12751 (En/Dis)able speculative scheduling of the instructions that
12752 are dependent on the control speculative loads.
12753 This is effective only with @option{-msched-control-spec} enabled.
12754 The default is 'enable'.
12756 @item -mno-sched-prefer-non-data-spec-insns
12757 @itemx -msched-prefer-non-data-spec-insns
12758 @opindex mno-sched-prefer-non-data-spec-insns
12759 @opindex msched-prefer-non-data-spec-insns
12760 If enabled, data speculative instructions will be chosen for schedule
12761 only if there are no other choices at the moment. This will make
12762 the use of the data speculation much more conservative.
12763 The default is 'disable'.
12765 @item -mno-sched-prefer-non-control-spec-insns
12766 @itemx -msched-prefer-non-control-spec-insns
12767 @opindex mno-sched-prefer-non-control-spec-insns
12768 @opindex msched-prefer-non-control-spec-insns
12769 If enabled, control speculative instructions will be chosen for schedule
12770 only if there are no other choices at the moment. This will make
12771 the use of the control speculation much more conservative.
12772 The default is 'disable'.
12774 @item -mno-sched-count-spec-in-critical-path
12775 @itemx -msched-count-spec-in-critical-path
12776 @opindex mno-sched-count-spec-in-critical-path
12777 @opindex msched-count-spec-in-critical-path
12778 If enabled, speculative dependencies will be considered during
12779 computation of the instructions priorities. This will make the use of the
12780 speculation a bit more conservative.
12781 The default is 'disable'.
12783 @item -msched-spec-ldc
12784 @opindex msched-spec-ldc
12785 Use a simple data speculation check. This option is on by default.
12787 @item -msched-control-spec-ldc
12788 @opindex msched-spec-ldc
12789 Use a simple check for control speculation. This option is on by default.
12791 @item -msched-stop-bits-after-every-cycle
12792 @opindex msched-stop-bits-after-every-cycle
12793 Place a stop bit after every cycle when scheduling. This option is on
12796 @item -msched-fp-mem-deps-zero-cost
12797 @opindex msched-fp-mem-deps-zero-cost
12798 Assume that floating-point stores and loads are not likely to cause a conflict
12799 when placed into the same instruction group. This option is disabled by
12802 @item -msel-sched-dont-check-control-spec
12803 @opindex msel-sched-dont-check-control-spec
12804 Generate checks for control speculation in selective scheduling.
12805 This flag is disabled by default.
12807 @item -msched-max-memory-insns=@var{max-insns}
12808 @opindex msched-max-memory-insns
12809 Limit on the number of memory insns per instruction group, giving lower
12810 priority to subsequent memory insns attempting to schedule in the same
12811 instruction group. Frequently useful to prevent cache bank conflicts.
12812 The default value is 1.
12814 @item -msched-max-memory-insns-hard-limit
12815 @opindex msched-max-memory-insns-hard-limit
12816 Disallow more than `msched-max-memory-insns' in instruction group.
12817 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
12818 when limit is reached but may still schedule memory operations.
12822 @node IA-64/VMS Options
12823 @subsection IA-64/VMS Options
12825 These @samp{-m} options are defined for the IA-64/VMS implementations:
12828 @item -mvms-return-codes
12829 @opindex mvms-return-codes
12830 Return VMS condition codes from main. The default is to return POSIX
12831 style condition (e.g.@ error) codes.
12833 @item -mdebug-main=@var{prefix}
12834 @opindex mdebug-main=@var{prefix}
12835 Flag the first routine whose name starts with @var{prefix} as the main
12836 routine for the debugger.
12840 Default to 64bit memory allocation routines.
12844 @subsection LM32 Options
12845 @cindex LM32 options
12847 These @option{-m} options are defined for the Lattice Mico32 architecture:
12850 @item -mbarrel-shift-enabled
12851 @opindex mbarrel-shift-enabled
12852 Enable barrel-shift instructions.
12854 @item -mdivide-enabled
12855 @opindex mdivide-enabled
12856 Enable divide and modulus instructions.
12858 @item -mmultiply-enabled
12859 @opindex multiply-enabled
12860 Enable multiply instructions.
12862 @item -msign-extend-enabled
12863 @opindex msign-extend-enabled
12864 Enable sign extend instructions.
12866 @item -muser-enabled
12867 @opindex muser-enabled
12868 Enable user-defined instructions.
12873 @subsection M32C Options
12874 @cindex M32C options
12877 @item -mcpu=@var{name}
12879 Select the CPU for which code is generated. @var{name} may be one of
12880 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
12881 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
12882 the M32C/80 series.
12886 Specifies that the program will be run on the simulator. This causes
12887 an alternate runtime library to be linked in which supports, for
12888 example, file I/O@. You must not use this option when generating
12889 programs that will run on real hardware; you must provide your own
12890 runtime library for whatever I/O functions are needed.
12892 @item -memregs=@var{number}
12894 Specifies the number of memory-based pseudo-registers GCC will use
12895 during code generation. These pseudo-registers will be used like real
12896 registers, so there is a tradeoff between GCC's ability to fit the
12897 code into available registers, and the performance penalty of using
12898 memory instead of registers. Note that all modules in a program must
12899 be compiled with the same value for this option. Because of that, you
12900 must not use this option with the default runtime libraries gcc
12905 @node M32R/D Options
12906 @subsection M32R/D Options
12907 @cindex M32R/D options
12909 These @option{-m} options are defined for Renesas M32R/D architectures:
12914 Generate code for the M32R/2@.
12918 Generate code for the M32R/X@.
12922 Generate code for the M32R@. This is the default.
12924 @item -mmodel=small
12925 @opindex mmodel=small
12926 Assume all objects live in the lower 16MB of memory (so that their addresses
12927 can be loaded with the @code{ld24} instruction), and assume all subroutines
12928 are reachable with the @code{bl} instruction.
12929 This is the default.
12931 The addressability of a particular object can be set with the
12932 @code{model} attribute.
12934 @item -mmodel=medium
12935 @opindex mmodel=medium
12936 Assume objects may be anywhere in the 32-bit address space (the compiler
12937 will generate @code{seth/add3} instructions to load their addresses), and
12938 assume all subroutines are reachable with the @code{bl} instruction.
12940 @item -mmodel=large
12941 @opindex mmodel=large
12942 Assume objects may be anywhere in the 32-bit address space (the compiler
12943 will generate @code{seth/add3} instructions to load their addresses), and
12944 assume subroutines may not be reachable with the @code{bl} instruction
12945 (the compiler will generate the much slower @code{seth/add3/jl}
12946 instruction sequence).
12949 @opindex msdata=none
12950 Disable use of the small data area. Variables will be put into
12951 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
12952 @code{section} attribute has been specified).
12953 This is the default.
12955 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
12956 Objects may be explicitly put in the small data area with the
12957 @code{section} attribute using one of these sections.
12959 @item -msdata=sdata
12960 @opindex msdata=sdata
12961 Put small global and static data in the small data area, but do not
12962 generate special code to reference them.
12965 @opindex msdata=use
12966 Put small global and static data in the small data area, and generate
12967 special instructions to reference them.
12971 @cindex smaller data references
12972 Put global and static objects less than or equal to @var{num} bytes
12973 into the small data or bss sections instead of the normal data or bss
12974 sections. The default value of @var{num} is 8.
12975 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
12976 for this option to have any effect.
12978 All modules should be compiled with the same @option{-G @var{num}} value.
12979 Compiling with different values of @var{num} may or may not work; if it
12980 doesn't the linker will give an error message---incorrect code will not be
12985 Makes the M32R specific code in the compiler display some statistics
12986 that might help in debugging programs.
12988 @item -malign-loops
12989 @opindex malign-loops
12990 Align all loops to a 32-byte boundary.
12992 @item -mno-align-loops
12993 @opindex mno-align-loops
12994 Do not enforce a 32-byte alignment for loops. This is the default.
12996 @item -missue-rate=@var{number}
12997 @opindex missue-rate=@var{number}
12998 Issue @var{number} instructions per cycle. @var{number} can only be 1
13001 @item -mbranch-cost=@var{number}
13002 @opindex mbranch-cost=@var{number}
13003 @var{number} can only be 1 or 2. If it is 1 then branches will be
13004 preferred over conditional code, if it is 2, then the opposite will
13007 @item -mflush-trap=@var{number}
13008 @opindex mflush-trap=@var{number}
13009 Specifies the trap number to use to flush the cache. The default is
13010 12. Valid numbers are between 0 and 15 inclusive.
13012 @item -mno-flush-trap
13013 @opindex mno-flush-trap
13014 Specifies that the cache cannot be flushed by using a trap.
13016 @item -mflush-func=@var{name}
13017 @opindex mflush-func=@var{name}
13018 Specifies the name of the operating system function to call to flush
13019 the cache. The default is @emph{_flush_cache}, but a function call
13020 will only be used if a trap is not available.
13022 @item -mno-flush-func
13023 @opindex mno-flush-func
13024 Indicates that there is no OS function for flushing the cache.
13028 @node M680x0 Options
13029 @subsection M680x0 Options
13030 @cindex M680x0 options
13032 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
13033 The default settings depend on which architecture was selected when
13034 the compiler was configured; the defaults for the most common choices
13038 @item -march=@var{arch}
13040 Generate code for a specific M680x0 or ColdFire instruction set
13041 architecture. Permissible values of @var{arch} for M680x0
13042 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
13043 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
13044 architectures are selected according to Freescale's ISA classification
13045 and the permissible values are: @samp{isaa}, @samp{isaaplus},
13046 @samp{isab} and @samp{isac}.
13048 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
13049 code for a ColdFire target. The @var{arch} in this macro is one of the
13050 @option{-march} arguments given above.
13052 When used together, @option{-march} and @option{-mtune} select code
13053 that runs on a family of similar processors but that is optimized
13054 for a particular microarchitecture.
13056 @item -mcpu=@var{cpu}
13058 Generate code for a specific M680x0 or ColdFire processor.
13059 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
13060 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
13061 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
13062 below, which also classifies the CPUs into families:
13064 @multitable @columnfractions 0.20 0.80
13065 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
13066 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
13067 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
13068 @item @samp{5206e} @tab @samp{5206e}
13069 @item @samp{5208} @tab @samp{5207} @samp{5208}
13070 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
13071 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
13072 @item @samp{5216} @tab @samp{5214} @samp{5216}
13073 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
13074 @item @samp{5225} @tab @samp{5224} @samp{5225}
13075 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
13076 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
13077 @item @samp{5249} @tab @samp{5249}
13078 @item @samp{5250} @tab @samp{5250}
13079 @item @samp{5271} @tab @samp{5270} @samp{5271}
13080 @item @samp{5272} @tab @samp{5272}
13081 @item @samp{5275} @tab @samp{5274} @samp{5275}
13082 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
13083 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
13084 @item @samp{5307} @tab @samp{5307}
13085 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
13086 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
13087 @item @samp{5407} @tab @samp{5407}
13088 @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}
13091 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
13092 @var{arch} is compatible with @var{cpu}. Other combinations of
13093 @option{-mcpu} and @option{-march} are rejected.
13095 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
13096 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
13097 where the value of @var{family} is given by the table above.
13099 @item -mtune=@var{tune}
13101 Tune the code for a particular microarchitecture, within the
13102 constraints set by @option{-march} and @option{-mcpu}.
13103 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
13104 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
13105 and @samp{cpu32}. The ColdFire microarchitectures
13106 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
13108 You can also use @option{-mtune=68020-40} for code that needs
13109 to run relatively well on 68020, 68030 and 68040 targets.
13110 @option{-mtune=68020-60} is similar but includes 68060 targets
13111 as well. These two options select the same tuning decisions as
13112 @option{-m68020-40} and @option{-m68020-60} respectively.
13114 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
13115 when tuning for 680x0 architecture @var{arch}. It also defines
13116 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
13117 option is used. If gcc is tuning for a range of architectures,
13118 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
13119 it defines the macros for every architecture in the range.
13121 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
13122 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
13123 of the arguments given above.
13129 Generate output for a 68000. This is the default
13130 when the compiler is configured for 68000-based systems.
13131 It is equivalent to @option{-march=68000}.
13133 Use this option for microcontrollers with a 68000 or EC000 core,
13134 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
13138 Generate output for a 68010. This is the default
13139 when the compiler is configured for 68010-based systems.
13140 It is equivalent to @option{-march=68010}.
13146 Generate output for a 68020. This is the default
13147 when the compiler is configured for 68020-based systems.
13148 It is equivalent to @option{-march=68020}.
13152 Generate output for a 68030. This is the default when the compiler is
13153 configured for 68030-based systems. It is equivalent to
13154 @option{-march=68030}.
13158 Generate output for a 68040. This is the default when the compiler is
13159 configured for 68040-based systems. It is equivalent to
13160 @option{-march=68040}.
13162 This option inhibits the use of 68881/68882 instructions that have to be
13163 emulated by software on the 68040. Use this option if your 68040 does not
13164 have code to emulate those instructions.
13168 Generate output for a 68060. This is the default when the compiler is
13169 configured for 68060-based systems. It is equivalent to
13170 @option{-march=68060}.
13172 This option inhibits the use of 68020 and 68881/68882 instructions that
13173 have to be emulated by software on the 68060. Use this option if your 68060
13174 does not have code to emulate those instructions.
13178 Generate output for a CPU32. This is the default
13179 when the compiler is configured for CPU32-based systems.
13180 It is equivalent to @option{-march=cpu32}.
13182 Use this option for microcontrollers with a
13183 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
13184 68336, 68340, 68341, 68349 and 68360.
13188 Generate output for a 520X ColdFire CPU@. This is the default
13189 when the compiler is configured for 520X-based systems.
13190 It is equivalent to @option{-mcpu=5206}, and is now deprecated
13191 in favor of that option.
13193 Use this option for microcontroller with a 5200 core, including
13194 the MCF5202, MCF5203, MCF5204 and MCF5206.
13198 Generate output for a 5206e ColdFire CPU@. The option is now
13199 deprecated in favor of the equivalent @option{-mcpu=5206e}.
13203 Generate output for a member of the ColdFire 528X family.
13204 The option is now deprecated in favor of the equivalent
13205 @option{-mcpu=528x}.
13209 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
13210 in favor of the equivalent @option{-mcpu=5307}.
13214 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
13215 in favor of the equivalent @option{-mcpu=5407}.
13219 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
13220 This includes use of hardware floating point instructions.
13221 The option is equivalent to @option{-mcpu=547x}, and is now
13222 deprecated in favor of that option.
13226 Generate output for a 68040, without using any of the new instructions.
13227 This results in code which can run relatively efficiently on either a
13228 68020/68881 or a 68030 or a 68040. The generated code does use the
13229 68881 instructions that are emulated on the 68040.
13231 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
13235 Generate output for a 68060, without using any of the new instructions.
13236 This results in code which can run relatively efficiently on either a
13237 68020/68881 or a 68030 or a 68040. The generated code does use the
13238 68881 instructions that are emulated on the 68060.
13240 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
13244 @opindex mhard-float
13246 Generate floating-point instructions. This is the default for 68020
13247 and above, and for ColdFire devices that have an FPU@. It defines the
13248 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
13249 on ColdFire targets.
13252 @opindex msoft-float
13253 Do not generate floating-point instructions; use library calls instead.
13254 This is the default for 68000, 68010, and 68832 targets. It is also
13255 the default for ColdFire devices that have no FPU.
13261 Generate (do not generate) ColdFire hardware divide and remainder
13262 instructions. If @option{-march} is used without @option{-mcpu},
13263 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
13264 architectures. Otherwise, the default is taken from the target CPU
13265 (either the default CPU, or the one specified by @option{-mcpu}). For
13266 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
13267 @option{-mcpu=5206e}.
13269 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
13273 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13274 Additionally, parameters passed on the stack are also aligned to a
13275 16-bit boundary even on targets whose API mandates promotion to 32-bit.
13279 Do not consider type @code{int} to be 16 bits wide. This is the default.
13282 @itemx -mno-bitfield
13283 @opindex mnobitfield
13284 @opindex mno-bitfield
13285 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
13286 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
13290 Do use the bit-field instructions. The @option{-m68020} option implies
13291 @option{-mbitfield}. This is the default if you use a configuration
13292 designed for a 68020.
13296 Use a different function-calling convention, in which functions
13297 that take a fixed number of arguments return with the @code{rtd}
13298 instruction, which pops their arguments while returning. This
13299 saves one instruction in the caller since there is no need to pop
13300 the arguments there.
13302 This calling convention is incompatible with the one normally
13303 used on Unix, so you cannot use it if you need to call libraries
13304 compiled with the Unix compiler.
13306 Also, you must provide function prototypes for all functions that
13307 take variable numbers of arguments (including @code{printf});
13308 otherwise incorrect code will be generated for calls to those
13311 In addition, seriously incorrect code will result if you call a
13312 function with too many arguments. (Normally, extra arguments are
13313 harmlessly ignored.)
13315 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
13316 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
13320 Do not use the calling conventions selected by @option{-mrtd}.
13321 This is the default.
13324 @itemx -mno-align-int
13325 @opindex malign-int
13326 @opindex mno-align-int
13327 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
13328 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
13329 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
13330 Aligning variables on 32-bit boundaries produces code that runs somewhat
13331 faster on processors with 32-bit busses at the expense of more memory.
13333 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
13334 align structures containing the above types differently than
13335 most published application binary interface specifications for the m68k.
13339 Use the pc-relative addressing mode of the 68000 directly, instead of
13340 using a global offset table. At present, this option implies @option{-fpic},
13341 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
13342 not presently supported with @option{-mpcrel}, though this could be supported for
13343 68020 and higher processors.
13345 @item -mno-strict-align
13346 @itemx -mstrict-align
13347 @opindex mno-strict-align
13348 @opindex mstrict-align
13349 Do not (do) assume that unaligned memory references will be handled by
13353 Generate code that allows the data segment to be located in a different
13354 area of memory from the text segment. This allows for execute in place in
13355 an environment without virtual memory management. This option implies
13358 @item -mno-sep-data
13359 Generate code that assumes that the data segment follows the text segment.
13360 This is the default.
13362 @item -mid-shared-library
13363 Generate code that supports shared libraries via the library ID method.
13364 This allows for execute in place and shared libraries in an environment
13365 without virtual memory management. This option implies @option{-fPIC}.
13367 @item -mno-id-shared-library
13368 Generate code that doesn't assume ID based shared libraries are being used.
13369 This is the default.
13371 @item -mshared-library-id=n
13372 Specified the identification number of the ID based shared library being
13373 compiled. Specifying a value of 0 will generate more compact code, specifying
13374 other values will force the allocation of that number to the current
13375 library but is no more space or time efficient than omitting this option.
13381 When generating position-independent code for ColdFire, generate code
13382 that works if the GOT has more than 8192 entries. This code is
13383 larger and slower than code generated without this option. On M680x0
13384 processors, this option is not needed; @option{-fPIC} suffices.
13386 GCC normally uses a single instruction to load values from the GOT@.
13387 While this is relatively efficient, it only works if the GOT
13388 is smaller than about 64k. Anything larger causes the linker
13389 to report an error such as:
13391 @cindex relocation truncated to fit (ColdFire)
13393 relocation truncated to fit: R_68K_GOT16O foobar
13396 If this happens, you should recompile your code with @option{-mxgot}.
13397 It should then work with very large GOTs. However, code generated with
13398 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
13399 the value of a global symbol.
13401 Note that some linkers, including newer versions of the GNU linker,
13402 can create multiple GOTs and sort GOT entries. If you have such a linker,
13403 you should only need to use @option{-mxgot} when compiling a single
13404 object file that accesses more than 8192 GOT entries. Very few do.
13406 These options have no effect unless GCC is generating
13407 position-independent code.
13411 @node M68hc1x Options
13412 @subsection M68hc1x Options
13413 @cindex M68hc1x options
13415 These are the @samp{-m} options defined for the 68hc11 and 68hc12
13416 microcontrollers. The default values for these options depends on
13417 which style of microcontroller was selected when the compiler was configured;
13418 the defaults for the most common choices are given below.
13425 Generate output for a 68HC11. This is the default
13426 when the compiler is configured for 68HC11-based systems.
13432 Generate output for a 68HC12. This is the default
13433 when the compiler is configured for 68HC12-based systems.
13439 Generate output for a 68HCS12.
13441 @item -mauto-incdec
13442 @opindex mauto-incdec
13443 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
13450 Enable the use of 68HC12 min and max instructions.
13453 @itemx -mno-long-calls
13454 @opindex mlong-calls
13455 @opindex mno-long-calls
13456 Treat all calls as being far away (near). If calls are assumed to be
13457 far away, the compiler will use the @code{call} instruction to
13458 call a function and the @code{rtc} instruction for returning.
13462 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13464 @item -msoft-reg-count=@var{count}
13465 @opindex msoft-reg-count
13466 Specify the number of pseudo-soft registers which are used for the
13467 code generation. The maximum number is 32. Using more pseudo-soft
13468 register may or may not result in better code depending on the program.
13469 The default is 4 for 68HC11 and 2 for 68HC12.
13473 @node MCore Options
13474 @subsection MCore Options
13475 @cindex MCore options
13477 These are the @samp{-m} options defined for the Motorola M*Core
13483 @itemx -mno-hardlit
13485 @opindex mno-hardlit
13486 Inline constants into the code stream if it can be done in two
13487 instructions or less.
13493 Use the divide instruction. (Enabled by default).
13495 @item -mrelax-immediate
13496 @itemx -mno-relax-immediate
13497 @opindex mrelax-immediate
13498 @opindex mno-relax-immediate
13499 Allow arbitrary sized immediates in bit operations.
13501 @item -mwide-bitfields
13502 @itemx -mno-wide-bitfields
13503 @opindex mwide-bitfields
13504 @opindex mno-wide-bitfields
13505 Always treat bit-fields as int-sized.
13507 @item -m4byte-functions
13508 @itemx -mno-4byte-functions
13509 @opindex m4byte-functions
13510 @opindex mno-4byte-functions
13511 Force all functions to be aligned to a four byte boundary.
13513 @item -mcallgraph-data
13514 @itemx -mno-callgraph-data
13515 @opindex mcallgraph-data
13516 @opindex mno-callgraph-data
13517 Emit callgraph information.
13520 @itemx -mno-slow-bytes
13521 @opindex mslow-bytes
13522 @opindex mno-slow-bytes
13523 Prefer word access when reading byte quantities.
13525 @item -mlittle-endian
13526 @itemx -mbig-endian
13527 @opindex mlittle-endian
13528 @opindex mbig-endian
13529 Generate code for a little endian target.
13535 Generate code for the 210 processor.
13539 Assume that run-time support has been provided and so omit the
13540 simulator library (@file{libsim.a)} from the linker command line.
13542 @item -mstack-increment=@var{size}
13543 @opindex mstack-increment
13544 Set the maximum amount for a single stack increment operation. Large
13545 values can increase the speed of programs which contain functions
13546 that need a large amount of stack space, but they can also trigger a
13547 segmentation fault if the stack is extended too much. The default
13553 @subsection MeP Options
13554 @cindex MeP options
13560 Enables the @code{abs} instruction, which is the absolute difference
13561 between two registers.
13565 Enables all the optional instructions - average, multiply, divide, bit
13566 operations, leading zero, absolute difference, min/max, clip, and
13572 Enables the @code{ave} instruction, which computes the average of two
13575 @item -mbased=@var{n}
13577 Variables of size @var{n} bytes or smaller will be placed in the
13578 @code{.based} section by default. Based variables use the @code{$tp}
13579 register as a base register, and there is a 128 byte limit to the
13580 @code{.based} section.
13584 Enables the bit operation instructions - bit test (@code{btstm}), set
13585 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
13586 test-and-set (@code{tas}).
13588 @item -mc=@var{name}
13590 Selects which section constant data will be placed in. @var{name} may
13591 be @code{tiny}, @code{near}, or @code{far}.
13595 Enables the @code{clip} instruction. Note that @code{-mclip} is not
13596 useful unless you also provide @code{-mminmax}.
13598 @item -mconfig=@var{name}
13600 Selects one of the build-in core configurations. Each MeP chip has
13601 one or more modules in it; each module has a core CPU and a variety of
13602 coprocessors, optional instructions, and peripherals. The
13603 @code{MeP-Integrator} tool, not part of GCC, provides these
13604 configurations through this option; using this option is the same as
13605 using all the corresponding command line options. The default
13606 configuration is @code{default}.
13610 Enables the coprocessor instructions. By default, this is a 32-bit
13611 coprocessor. Note that the coprocessor is normally enabled via the
13612 @code{-mconfig=} option.
13616 Enables the 32-bit coprocessor's instructions.
13620 Enables the 64-bit coprocessor's instructions.
13624 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
13628 Causes constant variables to be placed in the @code{.near} section.
13632 Enables the @code{div} and @code{divu} instructions.
13636 Generate big-endian code.
13640 Generate little-endian code.
13642 @item -mio-volatile
13643 @opindex mio-volatile
13644 Tells the compiler that any variable marked with the @code{io}
13645 attribute is to be considered volatile.
13649 Causes variables to be assigned to the @code{.far} section by default.
13653 Enables the @code{leadz} (leading zero) instruction.
13657 Causes variables to be assigned to the @code{.near} section by default.
13661 Enables the @code{min} and @code{max} instructions.
13665 Enables the multiplication and multiply-accumulate instructions.
13669 Disables all the optional instructions enabled by @code{-mall-opts}.
13673 Enables the @code{repeat} and @code{erepeat} instructions, used for
13674 low-overhead looping.
13678 Causes all variables to default to the @code{.tiny} section. Note
13679 that there is a 65536 byte limit to this section. Accesses to these
13680 variables use the @code{%gp} base register.
13684 Enables the saturation instructions. Note that the compiler does not
13685 currently generate these itself, but this option is included for
13686 compatibility with other tools, like @code{as}.
13690 Link the SDRAM-based runtime instead of the default ROM-based runtime.
13694 Link the simulator runtime libraries.
13698 Link the simulator runtime libraries, excluding built-in support
13699 for reset and exception vectors and tables.
13703 Causes all functions to default to the @code{.far} section. Without
13704 this option, functions default to the @code{.near} section.
13706 @item -mtiny=@var{n}
13708 Variables that are @var{n} bytes or smaller will be allocated to the
13709 @code{.tiny} section. These variables use the @code{$gp} base
13710 register. The default for this option is 4, but note that there's a
13711 65536 byte limit to the @code{.tiny} section.
13716 @subsection MIPS Options
13717 @cindex MIPS options
13723 Generate big-endian code.
13727 Generate little-endian code. This is the default for @samp{mips*el-*-*}
13730 @item -march=@var{arch}
13732 Generate code that will run on @var{arch}, which can be the name of a
13733 generic MIPS ISA, or the name of a particular processor.
13735 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
13736 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
13737 The processor names are:
13738 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
13739 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
13740 @samp{5kc}, @samp{5kf},
13742 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
13743 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
13744 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
13745 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
13746 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
13747 @samp{loongson2e}, @samp{loongson2f},
13751 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
13752 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
13753 @samp{rm7000}, @samp{rm9000},
13754 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
13757 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
13758 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
13760 The special value @samp{from-abi} selects the
13761 most compatible architecture for the selected ABI (that is,
13762 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
13764 Native Linux/GNU toolchains also support the value @samp{native},
13765 which selects the best architecture option for the host processor.
13766 @option{-march=native} has no effect if GCC does not recognize
13769 In processor names, a final @samp{000} can be abbreviated as @samp{k}
13770 (for example, @samp{-march=r2k}). Prefixes are optional, and
13771 @samp{vr} may be written @samp{r}.
13773 Names of the form @samp{@var{n}f2_1} refer to processors with
13774 FPUs clocked at half the rate of the core, names of the form
13775 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
13776 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
13777 processors with FPUs clocked a ratio of 3:2 with respect to the core.
13778 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
13779 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
13780 accepted as synonyms for @samp{@var{n}f1_1}.
13782 GCC defines two macros based on the value of this option. The first
13783 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
13784 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
13785 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
13786 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
13787 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
13789 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
13790 above. In other words, it will have the full prefix and will not
13791 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
13792 the macro names the resolved architecture (either @samp{"mips1"} or
13793 @samp{"mips3"}). It names the default architecture when no
13794 @option{-march} option is given.
13796 @item -mtune=@var{arch}
13798 Optimize for @var{arch}. Among other things, this option controls
13799 the way instructions are scheduled, and the perceived cost of arithmetic
13800 operations. The list of @var{arch} values is the same as for
13803 When this option is not used, GCC will optimize for the processor
13804 specified by @option{-march}. By using @option{-march} and
13805 @option{-mtune} together, it is possible to generate code that will
13806 run on a family of processors, but optimize the code for one
13807 particular member of that family.
13809 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
13810 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
13811 @samp{-march} ones described above.
13815 Equivalent to @samp{-march=mips1}.
13819 Equivalent to @samp{-march=mips2}.
13823 Equivalent to @samp{-march=mips3}.
13827 Equivalent to @samp{-march=mips4}.
13831 Equivalent to @samp{-march=mips32}.
13835 Equivalent to @samp{-march=mips32r2}.
13839 Equivalent to @samp{-march=mips64}.
13843 Equivalent to @samp{-march=mips64r2}.
13848 @opindex mno-mips16
13849 Generate (do not generate) MIPS16 code. If GCC is targetting a
13850 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
13852 MIPS16 code generation can also be controlled on a per-function basis
13853 by means of @code{mips16} and @code{nomips16} attributes.
13854 @xref{Function Attributes}, for more information.
13856 @item -mflip-mips16
13857 @opindex mflip-mips16
13858 Generate MIPS16 code on alternating functions. This option is provided
13859 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
13860 not intended for ordinary use in compiling user code.
13862 @item -minterlink-mips16
13863 @itemx -mno-interlink-mips16
13864 @opindex minterlink-mips16
13865 @opindex mno-interlink-mips16
13866 Require (do not require) that non-MIPS16 code be link-compatible with
13869 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
13870 it must either use a call or an indirect jump. @option{-minterlink-mips16}
13871 therefore disables direct jumps unless GCC knows that the target of the
13872 jump is not MIPS16.
13884 Generate code for the given ABI@.
13886 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
13887 generates 64-bit code when you select a 64-bit architecture, but you
13888 can use @option{-mgp32} to get 32-bit code instead.
13890 For information about the O64 ABI, see
13891 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
13893 GCC supports a variant of the o32 ABI in which floating-point registers
13894 are 64 rather than 32 bits wide. You can select this combination with
13895 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
13896 and @samp{mfhc1} instructions and is therefore only supported for
13897 MIPS32R2 processors.
13899 The register assignments for arguments and return values remain the
13900 same, but each scalar value is passed in a single 64-bit register
13901 rather than a pair of 32-bit registers. For example, scalar
13902 floating-point values are returned in @samp{$f0} only, not a
13903 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
13904 remains the same, but all 64 bits are saved.
13907 @itemx -mno-abicalls
13909 @opindex mno-abicalls
13910 Generate (do not generate) code that is suitable for SVR4-style
13911 dynamic objects. @option{-mabicalls} is the default for SVR4-based
13916 Generate (do not generate) code that is fully position-independent,
13917 and that can therefore be linked into shared libraries. This option
13918 only affects @option{-mabicalls}.
13920 All @option{-mabicalls} code has traditionally been position-independent,
13921 regardless of options like @option{-fPIC} and @option{-fpic}. However,
13922 as an extension, the GNU toolchain allows executables to use absolute
13923 accesses for locally-binding symbols. It can also use shorter GP
13924 initialization sequences and generate direct calls to locally-defined
13925 functions. This mode is selected by @option{-mno-shared}.
13927 @option{-mno-shared} depends on binutils 2.16 or higher and generates
13928 objects that can only be linked by the GNU linker. However, the option
13929 does not affect the ABI of the final executable; it only affects the ABI
13930 of relocatable objects. Using @option{-mno-shared} will generally make
13931 executables both smaller and quicker.
13933 @option{-mshared} is the default.
13939 Assume (do not assume) that the static and dynamic linkers
13940 support PLTs and copy relocations. This option only affects
13941 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
13942 has no effect without @samp{-msym32}.
13944 You can make @option{-mplt} the default by configuring
13945 GCC with @option{--with-mips-plt}. The default is
13946 @option{-mno-plt} otherwise.
13952 Lift (do not lift) the usual restrictions on the size of the global
13955 GCC normally uses a single instruction to load values from the GOT@.
13956 While this is relatively efficient, it will only work if the GOT
13957 is smaller than about 64k. Anything larger will cause the linker
13958 to report an error such as:
13960 @cindex relocation truncated to fit (MIPS)
13962 relocation truncated to fit: R_MIPS_GOT16 foobar
13965 If this happens, you should recompile your code with @option{-mxgot}.
13966 It should then work with very large GOTs, although it will also be
13967 less efficient, since it will take three instructions to fetch the
13968 value of a global symbol.
13970 Note that some linkers can create multiple GOTs. If you have such a
13971 linker, you should only need to use @option{-mxgot} when a single object
13972 file accesses more than 64k's worth of GOT entries. Very few do.
13974 These options have no effect unless GCC is generating position
13979 Assume that general-purpose registers are 32 bits wide.
13983 Assume that general-purpose registers are 64 bits wide.
13987 Assume that floating-point registers are 32 bits wide.
13991 Assume that floating-point registers are 64 bits wide.
13994 @opindex mhard-float
13995 Use floating-point coprocessor instructions.
13998 @opindex msoft-float
13999 Do not use floating-point coprocessor instructions. Implement
14000 floating-point calculations using library calls instead.
14002 @item -msingle-float
14003 @opindex msingle-float
14004 Assume that the floating-point coprocessor only supports single-precision
14007 @item -mdouble-float
14008 @opindex mdouble-float
14009 Assume that the floating-point coprocessor supports double-precision
14010 operations. This is the default.
14016 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
14017 implement atomic memory built-in functions. When neither option is
14018 specified, GCC will use the instructions if the target architecture
14021 @option{-mllsc} is useful if the runtime environment can emulate the
14022 instructions and @option{-mno-llsc} can be useful when compiling for
14023 nonstandard ISAs. You can make either option the default by
14024 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
14025 respectively. @option{--with-llsc} is the default for some
14026 configurations; see the installation documentation for details.
14032 Use (do not use) revision 1 of the MIPS DSP ASE@.
14033 @xref{MIPS DSP Built-in Functions}. This option defines the
14034 preprocessor macro @samp{__mips_dsp}. It also defines
14035 @samp{__mips_dsp_rev} to 1.
14041 Use (do not use) revision 2 of the MIPS DSP ASE@.
14042 @xref{MIPS DSP Built-in Functions}. This option defines the
14043 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
14044 It also defines @samp{__mips_dsp_rev} to 2.
14047 @itemx -mno-smartmips
14048 @opindex msmartmips
14049 @opindex mno-smartmips
14050 Use (do not use) the MIPS SmartMIPS ASE.
14052 @item -mpaired-single
14053 @itemx -mno-paired-single
14054 @opindex mpaired-single
14055 @opindex mno-paired-single
14056 Use (do not use) paired-single floating-point instructions.
14057 @xref{MIPS Paired-Single Support}. This option requires
14058 hardware floating-point support to be enabled.
14064 Use (do not use) MIPS Digital Media Extension instructions.
14065 This option can only be used when generating 64-bit code and requires
14066 hardware floating-point support to be enabled.
14071 @opindex mno-mips3d
14072 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
14073 The option @option{-mips3d} implies @option{-mpaired-single}.
14079 Use (do not use) MT Multithreading instructions.
14083 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
14084 an explanation of the default and the way that the pointer size is
14089 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
14091 The default size of @code{int}s, @code{long}s and pointers depends on
14092 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
14093 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
14094 32-bit @code{long}s. Pointers are the same size as @code{long}s,
14095 or the same size as integer registers, whichever is smaller.
14101 Assume (do not assume) that all symbols have 32-bit values, regardless
14102 of the selected ABI@. This option is useful in combination with
14103 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
14104 to generate shorter and faster references to symbolic addresses.
14108 Put definitions of externally-visible data in a small data section
14109 if that data is no bigger than @var{num} bytes. GCC can then access
14110 the data more efficiently; see @option{-mgpopt} for details.
14112 The default @option{-G} option depends on the configuration.
14114 @item -mlocal-sdata
14115 @itemx -mno-local-sdata
14116 @opindex mlocal-sdata
14117 @opindex mno-local-sdata
14118 Extend (do not extend) the @option{-G} behavior to local data too,
14119 such as to static variables in C@. @option{-mlocal-sdata} is the
14120 default for all configurations.
14122 If the linker complains that an application is using too much small data,
14123 you might want to try rebuilding the less performance-critical parts with
14124 @option{-mno-local-sdata}. You might also want to build large
14125 libraries with @option{-mno-local-sdata}, so that the libraries leave
14126 more room for the main program.
14128 @item -mextern-sdata
14129 @itemx -mno-extern-sdata
14130 @opindex mextern-sdata
14131 @opindex mno-extern-sdata
14132 Assume (do not assume) that externally-defined data will be in
14133 a small data section if that data is within the @option{-G} limit.
14134 @option{-mextern-sdata} is the default for all configurations.
14136 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
14137 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
14138 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
14139 is placed in a small data section. If @var{Var} is defined by another
14140 module, you must either compile that module with a high-enough
14141 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
14142 definition. If @var{Var} is common, you must link the application
14143 with a high-enough @option{-G} setting.
14145 The easiest way of satisfying these restrictions is to compile
14146 and link every module with the same @option{-G} option. However,
14147 you may wish to build a library that supports several different
14148 small data limits. You can do this by compiling the library with
14149 the highest supported @option{-G} setting and additionally using
14150 @option{-mno-extern-sdata} to stop the library from making assumptions
14151 about externally-defined data.
14157 Use (do not use) GP-relative accesses for symbols that are known to be
14158 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
14159 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
14162 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
14163 might not hold the value of @code{_gp}. For example, if the code is
14164 part of a library that might be used in a boot monitor, programs that
14165 call boot monitor routines will pass an unknown value in @code{$gp}.
14166 (In such situations, the boot monitor itself would usually be compiled
14167 with @option{-G0}.)
14169 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
14170 @option{-mno-extern-sdata}.
14172 @item -membedded-data
14173 @itemx -mno-embedded-data
14174 @opindex membedded-data
14175 @opindex mno-embedded-data
14176 Allocate variables to the read-only data section first if possible, then
14177 next in the small data section if possible, otherwise in data. This gives
14178 slightly slower code than the default, but reduces the amount of RAM required
14179 when executing, and thus may be preferred for some embedded systems.
14181 @item -muninit-const-in-rodata
14182 @itemx -mno-uninit-const-in-rodata
14183 @opindex muninit-const-in-rodata
14184 @opindex mno-uninit-const-in-rodata
14185 Put uninitialized @code{const} variables in the read-only data section.
14186 This option is only meaningful in conjunction with @option{-membedded-data}.
14188 @item -mcode-readable=@var{setting}
14189 @opindex mcode-readable
14190 Specify whether GCC may generate code that reads from executable sections.
14191 There are three possible settings:
14194 @item -mcode-readable=yes
14195 Instructions may freely access executable sections. This is the
14198 @item -mcode-readable=pcrel
14199 MIPS16 PC-relative load instructions can access executable sections,
14200 but other instructions must not do so. This option is useful on 4KSc
14201 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
14202 It is also useful on processors that can be configured to have a dual
14203 instruction/data SRAM interface and that, like the M4K, automatically
14204 redirect PC-relative loads to the instruction RAM.
14206 @item -mcode-readable=no
14207 Instructions must not access executable sections. This option can be
14208 useful on targets that are configured to have a dual instruction/data
14209 SRAM interface but that (unlike the M4K) do not automatically redirect
14210 PC-relative loads to the instruction RAM.
14213 @item -msplit-addresses
14214 @itemx -mno-split-addresses
14215 @opindex msplit-addresses
14216 @opindex mno-split-addresses
14217 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
14218 relocation operators. This option has been superseded by
14219 @option{-mexplicit-relocs} but is retained for backwards compatibility.
14221 @item -mexplicit-relocs
14222 @itemx -mno-explicit-relocs
14223 @opindex mexplicit-relocs
14224 @opindex mno-explicit-relocs
14225 Use (do not use) assembler relocation operators when dealing with symbolic
14226 addresses. The alternative, selected by @option{-mno-explicit-relocs},
14227 is to use assembler macros instead.
14229 @option{-mexplicit-relocs} is the default if GCC was configured
14230 to use an assembler that supports relocation operators.
14232 @item -mcheck-zero-division
14233 @itemx -mno-check-zero-division
14234 @opindex mcheck-zero-division
14235 @opindex mno-check-zero-division
14236 Trap (do not trap) on integer division by zero.
14238 The default is @option{-mcheck-zero-division}.
14240 @item -mdivide-traps
14241 @itemx -mdivide-breaks
14242 @opindex mdivide-traps
14243 @opindex mdivide-breaks
14244 MIPS systems check for division by zero by generating either a
14245 conditional trap or a break instruction. Using traps results in
14246 smaller code, but is only supported on MIPS II and later. Also, some
14247 versions of the Linux kernel have a bug that prevents trap from
14248 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
14249 allow conditional traps on architectures that support them and
14250 @option{-mdivide-breaks} to force the use of breaks.
14252 The default is usually @option{-mdivide-traps}, but this can be
14253 overridden at configure time using @option{--with-divide=breaks}.
14254 Divide-by-zero checks can be completely disabled using
14255 @option{-mno-check-zero-division}.
14260 @opindex mno-memcpy
14261 Force (do not force) the use of @code{memcpy()} for non-trivial block
14262 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
14263 most constant-sized copies.
14266 @itemx -mno-long-calls
14267 @opindex mlong-calls
14268 @opindex mno-long-calls
14269 Disable (do not disable) use of the @code{jal} instruction. Calling
14270 functions using @code{jal} is more efficient but requires the caller
14271 and callee to be in the same 256 megabyte segment.
14273 This option has no effect on abicalls code. The default is
14274 @option{-mno-long-calls}.
14280 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
14281 instructions, as provided by the R4650 ISA@.
14284 @itemx -mno-fused-madd
14285 @opindex mfused-madd
14286 @opindex mno-fused-madd
14287 Enable (disable) use of the floating point multiply-accumulate
14288 instructions, when they are available. The default is
14289 @option{-mfused-madd}.
14291 When multiply-accumulate instructions are used, the intermediate
14292 product is calculated to infinite precision and is not subject to
14293 the FCSR Flush to Zero bit. This may be undesirable in some
14298 Tell the MIPS assembler to not run its preprocessor over user
14299 assembler files (with a @samp{.s} suffix) when assembling them.
14302 @itemx -mno-fix-r4000
14303 @opindex mfix-r4000
14304 @opindex mno-fix-r4000
14305 Work around certain R4000 CPU errata:
14308 A double-word or a variable shift may give an incorrect result if executed
14309 immediately after starting an integer division.
14311 A double-word or a variable shift may give an incorrect result if executed
14312 while an integer multiplication is in progress.
14314 An integer division may give an incorrect result if started in a delay slot
14315 of a taken branch or a jump.
14319 @itemx -mno-fix-r4400
14320 @opindex mfix-r4400
14321 @opindex mno-fix-r4400
14322 Work around certain R4400 CPU errata:
14325 A double-word or a variable shift may give an incorrect result if executed
14326 immediately after starting an integer division.
14330 @itemx -mno-fix-r10000
14331 @opindex mfix-r10000
14332 @opindex mno-fix-r10000
14333 Work around certain R10000 errata:
14336 @code{ll}/@code{sc} sequences may not behave atomically on revisions
14337 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
14340 This option can only be used if the target architecture supports
14341 branch-likely instructions. @option{-mfix-r10000} is the default when
14342 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
14346 @itemx -mno-fix-vr4120
14347 @opindex mfix-vr4120
14348 Work around certain VR4120 errata:
14351 @code{dmultu} does not always produce the correct result.
14353 @code{div} and @code{ddiv} do not always produce the correct result if one
14354 of the operands is negative.
14356 The workarounds for the division errata rely on special functions in
14357 @file{libgcc.a}. At present, these functions are only provided by
14358 the @code{mips64vr*-elf} configurations.
14360 Other VR4120 errata require a nop to be inserted between certain pairs of
14361 instructions. These errata are handled by the assembler, not by GCC itself.
14364 @opindex mfix-vr4130
14365 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
14366 workarounds are implemented by the assembler rather than by GCC,
14367 although GCC will avoid using @code{mflo} and @code{mfhi} if the
14368 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
14369 instructions are available instead.
14372 @itemx -mno-fix-sb1
14374 Work around certain SB-1 CPU core errata.
14375 (This flag currently works around the SB-1 revision 2
14376 ``F1'' and ``F2'' floating point errata.)
14378 @item -mr10k-cache-barrier=@var{setting}
14379 @opindex mr10k-cache-barrier
14380 Specify whether GCC should insert cache barriers to avoid the
14381 side-effects of speculation on R10K processors.
14383 In common with many processors, the R10K tries to predict the outcome
14384 of a conditional branch and speculatively executes instructions from
14385 the ``taken'' branch. It later aborts these instructions if the
14386 predicted outcome was wrong. However, on the R10K, even aborted
14387 instructions can have side effects.
14389 This problem only affects kernel stores and, depending on the system,
14390 kernel loads. As an example, a speculatively-executed store may load
14391 the target memory into cache and mark the cache line as dirty, even if
14392 the store itself is later aborted. If a DMA operation writes to the
14393 same area of memory before the ``dirty'' line is flushed, the cached
14394 data will overwrite the DMA-ed data. See the R10K processor manual
14395 for a full description, including other potential problems.
14397 One workaround is to insert cache barrier instructions before every memory
14398 access that might be speculatively executed and that might have side
14399 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
14400 controls GCC's implementation of this workaround. It assumes that
14401 aborted accesses to any byte in the following regions will not have
14406 the memory occupied by the current function's stack frame;
14409 the memory occupied by an incoming stack argument;
14412 the memory occupied by an object with a link-time-constant address.
14415 It is the kernel's responsibility to ensure that speculative
14416 accesses to these regions are indeed safe.
14418 If the input program contains a function declaration such as:
14424 then the implementation of @code{foo} must allow @code{j foo} and
14425 @code{jal foo} to be executed speculatively. GCC honors this
14426 restriction for functions it compiles itself. It expects non-GCC
14427 functions (such as hand-written assembly code) to do the same.
14429 The option has three forms:
14432 @item -mr10k-cache-barrier=load-store
14433 Insert a cache barrier before a load or store that might be
14434 speculatively executed and that might have side effects even
14437 @item -mr10k-cache-barrier=store
14438 Insert a cache barrier before a store that might be speculatively
14439 executed and that might have side effects even if aborted.
14441 @item -mr10k-cache-barrier=none
14442 Disable the insertion of cache barriers. This is the default setting.
14445 @item -mflush-func=@var{func}
14446 @itemx -mno-flush-func
14447 @opindex mflush-func
14448 Specifies the function to call to flush the I and D caches, or to not
14449 call any such function. If called, the function must take the same
14450 arguments as the common @code{_flush_func()}, that is, the address of the
14451 memory range for which the cache is being flushed, the size of the
14452 memory range, and the number 3 (to flush both caches). The default
14453 depends on the target GCC was configured for, but commonly is either
14454 @samp{_flush_func} or @samp{__cpu_flush}.
14456 @item mbranch-cost=@var{num}
14457 @opindex mbranch-cost
14458 Set the cost of branches to roughly @var{num} ``simple'' instructions.
14459 This cost is only a heuristic and is not guaranteed to produce
14460 consistent results across releases. A zero cost redundantly selects
14461 the default, which is based on the @option{-mtune} setting.
14463 @item -mbranch-likely
14464 @itemx -mno-branch-likely
14465 @opindex mbranch-likely
14466 @opindex mno-branch-likely
14467 Enable or disable use of Branch Likely instructions, regardless of the
14468 default for the selected architecture. By default, Branch Likely
14469 instructions may be generated if they are supported by the selected
14470 architecture. An exception is for the MIPS32 and MIPS64 architectures
14471 and processors which implement those architectures; for those, Branch
14472 Likely instructions will not be generated by default because the MIPS32
14473 and MIPS64 architectures specifically deprecate their use.
14475 @item -mfp-exceptions
14476 @itemx -mno-fp-exceptions
14477 @opindex mfp-exceptions
14478 Specifies whether FP exceptions are enabled. This affects how we schedule
14479 FP instructions for some processors. The default is that FP exceptions are
14482 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
14483 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
14486 @item -mvr4130-align
14487 @itemx -mno-vr4130-align
14488 @opindex mvr4130-align
14489 The VR4130 pipeline is two-way superscalar, but can only issue two
14490 instructions together if the first one is 8-byte aligned. When this
14491 option is enabled, GCC will align pairs of instructions that it
14492 thinks should execute in parallel.
14494 This option only has an effect when optimizing for the VR4130.
14495 It normally makes code faster, but at the expense of making it bigger.
14496 It is enabled by default at optimization level @option{-O3}.
14501 Enable (disable) generation of @code{synci} instructions on
14502 architectures that support it. The @code{synci} instructions (if
14503 enabled) will be generated when @code{__builtin___clear_cache()} is
14506 This option defaults to @code{-mno-synci}, but the default can be
14507 overridden by configuring with @code{--with-synci}.
14509 When compiling code for single processor systems, it is generally safe
14510 to use @code{synci}. However, on many multi-core (SMP) systems, it
14511 will not invalidate the instruction caches on all cores and may lead
14512 to undefined behavior.
14514 @item -mrelax-pic-calls
14515 @itemx -mno-relax-pic-calls
14516 @opindex mrelax-pic-calls
14517 Try to turn PIC calls that are normally dispatched via register
14518 @code{$25} into direct calls. This is only possible if the linker can
14519 resolve the destination at link-time and if the destination is within
14520 range for a direct call.
14522 @option{-mrelax-pic-calls} is the default if GCC was configured to use
14523 an assembler and a linker that supports the @code{.reloc} assembly
14524 directive and @code{-mexplicit-relocs} is in effect. With
14525 @code{-mno-explicit-relocs}, this optimization can be performed by the
14526 assembler and the linker alone without help from the compiler.
14528 @item -mmcount-ra-address
14529 @itemx -mno-mcount-ra-address
14530 @opindex mmcount-ra-address
14531 @opindex mno-mcount-ra-address
14532 Emit (do not emit) code that allows @code{_mcount} to modify the
14533 calling function's return address. When enabled, this option extends
14534 the usual @code{_mcount} interface with a new @var{ra-address}
14535 parameter, which has type @code{intptr_t *} and is passed in register
14536 @code{$12}. @code{_mcount} can then modify the return address by
14537 doing both of the following:
14540 Returning the new address in register @code{$31}.
14542 Storing the new address in @code{*@var{ra-address}},
14543 if @var{ra-address} is nonnull.
14546 The default is @option{-mno-mcount-ra-address}.
14551 @subsection MMIX Options
14552 @cindex MMIX Options
14554 These options are defined for the MMIX:
14558 @itemx -mno-libfuncs
14560 @opindex mno-libfuncs
14561 Specify that intrinsic library functions are being compiled, passing all
14562 values in registers, no matter the size.
14565 @itemx -mno-epsilon
14567 @opindex mno-epsilon
14568 Generate floating-point comparison instructions that compare with respect
14569 to the @code{rE} epsilon register.
14571 @item -mabi=mmixware
14573 @opindex mabi=mmixware
14575 Generate code that passes function parameters and return values that (in
14576 the called function) are seen as registers @code{$0} and up, as opposed to
14577 the GNU ABI which uses global registers @code{$231} and up.
14579 @item -mzero-extend
14580 @itemx -mno-zero-extend
14581 @opindex mzero-extend
14582 @opindex mno-zero-extend
14583 When reading data from memory in sizes shorter than 64 bits, use (do not
14584 use) zero-extending load instructions by default, rather than
14585 sign-extending ones.
14588 @itemx -mno-knuthdiv
14590 @opindex mno-knuthdiv
14591 Make the result of a division yielding a remainder have the same sign as
14592 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
14593 remainder follows the sign of the dividend. Both methods are
14594 arithmetically valid, the latter being almost exclusively used.
14596 @item -mtoplevel-symbols
14597 @itemx -mno-toplevel-symbols
14598 @opindex mtoplevel-symbols
14599 @opindex mno-toplevel-symbols
14600 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
14601 code can be used with the @code{PREFIX} assembly directive.
14605 Generate an executable in the ELF format, rather than the default
14606 @samp{mmo} format used by the @command{mmix} simulator.
14608 @item -mbranch-predict
14609 @itemx -mno-branch-predict
14610 @opindex mbranch-predict
14611 @opindex mno-branch-predict
14612 Use (do not use) the probable-branch instructions, when static branch
14613 prediction indicates a probable branch.
14615 @item -mbase-addresses
14616 @itemx -mno-base-addresses
14617 @opindex mbase-addresses
14618 @opindex mno-base-addresses
14619 Generate (do not generate) code that uses @emph{base addresses}. Using a
14620 base address automatically generates a request (handled by the assembler
14621 and the linker) for a constant to be set up in a global register. The
14622 register is used for one or more base address requests within the range 0
14623 to 255 from the value held in the register. The generally leads to short
14624 and fast code, but the number of different data items that can be
14625 addressed is limited. This means that a program that uses lots of static
14626 data may require @option{-mno-base-addresses}.
14628 @item -msingle-exit
14629 @itemx -mno-single-exit
14630 @opindex msingle-exit
14631 @opindex mno-single-exit
14632 Force (do not force) generated code to have a single exit point in each
14636 @node MN10300 Options
14637 @subsection MN10300 Options
14638 @cindex MN10300 options
14640 These @option{-m} options are defined for Matsushita MN10300 architectures:
14645 Generate code to avoid bugs in the multiply instructions for the MN10300
14646 processors. This is the default.
14648 @item -mno-mult-bug
14649 @opindex mno-mult-bug
14650 Do not generate code to avoid bugs in the multiply instructions for the
14651 MN10300 processors.
14655 Generate code which uses features specific to the AM33 processor.
14659 Do not generate code which uses features specific to the AM33 processor. This
14662 @item -mreturn-pointer-on-d0
14663 @opindex mreturn-pointer-on-d0
14664 When generating a function which returns a pointer, return the pointer
14665 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
14666 only in a0, and attempts to call such functions without a prototype
14667 would result in errors. Note that this option is on by default; use
14668 @option{-mno-return-pointer-on-d0} to disable it.
14672 Do not link in the C run-time initialization object file.
14676 Indicate to the linker that it should perform a relaxation optimization pass
14677 to shorten branches, calls and absolute memory addresses. This option only
14678 has an effect when used on the command line for the final link step.
14680 This option makes symbolic debugging impossible.
14683 @node PDP-11 Options
14684 @subsection PDP-11 Options
14685 @cindex PDP-11 Options
14687 These options are defined for the PDP-11:
14692 Use hardware FPP floating point. This is the default. (FIS floating
14693 point on the PDP-11/40 is not supported.)
14696 @opindex msoft-float
14697 Do not use hardware floating point.
14701 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
14705 Return floating-point results in memory. This is the default.
14709 Generate code for a PDP-11/40.
14713 Generate code for a PDP-11/45. This is the default.
14717 Generate code for a PDP-11/10.
14719 @item -mbcopy-builtin
14720 @opindex mbcopy-builtin
14721 Use inline @code{movmemhi} patterns for copying memory. This is the
14726 Do not use inline @code{movmemhi} patterns for copying memory.
14732 Use 16-bit @code{int}. This is the default.
14738 Use 32-bit @code{int}.
14741 @itemx -mno-float32
14743 @opindex mno-float32
14744 Use 64-bit @code{float}. This is the default.
14747 @itemx -mno-float64
14749 @opindex mno-float64
14750 Use 32-bit @code{float}.
14754 Use @code{abshi2} pattern. This is the default.
14758 Do not use @code{abshi2} pattern.
14760 @item -mbranch-expensive
14761 @opindex mbranch-expensive
14762 Pretend that branches are expensive. This is for experimenting with
14763 code generation only.
14765 @item -mbranch-cheap
14766 @opindex mbranch-cheap
14767 Do not pretend that branches are expensive. This is the default.
14771 Generate code for a system with split I&D@.
14775 Generate code for a system without split I&D@. This is the default.
14779 Use Unix assembler syntax. This is the default when configured for
14780 @samp{pdp11-*-bsd}.
14784 Use DEC assembler syntax. This is the default when configured for any
14785 PDP-11 target other than @samp{pdp11-*-bsd}.
14788 @node picoChip Options
14789 @subsection picoChip Options
14790 @cindex picoChip options
14792 These @samp{-m} options are defined for picoChip implementations:
14796 @item -mae=@var{ae_type}
14798 Set the instruction set, register set, and instruction scheduling
14799 parameters for array element type @var{ae_type}. Supported values
14800 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
14802 @option{-mae=ANY} selects a completely generic AE type. Code
14803 generated with this option will run on any of the other AE types. The
14804 code will not be as efficient as it would be if compiled for a specific
14805 AE type, and some types of operation (e.g., multiplication) will not
14806 work properly on all types of AE.
14808 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
14809 for compiled code, and is the default.
14811 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
14812 option may suffer from poor performance of byte (char) manipulation,
14813 since the DSP AE does not provide hardware support for byte load/stores.
14815 @item -msymbol-as-address
14816 Enable the compiler to directly use a symbol name as an address in a
14817 load/store instruction, without first loading it into a
14818 register. Typically, the use of this option will generate larger
14819 programs, which run faster than when the option isn't used. However, the
14820 results vary from program to program, so it is left as a user option,
14821 rather than being permanently enabled.
14823 @item -mno-inefficient-warnings
14824 Disables warnings about the generation of inefficient code. These
14825 warnings can be generated, for example, when compiling code which
14826 performs byte-level memory operations on the MAC AE type. The MAC AE has
14827 no hardware support for byte-level memory operations, so all byte
14828 load/stores must be synthesized from word load/store operations. This is
14829 inefficient and a warning will be generated indicating to the programmer
14830 that they should rewrite the code to avoid byte operations, or to target
14831 an AE type which has the necessary hardware support. This option enables
14832 the warning to be turned off.
14836 @node PowerPC Options
14837 @subsection PowerPC Options
14838 @cindex PowerPC options
14840 These are listed under @xref{RS/6000 and PowerPC Options}.
14842 @node RS/6000 and PowerPC Options
14843 @subsection IBM RS/6000 and PowerPC Options
14844 @cindex RS/6000 and PowerPC Options
14845 @cindex IBM RS/6000 and PowerPC Options
14847 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
14854 @itemx -mno-powerpc
14855 @itemx -mpowerpc-gpopt
14856 @itemx -mno-powerpc-gpopt
14857 @itemx -mpowerpc-gfxopt
14858 @itemx -mno-powerpc-gfxopt
14860 @itemx -mno-powerpc64
14864 @itemx -mno-popcntb
14866 @itemx -mno-popcntd
14874 @itemx -mno-hard-dfp
14878 @opindex mno-power2
14880 @opindex mno-powerpc
14881 @opindex mpowerpc-gpopt
14882 @opindex mno-powerpc-gpopt
14883 @opindex mpowerpc-gfxopt
14884 @opindex mno-powerpc-gfxopt
14885 @opindex mpowerpc64
14886 @opindex mno-powerpc64
14890 @opindex mno-popcntb
14892 @opindex mno-popcntd
14898 @opindex mno-mfpgpr
14900 @opindex mno-hard-dfp
14901 GCC supports two related instruction set architectures for the
14902 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
14903 instructions supported by the @samp{rios} chip set used in the original
14904 RS/6000 systems and the @dfn{PowerPC} instruction set is the
14905 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
14906 the IBM 4xx, 6xx, and follow-on microprocessors.
14908 Neither architecture is a subset of the other. However there is a
14909 large common subset of instructions supported by both. An MQ
14910 register is included in processors supporting the POWER architecture.
14912 You use these options to specify which instructions are available on the
14913 processor you are using. The default value of these options is
14914 determined when configuring GCC@. Specifying the
14915 @option{-mcpu=@var{cpu_type}} overrides the specification of these
14916 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
14917 rather than the options listed above.
14919 The @option{-mpower} option allows GCC to generate instructions that
14920 are found only in the POWER architecture and to use the MQ register.
14921 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
14922 to generate instructions that are present in the POWER2 architecture but
14923 not the original POWER architecture.
14925 The @option{-mpowerpc} option allows GCC to generate instructions that
14926 are found only in the 32-bit subset of the PowerPC architecture.
14927 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
14928 GCC to use the optional PowerPC architecture instructions in the
14929 General Purpose group, including floating-point square root. Specifying
14930 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
14931 use the optional PowerPC architecture instructions in the Graphics
14932 group, including floating-point select.
14934 The @option{-mmfcrf} option allows GCC to generate the move from
14935 condition register field instruction implemented on the POWER4
14936 processor and other processors that support the PowerPC V2.01
14938 The @option{-mpopcntb} option allows GCC to generate the popcount and
14939 double precision FP reciprocal estimate instruction implemented on the
14940 POWER5 processor and other processors that support the PowerPC V2.02
14942 The @option{-mpopcntd} option allows GCC to generate the popcount
14943 instruction implemented on the POWER7 processor and other processors
14944 that support the PowerPC V2.06 architecture.
14945 The @option{-mfprnd} option allows GCC to generate the FP round to
14946 integer instructions implemented on the POWER5+ processor and other
14947 processors that support the PowerPC V2.03 architecture.
14948 The @option{-mcmpb} option allows GCC to generate the compare bytes
14949 instruction implemented on the POWER6 processor and other processors
14950 that support the PowerPC V2.05 architecture.
14951 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
14952 general purpose register instructions implemented on the POWER6X
14953 processor and other processors that support the extended PowerPC V2.05
14955 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
14956 point instructions implemented on some POWER processors.
14958 The @option{-mpowerpc64} option allows GCC to generate the additional
14959 64-bit instructions that are found in the full PowerPC64 architecture
14960 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
14961 @option{-mno-powerpc64}.
14963 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
14964 will use only the instructions in the common subset of both
14965 architectures plus some special AIX common-mode calls, and will not use
14966 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
14967 permits GCC to use any instruction from either architecture and to
14968 allow use of the MQ register; specify this for the Motorola MPC601.
14970 @item -mnew-mnemonics
14971 @itemx -mold-mnemonics
14972 @opindex mnew-mnemonics
14973 @opindex mold-mnemonics
14974 Select which mnemonics to use in the generated assembler code. With
14975 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
14976 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
14977 assembler mnemonics defined for the POWER architecture. Instructions
14978 defined in only one architecture have only one mnemonic; GCC uses that
14979 mnemonic irrespective of which of these options is specified.
14981 GCC defaults to the mnemonics appropriate for the architecture in
14982 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
14983 value of these option. Unless you are building a cross-compiler, you
14984 should normally not specify either @option{-mnew-mnemonics} or
14985 @option{-mold-mnemonics}, but should instead accept the default.
14987 @item -mcpu=@var{cpu_type}
14989 Set architecture type, register usage, choice of mnemonics, and
14990 instruction scheduling parameters for machine type @var{cpu_type}.
14991 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
14992 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
14993 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
14994 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
14995 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
14996 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
14997 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{ec603e}, @samp{G3},
14998 @samp{G4}, @samp{G5}, @samp{titan}, @samp{power}, @samp{power2}, @samp{power3},
14999 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x},
15000 @samp{power7}, @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
15001 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
15003 @option{-mcpu=common} selects a completely generic processor. Code
15004 generated under this option will run on any POWER or PowerPC processor.
15005 GCC will use only the instructions in the common subset of both
15006 architectures, and will not use the MQ register. GCC assumes a generic
15007 processor model for scheduling purposes.
15009 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
15010 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
15011 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
15012 types, with an appropriate, generic processor model assumed for
15013 scheduling purposes.
15015 The other options specify a specific processor. Code generated under
15016 those options will run best on that processor, and may not run at all on
15019 The @option{-mcpu} options automatically enable or disable the
15022 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
15023 -mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol
15024 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
15025 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx}
15027 The particular options set for any particular CPU will vary between
15028 compiler versions, depending on what setting seems to produce optimal
15029 code for that CPU; it doesn't necessarily reflect the actual hardware's
15030 capabilities. If you wish to set an individual option to a particular
15031 value, you may specify it after the @option{-mcpu} option, like
15032 @samp{-mcpu=970 -mno-altivec}.
15034 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
15035 not enabled or disabled by the @option{-mcpu} option at present because
15036 AIX does not have full support for these options. You may still
15037 enable or disable them individually if you're sure it'll work in your
15040 @item -mtune=@var{cpu_type}
15042 Set the instruction scheduling parameters for machine type
15043 @var{cpu_type}, but do not set the architecture type, register usage, or
15044 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
15045 values for @var{cpu_type} are used for @option{-mtune} as for
15046 @option{-mcpu}. If both are specified, the code generated will use the
15047 architecture, registers, and mnemonics set by @option{-mcpu}, but the
15048 scheduling parameters set by @option{-mtune}.
15050 @item -mcmodel=small
15051 @opindex mcmodel=small
15052 Generate PowerPC64 code for the small model: The TOC is limited to
15055 @item -mcmodel=large
15056 @opindex mcmodel=large
15057 Generate PowerPC64 code for the large model: The TOC may be up to 4G
15058 in size. Other data and code is only limited by the 64-bit address
15062 @itemx -mno-altivec
15064 @opindex mno-altivec
15065 Generate code that uses (does not use) AltiVec instructions, and also
15066 enable the use of built-in functions that allow more direct access to
15067 the AltiVec instruction set. You may also need to set
15068 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
15074 @opindex mno-vrsave
15075 Generate VRSAVE instructions when generating AltiVec code.
15077 @item -mgen-cell-microcode
15078 @opindex mgen-cell-microcode
15079 Generate Cell microcode instructions
15081 @item -mwarn-cell-microcode
15082 @opindex mwarn-cell-microcode
15083 Warning when a Cell microcode instruction is going to emitted. An example
15084 of a Cell microcode instruction is a variable shift.
15087 @opindex msecure-plt
15088 Generate code that allows ld and ld.so to build executables and shared
15089 libraries with non-exec .plt and .got sections. This is a PowerPC
15090 32-bit SYSV ABI option.
15094 Generate code that uses a BSS .plt section that ld.so fills in, and
15095 requires .plt and .got sections that are both writable and executable.
15096 This is a PowerPC 32-bit SYSV ABI option.
15102 This switch enables or disables the generation of ISEL instructions.
15104 @item -misel=@var{yes/no}
15105 This switch has been deprecated. Use @option{-misel} and
15106 @option{-mno-isel} instead.
15112 This switch enables or disables the generation of SPE simd
15118 @opindex mno-paired
15119 This switch enables or disables the generation of PAIRED simd
15122 @item -mspe=@var{yes/no}
15123 This option has been deprecated. Use @option{-mspe} and
15124 @option{-mno-spe} instead.
15130 Generate code that uses (does not use) vector/scalar (VSX)
15131 instructions, and also enable the use of built-in functions that allow
15132 more direct access to the VSX instruction set.
15134 @item -mfloat-gprs=@var{yes/single/double/no}
15135 @itemx -mfloat-gprs
15136 @opindex mfloat-gprs
15137 This switch enables or disables the generation of floating point
15138 operations on the general purpose registers for architectures that
15141 The argument @var{yes} or @var{single} enables the use of
15142 single-precision floating point operations.
15144 The argument @var{double} enables the use of single and
15145 double-precision floating point operations.
15147 The argument @var{no} disables floating point operations on the
15148 general purpose registers.
15150 This option is currently only available on the MPC854x.
15156 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
15157 targets (including GNU/Linux). The 32-bit environment sets int, long
15158 and pointer to 32 bits and generates code that runs on any PowerPC
15159 variant. The 64-bit environment sets int to 32 bits and long and
15160 pointer to 64 bits, and generates code for PowerPC64, as for
15161 @option{-mpowerpc64}.
15164 @itemx -mno-fp-in-toc
15165 @itemx -mno-sum-in-toc
15166 @itemx -mminimal-toc
15168 @opindex mno-fp-in-toc
15169 @opindex mno-sum-in-toc
15170 @opindex mminimal-toc
15171 Modify generation of the TOC (Table Of Contents), which is created for
15172 every executable file. The @option{-mfull-toc} option is selected by
15173 default. In that case, GCC will allocate at least one TOC entry for
15174 each unique non-automatic variable reference in your program. GCC
15175 will also place floating-point constants in the TOC@. However, only
15176 16,384 entries are available in the TOC@.
15178 If you receive a linker error message that saying you have overflowed
15179 the available TOC space, you can reduce the amount of TOC space used
15180 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
15181 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
15182 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
15183 generate code to calculate the sum of an address and a constant at
15184 run-time instead of putting that sum into the TOC@. You may specify one
15185 or both of these options. Each causes GCC to produce very slightly
15186 slower and larger code at the expense of conserving TOC space.
15188 If you still run out of space in the TOC even when you specify both of
15189 these options, specify @option{-mminimal-toc} instead. This option causes
15190 GCC to make only one TOC entry for every file. When you specify this
15191 option, GCC will produce code that is slower and larger but which
15192 uses extremely little TOC space. You may wish to use this option
15193 only on files that contain less frequently executed code.
15199 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
15200 @code{long} type, and the infrastructure needed to support them.
15201 Specifying @option{-maix64} implies @option{-mpowerpc64} and
15202 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
15203 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
15206 @itemx -mno-xl-compat
15207 @opindex mxl-compat
15208 @opindex mno-xl-compat
15209 Produce code that conforms more closely to IBM XL compiler semantics
15210 when using AIX-compatible ABI@. Pass floating-point arguments to
15211 prototyped functions beyond the register save area (RSA) on the stack
15212 in addition to argument FPRs. Do not assume that most significant
15213 double in 128-bit long double value is properly rounded when comparing
15214 values and converting to double. Use XL symbol names for long double
15217 The AIX calling convention was extended but not initially documented to
15218 handle an obscure K&R C case of calling a function that takes the
15219 address of its arguments with fewer arguments than declared. IBM XL
15220 compilers access floating point arguments which do not fit in the
15221 RSA from the stack when a subroutine is compiled without
15222 optimization. Because always storing floating-point arguments on the
15223 stack is inefficient and rarely needed, this option is not enabled by
15224 default and only is necessary when calling subroutines compiled by IBM
15225 XL compilers without optimization.
15229 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
15230 application written to use message passing with special startup code to
15231 enable the application to run. The system must have PE installed in the
15232 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
15233 must be overridden with the @option{-specs=} option to specify the
15234 appropriate directory location. The Parallel Environment does not
15235 support threads, so the @option{-mpe} option and the @option{-pthread}
15236 option are incompatible.
15238 @item -malign-natural
15239 @itemx -malign-power
15240 @opindex malign-natural
15241 @opindex malign-power
15242 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
15243 @option{-malign-natural} overrides the ABI-defined alignment of larger
15244 types, such as floating-point doubles, on their natural size-based boundary.
15245 The option @option{-malign-power} instructs GCC to follow the ABI-specified
15246 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
15248 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
15252 @itemx -mhard-float
15253 @opindex msoft-float
15254 @opindex mhard-float
15255 Generate code that does not use (uses) the floating-point register set.
15256 Software floating point emulation is provided if you use the
15257 @option{-msoft-float} option, and pass the option to GCC when linking.
15259 @item -msingle-float
15260 @itemx -mdouble-float
15261 @opindex msingle-float
15262 @opindex mdouble-float
15263 Generate code for single or double-precision floating point operations.
15264 @option{-mdouble-float} implies @option{-msingle-float}.
15267 @opindex msimple-fpu
15268 Do not generate sqrt and div instructions for hardware floating point unit.
15272 Specify type of floating point unit. Valid values are @var{sp_lite}
15273 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
15274 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
15275 and @var{dp_full} (equivalent to -mdouble-float).
15278 @opindex mxilinx-fpu
15279 Perform optimizations for floating point unit on Xilinx PPC 405/440.
15282 @itemx -mno-multiple
15284 @opindex mno-multiple
15285 Generate code that uses (does not use) the load multiple word
15286 instructions and the store multiple word instructions. These
15287 instructions are generated by default on POWER systems, and not
15288 generated on PowerPC systems. Do not use @option{-mmultiple} on little
15289 endian PowerPC systems, since those instructions do not work when the
15290 processor is in little endian mode. The exceptions are PPC740 and
15291 PPC750 which permit the instructions usage in little endian mode.
15296 @opindex mno-string
15297 Generate code that uses (does not use) the load string instructions
15298 and the store string word instructions to save multiple registers and
15299 do small block moves. These instructions are generated by default on
15300 POWER systems, and not generated on PowerPC systems. Do not use
15301 @option{-mstring} on little endian PowerPC systems, since those
15302 instructions do not work when the processor is in little endian mode.
15303 The exceptions are PPC740 and PPC750 which permit the instructions
15304 usage in little endian mode.
15309 @opindex mno-update
15310 Generate code that uses (does not use) the load or store instructions
15311 that update the base register to the address of the calculated memory
15312 location. These instructions are generated by default. If you use
15313 @option{-mno-update}, there is a small window between the time that the
15314 stack pointer is updated and the address of the previous frame is
15315 stored, which means code that walks the stack frame across interrupts or
15316 signals may get corrupted data.
15318 @item -mavoid-indexed-addresses
15319 @itemx -mno-avoid-indexed-addresses
15320 @opindex mavoid-indexed-addresses
15321 @opindex mno-avoid-indexed-addresses
15322 Generate code that tries to avoid (not avoid) the use of indexed load
15323 or store instructions. These instructions can incur a performance
15324 penalty on Power6 processors in certain situations, such as when
15325 stepping through large arrays that cross a 16M boundary. This option
15326 is enabled by default when targetting Power6 and disabled otherwise.
15329 @itemx -mno-fused-madd
15330 @opindex mfused-madd
15331 @opindex mno-fused-madd
15332 Generate code that uses (does not use) the floating point multiply and
15333 accumulate instructions. These instructions are generated by default if
15334 hardware floating is used.
15340 Generate code that uses (does not use) the half-word multiply and
15341 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
15342 These instructions are generated by default when targetting those
15349 Generate code that uses (does not use) the string-search @samp{dlmzb}
15350 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
15351 generated by default when targetting those processors.
15353 @item -mno-bit-align
15355 @opindex mno-bit-align
15356 @opindex mbit-align
15357 On System V.4 and embedded PowerPC systems do not (do) force structures
15358 and unions that contain bit-fields to be aligned to the base type of the
15361 For example, by default a structure containing nothing but 8
15362 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
15363 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
15364 the structure would be aligned to a 1 byte boundary and be one byte in
15367 @item -mno-strict-align
15368 @itemx -mstrict-align
15369 @opindex mno-strict-align
15370 @opindex mstrict-align
15371 On System V.4 and embedded PowerPC systems do not (do) assume that
15372 unaligned memory references will be handled by the system.
15374 @item -mrelocatable
15375 @itemx -mno-relocatable
15376 @opindex mrelocatable
15377 @opindex mno-relocatable
15378 On embedded PowerPC systems generate code that allows (does not allow)
15379 the program to be relocated to a different address at runtime. If you
15380 use @option{-mrelocatable} on any module, all objects linked together must
15381 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
15383 @item -mrelocatable-lib
15384 @itemx -mno-relocatable-lib
15385 @opindex mrelocatable-lib
15386 @opindex mno-relocatable-lib
15387 On embedded PowerPC systems generate code that allows (does not allow)
15388 the program to be relocated to a different address at runtime. Modules
15389 compiled with @option{-mrelocatable-lib} can be linked with either modules
15390 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
15391 with modules compiled with the @option{-mrelocatable} options.
15397 On System V.4 and embedded PowerPC systems do not (do) assume that
15398 register 2 contains a pointer to a global area pointing to the addresses
15399 used in the program.
15402 @itemx -mlittle-endian
15404 @opindex mlittle-endian
15405 On System V.4 and embedded PowerPC systems compile code for the
15406 processor in little endian mode. The @option{-mlittle-endian} option is
15407 the same as @option{-mlittle}.
15410 @itemx -mbig-endian
15412 @opindex mbig-endian
15413 On System V.4 and embedded PowerPC systems compile code for the
15414 processor in big endian mode. The @option{-mbig-endian} option is
15415 the same as @option{-mbig}.
15417 @item -mdynamic-no-pic
15418 @opindex mdynamic-no-pic
15419 On Darwin and Mac OS X systems, compile code so that it is not
15420 relocatable, but that its external references are relocatable. The
15421 resulting code is suitable for applications, but not shared
15424 @item -mprioritize-restricted-insns=@var{priority}
15425 @opindex mprioritize-restricted-insns
15426 This option controls the priority that is assigned to
15427 dispatch-slot restricted instructions during the second scheduling
15428 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
15429 @var{no/highest/second-highest} priority to dispatch slot restricted
15432 @item -msched-costly-dep=@var{dependence_type}
15433 @opindex msched-costly-dep
15434 This option controls which dependences are considered costly
15435 by the target during instruction scheduling. The argument
15436 @var{dependence_type} takes one of the following values:
15437 @var{no}: no dependence is costly,
15438 @var{all}: all dependences are costly,
15439 @var{true_store_to_load}: a true dependence from store to load is costly,
15440 @var{store_to_load}: any dependence from store to load is costly,
15441 @var{number}: any dependence which latency >= @var{number} is costly.
15443 @item -minsert-sched-nops=@var{scheme}
15444 @opindex minsert-sched-nops
15445 This option controls which nop insertion scheme will be used during
15446 the second scheduling pass. The argument @var{scheme} takes one of the
15448 @var{no}: Don't insert nops.
15449 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
15450 according to the scheduler's grouping.
15451 @var{regroup_exact}: Insert nops to force costly dependent insns into
15452 separate groups. Insert exactly as many nops as needed to force an insn
15453 to a new group, according to the estimated processor grouping.
15454 @var{number}: Insert nops to force costly dependent insns into
15455 separate groups. Insert @var{number} nops to force an insn to a new group.
15458 @opindex mcall-sysv
15459 On System V.4 and embedded PowerPC systems compile code using calling
15460 conventions that adheres to the March 1995 draft of the System V
15461 Application Binary Interface, PowerPC processor supplement. This is the
15462 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
15464 @item -mcall-sysv-eabi
15466 @opindex mcall-sysv-eabi
15467 @opindex mcall-eabi
15468 Specify both @option{-mcall-sysv} and @option{-meabi} options.
15470 @item -mcall-sysv-noeabi
15471 @opindex mcall-sysv-noeabi
15472 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
15474 @item -mcall-aixdesc
15476 On System V.4 and embedded PowerPC systems compile code for the AIX
15480 @opindex mcall-linux
15481 On System V.4 and embedded PowerPC systems compile code for the
15482 Linux-based GNU system.
15486 On System V.4 and embedded PowerPC systems compile code for the
15487 Hurd-based GNU system.
15489 @item -mcall-freebsd
15490 @opindex mcall-freebsd
15491 On System V.4 and embedded PowerPC systems compile code for the
15492 FreeBSD operating system.
15494 @item -mcall-netbsd
15495 @opindex mcall-netbsd
15496 On System V.4 and embedded PowerPC systems compile code for the
15497 NetBSD operating system.
15499 @item -mcall-openbsd
15500 @opindex mcall-netbsd
15501 On System V.4 and embedded PowerPC systems compile code for the
15502 OpenBSD operating system.
15504 @item -maix-struct-return
15505 @opindex maix-struct-return
15506 Return all structures in memory (as specified by the AIX ABI)@.
15508 @item -msvr4-struct-return
15509 @opindex msvr4-struct-return
15510 Return structures smaller than 8 bytes in registers (as specified by the
15513 @item -mabi=@var{abi-type}
15515 Extend the current ABI with a particular extension, or remove such extension.
15516 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
15517 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
15521 Extend the current ABI with SPE ABI extensions. This does not change
15522 the default ABI, instead it adds the SPE ABI extensions to the current
15526 @opindex mabi=no-spe
15527 Disable Booke SPE ABI extensions for the current ABI@.
15529 @item -mabi=ibmlongdouble
15530 @opindex mabi=ibmlongdouble
15531 Change the current ABI to use IBM extended precision long double.
15532 This is a PowerPC 32-bit SYSV ABI option.
15534 @item -mabi=ieeelongdouble
15535 @opindex mabi=ieeelongdouble
15536 Change the current ABI to use IEEE extended precision long double.
15537 This is a PowerPC 32-bit Linux ABI option.
15540 @itemx -mno-prototype
15541 @opindex mprototype
15542 @opindex mno-prototype
15543 On System V.4 and embedded PowerPC systems assume that all calls to
15544 variable argument functions are properly prototyped. Otherwise, the
15545 compiler must insert an instruction before every non prototyped call to
15546 set or clear bit 6 of the condition code register (@var{CR}) to
15547 indicate whether floating point values were passed in the floating point
15548 registers in case the function takes a variable arguments. With
15549 @option{-mprototype}, only calls to prototyped variable argument functions
15550 will set or clear the bit.
15554 On embedded PowerPC systems, assume that the startup module is called
15555 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
15556 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
15561 On embedded PowerPC systems, assume that the startup module is called
15562 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
15567 On embedded PowerPC systems, assume that the startup module is called
15568 @file{crt0.o} and the standard C libraries are @file{libads.a} and
15571 @item -myellowknife
15572 @opindex myellowknife
15573 On embedded PowerPC systems, assume that the startup module is called
15574 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
15579 On System V.4 and embedded PowerPC systems, specify that you are
15580 compiling for a VxWorks system.
15584 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
15585 header to indicate that @samp{eabi} extended relocations are used.
15591 On System V.4 and embedded PowerPC systems do (do not) adhere to the
15592 Embedded Applications Binary Interface (eabi) which is a set of
15593 modifications to the System V.4 specifications. Selecting @option{-meabi}
15594 means that the stack is aligned to an 8 byte boundary, a function
15595 @code{__eabi} is called to from @code{main} to set up the eabi
15596 environment, and the @option{-msdata} option can use both @code{r2} and
15597 @code{r13} to point to two separate small data areas. Selecting
15598 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
15599 do not call an initialization function from @code{main}, and the
15600 @option{-msdata} option will only use @code{r13} to point to a single
15601 small data area. The @option{-meabi} option is on by default if you
15602 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
15605 @opindex msdata=eabi
15606 On System V.4 and embedded PowerPC systems, put small initialized
15607 @code{const} global and static data in the @samp{.sdata2} section, which
15608 is pointed to by register @code{r2}. Put small initialized
15609 non-@code{const} global and static data in the @samp{.sdata} section,
15610 which is pointed to by register @code{r13}. Put small uninitialized
15611 global and static data in the @samp{.sbss} section, which is adjacent to
15612 the @samp{.sdata} section. The @option{-msdata=eabi} option is
15613 incompatible with the @option{-mrelocatable} option. The
15614 @option{-msdata=eabi} option also sets the @option{-memb} option.
15617 @opindex msdata=sysv
15618 On System V.4 and embedded PowerPC systems, put small global and static
15619 data in the @samp{.sdata} section, which is pointed to by register
15620 @code{r13}. Put small uninitialized global and static data in the
15621 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
15622 The @option{-msdata=sysv} option is incompatible with the
15623 @option{-mrelocatable} option.
15625 @item -msdata=default
15627 @opindex msdata=default
15629 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
15630 compile code the same as @option{-msdata=eabi}, otherwise compile code the
15631 same as @option{-msdata=sysv}.
15634 @opindex msdata=data
15635 On System V.4 and embedded PowerPC systems, put small global
15636 data in the @samp{.sdata} section. Put small uninitialized global
15637 data in the @samp{.sbss} section. Do not use register @code{r13}
15638 to address small data however. This is the default behavior unless
15639 other @option{-msdata} options are used.
15643 @opindex msdata=none
15645 On embedded PowerPC systems, put all initialized global and static data
15646 in the @samp{.data} section, and all uninitialized data in the
15647 @samp{.bss} section.
15649 @item -mblock-move-inline-limit=@var{num}
15650 @opindex mblock-move-inline-limit
15651 Inline all block moves (such as calls to @code{memcpy} or structure
15652 copies) less than or equal to @var{num} bytes. The minimum value for
15653 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
15654 targets. The default value is target-specific.
15658 @cindex smaller data references (PowerPC)
15659 @cindex .sdata/.sdata2 references (PowerPC)
15660 On embedded PowerPC systems, put global and static items less than or
15661 equal to @var{num} bytes into the small data or bss sections instead of
15662 the normal data or bss section. By default, @var{num} is 8. The
15663 @option{-G @var{num}} switch is also passed to the linker.
15664 All modules should be compiled with the same @option{-G @var{num}} value.
15667 @itemx -mno-regnames
15669 @opindex mno-regnames
15670 On System V.4 and embedded PowerPC systems do (do not) emit register
15671 names in the assembly language output using symbolic forms.
15674 @itemx -mno-longcall
15676 @opindex mno-longcall
15677 By default assume that all calls are far away so that a longer more
15678 expensive calling sequence is required. This is required for calls
15679 further than 32 megabytes (33,554,432 bytes) from the current location.
15680 A short call will be generated if the compiler knows
15681 the call cannot be that far away. This setting can be overridden by
15682 the @code{shortcall} function attribute, or by @code{#pragma
15685 Some linkers are capable of detecting out-of-range calls and generating
15686 glue code on the fly. On these systems, long calls are unnecessary and
15687 generate slower code. As of this writing, the AIX linker can do this,
15688 as can the GNU linker for PowerPC/64. It is planned to add this feature
15689 to the GNU linker for 32-bit PowerPC systems as well.
15691 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
15692 callee, L42'', plus a ``branch island'' (glue code). The two target
15693 addresses represent the callee and the ``branch island''. The
15694 Darwin/PPC linker will prefer the first address and generate a ``bl
15695 callee'' if the PPC ``bl'' instruction will reach the callee directly;
15696 otherwise, the linker will generate ``bl L42'' to call the ``branch
15697 island''. The ``branch island'' is appended to the body of the
15698 calling function; it computes the full 32-bit address of the callee
15701 On Mach-O (Darwin) systems, this option directs the compiler emit to
15702 the glue for every direct call, and the Darwin linker decides whether
15703 to use or discard it.
15705 In the future, we may cause GCC to ignore all longcall specifications
15706 when the linker is known to generate glue.
15708 @item -mtls-markers
15709 @itemx -mno-tls-markers
15710 @opindex mtls-markers
15711 @opindex mno-tls-markers
15712 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
15713 specifying the function argument. The relocation allows ld to
15714 reliably associate function call with argument setup instructions for
15715 TLS optimization, which in turn allows gcc to better schedule the
15720 Adds support for multithreading with the @dfn{pthreads} library.
15721 This option sets flags for both the preprocessor and linker.
15726 This option will enable GCC to use the reciprocal estimate and
15727 reciprocal square root estimate instructions with additional
15728 Newton-Raphson steps to increase precision instead of doing a divide or
15729 square root and divide for floating point arguments. You should use
15730 the @option{-ffast-math} option when using @option{-mrecip} (or at
15731 least @option{-funsafe-math-optimizations},
15732 @option{-finite-math-only}, @option{-freciprocal-math} and
15733 @option{-fno-trapping-math}). Note that while the throughput of the
15734 sequence is generally higher than the throughput of the non-reciprocal
15735 instruction, the precision of the sequence can be decreased by up to 2
15736 ulp (i.e. the inverse of 1.0 equals 0.99999994) for reciprocal square
15739 @item -mrecip=@var{opt}
15740 @opindex mrecip=opt
15741 This option allows to control which reciprocal estimate instructions
15742 may be used. @var{opt} is a comma separated list of options, that may
15743 be preceeded by a @code{!} to invert the option:
15744 @code{all}: enable all estimate instructions,
15745 @code{default}: enable the default instructions, equvalent to @option{-mrecip},
15746 @code{none}: disable all estimate instructions, equivalent to @option{-mno-recip};
15747 @code{div}: enable the reciprocal approximation instructions for both single and double precision;
15748 @code{divf}: enable the single precision reciprocal approximation instructions;
15749 @code{divd}: enable the double precision reciprocal approximation instructions;
15750 @code{rsqrt}: enable the reciprocal square root approximation instructions for both single and double precision;
15751 @code{rsqrtf}: enable the single precision reciprocal square root approximation instructions;
15752 @code{rsqrtd}: enable the double precision reciprocal square root approximation instructions;
15754 So for example, @option{-mrecip=all,!rsqrtd} would enable the
15755 all of the reciprocal estimate instructions, except for the
15756 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
15757 which handle the double precision reciprocal square root calculations.
15759 @item -mrecip-precision
15760 @itemx -mno-recip-precision
15761 @opindex mrecip-precision
15762 Assume (do not assume) that the reciprocal estimate instructions
15763 provide higher precision estimates than is mandated by the powerpc
15764 ABI. Selecting @option{-mcpu=power6} or @option{-mcpu=power7}
15765 automatically selects @option{-mrecip-precision}. The double
15766 precision square root estimate instructions are not generated by
15767 default on low precision machines, since they do not provide an
15768 estimate that converges after three steps.
15772 @subsection RX Options
15775 These command line options are defined for RX targets:
15778 @item -m64bit-doubles
15779 @itemx -m32bit-doubles
15780 @opindex m64bit-doubles
15781 @opindex m32bit-doubles
15782 Make the @code{double} data type be 64-bits (@option{-m64bit-doubles})
15783 or 32-bits (@option{-m32bit-doubles}) in size. The default is
15784 @option{-m32bit-doubles}. @emph{Note} RX floating point hardware only
15785 works on 32-bit values, which is why the default is
15786 @option{-m32bit-doubles}.
15792 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
15793 floating point hardware. The default is enabled for the @var{RX600}
15794 series and disabled for the @var{RX200} series.
15796 Floating point instructions will only be generated for 32-bit floating
15797 point values however, so if the @option{-m64bit-doubles} option is in
15798 use then the FPU hardware will not be used for doubles.
15800 @emph{Note} If the @option{-fpu} option is enabled then
15801 @option{-funsafe-math-optimizations} is also enabled automatically.
15802 This is because the RX FPU instructions are themselves unsafe.
15804 @item -mcpu=@var{name}
15805 @itemx -patch=@var{name}
15808 Selects the type of RX CPU to be targeted. Currently three types are
15809 supported, the generic @var{RX600} and @var{RX200} series hardware and
15810 the specific @var{RX610} cpu. The default is @var{RX600}.
15812 The only difference between @var{RX600} and @var{RX610} is that the
15813 @var{RX610} does not support the @code{MVTIPL} instruction.
15815 The @var{RX200} series does not have a hardware floating point unit
15816 and so @option{-nofpu} is enabled by default when this type is
15819 @item -mbig-endian-data
15820 @itemx -mlittle-endian-data
15821 @opindex mbig-endian-data
15822 @opindex mlittle-endian-data
15823 Store data (but not code) in the big-endian format. The default is
15824 @option{-mlittle-endian-data}, ie to store data in the little endian
15827 @item -msmall-data-limit=@var{N}
15828 @opindex msmall-data-limit
15829 Specifies the maximum size in bytes of global and static variables
15830 which can be placed into the small data area. Using the small data
15831 area can lead to smaller and faster code, but the size of area is
15832 limited and it is up to the programmer to ensure that the area does
15833 not overflow. Also when the small data area is used one of the RX's
15834 registers (@code{r13}) is reserved for use pointing to this area, so
15835 it is no longer available for use by the compiler. This could result
15836 in slower and/or larger code if variables which once could have been
15837 held in @code{r13} are now pushed onto the stack.
15839 Note, common variables (variables which have not been initialised) and
15840 constants are not placed into the small data area as they are assigned
15841 to other sections in the output executable.
15843 The default value is zero, which disables this feature. Note, this
15844 feature is not enabled by default with higher optimization levels
15845 (@option{-O2} etc) because of the potentially detrimental effects of
15846 reserving register @code{r13}. It is up to the programmer to
15847 experiment and discover whether this feature is of benefit to their
15854 Use the simulator runtime. The default is to use the libgloss board
15857 @item -mas100-syntax
15858 @itemx -mno-as100-syntax
15859 @opindex mas100-syntax
15860 @opindex mno-as100-syntax
15861 When generating assembler output use a syntax that is compatible with
15862 Renesas's AS100 assembler. This syntax can also be handled by the GAS
15863 assembler but it has some restrictions so generating it is not the
15866 @item -mmax-constant-size=@var{N}
15867 @opindex mmax-constant-size
15868 Specifies the maximum size, in bytes, of a constant that can be used as
15869 an operand in a RX instruction. Although the RX instruction set does
15870 allow constants of up to 4 bytes in length to be used in instructions,
15871 a longer value equates to a longer instruction. Thus in some
15872 circumstances it can be beneficial to restrict the size of constants
15873 that are used in instructions. Constants that are too big are instead
15874 placed into a constant pool and referenced via register indirection.
15876 The value @var{N} can be between 0 and 4. A value of 0 (the default)
15877 or 4 means that constants of any size are allowed.
15881 Enable linker relaxation. Linker relaxation is a process whereby the
15882 linker will attempt to reduce the size of a program by finding shorter
15883 versions of various instructions. Disabled by default.
15885 @item -mint-register=@var{N}
15886 @opindex mint-register
15887 Specify the number of registers to reserve for fast interrupt handler
15888 functions. The value @var{N} can be between 0 and 4. A value of 1
15889 means that register @code{r13} will be reserved for the exclusive use
15890 of fast interrupt handlers. A value of 2 reserves @code{r13} and
15891 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
15892 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
15893 A value of 0, the default, does not reserve any registers.
15895 @item -msave-acc-in-interrupts
15896 @opindex msave-acc-in-interrupts
15897 Specifies that interrupt handler functions should preserve the
15898 accumulator register. This is only necessary if normal code might use
15899 the accumulator register, for example because it performs 64-bit
15900 multiplications. The default is to ignore the accumulator as this
15901 makes the interrupt handlers faster.
15905 @emph{Note:} The generic GCC command line @option{-ffixed-@var{reg}}
15906 has special significance to the RX port when used with the
15907 @code{interrupt} function attribute. This attribute indicates a
15908 function intended to process fast interrupts. GCC will will ensure
15909 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
15910 and/or @code{r13} and only provided that the normal use of the
15911 corresponding registers have been restricted via the
15912 @option{-ffixed-@var{reg}} or @option{-mint-register} command line
15915 @node S/390 and zSeries Options
15916 @subsection S/390 and zSeries Options
15917 @cindex S/390 and zSeries Options
15919 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
15923 @itemx -msoft-float
15924 @opindex mhard-float
15925 @opindex msoft-float
15926 Use (do not use) the hardware floating-point instructions and registers
15927 for floating-point operations. When @option{-msoft-float} is specified,
15928 functions in @file{libgcc.a} will be used to perform floating-point
15929 operations. When @option{-mhard-float} is specified, the compiler
15930 generates IEEE floating-point instructions. This is the default.
15933 @itemx -mno-hard-dfp
15935 @opindex mno-hard-dfp
15936 Use (do not use) the hardware decimal-floating-point instructions for
15937 decimal-floating-point operations. When @option{-mno-hard-dfp} is
15938 specified, functions in @file{libgcc.a} will be used to perform
15939 decimal-floating-point operations. When @option{-mhard-dfp} is
15940 specified, the compiler generates decimal-floating-point hardware
15941 instructions. This is the default for @option{-march=z9-ec} or higher.
15943 @item -mlong-double-64
15944 @itemx -mlong-double-128
15945 @opindex mlong-double-64
15946 @opindex mlong-double-128
15947 These switches control the size of @code{long double} type. A size
15948 of 64bit makes the @code{long double} type equivalent to the @code{double}
15949 type. This is the default.
15952 @itemx -mno-backchain
15953 @opindex mbackchain
15954 @opindex mno-backchain
15955 Store (do not store) the address of the caller's frame as backchain pointer
15956 into the callee's stack frame.
15957 A backchain may be needed to allow debugging using tools that do not understand
15958 DWARF-2 call frame information.
15959 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
15960 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
15961 the backchain is placed into the topmost word of the 96/160 byte register
15964 In general, code compiled with @option{-mbackchain} is call-compatible with
15965 code compiled with @option{-mmo-backchain}; however, use of the backchain
15966 for debugging purposes usually requires that the whole binary is built with
15967 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
15968 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
15969 to build a linux kernel use @option{-msoft-float}.
15971 The default is to not maintain the backchain.
15973 @item -mpacked-stack
15974 @itemx -mno-packed-stack
15975 @opindex mpacked-stack
15976 @opindex mno-packed-stack
15977 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
15978 specified, the compiler uses the all fields of the 96/160 byte register save
15979 area only for their default purpose; unused fields still take up stack space.
15980 When @option{-mpacked-stack} is specified, register save slots are densely
15981 packed at the top of the register save area; unused space is reused for other
15982 purposes, allowing for more efficient use of the available stack space.
15983 However, when @option{-mbackchain} is also in effect, the topmost word of
15984 the save area is always used to store the backchain, and the return address
15985 register is always saved two words below the backchain.
15987 As long as the stack frame backchain is not used, code generated with
15988 @option{-mpacked-stack} is call-compatible with code generated with
15989 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
15990 S/390 or zSeries generated code that uses the stack frame backchain at run
15991 time, not just for debugging purposes. Such code is not call-compatible
15992 with code compiled with @option{-mpacked-stack}. Also, note that the
15993 combination of @option{-mbackchain},
15994 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
15995 to build a linux kernel use @option{-msoft-float}.
15997 The default is to not use the packed stack layout.
16000 @itemx -mno-small-exec
16001 @opindex msmall-exec
16002 @opindex mno-small-exec
16003 Generate (or do not generate) code using the @code{bras} instruction
16004 to do subroutine calls.
16005 This only works reliably if the total executable size does not
16006 exceed 64k. The default is to use the @code{basr} instruction instead,
16007 which does not have this limitation.
16013 When @option{-m31} is specified, generate code compliant to the
16014 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
16015 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
16016 particular to generate 64-bit instructions. For the @samp{s390}
16017 targets, the default is @option{-m31}, while the @samp{s390x}
16018 targets default to @option{-m64}.
16024 When @option{-mzarch} is specified, generate code using the
16025 instructions available on z/Architecture.
16026 When @option{-mesa} is specified, generate code using the
16027 instructions available on ESA/390. Note that @option{-mesa} is
16028 not possible with @option{-m64}.
16029 When generating code compliant to the GNU/Linux for S/390 ABI,
16030 the default is @option{-mesa}. When generating code compliant
16031 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
16037 Generate (or do not generate) code using the @code{mvcle} instruction
16038 to perform block moves. When @option{-mno-mvcle} is specified,
16039 use a @code{mvc} loop instead. This is the default unless optimizing for
16046 Print (or do not print) additional debug information when compiling.
16047 The default is to not print debug information.
16049 @item -march=@var{cpu-type}
16051 Generate code that will run on @var{cpu-type}, which is the name of a system
16052 representing a certain processor type. Possible values for
16053 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
16054 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
16055 When generating code using the instructions available on z/Architecture,
16056 the default is @option{-march=z900}. Otherwise, the default is
16057 @option{-march=g5}.
16059 @item -mtune=@var{cpu-type}
16061 Tune to @var{cpu-type} everything applicable about the generated code,
16062 except for the ABI and the set of available instructions.
16063 The list of @var{cpu-type} values is the same as for @option{-march}.
16064 The default is the value used for @option{-march}.
16067 @itemx -mno-tpf-trace
16068 @opindex mtpf-trace
16069 @opindex mno-tpf-trace
16070 Generate code that adds (does not add) in TPF OS specific branches to trace
16071 routines in the operating system. This option is off by default, even
16072 when compiling for the TPF OS@.
16075 @itemx -mno-fused-madd
16076 @opindex mfused-madd
16077 @opindex mno-fused-madd
16078 Generate code that uses (does not use) the floating point multiply and
16079 accumulate instructions. These instructions are generated by default if
16080 hardware floating point is used.
16082 @item -mwarn-framesize=@var{framesize}
16083 @opindex mwarn-framesize
16084 Emit a warning if the current function exceeds the given frame size. Because
16085 this is a compile time check it doesn't need to be a real problem when the program
16086 runs. It is intended to identify functions which most probably cause
16087 a stack overflow. It is useful to be used in an environment with limited stack
16088 size e.g.@: the linux kernel.
16090 @item -mwarn-dynamicstack
16091 @opindex mwarn-dynamicstack
16092 Emit a warning if the function calls alloca or uses dynamically
16093 sized arrays. This is generally a bad idea with a limited stack size.
16095 @item -mstack-guard=@var{stack-guard}
16096 @itemx -mstack-size=@var{stack-size}
16097 @opindex mstack-guard
16098 @opindex mstack-size
16099 If these options are provided the s390 back end emits additional instructions in
16100 the function prologue which trigger a trap if the stack size is @var{stack-guard}
16101 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
16102 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
16103 the frame size of the compiled function is chosen.
16104 These options are intended to be used to help debugging stack overflow problems.
16105 The additionally emitted code causes only little overhead and hence can also be
16106 used in production like systems without greater performance degradation. The given
16107 values have to be exact powers of 2 and @var{stack-size} has to be greater than
16108 @var{stack-guard} without exceeding 64k.
16109 In order to be efficient the extra code makes the assumption that the stack starts
16110 at an address aligned to the value given by @var{stack-size}.
16111 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
16114 @node Score Options
16115 @subsection Score Options
16116 @cindex Score Options
16118 These options are defined for Score implementations:
16123 Compile code for big endian mode. This is the default.
16127 Compile code for little endian mode.
16131 Disable generate bcnz instruction.
16135 Enable generate unaligned load and store instruction.
16139 Enable the use of multiply-accumulate instructions. Disabled by default.
16143 Specify the SCORE5 as the target architecture.
16147 Specify the SCORE5U of the target architecture.
16151 Specify the SCORE7 as the target architecture. This is the default.
16155 Specify the SCORE7D as the target architecture.
16159 @subsection SH Options
16161 These @samp{-m} options are defined for the SH implementations:
16166 Generate code for the SH1.
16170 Generate code for the SH2.
16173 Generate code for the SH2e.
16177 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
16178 that the floating-point unit is not used.
16180 @item -m2a-single-only
16181 @opindex m2a-single-only
16182 Generate code for the SH2a-FPU, in such a way that no double-precision
16183 floating point operations are used.
16186 @opindex m2a-single
16187 Generate code for the SH2a-FPU assuming the floating-point unit is in
16188 single-precision mode by default.
16192 Generate code for the SH2a-FPU assuming the floating-point unit is in
16193 double-precision mode by default.
16197 Generate code for the SH3.
16201 Generate code for the SH3e.
16205 Generate code for the SH4 without a floating-point unit.
16207 @item -m4-single-only
16208 @opindex m4-single-only
16209 Generate code for the SH4 with a floating-point unit that only
16210 supports single-precision arithmetic.
16214 Generate code for the SH4 assuming the floating-point unit is in
16215 single-precision mode by default.
16219 Generate code for the SH4.
16223 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
16224 floating-point unit is not used.
16226 @item -m4a-single-only
16227 @opindex m4a-single-only
16228 Generate code for the SH4a, in such a way that no double-precision
16229 floating point operations are used.
16232 @opindex m4a-single
16233 Generate code for the SH4a assuming the floating-point unit is in
16234 single-precision mode by default.
16238 Generate code for the SH4a.
16242 Same as @option{-m4a-nofpu}, except that it implicitly passes
16243 @option{-dsp} to the assembler. GCC doesn't generate any DSP
16244 instructions at the moment.
16248 Compile code for the processor in big endian mode.
16252 Compile code for the processor in little endian mode.
16256 Align doubles at 64-bit boundaries. Note that this changes the calling
16257 conventions, and thus some functions from the standard C library will
16258 not work unless you recompile it first with @option{-mdalign}.
16262 Shorten some address references at link time, when possible; uses the
16263 linker option @option{-relax}.
16267 Use 32-bit offsets in @code{switch} tables. The default is to use
16272 Enable the use of bit manipulation instructions on SH2A.
16276 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
16277 alignment constraints.
16281 Comply with the calling conventions defined by Renesas.
16285 Comply with the calling conventions defined by Renesas.
16289 Comply with the calling conventions defined for GCC before the Renesas
16290 conventions were available. This option is the default for all
16291 targets of the SH toolchain except for @samp{sh-symbianelf}.
16294 @opindex mnomacsave
16295 Mark the @code{MAC} register as call-clobbered, even if
16296 @option{-mhitachi} is given.
16300 Increase IEEE-compliance of floating-point code.
16301 At the moment, this is equivalent to @option{-fno-finite-math-only}.
16302 When generating 16 bit SH opcodes, getting IEEE-conforming results for
16303 comparisons of NANs / infinities incurs extra overhead in every
16304 floating point comparison, therefore the default is set to
16305 @option{-ffinite-math-only}.
16307 @item -minline-ic_invalidate
16308 @opindex minline-ic_invalidate
16309 Inline code to invalidate instruction cache entries after setting up
16310 nested function trampolines.
16311 This option has no effect if -musermode is in effect and the selected
16312 code generation option (e.g. -m4) does not allow the use of the icbi
16314 If the selected code generation option does not allow the use of the icbi
16315 instruction, and -musermode is not in effect, the inlined code will
16316 manipulate the instruction cache address array directly with an associative
16317 write. This not only requires privileged mode, but it will also
16318 fail if the cache line had been mapped via the TLB and has become unmapped.
16322 Dump instruction size and location in the assembly code.
16325 @opindex mpadstruct
16326 This option is deprecated. It pads structures to multiple of 4 bytes,
16327 which is incompatible with the SH ABI@.
16331 Optimize for space instead of speed. Implied by @option{-Os}.
16334 @opindex mprefergot
16335 When generating position-independent code, emit function calls using
16336 the Global Offset Table instead of the Procedure Linkage Table.
16340 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
16341 if the inlined code would not work in user mode.
16342 This is the default when the target is @code{sh-*-linux*}.
16344 @item -multcost=@var{number}
16345 @opindex multcost=@var{number}
16346 Set the cost to assume for a multiply insn.
16348 @item -mdiv=@var{strategy}
16349 @opindex mdiv=@var{strategy}
16350 Set the division strategy to use for SHmedia code. @var{strategy} must be
16351 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
16352 inv:call2, inv:fp .
16353 "fp" performs the operation in floating point. This has a very high latency,
16354 but needs only a few instructions, so it might be a good choice if
16355 your code has enough easily exploitable ILP to allow the compiler to
16356 schedule the floating point instructions together with other instructions.
16357 Division by zero causes a floating point exception.
16358 "inv" uses integer operations to calculate the inverse of the divisor,
16359 and then multiplies the dividend with the inverse. This strategy allows
16360 cse and hoisting of the inverse calculation. Division by zero calculates
16361 an unspecified result, but does not trap.
16362 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
16363 have been found, or if the entire operation has been hoisted to the same
16364 place, the last stages of the inverse calculation are intertwined with the
16365 final multiply to reduce the overall latency, at the expense of using a few
16366 more instructions, and thus offering fewer scheduling opportunities with
16368 "call" calls a library function that usually implements the inv:minlat
16370 This gives high code density for m5-*media-nofpu compilations.
16371 "call2" uses a different entry point of the same library function, where it
16372 assumes that a pointer to a lookup table has already been set up, which
16373 exposes the pointer load to cse / code hoisting optimizations.
16374 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
16375 code generation, but if the code stays unoptimized, revert to the "call",
16376 "call2", or "fp" strategies, respectively. Note that the
16377 potentially-trapping side effect of division by zero is carried by a
16378 separate instruction, so it is possible that all the integer instructions
16379 are hoisted out, but the marker for the side effect stays where it is.
16380 A recombination to fp operations or a call is not possible in that case.
16381 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
16382 that the inverse calculation was nor separated from the multiply, they speed
16383 up division where the dividend fits into 20 bits (plus sign where applicable),
16384 by inserting a test to skip a number of operations in this case; this test
16385 slows down the case of larger dividends. inv20u assumes the case of a such
16386 a small dividend to be unlikely, and inv20l assumes it to be likely.
16388 @item -maccumulate-outgoing-args
16389 @opindex maccumulate-outgoing-args
16390 Reserve space once for outgoing arguments in the function prologue rather
16391 than around each call. Generally beneficial for performance and size. Also
16392 needed for unwinding to avoid changing the stack frame around conditional code.
16394 @item -mdivsi3_libfunc=@var{name}
16395 @opindex mdivsi3_libfunc=@var{name}
16396 Set the name of the library function used for 32 bit signed division to
16397 @var{name}. This only affect the name used in the call and inv:call
16398 division strategies, and the compiler will still expect the same
16399 sets of input/output/clobbered registers as if this option was not present.
16401 @item -mfixed-range=@var{register-range}
16402 @opindex mfixed-range
16403 Generate code treating the given register range as fixed registers.
16404 A fixed register is one that the register allocator can not use. This is
16405 useful when compiling kernel code. A register range is specified as
16406 two registers separated by a dash. Multiple register ranges can be
16407 specified separated by a comma.
16409 @item -madjust-unroll
16410 @opindex madjust-unroll
16411 Throttle unrolling to avoid thrashing target registers.
16412 This option only has an effect if the gcc code base supports the
16413 TARGET_ADJUST_UNROLL_MAX target hook.
16415 @item -mindexed-addressing
16416 @opindex mindexed-addressing
16417 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
16418 This is only safe if the hardware and/or OS implement 32 bit wrap-around
16419 semantics for the indexed addressing mode. The architecture allows the
16420 implementation of processors with 64 bit MMU, which the OS could use to
16421 get 32 bit addressing, but since no current hardware implementation supports
16422 this or any other way to make the indexed addressing mode safe to use in
16423 the 32 bit ABI, the default is -mno-indexed-addressing.
16425 @item -mgettrcost=@var{number}
16426 @opindex mgettrcost=@var{number}
16427 Set the cost assumed for the gettr instruction to @var{number}.
16428 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
16432 Assume pt* instructions won't trap. This will generally generate better
16433 scheduled code, but is unsafe on current hardware. The current architecture
16434 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
16435 This has the unintentional effect of making it unsafe to schedule ptabs /
16436 ptrel before a branch, or hoist it out of a loop. For example,
16437 __do_global_ctors, a part of libgcc that runs constructors at program
16438 startup, calls functions in a list which is delimited by @minus{}1. With the
16439 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
16440 That means that all the constructors will be run a bit quicker, but when
16441 the loop comes to the end of the list, the program crashes because ptabs
16442 loads @minus{}1 into a target register. Since this option is unsafe for any
16443 hardware implementing the current architecture specification, the default
16444 is -mno-pt-fixed. Unless the user specifies a specific cost with
16445 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
16446 this deters register allocation using target registers for storing
16449 @item -minvalid-symbols
16450 @opindex minvalid-symbols
16451 Assume symbols might be invalid. Ordinary function symbols generated by
16452 the compiler will always be valid to load with movi/shori/ptabs or
16453 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
16454 to generate symbols that will cause ptabs / ptrel to trap.
16455 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
16456 It will then prevent cross-basic-block cse, hoisting and most scheduling
16457 of symbol loads. The default is @option{-mno-invalid-symbols}.
16460 @node SPARC Options
16461 @subsection SPARC Options
16462 @cindex SPARC options
16464 These @samp{-m} options are supported on the SPARC:
16467 @item -mno-app-regs
16469 @opindex mno-app-regs
16471 Specify @option{-mapp-regs} to generate output using the global registers
16472 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
16475 To be fully SVR4 ABI compliant at the cost of some performance loss,
16476 specify @option{-mno-app-regs}. You should compile libraries and system
16477 software with this option.
16480 @itemx -mhard-float
16482 @opindex mhard-float
16483 Generate output containing floating point instructions. This is the
16487 @itemx -msoft-float
16489 @opindex msoft-float
16490 Generate output containing library calls for floating point.
16491 @strong{Warning:} the requisite libraries are not available for all SPARC
16492 targets. Normally the facilities of the machine's usual C compiler are
16493 used, but this cannot be done directly in cross-compilation. You must make
16494 your own arrangements to provide suitable library functions for
16495 cross-compilation. The embedded targets @samp{sparc-*-aout} and
16496 @samp{sparclite-*-*} do provide software floating point support.
16498 @option{-msoft-float} changes the calling convention in the output file;
16499 therefore, it is only useful if you compile @emph{all} of a program with
16500 this option. In particular, you need to compile @file{libgcc.a}, the
16501 library that comes with GCC, with @option{-msoft-float} in order for
16504 @item -mhard-quad-float
16505 @opindex mhard-quad-float
16506 Generate output containing quad-word (long double) floating point
16509 @item -msoft-quad-float
16510 @opindex msoft-quad-float
16511 Generate output containing library calls for quad-word (long double)
16512 floating point instructions. The functions called are those specified
16513 in the SPARC ABI@. This is the default.
16515 As of this writing, there are no SPARC implementations that have hardware
16516 support for the quad-word floating point instructions. They all invoke
16517 a trap handler for one of these instructions, and then the trap handler
16518 emulates the effect of the instruction. Because of the trap handler overhead,
16519 this is much slower than calling the ABI library routines. Thus the
16520 @option{-msoft-quad-float} option is the default.
16522 @item -mno-unaligned-doubles
16523 @itemx -munaligned-doubles
16524 @opindex mno-unaligned-doubles
16525 @opindex munaligned-doubles
16526 Assume that doubles have 8 byte alignment. This is the default.
16528 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
16529 alignment only if they are contained in another type, or if they have an
16530 absolute address. Otherwise, it assumes they have 4 byte alignment.
16531 Specifying this option avoids some rare compatibility problems with code
16532 generated by other compilers. It is not the default because it results
16533 in a performance loss, especially for floating point code.
16535 @item -mno-faster-structs
16536 @itemx -mfaster-structs
16537 @opindex mno-faster-structs
16538 @opindex mfaster-structs
16539 With @option{-mfaster-structs}, the compiler assumes that structures
16540 should have 8 byte alignment. This enables the use of pairs of
16541 @code{ldd} and @code{std} instructions for copies in structure
16542 assignment, in place of twice as many @code{ld} and @code{st} pairs.
16543 However, the use of this changed alignment directly violates the SPARC
16544 ABI@. Thus, it's intended only for use on targets where the developer
16545 acknowledges that their resulting code will not be directly in line with
16546 the rules of the ABI@.
16548 @item -mimpure-text
16549 @opindex mimpure-text
16550 @option{-mimpure-text}, used in addition to @option{-shared}, tells
16551 the compiler to not pass @option{-z text} to the linker when linking a
16552 shared object. Using this option, you can link position-dependent
16553 code into a shared object.
16555 @option{-mimpure-text} suppresses the ``relocations remain against
16556 allocatable but non-writable sections'' linker error message.
16557 However, the necessary relocations will trigger copy-on-write, and the
16558 shared object is not actually shared across processes. Instead of
16559 using @option{-mimpure-text}, you should compile all source code with
16560 @option{-fpic} or @option{-fPIC}.
16562 This option is only available on SunOS and Solaris.
16564 @item -mcpu=@var{cpu_type}
16566 Set the instruction set, register set, and instruction scheduling parameters
16567 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
16568 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
16569 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
16570 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
16571 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
16573 Default instruction scheduling parameters are used for values that select
16574 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
16575 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
16577 Here is a list of each supported architecture and their supported
16582 v8: supersparc, hypersparc
16583 sparclite: f930, f934, sparclite86x
16585 v9: ultrasparc, ultrasparc3, niagara, niagara2
16588 By default (unless configured otherwise), GCC generates code for the V7
16589 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
16590 additionally optimizes it for the Cypress CY7C602 chip, as used in the
16591 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
16592 SPARCStation 1, 2, IPX etc.
16594 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
16595 architecture. The only difference from V7 code is that the compiler emits
16596 the integer multiply and integer divide instructions which exist in SPARC-V8
16597 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
16598 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
16601 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
16602 the SPARC architecture. This adds the integer multiply, integer divide step
16603 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
16604 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
16605 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
16606 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
16607 MB86934 chip, which is the more recent SPARClite with FPU@.
16609 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
16610 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
16611 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
16612 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
16613 optimizes it for the TEMIC SPARClet chip.
16615 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
16616 architecture. This adds 64-bit integer and floating-point move instructions,
16617 3 additional floating-point condition code registers and conditional move
16618 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
16619 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
16620 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
16621 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
16622 @option{-mcpu=niagara}, the compiler additionally optimizes it for
16623 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
16624 additionally optimizes it for Sun UltraSPARC T2 chips.
16626 @item -mtune=@var{cpu_type}
16628 Set the instruction scheduling parameters for machine type
16629 @var{cpu_type}, but do not set the instruction set or register set that the
16630 option @option{-mcpu=@var{cpu_type}} would.
16632 The same values for @option{-mcpu=@var{cpu_type}} can be used for
16633 @option{-mtune=@var{cpu_type}}, but the only useful values are those
16634 that select a particular cpu implementation. Those are @samp{cypress},
16635 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
16636 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
16637 @samp{ultrasparc3}, @samp{niagara}, and @samp{niagara2}.
16642 @opindex mno-v8plus
16643 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
16644 difference from the V8 ABI is that the global and out registers are
16645 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
16646 mode for all SPARC-V9 processors.
16652 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
16653 Visual Instruction Set extensions. The default is @option{-mno-vis}.
16656 These @samp{-m} options are supported in addition to the above
16657 on SPARC-V9 processors in 64-bit environments:
16660 @item -mlittle-endian
16661 @opindex mlittle-endian
16662 Generate code for a processor running in little-endian mode. It is only
16663 available for a few configurations and most notably not on Solaris and Linux.
16669 Generate code for a 32-bit or 64-bit environment.
16670 The 32-bit environment sets int, long and pointer to 32 bits.
16671 The 64-bit environment sets int to 32 bits and long and pointer
16674 @item -mcmodel=medlow
16675 @opindex mcmodel=medlow
16676 Generate code for the Medium/Low code model: 64-bit addresses, programs
16677 must be linked in the low 32 bits of memory. Programs can be statically
16678 or dynamically linked.
16680 @item -mcmodel=medmid
16681 @opindex mcmodel=medmid
16682 Generate code for the Medium/Middle code model: 64-bit addresses, programs
16683 must be linked in the low 44 bits of memory, the text and data segments must
16684 be less than 2GB in size and the data segment must be located within 2GB of
16687 @item -mcmodel=medany
16688 @opindex mcmodel=medany
16689 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
16690 may be linked anywhere in memory, the text and data segments must be less
16691 than 2GB in size and the data segment must be located within 2GB of the
16694 @item -mcmodel=embmedany
16695 @opindex mcmodel=embmedany
16696 Generate code for the Medium/Anywhere code model for embedded systems:
16697 64-bit addresses, the text and data segments must be less than 2GB in
16698 size, both starting anywhere in memory (determined at link time). The
16699 global register %g4 points to the base of the data segment. Programs
16700 are statically linked and PIC is not supported.
16703 @itemx -mno-stack-bias
16704 @opindex mstack-bias
16705 @opindex mno-stack-bias
16706 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
16707 frame pointer if present, are offset by @minus{}2047 which must be added back
16708 when making stack frame references. This is the default in 64-bit mode.
16709 Otherwise, assume no such offset is present.
16712 These switches are supported in addition to the above on Solaris:
16717 Add support for multithreading using the Solaris threads library. This
16718 option sets flags for both the preprocessor and linker. This option does
16719 not affect the thread safety of object code produced by the compiler or
16720 that of libraries supplied with it.
16724 Add support for multithreading using the POSIX threads library. This
16725 option sets flags for both the preprocessor and linker. This option does
16726 not affect the thread safety of object code produced by the compiler or
16727 that of libraries supplied with it.
16731 This is a synonym for @option{-pthreads}.
16735 @subsection SPU Options
16736 @cindex SPU options
16738 These @samp{-m} options are supported on the SPU:
16742 @itemx -merror-reloc
16743 @opindex mwarn-reloc
16744 @opindex merror-reloc
16746 The loader for SPU does not handle dynamic relocations. By default, GCC
16747 will give an error when it generates code that requires a dynamic
16748 relocation. @option{-mno-error-reloc} disables the error,
16749 @option{-mwarn-reloc} will generate a warning instead.
16752 @itemx -munsafe-dma
16754 @opindex munsafe-dma
16756 Instructions which initiate or test completion of DMA must not be
16757 reordered with respect to loads and stores of the memory which is being
16758 accessed. Users typically address this problem using the volatile
16759 keyword, but that can lead to inefficient code in places where the
16760 memory is known to not change. Rather than mark the memory as volatile
16761 we treat the DMA instructions as potentially effecting all memory. With
16762 @option{-munsafe-dma} users must use the volatile keyword to protect
16765 @item -mbranch-hints
16766 @opindex mbranch-hints
16768 By default, GCC will generate a branch hint instruction to avoid
16769 pipeline stalls for always taken or probably taken branches. A hint
16770 will not be generated closer than 8 instructions away from its branch.
16771 There is little reason to disable them, except for debugging purposes,
16772 or to make an object a little bit smaller.
16776 @opindex msmall-mem
16777 @opindex mlarge-mem
16779 By default, GCC generates code assuming that addresses are never larger
16780 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
16781 a full 32 bit address.
16786 By default, GCC links against startup code that assumes the SPU-style
16787 main function interface (which has an unconventional parameter list).
16788 With @option{-mstdmain}, GCC will link your program against startup
16789 code that assumes a C99-style interface to @code{main}, including a
16790 local copy of @code{argv} strings.
16792 @item -mfixed-range=@var{register-range}
16793 @opindex mfixed-range
16794 Generate code treating the given register range as fixed registers.
16795 A fixed register is one that the register allocator can not use. This is
16796 useful when compiling kernel code. A register range is specified as
16797 two registers separated by a dash. Multiple register ranges can be
16798 specified separated by a comma.
16804 Compile code assuming that pointers to the PPU address space accessed
16805 via the @code{__ea} named address space qualifier are either 32 or 64
16806 bits wide. The default is 32 bits. As this is an ABI changing option,
16807 all object code in an executable must be compiled with the same setting.
16809 @item -maddress-space-conversion
16810 @itemx -mno-address-space-conversion
16811 @opindex maddress-space-conversion
16812 @opindex mno-address-space-conversion
16813 Allow/disallow treating the @code{__ea} address space as superset
16814 of the generic address space. This enables explicit type casts
16815 between @code{__ea} and generic pointer as well as implicit
16816 conversions of generic pointers to @code{__ea} pointers. The
16817 default is to allow address space pointer conversions.
16819 @item -mcache-size=@var{cache-size}
16820 @opindex mcache-size
16821 This option controls the version of libgcc that the compiler links to an
16822 executable and selects a software-managed cache for accessing variables
16823 in the @code{__ea} address space with a particular cache size. Possible
16824 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
16825 and @samp{128}. The default cache size is 64KB.
16827 @item -matomic-updates
16828 @itemx -mno-atomic-updates
16829 @opindex matomic-updates
16830 @opindex mno-atomic-updates
16831 This option controls the version of libgcc that the compiler links to an
16832 executable and selects whether atomic updates to the software-managed
16833 cache of PPU-side variables are used. If you use atomic updates, changes
16834 to a PPU variable from SPU code using the @code{__ea} named address space
16835 qualifier will not interfere with changes to other PPU variables residing
16836 in the same cache line from PPU code. If you do not use atomic updates,
16837 such interference may occur; however, writing back cache lines will be
16838 more efficient. The default behavior is to use atomic updates.
16841 @itemx -mdual-nops=@var{n}
16842 @opindex mdual-nops
16843 By default, GCC will insert nops to increase dual issue when it expects
16844 it to increase performance. @var{n} can be a value from 0 to 10. A
16845 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
16846 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
16848 @item -mhint-max-nops=@var{n}
16849 @opindex mhint-max-nops
16850 Maximum number of nops to insert for a branch hint. A branch hint must
16851 be at least 8 instructions away from the branch it is effecting. GCC
16852 will insert up to @var{n} nops to enforce this, otherwise it will not
16853 generate the branch hint.
16855 @item -mhint-max-distance=@var{n}
16856 @opindex mhint-max-distance
16857 The encoding of the branch hint instruction limits the hint to be within
16858 256 instructions of the branch it is effecting. By default, GCC makes
16859 sure it is within 125.
16862 @opindex msafe-hints
16863 Work around a hardware bug which causes the SPU to stall indefinitely.
16864 By default, GCC will insert the @code{hbrp} instruction to make sure
16865 this stall won't happen.
16869 @node System V Options
16870 @subsection Options for System V
16872 These additional options are available on System V Release 4 for
16873 compatibility with other compilers on those systems:
16878 Create a shared object.
16879 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
16883 Identify the versions of each tool used by the compiler, in a
16884 @code{.ident} assembler directive in the output.
16888 Refrain from adding @code{.ident} directives to the output file (this is
16891 @item -YP,@var{dirs}
16893 Search the directories @var{dirs}, and no others, for libraries
16894 specified with @option{-l}.
16896 @item -Ym,@var{dir}
16898 Look in the directory @var{dir} to find the M4 preprocessor.
16899 The assembler uses this option.
16900 @c This is supposed to go with a -Yd for predefined M4 macro files, but
16901 @c the generic assembler that comes with Solaris takes just -Ym.
16905 @subsection V850 Options
16906 @cindex V850 Options
16908 These @samp{-m} options are defined for V850 implementations:
16912 @itemx -mno-long-calls
16913 @opindex mlong-calls
16914 @opindex mno-long-calls
16915 Treat all calls as being far away (near). If calls are assumed to be
16916 far away, the compiler will always load the functions address up into a
16917 register, and call indirect through the pointer.
16923 Do not optimize (do optimize) basic blocks that use the same index
16924 pointer 4 or more times to copy pointer into the @code{ep} register, and
16925 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
16926 option is on by default if you optimize.
16928 @item -mno-prolog-function
16929 @itemx -mprolog-function
16930 @opindex mno-prolog-function
16931 @opindex mprolog-function
16932 Do not use (do use) external functions to save and restore registers
16933 at the prologue and epilogue of a function. The external functions
16934 are slower, but use less code space if more than one function saves
16935 the same number of registers. The @option{-mprolog-function} option
16936 is on by default if you optimize.
16940 Try to make the code as small as possible. At present, this just turns
16941 on the @option{-mep} and @option{-mprolog-function} options.
16943 @item -mtda=@var{n}
16945 Put static or global variables whose size is @var{n} bytes or less into
16946 the tiny data area that register @code{ep} points to. The tiny data
16947 area can hold up to 256 bytes in total (128 bytes for byte references).
16949 @item -msda=@var{n}
16951 Put static or global variables whose size is @var{n} bytes or less into
16952 the small data area that register @code{gp} points to. The small data
16953 area can hold up to 64 kilobytes.
16955 @item -mzda=@var{n}
16957 Put static or global variables whose size is @var{n} bytes or less into
16958 the first 32 kilobytes of memory.
16962 Specify that the target processor is the V850.
16965 @opindex mbig-switch
16966 Generate code suitable for big switch tables. Use this option only if
16967 the assembler/linker complain about out of range branches within a switch
16972 This option will cause r2 and r5 to be used in the code generated by
16973 the compiler. This setting is the default.
16975 @item -mno-app-regs
16976 @opindex mno-app-regs
16977 This option will cause r2 and r5 to be treated as fixed registers.
16981 Specify that the target processor is the V850E2V3. The preprocessor
16982 constants @samp{__v850e2v3__} will be defined if
16983 this option is used.
16987 Specify that the target processor is the V850E2. The preprocessor
16988 constants @samp{__v850e2__} will be defined if
16992 Specify that the target processor is the V850E1. The preprocessor
16993 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
16997 Specify that the target processor is the V850E@. The preprocessor
16998 constant @samp{__v850e__} will be defined if this option is used.
17000 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
17001 nor @option{-mv850e2} nor @option{-mv850e2v3}
17002 are defined then a default target processor will be chosen and the
17003 relevant @samp{__v850*__} preprocessor constant will be defined.
17005 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
17006 defined, regardless of which processor variant is the target.
17008 @item -mdisable-callt
17009 @opindex mdisable-callt
17010 This option will suppress generation of the CALLT instruction for the
17011 v850e, v850e1, v850e2 and v850e2v3 flavors of the v850 architecture. The default is
17012 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
17017 @subsection VAX Options
17018 @cindex VAX options
17020 These @samp{-m} options are defined for the VAX:
17025 Do not output certain jump instructions (@code{aobleq} and so on)
17026 that the Unix assembler for the VAX cannot handle across long
17031 Do output those jump instructions, on the assumption that you
17032 will assemble with the GNU assembler.
17036 Output code for g-format floating point numbers instead of d-format.
17039 @node VxWorks Options
17040 @subsection VxWorks Options
17041 @cindex VxWorks Options
17043 The options in this section are defined for all VxWorks targets.
17044 Options specific to the target hardware are listed with the other
17045 options for that target.
17050 GCC can generate code for both VxWorks kernels and real time processes
17051 (RTPs). This option switches from the former to the latter. It also
17052 defines the preprocessor macro @code{__RTP__}.
17055 @opindex non-static
17056 Link an RTP executable against shared libraries rather than static
17057 libraries. The options @option{-static} and @option{-shared} can
17058 also be used for RTPs (@pxref{Link Options}); @option{-static}
17065 These options are passed down to the linker. They are defined for
17066 compatibility with Diab.
17069 @opindex Xbind-lazy
17070 Enable lazy binding of function calls. This option is equivalent to
17071 @option{-Wl,-z,now} and is defined for compatibility with Diab.
17075 Disable lazy binding of function calls. This option is the default and
17076 is defined for compatibility with Diab.
17079 @node x86-64 Options
17080 @subsection x86-64 Options
17081 @cindex x86-64 options
17083 These are listed under @xref{i386 and x86-64 Options}.
17085 @node i386 and x86-64 Windows Options
17086 @subsection i386 and x86-64 Windows Options
17087 @cindex i386 and x86-64 Windows Options
17089 These additional options are available for Windows targets:
17094 This option is available for Cygwin and MinGW targets. It
17095 specifies that a console application is to be generated, by
17096 instructing the linker to set the PE header subsystem type
17097 required for console applications.
17098 This is the default behavior for Cygwin and MinGW targets.
17102 This option is available for Cygwin targets. It specifies that
17103 the Cygwin internal interface is to be used for predefined
17104 preprocessor macros, C runtime libraries and related linker
17105 paths and options. For Cygwin targets this is the default behavior.
17106 This option is deprecated and will be removed in a future release.
17109 @opindex mno-cygwin
17110 This option is available for Cygwin targets. It specifies that
17111 the MinGW internal interface is to be used instead of Cygwin's, by
17112 setting MinGW-related predefined macros and linker paths and default
17114 This option is deprecated and will be removed in a future release.
17118 This option is available for Cygwin and MinGW targets. It
17119 specifies that a DLL - a dynamic link library - is to be
17120 generated, enabling the selection of the required runtime
17121 startup object and entry point.
17123 @item -mnop-fun-dllimport
17124 @opindex mnop-fun-dllimport
17125 This option is available for Cygwin and MinGW targets. It
17126 specifies that the dllimport attribute should be ignored.
17130 This option is available for MinGW targets. It specifies
17131 that MinGW-specific thread support is to be used.
17135 This option is available for mingw-w64 targets. It specifies
17136 that the UNICODE macro is getting pre-defined and that the
17137 unicode capable runtime startup code is chosen.
17141 This option is available for Cygwin and MinGW targets. It
17142 specifies that the typical Windows pre-defined macros are to
17143 be set in the pre-processor, but does not influence the choice
17144 of runtime library/startup code.
17148 This option is available for Cygwin and MinGW targets. It
17149 specifies that a GUI application is to be generated by
17150 instructing the linker to set the PE header subsystem type
17153 @item -fno-set-stack-executable
17154 @opindex fno-set-stack-executable
17155 This option is available for MinGW targets. It specifies that
17156 the executable flag for stack used by nested functions isn't
17157 set. This is necessary for binaries running in kernel mode of
17158 Windows, as there the user32 API, which is used to set executable
17159 privileges, isn't available.
17161 @item -mpe-aligned-commons
17162 @opindex mpe-aligned-commons
17163 This option is available for Cygwin and MinGW targets. It
17164 specifies that the GNU extension to the PE file format that
17165 permits the correct alignment of COMMON variables should be
17166 used when generating code. It will be enabled by default if
17167 GCC detects that the target assembler found during configuration
17168 supports the feature.
17171 See also under @ref{i386 and x86-64 Options} for standard options.
17173 @node Xstormy16 Options
17174 @subsection Xstormy16 Options
17175 @cindex Xstormy16 Options
17177 These options are defined for Xstormy16:
17182 Choose startup files and linker script suitable for the simulator.
17185 @node Xtensa Options
17186 @subsection Xtensa Options
17187 @cindex Xtensa Options
17189 These options are supported for Xtensa targets:
17193 @itemx -mno-const16
17195 @opindex mno-const16
17196 Enable or disable use of @code{CONST16} instructions for loading
17197 constant values. The @code{CONST16} instruction is currently not a
17198 standard option from Tensilica. When enabled, @code{CONST16}
17199 instructions are always used in place of the standard @code{L32R}
17200 instructions. The use of @code{CONST16} is enabled by default only if
17201 the @code{L32R} instruction is not available.
17204 @itemx -mno-fused-madd
17205 @opindex mfused-madd
17206 @opindex mno-fused-madd
17207 Enable or disable use of fused multiply/add and multiply/subtract
17208 instructions in the floating-point option. This has no effect if the
17209 floating-point option is not also enabled. Disabling fused multiply/add
17210 and multiply/subtract instructions forces the compiler to use separate
17211 instructions for the multiply and add/subtract operations. This may be
17212 desirable in some cases where strict IEEE 754-compliant results are
17213 required: the fused multiply add/subtract instructions do not round the
17214 intermediate result, thereby producing results with @emph{more} bits of
17215 precision than specified by the IEEE standard. Disabling fused multiply
17216 add/subtract instructions also ensures that the program output is not
17217 sensitive to the compiler's ability to combine multiply and add/subtract
17220 @item -mserialize-volatile
17221 @itemx -mno-serialize-volatile
17222 @opindex mserialize-volatile
17223 @opindex mno-serialize-volatile
17224 When this option is enabled, GCC inserts @code{MEMW} instructions before
17225 @code{volatile} memory references to guarantee sequential consistency.
17226 The default is @option{-mserialize-volatile}. Use
17227 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
17229 @item -mforce-no-pic
17230 @opindex mforce-no-pic
17231 For targets, like GNU/Linux, where all user-mode Xtensa code must be
17232 position-independent code (PIC), this option disables PIC for compiling
17235 @item -mtext-section-literals
17236 @itemx -mno-text-section-literals
17237 @opindex mtext-section-literals
17238 @opindex mno-text-section-literals
17239 Control the treatment of literal pools. The default is
17240 @option{-mno-text-section-literals}, which places literals in a separate
17241 section in the output file. This allows the literal pool to be placed
17242 in a data RAM/ROM, and it also allows the linker to combine literal
17243 pools from separate object files to remove redundant literals and
17244 improve code size. With @option{-mtext-section-literals}, the literals
17245 are interspersed in the text section in order to keep them as close as
17246 possible to their references. This may be necessary for large assembly
17249 @item -mtarget-align
17250 @itemx -mno-target-align
17251 @opindex mtarget-align
17252 @opindex mno-target-align
17253 When this option is enabled, GCC instructs the assembler to
17254 automatically align instructions to reduce branch penalties at the
17255 expense of some code density. The assembler attempts to widen density
17256 instructions to align branch targets and the instructions following call
17257 instructions. If there are not enough preceding safe density
17258 instructions to align a target, no widening will be performed. The
17259 default is @option{-mtarget-align}. These options do not affect the
17260 treatment of auto-aligned instructions like @code{LOOP}, which the
17261 assembler will always align, either by widening density instructions or
17262 by inserting no-op instructions.
17265 @itemx -mno-longcalls
17266 @opindex mlongcalls
17267 @opindex mno-longcalls
17268 When this option is enabled, GCC instructs the assembler to translate
17269 direct calls to indirect calls unless it can determine that the target
17270 of a direct call is in the range allowed by the call instruction. This
17271 translation typically occurs for calls to functions in other source
17272 files. Specifically, the assembler translates a direct @code{CALL}
17273 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
17274 The default is @option{-mno-longcalls}. This option should be used in
17275 programs where the call target can potentially be out of range. This
17276 option is implemented in the assembler, not the compiler, so the
17277 assembly code generated by GCC will still show direct call
17278 instructions---look at the disassembled object code to see the actual
17279 instructions. Note that the assembler will use an indirect call for
17280 every cross-file call, not just those that really will be out of range.
17283 @node zSeries Options
17284 @subsection zSeries Options
17285 @cindex zSeries options
17287 These are listed under @xref{S/390 and zSeries Options}.
17289 @node Code Gen Options
17290 @section Options for Code Generation Conventions
17291 @cindex code generation conventions
17292 @cindex options, code generation
17293 @cindex run-time options
17295 These machine-independent options control the interface conventions
17296 used in code generation.
17298 Most of them have both positive and negative forms; the negative form
17299 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
17300 one of the forms is listed---the one which is not the default. You
17301 can figure out the other form by either removing @samp{no-} or adding
17305 @item -fbounds-check
17306 @opindex fbounds-check
17307 For front-ends that support it, generate additional code to check that
17308 indices used to access arrays are within the declared range. This is
17309 currently only supported by the Java and Fortran front-ends, where
17310 this option defaults to true and false respectively.
17314 This option generates traps for signed overflow on addition, subtraction,
17315 multiplication operations.
17319 This option instructs the compiler to assume that signed arithmetic
17320 overflow of addition, subtraction and multiplication wraps around
17321 using twos-complement representation. This flag enables some optimizations
17322 and disables others. This option is enabled by default for the Java
17323 front-end, as required by the Java language specification.
17326 @opindex fexceptions
17327 Enable exception handling. Generates extra code needed to propagate
17328 exceptions. For some targets, this implies GCC will generate frame
17329 unwind information for all functions, which can produce significant data
17330 size overhead, although it does not affect execution. If you do not
17331 specify this option, GCC will enable it by default for languages like
17332 C++ which normally require exception handling, and disable it for
17333 languages like C that do not normally require it. However, you may need
17334 to enable this option when compiling C code that needs to interoperate
17335 properly with exception handlers written in C++. You may also wish to
17336 disable this option if you are compiling older C++ programs that don't
17337 use exception handling.
17339 @item -fnon-call-exceptions
17340 @opindex fnon-call-exceptions
17341 Generate code that allows trapping instructions to throw exceptions.
17342 Note that this requires platform-specific runtime support that does
17343 not exist everywhere. Moreover, it only allows @emph{trapping}
17344 instructions to throw exceptions, i.e.@: memory references or floating
17345 point instructions. It does not allow exceptions to be thrown from
17346 arbitrary signal handlers such as @code{SIGALRM}.
17348 @item -funwind-tables
17349 @opindex funwind-tables
17350 Similar to @option{-fexceptions}, except that it will just generate any needed
17351 static data, but will not affect the generated code in any other way.
17352 You will normally not enable this option; instead, a language processor
17353 that needs this handling would enable it on your behalf.
17355 @item -fasynchronous-unwind-tables
17356 @opindex fasynchronous-unwind-tables
17357 Generate unwind table in dwarf2 format, if supported by target machine. The
17358 table is exact at each instruction boundary, so it can be used for stack
17359 unwinding from asynchronous events (such as debugger or garbage collector).
17361 @item -fpcc-struct-return
17362 @opindex fpcc-struct-return
17363 Return ``short'' @code{struct} and @code{union} values in memory like
17364 longer ones, rather than in registers. This convention is less
17365 efficient, but it has the advantage of allowing intercallability between
17366 GCC-compiled files and files compiled with other compilers, particularly
17367 the Portable C Compiler (pcc).
17369 The precise convention for returning structures in memory depends
17370 on the target configuration macros.
17372 Short structures and unions are those whose size and alignment match
17373 that of some integer type.
17375 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
17376 switch is not binary compatible with code compiled with the
17377 @option{-freg-struct-return} switch.
17378 Use it to conform to a non-default application binary interface.
17380 @item -freg-struct-return
17381 @opindex freg-struct-return
17382 Return @code{struct} and @code{union} values in registers when possible.
17383 This is more efficient for small structures than
17384 @option{-fpcc-struct-return}.
17386 If you specify neither @option{-fpcc-struct-return} nor
17387 @option{-freg-struct-return}, GCC defaults to whichever convention is
17388 standard for the target. If there is no standard convention, GCC
17389 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
17390 the principal compiler. In those cases, we can choose the standard, and
17391 we chose the more efficient register return alternative.
17393 @strong{Warning:} code compiled with the @option{-freg-struct-return}
17394 switch is not binary compatible with code compiled with the
17395 @option{-fpcc-struct-return} switch.
17396 Use it to conform to a non-default application binary interface.
17398 @item -fshort-enums
17399 @opindex fshort-enums
17400 Allocate to an @code{enum} type only as many bytes as it needs for the
17401 declared range of possible values. Specifically, the @code{enum} type
17402 will be equivalent to the smallest integer type which has enough room.
17404 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
17405 code that is not binary compatible with code generated without that switch.
17406 Use it to conform to a non-default application binary interface.
17408 @item -fshort-double
17409 @opindex fshort-double
17410 Use the same size for @code{double} as for @code{float}.
17412 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
17413 code that is not binary compatible with code generated without that switch.
17414 Use it to conform to a non-default application binary interface.
17416 @item -fshort-wchar
17417 @opindex fshort-wchar
17418 Override the underlying type for @samp{wchar_t} to be @samp{short
17419 unsigned int} instead of the default for the target. This option is
17420 useful for building programs to run under WINE@.
17422 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
17423 code that is not binary compatible with code generated without that switch.
17424 Use it to conform to a non-default application binary interface.
17427 @opindex fno-common
17428 In C code, controls the placement of uninitialized global variables.
17429 Unix C compilers have traditionally permitted multiple definitions of
17430 such variables in different compilation units by placing the variables
17432 This is the behavior specified by @option{-fcommon}, and is the default
17433 for GCC on most targets.
17434 On the other hand, this behavior is not required by ISO C, and on some
17435 targets may carry a speed or code size penalty on variable references.
17436 The @option{-fno-common} option specifies that the compiler should place
17437 uninitialized global variables in the data section of the object file,
17438 rather than generating them as common blocks.
17439 This has the effect that if the same variable is declared
17440 (without @code{extern}) in two different compilations,
17441 you will get a multiple-definition error when you link them.
17442 In this case, you must compile with @option{-fcommon} instead.
17443 Compiling with @option{-fno-common} is useful on targets for which
17444 it provides better performance, or if you wish to verify that the
17445 program will work on other systems which always treat uninitialized
17446 variable declarations this way.
17450 Ignore the @samp{#ident} directive.
17452 @item -finhibit-size-directive
17453 @opindex finhibit-size-directive
17454 Don't output a @code{.size} assembler directive, or anything else that
17455 would cause trouble if the function is split in the middle, and the
17456 two halves are placed at locations far apart in memory. This option is
17457 used when compiling @file{crtstuff.c}; you should not need to use it
17460 @item -fverbose-asm
17461 @opindex fverbose-asm
17462 Put extra commentary information in the generated assembly code to
17463 make it more readable. This option is generally only of use to those
17464 who actually need to read the generated assembly code (perhaps while
17465 debugging the compiler itself).
17467 @option{-fno-verbose-asm}, the default, causes the
17468 extra information to be omitted and is useful when comparing two assembler
17471 @item -frecord-gcc-switches
17472 @opindex frecord-gcc-switches
17473 This switch causes the command line that was used to invoke the
17474 compiler to be recorded into the object file that is being created.
17475 This switch is only implemented on some targets and the exact format
17476 of the recording is target and binary file format dependent, but it
17477 usually takes the form of a section containing ASCII text. This
17478 switch is related to the @option{-fverbose-asm} switch, but that
17479 switch only records information in the assembler output file as
17480 comments, so it never reaches the object file.
17484 @cindex global offset table
17486 Generate position-independent code (PIC) suitable for use in a shared
17487 library, if supported for the target machine. Such code accesses all
17488 constant addresses through a global offset table (GOT)@. The dynamic
17489 loader resolves the GOT entries when the program starts (the dynamic
17490 loader is not part of GCC; it is part of the operating system). If
17491 the GOT size for the linked executable exceeds a machine-specific
17492 maximum size, you get an error message from the linker indicating that
17493 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
17494 instead. (These maximums are 8k on the SPARC and 32k
17495 on the m68k and RS/6000. The 386 has no such limit.)
17497 Position-independent code requires special support, and therefore works
17498 only on certain machines. For the 386, GCC supports PIC for System V
17499 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
17500 position-independent.
17502 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
17507 If supported for the target machine, emit position-independent code,
17508 suitable for dynamic linking and avoiding any limit on the size of the
17509 global offset table. This option makes a difference on the m68k,
17510 PowerPC and SPARC@.
17512 Position-independent code requires special support, and therefore works
17513 only on certain machines.
17515 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
17522 These options are similar to @option{-fpic} and @option{-fPIC}, but
17523 generated position independent code can be only linked into executables.
17524 Usually these options are used when @option{-pie} GCC option will be
17525 used during linking.
17527 @option{-fpie} and @option{-fPIE} both define the macros
17528 @code{__pie__} and @code{__PIE__}. The macros have the value 1
17529 for @option{-fpie} and 2 for @option{-fPIE}.
17531 @item -fno-jump-tables
17532 @opindex fno-jump-tables
17533 Do not use jump tables for switch statements even where it would be
17534 more efficient than other code generation strategies. This option is
17535 of use in conjunction with @option{-fpic} or @option{-fPIC} for
17536 building code which forms part of a dynamic linker and cannot
17537 reference the address of a jump table. On some targets, jump tables
17538 do not require a GOT and this option is not needed.
17540 @item -ffixed-@var{reg}
17542 Treat the register named @var{reg} as a fixed register; generated code
17543 should never refer to it (except perhaps as a stack pointer, frame
17544 pointer or in some other fixed role).
17546 @var{reg} must be the name of a register. The register names accepted
17547 are machine-specific and are defined in the @code{REGISTER_NAMES}
17548 macro in the machine description macro file.
17550 This flag does not have a negative form, because it specifies a
17553 @item -fcall-used-@var{reg}
17554 @opindex fcall-used
17555 Treat the register named @var{reg} as an allocable register that is
17556 clobbered by function calls. It may be allocated for temporaries or
17557 variables that do not live across a call. Functions compiled this way
17558 will not save and restore the register @var{reg}.
17560 It is an error to used this flag with the frame pointer or stack pointer.
17561 Use of this flag for other registers that have fixed pervasive roles in
17562 the machine's execution model will produce disastrous results.
17564 This flag does not have a negative form, because it specifies a
17567 @item -fcall-saved-@var{reg}
17568 @opindex fcall-saved
17569 Treat the register named @var{reg} as an allocable register saved by
17570 functions. It may be allocated even for temporaries or variables that
17571 live across a call. Functions compiled this way will save and restore
17572 the register @var{reg} if they use it.
17574 It is an error to used this flag with the frame pointer or stack pointer.
17575 Use of this flag for other registers that have fixed pervasive roles in
17576 the machine's execution model will produce disastrous results.
17578 A different sort of disaster will result from the use of this flag for
17579 a register in which function values may be returned.
17581 This flag does not have a negative form, because it specifies a
17584 @item -fpack-struct[=@var{n}]
17585 @opindex fpack-struct
17586 Without a value specified, pack all structure members together without
17587 holes. When a value is specified (which must be a small power of two), pack
17588 structure members according to this value, representing the maximum
17589 alignment (that is, objects with default alignment requirements larger than
17590 this will be output potentially unaligned at the next fitting location.
17592 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
17593 code that is not binary compatible with code generated without that switch.
17594 Additionally, it makes the code suboptimal.
17595 Use it to conform to a non-default application binary interface.
17597 @item -finstrument-functions
17598 @opindex finstrument-functions
17599 Generate instrumentation calls for entry and exit to functions. Just
17600 after function entry and just before function exit, the following
17601 profiling functions will be called with the address of the current
17602 function and its call site. (On some platforms,
17603 @code{__builtin_return_address} does not work beyond the current
17604 function, so the call site information may not be available to the
17605 profiling functions otherwise.)
17608 void __cyg_profile_func_enter (void *this_fn,
17610 void __cyg_profile_func_exit (void *this_fn,
17614 The first argument is the address of the start of the current function,
17615 which may be looked up exactly in the symbol table.
17617 This instrumentation is also done for functions expanded inline in other
17618 functions. The profiling calls will indicate where, conceptually, the
17619 inline function is entered and exited. This means that addressable
17620 versions of such functions must be available. If all your uses of a
17621 function are expanded inline, this may mean an additional expansion of
17622 code size. If you use @samp{extern inline} in your C code, an
17623 addressable version of such functions must be provided. (This is
17624 normally the case anyways, but if you get lucky and the optimizer always
17625 expands the functions inline, you might have gotten away without
17626 providing static copies.)
17628 A function may be given the attribute @code{no_instrument_function}, in
17629 which case this instrumentation will not be done. This can be used, for
17630 example, for the profiling functions listed above, high-priority
17631 interrupt routines, and any functions from which the profiling functions
17632 cannot safely be called (perhaps signal handlers, if the profiling
17633 routines generate output or allocate memory).
17635 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
17636 @opindex finstrument-functions-exclude-file-list
17638 Set the list of functions that are excluded from instrumentation (see
17639 the description of @code{-finstrument-functions}). If the file that
17640 contains a function definition matches with one of @var{file}, then
17641 that function is not instrumented. The match is done on substrings:
17642 if the @var{file} parameter is a substring of the file name, it is
17643 considered to be a match.
17646 @code{-finstrument-functions-exclude-file-list=/bits/stl,include/sys}
17647 will exclude any inline function defined in files whose pathnames
17648 contain @code{/bits/stl} or @code{include/sys}.
17650 If, for some reason, you want to include letter @code{','} in one of
17651 @var{sym}, write @code{'\,'}. For example,
17652 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
17653 (note the single quote surrounding the option).
17655 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
17656 @opindex finstrument-functions-exclude-function-list
17658 This is similar to @code{-finstrument-functions-exclude-file-list},
17659 but this option sets the list of function names to be excluded from
17660 instrumentation. The function name to be matched is its user-visible
17661 name, such as @code{vector<int> blah(const vector<int> &)}, not the
17662 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
17663 match is done on substrings: if the @var{sym} parameter is a substring
17664 of the function name, it is considered to be a match. For C99 and C++
17665 extended identifiers, the function name must be given in UTF-8, not
17666 using universal character names.
17668 @item -fstack-check
17669 @opindex fstack-check
17670 Generate code to verify that you do not go beyond the boundary of the
17671 stack. You should specify this flag if you are running in an
17672 environment with multiple threads, but only rarely need to specify it in
17673 a single-threaded environment since stack overflow is automatically
17674 detected on nearly all systems if there is only one stack.
17676 Note that this switch does not actually cause checking to be done; the
17677 operating system or the language runtime must do that. The switch causes
17678 generation of code to ensure that they see the stack being extended.
17680 You can additionally specify a string parameter: @code{no} means no
17681 checking, @code{generic} means force the use of old-style checking,
17682 @code{specific} means use the best checking method and is equivalent
17683 to bare @option{-fstack-check}.
17685 Old-style checking is a generic mechanism that requires no specific
17686 target support in the compiler but comes with the following drawbacks:
17690 Modified allocation strategy for large objects: they will always be
17691 allocated dynamically if their size exceeds a fixed threshold.
17694 Fixed limit on the size of the static frame of functions: when it is
17695 topped by a particular function, stack checking is not reliable and
17696 a warning is issued by the compiler.
17699 Inefficiency: because of both the modified allocation strategy and the
17700 generic implementation, the performances of the code are hampered.
17703 Note that old-style stack checking is also the fallback method for
17704 @code{specific} if no target support has been added in the compiler.
17706 @item -fstack-limit-register=@var{reg}
17707 @itemx -fstack-limit-symbol=@var{sym}
17708 @itemx -fno-stack-limit
17709 @opindex fstack-limit-register
17710 @opindex fstack-limit-symbol
17711 @opindex fno-stack-limit
17712 Generate code to ensure that the stack does not grow beyond a certain value,
17713 either the value of a register or the address of a symbol. If the stack
17714 would grow beyond the value, a signal is raised. For most targets,
17715 the signal is raised before the stack overruns the boundary, so
17716 it is possible to catch the signal without taking special precautions.
17718 For instance, if the stack starts at absolute address @samp{0x80000000}
17719 and grows downwards, you can use the flags
17720 @option{-fstack-limit-symbol=__stack_limit} and
17721 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
17722 of 128KB@. Note that this may only work with the GNU linker.
17724 @item -fleading-underscore
17725 @opindex fleading-underscore
17726 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
17727 change the way C symbols are represented in the object file. One use
17728 is to help link with legacy assembly code.
17730 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
17731 generate code that is not binary compatible with code generated without that
17732 switch. Use it to conform to a non-default application binary interface.
17733 Not all targets provide complete support for this switch.
17735 @item -ftls-model=@var{model}
17736 @opindex ftls-model
17737 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
17738 The @var{model} argument should be one of @code{global-dynamic},
17739 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
17741 The default without @option{-fpic} is @code{initial-exec}; with
17742 @option{-fpic} the default is @code{global-dynamic}.
17744 @item -fvisibility=@var{default|internal|hidden|protected}
17745 @opindex fvisibility
17746 Set the default ELF image symbol visibility to the specified option---all
17747 symbols will be marked with this unless overridden within the code.
17748 Using this feature can very substantially improve linking and
17749 load times of shared object libraries, produce more optimized
17750 code, provide near-perfect API export and prevent symbol clashes.
17751 It is @strong{strongly} recommended that you use this in any shared objects
17754 Despite the nomenclature, @code{default} always means public ie;
17755 available to be linked against from outside the shared object.
17756 @code{protected} and @code{internal} are pretty useless in real-world
17757 usage so the only other commonly used option will be @code{hidden}.
17758 The default if @option{-fvisibility} isn't specified is
17759 @code{default}, i.e., make every
17760 symbol public---this causes the same behavior as previous versions of
17763 A good explanation of the benefits offered by ensuring ELF
17764 symbols have the correct visibility is given by ``How To Write
17765 Shared Libraries'' by Ulrich Drepper (which can be found at
17766 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
17767 solution made possible by this option to marking things hidden when
17768 the default is public is to make the default hidden and mark things
17769 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
17770 and @code{__attribute__ ((visibility("default")))} instead of
17771 @code{__declspec(dllexport)} you get almost identical semantics with
17772 identical syntax. This is a great boon to those working with
17773 cross-platform projects.
17775 For those adding visibility support to existing code, you may find
17776 @samp{#pragma GCC visibility} of use. This works by you enclosing
17777 the declarations you wish to set visibility for with (for example)
17778 @samp{#pragma GCC visibility push(hidden)} and
17779 @samp{#pragma GCC visibility pop}.
17780 Bear in mind that symbol visibility should be viewed @strong{as
17781 part of the API interface contract} and thus all new code should
17782 always specify visibility when it is not the default ie; declarations
17783 only for use within the local DSO should @strong{always} be marked explicitly
17784 as hidden as so to avoid PLT indirection overheads---making this
17785 abundantly clear also aids readability and self-documentation of the code.
17786 Note that due to ISO C++ specification requirements, operator new and
17787 operator delete must always be of default visibility.
17789 Be aware that headers from outside your project, in particular system
17790 headers and headers from any other library you use, may not be
17791 expecting to be compiled with visibility other than the default. You
17792 may need to explicitly say @samp{#pragma GCC visibility push(default)}
17793 before including any such headers.
17795 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
17796 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
17797 no modifications. However, this means that calls to @samp{extern}
17798 functions with no explicit visibility will use the PLT, so it is more
17799 effective to use @samp{__attribute ((visibility))} and/or
17800 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
17801 declarations should be treated as hidden.
17803 Note that @samp{-fvisibility} does affect C++ vague linkage
17804 entities. This means that, for instance, an exception class that will
17805 be thrown between DSOs must be explicitly marked with default
17806 visibility so that the @samp{type_info} nodes will be unified between
17809 An overview of these techniques, their benefits and how to use them
17810 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
17812 @item -fstrict-volatile-bitfields
17813 This option should be used if accesses to volatile bitfields (or other
17814 structure fields, although the compiler usually honors those types
17815 anyway) should use a single access in a mode of the same size as the
17816 container's type, aligned to a natural alignment if possible. For
17817 example, targets with memory-mapped peripheral registers might require
17818 all such accesses to be 16 bits wide; with this flag the user could
17819 declare all peripheral bitfields as ``unsigned short'' (assuming short
17820 is 16 bits on these targets) to force GCC to use 16 bit accesses
17821 instead of, perhaps, a more efficient 32 bit access.
17823 If this option is disabled, the compiler will use the most efficient
17824 instruction. In the previous example, that might be a 32-bit load
17825 instruction, even though that will access bytes that do not contain
17826 any portion of the bitfield, or memory-mapped registers unrelated to
17827 the one being updated.
17829 If the target requires strict alignment, and honoring the container
17830 type would require violating this alignment, a warning is issued.
17831 However, the access happens as the user requested, under the
17832 assumption that the user knows something about the target hardware
17833 that GCC is unaware of.
17835 The default value of this option is determined by the application binary
17836 interface for the target processor.
17842 @node Environment Variables
17843 @section Environment Variables Affecting GCC
17844 @cindex environment variables
17846 @c man begin ENVIRONMENT
17847 This section describes several environment variables that affect how GCC
17848 operates. Some of them work by specifying directories or prefixes to use
17849 when searching for various kinds of files. Some are used to specify other
17850 aspects of the compilation environment.
17852 Note that you can also specify places to search using options such as
17853 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
17854 take precedence over places specified using environment variables, which
17855 in turn take precedence over those specified by the configuration of GCC@.
17856 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
17857 GNU Compiler Collection (GCC) Internals}.
17862 @c @itemx LC_COLLATE
17864 @c @itemx LC_MONETARY
17865 @c @itemx LC_NUMERIC
17870 @c @findex LC_COLLATE
17871 @findex LC_MESSAGES
17872 @c @findex LC_MONETARY
17873 @c @findex LC_NUMERIC
17877 These environment variables control the way that GCC uses
17878 localization information that allow GCC to work with different
17879 national conventions. GCC inspects the locale categories
17880 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
17881 so. These locale categories can be set to any value supported by your
17882 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
17883 Kingdom encoded in UTF-8.
17885 The @env{LC_CTYPE} environment variable specifies character
17886 classification. GCC uses it to determine the character boundaries in
17887 a string; this is needed for some multibyte encodings that contain quote
17888 and escape characters that would otherwise be interpreted as a string
17891 The @env{LC_MESSAGES} environment variable specifies the language to
17892 use in diagnostic messages.
17894 If the @env{LC_ALL} environment variable is set, it overrides the value
17895 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
17896 and @env{LC_MESSAGES} default to the value of the @env{LANG}
17897 environment variable. If none of these variables are set, GCC
17898 defaults to traditional C English behavior.
17902 If @env{TMPDIR} is set, it specifies the directory to use for temporary
17903 files. GCC uses temporary files to hold the output of one stage of
17904 compilation which is to be used as input to the next stage: for example,
17905 the output of the preprocessor, which is the input to the compiler
17908 @item GCC_EXEC_PREFIX
17909 @findex GCC_EXEC_PREFIX
17910 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
17911 names of the subprograms executed by the compiler. No slash is added
17912 when this prefix is combined with the name of a subprogram, but you can
17913 specify a prefix that ends with a slash if you wish.
17915 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
17916 an appropriate prefix to use based on the pathname it was invoked with.
17918 If GCC cannot find the subprogram using the specified prefix, it
17919 tries looking in the usual places for the subprogram.
17921 The default value of @env{GCC_EXEC_PREFIX} is
17922 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
17923 the installed compiler. In many cases @var{prefix} is the value
17924 of @code{prefix} when you ran the @file{configure} script.
17926 Other prefixes specified with @option{-B} take precedence over this prefix.
17928 This prefix is also used for finding files such as @file{crt0.o} that are
17931 In addition, the prefix is used in an unusual way in finding the
17932 directories to search for header files. For each of the standard
17933 directories whose name normally begins with @samp{/usr/local/lib/gcc}
17934 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
17935 replacing that beginning with the specified prefix to produce an
17936 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
17937 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
17938 These alternate directories are searched first; the standard directories
17939 come next. If a standard directory begins with the configured
17940 @var{prefix} then the value of @var{prefix} is replaced by
17941 @env{GCC_EXEC_PREFIX} when looking for header files.
17943 @item COMPILER_PATH
17944 @findex COMPILER_PATH
17945 The value of @env{COMPILER_PATH} is a colon-separated list of
17946 directories, much like @env{PATH}. GCC tries the directories thus
17947 specified when searching for subprograms, if it can't find the
17948 subprograms using @env{GCC_EXEC_PREFIX}.
17951 @findex LIBRARY_PATH
17952 The value of @env{LIBRARY_PATH} is a colon-separated list of
17953 directories, much like @env{PATH}. When configured as a native compiler,
17954 GCC tries the directories thus specified when searching for special
17955 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
17956 using GCC also uses these directories when searching for ordinary
17957 libraries for the @option{-l} option (but directories specified with
17958 @option{-L} come first).
17962 @cindex locale definition
17963 This variable is used to pass locale information to the compiler. One way in
17964 which this information is used is to determine the character set to be used
17965 when character literals, string literals and comments are parsed in C and C++.
17966 When the compiler is configured to allow multibyte characters,
17967 the following values for @env{LANG} are recognized:
17971 Recognize JIS characters.
17973 Recognize SJIS characters.
17975 Recognize EUCJP characters.
17978 If @env{LANG} is not defined, or if it has some other value, then the
17979 compiler will use mblen and mbtowc as defined by the default locale to
17980 recognize and translate multibyte characters.
17984 Some additional environments variables affect the behavior of the
17987 @include cppenv.texi
17991 @node Precompiled Headers
17992 @section Using Precompiled Headers
17993 @cindex precompiled headers
17994 @cindex speed of compilation
17996 Often large projects have many header files that are included in every
17997 source file. The time the compiler takes to process these header files
17998 over and over again can account for nearly all of the time required to
17999 build the project. To make builds faster, GCC allows users to
18000 `precompile' a header file; then, if builds can use the precompiled
18001 header file they will be much faster.
18003 To create a precompiled header file, simply compile it as you would any
18004 other file, if necessary using the @option{-x} option to make the driver
18005 treat it as a C or C++ header file. You will probably want to use a
18006 tool like @command{make} to keep the precompiled header up-to-date when
18007 the headers it contains change.
18009 A precompiled header file will be searched for when @code{#include} is
18010 seen in the compilation. As it searches for the included file
18011 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
18012 compiler looks for a precompiled header in each directory just before it
18013 looks for the include file in that directory. The name searched for is
18014 the name specified in the @code{#include} with @samp{.gch} appended. If
18015 the precompiled header file can't be used, it is ignored.
18017 For instance, if you have @code{#include "all.h"}, and you have
18018 @file{all.h.gch} in the same directory as @file{all.h}, then the
18019 precompiled header file will be used if possible, and the original
18020 header will be used otherwise.
18022 Alternatively, you might decide to put the precompiled header file in a
18023 directory and use @option{-I} to ensure that directory is searched
18024 before (or instead of) the directory containing the original header.
18025 Then, if you want to check that the precompiled header file is always
18026 used, you can put a file of the same name as the original header in this
18027 directory containing an @code{#error} command.
18029 This also works with @option{-include}. So yet another way to use
18030 precompiled headers, good for projects not designed with precompiled
18031 header files in mind, is to simply take most of the header files used by
18032 a project, include them from another header file, precompile that header
18033 file, and @option{-include} the precompiled header. If the header files
18034 have guards against multiple inclusion, they will be skipped because
18035 they've already been included (in the precompiled header).
18037 If you need to precompile the same header file for different
18038 languages, targets, or compiler options, you can instead make a
18039 @emph{directory} named like @file{all.h.gch}, and put each precompiled
18040 header in the directory, perhaps using @option{-o}. It doesn't matter
18041 what you call the files in the directory, every precompiled header in
18042 the directory will be considered. The first precompiled header
18043 encountered in the directory that is valid for this compilation will
18044 be used; they're searched in no particular order.
18046 There are many other possibilities, limited only by your imagination,
18047 good sense, and the constraints of your build system.
18049 A precompiled header file can be used only when these conditions apply:
18053 Only one precompiled header can be used in a particular compilation.
18056 A precompiled header can't be used once the first C token is seen. You
18057 can have preprocessor directives before a precompiled header; you can
18058 even include a precompiled header from inside another header, so long as
18059 there are no C tokens before the @code{#include}.
18062 The precompiled header file must be produced for the same language as
18063 the current compilation. You can't use a C precompiled header for a C++
18067 The precompiled header file must have been produced by the same compiler
18068 binary as the current compilation is using.
18071 Any macros defined before the precompiled header is included must
18072 either be defined in the same way as when the precompiled header was
18073 generated, or must not affect the precompiled header, which usually
18074 means that they don't appear in the precompiled header at all.
18076 The @option{-D} option is one way to define a macro before a
18077 precompiled header is included; using a @code{#define} can also do it.
18078 There are also some options that define macros implicitly, like
18079 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
18082 @item If debugging information is output when using the precompiled
18083 header, using @option{-g} or similar, the same kind of debugging information
18084 must have been output when building the precompiled header. However,
18085 a precompiled header built using @option{-g} can be used in a compilation
18086 when no debugging information is being output.
18088 @item The same @option{-m} options must generally be used when building
18089 and using the precompiled header. @xref{Submodel Options},
18090 for any cases where this rule is relaxed.
18092 @item Each of the following options must be the same when building and using
18093 the precompiled header:
18095 @gccoptlist{-fexceptions}
18098 Some other command-line options starting with @option{-f},
18099 @option{-p}, or @option{-O} must be defined in the same way as when
18100 the precompiled header was generated. At present, it's not clear
18101 which options are safe to change and which are not; the safest choice
18102 is to use exactly the same options when generating and using the
18103 precompiled header. The following are known to be safe:
18105 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
18106 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
18107 -fsched-verbose=<number> -fschedule-insns -fvisibility= @gol
18112 For all of these except the last, the compiler will automatically
18113 ignore the precompiled header if the conditions aren't met. If you
18114 find an option combination that doesn't work and doesn't cause the
18115 precompiled header to be ignored, please consider filing a bug report,
18118 If you do use differing options when generating and using the
18119 precompiled header, the actual behavior will be a mixture of the
18120 behavior for the options. For instance, if you use @option{-g} to
18121 generate the precompiled header but not when using it, you may or may
18122 not get debugging information for routines in the precompiled header.